Reconstructing the ancestral butterfly eye: focus on the opsins

The butterfly eye is a marvel of evolution. Butterfly vision, like that of other insects, is based on three major classes of photoreceptor, with peak sensitivity (λ_max) in the ultraviolet (UV, 300–400 nm), blue (B, 400–500) and long wavelength (LW, 500–600 nm) part of the light spectrum. At the molecular level, these visual pigments are composed of a retinal-based chromophore (e.g. 11-cis-3-hydroxyretinal) (Smith and Goldsmith, 1990) attached by a Schiffbase linkage to an opsin protein. The spectral tuning of the visual pigment wavelength of peak absorbance, λ_max, is achieved through the interaction of the chromophore with critical amino acid residues within the opsin. Changes in the polarity of amino acids in the chromophore-binding pocket of opsins, for example, affect the distribution of electrons in the π-electron system of the chromophore, producing a diversity of λ_max values(Honig et al., 1976). In the case of the butterfly visual pigments, the three major spectral classes are encoded by ancient duplications, which produced distinct UV, B and LW opsin genes.

Unlike those of bees, the eyes of butterflies are anatomically and physiologically diverse (Peitsch et al.,1992; Briscoe and Chittka,2001; Stavenga and Arikawa,2006; Frentiu et al.,2007b; Frentiu et al.,2007a). This is due in part to lineage-specific opsin gene duplications, as well as to changes in the kind and distribution of lateral filtering pigments (Arikawa and Stavenga,1997; Stavenga,2002).

To understand this diversity, this review focuses on the eyes of four species of butterfly for which we have the most complete information, Danaus plexippus Linnaeus, Lycaena rubidus Behr, Pieris rapae crucivora Boisduval and Papilio xuthus Linnaeus, and whose eyes serve as prototypes for the butterfly families Nymphalidae, Lycaenidae,Pieridae and Papilionidae, respectively. Although the importance of non-opsin-based filtering pigments for modifying photoreceptor sensitivity to light is mentioned, it is not yet clear whether the filtering pigments in different families are homologous or not; therefore I have intentionally spotlighted the evolution of the opsins themselves because of the homologous relationship between opsin protein sequences between butterfly families and the spectral range of vision. By analyzing the physiologically determined photopigment absorbance spectra and the phylogenetic relationship of opsin gene sequences, overlaid with a character map of opsin expression patterns, I propose a novel model of the evolutionary events leading from an ancestral butterfly eye to the radiation of eye types observed today. And with an eye to the future, I emphasize a few fertile areas for further investigation.

Anatomically, the basic unit of the butterfly compound eye is the ommatidium. An ommatidium is composed of nine photoreceptor cells (R1–9)along with primary and secondary pigment cells. An ommatidium from the simplest eye, that of nymphalid butterflies, consists of two tiers composed,respectively, of eight photoreceptor cells (R1–8) that are elongated,and a tiny R9 cell that sits just above the basement membrane(Fig. 1A). Light passing through the cornea is focused by the crystalline cone onto the rhabdom. The fused rhabdom is composed of nine rhabdomeres, the microvillous membranes protruded by the individual photoreceptor cells, in which the visual pigments are embedded.

The ommatidia of the main retina of the nymphalid eye are divided into three subtypes with respect to opsin mRNA expression, with R1 and R2 photoreceptor cells expressing UV–UV, UV–B or B–B, while the outer R3–8 cells express the LW opsin(Fig. 1B)(Briscoe et al., 2003). The monarch butterfly, Danaus plexippus, is a nymphalid butterfly whose eye typifies this pattern of opsin mRNA expression(Fig. 1C–E)(Sauman et al., 2005). The identity of the opsin expressed in the R9 cell of monarchs is unknown, but is likely to be the same as in another nymphalid Vanessa cardui, in which R9 expresses the LW opsin (Briscoe et al., 2003).

