A seed flying like a bullet: ballistic seed dispersal in Chinese witch-hazel (Hamamelis mollis OLIV., Hamamelidaceae)

Abstract

The fruits of Chinese witch-hazel (Hamamelis mollis, Hamamelidaceae) act as ‘drying squeeze catapults', shooting their seeds several metres away. During desiccation, the exocarp shrinks and splits open, and subsequent endocarp deformation is a complex three-dimensional shape change, including formation of dehiscence lines, opening of the apical part and formation of a constriction at the middle part. Owing to the constriction forming, mechanical pressure is increasingly applied on the seed until ejection. We describe a structural latch system consisting of connective cellular structures between endocarp and seed, which break with a distinct cracking sound upon ejection. A maximum seed velocity of 12.3 m s⁻¹, maximum launch acceleration of 19 853 m s⁻² (approx. 2000g) and maximum seed rotational velocity of 25 714 min⁻¹ were measured. We argue that miniscule morphological differences between the inner endocarp surface and seed, which features a notable ridge, are responsible for putting spin on the seed. This hypothesis is further corroborated by the observation that there is no preferential seed rotation direction among fruits. Our findings show that H. mollis has evolved similar mechanisms for stabilizing a ‘shot out’ seed as humans use for stabilizing rifle bullets and are discussed in an ecological (dispersal biology), biomechanical (seed ballistics) and functional–morphological (fine-tuning and morphospace of functional endocarps) contexts, and promising additional aspects for future studies are proposed.

1. Introduction

Shooting mechanisms evolved multiple times across several organismal kingdoms and play crucial roles, for example in reproduction, prey capture and defence (reviewed by Sakes et al. [1]). Plants with ballistochory, i.e. which forcibly eject their seeds, have evolved structural means optimizing the seed launch angle, the release of elastic energy stored in the fruits and the reduction of drag during seed flight [2–8], leading to further dispersal distances. In witch-hazels (Hamamelis spp., Hamamelidaceae) (figure 1a), the fruits act as ‘drying squeeze catapults' [1,9] for shooting the seeds.

Figure 1.

Fruits and seeds in H. mollis. (a) Closed and dehisced fruits on a branch. In the dehisced fruits, the seeds have already been ejected. (b) Seeds photographed from different perspectives; the micropyle is indicated. The inset shows a seed in frontal view (the micropyle is on top), with its contour being retraced (dotted yellow line) for highlighting the ridge and resulting mono-symmetry. (c,d) SEM images of the smooth seed surface (testa). (c) Front side of seed, micropyle not shown. (d) Rear end of seed. (Online version in colour.)

The loculicidally deshiscent fruits of the witch-hazel consist of two carpels each containing one seed. The individual carpel consists of a fleshy exocarp and a tough endocarp [10,11]. The motion of the endocarp, which leads to seed ejection, is rendered possible by differential swelling/shrinking properties of its tissue bilayer architecture [12]. It forms two halves, which open and deform due to desiccation and thereby exert pressure on the seeds [13], which eventually become ejected ‘…with as much force as if these were spent shot from a gun’ [14]. Seeds from Hamamelis virginiana can reportedly fly as far as approximately 6 m [10,14,15].

To gain a better understanding of the ejection mechanism, we investigated fruit architecture and seed release in Chinese witch-hazel (Hamamelis mollis OLIV.), which is cultivated in the Botanic Garden Freiburg. Hamamelis mollis is an up to 8 m tall arborescent shrub native to eastern and southern China, where it grows in thickets and forests at altitudes between 300 and 800 m [16]. Specifically, we asked how the combination of slow fruit opening, slow endocarp deformation and fast seed ejection are interconnected, and how the sudden release of the seed is rendered possible. Moreover, we investigated the ballistics of ejected seeds to address the question of how far shooting distances are achieved.

2. Material and methods

2.1. Plant material

We harvested fruits from a H. mollis plant cultivated outdoors in the Botanic Garden Freiburg and stored together with moist paper towels in glass jars for avoiding desiccation-induced fruit opening and seed release.

2.2. Seed morphometry and mass

Seed morphometry was investigated using a SZX9 stereo microscope with a ColorView II digital camera and the Cell^D software (v. 2.6) (Olympus Corp., Tokyo, Japan). The images were subsequently analysed with the software Fiji/ImageJ [17]. Seed masses were determined with an analytical balance (220-5DM, Kern & Sohn GmbH, Balingen-Frommern, Germany).

2.3. Functional–morphological investigations

Seeds and air-dried (open), cleared and manually halved endocarps were investigated with a scanning electron microscope (SEM) (LEO 435 VP, Leica, Wiesbaden, Germany). Preparation involved mounting on aluminium stubs using conductive double-sided adhesive tabs (Plano GmbH, Wetzlar, Germany) and gold coating (approx. 15 nm) with a Sputter Coater 108 auto (Cressington Scientific Instruments Ltd, Watford, UK).

