ELSEVIER Arthropod Structure & Development 35 (2006) 293-305 ARTHROPOD STRUCTURE & DEVELOPMENT www.elsevier.com/locate/asd Ocellar optics in nocturnal and diurnal bees and wasps Eric J. Warranta'*, Almut Kelber a, Rita Wallen a, William T. Wcislo b " Department of Cell & Organism Biology, Zoology Building, University of Lund, Helgonavagen 3, S-22362 Lund, Sweden Smithsonian Tropical Research Institute, Apartado 0843-03092 Balboa, Panama Received 1 July 2006; accepted 2 August 2006 Abstract Nocturnal bees, wasps and ants have considerably larger ocelli than their diurnal relatives, suggesting an active role in vision at night. In a first step to understanding what this role might be, the morphology and physiological optics of ocelli were investigated in three tropical rainforest species ? the nocturnal sweat bee Megalopta genalis, the nocturnal paper wasp Apoica pallens and the diurnal paper wasp Polistes occidentalis ? using hanging-drop techniques and standard histological methods. Ocellar image quality, in addition to lens focal length and back focal distance, was determined in all three species. During flight, the ocellar receptive fields of both nocturnal species are centred very dorsally, possibly in order to maximise sensitivity to the narrow dorsal field of light that enters through gaps in the rainforest canopy. Since all ocelli investigated had a slightly oval shape, images were found to be astigmatic: images formed by the major axis of the ocellus were located further from the proximal surface of the lens than images formed by the minor axis. Despite being astigmatic, images formed at either focal plane were reasonably sharp in all ocelli investigated. When compared to the position of the retina below the lens, measurements of back focal distance reveal that the ocelli of Megalopta are highly underfocused and unable to resolve spatial detail. This together with their very large and tightly packed rhabdoms suggests a role in making sensitive measurements of ambient light intensity. In contrast, the ocelli of the two wasps form images near the proximal boundary of the retina, suggesting the potential for modest resolving power. In light of these results, possible roles for ocelli in nocturnal bees and wasps are discussed, including the hypothesis that they might be involved in nocturnal homing and navigation, using two main cues: the spatial pattern of bright patches of daylight visible through the rainforest canopy, and compass information obtained from polarised skylight (from the setting sun or the moon) that penetrates these patches. ? 2006 Elsevier Ltd. All rights reserved. Keywords: Ocellus; Nocturnal vision; Bee; Wasp; Optics; Resolution; Sensitivity 1. Introduction Bees and wasps are highly visual insects that rely on their compound eyes for a variety of behavioural tasks, including the discrimination of flower shape and colour, the identifica- tion of prey, the extraction of compass information from polar- ised skylight, the recognition of learned terrestrial landmarks, and the stabilisation and control of flight using information derived from the optical flow field. Nearly all species of bees and wasps execute these tasks exclusively in bright daylight. However, in the world's tropical rainforests several Corresponding author. Tel.: +46 46 222 9341; fax: +46 46 222 4425. E-mail address: eric.warrant@cob.lu.se (E.J. Warrant). species provide notable exceptions to this rule. Due to the pressures of predation and competition for limited food resources, some groups of bees and wasps have evolved a noc- turnal lifestyle (Cockerell, 1923; Richards, 1978; Roubik, 1992; Wcislo et al., 2004). Remarkably, despite experiencing light levels 100 million times dimmer than their day-active relatives, these species are still able to forage and orient using visual cues (Warrant et al., 2004). All hymenopteran insects ? both diurnal and nocturnal ? possess apposition compound eyes, an eye design that is better suited to vision in bright light. Apposition eyes are found in most day-active insects, including flies, butterflies and dragon- flies, but their poor sensitivity to light makes them rare in nocturnal insects with demanding visual requirements. Such nocturnal insects typically have considerably more sensitive 1467-8039/$ - see front matter ? 