Spectral properties of identified polarized-light sensitive interneurons in the brain of the desert locust Schistocerca gregaria

For long-distance navigation many animals rely on external compass cues,such as the position of the sun, the moon, or the earth's magnetic field to maintain navigational directions (Rossel and Wehner, 1984; Rossel and Wehner, 1986; Lohmann and Lohmann, 1996; Wiltschko and Wiltschko, 1996; Wehner,1997; Wehner,2003; Dacke et al.,2003a; Dacke et al.,2003b; Mouritsen and Ritz,2005). Sun compass navigation is a particularly common strategy. In sun compass navigation, the animals adjust their navigational direction at a certain angle to the solar azimuth, the horizontal component of the sun's position in the sky. Celestial cues other than direct sunlight, however, are also useful as a reference to the sun, especially when the sun is not visible at dawn or dusk or when it is hidden behind clouds or large objects. Scattering of sunlight in the atmosphere results in a polarization pattern, in a spectral gradient, and in an intensity gradient along the sunlit sky. Behavioral experiments have demonstrated that desert ants, monarch butterflies, dung beetles and honeybees use the celestial polarization pattern as a cue for navigation (Rossel and Wehner, 1986; Dacke et al.,2003a; Dacke et al.,2003b; Wehner,2003; Saumann et al., 2005) (but see Stalleicken et al., 2005). Desert ants and bees, in addition, can also navigate based on the spectral gradient in the sky (Rossel and Wehner,1984; Wehner,1997).

The neuronal basis of polarized-light vision has been studied in several insect species. Polarized light is perceived by a small dorsal rim area (DRA)in the compound eye. Photoreceptors of the DRA show striking adaptations for detection of polarized light: they are homochromatic, have microvilli that are highly aligned in parallel, and often have wide receptive fields(Labhart and Meyer, 1999; Dacke et al., 2002; Homberg and Paech, 2002). As a result, photoreceptors in the DRA show high polarization sensitivity(Labhart and Meyer, 1999; Dacke et al., 2002; Stalleicken et al., 2006). Their axons project to dorsal areas in the lamina and medulla(Blum and Labhart, 2000; Homberg and Paech, 2002) and provide input to polarized-light sensitive interneurons (POL neurons). POL neurons show sinusoidal modulation of spiking activity depending on the e-vector angle of polarized light. Various types of POL neuron have been characterized in the optic lobe of several species(Labhart, 1988; Homberg and Würden, 1997; Labhart, 2000; Labhart et al., 2001; Loesel and Homberg, 2001; Pfeiffer et al., 2005). Among these, POL-1 neurons of crickets with ramifications in the medulla have been studied particularly well (Labhart et al.,2001; Wehner and Labhart,2006). Both in the field cricket and in the desert locust, POL neurons have also been reported in the central complex(Vitzthum et al., 2002; Sakura and Labhart, 2005), a brain area involved in visual memory and spatial orientation(Strauss, 2002; Liu et al., 2006). Central-complex neurons have receptive fields oriented toward the zenith and display a wide range of e-vector tunings(Homberg, 2004; Sakura and Labhart, 2005). In the desert locust Schistocerca gregaria these neurons are sensitive not only to polarized light but also to unpolarized light(Vitzthum et al., 2002). In contrast, the POL-1 neurons in crickets are not sensitive to unpolarized light(Labhart, 1988). This difference suggests that the sky navigation system of the desert locust might be different in certain respects from that of the field cricket.

In desert locusts, the anterior optic tubercle (AOTu) is a relay station in the polarization vision pathway from the compound eye to the central complex(Homberg et al., 2003). It receives input from line tangential neurons of the medulla. These neurons have dendritic ramifications in the dorsal rim area of the medulla and axonal projections through the anterior optic tract to the AOTu(Homberg et al., 2003). Recently, four types of POL neuron were identified in the AOTu of the locust(Pfeiffer et al., 2005). Two of them, LoTu1 (Fig. 1A) and TuTu1, innervate the lower units of the AOTu bilaterally. Both cell types respond to polarized light and to unpolarized light like POL neurons of the central complex (Fig. 1B,C). LoTu1 and TuTu1 show distinct e-vector tuning and receive polarized-light input exclusively (LoTu1) or largely (TuTu1) via the ipsilateral eye. Their responses to unpolarized light depend on stimulus position(Fig. 1C). Two other types of POL neuron in the AOTu have been studied less well. These neurons (TuLAL1a and TuLAL1b) project to the lateral accessory lobe and provide input to POL neurons of the central complex.