Light microscopy (LM) investigations of closed fruits involved the following preparation steps: (i) embedding with Technovit7100 (Heraeus Kulzer GmbH, Wehrheim, Germany), (ii) production of 3–4 µm thick semi-thin longitudinal sections with a custom-made rotating microtome, (iii) toluidine blue staining by infiltration for 1 min in toluidine blue O (C.I. 52040, Merck KGaA, Darmstadt, Germany) and subsequent 1 min washing with de-ionized water, and (iv) sealing of the microscopy slides with Entellan (Merck KGaA, Darmstadt, Germany). An Olympus BX61 light microscope equipped with a DP71 digital camera and the software Cell^P (v. 2.8) (Olympus Corp., Tokyo, Japan) were used.

We additionally dried one complete, closed fruit at room temperature until the fruit with the respective endocarps were split open, but the seeds not yet released. In this state, the fruit was imaged using magnetic resonance imaging (MRI), which was performed on a 7 T Bruker Biospec 70/20 small animal scanner using a surface 2 cm cryo-coil (both from Bruker Biospin, Ettlingen, Germany) (see also [18,19]). A three-dimensional (3D) UTE (ultra-short echotime) sequence was chosen for high-resolution 3D imaging as it provides a suitable signal-to-noise ratio for specimen with short T2* relaxation time. A field of view of 2 × 2 × 2 cm³ (matrix size 300 × 300 × 300) with image resolution of 67 × 67 × 67 µm³ was chosen, with total imaging time of 25 min and 32 s. The repetition time was 5.5 ms, the echo time was set to 15 µs and a flip angle of 5° was applied for the acquisition of 282 364 projections. The acquired images were post-processed using Avizo software (v. 9.2.0.; © 2018 Thermo Fisher Scientific). A median filter (neighbourhood = 6 and iterations = 1) was applied for sharpening tissue boundaries and reducing image noise. This step simplifies an initial semi-automatic segmentation of various fruit tissues, the exo- and mesocarps, endocarps, testa and embryos, using the watershed algorithm of the software. Subsequently, the segmentation results were manually corrected on the basis of the unfiltered image raw data using the paintbrush tool to ensure a detailed and accurate segmentation of the tissues. Manual image segmentation can lead to the striation pattern on the surface of the 3D data representations (cf. figure 3), as individual two-dimensional tomograms are used as the basis for manual image segmentation. An additional careful manual segmentation of the zone in which endocarp and seed testa are in direct contact (the contact zone) was performed. The contact zone is not a distinct line of cells or tissue borders in the obtained MR image, since the relatively low image resolution (voxel size) of 67 µm³ results in a blurring of tissue borders as the signals of different tissue types admix within a single image voxel (partial volume artefact). Higher image resolution lowers the signal-to-noise ratio, such that smaller voxels would require micro-coils and/or prolongation of image acquisition time. However, as entire, structurally and functionally intact fruits needed to be imaged, which requires for larger RF coils and prolonged image acquisition times, the analyses would result in motion artefacts. Hence, only a zone in which the greyscale values of endocarp and testa admix (contact zone of 1–2 voxels in thickness) could be segmented (region in which endocarp and testa are in contact) and not a distinct tissue or cell layer as it is the case for the connection zone (region of tissue adhesion, see Results).

Figure 3.

MRI analysis of a whole fruit in the dry, open state shortly before seed ejection. The scale bar applies to all images. em, embryo; en, endocarp; ex, exocarp; t, testa. (a) Cross-section through the middle part of the fruit (compare with (c) for orientation), showing both carpels. Note that the right carpel does not contain a viable seed (cf. figure 4a,c). The exocarp, endocarp, testa and embryo are indicated in the left carpel. (b) Schematic of the endocarp and seed, as seen in the cross-section of the left carpel in (a). The ridge of the seed faces to the left. The endocarp is split open and exerts pressure on the seed by bending deformation, thereby forming a constriction (cf. figure 4c). Solid white lines indicate the contact zones between seed and endocarp. The blue arrows indicate viewing directions for (c–i). (c,d) Quasi-3D data representations (3D models) of the left endocarp and the respective seed, as seen from different inclined lateral directions (compare with (b) for orientation). The seed ridge is indicated. In (c), the cross-section is added to highlight the location where the image in (a) was taken. (e–i) Quasi-3D data representations (3D models) of the contact zone (highlighted in red with solid white perimeter lines) between the seed and the endocarp (both transparent), as seen from different lateral directions (compare with (b) for orientation). The seed ridge is indicated. In (i), the inclined lateral view reveals the general mono-symmetrical set-up of the contact zone, which is furthermore characterized by a bowl-shaped lower part (here clasping around the seed) and four ‘arms’, each running along the endocarp dehiscence lines. (Online version in colour.)

