Journal of Comparative Physiology A

, Volume 195, Issue 4, pp 401–408

Evolution of color vision in pierid butterflies: blue opsin duplication, ommatidial heterogeneity and eye regionalization in Colias erate


  • Hiroko Awata
    • Laboratory of NeuroethologySokendai (The Graduate University for Advanced Science)
  • Motohiro Wakakuwa
    • Laboratory of NeuroethologySokendai (The Graduate University for Advanced Science)
    • Department of Biophysics, Graduate School of ScienceKyoto University
    • Laboratory of NeuroethologySokendai (The Graduate University for Advanced Science)
Original Paper

DOI: 10.1007/s00359-009-0418-7

Cite this article as:
Awata, H., Wakakuwa, M. & Arikawa, K. J Comp Physiol A (2009) 195: 401. doi:10.1007/s00359-009-0418-7


This paper documents the molecular organization of the eye of the Eastern Pale Clouded Yellow butterfly, Colias erate (Pieridae). We cloned four cDNAs encoding visual pigment opsins, corresponding to one ultraviolet, two blue and one long wavelength-absorbing visual pigments. Duplication of the blue visual pigment class occurs also in another pierid species, Pieris rapae, suggesting that blue duplication is a general feature in the family Pieridae. We localized the opsin mRNAs in the Colias retina by in situ hybridization. Among the nine photoreceptor cells in an ommatidium, R1-9, we found that R3-8 expressed the long wavelength class mRNA in all ommatidia. R1 and R2 expressed mRNAs of the short wavelength opsins in three fixed combinations, corresponding to three types of ommatidia. While the duplicated blue opsins in Pieris are separately expressed in two subsets of R1-2 photoreceptors, one blue sensitive and another violet sensitive, those of Colias appear to be always coexpressed.


Visual pigmentPhotoreceptorCompound eyeInsectColor vision


Color vision is an important visual function shared by a vast variety of animals. Some butterflies do have color vision and use it upon food search and oviposition (Kelber 1999; Kelber and Pfaff 1999; Kinoshita et al. 1999, 2008; Kinoshita and Arikawa 2000; Zaccardi et al. 2006). Their eyes seem to be well designed for discriminating colors with a set of photoreceptors of several different spectral sensitivities. Although the eyes of many butterflies share basic cellular organization, detailed anatomical and physiological studies have revealed an extensive variation between species. Therefore, the visual system of butterflies has been extensively studied to understand the evolution and mechanism underlying color vision (Briscoe and Chittka 2001; Arikawa et al. 2004; Kelber 2006).

Butterfly eyes are composed of several thousands of ommatidia, each containing nine photoreceptor cells, R1-9 (Yagi and Koyama 1963; Gordon 1977; Kolb 1978). These cells bear photoreceptive microvilli, together forming a three-tiered fused rhabdom in a majority of butterflies. The distal tier is composed of the microvilli of R1-4, whereas the proximal tier is made up of the microvilli of R5-8. The basal photoreceptor R9 bears microvilli immediately distal to the basement membrane.

The first group of butterflies whose anatomy and physiology of eyes was studied in detail was the genus Papilio (Papilionidae). The eyes of Papilio are furnished with at least six classes of spectral receptors that are embedded in the ommatidia in three fixed combinations (Arikawa 2003). The ommatidial heterogeneity can be anatomically identified by their characteristic yellow or red pigmentation around the rhabdom. A subset of red-pigmented ommatidia bears another pigment fluorescing under ultraviolet (UV) light. All of these pigments function as spectral filters for the photoreceptors, causing their spectral sensitivity to deviate substantially from the absorption spectrum of the expressed visual pigments.