The spectral sensitivity of POL neurons in the locust is not known. In the DRA of S. gregaria, there are blue receptors peaking at 450 nm with high sensitivity to polarized light (polarization sensitivity, PS=6.92) and UV receptors peaking at 320 nm with low sensitivity to polarized light (PS=2.04)(Eggers and Gewecke, 1993). In the rest of compound eye, there are no published data from S. gregaria, but three types of spectral receptors were identified in Locusta migratoria (Vishnevskaya and Shura-Bura, 1990). These types of photoreceptor are maximally sensitive at 360 nm (UV receptor), 430 nm (blue receptor), and 530 nm (green receptor). Which type of spectral receptor dominates the responses of POL neurons in the desert locust? POL neurons in crickets and desert ants are monochromatic and tuned to the spectral sensitivity of polarized-light sensitive photoreceptors in the DRA(Labhart, 1988; Labhart, 2000). Is this also true for the locust?

In the present study, we focus on the two previously characterized bilateral POL neurons of the AOTu, LoTu1 and TuTu1. We penetrated these neurons and analyzed their spectral responses to polarized light and to unpolarized light by using a set of monochromatic filters. We show that both types of neuron are sensitive to polarized blue light. In response to unpolarized light, the spectral responses depend on stimulus position and show antagonism in the response to UV and green light. These responses might be an adaptation to the spectral gradient in the sky.

We used mature female desert locusts Schistocerca gregariaForskål within 1–3 weeks after imaginal moult. Locusts were reared in crowded colonies at 28°C under a light regime of 12 h:12 h light:dark at the University of Marburg. They were fed with fresh wheat leaves and wheat flakes.

Polarized and unpolarized monochromatic light were used for stimulation. Both types of light stimuli were provided by passing the light of a 75 W xenon lamp through a set of narrow band interference filters, neutral density filters, and a circular neutral density wedge spanning 5 log units of intensity. The interference filters and a shutter were controlled by a shutter controller (Lambda 10-2, Sutter Instruments, Novato, CA, USA). The neutral density wedge was adjusted by a custom-built control unit. Both devices were driven by a custom-built program.

Unpolarized monochromatic light was produced by passing light through one of nine interference filters with a spectral range between 330 and 600 nm(330FS10–600FS10; LOT Oriel, Darmstadt, Germany). The maximum intensity of monochromatic light was adjusted to equal photon flux at either 16.5×10¹² or 10.6×10¹² photons cm^–2 s^–1. The intensity of the unpolarized monochromatic light was changed within a range of 3 log units with neutral density filters. Calibration of intensities was carried out with a radiometer(P-9201, Gigahertz-Optik, Puchheim, Germany). The duration of monochromatic light stimuli was either 500 ms or 1 s, separated by 1–3 s of darkness. Light passed through a UV-transmitting quartz light guide attached to a perimeter and was seen by the locust at a distance of 10 cm from the locust's head. The angular extent of the stimuli at the locust's eye was 2°. By moving the light guide along the perimeter, the light stimuli were administered from four directions: from dorsal (zenith), from frontal(elevation about 50°), and from lateral to the right or left eye at an elevation of 30–45°.

Polarized light was produced by inserting a UV transmitting polarizer(HNP'B, Polaroid, Cambridge, MA, USA) between the light guide and the animal. During stimulation, the polarizer was rotated through 360° in either direction at 20 or 21.8° s^–1. We stimulated with both`white' and monochromatic polarized lights (UV, 330 nm; blue, 450 nm; green,530 nm). The maximum intensity of each polarized monochromatic light stimulus at the surface of the locust's eye was adjusted to either 9.0×10¹² or 10.6×10¹² photons cm^–2 s^–1. The intensity of each polarized monochromatic light was changed within a range of 4 log units with neutral density filters.