2.4. Analyses of fruit and endocarp movements

One whole closed fruit, four cleared endocarps and one fruit with halved exo- and endocarp (by trimming with a scalpel until the seed was exposed to approx. 50% of total seed length prior to ejection) were each fixed with putty and air-dried at an ambient temperature of approximately 23°C and approximately 50% air humidity in our microscopy laboratory. We recorded the respective movements with the same equipment as described in §2.2). The recording speeds (frame rates) of the time-lapse studies were one frame per 3 min (whole fruit), one frame per 2 or 5 min (cleared endocarps) and one frame per 2.5 min (halved fruit). We subsequently analysed the images with Fiji/ImageJ.

2.5. High-speed analyses of seed ejection

A high-speed camera with 8 GB ring buffer (Motion Pro Y4, IDT Inc., Tallahassee, FL, USA), a halogen lamp (C12, 1250 W, Hedler Systemlicht GmbH, Runkel, Rermany) and a ‘Makro-Planar T* 2/100 mm ZF’ camera objective lens (Carl Zeiss AG, Oberkochen, Germany) were used for recording the fast seed release events. Fruits and cleared endocarps were fixed with putty on a wooden beam. Millimetre paper served as scale. The software Motion Studio (v. 2.12.02; IDT Inc., Tallahassee, FL, USA) was used for data acquisition and for optical triggering. Once the endocarp halves split open due to desiccation, the partially exposed seeds were gently marked with a white dot (Tipp-Ex^® Korrekturflüssigkeit Rapid, Tipp-Ex GmbH & Co. KG, Munich, Germany). When the dot started to move out of a frame section which was previously defined in Motion Studio (i.e. when the seed becomes ejected), the camera recorded the seed release and flight. We used the white dots also for the quantification of seed rotation during flight.

We recorded seed release events and flights from lateral views (horizontal measuring range: 90 mm) with a frame capture rate of 30 000 fps. Seeds were ejected from whole fruits (test series A, n = 10) and from cleared intact endocarps (test series B, n = 9). Fiji/ImageJ and Excel 2007 (Microsoft Corporation, Redmont, USA) were used for calculations of seed velocity (for the full distance), launch acceleration and rotational velocity (only for seeds showing at least one full rotation in the respective recordings). We also noted the directions of seed rotation using the following terminology: clockwise rotating, counter-clockwise rotating and wobbling. Additionally, one seed release event from an un-manipulated endocarp was compared with two release events from cleared and trimmed (i.e. shortened with a scalpel) endocarps. In one of these trimmed endocarps, the seed was exposed to approximately 64% of total seed length prior to ejection; in the other one, the seed was exposed to approximately 85% of total seed length.

We also recorded one seed release event in frontal view with 20 000 fps for analysing the endocarp behaviour during seed ejection. To achieve a good view on the location of endocarp–seed interaction, we trimmed the respective single endocarp with a scalpel until the seed was exposed to approximately 50% of total seed length prior to ejection (similarly as described in §2.4).

2.6. Statistics

We tested data of the test series A and B for normal distribution with the Shapiro–Wilk test, for variance homogeneity with the Levene test and possible significant differences were determined with the T-test using the software GNU R v. 1.1.442 [20].

3. Results

3.1. Seed morphometry and mass

Seeds are hard, spindle-shaped and each with a notable ridge in direction to the outer face of the endocarp, inducing a mono-symmetry in frontal view (figure 1b). Their front sides (facing in flight course during ejection) with the micropyles are rounded, whereas the rear ends are tapered. They are 9.4 ± 0.3 mm long (n = 19) and 4.4 ± 0.1 mm wide (at their widest parts) (n = 16), with a mean length/width ratio of 2.1 ± 0.1. Seeds have a mass of 76.8 ± 5.0 mg (n = 19). Our SEM observations reveal that the testa is smooth without any noticeable surface sculpturing (figure 1c,d). Detailed results on seed morphometry and mass are presented in electronic supplementary material, table S1.

3.2. Endocarp functional morphology

A dense network of prosenchymatous cells (fibres) characterizes the rough outer surface of the endocarp (figure 2a). By contrast, its upper inner surface is very smooth (figure 2b,c). Only at the basal part of the inner endocarp surface, extending up to approximately its middle, a dense arrangement of filigree, thread-like surface structures (figure 2b,d) or larger cellular structures (electronic supplementary material, figure S1a,b) can be observed. Our LM investigation of a longitudinal section of a mature, not yet opened fruit (i.e. with intact exo- and endocarp and seed) reveals that the seed is connected to the endocarp only at its middle to lower part (figure 2e). A higher LM magnification shows that the connection zone (region of tissue adhesion) between seed and endocarp is characterized by the presence of a multitude of densely arranged cellular structures (figure 2f). At regions, where the connection zone is partly ruptured prior to seed release, the remnants of this zone are predominantly adhering to the endocarp surface (electronic supplementary material, figure S1c).

Figure 2.