Recent studies have accumulated data on the genetic background of visual pigment expression (Wakakuwa et al. 2006; Briscoe 2008). Insect eyes basically express three distinct opsins, belonging to the UV, blue (B) and long wavelength (L) absorbing visual pigment classes. Normally, a photoreceptor expresses a single opsin. The basic organization of the butterfly retina appears to be that R1 and R2 express the UV or B opsin in three combinations, UV–B, B–B and UV–UV, while R3-8 (and probably also R9) express the L opsin. However, the number of visual pigments is variable among species due to gene duplications. For example, pierids (Arikawa et al. 2005) and lycaenids (Sison-Mangus et al. 2006) have multiple B opsins, while multiple L opsins are found in Papilio (Kitamoto et al. 1998; Briscoe 2000), Apodemia mormo (Riodinidae) and Helmeuptychia helmes (Nymphalidae) (Frentiu et al. 2007). Multiple UV opsins have not yet been identified in butterflies, but they exist in Drosophila (Salcedo et al. 2003). The variability of spectral organization of eyes reflects the evolution and visual ecology of each species. To establish broader views on the evolution and neuronal mechanisms underlying color vision, comparative studies on carefully selected species are particularly important.

The family Pieridae contains four subfamilies, Pierinae, Coliadinae, Dismorphiinae and Pseudopontiinae (Braby 2006). We have found that Pieris rapae, a member of the Pierinae, has four visual pigments: they are UV (PrUV), violet (PrV), blue (PrB) and L (PrL)-absorbing visual pigments (Wakakuwa et al. 2004; Arikawa et al. 2005). Both PrV and PrB cluster in the clade of insect B opsins, and are expressed either in R1 or R2 photoreceptors, which consequently are violet or blue sensitive. To see whether this pattern is shared by other pierid species, we initiated an analysis of the eye of a member of the subfamily Coliadinae, the Eastern Pale Clouded Yellow butterfly, Colias erate. In Colias, we also found duplicated B opsins, but their expression pattern is simpler than in Pieris.

Materials and methods


Adults of the Eastern Pale Clouded Yellow butterfly, Colias erate Esper, were obtained from a laboratory culture. The culture was derived from eggs laid by females captured around the Sokendai Hayama Campus. The hatched larvae were fed with fresh clover leaves at 26°C and maintained under a 14 h light:10 h dark cycle.

Cloning and phylogenetic analysis

Poly-A RNA was extracted from eyes using QuickPrep micro mRNA purification kit (GE Healthcare, Uppsala, Sweden). To amplify fragments of cDNA encoding opsins of the UV, B and L classes, we carried out RT-PCR using degenerate primers designed based on consensus sequences of lepidopteran opsins so far identified. The full-length cDNAs were obtained by 5′- and 3′-RACE method. Both PCR and RACE products were purified, cloned using TOPO TA cloning kit (Invitrogen, Carlsbad, CA) and sequenced using ABI3130xl and BigDye terminator v1.1 (Applied Biosystems, Warrington, UK). The obtained sequences were aligned with others and processed for phylogenetic analysis by the neighbor-joining (NJ) and the maximum parsimony (MP) protocols using the MEGA 4.0.1 software (Tamura et al. 2007) and also by the maximum likelihood (ML) protocol using the PHYML 2.4.4 software (Guindon et al. 2005).

In situ hybridization

Isolated eyes immersed in 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.2) were microwave-irradiated six times for 5 s each (total 30 s) and then fixed for 20 min at 4°C. After dehydrating with graded ethanol, the eyes were sequentially infiltrated with terpineol and xylene, embedded into Paraplast (Sigma-Aldrich, St Louis, MO, USA) and sectioned at 8–10 μm thickness. The sections were deparaffinized with xylene, treated with 10 μg/ml proteinase K in PBS for 5 min at 37°C and acetylated with 0.25% acetic acid in 0.1 M triethanolamine (pH 8.0) for 10 min prior to hybridization.

Antisense RNA probes were synthesized from linearized plasmid carrying partial sequence of identified opsin mRNAs by in vitro transcription using digoxigenin-UTP (Roche, Mannheim, Germany). The probes were denatured at 70°C for 10 min before adding hybridization solution (300 mM NaCl, 2.5 mM EDTA, 200 mM Tris–HCl, pH 8.0, 50% formamide, 10% dextran sulfate, 1 mg/ml yeast tRNA, and 1X Denhardt’s solution), and again treated at 90°C for 5 min after being diluted with the hybridization solution at the final concentration of 0.5 μg/ml.