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.asd.2006.08.012 294 EJ. Warrant et al. I Arthropod Structure & Development 35 (2006) 293?305 superposition compound eyes. The apposition eyes of the noc- turnal sweat bee Megalopta genalis, an insect capable of learn- ing visual landmarks at starlight intensities, have thus generated considerable interest. However, despite being 30 times more sensitive to light than those of diurnal bees (Greiner et al., 2004a; Warrant et al., 2004), Megaloptds appo- sition eyes are still insufficiently sensitive on their own to explain the bee's impressive nocturnal visual abilities. Recent work has shown that nocturnal vision in Megalopta can be ex- plained if neural summation of visual signals in space and time occurs at a higher level in the visual system, a hypothesis that is supported by anatomical and theoretical evidence (Greiner et al., 2004b, 2005; Theobald et al., 2006). In addition to the compound eyes, many insects ? includ- ing bees and wasps ? possess two or three conspicuous ocelli on the dorsal surface of the head between the eyes. Ocelli are single-lens eyes of the camera type, and their exact role is still a matter of conjecture. They most likely have different roles in different insects, and there is now good evidence that they sup- port a variety of behavioural tasks including flight stabilisa- tion, navigation and orientation, absolute intensity measurement and neurosecretion (for reviews see Goodman, 1981; Wehner, 1987; Mizunami, 1994). The properties of ocelli are well suited to these tasks. Compared to the com- pound eyes the ocellar pathways are fast, with large neurons and few synapses, ensuring that information from the ocelli can reach thoracic motor centres rapidly (Guy et al., 1979). This high temporal resolution is, however, in stark contrast to their spatial resolution. Optical measurements of back focal distance in a variety of insects (Homann, 1924) ? particularly in flies (Cornwell, 1955; Schuppe and Hengstenberg, 1993) and locusts (Parry, 1947; Comwell, 1955; Wilson, 1975) - have all indicated that the plane of best focus of the ocellar lens lies a considerable distance behind the retina. This under- focusing ensures that the photoreceptors receive a very blurry image. Moreover, ocelli typically have a slightly oval shape that can cause astigmatism, degrading the image even further (Schuppe and Hengstenberg, 1993) ? images focused by the long (major) axis of an oval ocellus are likely to be located further from the back surface of the lens than images focused by the short (minor) axis. Even if a perfectly crisp image could be focused on the distal photoreceptor tips, the wiring of the retina would still prevent good spatial resolution. Firstly, the arrangement of photoreceptors in the ocellar retina is typically disorganised, and not arranged in the tight sampling matrix that might be expected for high spatial resolution. Secondly, the outputs of these photoreceptors are then summed within large groups by second-order L-neurons ? in the extreme case of cockroaches, for instance, over 10000 photoreceptors synapse onto just four L-neurons (Toh and Tateda, 1991), ef- fectively reducing the visual space sampled by the ocellus to just four image pixels. Thus, most ocelli so far studied are coarse, fast and sensi- tive organs of vision, well suited for rapidly signalling changes in light intensity but not for resolving fine spatial detail. These properties ? in combination with their broad partially-overlap- ping visual fields covering the forward horizon and the dome of the sky (Wilson, 1975) ? make them ideal for detecting the changes of light intensity that would be experienced when in- sects change pitch or roll during flight, a task now considered to be their most common function. However, as Mizunami (1993, 1995) rightly points out, the structure and wiring of ocellar systems is so varied within the insects, that many other roles are likely. A case in point is the recently described me- dian ocellus of dragonflies (Stange et al., 2002; van Kleef et al., 2005; Berry et al., in press). Optical and electrophysio- logical measurements have shown that these ocelli are capable of resolving the sharp boundary between sky and land, rather than simply the brightening and darkening associated with se- quentially viewing one after the other. The optics of the me- dian ocellar lens provides an image that has good spatial resolution in the vertical direction (Stange et al., 2002), and this resolution is preserved by both the photoreceptors (van Kleef et al., 2005) and the second-order L-cells to which they connect (Berry et al., in press). These features allow an accurate one-dimensional analysis of the horizon over a wide angular azimuth, and this may be used for fine control of flight attitude. What role might the ocelli play in a nocturnal bee or wasp that flies in a cluttered rainforest environment where the hori- zon is obscured? Kerfoot (1967) has shown that the ocelli of nocturnal bees are larger than those of crepuscular bees, and that these in turn are larger than those of diurnal bees, suggest- ing that ocelli may play an important role in dim light. Indeed, the ocelli of bumblebees allow them to navigate using polar- ised skylight at dusk when terrestrial landmarks have become too dim for the compound eyes to recognise (Wellington, 1974). Can the ocelli of a nocturnal bee or wasp help to stabi- lise flight in a cluttered rainforest at night, or to aid in naviga- tion? A first step in answering this question is to study the imaging properties of their ocelli. Are they capable of spatial resolution, like the median ocellus of the dragonfly? Or are they simply highly sensitive light detectors? The goal of the present paper is to answer these questions by studying the large ocelli of the nocturnal sweat bee M. genalis (Halictidae) and the nocturnal paper wasp Apoica pollens (Vespidae). Both species are active at night in the tangled tropical rainforest, and both have demanding visual behaviours, including hom- ing. As a comparison, the ocelli of a close diurnal relative to Apoica ? the day-active tropical paper wasp Polistes occiden- talis (Vespidae) ? will also be studied. 