Functional endocarp morphology and endocarp–seed connection zone (the ‘latch’). SEM figures (a–d) show a cleared and halved endocarp. LM figures (e,f) are longitudinal sections of a closed, whole fruit. (a) The rough outer surface of the endocarp features a dense network of prosenchymatous cells (fibres). Owing to the desiccation-driven endocarp deformation, dehiscence lines and cracks occur on the endocarp. (b) The inside of the halved endocarp, showing parts of the constriction zone (indicated in white), where mechanical pressure is exerted on the seed until ejection. The upper part of the inner endocarp surface is very smooth. A multitude of thread-like remnants can be seen at the middle and bottom parts (highlighted by a dotted yellow line). (c) Detail of the smooth upper, inner endocarp surface. (d) Detail of the lower inner endocarp surface with thread-like structures. (e) The fleshy exocarp, the fibrous endocarp and the seed are well visible in the closed fruit. The seed surface is connected to the inner endocarp surface only at its middle and lower parts. A dotted yellow line indicates the connection zone. (f) Higher magnification of the endocarp–seed connection zone (partly indicated by dotted yellow line), which is characterized by a multitude of cellular structures. (Online version in colour.)

Our MRI analysis of an open fruit shortly before seed release shows that the endocarps are strongly bent (as seen in the cross-section, figure 3a,b), thereby forming the constriction at the middle region (cf. figure 2b). The seed is situated inside this deformed capsule-like structure, with its ridge (cf. figure 1b) pointing towards the lateral dehiscence line (figure 3b–d). The endocarp is in contact with the seed (solid white lines in figure 3b) and exerts pressure on it. Our 3D data analysis (data segmentation and 3D representation) of this contact zone (figure 3e–i) shows that it is bowl-shaped at the lower endocarp region, thereby clasping around the seed. In the upper regions, the contact zone splits up into four pointed ‘arms’, each running along the edges of the dehiscence lines (where none of the above-mentioned cellular connective structures were found, cf. figure 2b). The contact zone ‘arms’, which run along the endocarp dehiscence line at the side of the seed ridge, do not extend as much along the endocarp as the other two ‘arms’. By this, the whole contact zone appears as slightly mono-symmetrical, and additional small morphological differences on both ‘mirrored sides’ exist (figure 3i).

3.3. Fruit dehiscence and endocarp deformation

The intact, wet fruit of H. mollis shrinks markedly during desiccation. Cracks eventually appear in the fleshy exocarp at its distal part and the whole fruit splits open. The two tough endocarps become visible, which continue to deform until seed ejection occurs. In our movie, ejection takes place after 1416 min (approx. 23.5 h) (figure 4a; electronic supplementary material, movie S1).

Figure 4.

Desiccation-driven fruit and endocarp movements and seed release. (a) Opening sequence of a whole fruit, consisting of two carpels, during drying (electronic supplementary material, movie S1). At t = 612 min, cracks appear on the fleshy exocarp. At t = 1416 min, the seed from the right carpel is ejected. Note that the left carpel does not contain a viable seed (see also (c) and figure 3a for other fruits with only one viable seed). (b) Different steps of the endocarp motion (electronic supplementary material, movie S2). One dehiscence line and a crack are indicated at t = 40 min. Seed release occurs at t = 114 min. In the first frame after release at t = 116 min, the opening angle of the endocarp is approximately 45°. The motion continues until a maximum state of opening is achieved at t = 260 min (approx. 55°, indicated by solid white lines). (c) Top view on a halved fruit, with the seed still attached (ridge indicated). Initially, the distance between the lateral parts of the endocarp is taken as 100%. At t = 500 min, the constriction has formed and the distance has shrunken to approximately 95%. Mechanical pressure is applied on the seed. This state, shortly before seed ejection, corresponds approximately to the state of the whole fruit used in our MRI investigations (figure 3). Immediately after seed release at t = 502.5 min, the distance is decreased to approximately 76%. The motion continues until a maximum constriction is formed at t = 902.5 min, with a distance corresponding to approximately 54% of its initial value. (Online version in colour.)

The desiccation-induced endocarp motion is a complex 3D deformation including an opening movement (i.e. the apical parts of the endocarp move away from each other), formation of dehiscence lines and additional cracks, and formation of a constriction at the middle part of each endocarp. Except for the formation of dehiscence lines and cracks, this passive-nastic motion is fully reversible, as open endocarps re-close under water. In the four cleared endocarps (i.e. without exocarps) we recorded, seed release occurred within 116–264 min. After release, the endocarp motion proceeds further until a maximum state of opening is achieved (electronic supplementary material, movie S2). In the example presented in figure 4b, the opening angle is approximately 45° at the time of seed ejection and approximately 55° at the end of the motion.

Our recording of a halved fruit with the seed still firmly attached allows us to quantify the formation of the constriction (figure 4c; electronic supplementary material, movie S3). At the beginning, when the structures are still fully wet, the distance between the lateral parts of the endocarp, where the dehiscence lines occur during desiccation, is set as 100%. Owing to water loss, the endocarp deforms over time and this distance gets smaller, i.e. the constriction forms. In the last recorded frame before seed ejection at t = 500 min, it has shrunken to approximately 95% of its initial length. In the next frame, the seed has been ejected and the distance has further shrunken to approximately 76%. The motion proceeds until a maximum constriction has formed, with the distance being approximately 54% of its initial length, at t = 902.5 min.