In addition to the ommatidia of the main retina, butterfly eyes also contain a dorsal rim area (DRA), like that found in other insects(Labhart and Meyer, 1999). In monarchs, the DRA R1–8 cells express the UV opsin exclusively(Sauman et al., 2005); these ommatidia contain rhabdoms whose microvilli are anatomically specialized to detect polarized skylight (Reppert et al.,2004). This monochromatic pattern of opsin expression in the DRA is important for efficient perception of polarized light, by avoiding interference with color information(Labhart and Meyer, 1999).

The wavelength of peak absorbance, λ_max, of visual pigments can be directly measured using epi-microspectrophotometry, or derived from the spectral sensitivity curves of intracellular recordings of individual photoreceptor cells. In both cases, the experimental data are matched to an idealized rhodopsin template (Stavenga et al., 1993; Palacios et al.,1996), and a computational model is then applied to find the best estimate of λ_max.

Based on such measurements, the spectral diversity of visual pigments in the eyes of butterflies is striking. As already alluded to, the eye of the monarch contains the smallest number of butterfly photopigments (P) with one UV (λ_max=340 nm or P340), one B (P435) and one LW (P545)(Stalleicken et al., 2006; Frentiu et al., 2007a)(Fig. 2A). The lycaenid Lycaena rubidus eye contains four photopigments: one UV (P360), two B(P437 and P500) and one LW (P568) (Bernard and Remington, 1991) (Fig. 2B). The papilionid Papilio xuthus eye contains the largest number of butterfly photopigments so far described, with one UV(P360), one B (P460) and three LW (P515, P530 and P575)(Arikawa, 2003)(Fig. 2C). The eye of the pierid Pieris rapae contains four photopigments: one UV (P360), two B(P425 and P453) and one LW (P563) (Qiu and Arikawa, 2003a; Qiu and Arikawa, 2003b) (Fig. 2D).

The opsin cDNAs encoding each of the visual pigments have been isolated(Kitamoto et al., 1998; Kitamoto et al., 2000; Wakakuwa et al., 2004; Arikawa et al., 2005; Sauman et al., 2005; Sison-Mangus et al., 2006; Frentiu et al., 2007a), and with these data a gene tree has been constructed for each of the three spectral classes of opsin: UV, B and LW(Fig. 3A–C,respectively). Species from all five butterfly families, Papilionidae(Papilio xuthus and P. glaucus Linnaeus), Pieridae(Pieris rapae), Nymphalidae (Danaus plexippus), Lycaenidae(Lycaena rubidus) and Riodinidae (Apodemia mormo Felder and Felder) have been included to illustrate the astonishing fact that independent duplications of either the B or LW opsin have occurred in each family,rendering the eyes of butterflies the most diverse yet characterized among insects.

For each of the butterfly family examples, only one UV opsin gene has been isolated, including from the riodinid Apodemia mormo, presented here for the first time (Fig. 3A). By contrast, duplicate B opsin genes have been isolated from P. rapae, encoding blue-absorbing (P453) and violet (V)-absorbing (P425)photopigments (Arikawa et al.,2005), and from L. rubidus, encoding blue-absorbing(P437) and blue–green-absorbing (P500) photopigments(Sison-Mangus et al., 2006). The bootstrap values of the relevant nodes and branching order of the tree,which nicely matches the current phylogeny of butterfly families based on independent molecular and morphological markers (see below)(Wahlberg et al., 2005),indicate that the duplication of the blue opsin gene in pierids and in lycaenids occurred independently.

