A ballistic pollen dispersal strategy based on stylar oscillation of Hypochaeris radicata (Asteraceae)

Animal-mediated pollination is a complex interaction between plants and pollinators with conflicting interests. Plants strive for efficient and reliable pollen dispersal, whereas pollinators seek floral rewards that can be harvested as efficiently as possible (van der Kooi et al., 2021; Westerkamp, 1996). Plants employ various strategies to attract pollinators and ensure efficient pollen dispersal. One common approach is to offer floral rewards, such as nectar and pollen, to entice pollinators and facilitate direct physical contact between the exposed pollen grains and the floral visitors. To enhance this process, many plants have diversified their morphologies to include features such as a corolla tube (Gurung et al., 2021; Katzer et al., 2019), staminal lever (Claßen-Bockhoff et al., 2004), trapping device (Bröderbauer et al., 2013) or even sexual deception through the use of petals (Vereecken et al., 2012). In contrast, there are other dynamic pollination strategies that do not involve direct contact between the presented pollen and pollinators, such as explosive pollen release (Edwards et al., 2005; Switzer et al., 2018; Taylor et al., 2006) and buzz pollination (Michener, 1962; King et al., 1996; King and Buchmann, 2011; Vallejo-Marín, 2019, 2022; Jankauski et al., 2022). In this paper, we focus on a ubiquitous floral feature of Asteraceae, with the aim of shedding light on previously unexplored dynamic pollen dispersal mechanics in the plant family.

The Asteraceae family, with over 23,000 species making up roughly 10% of all flowering plants, is widely distributed on every continent except Antarctica (Barreda et al., 2015). These plants are characterized by their tightly packed inflorescences, or flower heads, composed of numerous individual flowers, known as florets (Fig. 1). Pollination in the Asteraceae family is initiated by the release of pollen grains into the anther tube, which pushes or brushes the pollen out and exposes it at the stylar surface (Erbar and Leins, 2015; Leins and Erbar, 2006). The flowers of Asteraceae are visited by a diverse range of pollinators, and thus they are considered to be generalists (Alarcón et al., 2008). However, some visitors utilize their elongated mouthparts to obtain nectar without frequently contacting the exposed pollen grains (Fig. 2F–H), potentially leading to insufficient pollen transfer for successful pollination. Corbiculate bees have refined pollen-collection methods using adhesive saliva and a pollen-collecting apparatus (Parker et al., 2015; Matherne et al., 2021; Roberts and Vallespir, 2015; Snodgrass, 1910), but without strategies to prevent excess pollen collection, the costs of attracting such pollinators can outweigh the benefits. While dynamic pollination strategies, such as explosive pollen release and buzz pollination, have been documented in other plant families, they have not been documented in Asteraceae. In this study, we examined the feasibility of transferring pollen to pollinators without direct physical contact through the oscillation of a pollen-bearing style. Our objective was to uncover a previously undocumented mode of pollen dispersal in a typical flower of the Asteraceae family through quantitative experiments and numerical simulations.

Hypochaeris radicata (Asteraceae) was previously adopted as a model species to investigate pollen adhesion (Ito and Gorb, 2019a,b). In this study, we used this species to examine the motion of pollen-bearing styles and the resulting pollen dispersal. Hypochaeris radicata is a perennial plant, native to Europe, and is a cosmopolitan invasive species occurring in a wide range of temperate zones, including America, Japan and Australia (Doi et al., 2006). Hypochaeris radicata is known to be self-incompatible. This means that the successful transport of pollen grains to different individual plants of the same species is necessary to enable its healthy reproduction (Picó et al., 2004). Flowering stems of H. radicata were collected in Kiel, Germany, for experimentation. They were placed in water until the youngest florets exposed fresh pollen.

To observe the pollen-collecting behaviors of pollinators on flowerheads of H. radicata, we filmed videos in slow motion (120 frames s⁻¹) using an iPhone 7 (Apple Inc.) together with a ×30 magnifying glass (Fig. 2C–E).

