Twilight orientation to polarised light in the crepuscular dung beetle Scarabaeus zambesianus

During twilight, the sky completely changes its appearance. The intensity of light drops by 7–9 log units and the red glow from the setting sun is slowly replaced by the pale star shine from other more distant stars. This marked decline in illuminance accelerates when the sun is still visible above the horizon and marks the onset of twilight. It lasts until the sun has disappeared some 20° below the horizon(Rozenberg, 1966). On some evenings, this period of transition is accompanied by the rise of the moon.

Also present in the sky, but not visible to humans, is the pattern of polarised skylight centred on the sun. During the course of the day, this pattern of polarised light changes its appearance with the apparent movement of the sun. During twilight, the pattern is most simple, with the light of the whole sky polarised in one direction. The zenith of the sky now has the highest degree of polarisation of the day, reaching between 70% and 80%polarisation (Brines and Gould,1982; Wehner,1989). This high degree of polarization stretches in a band across the sky from south to north. The remainder of the skylight is polarised in a parallel direction with falling degrees of polarisation towards the sun and the anti-sun. On nights with a full moon, a similar pattern of polarised light will also form around this source of light(Gál et al., 2001).

In the compound eyes of beetles and other arthropods, the microvilli of the retinula cells form light-absorbing rhabdoms. In polarisation-sensitive insects, the arrangement of these structures follows a common pattern; the microvilli in each rhabdomere are organised in only one of two orthogonal directions (for a review, see Labhart and Meyer, 1999). With a maximum sensitivity to light polarised parallel to the direction of the microvilli(Goldsmith and Wehner, 1977; Hardie, 1984; Israelachvili and Wilson,1976), this arrangement tunes the two groups of receptors to orthogonal planes of polarisation. An opponency between the two sets of receptors will not only enhance the polarisation contrast but will also make the system independent of the light intensity of the stimulus(Labhart, 1988; Nilsson and Warrant, 1999). This rhabdom design is generally confined to a narrow strip at the dorsal rim of the eye, termed the dorsal rim area (DRA; Labhart, 1980). Within this area, there are often additional specialisations to facilitate the perception of the polarised light in the sky. Examples of such are a lack of screening pigments or poor lens optics (Aepli et al.,1985; Burghause,1979; Labhart et al.,1992; Meyer and Labhart, 1981, 1993; Ukhanov et al., 1996).

As the day slips towards night, many animals begin – or end –their activity. The dung beetle Scarabaeus zambesianus starts to fly at around sunset with the prospect of finding fresh dung. Once found, it forms a ball of dung and rolls it off at high speed in a line as straight as the terrain will allow. This is supposedly done to avoid competition at and around the food source. The ball is finally buried in a suitable place to be consumed in secure solitude, either by the beetle itself or by a beetle larva(Hanski and Cambefort, 1991). While rolling, the beetle inevitably has to rely upon some sort of reference to stay on route. For many animals, this cue is the polarisation pattern of skylight, a stimulus well known to be used for orientation (see Waterman, 1981).

Whereas diurnal polarised light orientation has been thoroughly explored,twilight orientation remains a rarely investigated topic. An exception is the use of skylight polarization in dusk-migrating birds (Helbig, 1990, 1991; Moore and Philips, 1988;Philips and Moore, 1992), cockchafers(Labhart et al., 1992) and spiders (Dacke et al., 1999, 2001). The mechanism behind the perception of polarised light in birds, however, remains unsolved, and some studies even argue against a use of skylight polarisation as a cue(Coemans et al., 1994). Dusk-active bumblebees have also been suggested to navigate using the polarised light pattern at dusk, when the surroundings are too dim to distinguish terrestrial landmarks(Wellington, 1974). The morphological basis of this behaviour has, to our knowledge, not yet been investigated.

In the present study, we combine the results of behavioural studies of polarised light orientation in the field with those of morphological studies of the eyes of the crepuscular dung beetle Scarabaeus zambesianus. By restricting our experiments to moonless evenings, the beetles could only use direct sunlight and the polarized light pattern of skylight for celestial orientation. We show that polarised light in the sky is used by beetles for orientation to roll balls in straight paths, and we identify the receptors used to perceive this skylight cue. Based on these results, we also question whether the fan-shaped arrangement of analysers in the DRA is an adaptation to polarised light orientation.