The situation for the LW opsins, with respect to the frequency of gene duplication, is more complex. The eyes of pierids, nymphalids and lycaenids express only one LW opsin gene encoding photopigments of P563, P545 and P568,respectively (Wakakuwa et al.,2004; Sauman et al.,2005; Sison-Mangus et al.,2006), while the papilionid eye expresses three (P515, P530 and P575) (Kitamoto et al., 1998)and the riodinid eye expresses two LW opsin genes (P505 and P600)(Frentiu et al., 2007a)(Fig. 3C). To add more complexity, at least one additional near full-length LW opsin cDNA, PglRh4, for a total of four LW opsin genes, has been isolated from head tissue of Papilio glaucus(Briscoe, 1998; Briscoe, 2000) – an opsin that is not apparently expressed in the eye (A.D.B., unpublished data)(Lampel et al., 2005). The presence of two LW opsin genes in the genome of the silkmoth Bombyx mori, and the apparent close homology of the sphingid moth Manduca sexta opsin gene MANOP1 with one of them(Fig. 3C), suggests that M. sexta too may contain a second LW opsin, which is a homolog of the silkmoth larval `brain opsin' gene Boceropsin(Shimizu et al., 2001).

The fate of duplicate genes is typically classified into one of two categories: subfunctionalization (`division of labor'), in which one or both of the paralogs acquires a more restricted expression pattern and/or biological activity than the common ancestral gene, and neofunctionalization,in which the paralogs are free to acquire new expression patterns and/or functions (Force et al.,1999). Opsin duplication in butterflies can fall into either category (Briscoe, 2001). Importantly, duplication of butterfly opsin genes has led to the evolution of novel λ_max values that ultimately impact on behavior(Kelber, 1999; Kelber and Pfaff, 1999; Kinoshita et al., 1999).

In Pieris rapae, LW opsin expression is identical to that of the nymphalid eye in that its single LW opsin transcript is expressed in R3–8 (Wakakuwa et al.,2004). In addition, the expression of the UV opsin in Pieris also follows that of the nymphalid in that it is found in either UV–UV or UV–B combinations in R1 and R2. In what appears to be a complete departure from the nymphalid ground plan for the eye, however,the B–B ommatidial subtype has been completely replaced in P. rapae ventral eye by a V–V ommatidial class that is the product of a B opsin gene duplication (Fig. 4). Since the violet-absorbing visual pigment is encoded by a duplicate B opsin (Fig. 3B),and the pattern of B opsin expression in other butterflies considered so far is restricted to the R1 and R2 cells, the V–V class of ommatidium can be interpreted as an example of subfunctionalization of the ancestral B opsin domain of expression, into a more limited expression pattern.

The pattern of duplicate B opsin expression in Lycaena rubidusillustrates both neofunctionalization and subfunctionalization and leads to the situation in which male and female butterflies literally see the world through different eyes. In the compound eye of this animal, four opsin genes are expressed – one UV (UVRh), duplicate B, B1 (BRh1)and B2 (BRh2), and one LW (LWRh) opsin(Fig. 3). It is the dorsal eye of L. rubidus that is sexually dimorphic. Unlike the papilionid,pierid or nymphalid eye, in which R3–8 express one or more LW opsins, in L. rubidus males, R3–8 exclusively express the B1 (dark blue ovals) opsin, while in females, R3–8 co-express the B1 and LW (dark blue–orange ovals) opsins (Fig. 4), an example of neofunctionalization of the B1 opsin domain of expression. In addition, the R1 and R2 cells of the dorsal eye of both males and females express UV–UV almost exclusively, with a minor number of ommatidia expressing the UV–B1 or B1–B1 combination(Sison-Mangus et al., 2006). The highly territorial male L. rubidus(Bernard and Remington, 1991)probably use their dorsal eye for dichromatic color vision and the detection of moving objects, such as airborne males.

The ventral eye, by contrast, has a pattern of LW opsin expression that is more consistent with the other butterfly families examined: R3–8 of the ventral retina express the LW opsin. It is here, though, that subfunctionalization of the second blue opsin duplicate gene, B2 (light blue ovals, Fig. 4), is evident. The ventral retina contains six classes of ommatidia that differ according to the opsins expressed in R1 and R2: UV–UV, UV–B1, UV–B2,B1–B2, B1–B1 or B2–B2(Fig. 4).

