Random Array of Colour Filters in the Eyes of Butterflies

Butterflies are often admired for their spectacular colour patterns. They are therefore generally assumed to possess colour vision. Behavioural experiments on butterflies have indicated that they can indeed perceive colour contrast (for example Hidaka and Yamashita, 1975; Ilse, 1928, 1937, 1941; Kolb and Scherer, 1982; Wehner, 1981). Compared with bees, in which extensive behavioural studies have demonstrated the ability to see colour (Menzel and Backhaus, 1989), and flies, which also possess a clear colour discrimination system (Fukushi, 1989; Troje, 1993), convincing behavioural evidence for colour vision in butterflies is lacking.

A basic physiological requirement for colour vision is the existence of a set of different spectral receptor types in the retina. This is certainly fulfilled in a number of butterfly species, as shown by anatomical (Ribi, 1978, 1987), optical (Bernard, 1979; Bernard and Remington, 1991) and electrophysiological (Eguchi et al. 1982; Horridge et al. 1984; Kinoshita et al. 1997; Matic 1983; Shimohigashi and Tominaga, 1991; Steiner et al. 1987) studies. Presently, the best characterized retina of a butterfly, in terms of the spectral properties of the photoreceptors, is that of the Japanese yellow swallowtail butterfly Papilio xuthus. We have demonstrated previously that the retina contains at least five types of spectral receptors, peaking in the ultraviolet, violet, blue, green and red wavelength regions (Arikawa et al. 1987). Subsequent work has demonstrated that the nine photoreceptors of an ommatidium, R1–R9 (see Fig. 1), are spectrally heterogeneous. R1 and R2 are ultraviolet, violet or blue receptors, R3 and R4 are green receptors (Bandai et al. 1992) and R5–R9 are either green or red receptors (Arikawa and Uchiyama, 1996).

Within this diverse population of photoreceptors, the nature and composition of spectral receptors in the ommatidia has yet to be elucidated. A more complete characterization of the ommatidia and their arrangement over the retina is the aim of the present paper. Histological, electrophysiological and optical experiments on the compound eye of P. xuthus have identified three different types of ommatidia, randomly distributed over the retina. This random distribution may be crucial for colour vision.

The Japanese yellow swallowtail butterfly Papilio xuthus Linnaeus was held as a laboratory stock culture. Other species were collected in Yokohama, Japan.

For light microscopy, the compound eyes were isolated from the head, and fixed in 4 % paraformaldehyde in 0.1 mol l⁻¹ sodium cacodylate buffer at pH 7.2 (CB) for 30 min. The eyes were subsequently dehydrated through an acetone series, embedded in Epon, and sectioned at 10 µm using a rotary microtome. The sections were observed without staining.

For electron microscopy, the eyes were fixed in 2 % glutaraldehyde and 2 % paraformaldehyde in CB for 2 h, and further fixed in 2 % OsO₄ in CB for 2 h. After being dehydrated, the tissues were embedded in Epon. Ultrathin sections were stained with uranyl acetate and lead citrate.

To observe ommatidial transmission, a fresh compound eye was cut at the level of the proximal photoreceptors using a vibrating microtome. The slice was then mounted, with the corneal side up, on the stage of an epifluorescent microscope (BX-60, Olympus). The focus was at the corneal surface. First, the transmitted light was photographed with white light illumination, from below. Subsequently, the ultraviolet-induced fluorescence was photographed (objective 4×, numerical aperture 0.16).

A butterfly was mounted in a Faraday cage, and a glass micropipette, filled with 5 % Lucifer Yellow CH aqueous solution (resistance 150 MΩ), was inserted into the retina through a hole made in the dorsal cornea. After impalement of the electrode into a single photoreceptor, the spectral sensitivity of the cell was determined by axial stimulation with a point source delivering an equiquantal series of monochromatic lights with wavelengths ranging from 290 to 700 nm (maximal intensity 5.0×10¹¹ photons cm⁻² s⁻¹ at the corneal surface). After injecting Lucifer Yellow into the photoreceptor, by applying 2–5 nA of hyperpolarizing d.c. current for 5–10 min, the eye was prepared for light microscopy as described above. The Lucifer-Yellow-injected photoreceptors were then identified in 10 µm sections under violet epi-illumination.