2. Materials and methods 2.1. Animals All wasps and bees used in this study were collected on Barro Colorado Island in the Panama Canal (a tropical rainfor- est field station of the Smithsonian Tropical Research Institute, Panama City). Nocturnal polistine paper wasps (A. pallens) and halictid bees (M. genalis) were collected at night in the rainforest on white sheets illuminated by ultraviolet-enriched light. The diurnal halictid bee shown in Fig. 2 was kindly E.J. Warrant et al. I Arthropod Structure & Development 35 (2006) 293?305 295 loaned to us by Dr. David Roubik (Smithsonian Tropical Re- search Institute), who tentatively identified the species as Au- gochloropsis (Paraugochloropsis) cf. Fuscognatha. Diurnal paper wasps (P. occidentalis) were collected from nests built under the eaves of research buildings on the island. 2.2. Histology Light microscopy, and transmission and scanning electron microscopy, were performed using standard methods. Whole eyes were placed for 2 h at 4 ?C in standard fixative (2.5% glu- teraldehyde and 2% paraformaldehyde in phosphate buffer (pH 7.2)). Following a buffer rinse, eyes were then added to 2% OsOj for 1 h. Dehydration was performed in an alcohol se- ries and eyes were embedded in Araldite. Ultrathin sections for electron microscopy were stained with lead citrate and ura- nyl acetate. Thin (3.5 um) sections for light microscopy were stained with toluidine blue. Rhabdom cross-sectional areas were calculated from light and electron microscope sections using NIH Image. 2.3. Determination of ocellar receptive field centres The small end was cut from a plastic pipette tip leaving an opening large enough for a bee or wasp head to protrude through. The insect was fixed in position by gluing the mouth- parts to the tube with dental wax, and this preparation was then mounted at the centre of curvature of a Leitz goniometer. The goniometer was placed onto the foot-plate of an Askania ma- croscope. The insect was then manipulated so that the long axis of the eye edge was parallel to the plane of the stage. The head was further manipulated so that (1) the origin of the three goniometer axes was in the centre of the head, and (2) the three goniometer axes were lined up with the dor- sal?ventral (yaw), anterior?posterior (roll), and left?right (pitch) axes, respectively, of the insect's head. With the stage horizontal, the head then looked vertically upwards into the objective of the macroscope, and when observed in this posi- tion, the head was oriented exactly anteriorly (from the ani- mal's point of view). The goniometer allowed us to tilt the stage (and thus the head) in defined angular steps of latitude and longitude, with latitude = 0? and longitude = 0? defined as the anterior orientation described above ("Ap" in Fig. 5A). Dorsal ("Dp") corresponds to a latitude of +90? and lateral ("Lp") to a latitude of 0? and a longitude of ?90?. To illuminate the eyes we introduced a half-silvered mirror, angled at 45?, just beneath the objective of the macroscope. Collimated white light (from a halogen source) was directed laterally to the mirror so that the eyes were illuminated and viewed along the same axis ("orthodromic illumination"). A bright spot of light ? the reflection of the white halogen source ? was then visible on the surface of the ocellus. To determine ocellar receptive field centres, the goniometer stage was tilted both in longitude and latitude until the reflected spot was judged to be at the absolute centre of the ocellus. The coordinate of latitude and longitude required to do this was taken as the ocellar receptive field centre. 2.4. Optical measurements of ocellar image quality The back focal distances and focal lengths of ocellar lenses were measured using a modification of Homann's (1924) hanging-drop method. A small piece of cuticle containing ei- ther a lateral or median ocellus was carefully dissected from the head capsule, placed in a petri dish of saline and lightly cleaned from tissue and pigment using a small paintbrush. It was then placed external side outwards in a tiny drop of phys- iological saline (refractive index = 1.34) that was placed on the centre of a microscope cover slip. An o-ring was waxed to a conventional microscope glass, after which the upper sur- face of the o-ring was lightly greased with Vaseline. The cover slip was then turned upside down and placed onto the greased