Polarisation vision in the dark: green-sensitive photoreceptors in the nocturnal ball-rolling dung beetle Escarabaeus satyrus

As sunlight interacts with small particules in our atmosphere, a pattern of linearly polarised light forms across the sky. Many diurnal animals, including fish (Novales Flamarique, 2019), birds (Muheim et al., 2009), spiders (Dacke et al., 2001) and insects (Heinze, 2014; Mathejczyk and Wernet, 2017; Rossel and Wehner, 1984; Rossel et al., 1978; Seidl et al., 2006; Wehner and Müller, 2006; Weir and Dickinson, 2012) extract directional information from this spatially predictable pattern to navigate their environment. A similar pattern of polarised light – albeit up to several million times dimmer – is produced around the moon (Heinze, 2014). This pattern offers a potential directional reference for those that need to find their way at night. To date, only dung beetles (Scarabaeidae) are known to steer according to this dimmer version of celestial polarised light (Dacke et al., 2003a, 2011) but nocturnal species across the animal kingdom may share this ability.

In insects, polarised light is typically detected by two sets of photoreceptors, sharing a common field of view, with orthogonally oriented microvilli along the entire length of the rhabdom (Blum and Labhart, 2000; Homberg and Paech, 2002; Labhart and Meyer, 1999; Labhart et al., 2009; Wernet et al., 2012; Wunderer and Smola, 1982). When used to inform orientation and navigation, these specialised photoreceptors are found in the dorsal-most region of the compound eyes, in the so-called dorsal rim area (DRA). Additional anatomical specialisations such as widened and shortened rhabdoms – to enlarge receptive fields and reduce self-screening, respectively – enhance the sensitivity to polarised light (reviewed in Labhart and Meyer, 1999).

The spectral sensitivity of DRA photoreceptors varies across species. The DRAs of honeybees (Labhart, 1980), bumblebees (Meyer-Rochow, 1981), ants (Labhart, 1986), flies (Fortini and Rubin, 1991; Wernet et al., 2003), butterflies (Stalleicken et al., 2006) and diurnal dung beetle species (Dacke et al., 2002; Khaldy et al., 2022) are equipped with photoreceptors primarily sensitive to the ultraviolet (UV) component of light. This widespread perception of celestial polarised light in the UV is rather counterintuitive, as the degree of linear polarisation, as well as the intensity of skylight is considerably lower in the UV than in the blue or green region of the spectrum (Barta and Horváth, 2004; Brines and Gould, 1982; Coulson, 1988). However, the shorter the wavelength, the greater the proportion of celestial polarisation that can be used by animals under cloudy-sky conditions (Barta and Horváth, 2004; Brines and Gould, 1982; Pomozi et al., 2001). The UV-sensitive DRAs of diurnal insects are thus probably adapted to also support the perception of the celestial polarised light pattern under cloudy conditions. While blue-sensitive DRA photoreceptors can be found in crickets (Labhart et al., 1984; Zufall et al., 1989), desert locusts (Eggers and Gewecke, 1993; Schmeling et al., 2014) and cockroaches (Loesel and Homberg, 2001), green-sensitive photoreceptors have so far been reported only for beetles (Bisch, 1999; Khaldy et al., 2022; Labhart et al., 1992) and moths (Belušič et al., 2017). The nocturnal activity of many of these insects suggests that the sensitivity to polarised light in the blue or green spectrum of light is an adaptation to the perception of the celestial polarisation pattern during the darker hours of the day (Hegedüs et al., 2006; Herzmann and Labhart, 1989; Homberg et al., 2011), but the functional significance of this warrants further investigation.

To steer across the savanna, the nocturnal dung beetle Escarabaeus satyrus (Boheman 1860) extracts directional information from the moon, the stars and the pattern of polarised light that spans the night sky (Dacke et al., 2004, 2011, 2013). This beetle's relatively large eyes can be expected to support a high sensitivity to light (Byrne and Dacke, 2011; Tocco et al., 2021), but the specific adaptations that enable them to perceive the dim polarisation patten of the night sky have not been well described. To address this, we first characterised the spectral sensitivity of the photoreceptors in the DRA of the beetle and confirmed its navigational relevance. To understand the adaptive value of green spectral sensitivity for the analysis of the celestial polarisation pattern, we also obtained the polarisation properties of the blue and green components in the South African night sky in the natural habitat of the beetle. A characterisation of the relative photon catches revealed that green-sensitive DRA photoreceptors should catch an order of magnitude more photons from the night sky compared to UV-sensitive photoreceptors in the main retina (which also contains green-sensitive photoreceptors). Finally, we also identified a range of morphological adaptations in the DRA of E. satyrus for increased sensitivity to light.