To study the histological relevance of the distribution of spectral receptors, we first prepared light microscope sections of the P. xuthus retina. We discovered in serial sections that both R3 and R4 and R5–R8 photoreceptors in any one ommatidium contain either yellow or red pigment near the rhabdom (Figs 1, 2A). Further electron microscopy demonstrated that the yellow and red pigmentation is made up of pigment granules, residing in the cell body 1–1.5 µm from the edge of the rhabdom (Fig. 2B). R1 and R2 of all ommatidia have a purple pigmentation positioned at the distal tip of the ommatidia. The anatomical results are summarized schematically in Fig. 1.

The rhabdom diameter is approximately 2.0 µm (Fig. 2B) and it therefore functions as an optical waveguide (for a review, see van Hateren, 1989). Consequently, the light-absorbing pigment surrounding the rhabdom will act as a colour filter (Stavenga, 1989; Marshall et al. 1991). This effect can be directly visualized by cutting a fresh compound eye with a vibrating microtome at the level of the proximal photoreceptors. In transmitted light, the ommatidia then appear either yellow or (more or less saturated) red (Fig. 3A). The purple pigment is shared by R1 and R2 in all ommatidia, so that the observed difference in coloration must be attributed to the yellow or red pigment in R3–R8 (Fig. 2A).

The ratio of yellow to red ommatidia is approximately 1:3. For 653 ommatidia in 10 transverse sections collected from four individuals, the ratio was 166:487; there was no significant variation among individuals. We further investigated whether the array of differently coloured ommatidia has any regularity. To do this, we counted the frequency of transition, for example from yellow to red or from red to red, along the three axes of the hexagonal lattice. It appeared that the transition frequency is independent of the frequency of the type of its neighbours and only reflects the absolute probability of the ommatidial type (10 micrographs yielded χ²=2.25, d.f.=4).

To investigate any possible correlation between the spectral receptor types and the pigmentation, we recorded spectral sensitivities from single photoreceptors and marked the cells by injecting Lucifer Yellow. Subsequently, we identified the pigmentation of the ommatidia to which the penetrated photoreceptors belonged by light microscope histology. We found that the proximal R5–R8 cells in the yellow-pigment-containing ommatidia, without exception, appear to be green-sensitive receptors. The R5–R8 cells in the red-pigment-containing ommatidia are always red-sensitive receptors.

Of course, it is virtually impossible to prove conclusively using electrophysiology that cells R5–R8 in a single ommatidium are always identical. A powerful alternative is offered by histological in situ hybridization to localize different mRNAs encoding visual pigment opsins. Our preliminary results indicate that photoreceptors R5–R8 in a single ommatidium always contain identical opsin mRNA (Kitamoto et al. 1996); that is, it is most likely that R5–R8 in the red-pigment-containing and the yellow-pigment-containing ommatidia are all red- and green-sensitive receptors, respectively.

We also investigated the intact eye of P. xuthus using fluorescence microscopy and found that a further aspect of retinal heterogeneity can be directly observed in vivo. Ultraviolet epi-illumination of the P. xuthus eye excites a distinct whitish emission in approximately one-quarter of the ommatidia in the ventral half of the eye (Fig. 3B). In six micrographs, 28 % of the ommatidia fluoresced. Again, their distribution appears to be random. Upon focusing at levels proximal to the cornea, waveguide mode patterns appear, indicating that the fluorescence originates from the rhabdom (Franceschini et al. 1981; Nilsson et al. 1988; van Hateren, 1989). Preliminary photochemical and biochemical experiments suggest that the fluorescing pigment is a retinoid, but conclusive results have yet to be obtained. The ultraviolet-fluorescing ommatidia are not seen in the dorsal half of the eye. Their presence, if they are there, is obscured by a distinct corneal fluorescence.