o-ring, thus creating an air-tight chamber containing the saline drop and its downward pointing ocellus. The microscope slide was mounted on the stage of a conventional light microscope (Leica) with condenser removed. Objects of known size (typ- ically patterns of dark stripes on translucent tracing paper) were placed on the foot of the microscope, over the lamp ap- erture. Images of these objects were focused by the ocellus within the saline drop. These images were then viewed with the 40 x objective, and photographed with a digital camera fit- ted to the microscope. The focal length/of each ocellus was calculated according to the following equation: f = So- li (1) where s0 is the distance between the striped object and the ocellus (127 mm), \0 is the spatial wavelength of the striped pattern (the distance between the centre of one stripe and the centre of the next: 4.53 mm) and X; is the spatial wave- length of the image of the striped pattern (mm). The optical back focal distance ? the distance from the back of the ocellar lens to the plane of best focus ? was mea- sured by first focusing upon small particles of debris attached to the back of the lens. The back focal distance was deter- mined by focussing upwards until the best image of the striped object was obtained. The change in focus (in micrometres) was measured using a micrometer gauge attached to the mi- croscope stage. This procedure was repeated at least 10 times and the values averaged. This mean value was corrected for the refractive index of the saline by multiplication by 1.34. Images collected from wasp and bee ocelli were found to be astigmatic: images focused by the long (major) axis of the oval ocellus were found to be located further from the back surface of the lens than images focused by the short (mi- nor) axis. Back focal distances were calculated for both image planes by using patterns whose stripes were either perpendic- ular (far image plane) or parallel (near image plane) to the long axis of the ocellus. 296 3. Results EJ. Warrant et al. I Arthropod Structure & Development 35 (2006) 293?305 3.2. Internal ocellar morphology 3.1. External ocellar morphology and location All bees and wasps have three ocelli on the dorsal head cap- sule between the eyes (Figs. 1?4). The ocelli of all species have a slightly oval shape (Table 1), although the median ocel- lus of Apoica is almost round (Fig. 3C, Table 1). The ocelli of nocturnal species appear to bulge prominently from the dorsal surface of the head (Figs. 1 and 3), whereas those of diurnal species are less conspicuous (Figs. 2 and 4). As previously de- scribed, nocturnal species have larger ocelli (Figs. 1 and 3) than diurnal species (Figs. 2 and 4), suggesting an active role in dim light. As a percentage of head diameter, the ocelli of nocturnal Apoica (Fig. 3) and Megalopta (Fig. 1) are about twice the size (12?14%) of those of diurnal Polistes (Fig. 4) and Augochloropsis (Fig. 2) (6?8%). The median ocellus of an immobilised specimen of the noc- turnal bee Megalopta has the centre of its receptive field centred frontally, about 10? above anterior on the equator of the eye ("Ap" in Fig. 5A), where "anterior" is defined as the direction perpendicular to the physical long axis of the head and compound eyes (see Materials and methods). The co- ordinate of the median ocellus receptive field centre (latitude, longitude) is thus [+10?, 0?]. The two lateral ocelli have re- ceptive fields centred 50? above the equator, and 60? laterally on either side (coordinates [+50?, +60?] and [+50?, -60?]). Immobilised specimens of Apoica reveal a distinctly more dor- sal placement of the ocelli (Fig. 5A), with receptive field cen- tres located at [+25?, 0?], [+60?, +100?] and [+60?, -100?] for the median and two lateral ocelli, respectively. Immobilised specimens, however, are unable to reveal the true orientations of the ocelli in a flying animal ? if the head is tilted upwards or downwards during flight, the recep- tive field centres shown in Fig. 5A will likewise be tilted. To account for this possibility, film sequences of flying Mega- lopta were analysed (Kelber, unpublished data), and head an- gles relative to true horizontal ("At" in Fig. 5B) were calculated for a number of bees. In free flight Megalopta tilts its head upwards by 52?, meaning that its ocelli have a con- siderably more dorsal orientation than indicated in Fig. 5A (see Fig. 5B). This substantial head tilt during flight is typ- ical of many bees (Kelber, unpublished data). Even though we did not have the opportunity to film Apoica in free flight, inspections of photographs of two species of wasps in flight ? Sceliphron caementarium (Sphecidae) and Polistes metri- cus (Vespidae) ? reveal a considerably lower upward head tilt: 9.3? and 11.4?, respectively (Dalton, 1975). If we as- sume that Apoica employs a similar head tilt during flight (say 10?), then the receptive field centres of its two lateral ocelli have a very similar dorsal location