Adult male and female Scarabaeus zambesianus Péringuey(Coleoptera: Scarabaeidae) were collected at Cinergy Game Farm, Naboomspruit District, Mpumalanga Province, South Africa, 24°38′ S,28°45′ E. Field experiments were conducted in the same area on evenings when the beetles were spontaneously active. Flying beetles were attracted to the experimental site by the use of 1 litre of fresh pig dung,placed in the centre of an arena 3 m in diameter. The period of activity was determined from the time period during which the beetles arrived at the dung pile. The sandy ground in the arena was carefully levelled and any objects that could function as landmarks were removed. All manipulations on the beetles were performed within this circle, as soon as the beetle had formed a ball and started to roll it away from the pile of dung. To ensure that no beetles were ever used twice in the same experiment, the beetles were individually marked with a number on the thorax using a Liquid-paper™pen. The number of beetles in each experiment is thus equivalent to the number of runs. Laboratory experiments were conducted in Lund, Sweden within two weeks of capturing the beetles. Adult male and female Pachysoma striatum Castelnau (Coleoptera: Scarabaeidae) were collected in Kleinzee,Namaqualand, South Africa, 29°33′ S, 17°10′ E.

In a simple set of experiments, the beetles were allowed to roll along their chosen course with an undisturbed view of the evening sky. A semitransparent Perspex barrier (7 cm high × 30 cm wide) was placed perpendicular to the path of the beetle, forcing the beetle to move around it. The deviation after the obstacle, from the path taken before the obstacle, was measured with a protractor using the tracks made by the beetles in the fine sand as indicators of the path. All measurements were taken within the first hour after sunset on moonless nights.

In a second experiment, a polarising filter (Polaroid HN 22) with a circular diameter of 42 cm was mounted in a holder with four symmetrically placed legs, each with a height of 10 cm. With the help of a magnetic compass fitted on the holder, the filter was placed under the open sky over the expected path of the beetle with its e-vector transmission axis oriented in a west-easterly direction (Fig. 1). Thus, as the beetle entered the area below the filter, the south-northerly oriented polarised light pattern of evening skylight appeared to switch by 90°. The exact orientation of the filter had to be adjusted according to the azimuth of the setting sun. In our experiments, the sun set at 250° east of north and the transmission axis of the filter was oriented 70–250° east of north to simulate a 90° switch of the polarised twilight pattern of skylight. The reaction of the beetles to this switch was filmed using a Sony video camera equipped with `night-shot' and was later analysed in the laboratory for tracing the path of the beetle. Both the empty holder and the holder with the filter with its e-vector transmission axis oriented in a south-northerly direction were placed over the expected path of the beetle to act as two controls. The reaction of the beetles to these controls was recorded and analysed as described above.

In the laboratory, we used the same filter and holder as described above,but the filter was now rotated within the holder. An Osram Ultra-Vitalux lamp(300 W) was used as a light source, centred above the filter. Tracing paper was placed on top of the polarising filter to present an extended polarised stimulus. A beetle and its ball were placed in an arena (70 cm×100 cm)covered with fine sand. As soon as the beetle had rolled 5 cm in under the filter, the filter was rotated through 90°. The angle turned by the beetle, in response to the filter being turned, was measured by analysing the track made by the beetle in the fine sand. As a control, the track made by the beetles under a stationary filter was also recorded. The direction taken towards the centre of the filter was compared with the direction taken after this point.

Sections for light and electron microscopy were prepared using conventional techniques. The eyes were dissected and fixed for 72 h at 8° in a fixative containing 2.5% glutaraldehyde, 2% paraformaldehyde and 0.1 mmol l^–1 EGTA in 0.1 mol l^–1 cacodylate buffer. The buffer was adjusted to pH 7.2. Following postfixation with 1%OsO₄ in 0.1 mol l^–1 cacodylate buffer for 2.5 h at room temperature, the tissue was dehydrated in an ethanol and propylene oxide series and embedded in Epon resin. A possible twist of the DRA rhabdoms was analysed by 0.5 μm cross-sections through the depth of the DRA. In these sections, a single rhabdom was identified during sectioning with the help of irregularities in the retina. The orientation of the transverse-axis of this rhabdom and neighbouring rhabdoms was determined every five sections. Ultra-thin sections (50 nm) were taken and prepared for electron microscopy with 1% uranyl acetate and lead citrate. Animals used for scanning electron microscopy were air-dried and sputter coated with gold–palladium(40/60).