3.4. Seed release and flight

In both test series A and B, seeds became ejected with an audible cracking sound. Flying seeds show, in most cases, distinct clockwise or counter-clockwise spins. We found no statistical difference in our comparative measurements of seed velocity, launch acceleration and rotational velocity between test series A (complete fruits) and B (cleared endocarps) (detailed results are presented in electronic supplementary material, table S2). Electronic supplementary material, movie S4 shows one exemplary seed release event from test series B. The highest seed velocity of 12.3 m s⁻¹ was measured in test series A (mean: 9.3 ± 2.0 m s⁻¹; n = 10), results of which otherwise do not differ significantly (p = 0.6301) from those of test series B (mean: 9.5 ± 1.0 m s⁻¹; n = 9) (figure 5a). The highest seed kinetic energy of 5.9 mJ (as calculated from seed mass and corresponding velocity, see electronic supplementary material, tables S1 and S2) also was found in test series A (mean: 3.5 ± 1.4 mJ; n = 10). The mean seed kinetic energy for test series B is 3.4 ± 0.7 mJ (n = 9). The highest launch acceleration of 19 853 m s⁻² (approx. 2000g) was measured in test series B (mean: 13 385 ± 3724 m s⁻²; n = 9), results of which otherwise do not differ significantly (p = 0.2485) from those of test series A (mean: 11 664 ± 4858 m s⁻²; n = 10) (figure 5b). The highest seed rotational velocity of 25 714 min⁻¹ (approx. 429 Hz) was measured in test series A (mean: 15 862 ± 5617 min⁻¹; n = 7), results of which do otherwise not differ significantly (p = 0.6449) from those of test series B (mean: 17 152 ± 4556 min⁻¹; n = 8) (figure 5c). In three seed release events from test series A and from one event from series B, we could not calculate the seed rotational velocities, as the respective seeds did not perform one full rotation along the measuring distance (electronic supplementary material, table S2).

Figure 5.

Boxplots of velocity, launch acceleration and rotational velocity of seeds ejected from whole fruits (test series A) and from cleared, intact endocarps (test series B). The sample sizes are indicated. (a) Seed velocity over the whole measured distance, (b) seed launch acceleration and (c) seed rotational velocity. In all comparative measurements (test series A and B in (a), (b), and (c)), no significant differences exist (p > 0.1).

In the experiment with manipulated endocarps, we observed strikingly different seed flight behaviours (figure 6). The seed ejected from an intact endocarp travelled 72.3 mm (100%) over the time span of 7.33 ms (corresponding to a velocity over this distance of 9.85 m s⁻¹). The seeds from the trimmed endocarps travelled 52 mm (72% of the distance found for the intact endocarp, corresponding to 7.09 m s⁻¹ over this distance) in the case in which the seed was exposed by approximately 64% prior to ejection. A travelling distance of 28.2 mm (39% of the distance found for the intact endocarp, corresponding to 3.85 m s⁻¹ over this distance) was measured in the case in which the seed was exposed by approximately 85% prior to ejection. Moreover, distinct wobbling behaviour, i.e. tilting over the transverse axis, is visible in the seeds ejected from manipulated endocarps. Only in the endocarp experiment with 64% seed exposure prior to release, a seed rotational velocity of 5538 min⁻¹ could be calculated, which is markedly lower than the minimal values measured in test series A and B for the intact endocarp.

Figure 6.

Comparative seed ejection analyses with intact and manipulated, cleared endocarps. In the intact endocarp (a), the ejected seed shows only slight wobbling behaviour and travels a distance of 72.2 mm during 7.33 ms (corresponding to a velocity over this distance of 9.85 m s⁻¹). Wobbling behaviour is distinctly visible in the release experiment where the seed was exposed to 64% prior to release (b). In addition, the distance travelled during the given time span of 7.33 ms is smaller here (52 mm, corresponding to a velocity over this distance of 7.09 m s⁻¹). The distance travelled (28.2 mm) and the corresponding velocity over this distance of 3.85 m s⁻¹ are smallest in the experiment where the seed was exposed to 85% prior to release (c). A very prominent wobbling behaviour can be observed in the latter case.

The frontal recording of a trimmed single endocarp (seed exposure approx. 50% relative to total seed length prior to ejection) (electronic supplementary material, movie S5) reveals a sudden deformation of the manipulated endocarp structure upon seed release, combined with a distinct recoil behaviour of the whole endocarp. To be more precise, the edges of the endocarp along the lateral dehiscence lines, which are in contact with the seed, suddenly lose their grip on the seed surface and slip off simultaneously within 0.2 ms. The area of constriction rapidly constricts further, leading to the forcible ejection of the seed which begins to rotate.