Nocturnal ball-rolling dung beetles, of the species Escarabaeus satyrus, were collected at Stonehenge game farm (24.32°E, 26.39°S) in South Africa and transported to Lund University, Sweden. Here the beetles were kept in large soil-filled plastic bins in a climate-controlled room (at 26°C and a 12 h light:12 h dark regime) and fed with fresh horse dung ad libitum.

Extracted beetle eyes were fixed for 48 h with 2% paraformaldehyde, 2% glutaraldehyde, 0.1 mmol EGTA and 0.1 mmol sucrose in 0.1 mol sodium cacodylate buffer (pH 7.2) at room temperature. The samples were processed through a series of washes in sodium cacodylate buffer and post-fixed in 1% OsO₄ in distilled H₂O at 4°C for 2.5 h. This was followed by a dehydration process in an increasing ethanol (70%, 2×10 min; 96%, 2×10 min; >99%, 2×15 min) and acetone (acetone, 2×20 min; 2:1 acetone/epon, 1×30 min; 1:1 acetone/epon, overnight) series, and transfer to and embedding in fresh Epoxy resin. Semi-thin (1 μm) and ultra-thin (50 nm) sections of the eye (RMC Powertome PT XL, Boeckeler, USA) were stained with Toluidine Blue (for light microscopy; Axiophot, Zeiss, Germany) or 1% uranyl acetate and lead citrate (for electron microscopy; JEM-1400plus, Japan). For scanning electron microscopy (Hitachi, SU3500, US), the beetles were air-dried and sputter-coated with gold–palladium (40/60).

Each beetle was cold-anaesthetised, immobilised with beeswax and resin and mounted on a goniometric XYZ-stage that carried a micromanipulator (Sensapex, Oulu, Finland). Microelectrodes, pulled from 1 mm diameter borosilicate capillaries on a horizontal puller P-2000 (capillaries and puller Sutter, Novato, CA, USA) were filled with 3 mol l⁻¹ KCl (resistance 100–150 MΩ) and inserted into the photoreceptors of the expected polarisation-sensitive dorsal rim area region through a small triangular hole in the cornea of the dorsal eye. A 50 μm diameter Ag/AgCl reference electrode was inserted into the head capsule next to the eye. The beetles were kept in darkness for 20 min at room temperature prior to the recordings. The electric signal was amplified (SEC 10 LX amplifier, Npi electronic, Tamm, Germany; Cyber Amp 320, Axon Instruments, Union City, CA, USA) and digitised (Micro 1401, CED, Cambridge, UK). Spectral stimulation was provided with an LED array (‘LED synth’; Belušič et al., 2016) or a xenon arc lamp (XBO, Cairn Research Ltd, Faversham, UK) filtered with a monochromator (B&M, Limburg, Germany) tuned to emit equal numbers of photons (max. 10¹⁵ quanta s⁻¹ cm⁻² at 10 nm bandwidth) at all wavelengths (‘isoquantal’ mode).

A quick stimulation with the LED synth allowed for an initial categorisation of the spectral sensitivity of each cell (peak at 365 or 520 nm). An intensity-response series and a sequence of responses to polarised light were recorded at the peak sensitivity wavelength. Polarised light was created by shining monochromatic light through a polarisation filter (OUV2500, Knight Optical, Harrietsham, UK) mounted on a motorised rotator (Qioptiq, Göttingen, Germany). Next, for the cells held for 6 min or more (n=31), a detailed spectral scan with the monochromator (see above) was performed. The spectral responses and responses to polarised light were transformed to sensitivity by means of an intensity–response function and a reverse Hill transformation (Belušič et al., 2017). A cos² function was fitted to the polarisation sensitivity data and the PS ratio of the photoreceptors was calculated as the ratio between the function maximum and minimum, i.e. PS=S_max/S_min (Bernard and Wehner, 1977).