The heat-treated probe was applied to the sections at 45°C overnight. After washing in 2X SSC at 55°C for 15 min and then 50% formaldehyde-2X SSC at 55°C for 2 h, the sections were subsequently treated with 10 μg/ml RNase A for 1 h at 37°C and with 0.5% blocking solution (Roche, Mannheim, Germany) for 30 min at room temperature. The hybridized probes were detected using anti-digoxigenin-AP Fab fragments (Roche, Mannheim, Germany) and then visualized using 4-nitroblue tetrazolium chloride and 5-bromo-4-chloro-3-indolyl phosphate.

Light microscopy

Isolated eyes were prefixed in 2.5% glutaraldehyde, 2% paraformaldehyde in 0.1 M sodium cacodylate buffer (pH 7.4, CB) for 30 min at room temperature. After washing with CB, the eyes were postfixed in 2% osmium tetroxide in 0.1 M CB for 2 h at room temperature. The eyes were dehydrated with an acetone series and embedded in Quetol 812. Semi-thin sections were cut with a diamond knife.


Four opsin mRNAs

Starting from mRNA extracted from the retinal homogenate, we cloned four cDNAs encoding visual pigment opsins. They all shared seven trans-membrane domains and a lysine residue for binding retinal chromophore.

Phylogenetic analyses of cDNA sequences using the NJ protocol revealed that one opsin belongs to the clade of the UV-absorbing class (Fig. 1a), two belong to the B-absorbing class (Fig. 1b) and one to the L-absorbing class (Fig. 1c). Duplication of B opsins has been found in P. rapae, where the duplicated B opsins correspond to a violet (PrV, R425) and a blue (PrB, R453)-absorbing visual pigment. However, both B opsins of Colias clustered with PrV, but not with PrB: we therefore termed the Colias B opsins CeV1 and CeV2. We similarly termed the opsins of the UV and the L classes CeUV and CeL, respectively. The clustering pattern was consistent when analyzed using the MP and ML protocols (data not shown). Figure 1d shows similarities of amino acid sequences between the five B opsins so far identified in Pieridae. CeV1 is almost identical to the Colias philodice violet (CpV), but other pairs exhibit similarities between 72.5 and 85.2%. We carefully searched for additional B opsins in C. erate that would cluster with the Pieris PrB, using several other primer sets designed for amplifying B opsins of lepidoptera, but our search has remained unsuccessful.
Fig. 1

Phylogeny of butterfly opsins corresponding to the ultraviolet (a), blue (b) and long wavelength-absorbing (c) visual pigments. The neighbor joining (NJ) method was applied to all nucleotide triplets of the entire coding region using the Tamura-Nei substitution model of evolution. Opsins of a hawkmoth Manduca sexta (Sphingidae) were used as the outgroup. Each molecule was designated with the species name, the opsin term and the DDBJ accession numbers. Numbers represent bootstrap values based on 1,000 replicate analyses. Values larger than 50% are shown. (d) A matrix of amino acid homology of B opsins of Colias and Pieris

Distribution of opsin mRNAs in the retina

We localized the opsin mRNAs in the retina by in situ hybridization. We designed antisense cRNA probes hybridizing to a partial coding region (CeUV and CeL), the 3′ end of the coding region plus the following 3′ UTR (CeV2) or only 3′ UTR (CeV1). The lengths of the probes for CeUV, CeV1, CeV2, and CeL were 666, 553, 492, 873 nucleotides, respectively. We confirmed by dot blot analysis that any of the probes do not cross-hybridize with other templates (data not shown).

In longitudinal sections, we found that all four probes labeled the retina from the dorsal to the ventral edge (Fig. 2a–d). The CeUV, CeV1 and CeV2 labeled only in the distal tier (Fig. 2a–c), whereas the CeL probe labeled the entire length of the retinal layer (Fig. 2d). The distal and proximal tiers of the rhabdom are constructed with R1-4 and R5-8, respectively, as shown in Fig. 2e.
Fig. 2

