© INRA/DIB-AGIB/EDP Sciences, 2010

1. INTRODUCTION

Like that of all angiosperms, the pollen of oilseed rape (Brassica napus) is coated with pollenkitt, a complex mixture of lipidic, viscous substances (Stead et al., 1979; Piffanelli et al., 1998; Pacini, 2000). Due to this pollenkit, the grains are released as clumps which stick to the anthers and are not readily dislodged, even under high wind velocity (Eisikowitch, 1981; Pierre, unpubl. data).

The role of bees (Apoidea) as pollinators effecting direct pollen transfer when visiting flowers is well established. The relative importance of their pollinating activity varies primarily with the breeding system of the plant. In the case of open pollinated oilseed rape, the seed production in the standard male-fertile (MF) plants is the combined result of self-pollination, wind pollination by neighbouring plants and pollen transfer by bees (Pierre et al., in prep.) whereas for hybrid seed production bees are essential to transfer pollen from MF to male-sterile (MS) because both type of plants are physically separated within fields and pollen transfer by wind is limited to very short distances (Mesquida and Renard, 1982; Mesquida et al., 1988; Lavigne et al., 1998; Pierre et al., in prep.). Bees could also indirectly take part in pollen dispersal and pollination by making some of the pollen on their body airborne during their flight and grooming (Free and Nuttal, 1968; Free and Ferguson, ; Morh and Jay, 1988). Indeed they could pollinate flowers without touching them. Such a mechanism has been hypothesized by several authors (Mamood and Schmidt, 1991; Vaissière et al., 1996) and its incidence has even been taken into account in models (Rademaker et al., 1997), yet it remains to be demonstrated experimentally.

The aim of our study was thus to determine whether bees could make oilseed rape pollen airborne and to determine the pollinating ability of such airborne pollen. To this end, a cage experiment, with or without bees, was conducted to measure (i) the quantity of airborne pollen released by honeybees, and (ii) the pollination efficiency of such airborne pollen.

2. MATERIAL AND METHODS

2.1. Experimental layout

The experiment was conducted in a meadow under insect-proof cages (3m × 3m × 2m high covered with netting with 1.2 × 1.2 mm openings). Six cages were planted with 4 contiguous rows of male fertile (MF) plants of the oilseed rape spring cultivar ’Tanto’ and 2 contiguous rows with the male sterile (MS) plants (cultivar ‘Fu-Tanto’) and two treatments were applied (3 cages with honeybees, 3 cages without honeybees). A seventh cage was sown with MS plants only and without honeybees to assess the extent of the pollination by airborne pollen that might have entered the cage while passing through the net (net control).

Two experiments were conducted in parallel: one to quantify the airborne pollen deposition on glass slides, the other to measure pollination efficacy. The foraging activity of the honeybees, the number of male flowers and climatic conditions, that are factors potentially influencing the airborne pollen production and pollination were also recorded.

2.2. Deposition of airborne pollen

Pollen released by honeybees was collected on four microscope glass slides coated with double-sided sticky tape. The slides were laid horizontally and located 1.1 m above the ground, corresponding to the average location of flowers. A piece of gauze was placed 10 cm above the slides, to allow comparison between pollen deposition and pollination of the bagged MS plants tested in the second experiment (see below). To minimise the risk of rain exposure, series of slides were collected after 7 periods of exposure (E1 to E7) of variable duration and under variable climatic conditions, some of which nevertheless included some rain. Because counting pollen grains is time-consuming, slides were exposed only in 4 of the cages planted with MS and MF plants (cages 1 and 2 with honey bees; cages 3 and 4 without). The number of grains deposited per unit area was recorded by counting them under a binocular microscope (× 500) over a central area (1.2 × 2.5 cm = 3 cm² on each slide and on each of the 4 slides exposed in each cage (4 replicates/cage/period). Because the environmental conditions during the 7 periods of exposure were very different, the analysis of data was made separately for each period, using a two-way ANOVA model with “cages” nested under “treatment”. Before analysis, the pollen count data were transformed (square root) and the normality was verified.

2.3. Assessment of airborne pollen on plant pollination

At the onset of flowering, 5 MF and 10 MS plants of similar size and distributed throughout the covered area were selected and tagged in each cage. The five MF and five of the MS plants were left to be open-pollinated, while the remaining five MS plants were bagged under gauze (hexagonal openings, 2 mm diameter) to prevent bee visits. Three of the 6 cages then received a small colony of honeybees (Apis mellifera L.), while the remaining cages were used as control without bees. Before the introduction of honeybees, open flowers were removed from the tagged plants in all cages and plant densities were equalized among the 6 cages by removing a few plants as necessary. The experiment was performed over 13 days during the main flowering period, i.e. from the onset of full-flowering of MS plants to the first main drop of petals, after which the beehives were removed. All buds remaining after the withdrawal of the beehives were removed from the tagged plants in all 6 cages, so that pollination could be assessed only on the flowers that opened during the 13 days when honeybees were present.