Intriguingly, while butterflies in the genus Papilio use duplicate LW opsins to see green (Kelber,1999), the lycaenid Polyommatus icarus, uses its duplicate B2 opsin, in conjunction with its LW opsin, to see in the green part of the spectrum extending up to 560 nm(Sison-Mangus et al., 2008). This suggests that natural selection has hit upon alternative strategies for color vision in the green range in lycaenid and papilionid butterflies.

The basic pattern of photoreceptor cell morphology and opsin mRNA expression exemplified by the nymphalid eye is also complicated in the swallowtail butterfly Papilio by alterations in the shapes of the photoreceptor cells and by the expression of the three duplicate LW opsins. For example, the ommatidium of the papilionid eye is organized into three tiers rather than two: a distal tier in which the cell bodies of the R1–4 cells are the widest, a middle tier in which the cell bodies of the R5–8 cells are the widest, and a third tier in which the cell body of the tiny R9 cell is located (Fig. 5).

In Papilio xuthus, the butterfly eye about which we know the most,three LW opsin genes (PxRh1, PxRh2 and PxRh3) are also expressed in the eye. In this butterfly, as in lycaenids, the main retina is divided into two parts: the dorsal eye and the ventral eye. Like nymphalids,the R1 and R2 cells of both the dorsal and ventral eye express UV–UV,UV–B or B–B opsin mRNAs. Unlike nymphalids, the R3 and R4 cells of the dorsal eye of P. xuthus express the PxRh2 opsin mRNA(orange ovals, Fig. 4), while in the ventral eye, the R3 and R4 cells co-express PxRh1 and PxRh2 (green–orange ovals, Fig. 4)(Kitamoto et al., 1998).

Strikingly, the expression of opsins in R5–8 is coordinated with the expression of opsins in R1 and R2 (Arikawa,2003). In the dorsal eye, one subtype contains B–B in R1 and R2 and PxRh2 (orange ovals) in R5–8, another subtype contains UV–B in R1 and R2 and PxRh3 (yellow ovals) in R5–8 and a third subtype contains UV–UV in R1 and R2 and PxRh2 and PxRh3 (orange–yellow) in R5–8. The ventral eye contains the same three ommatidial subtypes as in the dorsal eye, except as noted above, the PxRh1 opsin is also co-expressed with the PxRh2opsin in R3 and R4. Altogether, there are at least six classes of ommatidia with respect to opsin expression in the main retina of P. xuthus(Fig. 4).

Interestingly, the most phylogenetically basal of the P. xuthus LW opsin gene duplicates expressed in the eye, PxRh2(Fig. 3), is also the one that has the widest expression in the R3–8 cells. And, like the LW opsin of V. cardui (Briscoe et al.,2003) and P. glaucus(Fig. 5K,L), it is also expressed in the R9 cell. This pattern is consistent with an ancestral function for PxRh2 and a subfunctionalization of the more recent duplicates, PxRh1 and PxRh3(Briscoe, 2000), into a more limited domain of expression. It is also worth noting that PxRh3exemplifies neofunctionalization, as it encodes P575, whose absorbance spectrum maximum is quite red shifted compared with PxRh2, which encodes P530 [see discussion in Briscoe(Briscoe, 2000)].