The seeds do not have a striking ‘preferential’ rotation direction (electronic supplementary material, table S3). In test series A, four seeds (40%) show a clockwise rotating (cwr) during flight, five seeds (50%) a counter-clockwise rotating (c-cwr) and one seed a wobbling behaviour (10%). In test series B, we found four cwr (approx. 44.4%) and five c-cwr seeds (approx. 55.6%). Altogether (A&B), approximately 42.1% of the seeds perform cwr, approximately 52.6% c-cwr and approximately 5.3% wobbling.

4. Discussion

4.1. Fruit and endocarp motions

Both the exo- and endocarp of the H. mollis fruit show desiccation-driven motions when transferred from a wet into a dry environment. Whereas the fleshy exocarp displays a shrinking behaviour until it splits open (figure 4a; electronic supplementary material, movie S1), the movement of the endocarp is more complex and involves simultaneous opening and constriction deformation (figure 4b,c; electronic supplementary material, movie S2). In the recordings in our laboratory, the opening time of the whole fruit is approximately 5–12 times longer than the opening times of the four analysed, cleared individual endocarps. The opening duration of the halved fruit gave intermediate values (figure 4c). This indicates that the exocarp hinders desiccation and, thereby, endocarp movement to some extent and delays seed release. We hypothesize that the splitting of the exocarp induces endocarp desiccation and consequent motion, which would thereby drive the overall fruit opening sequence, leading to seed release.

Eichholz [12] describes the functional morphology of the H. virginiana endocarp in relation to its passive-nastic motion. According to his comprehensive study, the outer endocarp layer consists of prosenchymatous cells, which span the outer surface in a fan-like manner, i.e. they are aligned horizontally in the middle, whereas more upwards and downwards they appear with gradually increasing angles inclined against the horizontal. The inner prosenchymatous cells are thinner and layered vertically. The desiccation-driven contraction of the outer cells in their transverse direction is 12–15%, but insignificant in the longitudinal direction. The inner cells shrink by 8–10% in their transverse direction, but insignificantly in the longitudinal direction. This anatomical set-up and the different shrinking behaviours result in a vertical straightening of the H. virginiana endocarp tips upon desiccation (leading to the opening motion) and, concomitantly, also in an inward horizontal curvature change at the lower endocarp part (leading to constriction formation), by which the seed is ultimately being pressed out. Due to strong similarities in the general set-up of the endocarp (cf. figure 2) and its motion (cf. figure 4), we assume the same motion principles to be at play in the here investigated H. mollis. Generally, desiccation-driven passive-nastic plant movements are often involved in seed release and dispersal mechanisms (reviewed in [21,22]). It is interesting to note that here a rather slow hydraulic actuation mechanism initiates a very fast seed release without the ‘explosive character’ as known from other fruits, e.g. those of the dynamite tree (Hura crepitans) [3]. The development of ‘drying squeeze catapults' as investigated here in a species of the Hamamelidaceae is otherwise known from members of Euphorbiaceae, Oxalidaceae, Rutaceae, Schisandraceae and Violaceae (reviewed by Sakes et al. [1]).

The observation of the smooth and continuous endocarp motion before as well as after seed ejection (figure 4b,c) indicates that mechanical pressure is increasingly exerted on the seeds until they become ejected. Owing to the finding that the constriction proceeds notably further after ejection (figure 4c; electronic supplementary material, movie S2), one may speculate that the pressure threshold for seed release is probably not kinematically and mechanically very fine-tuned (i.e. by achieving a certain value of endocarp constriction and respective mechanical pressure), but probably rather optimized for functional reliability. However, this remains to be examined in detail and constitutes a promising aspect for future studies (cf. [23]).

The seed launch is initiated by the slipping of the endocarp off the seed (electronic supplementary material, movie S5). Therefore, the smooth seed surface (figure 1b–d) in combination with the also smooth upper inner endocarp surface (figure 2b,c) are presumably structural prerequisites for the correct functioning of the ‘squeeze catapult’, which is assumed to be impeded or even prevented if these surfaces would possess prominent sculpturing. Electronic supplementary material, movie S5 shows the tissue relaxation (i.e. the rapid progression of the constriction within 0.2 ms) in the endocarp upon seed release. This is further supported by figure 4c, where a decrease in the distance between the lateral endocarp parts from approximately 95% of initial length shortly before seed release to approximately 76% after release were measured. This ultra-fast relaxation ranks among the fastest plant movements known so far, which are actuated by passive elastic instabilities resulting from pre-stressing [24–28]. However, the observed endocarps investigated here were trimmed and, therefore, we could not investigate the full 3D relaxation motions. Spatially and mechanistically complex shape-changing plant systems are generally little investigated in detail until now, e.g. in the snap-traps of the carnivorous Venus flytrap and Waterwheel plant, where active hydraulics as well as release of pre-stress are at play [25,28].