The sun at Naboomspruit set at 250° east of north between 18.58 h and 19.00 h at the time of our experiments. On evenings warm enough for the beetles to fly, they became active some time between 10 min before and 10 min after sunset. When no moon was available, they were active for 40–50 min, while on evenings with a high moon they stayed active for another 40–50 min. On completely overcast nights, no activity could be measured.

S. zambesianus starts forming a ball of dung immediately after landing on a pile of fresh dung. It does so by the use if its flat front legs. As soon as the ball is ready, the beetle pushes it off the pile and climbs on top of it. Here, it cleans its eyes, stretches its head out and performs one or two rotatory (yaw) body movements before it climbs down and starts rolling,head down, pushing the ball with its hind legs. If the beetle encounters a branch or a deep hole that makes it fall over, it often repeats its little dance on top of the ball before it starts rolling again. The ball is finally buried at a suitable spot some 5–30 m away. All balls rolled in this study were feeding balls to be consumed by the beetle itself. Once the ball of dung is depleted, the beetle will leave the burial spot never to return again.

With a full view of the moonless twilight sky, the beetles (N=15)negotiate an obstacle placed in their way by continuously pushing the ball sideways. As they reach the end of the obstacle they again set out parallel to the course they started with an absolute mean angular deviation from the original running direction of 16.9±11.5° (mean ± S.D.). Of 15 beetles, seven turned to the left from the original running direction and eight turned to the right (Fig. 2). Taking the directions of the deviations into consideration,where a left turn is assigned a `minus' and a right turn is assigned a `plus',there is no bias to the left or to the right (–0.2±20.3°). This experiment clearly shows that, in general, beetles maintain their chosen course, deviating from it only temporarily when forced to do so.

In the second experiment, the natural polarisation pattern was manipulated via rotation by 90° as beetles (N=26) rolled their ball in under a polarising filter. The beetles continued to roll along their chosen course until they were at least 5 cm in under the filter. Without any delay,they then turned in response to the rotated polarisation pattern(Fig. 1). A mean turn of 80.9±15.8° was close to the expected 90°. A small turn of 6.7±5.0°, recorded in response to rolling in under a polarising filter placed with its e-vector transmission axis oriented parallel to the e-vector of evening skylight, was not significantly different(t-test, P>0.05) from the response when rolling in under the empty holder of the filter (4.6±4.5°).

In the laboratory, where the beetles (N=8) were exposed to a polarising filter rotated through 90°, the turn was smaller, with a mean value of 61.4±16.1°. With a non-rotated filter, no turn was recorded.

The eyes of S. zambesianus are divided into dorsal and ventral eyes by a cuticular ridge, the canthus, which projects from the edge of the`cheek' (Fig. 3). The dorsal eye is the smaller of the two. Each of these eyes acts independently as a superposition eye. The surface of the eye is perfectly smooth without any visible facets. The thick corneal lens of each ommatidium is attached to a crystalline cone, beneath which is a clear zone and a 120 μm-long rhabdom. No screening pigments, but a tracheal tapetum, can be found between the rhabdoms as far as half way up the rhabdom. For the structures mentioned above, no differences were observed among different regions from the dorsal and ventral eyes. Microvilli from seven of eight retinula cells form the rhabdom in both eyes, but the arrangement of these structures varies across the eye. In the ventral eye and the ventral half of the dorsal eye, the microvilli form a flower-shaped rhabdom(Fig. 4B,C). Here, the microvilli run in different directions in different rhabdomeres.