In addition to the visual pigments, all four butterfly eyes profiled here contain one or more combinations of yellow, orange or red filtering pigments(Arikawa et al., 1999b; Arikawa, 2003; Stavenga and Arikawa, 2006). Even more remarkable is that the eye of Pieris rapae crucivora is sexually dimorphic with respect to the expression of a blue-absorbing filtering pigment that is present only in the eye of males and is co-expressed with the violet opsin (Arikawa et al.,2005). Moreover, some butterfly eyes not reviewed here lack filtering pigments entirely (Briscoe and Bernard, 2005). Lateral filtering pigments coat the rhabdom, and filter short wavelength light, thus shifting the sensitivity of the visual pigments to the longer wavelengths. Examples of such an effect for red filtering pigments have been shown electrophysiologically in Heliconius erato (Bernhard et al.,1970; Swihart and Gordon,1971; Struwe,1972), Papilio xuthus(Arikawa et al., 1987; Arikawa et al., 1999a) and Pieris rapae crucivora (Qiu et al., 2002; Wakakuwa et al.,2004), and for the blue-absorbing filtering pigment of P. rapae (Arikawa et al.,2005).

The biological significance of filtering pigments is that while they do not expand the total visual range of the animal (this is dependent on the visual pigments themselves), they may expand the animal's color vision range. The red filtering pigments in the eyes of butterflies, however, do not appear to have a unified function. For instance, behavioral tests have shown that the nymphalid Heliconius erato uses a heterogeneously expressed red filtering pigment together with a single LW opsin to produce expanded color vision in the long wavelength range when foraging(Zaccardi et al., 2006), but the lycaenid Polyommatus icarus, which also has a heterogeneously expressed red filtering pigment in its eye, does not appear to have expanded color vision, at least when tested in the context of feeding(Sison-Mangus et al.,2008).

Providing a clearer role for filtering pigments in color vision will require further behavioral testing in a larger collection of butterfly species and, in some cases, such as the red receptor of Papilio, which co-expresses a red-filtering pigment with a red-absorbing visual pigment(Arikawa, 2003), it will be difficult to disentangle the impact of the filtering pigment on color vision from that of their coordinately expressed opsin.

The phylogeny of the butterflies provides a framework for reconstructing the pattern of opsin expression in the ancestral eye. Within the true butterflies (Papilionoidea) the current understanding of familial relationships is Papilionidae+{Pieridae+[Nymphalidae+(Lycaenidae+Riodinidae)]} (Wahlberg et al., 2005), where papilionid and pierid butterflies represent the most basal lineages, and nymphalid, lycaenid and riodinid the most derived(Fig. 6). The inferred instances of lineage-specific opsin gene duplications(Fig. 3) together with the extant pattern of opsin mRNA expression (summarized in Fig. 4) can be mapped onto the butterfly family species tree to infer the most parsimonious ancestral state,using the sphingid moth Manduca sexta as an outgroup. Remarkably, the ancestral state is not represented by the most basal butterfly lineages but,instead, is manifested in the nymphalids.

Fig. 6A shows the most parsimonious reconstruction of the ancestral pattern of opsin mRNA expression in the R1 and R2 cells. Using this reconstruction, the ancestor of all butterfly eyes had merely one UV (gray box and ovals) and one B opsin (dark blue box and ovals) gene expressed in a non-overlapping fashion in the R1 and R2 cells – a pattern shared with the opsin expression in the main retina of sphingid moths (White et al.,2003) and bees (Spaethe and Briscoe, 2005; Wakakuwa et al., 2005) and consistent with the existence of only one UV and one B opsin gene in the genome of the silkmoth, Bombyx mori(Mita et al., 2004; Xia et al., 2004).

Subsequently, sometime after the divergence of the pierid from the papilionid and nymphalid+(lycaenid+riodinid) lineages, a B opsin gene duplication arose, presumably prior to the radiation of the Pierinae+Coliadinae subfamilies, since a homologue of the P. rapae(Pierinae) V opsin has been cloned from Colias philodice(Sison-Mangus et al., 2006), a species in one of the other major pierid subfamilies, Coliadinae. [The two more basal subfamilies, Pseudopontiinae and Dismorphiinae(Braby et al., 2006; Chew and Watt, 2006) have not yet been assayed for opsins.]