The impaled green-sensitive cells were oriented towards a back-projection screen (ST-Pro-X, Screen-Tech e.K., Hohenaspe, Germany) and stimulated with a DLP (digital light processing) projector (LightCrafter 4500, Texas Instruments, USA) emitting light at ∼530 nm at 220 Hz refresh rate. The receptive field centre was first located at the maximal response coordinates upon the passage of a horizontal and vertical bar across the screen. The cells were then stimulated with 1.4×1.4 deg green square stimuli, flashing on a 26.6×26.6 deg grid. The stimulus was generated with PsychoPy software (Peirce et al., 2019). Voltage responses were transformed into sensitivity, normalised and the resulting 2D spatial sensitivity matrix was fitted with an ellipsoid 2D-Gaussian curve with three free parameters – standard deviations (σ_1,2) and maximal radius angle – yielding two acceptance angles (Δρ_1,2), calculated as the full width at half maximum Δρ_1,2=2.355 σ_1,2.

To calculate the relative photon catch of green-sensitive photoreceptors in the DRA and green- and UV-sensitive photoreceptors in the main retina – at different regions of the spectrum, we used the defined spectral sensitivities of each photoreceptor and irradiance spectra of sunlit, moonlit and starlit skies as reported by Johnsen et al. (2006).

Twenty-one circularly arranged light-emitting diodes (LEDs) with a maximum emission peak at 520 nm (Lumileds, San Jose, CA, USA) were mounted on a square-shaped aluminium plate (60×60 cm) placed 60 cm above a 60 cm diameter sand-painted circular arena (Fig. 1A, see also Khaldy et al., 2022). To diffuse the light, 10 sheets of sandblasted plexiglass (60×60×0.3 cm, Plexiglas Solar 2458, EBLA-GmbH, Appenweier, Germany) were placed at 1 cm intervals starting from 7.5 cm below the LEDs. A 60 cm diameter polarisation filter (BVO UV Polarizer, Bolder Vision Optik, Boulder, CO, USA) was placed below the 10th sheet of Plexiglas. The intensity of the 100% linearly polarised green light was adjusted by neutral density filters (LEE filters) placed in between the LED layout and diffusers to a photon flux of 6×10⁸ photons cm⁻² s⁻¹ (resembling the irradiance of green spectrum of a moonlit sky; Johnsen et al., 2006) as measured from the centre of the arena (QE65000; Ocean Optics, Dunedin, FL, USA).

A beetle was placed alongside its dung ball in the centre of the circular experimental arena and allowed to roll its ball to the arena perimeter. Here, the exit bearing was noted, the beetle was removed from its ball and placed back to the center of the arena with its ball. This procedure was repeated another four times, resulting in five exits per beetle. Beetles for which the angular distribution of the first five exit angles did not differ from a uniform distribution (P>0.1, Rayleigh test on axial data) were excluded from further experimentation. Beetles that in this way were considered motivated to orient under the highly artificial experimental conditions presented to them were allowed to exit the arena an additional five times, resulting in 10 exits per beetle. After the fifth exit, the linear polarisation filter was either turned by 90 deg (test) or kept in place (control) (Fig. 1A). The initial orientation of the overhead polarisation filter was alternated between an e-vector alignment of 0–180 deg and 90–270 deg with respect to the 0 deg mark on the circular arena.

The initial bearing selection of the population in relation to the e-vector of light was obtained from each beetle's first exit point under the polarisation filter (Fig. 1B). A Rayleigh′s uniformity test for axial data was used to test for uniformity of the distribution (Batschelet, 1981). The angular response to a 90 deg rotation of the overhead polarisation filter (test), was calculated from the differences of the mean bearing for exits 1–5 and the mean bearing for exits 6–10 for each beetle. As a control, the corresponding angular change in direction (between the mean bearing for exits 1–5 and the mean bearing for exits 6–10) was obtained for beetles exiting the arena when the polarisation filter remained in place. A V-test for axial data was applied with an expected mean of 0–180 deg for the control group and 90–270 deg for the test group. Circular statistics were performed using Oriana 4.0 (Kovach Computing Services, Anglesey, UK).