In situ hybridization results (ad longitudinal sections, fj transverse sections of the dorsal region, ko transverse sections of the ventral eye region) with a schematic diagram of a Colias ommatidium (e). The sections were labeled by probes specific to the CeUV (a, f, k), CeV1 (b, g, l), CeV2 (c, h, m) and CeL (d, i, j, n, o) mRNAs. Two sets of serial sections through the distal tier of the dorsal (fh) and the ventral (km) regions reveal three types of ommatidia, indicated by solid circles (type I), dotted circles (type II) and broken circles (type III). Bars 200 μm (ad, shown in a), 20 μm (fo, shown in k). 1–9, photoreceptors R1-9; BM basement membrane, L lamina, M medulla, R retina

To identify the photoreceptors expressing each opsin mRNA, we labeled serial transverse sections of the dorsal (Fig. 2f–j) and the ventral (Fig. 2k–o) regions. We found that the CeUV, CeV1 and CeV2 probes labeled either R1 or R2 (Fig. 2f–h, k–m). Rather surprisingly, the CeV1 and CeV2 labels always colocalized. According to the labeling pattern of the CeUV and CeV1/V2 probes, we identified three ommatidial types. Type I ommatidia had either R1 or R2 labeled by the CeUV probe, and the other photoreceptor (R2 or R1) was then labeled by the CeV1/V2 probes (Fig. 2f–h, k–m, solid circles); the photoreceptor labeled by the CeUV probe looked smaller than the CeV1/V2 probe-labeled photoreceptor in some cases. Type II ommatidia have both R1 and R2 labeled by the CeV1/V2 probes (Fig. 2f–h, k–m, dotted circles). The type II ommatidia in the dorsal eye region are labeled stronger (Fig. 2f–h, dorsal subtype) than those in the ventral eye region (Fig. 2k–m, ventral subtype). The cells labeled stronger may contain more visual pigment molecules, but no direct evidence for this is available so far. Type III ommatidia have R1 and R2 both labeled by the CeUV probe (broken circles). We found that the R3-8 photoreceptors in all ommatidia were labeled by the CeL probe. We could not find clear labeling with any probe in the R9 photoreceptors.

Dorso-ventral specialization

We analyzed the in situ hybridization results to quantify the dorso-ventral specialization in the distribution of the ommatidial types, including the two orientation subtypes of type I and the dorsal- and ventral-subtypes of type II, by determining the identity of 4095 ommatidia in micrographs from five individuals. After defining the border between the dorsal and ventral eye regions, using the dorsal and the ventral subtypes of type II ommatidia (Fig. 3a), we quantified the number of each type in bands of six horizontal rows (Fig. 3b). While about half of the ommatidia are type I throughout the eye, the proportions of types II and III gradually change: the ventral region contains up to 50% of type II ommatidia, while in the most dorsal region up to 30% of the ommatidia are type III.
Fig. 3

Eye regionalization. a Reconstructed representation of a transverse section with color-coded ommatidial types; type I (red and orange), type II (blue and light blue), type III (ivory). b Distribution of each type of ommatidia in the eye based on 4,095 ommatidia from five individuals. Each bar represents the ratio of ommatidia contained in bands of six rows of ommatidia. The border between the dorsal and the ventral region was determined by the existence of type II ommatidia whose R1 and R2 were strongly labeled with the CeV1/V2 probes. Type I contains two subtypes, one with R1 being labeled with the CeUV probe (red-coded) and the other with R2 labeled with the CeUV probe (orange-coded). Range bars show standard deviation

Perirhabdomal pigmentation

Unstained light microscopic histology also revealed three types of ommatidia. All ventral ommatidia appeared to have four reddish spots around the rhabdom, which are pigment clusters in R5-8 photoreceptors. The pigment clusters are arranged in either a trapezoidal, square or rectangular pattern (Fig. 4). Analyzing 209 ommatidia in the ventral region of an eye, we found that about 48% had the trapezoidal pattern, 39% the square, and 13% the rectangular pattern (Table 1). The trapezoidal pattern occurred in about half of the ommatidia, which had a smaller cell body on the side of the shorter edge of the trapezoid (arrow in Fig. 4). Most likely these ommatidia are the type I ommatidia, and the smaller cells are the CeUV-expressing photoreceptors (Figs. 2, 3). The occupation ratio indicates that the square type (ca. 40%) and the rectangular type (ca. 10%) correspond to types II and III ommatidia, respectively. Table 1 summarizes the histological results.
Fig. 4