After cropping legs and abdomen, locusts were fixed to a metal holder with a wax-rosin mixture. The head capsule was opened frontally to expose the brain. The metal holder was mounted in the center of a Faraday cage. Sharp glass microelectrodes filled with 1 mol l^–1 KCl (resistance about 50–150 MΩ) and 4% Neurobiotin in 1 mol l^–1KCl (Vector Laboratories, Burlingame, CA, USA) at the tip were inserted in the vicinity of the AOTu. After successful impalement we first stimulated with polarized light. If the cell was polarization-sensitive, we measured the spectral response properties to polarized light and to unpolarized light. Action potentials were amplified with a custom-made amplifier, monitored with an oscilloscope (Hameg HM 205–2; Hameg, Frankfurt/Main, Germany),digitized at 25 kHz with a Digidata 1322A (Molecular Devices, Sunnyvale, CA,USA) and stored on a personal computer using Clampex 9.2 (Molecular Devices). After recording, the neurons were injected with Neurobiotin by administering positive currents of 1–3 nA for 5–60 s.

The responses to unpolarized monochromatic light were evaluated by measuring the mean spiking rate during 1 s time intervals before the onset of the stimulus, during the stimulus and after the stimulus, using a semiautomatic script. Two-sided student's t-tests were used to determine statistical differences in the responses to different colors.

After the Neurobiotin injection, the locust was kept at room temperature for at least 20 min to allow for diffusion of the tracer. The brain was dissected out of the head capsule and fixed overnight at 4°C in 4%paraformaldehyde, 0.25% glutaraldehyde, and 0.25% saturated picric acid in 0.1 mol l^–1 phosphate buffer pH 7.4 (PB). Brains were subsequently embedded in gelatin/albumin and fixed overnight in 8%formaldehyde in PB at 4°C. Sections of 35 μm were cut with a vibrating blade microtome (VT–1000S, Leica, Wetzlar, Germany). They were incubated for at least 18 h at room temperature in streptavidin conjugated to horseradish-peroxidase (Amersham Buchler, Brunswick, Germany) at a dilution of 1:200 in phosphate-buffered saline containing 0.5% Triton X-100. Sections were stained with 3,3′-diaminobenzidine tetrahydrochloride and nickel ammonium sulfate as described elsewhere(Vitzthum et al., 2002). Finally, the sections were mounted on glass microslides, dehydrated, cleared in xylene, and embedded in Entellan (Merck, Darmstadt, Germany) under glass coverslips. Neurons were reconstructed using a compound microscope with camera lucida attachment. The terms ipsilateral and contralateral refer to the position of the cell body.

We recorded from 32 polarization-sensitive interneurons with arborizations in the AOTu. Of these, 22 stable recordings were selected for analysis. All recordings could be classified as being from either LoTu1 (14 recordings) or TuTu1 (8 recordings) types of heterolateral interneurons of the AOTu described previously (Pfeiffer et al.,2005). The perikarya of both cell types lie in the inferior lateral protocerebrum.

Recordings from LoTu1 confirmed our previous findings on responses to polarized and unpolarized white light (Fig. 1) (Pfeiffer et al.,2005). Dorsal polarized light led to tonic excitation that was sinusoidally modulated in strength by the rotating e-vector(Fig. 1B). Unpolarized ipsilateral light led to an increase in spiking activity, while zenithal dorsal stimulation led to tonic inhibition. Strong frontal illumination caused an excitatory response or a phasic excitation followed by strong tonic inhibition (Fig. 1C).