We collected the newly opened florets from the flower heads and analyzed their morphology under a light microscope (Leica Microsystems, Wetzlar, Germany) (Fig. 1). In order to characterize the mechanical properties of the florets, we measured the spring constant of the following segments: (1) styles, (2) anther tubes and (3) filaments. Fig. 3A illustrates the experimental setup used for this purpose. Each freshly opened floret was first horizontally positioned between two wooden blocks. We fixed the floret specimens at different positions to measure the spring constant of their different segments (Fig. 3B, right): (1) the entire anther tube was fixed to test the style (fixation at f₁), (2) the basal part of the anther tube including the filaments was fixed to test the anther tube (fixation at f₂), and (3) the basal part of the petal was fixed to test the filaments (fixation at f₃). Floret specimens were deflected using a thin piece of metal mounted on a force transducer (10 g capacity; World Precision Instruments Inc., Sarasota, FL, USA). The deflections were always applied 1 mm distal to the fixation positions at a controlled displacement speed of 0.01 mm s⁻¹. The force required to deflect the specimens was continuously recorded using AcqKnowledge 3.7.0 software (Biopac Systems Ltd, Goleta, CA, USA). To measure the spring constant of the floret specimens, we always used the data from the initial part of the obtained force–displacement curves (limited to a displacement equal to 50 μm). In total, we tested 34 florets including 9 tests on styles, 10 on anther tubes and 15 on filaments. Each floret was subjected to a single test and was not used again.

Black circular polyethylene plates with a hole at the center were prepared for visualizing the distribution of pollen grains catapulted by stylar oscillations. Each newly opened floret was vertically fixed through a hole in the plate onto the vise to stand upright (Fig. 4). To examine whether the pollen dispersal caused by the stylar oscillation depends on the magnitude of an initial deflection, an insect pin attached to a micro-manipulator (World Precision Instruments) was brought into contact with either the style or the anther tube of the standing floret (Fig. 4). Contact with the anther tube resulted in a large deflection of the floret, whereas contact with the style caused a smaller deflection. The insect pin was kept moving horizontally until the deflected floret was released (Figs 2I–K and 4). After release, the floret started to oscillate and this led to the release of clumps of pollen grains, here called ‘dispersal units’ (DU), from the style (Fig. 2J,K; Movie 1). The oscillations were filmed by a high-speed camera (Olympus, Tokyo, Japan) at 5000 frames s⁻¹ and tracked using open-source tracking software (Tracker by Douglas Brown; https://physlets.org/tracker/) (Brown and Cox, 2009). In total, we analyzed the stylar oscillations of 55 specimens, among which 24 specimens were used to analyze the distribution and the number of pollen grains catapulted by the oscillations.

The adhesion of pollen grains has previously been measured for H. radicata on the stylar surface: median adhesion force, F_a, was 98 nN (N=50) (Ito and Gorb, 2019a). The mean (±s.d.) mass of individual pollen grains, m, was 15.0±0.4 ng (Ito and Gorb, 2019b). The number of pollen grains forming a single dispersal unit, n, was determined by dividing the number of dispersed grains by the number of dispersal units (n=12.9±5.3, mean±s.d., N=24). The number of contact points, c, was set as 2 to restrict major pollen dispersal to the first cycle of oscillations, as observed in experiments. Using these data, the critical acceleration that can detach a dispersal unit from the style was calculated, and is shown as a boxplot in Fig. 5C. In the numerical simulation, we used the median value of the critical acceleration (a_c=1005 m s⁻²).

We measured the stylar velocity, v, and acceleration, a, at different time frames using the recorded high-speed videos (N=55) (Fig. 6A, red asterisks). When the stylar acceleration was greater than the critical acceleration, the trajectory of the dispersal unit was simulated (Fig. 6A, solid curves). The initial position and velocity of each dispersal unit were set to be equal to those of the styles. The initial position of a dispersal unit (x, y) is referred to here as the detachment point of that unit (Fig. 6B).

Florets of H. radicata can be subdivided into three functional segments from top to bottom: an exposed distal style, an anther tube in the middle and five basal filaments (Figs 1 and 3B). Mechanical testing of the florets revealed distinct stiffness of the three segments. The style, which passes through the anther tube and is exposed at its distal segment, has the lowest spring constant (K=0.2±0.1 N m⁻¹, N=9) among other segments. It is embraced by the stiff anther tube, which features the highest spring constant (K=6.5±2.7 N m⁻¹, N=10). The anther tube is connected to the petal at its proximal part by five filaments, which altogether have an intermediate spring constant (K=2.31.9 N m⁻¹, N=15). The spring constants of the three segments are significantly different from each other (Tukey multiple comparisons of means, filaments versus style: P=0.01; anther tube versus style: P=2.0×10⁻⁷; anther tube versus filaments: P=1.3×10⁻⁴).