Pollination was measured on the bagged MS plants, using the open-pollinated MF and MS plants as controls. This is necessary because caging has been shown to induce lower seed yields. All tagged plants were harvested 16 days after bees were removed, enabling to assess fruit set and seed production without loosing seeds during harvest. In the cage with MS plants only (net control), excessive growth and flowering due to lack of pollination led plants to intermingle; this made it impossible to collect plants individually. Therefore, 50 main racemes chosen at random throughout this cage were harvested, one week later than in the other 6 cages to ensure ripening of the main racemes. Fruit set was calculated as the number of not-empty pods divided by the total number of pods. Seed production was determined as the number of seeds per non-empty pod. An average of 332 pods were harvested and examined per tagged plant (i.e. a total of about 1660 pods per cage), while in the cage with MS plants only, a total of 2788 pods were examined.

An overall analysis of variance of the data was made taking into account three factors (treatment, cage nested under treatment, and plant type nested under treatment and cage). Then, a separate analysis of the pollination for each plant type was conducted with a two-way analysis of variance, and a comparison of means between cages with and without honeybees (Tukey test). Before analysis, the data were transformed (respectively angular transformation of Bliss (1938) for fruit set and square root for number of seeds per pod) and the normality was verified.

2.4. Density of available flowers and foraging honeybees

Both factors were recorded under the 4 cages corresponding to the experiment made on the pollen deposition. The density of available MF flowers/m² was measured following Pierre et al. (1996) on 6 dates during the experiment. Meanwhile, the number of honeybees foraging on MF plants was registered for 3 min, from outside the cage (not to disturb bees), when the weather conditions were propitious (no rain and temperature higher than 12 °C). That lead to15 replicates distributed over 8 dates from full-flowering to the decrease of flowering. The measures were made quasi simultaneously on both cages (cage 1 = colony 1; cage 2 = colony 2). The foraging activity (number of foraging honeybees per 1000 flowers) was estimated from these data.

2.5. Climatic conditions

The duration of slide exposure and the climatic conditions during these (temperature, cumulative rain, relative humidity, cumulative wind run and maximal wind speed) were recorded.

3. RESULTS

3.1. Airborne pollen

The amount of pollen deposited was, on average over the 7 exposure periods, 26.5% higher in the cages with honeybees (cages 1 and 2; overall mean = 97.6) compared to the control cages (cages 3 and 4; overall mean = 71.82). However, the number of pollen grain was very variable in each cage and at each period and the difference between treatments was not always statistically significant at each period (Tab. I): it reached significance level in five of the seven periods, but failed to do so in periods 1 and 6.

Several abiotic (Tab. II) or biotic factors (Tab. III) could affect the number of pollen grain deposited on the glass slides such as the duration of exposure, the weather conditions during exposure and the density of MF flowers or the foraging activity of honeybees in each cage. However, Spearman rank correlation tests showed no clear relation between the amount of pollen deposition and environmental factors other than relative humidity and foraging activity. For instance, pollen deposition was significantly higher in cages 2, 3, and 4 when the air was dry (df = 5; r respectively –0.68, 0.67 and –0.71; P < 0.05). There was a significant relationship between the foraging activity and pollen deposition in cage 1 (df = 5; r = 0.75; P = 0.02), but not in cage 2 (df = 5; r = 0.51; P > 0.1) although the bee activity was twice higher in that cage (cage 1: 8.9 ± 1.1 foraging honeybees/1000 flowers; cage 2: 15.1 ± 1.5). The interaction between environmental factors and pollen deposition was rather complex, and difficult to predict. A detailed study of the data showed that during the first exposure period (E1), the number of MF flowers was high and the weather was very dry and windy. Under such conditions, the amount of pollen deposited on the slides was very high and there was no difference between the cages with and without honeybees (average of 322 ± 213 and 376 ± 84 pollen grains/cm², respectively; Tab. I). By contrast, during exposure period 6, the climatic conditions were characterized by a low mean temperature, a factor well known to reduce the foraging activity (confirmed by our data, Tab. III): this in turn can explain the absence of difference in pollen deposition between cages with or without bees at that date.

Figure 1 Percentage of Fruit set (number of pod/number of flowers; ± standard error) on open pollinated male-fertile or male-sterile plants and on bagged male-sterile plants under cages with or without honeybees. Letters indicate means that are different (P < 0.05) for each category of plants, using Tukey test.

Figure 2 Number of seeds per not empty pod (± standard error) on open pollinated male-fertile or male-sterile plants and on bagged male-sterile plants under cages with or without honeybees. Letters indicate means that are different (P < 0.05) for each category of plants, using Tukey test.

3.2. Plant pollination

The percentage of fruit set (Fig. 1), as well as seed set per not-empty pod (Fig. 2) was measured on the three kinds of plants (open pollinated MF and MS plants, bagged MS plants). The analysis of variance of the overall data showed effects of the three factors on fruit set (treatment: df = 2, P = 0.000; cage: df = 4, P = 0.064; plant: df = 12, P = 0.000) whereas only two factors (treatment and plant) had an impact on the number of seeds per not-empty pod (treatment: df = 2, P = 0.000; cage: df = 4, P = 0.727; plant: df = 12, P = 0.000).