In the present study we analyzed the responses of LoTu1 to monochromatic light stimuli. Dorsal polarized monochromatic light stimuli (UV, 330 nm; blue,450 nm; green, 530 nm) at 9.0×10⁹ photons cm^–2 s^–1 were tested in seven recordings from LoTu1. The background activity of these neurons was relatively low at 4.82±1.09 impulses s^–1 (mean ± s.e.m.). The excitatory response of LoTu1 to polarized blue light was stronger than the response to polarized UV and green light(Fig. 2A,B). In four recordings, green light elicited virtually no response(Fig. 2A), but in three other LoTu1 cells, the response to polarized UV light was smaller than the response to polarized green light. The mean spiking activity at Φ_maxduring stimulation with polarized blue light was about two times the activity at Φ_max when stimulating with UV light or green light(Fig. 2B). The mean spiking activity at Φ_max in response to polarized UV light did not differ significantly from the response to polarized green light. The response strength R (for definition of R, see Materials and methods)showed the same results (Fig. 2B). Fig. 2C,Dshows the intensity/response curves from six LoTu1s (mean background activity,8.65±1.82 impulses s^–1). At the lowest light intensity of logI=–4(10.6×10⁸ photons cm^–2 s^–1) LoTu1 did not show clear responses. The activity at Φ_max saturated at intensities between logI=–2 and –1.5, the response value R,in contrast, only near logI=0. In one LoTu1, the response value R at maximum intensity was smaller than the response strength at logI=–3 and –2.

The responses of LoTu1 to unpolarized monochromatic lights (UV, 350 nm;blue, 430 nm; green, 530 nm) were different depending on eye region and wavelength (Fig. 3). The responses to unpolarized light at the lowest light intensity(logI=–3) were quite small. Clear responses to unpolarized light stimuli were observed at intensities above logI=–2. LoTu1 was inhibited by blue light from the dorsal and from the contralateral side(Fig. 3A, arrows). At intensities of logI=–1 and logI=0, LoTu1 was inhibited by both UV and blue light. These inhibitions were followed by post-inhibitory rebound excitations after the offset of the stimulus(Fig. 3A, open arrowheads). The strength of these rebound excitations depended on the light intensity of the stimulus (Fig. 3A). The inhibitory responses to dorsal light stimulation were stronger than those to contralateral stimulation. When light stimuli were applied ipsilaterally,LoTu1 was inhibited by UV light (Fig. 3A, arrowhead, Fig. 4B) but was excited by green light(Fig. 3A, double arrowhead, Fig. 4B). At logI=0,however, one of the neurons was excited by ipsilateral UV and green light(Fig. 3A). This reversal in response to UV was possibly caused by masking of UV inhibition by a strong excitatory response to green.

The responses of LoTu1 to unpolarized light were consistent among four to six LoTu1s (Fig. 3B). When stimulated dorsally, all LoTu1s were inhibited by blue light. In response to ipsilateral stimulation, four LoTu1s were excited by green and were inhibited by UV light as shown in Fig. 3A. One of two other recordings from LoTu1s with no background activity showed an excitatory response to green light and no response to UV light. The other LoTu1 showed weak excitatory responses to UV light and green light. When light stimuli were presented from the contralateral side, three LoTu1s were inhibited by blue light, but the mean spiking activity during stimulation was not significantly lower than background activity. During ipsilateral stimulation with a series of nine monochromatic lights, LoTu1 showed clear spectral opponency (Fig. 3C). Stimulations at short wavelengths, from 330 nm to 430 nm,inhibited the neuron, whereas stimulations at long wavelengths, from 500 nm to 550 nm, excited the neuron.

In one LoTu1 neuron, we successfully recorded the responses to polarized blue light and to unpolarized lights, adjusted to the same intensities(Fig. 4). LoTu1 (background activity 5.23 impulses s^–1) started to be inhibited by unpolarized dorsal blue light at logI=–1.5(Fig. 4B, arrow). This inhibitory response increased with increasing light intensity. The post-inhibitory rebound excitation after the stimulus also increased depending on stimulus intensity (Fig. 4B,open arrowhead). The neuron was inhibited by unpolarized ipsilateral UV light(Fig. 4B, arrowhead). In contrast to the blue-light inhibition, this inhibitory response outlasted the stimulus and led to complete inhibition by UV light above a light intensity of logI=–1 (Fig. 4B, double open arrowhead). The neuron only gradually recovered to background spiking after more than 3 s following stimulation at highest intensity. The response to ipsilateral green light became apparent at a light intensity of logI=–1.5(Fig. 4B, double arrowhead). Comparison of the response/intensity curves for responses to unpolarized lights and to dorsal polarized light (Fig. 4C) shows that sensitivity to polarized light is already present at intensities below logI=–3. In contrast, clear responses to unpolarized light only occurred above a light intensity of logI=–1. These results show that LoTu1 is about 2.5 log units more sensitive to polarized light than to unpolarized light. The dynamic range of intensity coding for polarized and unpolarized light was very narrow and covered only about 1–1.5 log units.