Combined fluorescence and transmission microscopy of eye slices revealed that those ommatidia that fluoresce under ultraviolet excitation appear a less saturated red, i.e. very pale to pink, in transmitted light (Fig. 3A,B). Further electrophysiological experiments, combined with Lucifer Yellow staining and fluorescence microscopy, demonstrated that violet receptors only occur in the fluorescing ommatidia. Ultraviolet and blue receptors were found exclusively in the non-fluorescing ommatidia.

The sensitivity spectrum of the violet receptor is very sharp-peaked (Arikawa et al. 1987). This suggests the challenging hypothesis that the fluorescing pigment might act as a selective ultraviolet absorbance filter. The red receptor also has a narrower spectral sensitivity spectrum than predicted from a rhodopsin template (Arikawa et al. 1987). This is probably a direct consequence of the strong high-pass filtering by the red photoreceptor pigmentation (Arikawa et al. 1996).

Our studies on P. xuthus show that its retina is a mesh of at least three types of optically distinguishable ommatidia: one containing yellow pigment and two containing red pigment, one type of which contains an ultraviolet-fluorescing pigment. Together with the visual pigments of the photoreceptors, these pigments are crucial in shaping the spectral sensitivity of the five receptor types. How P. xuthus connects these receptors to achieve colour discrimination, which might even be pentachromatic, remains an intriguing problem. Behavioural experiments detailing the colour discrimination capabilities of P. xuthus are now in progress.

A mesh of different classes of ommatidia has been found previously in the retina of flies (Franceschini et al. 1981; Hardie, 1986; Chou et al. 1996), where the two central photoreceptors determine colour vision (Troje, 1993). The observed randomness of the retinal organization in P. xuthus is also shared with primates: a random distribution of M and L cones has, for instance, been demonstrated in talapoin monkeys (Cercopithecus talapoin) (Mollon and Bowmaker, 1992). In contrast, several fish species have a very regular retina (Lythgoe, 1979).

In a survey of the compound eyes of other butterflies (Fig. 4), we found a marked variation in pigmentation around the rhabdom, not only in the species belonging to the family Papilionidae (Papilio bianor, Papilio protenor, Papilio polytes, Graphium sarpedon, Parnassius glacialis) but also in Lycaenidae (Zizeeria maha, Lycaena phlaeas) and Satyridae (Neope goschkevitschii, Ypthima argus). These results are in agreement with previous optical studies on intact butterfly eyes using epi-illumination microscopy (Bernard and Miller, 1970; Miller, 1979). These studies revealed a multicoloured reflection from a tapetal mirror proximal to the rhabdom that exists in the eye of all butterflies except for the Papilionidae. Although no direct evidence is available so far, the difference in pigmentation and coloration of the tapetal reflection strongly indicates that the ommatidia are different in terms of the spectral receptor types they contain.

Compound eyes often exhibit considerable regionalization, because different parts of the eye are devoted to specific tasks, such as prey or mate recognition or polarization vision (Stavenga, 1992). Here, we have shown that, within restricted regions of the butterfly eye, ommatidial characteristics can differ considerably. Clearly, the concept that compound eyes, at least locally, consist of identical building blocks, the ommatidia, does not hold. The random organization of the spectral receptors in the retina of butterflies seems to be a universal feature, presumably because a diverse set of spectral receptors is essential for a highly developed colour vision system (Bernard, 1979; Bernard and Remington, 1991; Goldsmith, 1990).

We thank P. Brakefield, M. F. Land and G. D. Bernard for reading critically an earlier version of the manuscript. This work was supported by Grants from the Ministry of Education, Science, and Culture of Japan and the Uehara Memorial Foundation to K.A. and from the Japanese Society for the Promotion of Science (JSPS) to D.G.S.