In the dorsal half of the dorsal eyes, the microvilli of the seven cells run in only two directions, forming an almost heart-shaped structure in cross-section (Figs 4A, 5). In these specialised ommatidia, the microvilli of receptor cell 1 run parallel to those of cell 1 in neighbouring ommatidia but perpendicular to the microvilli of the remaining cells (2–7) of the ommatidium (for cell numbering, see Labhart et al., 1992). This special eye region, termed the dorsal rim area (DRA; Labhart 1980), extends for approximately 26 ommatidia along the dorsal rim of the eye. With a width of approximately 20 ommatidia in the centre, narrowing towards the ends, the DRA covers roughly 50% of the dorsal eye (Fig. 3A). Serial cross-sectioning through the entire length of the rhabdoms indicates that there is no rhabdomeric twist or big jumps in microvillar orientation. Microvillar misalignment strongly influences the polarisation sensitivity as it make the retinula cell less sensitive to one single direction of polarisation (Nilsson et al., 1987; Wehner et al.,1975). The transvere-axis of the DRA rhabdoms, defined by the microvillar direction of photoreceptor cells 2–7, is oriented in a dorsally converging fan-shaped pattern across the eye(Fig. 5). The border of the DRA can easily be found based on the changing appearance of the rhabdoms, but no sharp border can be seen in the orientation of the ommatidial rows. The flower-shaped rhabdoms in the remainder of the eye simply seem to follow the same orientation scheme as in the DRA (Fig. 5).

While a homing animal sets out to find a specific goal, the `leaving'beetle does quite the reverse when escaping from a specific point (the dung pile) as effectively as possible. Despite the contrasting intentions, the best and quickest solution to the two problems becomes the same: orientation along a straight line. When negotiating a barrier placed in its way, S. zambesianus is forced to move around it and deviate from this straight path. Once the beetle has reached the end of the barrier, it can again set out parallel to its previous course (Fig. 6A). The same manipulation on a homing animal will result in a new course to compensate for the sideways movement caused by the barrier(Fig. 6B; Frantsevich et al., 1977; Schmidt et al., 1992). The direction of this new course is obtained by path integration(Müller and Wehner,1988). If the path-integrating animal is instead picked up and transferred sideways, it cannot compensate for the sideways movement and behaves as the ball-rolling beetle (Fig. 6C; Wehner, 1982; Wehner and Wehner, 1986). An obvious difference between homing and leaving is, of course, that the leaving beetle has a set place to start, while the homing animal has a determined place to stop. This simplifies the task of orientation, and a parallel course after negotiating an obstacle will still effectively take the beetle away from the busy dung pile while a homing animal might be hopelessly lost.

For an animal active shortly after sunset, the visual cues that can be used for orientation are fairly limited. It is just dark enough for the stars to be visible and the moon is not always present. Terrestrial landmarks might be used but they will be hard to detect at these low light levels. The only cues left are the polarised light pattern present in the sky together with a brighter sky towards the setting sun. When the direction of the pattern of polarised skylight was artificially switched by 90°, the beetle accordingly changed its course. The same response, but less pronounced, can also be observed in the laboratory. The somewhat smaller turn of the beetle in the laboratory is likely to be explained by a decreased desire to roll when kept in captivity. Nevertheless, both experiments clearly indicate that the beetles perceive the e-vector of light and use it to roll their balls along a chosen path.

In the field, a polarising filter will artificially change the intensity of the skylight polarisation pattern. When the filter is placed with its e-vector transmission axis perpendicular to the dominant polarisation direction of skylight at dusk (north–south), the darkest part of the sky will now be perceived from the zenith (where the degree of polarisation is the highest)rather than from the east. West, however, will still remain the brightest part of the sky and could still serve as a compass cue for orientation. The recorded turn of the beetles under the filter thus shows that polarised light could well be the primary cue used by the beetle to maintain its bearing. The little dance performed on the top of the ball before S. zambesianusstarts to roll supports this idea. In other navigating insects, such rotations are believed to recalibrate the polarization compass before they start their journey (Wehner, 1997).

A course-stabilising function of polarisation vision has also been suggested in flies (von Philipsborn and Labhart, 1990; Wellington,1953; Wolf et al.,1980; Wunderer and Smola,1982) and crickets (Brunner and Labhart, 1987). Common to these insects, and most other polarisation-sensitive insects, is that the receptors used for celestial polarisation analysis are restricted to the DRA of the eye(Labhart and Meyer, 1999). This also holds true for S. zambesianus.