Skylight polarisation data was adapted from Foster et al. (2019). An image series was recorded through a photographic polariser (WR 72mm: Sigma) with the transmission axis oriented 0, 45, 90, 135 and 180 deg to the image's horisontal axis. This series was repeated in five directions for each sky condition: towards the zenith, as well as North, East, South and West towards the horizon. The image series were stitched together to create a single full-sky image and converted to estimated absolute spectral radiance for the camera's red, green and blue colour channels. The same image was also convolved with a 7 deg half-width Gaussian filter to approximate the receptive fields measured for DRA photoreceptors (Fig. 6A). For the blue channel (as a proxy for short wavelengths more generally; Pomozi et al., 2001) and the green channel, which matches well with the spectral sensitivity of the DRA photoreceptor in E. satyrus, pixel-wise polarisation was compared across a full moon, waning gibbous moon and waxing quarter moon recorded at the location where the beetles were collected.

Ball-rolling dung-beetle superposition compound eyes are commonly divided by a cuticular ridge, the canthus, into a smaller dorsal and a larger ventral section (Fig. 2A). In E. satyrus, the surface of the dorsal eyes is smooth without any external visual borders between the facets. The thick corneal lens of each ommatidium is followed by a crystalline cone, the clear zone and a light-sensitive retina (Fig. 2C,F). Serial cross-sectioning through the retina revealed two types of rhabdoms. In the dorsal-most 30% of the dorsal eye (505±37 ommatidia, n=2), the microvilli of seven untwisted retinular cells form a heart-shaped rhabdom with the microvilli of one of the seven rhabdomeres oriented perpendicular to that of the other six (Fig. 2D). In other dung beetle species, and in insects in general, this type of orthogonal arrangement is well known to support polarisation vision (Dacke et al., 2003b) and its full extension over the dorsal part of the eye is termed the dorsal rim area (DRA) (Labhart and Meyer, 1999). In the remaining 72% of the dorsal eye (ca. 1200 ommatidia, n=2), the microvilli of the seven retinular cells instead diverge in different directions, forming a flower-shaped rhabdom (Fig. 2E). The rhabdoms in the DRA are longer (106.1±8.5 μm, mean±s.d., t-test, P<0.001, n=9) and somewhat wider (12.3±2.1 μm) than those in the rest of the eye (82.6±4.6 μm and 10.2±0.3 μm, respectively (Fig 2B,F; Fig. 3A,B). This results in rhabdomeric volumes almost twice as large (assuming a cylindrical shape) in DRA retina compared with the main retina; 1.3×10⁴ mm³ versus 6.9×10³ µm³ (Fig. 3C). In E. satyrus, the rhabdoms are not optically isolated from each other by screening pigments (Fig. 2C–E). Instead, they are partly surrounded by a reflective sheath that extends to 34.2% of the length of the rhabdom in the DRA, and as far as 72.1% of the length of the rhabdom in the main retina (35.0±8.3 μm versus 59.2±4.0 μm in height, t-test, P<0.0001) (Fig. 2E,F; Fig. 3A). When the cornea and the clear zone were experimentally removed from the eye, the uppermost dorsal part of the dorsal eye appeared darker than the main retina (Fig. 2B). This optical effect is probably due to the relatively shorter tracheal sheath of the DRA and thus serves as a good, visual indicator of the same.

Intracellular photoreceptor recordings in the DRA of E. satyrus revealed one photoreceptor class, maximally sensitive to light in the green part of the spectrum (λ_max≈520 nm, n=7) (Fig. 4A). Recordings in the main retina rather indicated two types of spectrally distinct photoreceptors, one maximally sensitive to the ultraviolet (UV, λ_max≈365 nm, n=3) and one to the green (λ_max≈520 nm, n=8) part of the spectrum (Fig. 4B). The green-sensitive photoreceptors in the DRA and the main retina both had a slightly broader spectral sensitivity compared with the opsin template with λ_max=520 nm and a β-peak of sensitivity in the UV range with a variable amplitude (5–45% relative to the peak at 520 nm) (Fig. 4A,B). As expected from their rhabdomeric structure (Fig. 2D), the green-sensitive photoreceptors in the DRA were found to have a higher polarisation sensitivity (PS=7.3±5.7; mean±s.d., n=16) than the photoreceptors in the adjacent main retina (PS=1.2±0.2, n=15) (Fig. 4C,D). The observed variability in polarisation sensitivity values can most likely be ascribed to varying sizes of the targeted rhabdomeres (Immonen et al., 2014), as well as retinal depths and the proximity of the recording site to the lamina where the recordings were obtained. For the UV-sensitive photoreceptors of the main retina, a tail of sensitivity (less than 5% relative to the peak at 365 nm) could be observed between 400 and 600 nm (Fig. 4B).