The high-speed video analysis enabled us to investigate stylar oscillations. Fig. 2I–K shows three snapshots of the stylar oscillation from release to return. When the floret was deflected, the elastic energy was mainly stored in the proximal part of the flexible style and the filaments. However, no obvious deformation was observed in the anther tube (Fig. 2I). Upon the release, the style snapped back in the opposite direction of the applied displacement, initiating the first half-cycle of oscillation. The oscillation decayed quickly, and the style returned to the resting position (N=55, damping ratio ζ=0.182±0.05, frequency f=119±23 Hz) (Fig. 5B).

Based on the high-speed videos, we found that the first half-cycle caused the major pollen dispersal (Fig. 2K). The maximum acceleration in each half cycle decreased over time (Fig. 5B,C), so that the values were significantly different from each other (Tukey multiple comparisons of means, large deflections, a₁ versus a₂: P<0.001, a₂ versus a₃: P=0.0013, a₃ versus a₁: P<0.001; small deflections, a₁ versus a₂: P<2.2×10⁻⁶, a₂ versus a₃: P=0.0098, a₃ versus a₁: P<0.001).

As shown in Fig. 5C (left), for both large and small deflections, the maximum stylar acceleration in the first half-cycle, a₁, exceeded the critical acceleration, a_c. However, the maximum acceleration in the subsequent half-cycles (i.e. a₂ and a₃) largely overlapped with the critical acceleration, and the third peak, a₃, mostly became lower than the median critical acceleration.

The velocity corresponding to the maximum acceleration in the first half cycle, v₁, was by far the highest and significantly different from velocities corresponding to acceleration in the second and third peaks, v₂ and v₃ (Fig. 5C, right) (Tukey multiple comparisons of means, large deflections, v₁ versus v₂: P<0.001, v₂ versus v₃: P=0.97, v₃ versus v₁: P<0.001; small deflections, v₁ versus v₂: P<0.001, v₂ versus v₃: P=0.77, v₃ versus v₁: P<0.001).

We examined the relationship between the number of dispersed grains, n_d, and three potentially influential parameters: frequency, damping ratio and maximum acceleration of the styles (Fig. 5A). While no correlation was found between the frequency or damping ratio and the number of dispersed grains (Pearson's product-moment correlation, f versus n_d: t=−0.22, d.f.=22, P=0.82; ζ versus n_d: t=−0.74, d.f.=22, P=0.47), we found a significant correlation between maximum acceleration and the number of dispersed grains (Pearson's product-moment correlation, maximum a versus n_d: t=3.2, d.f.=22, P=0.004). The relatively low coefficient of determination (R²=0.32) can be attributed to variation in the number of pollen grains on the examined styles.

Large deflections resulted in significantly higher acceleration in the first two peaks, but not in the subsequent peaks (t-test, a₁: t=3.90, d.f.=52, P=0.00028; a₂: t=2.03, d.f.=52, P=0.047; a₃: t=0.96, d.f.=52, P=0.34) (Fig. 5C, left). They also led to a significantly higher number of dispersed grains, in comparison with small deflections (t-test, t=−2.34, d.f.=16, P=0.034). This is reflected by the positive correlation between the maximum acceleration in the first half-cycle, a₁, and the number of dispersed grains, n_d, demonstrated in Fig. 5A (see above). In contrast, large deflection caused higher corresponding velocity at only the first acceleration peak (t-test, v₁: t=6.10, d.f.=52, P<0.001; v₂: t=1.13, d.f.=52, P=0.26; v₃: t=1.66, d.f.=52, P=0.10) (Fig. 5C, right).

However, as shown in Fig. 7A, the different deflection magnitudes had no significant effect on pollen distribution (t-test, t=−0.96, d.f.=397, P=0.34). The stylar oscillation tended to catapult the dispersal units against the deflection direction with a median dispersal distance of 5.5 mm, for both large and small deflections. Surprisingly, 93.5% of the dispersal units landed within 15.8 mm of the origin, which is equal to the mean radius of flowerheads of H. radicata.

Although our simulations showed a broader peak of the dispersal distance (Fig. 7B), the simulated pollen distribution exhibited similar trends to the measurements: a tendency for pollen dispersal to be directed toward a pollinator and dispersal distances that were restrained within the size of the flowerhead.