4.2. The ‘latch’ for triggering seed ejection

Seed ejection in H. mollis is always accompanied by a distinctly audible cracking sound. By contrast, we had several instances in our experiments where one of the two carpels of a fruit did not contain a viable seed (cf. figures 3a and 4a,c), and the desiccation of the respective endocarps were not accompanied by seed release or cracking sound. Furthermore, when an empty endocarp re-closes under water and then opens again under dry conditions, no sound can be heard either. Shot seeds re-inserted into their original open endocarps, which then similarly re-closed under water and were subsequently dried in air to open, were not shot again and no cracking sound could be heard. The re-inserted seeds were not held firmly by their endocarps, but were dangling to some extent. By contrast, in intact endocarps (which have not yet opened before, e.g. the one we used for our MRI investigations), the respective seeds are always held very firmly (see contact zone in figure 3e–i). Furthermore, cracks do also not act as a trigger, as crack formation often occurred much earlier as seed ejection in our experiments (cf. figure 4b).

Our SEM and LM observations (figure 2; electronic supplementary material, figure S1) reveal a multitude of cellular connective structures at the lower to middle endocarp region, where the seed–endocarp contact zone is bowl-shaped (figure 3e–i). In the fresh state (LM figure 2e,f), with closed endocarp and unreleased seed, these connective structures form a dense array between the inner endocarp surface and the seed testa, which we assume to hold the seed firmly in place. These structures break upon seed release, and remnants of these structures are still visible afterwards (SEM figure 2b,d; electronic supplementary material, figure S1a,b). We interpret these cellular structures to form a functional latch, which breaks upon application of a critical mechanical pressure on the seed by the endocarp constriction, leading to the cracking sound, endocarp relaxation and finally seed ejection. Similar predetermined, rupturing latch regions for fast diaspore release initiation are known, for example, from the Sphagnum moss capsule cap [26] and Touch-Me-Not (Impatiens ssp.) fruits [6,29]. According to electronic supplementary material, figure S1c, the actual adhesion of the connection zone's cellular structures to the seed testa fails more easily than its adhesion to the endocarp. This presumably explains why, after seed release, we have found obvious remnants of the connection zone exclusively on endocarp surfaces (figure 2b,d; electronic supplementary material, figure S1a,b). Future studies could investigate the origin of the connective structures (e.g. if they are truly cellular, or constitute surface outgrowths, or consist of/contain glue), and tackle the question how the biomechanics of the latch interrelate with the force applied by the constriction, i.e. what is the critical tension that the connective zone can resist.

In this context, it is also worth discussing that the latch might be due to overcoming static friction along with conformational changes in the endocarp. Such a mechanism should theoretically be able to function repeatedly, in contrast with the above-discussed structural latch. As already stated, we were not able to induce repeated ejections of seeds re-inserted into their original endocarps. Furthermore, the clearly audible cracking sound in functional endocarps during seed ejection points towards a rupturing process. However, the fact that re-inserted seeds were not shot again may also be due to the complex 3D seed morphology and our handling and reinsertion during the experiments. Additionally, it remains also possible that the sound emission during seed ejection is of another origin as the proposed rupture of the connection zone. The definitive elucidation of the latch mechanism, which could also be a combination of the here described and discussed systems, would require combined force measurements and 3D structural and kinematical analyses during endocarp motion and seed ejection.

4.3. Seed ballistics and flight behaviour

The measured values for seed velocity, launch acceleration and rotational velocity do not differ significantly between the test series with complete fruits and cleared endocarps (figure 5). Accordingly, the endocarp constitutes the mechanical device for seed ejection, whereas the exocarp does not influence the actual ejection but is involved (probably passively, see discussion in §4.1) in the fruit opening process, and/or has other additional functions (e.g. protection).

Our experiments with trimmed endocarps revealed markedly reduced seed flight performances in terms of measured velocities over the respective distances (including the occurrence of distinct wobbling behaviour) (figure 6), which lead to reduced dispersal distances. Therefore, one may speculate that the structural integrity of the endocarp is mandatory for executing its complex 3D deformation behaviour and, therefore, for optimal seed dispersal. In this context, the interrelation of the mechanical fine-tuning (mentioned in §4.1) and the morphospace for fully functional endocarps constitute another promising future research topic.

From our results on seed velocity and mass (for A08, electronic supplementary material, table S1), the theoretical maximum seed flight distance can be calculated by taking into account the maximum release height (3.8 m) of a fruit found on the H. mollis plant in the Botanic Garden (temperature = 25°C, atmospheric pressure = 1 bar, air density = 1.168 kg m⁻³, kinetic viscosity of air = 18.5 µPa s). Accordingly, at an optimal angle of 39° relative to the horizontal (as calculated iteratively from 0 to 90° in 1° intervals), the seed would be able to fly as far as 18 m, which is three times the distance as reported for H. virginiana [10]. We assume that the very high rotational velocity of the seed (up to approx. 429 Hz) contributes to a stable flight and, thus, to the far flight distance. Additionally, the smooth seed surface, which may contribute to a reduced friction coefficient, and the streamlined shape of the seed itself (which is indeed similar to an American football) probably reduce drag during flight. In a recent study on the explosively launched seeds of the perennial herb Ruellia ciliatiflora (Acanthaceae) [8], it is demonstrated that by a very high backspin of 1660 Hz, the disc-shaped seeds (which are smaller and lighter than those of H. mollis) are gyroscopically stabilized and thereby reach flight distances of up to 7 m. In this context, it would be interesting to analyse similarly the drag forces for spinning, non-spinning and wobbling seeds for H. mollis and to calculate possible effects on energy costs for seed dispersal.