The receptive fields of the photoreceptors were elliptical and could be described with two angular radii, w_1,2. The radii in the DRA (w₁=6.4±0.7 deg; w₂=7.5±0.9 deg, n=5) did not differ significantly in length from the radii (w₁=5.9±0.5 deg; w₂=7.4±0.8 deg, n=8) in the green-sensitive photoreceptors in the main retina (t-test, P_w1=0.25, P_w2=0.95) (Fig. 4E,F).

When repeatedly released with their dung balls in the centre of the circular experimental arena under highly linearly polarised green light (λ_max=520 nm), the beetles tended to preferentially exit along bearings roughly aligned with the e-vector of the overhead light source (mean: 15.1 deg, 95% CI: 359.4–30.9 deg, P=0.003, n=15, Rayleigh test) (Fig. 1B). The 10 exit bearings recorded for each beetle also revealed a directed bimodal distribution (r_axial=0.63±0.15, n=30). When the filter was maintained in the same position for the two experimental sets of five consecutive exits (Fig. 5A), the angular change of direction was only 6.2 deg_axial (95% CI: 335.1–37.3 deg, P=0.025, V-test for axial data along the 0–180 deg axis, n=15). In contrast, when the filter was rotated by 90 deg after the first five exits (Fig. 5A) the angular changes of direction between the mean bearing of exits 1–5 versus exits 6–10 (Fig. 5B) resulted in a mean angular response of 108.9 deg_axial (95% CI: 92.4–125.2 deg, P<0.001, V-test for axial data along the 90–270 deg axis, n=15) (Fig. 5B). This change in bearing is significantly different to the control condition where the polariser remained in the same position (P=0.003, Mardia–Watson–Wheeler Test). Overall, these results clearly demonstrate that the nocturnal dung beetle E. satyrus can detect and orient to polarised light in the green region of the spectrum.

Skylight polarisation measurements for a moonlit sky (Fig. 6) close to a full moon revealed that, while the distributions of per-pixel degree of polarisation for the green and blue channels overlapped closely, modal values, representing the high degree of polarisation band, were slightly shifted towards lower values for the green channel as compared to the blue channel (0.60 and 0.65 respectively) (Fig. 6B). Similarly, a notable shift towards lower degrees of polarisation for the green channel compared to the blue channel was observed for the higher-elevation 93%-illuminated waning gibbous moon (modal values of 0.57 and 0.60) (Fig. 6C). The reduced moon fullness during the first quarter resulted in generally lower degrees of polarisation across the sky (Fig. 6D). In this case, the modal degree of polarisation around 0.15 for the green channel was close to the reported threshold for E. satyrus (Foster et al., 2019).

Calculations of relative photon catch for the green-sensitive photoreceptors in the DRA and the UV-sensitive photoreceptors in the main retina revealed that the green-sensitive photoreceptors would catch an order of magnitude more photons compared with the UV-sensitive photoreceptors under moonlit conditions (relative photon catch 9.26×10¹⁰ versus 9.78×10⁹ quanta per receptor) (Fig. 7). This holds true also under the day sky (relative photon catch 6.28×10¹³ versus 8.95×10¹² quanta per receptor) and a starlit sky on a moonless night (relative photon catch 1.96×10⁸ versus 4.48×10⁷ quanta per receptor) (Fig. 7).

The upper third of the dorsal eye in E. satyrus is internally characterised by heart-shaped rhabdoms, formed by two orthogonally arranged groups of retinal cells with straight and aligned microvilli (Fig. 2). This microvillar arrangement, that supports an analysis of the direction of the e-vector of light, decoupled from its intensity, is a common characteristic of visual systems sensitive to the polarisation of light (Labhart and Meyer, 1999; Nilsson et al., 1987; Labhart, 2016). Intracellular recordings from the photoreceptors in the upper third of the retina of E. satyrus also revealed a polarisation sensitivity (PS) of ≈7 (Fig. 4C). Polarisation sensitivities comparable to that of E. satyrus have also been identified in crepuscular crickets (PS≈8.3) (Labhart et al., 1984) and dung beetles (PS≈7.7) (Dacke et al., 2003b), as well as in a range of diurnal insects including honeybees (PS≈6.6) (Labhart, 1980), monarch butterflies (PS≈9.4) (Stalleicken et al., 2006) and ants (PS≈6.3) (Labhart, 1986).