Fruit set in open-pollinated MF plants was increased by the presence of honeybees in the cages (average fruit set of 75% and 68% in the cages with and without honeybees, respectively). The difference in the seed production per pod was also significant as the pods from the cages with honey bees had more than twice as many seeds (20.1 versus 9.9).

In open-pollinated MS plants, the effectiveness of honeybees for direct pollen transfer and the low efficiency of wind pollination were clearly confirmed. The fruit set and the seed content per not empty pod showed a 8-fold (69% versus 8.9%) and 5-fold (16.7 versus 3.2) increase, respectively, as a result of the presence of honeybees.

Fruit set in bagged MS plants was significantly higher in the 3 cages with honeybees than in those without honeybees (average of 21% and 3%, respectively). The seed production of these bagged MS plants was also significantly higher in presence of honeybees (average of 5.8 ± 0.7 and 1.7 ± 0.2 seeds/not-empty pod in the cages with and without honeybees, respectively). Furthermore, the comparison of the cages with honeybee colonies 1 and 2 indicated that the seed content per not-empty pod was less in the cage with the lower foraging activity (average of 4.3 and 8.0 seeds/not-empty pod in the cages with colony 1 and 2, respectively). Fruit set and seed content per not-empty pod in the cage with MS plants only were similar to those obtained on bagged MS plants in the cages without honeybees (fruit set = 2.9% and 1.7 seeds/not-empty pod). This indicated that some pollination by the airborne pollen coming from outside the cages was possible through the net covering the cage. Interestingly, these results also suggest that the gauze used to bag the MS plants did not noticeably reduce the pollination by airborne pollen.

4. DISCUSSION

The aim of this work was to determine if bees contributed to the release of airborne pollen, and whether such airborne pollen could result in effective pollination.

The first step was to evaluate the pollen deposition on glass slides as a result of both wind and honeybee foraging compared to that from wind alone. Several experimental difficulties arose in controlling the duration of slide exposure and the weather conditions during exposure, as in some cases rainfall could not be avoided. Moreover, in many situations, the glass slides had to remain exposed during the night or when honeybees did not forage, so that pollen deposition by wind alone could take place, especially when the atmosphere was dry. Such a variability in the conditions of exposure resulted in a high variability of the number of pollen grains deposited on the slides. Nevertheless, the results demonstrated that honeybee foraging generally induced a significant increase in the amount of airborne pollen. A correlation was found between the foraging activity of colony 1 and the amounts of pollen grain deposited at each period. Nevertheless no correlation was detected with colony 2, likely because this colony was more constantly active.

The second step of our study was to evaluate the incidence on pollination of this airborne pollen released by honeybees during their flights and grooming. This was calculated as the difference between the percentage of fruit set and the number of seeds per non-empty pod in MS plants bagged and exposed to honeybees and MS plants un-bagged and not exposed to honeybees. The differences amounted to ca. 17% for fruit set and 4 for seeds/not-empty pod. These figures may be overestimated, since (i) the density of foragers was higher in our experimental conditions than those commonly recorded on winter oilseed rape in the open (Pierre et al., 2003), and (ii) the efficiency of pollination by airborne pollen was measured after 13 consecutive days of exposure to honeybees, which is more than the total duration of exposure of slides (187.75 hours = 7.8 days with interruption phases).

From these results, obtained under cages that reduce wind speed (Mesquida and Renard, 1982), it is not possible to estimate the part of pollination due, in the field, to the release of pollen in the atmosphere by honeybees. Nevertheless, our results could explain why pollen grains are found in high quantities in the air over the crop (Williams, 1884), whereas the oilseed rape pollen grains are not easily disseminated by wind because they are stick together by their viscous pollenkit (Eisikowitch, 1981).

We suggest that, as they are foraging, honeybees manipulate the pollen aggregates and participate to their breaking up in smaller and consequently lighter parts. This could lead to an increase of the atmospheric pollen amount dropped onto plants. This interpretation of our results underlines that wind and insect are not only two separate additive factors but that foraging honeybees can facilitate the pollen dissemination by the wind. This pollination mode, the existence of which was discussed by Brantjes (1981) who called it “insect-assisted wind pollination” in the context of ecological considerations, is thus now established.

Acknowledgments

The authors thank Nathalie Pelotte and Marguerite Dedryver for their technical assistance. They also thank Pr. Jean-Sébastien Pierre for his valuation of the statistical analysis of the data.

References

All Tables

All Figures

	Figure 1 Percentage of Fruit set (number of pod/number of flowers; ± standard error) on open pollinated male-fertile or male-sterile plants and on bagged male-sterile plants under cages with or without honeybees. Letters indicate means that are different (P < 0.05) for each category of plants, using Tukey test.
In the text

	Figure 2 Number of seeds per not empty pod (± standard error) on open pollinated male-fertile or male-sterile plants and on bagged male-sterile plants under cages with or without honeybees. Letters indicate means that are different (P < 0.05) for each category of plants, using Tukey test.
In the text