to those of the lat- eral ocelli in flying Megalopta (Fig. 5B): around 50? above the horizon, and somewhat posterior (longitudes of ca. ?120?). Interestingly, the receptive field centre of the me- dian ocellus in Apoica is almost 30? more anterior than the receptive field centre of the median ocellus in Megalopta (Fig. 5B). The median ocellar lens in all three species overlies a retina of photoreceptors (Fig. 6). In the two nocturnal species, the distal retinal surface is positioned close to the proximal sur- face of the lens, and in Megalopta these are directly in contact (Fig. 6B). In Apoica the distal retinal surface is concentric with the proximal surface of the lens, with these surfaces being separated by an approximately 35 urn depth of corneageal cells (Fig. 6A). In the diurnal wasp Polistes, the retina is not concentric with the proximal lens surface (Fig. 6C). Unlike Polistes, neither Megalopta nor Apoica possesses screening pigments in the retina, an adaptation likely to improve light capture at night and thus sensitivity. In cross-section, the rhabdoms of the retina do not form an orderly matrix, but form an array that is somewhat disorganised. Rhabdoms in all species are highly elongated, and constructed from the rhabdomeres of two retinula cells whose dominant mi- crovillar directions are practically parallel to each other (Fig. 7). The rhabdoms of the nocturnal bee Megalopta are the largest of three species studied (17.5 x 1.3 urn: Table 1), followed by the rhabdoms of the nocturnal wasp Apoica (5.9 x 0.6 urn) and the diurnal wasp Polistes (4.6 x 0.6 um). The large rhabdoms of Megalopta are clearly an adaptation for improved sensitivity at night, and the interesting observation that the equally nocturnal Apoica has small rhabdoms suggests that some of their sensitivity may have been offered in favour of spatial resolution. This conclusion is reinforced by calculations of the occupation ratio of rhabdoms in the retina (the percentage cross-sectional area of rhabdoms within a unit cross-sectional area of retina). The occupation ratios of the wasps Apoica and Polistes are quite small (9.5% and 6.1%, respectively: Table 1). Even though Apoica is nocturnal, its rhabdom occupation ratio is only slightly larger than that of its diurnal relative. In contrast, the nocturnal bee Megalopta has a rhabdom occupation ratio over three times greater (33.6%: Table 1). 3.3. The optical properties of ocelli Images formed behind the lenses of the ocelli in all three species were astigmatic (Fig. 8, Table 1): images formed by the major axes of the oval ocelli (b, d, f in Fig. 8) were located further from the proximal surface of the lens than images formed by the minor axes (a, c, e in Fig. 8). Despite being a- stigmatic, images formed at either focal plane were reasonably sharp in all ocelli investigated. Ocellar focal lengths/were determined for each of the two astigmatic focal planes: /maj (for the ocellar major axis) and /min (for the ocellar minor axis). Their values in the lateral and median ocelli do not differ significantly in any of the three species (Table 1). Values of /maj for all ocelli are in the range 510?690 um in Megalopta, 430?460 um in Apoica and 180?220 um in Polistes. Corresponding values of /,?? are 370-600 um, 300-420 um and 150-200 um, respectively. Significant variations in measured focal lengths for median or lateral ocelli were found in Megalopta, but the variation was in most cases not as great in the two wasps (Table 1), E.J. Warrant et al. I Arthropod Structure & Development 35 (2006) 293?305 297 _-. .-::? Fig. 1. The ocelli of the nocturnal halictid bee Megalopta genalis, showing their position on the head (A) and their arrangement as a group (B). (C) The median ocellus. Scales: 1 mm (A), 200 u.m (B) and 100 um (C). Fig. 2. The ocelli of the diurnal halictid bee Augochloropsis (Paraugochlorop- sis) cf. Fuscognatha. Descriptions and scales as for Fig. 1. 2% EJ. Warrant et al. I Arthropod Structure & Development 35 (2006) 293?305 Fig. 3. The ocelli of the nocturnal paper wasp Apoica pallens. Descriptions and scales as for Fig. 1. Fig. 4. The ocelli of the diurnal paper wasp Polistes occidentals. Descriptions and scales as for Fig. 1. E.J. Warrant et al. I Arthropod Structure & Development 35 (2006) 293?305 299 Table 1 Optical and anatomical dimensions in the lateral and median ocelli Parameter Megalopta Apoica Polistes Median (3) Lateral (5) Median (3) Lateral (2) Median (2) Lateral (2) *min 410 ?60 398?63 416? 16 414 216?12 161 ?18 <2>maj 471 ?40 457 ? 66 427 ?9 486 273 ?8 218 ?21 "W^head 0.120 0.117 0.122 0.139 0.070 0.056 /mm 427 ? 59 500 ?106 357 ? 63 331 ?4 168 ?13 185 ?15 /maj 644 599?88 436 450 ?8 213 ?9 190 ?8 'mill 409 ? 14 447 ? 140 162 ?14 184 ?7 100 ?8 120?11 'maj 565 548 ?108 190 304 ? 25 170 ?19 185 ?15 (Lin 1.3 ?0.1 - 0.6 ?0.1 - 0.6 ?0.1 - 17.5 ?1.0 - 5.9 ?2.1 - 4.6 ?0.8 - OR% 33.6 - 9.5 - 6.1 - All dimensions are given in um. mm = minor diameter of the elliptical ocellus;