The location of the DRA is revealed by the shape of the rhabdoms. The two sets of receptors with parallel microvilli, oriented 90° to each other,can be found only in the rhabdoms within this dorsal area and these well satisfy the requirements for a polarisation opponent analyzer. A second important characteristic for high polarisation sensitivity is that the microvilli are well aligned along the length of the rhabdom(Nilsson et al., 1987; Wehner et al., 1975).

In some animals, the location of the DRA can be observed from the surface of the eye. Light-scattering cavities in the cornea, or differently shaped facets, discriminate this area from the rest of the eye(Aepli et al., 1985; Burghause, 1979; Labhart et al., 1992; Meyer and Labhart, 1981; Ukhanov et al., 1996). No such differences can be found in the cornea or are visible on the smooth, glassy surface of the eye in S. zambesianus; neither has it been reported in the dung beetle P. striatum(Dacke et al., 2002) or in any other morphological study of the eyes of dung beetles (Scarabaeidae) where an ability to detect the polarisation of light has been suggested (Gokan, 1989a,b,c, 1990; Gokan and Meyer-Rochow, 1990; Meyer-Rochow, 1978). Light-scattering cavities, or differently shaped facets to aid in polarised light detection, thus appear to be absent within the dung beetles(Scarabaeidae).

For polarised light orientation around sunset, the sensitivity of the detector becomes more and more critical as light levels fall continuously with the setting sun. Low light intensities will become a problem for the whole visual system, but here we focus our interest only on the DRA. The dung beetle P. striatum is also able to orient to polarised skylight but is active during the day in one of the brightest habitats on earth: the sun-flooded desert plain (Dacke et al.,2002; Scholtz,1989). How does the optical sensitivity of the DRA in S. zambesianus compare with that in this day-navigating species?

Do the beetles stop foraging when they can no longer perceive and orient to the e-vector of light? At this point, we can only hypothesise about whether this is the case. On a moonless night, the beetles cease their activity 40–50 min after sunset. This halt in activity coincides with the time of night when the light intensity drops dramatically and the degree of polarisation at the sky's zenith decreases from 45% to 5% within 15 min(Dave and Ramanathan, 1956). Crickets can detect the e-vector of strongly polarised light at intensities that are even lower than that from a clear moonless sky, and during the day they need no more than 5% polarisation to detect the direction of polarisation(Herzmann and Labhart, 1989; Labhart, 1996). For S. zambesianus, low light intensities will unfortunately coincide with low degrees of polarisation, and the critical threshold for orientation to the polarisation of twilight skylight will thus probably occur at higher intensities and degrees of polarisation than those recorded for crickets. On moonlit nights, the beetles stay active longer than on moonless nights,possibly using the moon as an orientation source when the polarised light pattern from the sun is no longer available.

S. zambesianus has the same ommatidial array as that found in the DRA of all diurnal (Burghause,1979; Dacke et al.,2002; Wehner,1982; Wunderer and Smola,1982) and crepuscular (Labhart et al., 1992) animals where this has been carefully mapped. This arrangement is remarkably stable between different insect groups, irrespective of whether they have superposition or apposition eyes. It is also stable between insects orienting to the polarisation of skylight at different times of the day. About 15 years ago, this fan-shaped arrangement of the ommatidial axis in the DRA was proposed to mimic the e-vector pattern of skylight(Wehner 1989), but even then the possibility was raised that the arrangement could be a natural consequence of the spherical shape of the eye. In S. zambesianus, we found no morphological difference in the arrangement of the ommatidial rows in the DRA from that in the rest of the eye. This supports the idea that the arrangement of the rhabdoms in the DRA is simply a consequence of the way the eye is built, rather than an adaptation to polarised light analysis. Adaptations for polarised light detection are more likely restricted to the optics of the eye,the orthogonal arrangement of the microvilli and in the way in which the signals from the polarisation analysers are pooled(Blum and Labhart, 2000; Labhart et al., 2001).

We thank Carina Rasmussen and Rita Wallén for expert technical assistance, and Dan Nilsson, Eric Warrant and Marcus Byrne for critically reading the manuscript. We also thank Mr David Alexander for letting us work on his farm and for the hospitality from him and his staff. The project was financed from the Swedish Science Research Council, The Royal Physiographic Society in Lund, SIDA and NRF of South Africa.