We recorded the responses to dorsal polarized monochromatic UV, blue and green light stimuli from three TuTu1 neurons. The mean background activity of these neurons was 21.03±6.24 impulses s^–1,considerably higher than the activity of LoTu1. TuTu1 neurons showed polarization-opponency in response to the rotating polarizer (see also Pfeiffer et al., 2005). This means that TuTu1 neurons were maximally excited at Φ_max and were maximally inhibited at an e-vector orientation orthogonal toΦ _max (Φ_min). The opponent response to polarized blue light was stronger than the responses to polarized UV and polarized green light (Fig. 5A,B). In two neurons, the response to polarized UV light was slightly stronger than that to polarized green light (Fig. 5A). The response strength R to polarized light was significantly higher at 450 nm (blue) than at 350 nm (UV) or at 530 nm (green)(Fig. 5B). In contrast, the neural activity at Φ_max was not significantly different between the responses to the three monochromatic polarized lights. Fig. 5C,D shows the intensity/response curves of three recordings from TuTu1. The neural activity at Φ_max and the response amplitude increased with increasing stimulus intensity. TuTu1 did not respond to polarized light at logI=–4. At a light intensity of logI=–2, both the activity at Φ_max and the response saturated.

We recorded the responses to unpolarized monochromatic lights from four TuTu1 neurons, but the responses were not consistent among the recordings. Fig. 6 shows two examples of the responses to unpolarized monochromatic lights applied to different eye regions. In Fig. 6A, TuTu1 showed spectral-opponency when light stimuli were applied from the ipsilateral direction. This TuTu1 was excited by UV light(Fig. 6A, double arrowhead) and was inhibited by green light (Fig. 6A, open arrowheads). When contralateral light stimuli were applied, the neuron showed an inhibitory response(Fig. 6A, arrow) with post-rebound excitation (Fig. 6A, open double arrowheads) to blue light. Another TuTu1 was inhibited by blue light at maximum intensity from dorsal and contralateral directions (Fig. 6B, arrows). This neuron was excited by both UV light and blue light coming from the ipsilateral side (Fig. 6B,arrowheads), but showed no clear response to green light. The third TuTu1 (not shown) was excited by UV light presented dorsally and ipsilaterally, but did not respond to blue light and green light from any directions. The last TuTu1 showed very weak responses that were similar in their properties to those of the neuron of Fig. 6A.

We have analyzed the spectral and polarization properties of two identified polarization-sensitive neurons, LoTu1 and TuTu1, in the brain of the desert locust. For unbiased signalling of e-vector orientations,polarization-sensitive interneurons in other species, such as the cricket,were shown to receive exclusive input from homochromatic photoreceptors(Labhart, 1988). In contrast,both LoTu1 and TuTu1 neurons of the locust receive input from homochromatic blue polarized-light sensitive photoreceptors (Figs 2, 5, 7), as well as from polarization insensitive UV and green receptors (Figs 3, 4, 6, 7). Firm conclusions on the biological significance of these wavelength-specific responses await further studies. An attractive hypothesis, however, is that through chromatic contrast signalling spectral gradients in the sky might contribute to sky compass coding in these two neurons.