The pattern of ancestral LW opsin expression can be reconstructed similarly, and doing so suggests that the common ancestor of all butterfly eyes expressed a single LW opsin in the R3–8 cells (and also probably in the R9 cell) (Fig. 6B). Under this scenario, sometime after the split between the lineage leading to Papilio and the other butterfly families, two rounds of LW opsin gene duplication occurred, followed by subfunctionalization of the newer paralogs in Papilio. Intriguingly, the topology of the LW opsin gene tree also implies that the LW opsin duplicated prior to the radiation of the riodinid and lycaenid subfamilies, and that the homolog of the riodinid LWRh1 opsin was lost in the lineage leading to lycaenids. Subsequently, in the case of the highly divergent, sexually dimorphic pattern of opsin expression in the R3–8 cells of L. rubidus, the co-expression of a B opsin, B1 in R3–8, along with the LW opsin, can be viewed as an intermediate step along the evolution of a new function for B1 opsin in the dorsal eye of males(Sison-Mangus et al.,2006).

One could imagine extending ancestral state reconstruction to other parts of the butterfly visual system to further understand the evolution of photoreceptor patterning, color vision, and even an integration of the evolution of the visual system with wing color. For instance, at present, the molecular identity, phylogenetic distribution and behavioral impact of filtering pigments in the eyes of butterflies are not well understood, but it seems likely that the ancestral butterfly eye contained heterogeneously expressed lateral filtering pigments, with secondary losses. Like the opsins,lateral filtering pigments may have evolved lineage-specific modifications and expression patterns that could have an impact on the color vision system. And,as has been pointed out, some of the pigments found in the brightly colored scales of butterfly wings are evolutionarily derived from eye pigments(Nijhout, 1991; Reed and Nagy, 2005). So it is possible that modifications of wing pigments co-evolved with the modifications of eye pigments.

Similarly, it is not known how the dorsal rim area ommatidium of the ancestral butterfly eye was patterned with respect to opsin expression. Unlike most moths, which are primarily nocturnal and may use polarized moonlight or starlight for navigation, butterflies are diurnally active and use the polarized skylight as a navigational cue(Froy et al., 2003; Reppert et al., 2004). If different opsins are expressed in the DRA ommatidia of different butterflies,this would imply that the polarized skylight cues used by butterflies are not uniform. In fact, it would not be surprising if this were the case, given the diversity of spectral classes of photoreceptors found in the DRA ommatidia of different insects (Labhart and Meyer,1999) and the apparent ease with which domains of opsin expression in butterflies have been altered. A hint that the DRA ommatidia themselves may have evolved in Lepidoptera is evident in the observation that while the R1–8 cells in monarch DRA ommatidia express only the UV opsin(Sauman et al., 2005), in the hawkmoth, Manduca sexta, only some of the photoreceptor cells in the DRA ommatidia express the UV opsin (P357), while other photoreceptor cells express none of the other (B and L) opsins cloned(White et al., 2003).

Finally, alterations in opsin expression patterns beg the question of whether or not the neural circuitry necessary to process this information has evolved. At present, correlation between opsin expression and photoreceptor axon terminals has only been attempted in the swallowtail(Takemura et al., 2005; Takemura and Arikawa, 2006). The collateral morphology of the axons of the R5–8 cells of P. xuthus varies according to which LW opsin they express. But what about in other butterflies that do not have duplicated LW opsins?

In conclusion, it is clear that the butterfly eye is one of the most beautifully honed instruments of evolution. Bringing to bear molecular tools that are now becoming available for non-model organisms (e.g. genetic transformation, custom microarrays and RNA interference) will allow a deeper understanding of the biological events involved in butterfly eye diversity that may be applicable to a broader range of biological processes.

Thanks to Lisa Nagy for providing a supportive research environment for the Papilio glaucus work, Marilou Sison-Mangus for technical assistance,Matthew McHenry for suggesting character mapping and, especially, Steven Reppert for helpful comments on the manuscript. This work was supported by NSF grants IBN-0346765 and IOS-0646060.