The flower-shaped rhabdoms in the rest of the eye of E. satyrus are composed of retinular cells with a wide range of microvillar directions. As observed for similar cellular morphologies in other species, the polarisation sensitivity in this part of the eye is drastically reduced (PS≈1.2) (Fig. 4E). Our results thus indicate that the DRA of E. satyrus provides the anatomical and physiological substrate for their documented ability to extract directional information from the nocturnal celestial polarisation pattern (Dacke et al., 2011).

The low intensity of lunar skylight presents a challenge for the perception of the celestial polarisation pattern at night, and the DRA of E. satyrus shows a range of morphological adaptations to meet this challenge (Fig. 2). The larger rhabdomeric volume (twice as large in the DRA compared with the rest of the eye) should increase the light sensitivity of the polarisation-sensitive receptors (Labhart et al., 1992) and improve the signal to noise ratio (Heras and Laughlin, 2017). DRAs with relatively longer rhabdoms compared with the rest of the eye have also been described in the dim-light active dung beetle S. zambesianus (Dacke et al., 2003b), the cockchafer Melolontha melolontha (Labhart et al., 1992) and in crickets (Burghause, 1979). In contrast, the rhabdoms of the DRA photoreceptors in other polarisation-sensitive insects are often shorter than those of the main retina (Labhart and Meyer, 1999; Nilsson et al., 1987). For example, in the diurnal ant Cataglyphis bicolor, the DRA rhabdoms are significantly shorter (70–75 μm) than the rhabdoms (120 μm) of the main retina (Herrling, 1976). Shorter rhabdoms favour polarisation sensitivity through reduced self-screening (Heinze, 2014; Labhart and Meyer, 1999; Nilsson et al., 1987) but for the analysis of the dim celestial pattern of polarised light in the night sky, the extra absorption by longer rhabdoms apparently outcompetes this effect.

The rhabdoms in the dorsal eyes of E. satyrus are partly isolated from each other by tracheal sheath (Fig. 2F). These light scattering structures – which are relatively shorter in the DRA compared with the rest of the eye – also act to increase the overall sensitivity of the photoreceptors by repeatedly reflecting incident light through the rhabdoms (Labhart et al., 1992; Warrant and McIntyre, 1991). In butterflies, these internal reflections only slightly depolarise the reflected light (Nilsson and Howard, 1989). However, the tapetum in eyes with superposition optics is different from that in butterflies and creates a mirrored box around the rhabdom. This box-like organisation of the shorter tracheal sheath in the DRA of E. satyrus might depolarise the incident light to a higher degree. Thus, the shorter tracheal sheath in the DRA is likely the result of a trade-off between an adaptation to enhance sensitivity, while maintaining a high degree of polarisation of incident light.

The resultant receptive fields of DRA photoreceptors, defined as ­6.4 deg (Fig. 4E), are not significantly larger than the 5.9 deg defined for the main retina of the dorsal eye. Although not directly addressed in this study, the unusually large DRA with ∼500 ommatidia covering 30% of the dorsal eye of the beetle (Fig. 2B), opens for the possibility of extensive spatial summation of the polarisation signal, for increased sensitivity, but again at the cost of spatial resolution (Greiner et al., 2005; Hardcastle et al., 2021; Kind et al., 2021; Klaus and Warrant, 2009; Sancer et al., 2019; Theobald et al., 2007). However, for the specific task of extracting directional information from the night sky, lower spatial resolution might be a benefit rather than a cost. This is because larger sampling units effectively reduce the effect of disruption in the celestial polarisation pattern caused by aerosols and clouds by removing the noise in finer spatial details (Labhart, 1999; Labhart et al., 2001). The crepuscular dung beetle S. zambezianus and the nocturnal cockchafer M. melolontha also holds large DRAs with ∼500 (calculated from Dacke et al., 2003b) and ∼280 (calculated from Labhart et al., 1992) ommatidia, respectively. The DRA is much less prominent in the nocturnal bee Megalopta genalis (Greiner et al., 2007), the nocturnal ant Myrmecia pyriformis (Reid, 2010) and the nocturnal corn borer moth Ostrinia nubilalis (Belušič et al., 2017), with the DRA consisting of ∼140, 120 and 100 ommatidia and covering only ∼2.4%, 1.7% and 3% of their eye, respectively.