In the DRA of the desert locust, UV receptors with low polarization sensitivity and blue receptors with high polarization sensitivity have been detected (Eggers and Gewecke,1993). LoTu1 and TuTu1 were most sensitive to dorsally presented polarized blue light and showed much lower sensitivity to polarized UV and green light (Fig. 2A,B, Fig. 5A,B). This result is consistent with the fact that only blue receptors in the locust DRA showed high polarization sensitivity (Eggers and Gewecke, 1993) and indicates that the polarized-light sensitivity of LoTu1 and TuTu1 neurons is based on blue receptors in the DRA. The polarization vision system of crickets, likewise, depends on blue photoreceptors, demonstrated behaviorally, in photoreceptor recordings, and in intracellular recordings from POL interneurons(Labhart, 1988) (reviewed by Labhart and Petzold, 1993). In contrast, polarization vision in hymenopteran species (honeybee, desert ant)relies on UV light (reviewed by Wehner and Labhart, 2006).

The absolute sensitivity for polarized light is similar in LoTu1 and TuTu1. The response threshold for polarized light was at a light intensity of logI=–3.5 to –3, and saturation of the response atΦ _max was reached at an intensity of logI=–2(Fig. 2C, Fig. 5C). This means that 1–1.5 log units above threshold, the responses of both POL neurons atΦ _max are intensity independent. While the response strength R of TuTu1 showed a similar intensity dependence(Fig. 5D), the R value of LoTu1 increased over 3–4 log units of light intensity and only reached saturation around logI=0(Fig. 2D). A likely reason for this may be the increasing contribution of a polarization-insensitive inhibition of LoTu1 by dorsal blue light above logI=–2, as shown in Fig. 4C. The cricket POL-1 neuron, in contrast, shows maximum response within 1 log unit of light intensity (Labhart, 1988; Labhart et al., 2001). Above a light level of about 3×10⁸ photons cm^–2s^–1 of blue light (443 nm), its e-vector response becomes intensity independent by receiving antagonistic input from photoreceptors with mutually orthogonal microvilli orientation. The POL-1 neuron is, therefore, at least 2 log units more sensitive to polarized light than the two locust neurons studied here.

The spectral responses of the bilateral POL neurons to unpolarized light are surprisingly complex. All three types of spectral receptors in the compound eye contribute to the unpolarized light response. The spectral responses are different at different stimulus positions, indicating that the set of spectral inputs contributing to the unpolarized light responses differ considerably depending on the eye region(Fig. 7).

LoTu1 and one subtype of TuTu1 receive inhibitory input from blue receptors in dorsal eye regions (Fig. 3, Fig. 6). The sensitivity to dorsally presented blue light may code for brightness of the blue sky. At very high light intensities, the inhibition in LoTu1 by blue light can strongly suppress the responses to polarized light and may even become apparent when using polarized light as the stimulus.

In addition, the spectral response to unpolarized light shows opponency in the response to UV light and green light in both bilateral POL neurons when light was presented from the ipsilateral side(Fig. 3, Fig. 6). Spectral opponency is a widespread phenomenon in color vision and has been demonstrated in neurons of the optic lobe of the honeybee (Kien and Menzel, 1977) and migratory locust(Osorio, 1986). Sustaining responses and narrow receptive fields of some green-UV color opponent neurons of the locust medulla suggested that they might play a role in maintaining flight posture relative to the horizon(Osorio, 1986). Likewise, the spectral opponency in LoTu1 and TuTu1 may not contribute to true color vision,but may rather serve to evaluate the spectral gradient in the sky. At positions near the sun, the chromatic contrast between long (green) and short(UV) wavelength light is high, but becomes smaller with increasing angular distance to the sun in the anti-solar hemisphere(Rossel and Wehner, 1984; Coemans et al., 1994). To process all spectral information in the sky, an animal may code the intensity of blue light as a reference and, at the same time, the difference of intensities between green light and UV light. To further substantiate the hypothesis that LoTu1 and TuTu1 neurons integrate polarization and chromatic contrast of the sky, it will be necessary, however, to examine the azimuthal dependence and receptive fields of the chromatic responses in more detail and to test the combined effects of polarized and chromatic stimuli on the responses of the neurons.