Fig. 6A shows the representative simulated trajectories of dispersal units based on the tracked stylar oscillations (red asterisks), initiated by small and large deflection of the styles. Fig. 6B shows estimated detachment points, where Eqn 3 was satisfied. Large deflections widened the range of detachment points, in comparison to small deflections. They also shifted the detachment points towards lower values in both the x- and y-axes, compared with the small deflections (Kruskal test, x-axis: χ²=29.9, P<0.001; y-axis: χ²=45.2, P<0.001). This means that the larger deflections caused pollen detachment closer to the ground and further from a pollinator than the smaller deflections.

This study focused on the oscillation of the style triggered by an initial deflection and its potential as a ballistic lever that catapults pollen towards a flower-visiting pollinator. As the pollen catapult is a passive spring-driven system, adaption to a variety of deflection magnitudes is likely to be key to the successful delivery of pollen to a pollinator and the minimization of waste for dispersed pollen. Below, we discuss the potential role of morphological and mechanical features of the reproductive part of a floret as well as the adhesive properties of pollen in the pollen catapult system.

Similar to other species in the Asteraceae family (Meiri and Dulberger, 1986), H. radicata florets can be regarded as composites made from three morphologically distinctive segments (Figs 1 and 3B). It is known that each of the segments plays an essential role in pollen presentation (Erbar and Leins, 2015, 1995; Leins and Erbar, 2006). Do they also contribute to catapulting the pollen towards visiting pollinators, after pollen presentation? To present pollen and successfully deliver it to visiting pollinators, the floret with the pollen-bearing style should meet these requirements: (1) high mechanical compliance during contact to store elastic energy and prevent damage and (2) catapult pollen towards a pollinator by adapting to varying magnitudes of initial deflections.

At the distal end of a floret, the flexibility of the pollen-bearing style enables high compliance under contact as well as energy storage for the pollen catapult after release (Fig. 2I–K). The middle segment of the floret is reinforced by the anther tube, which functions as a support for the slender style during contact and stabilizes the shape of the floret during non-contact periods. Previous studies have shown that the deformation of a flexible cantilever with a high aspect ratio occurs proximally to the deflection points (Rajabi et al., 2018). Therefore, in the case of physical contact at the style, the elastic energy would be stored in the broad proximal area of the floret. However, the reinforcement of the proximal segment of the floret by the anther tube prevents the deformation of this segment, limiting the amount of energy stored and released at the distal end of the style. This is likely to restrict excessive stylar acceleration, and thus reduce the number of catapulted pollen grains.

The other energy-storing segment, consisting of filaments, is located at the base of the floret. This segment, with its intermediate stiffness, functions as a flexible joint. It enhances the mechanical compliance of the whole structure in contact, and therefore enables adaptation to the different deflection magnitudes. When a large initial deflection is applied, the flexible joint helps to recline the whole floret (Figs 2I and 6A). As mentioned earlier, larger deflections lower the detachment points and shift them further from a pollinator (Fig. 6B). This is likely to counterbalance the increased initial velocity of catapulted dispersal units (Fig. 5C, v₁) so that pollen is still distributed within the same range of distances, independent of the initial deflection magnitude (Fig. 7A).

It is important to mention that the major pollen detachment occurs only in the first half-oscillation cycle. Hence, the general oscillatory frequency has very little influence on pollen detachment. To restrict the pollen detachment only to the first half-cycle, the critical acceleration, a_c, should be less than the first maximum acceleration, a₁, but greater than maximum acceleration in any other oscillation cycles (Fig. 5C, left).

As shown in Eqn 3, the critical acceleration depends not only on the pollen-style adhesion but also on the pollen–pollen adhesion, which governs the number of pollen grains in a dispersal unit. The weaker pollen–pollen adhesion, therefore, decreases the number of grains in a dispersal unit, in comparison to the n=13 measured here. As the inertial force is proportional to the mass of the dispersal unit, a smaller number of pollen grains in a dispersal unit can challenge its detachment from the style. If the number of grains is as small as n=3, the median critical acceleration would be greater than most of the measured maximum stylar accelerations, and this would completely incapacitate the pollen catapult. In contrast, higher pollen–pollen adhesion would result in a higher number of grains in the dispersal units. This would, in turn, lead to a prolonged and probably isotropic pollen dispersal (both towards and against the pollinator) and, therefore, increase the number of wasted pollen grains. Hence, for guided pollen dispersal towards a pollinator, pollen adhesion and the size of dispersal units should be kept within a certain range.