The photoreceptors in the DRA of E. satyrus are sensitive to a broad spectral range – from green to UV – with a peak in the green (520 nm) (Fig. 4A). To confirm the navigational relevance of this spectral sensitivity of the DRA photoreceptors, the beetles were made to orient under a linearly polarised artificial sky of this wavelength. The beetles did indeed orient with high precision under the green light, but preferentially along bearings that coincided with the e-vector of the artificial sky (Fig. 5B). In crickets, this type of polarotactic behaviour is induced when the animal, as in this study, is presented with a stimulus of a high degree of polarisation (Brunner and Labhart, 1987; Henze and Labhart, 2007). Earlier laboratory experiments reveal that E. satyrus can also orient to polarised light in the UV range of the spectrum (Foster et al., 2019). In this case, the perception of UV polarised light is most likely mediated through the secondary absorbance peak (β-band) of the green-sensitive DRA photoreceptors, but we cannot fully exclude the possibility that there is a small proportion of UV-sensitive photoreceptors in the DRA that escaped our recordings. It is worth noting that in Foster et al. (2019), the intensity level used to elicit a response to polarised UV light was three orders magnitude brighter than the UV content typically present in the moonlight sky (Johnsen et al., 2006).

The two types of photoreceptors in the main retina, one maximally sensitive to the green (λ_max≈520 nm) and one maximally sensitive to the UV (λ_max≈365 nm) range of the spectrum suggest a physiological substrate for a UV–green dichromatic colour vision system, as has been suggested in many other beetle species (Sharkey et al., 2017; Wu et al., 2020; Yilmaz et al., 2022). Considering that the opsin genes typically sensitive to ‘blue’ wavelengths were lost prior to the origin of modern beetles, approximately 300 million years ago, the identified sensitivity peaks in green and UV – but not in the blue – did not come as a surprise (Sharkey et al., 2017). The UV-sensitive photoreceptors typically exhibit response thresholds approximately one log unit lower than the green-sensitive photoreceptors (Fig. 7). This corresponds well to the lower availability of UV photons in the sky compared with the green photons. This balance of response thresholds against the spectrum of ambient illumination could potentially be important in colour vision.

The dry semi-arid habitats of E. satyrus regularly provide the beetles with clear night skies with celestial patterns of high degrees of polarisation (Foster et al., 2019). The radiance of the moonlit sky in the green (500–600 nm) part of the spectrum is 14% higher than in the blue (400–500 nm) and 68% higher than in the UV (300–400 nm) (calculated from Johnsen et al., 2006; see also Foster et al., 2019). That is, at night, there are relatively more photons available in the longer wavelengths of light. Indeed, this is also the primary part of the spectrum that underlies polarisation vision in E. satyrus. This stands in direct contrast to diurnal dung beetle species, which – like most day active navigators – rely primarily on the more robust ultraviolet (UV) component of the celestial polarisation pattern (Dacke et al., 2002; Khaldy et al., 2022). This indicates that, for the analysis of the dim celestial pattern of polarised light in the night sky, the benefits of the higher absorption of light in the green outcompetes the challenges imposed by the larger depolarising effect of clouds in these longer wavelengths of light (Seliger et al., 1994). While the blue-sensitive DRA photoreceptors of crickets support the detection of celestial polarised light under partly cloudy skies as well as during twilight and at night (Henze and Labhart, 2007; Herzmann and Labhart, 1989), the large green-sensitive DRA of E. satyrus seems to be built to primarily meet the challenge of low light intensities and allows the nocturnal beetles to orient with the same precision as their diurnal relatives even under crescent moon conditions (Dacke et al., 2011). In addition, as starlight spectrum has peaks in the green region of the spectrum (Johnsen et al., 2006), the green-sensitive photoreceptors of the unusually large DRA of E. satyrus could, in theory, also support their documented orientation to the Milky Way (Foster et al., 2018).

In summary, green-sensitive DRA photoreceptors, together with a range of morphological adaptations for increased sensitivity, allow the nocturnal ball-rolling dung beetle E. satyrus to increase the quality of their celestial compass readings from the night sky throughout the lunar cycle.

The authors are grateful to Willi Ribi at the Australian National University for kindly providing the cross-section images of the retina in Fig. 2D,E, Carina Rasmussen for assistance in obtaining the EM and light microscopy data and Marcus Byrne at the University of the Witwatersrand for assistance in collecting the beetles. We also thank the microscopy facility at the Department of Biology, Lund University.