Ommatidial pigmentation. Unstained plastic section through the distal tier of the ventral eye. Arrangement of four pigment spots around the rhabdom designates three types of ommatidia corresponding to type I (trapezoidal, red circle), type II (square, dotted circle) and type III (rectangular, broken circle) identified by in situ hybridization. Smaller cell body of type I (arrow)

Table 1

Three types of the ventral ommatidia in the Colias eye


Evolution of B opsins in Pieridae

In the retina of C. erate, we identified four opsin mRNAs, two of which clustered in the blue-absorbing (B) class of butterfly opsins. Because duplication of B opsins was found in another pierid species, P. rapae (Arikawa et al. 2005), we expected to find two B opsins also in C. erate, although only one B opsin was identified in the Clouded Sulfur, C. philodice (Sison-Mangus et al. 2006). Surprisingly, the two B opsins, CeV1 and CeV2, found in C. erate both clustered with one of the Pieris B opsins, PrV (Fig. 1b). A comparison of amino acid sequences indicated that CeV1 is an ortholog of CpV, the violet opsin of C. philodice (Fig. 1d). Because an extensive search in C. erate for additional B opsin clustering with the PrB was unsuccessful, we currently have to conclude that Colias does not have a PrB ortholog.

Assuming that C. erate and C. philodice lack PrB orthologs and that the speed of evolution of pierid B opsins is constant, we can hypothesize the evolution of the trait as follows. At the early stage of evolution of the family Pieridae, an ancestral B opsin gene duplicated in the common ancestor of the Pierinae and the Coliadinae, eventually forming PrB and PrV in P. rapae (Pierinae). In the lineage of Coliadinae, a duplicate that was ancestral to PrB was lost, and the other that was ancestral to PrV independently duplicated forming CeV1/CpV and CeV2. This hypothesis would be supported at least in part if C. philodice has a CeV2 ortholog.

Function of duplicated opsins in butterflies

The basic design of butterfly eyes is to have UV and B opsins in R1 and R2 in three fixed combinations, UV–B, B–B and UV–UV (Wakakuwa et al. 2006), but opsin gene duplication has modified the principle. In P. rapae, the B–B combination is replaced by V–V, where both R1 and R2 express PrV. PrB is expressed in the UV–B combination, together with PrUV. The differential expression of PrV and PrB divides the R1 and R2 into violet (λmax = 425 nm) and blue (λmax = 453 nm) receptors (Arikawa et al. 2005). The situation is completely different in C. erate: the paralogs CeV1 and CeV2 are always coexpressed in R1 and R2, as if they behave as a single unit.

There are more examples of opsin gene duplications and coexpression of multiple opsins in single photoreceptors. Papilio has three L opsin paralogs, PxL1, PxL2 and PxL3, which are differentially expressed in R3-8 photoreceptors in a characteristic pattern (Arikawa 2003). Differential expression of three L opsins divides the R3-8 into green (PxL2 expressed, λmax = 540 nm) and red (PxL3 expressed, λmax = 600 nm) receptors (Kitamoto et al. 1998; Arikawa et al. 1999). A subset of R5-8 coexpresses PxL2 and PxL3: these receptors have a characteristic broadband sensitivity. PxL1 is peculiar: it always exists together with PxL2 in R3 and R4 green receptors in the ventral ommatidia (Kitamoto et al. 1998). Furthermore, in females of the lycaenid butterfly Lycaena rubidus, one of the duplicated B opsins is coexpressed with the L opsin in the R3-8 photoreceptors of the dorsal ommatidia (Sison-Mangus et al. 2006). Drosophila has a set of ommatidia that coexpresses two UV opsins, Rh3 and Rh4, in a central photoreceptor R7 (Mazzoni et al. 2008).

The physiological consequences of coexpression of opsins in single photoreceptors can be seen in the broadband receptors of Papilio. Their spectral sensitivity can be well understood by postulating that both PxL2 (R515) and PxL3 (R575) visual pigments participate in the phototransduction process (Arikawa et al. 2003). Presumably, the same holds for the coexpressed visual pigments of Lycaena and Drosophila, because the coexpressed opsins are singly expressed elsewhere and function properly there.