Which neurons provide input to the bilateral POL neurons? Previous anatomical studies have shown that medulla line tangential neurons are the most promising candidates to provide visual input from the ipsilateral eye(Homberg et al., 2003). These neurons have small diameter processes along the dorso-ventral axis of the medulla – in addition to ramifications in the dorsal rim area of the medulla – and send direct processes to the lower unit of the AOTu. This morphology is ideally suited to integrate inputs from the DRA and the main retina of the compound eye. In both LoTu1 and TuTu1, ramifications in the ipsilateral AOTu are of smooth appearance and, therefore, most likely dendritic, while arborizations in the contralateral AOTu have a beaded or varicose appearance and are, therefore, most likely axonal(Pfeiffer et al., 2005). These morphologies suggest that the bilateral POL neurons receive information only from the ipsilateral eye. In fact, LoTu1 receives polarized-light input only through the ipsilateral eye, while polarization sensitivity in TuTu1 is dominated by ipsilateral eye input(Pfeiffer et al., 2005). In addition, our present study suggests that TuTu1 at least responds to unpolarized light perceived by both eyes. The information from the contralateral eye may in part originate from the counterpart LoTu1 and TuTu1 neurons of the other brain hemisphere. In addition, medulla line tangential neurons without ramifications in the DRA also project to the lower unit of AOTu (U.H., unpublished observation). These neurons may provide selectively unpolarized light inputs to the bilateral POL neurons in the AOTu.

The bilateral POL neurons may be suited to integrate information on the celestial polarization pattern and on the spectral gradient in the sky. Polarized light information at the zenith is used for coding the orientation of the body axis relative to the solar meridian. However, sky polarization alone is not sufficient to signal solar azimuth unambiguously, since e-vector orientations alone do not allow the animal to discriminate whether the sun occurs at an azimuth φ or at an azimuth φ+180°. POL neurons in the locust, as reported here, might resolve this problem by coding not only the e-vector angle of polarized light but also the spectral gradient in the sky. Locusts may, in addition, use intensity gradients in the sky for orientation, but whether and how intensity information is actually integrated in the compass orientation system remains to be seen. Behavioral experiments in homing bees have shown that intensity gradients do not provide substantial information for navigation (Rossel and Wehner, 1984).

The absolute sensitivity to polarized light is about 2 log units higher than the sensitivity to unpolarized light, as shown for LoTu1(Fig. 4C). This suggests that,depending on light intensities, polarization input or unpolarized light input may dominate the responses of the neurons. At low light intensities, before dawn and after sunset or under a partly cloudy sky, the polarization vision system may be more important for orientation behavior. However, when the sky is clear and the sun is directly visible, sky chromatic contrast may prevail and provide the relevant information for orientation.

This study is the first to address the possibility that both e-vector angle of polarized light and the spectral gradient in the sky are encoded in the same neural system underlying compass orientation in an insect. Our results support behavioral data indicating that celestial compass orientation in bees and desert ants relies on both the celestial polarized-light pattern and the spectral gradient in the sky (Wehner,1989). Observations in bees, ants, and pigeons suggest that these animals evaluate the intensity of long wavelength light with respect to the relatively isotropic UV background for sun compass navigation(Rossel and Wehner, 1984; Coemans et al., 1994; Wehner, 2003). Our findings of spectral opponency in the responses to unpolarized light fit these behavioral observations well and may be a first step in understanding how the integration of different celestial cues used for spatial orientation is organized in the brain.

 
• A

  spiking activity

 
• AOTu

  anterior optic tubercle

 
• DRA

  dorsal rim area of the compound eye

 
• I

  stimulus intensity

 
• K

  intensity eliciting 50% rA_max, resp. rR_max

 
• LoTu1

  lobula-tubercle neuron 1

 
• n

  exponent

 
• PB

  phosphate buffer

 
• POL

  polarized-light sensitive

 
• PS

  polarization sensivity

 
• R

  response value

 
• rA

  relative spiking activity

 
• rR

  relative response

 
• TuTu1

  tubercle-tubercle neuron 1

 
• UV

  ultraviolet

 
• Φmax, Φmin

  e-vector angle eliciting maximal/minimal spiking activity

We thank K. H. Herklotz for assistance with insect rearing, and Dr P. Galland for access to his radiometer. This work was supported by DFG grant HO 950/16-1 to U. Homberg and a fellowship by the Canon Foundation in Europe to M. Kinoshita.