Unlike wind-pollinated species, insect-pollinated ones are covered by a thick layer of a viscous oily substance called pollenkitt (Pacini and Hesse, 2005), which helps to form large dispersal units by bonding pollen grains together (Pacini and Hesse, 2005; Lisci et al., 1996). In addition to the general function of pollenkitt as a ‘pollen adhesive’, we recently showed that it inhibits pollen adhesion by weakening water capillary attraction (Ito and Gorb, 2019b). Therefore, we can assume that pollenkitt may play a role in maintaining pollen adhesion within a specific range. Further studies are needed to quantitatively investigate the role of pollenkitt in pollen–pollen adhesion.

Once pollen detachment occurs, pollen dispersal is mainly governed by the initial velocity of the dispersal units, rather than their acceleration. Hence, the velocity corresponding to each maximum stylar acceleration (Fig. 5C, right) is also key to guided pollen dispersal towards a pollinator.

Unlike the stylar acceleration peaks, the corresponding stylar velocities abruptly decayed, so that the velocities corresponding to the second and third maximum accelerations (i.e. v₂ and v₃, respectively) were not significantly different from each other (Fig. 5C, right). Only the first corresponding velocity, v₁, was significantly higher than the others, because of the phase shift between the stylar acceleration and velocity in the first half-cycle.

Asteraceae are known for attracting various insect pollinators (Alarcón et al., 2008). Their robust adaptability makes a number of them notorious invasive species (Thebaud and Abbott, 1995; Hao et al., 2011). To enable their versatile reproduction, Asteraceae may use multiple strategies to secure pollen transfer to diverse pollinators. Each pollinator species may have its own difficulties in terms of being a functional pollen vector for the benefit of plants. Bees, for example, are both pollen-transporting vectors and pollen consumers. Actively collected pollen grains are soaked with their saliva and packed in a specialized pollen-carrying apparatus, and are therefore no longer available for pollination (Parker et al., 2015; Thorp, 2000). Some pollinators, such as flies from the family Bombyliidae, with their specialized long mouthparts and slender limbs, less frequently make physical contact with styles on flowers, yet are able to obtain nectary rewards (Kastinger and Weber, 2001) (Fig. 2F–H).

In this paper, we discussed that the superficially unremarkable floret of H. radicata, by optimizing its (1) morphological and mechanical properties, (2) adhesive properties of pollen and (3) phase difference between the stylar acceleration and velocity, functions as a projectile tool to catapult pollen towards pollinators. The airborne pollen dispersal system is advantageous for delivering pollen to various body parts of pollinators, including their safe sites, as well as engaging the long-limbed insects beyond the physical reach of the styles (Macior, 1964; Koch et al., 2017). The attached pollen on the pollinators is potentially transferred to the stigma of other flowers by electrostatic attraction (Vaknin et al., 2001; Eisikowitch et al., 2005; Bowker and Crenshaw, 2007a,b) or grooming (Amador et al., 2017). It is also remarkable that more than 90% of the dispersal units landed within a region of the same size as that of the flowerhead. Therefore, even if the dispersal units fail to reach the target pollinator, they remain within the flowerhead, waiting to be transferred to forthcoming pollinators.

It is well known that florets of Asteraceae consecutively mature from the outer row towards the central row. For many species of this family, including H. radicata, this means that young and most pollen-bearing styles are situated in the center of the flowerhead (Harris, 1995). The tuned travel distance of the dispersal units, therefore, may explain the order of floret maturity in Asteraceae. The working principles of the pollen catapult in Asteraceae are rooted in a common floral feature of the family. If this mechanism is found to be present in other species within the family, it could potentially contribute to the reproductive success of one of the largest families of flowering plants. However, future research is necessary to determine the prevalence of similar pollen dispersal mechanics in other species within the family.

In this study, we investigated the pollen dispersal strategy resulting from the stylar oscillation of H. radicata. Based on the combination of high-speed motion analysis, mechanical tests and numerical simulations, we found that the morphologically and mechanically distinctive segments comprising the floret, as well as optimal pollen adhesion, potentially contribute to (1) catapulting pollen towards a visiting pollinator while (2) minimizing wasteful pollen dispersal by (3) adapting to different degrees of oscillation-inducing external deflection. As ballistic pollen dispersal arises from a standard floral morphology of Asteraceae, it can potentially play a role in the pollination of the most ubiquitous flowering family on the planet.

The authors thank Alexander Köhnsen for the help in motion tracking, and the YouTube channel ‘JustNature’ for permission to use video footage of a Bombyliidae. Yuka Ito is acknowledged for permission to use her iPhone for fieldwork, and Aria Ito for inspiration for this study.