The Colias CeV1/CeV2 and the Papilio PxL1 are unusual, as these opsins are never expressed singly. A challenging hypothesis for the Colias case is that CeV1 and CeV2 are functional only when they form heterodimers as a result of possible subfunctionalization of duplicated genes (Force et al. 1999). In fact, vertebrate visual pigments seem to exist as dimers in native disc membranes (Fotiadis et al. 2006), and some G protein-coupled receptors (GPCRs) require heterodimerization for retaining proper activity (Prinster et al. 2005). The case of Papilio PxL1 is even more puzzling because coexpressing PxL2 does not have to have PxL1 for proper function: there are a set of photoreceptors that exclusively expresses PxL2 (Arikawa 2003). Here the PxL1 mRNA may not even be translated.

Spectral heterogeneity of ommatidia

Heterogeneity of ommatidia has repeatedly been reported in insects including butterflies (Wakakuwa et al. 2006). Here we found a similar heterogeneity in the butterfly C. erate. Its eyes have at least three types of ommatidia, expressing four opsin mRNAs in three fixed combinations (Table 1). The differential expression of four opsins must create spectral heterogeneity of the ommatidia. Based on the previous analyses on the spectral sensitivities of Pieris photoreceptors (Qiu and Arikawa 2003a, b; Wakakuwa et al. 2004; Arikawa et al. 2005), it is rather easy to predict that the CeUV- and CeV1/V2-double expressing R1/R2 photoreceptors are UV and violet sensitive, respectively. The CeL expressing R3-8 photoreceptors are at least dimorphic depending on their locations. Those in the distal tier, R3 and R4, will have a spectral sensitivity similar to the absorption spectrum of the CeL visual pigment, and thus the R3 and R4 are probably green or yellow-green sensitive. However, the R5-8 in the proximal tier are most likely red sensitive due to the filtering effect of the red perirhabdomal pigment (Fig. 4) (Wakakuwa et al. 2004). These points have to be confirmed by electrophysiology as well as by in vitro expression of visual pigments and subsequent spectroscopic analysis.

Eye regionalization

The most pronounced regional specialization of compound eyes is often seen in the dorsal rim area in many insects, including butterflies. The area is dominated by short wavelength photoreceptors and usually plays a crucial role in detecting sky polarization for long distance navigation (Rossel 1989; Labhart and Meyer 1999; Homberg 2004; Sauman et al. 2005). We have not been successful in characterizing the ommatidia in the dorsal rim area in Colias, but the present in situ hybridization clearly demonstrates a more general regionalization along the dorso-ventral axis in the main part of the eye, found in a variety of butterflies by optical methods (Arikawa and Stavenga 1997; Stavenga et al. 2001; Stavenga 2002). We found that the expression density of CeV1/V2 mRNA in type II ommatidia distinctly marked the border between the dorsal and the ventral eye regions: the dorsal type II ommatidia were more strongly labeled than the ventral ones. Also we found more type III (UV–UV) ommatidia in the dorsal region, and type II (V–V) in the ventral region (Fig. 3b). The richness of type III ommatidia in the dorsal region indicates that the region has higher spatial resolution and sensitivity in the UV wavelength range. On the other hand, the ventral region has higher spatial resolution in the violet wavelength range. A similar biased distribution of photoreceptors expressing B opsin in the ventral region is found in honeybees (Wakakuwa et al. 2005), which is in accordance with better detection of blue targets by the ventral region of the eye (Giurfa et al. 1999). Assuming that the in situ hybridization labeling density is related to the absolute sensitivity of the receptor, we may hypothesize that in the dorsal region the violet receptors may be more sensitive in order to compensate for their lower density.


We thank Doekele Stavenga and Primoz Pirih for critical reading of the manuscript. This work was supported by the Grants-in-Aid for Scientific Research from JSPS and by a Project Grant from the Hayama Center for Advanced Studies of Sokendai to KA. MW was a JSPS research fellow.

Copyright information

© Springer-Verlag 2009