4.4. Initiation of seed rotation

In the R. ciliatiflora fruit, small hooks launch the seeds with a contact force that is offset from the seeds' centre of mass, thereby inducing gyration [8]. But how is the spin put on the seed in H. mollis, where no such evident jaculators exist? We found that clockwise or counter-clockwise seed rotating events are rather evenly distributed among our test series (42.1% versus 52.6%, respectively), plus one wobbling event (electronic supplementary material, table S3). Hence, as we could not detect any strict ‘preferential’ seed rotational behaviour, the structural features for rotation initiation apparently vary between fruits.

According to our MRI-based 3D reconstruction, we can generally describe the contact zone between seed and endocarp shortly before ejection as mono-symmetrical (figure 3i). However, this revealed mono-symmetry cannot explain the rotation of the seed around its longitudinal axis. A very suspicious structural feature of the seed is its ridge, which is held between the lateral dehiscence line of the endocarp (figures 1b, 3 and 4c). Upon critical pressure of the endocarp on the seed, the latch breaks (see 4.2), and the endocarp slips off the seed, which then begins to rotate (electronic supplementary material, movie S5). In accordance with the observed ‘non-preferential’ rotation direction, we ultimately hypothesize that small morphological differences on both sides of the seed ridge and/or on the respective endocarp surface dictate on which side of the seed ridge the endocarp slips off first, in which direction the seed is able to rotate while still being situated inside the endocarp, and how the seed is eventually being pressed out of the endocarp. Hence, according to this scenario, the endocarp parts along the dehiscence line with the seed ridge would act as jaculator structures for putting the spin on the seed.

In our MRI-based contact zone analysis, it is possible to see such small morphological differences by comparing both ‘mirrored’ sides. Such differences could, on the one hand, be artefacts of the method by itself, as imaging-resolution-based reconstructions might conceal morphological features that are smaller than the lower resolution limit (here 67 µm³) and could get lost in the partial volume effect. However, on the other hand, we are convinced that small morphological differences indeed exist between these two endocarp sides, as plant structures are never mirror-inverted to 100%.

Another possibility is the existence of an asymmetrical structural latch, with connection zones varying in dimensions between endocarp and seed, resulting in an uneven detachment of the seed from the inner endocarp surface upon release. As we did not follow this further, it remains to be investigated in detail in future approaches how the connection zone is formed and how its detailed symmetry is.

5. Conclusion

We have shown that H. mollis possesses an intriguing seed ejection mechanism, which allows seeds to be shot with high speed and high rotational velocity. We discuss the presence of a latch system for triggering seed release and emphasize that ‘morphological noise’ is likely to be responsible for seed rotation initiation. We hypothesize that these features increase the dispersal distance of H. mollis seeds, which may be a selective advantage as an understorey shrub [8,30]. While short-distance seed dispersal is responsible for the local dynamics of plant populations, long-distance dispersal shapes their large-scale dynamics, and both may be important for a plant species to persist (reviewed by Schurr et al. [31]). For example, a species growing in a fragmented landscape may face local extinction from occupied habitat fragments, which could only be balanced by long-distance dispersal to unoccupied fragments. This highlights that the adaptive significance of increased dispersal distances, as discussed here for H. mollis, remains to be examined in detail, e.g. in the context of population maintenance, gene flow (migration to another population) and colonization of new habitats (cf. [32]).

In conclusion, our analyses allow for first insights into the mechanics of ballistic dispersal mechanism in a species of Hamamelidaceae and show that H. mollis has evolved a similar mechanism for stabilizing a ‘shot out’ seed as humans use for stabilizing rifle or gun bullets. Comparative future experiments among different plant lineages are likely to lead to a deepened understanding of the evolution of this particular seed launch system.

Data accessibility

All data needed to evaluate the conclusions in the paper are present in the paper and/or the electronic supplementary material. Additional data related to this paper may be requested from the authors.

Authors' contributions

All authors contributed to conception and design, analysis and interpretation of data, drafting or revising the article. S.P. wrote the first manuscript draft.

Competing interests

We declare we have no competing interests.

Funding

S.P. and T.S. acknowledge funding by the Joint Research Network on Advanced Materials and Systems (JONAS). L.H. thanks the Joachim Herz Stiftung for support. T.S. and R.S. additionally acknowledge funding by the German Research Foundation (DFG) within the priority program SPP1420 ‘Biomimetic Materials Research: Functionality by Hierarchical Structuring of Materials'.