Identification and characterization of visual pigments in caecilians (Amphibia: Gymnophiona), an order of limbless vertebrates with rudimentary eyes Free

The Gymnophiona, commonly known as caecilians, are one of the three orders of extant Amphibia. They are readily distinguished from frogs (Anura) and salamanders (Urodela) by their annulated, limbless bodies, and are distinct in many other characters, including rudimentary eyes (e.g. Taylor, 1968; Wilkinson and Nussbaum, 2006). Reduction in the visual system of the Gymnophiona, associated with a mainly fossorial lifestyle, has long been recognized (see Norris and Hughes, 1918; Walls, 1942) but the extent to which the caecilian eye is functional remains uncertain (Wake, 1980; Wake, 1985; Himstedt, 1995; Himstedt, 1996). Adult caecilian eyes are relatively small, covered with skin and sometimes bone, lack features expected of well-developed vertebrate eyes and have several features apparently co-opted to novel sensory structures and functions (e.g. Himstedt and Simon, 1995). Relaxed molecular clock analysis suggests that the last common ancestor of extant caecilians occurred in the Triassic or Jurassic (Roelants et al., 2007; San Mauro, 2010) and that caecilians may therefore have possessed a reduced visual system for over 150 million years.

It is widely assumed that variations in the peak spectral sensitivities (λ_max) of the visual system represent adaptations to specific visual needs associated with particular habitats or lifestyles (Lythgoe, 1979; Peichl, 2005; Davies et al., 2009). Spectral sensitivity is determined either by changes in the tertiary structure of the opsin protein via amino acid substitutions and/or by the use of alternate chromophores derived from vitamin A1 or A2 (Crescitelli et al., 1972; Yokoyama, 2000; Bowmaker and Hunt, 2006; Hart and Hunt, 2007). Rods, which generally express the rh1 opsin gene and rod-specific components of the phototransduction cascade, are generally responsible for scotopic or dim light vision. By contrast, up to four classes of cone opsin genes exist in vertebrates that are associated with colour and photopic vision (Hunt et al., 2001a; Bowmaker and Hunt, 2006). The encoded proteins of these genes have λ_max values (based on a vitamin A1 chromophore) between about 360–440 nm (‘ultraviolet’ or ‘violet’, short wavelength-sensitive; the sws1 gene) 440–470 nm (‘blue’, short wavelength-sensitive; the sws2 gene), 470–520 nm (‘green’, middle wavelength-sensitive; the rh2 gene) and 500–570 nm (‘red’, long wavelength-sensitive; the lws gene).

Electroretinography has detected a single class of visual pigment in the caecilian Ichthyophis cf. kohtaoensis (Himstedt, 1995), and this was presumed to be a rod pigment based on its λ_max value at 502 nm. Consistent with this, fragments of rh1 coding sequence have been isolated from four species of morphologically and phylogenetically diverse caecilians (Frost et al., 2006; Venkatesh et al., 2001), suggesting that this class of visual pigment may be present throughout the Gymnophiona. Immunohistochemistry has also been performed on a similarly diverse group of caecilians using antibodies raised to rod opsin, transducin, and cone opsin (Nguyen, 2003). Strong reactions were observed with anti-rod antibodies and weaker reactions with one anti-cone antibody, although the latter reaction is most probably the result of cross-reactivity with the rod pigment (Nguyen, 2003; Foster et al., 1993). An anti-bovine transducin antibody also reacted positively, suggesting that the rod phototransduction pathway is intact in the retina of caecilians. Thus, a functional scotopic visual pathway would appear to be present in caecilians, even in species with rudimentary eyes concealed under bone, without any evidence for the presence of photopic vision.

The aim of this study was to identify and characterize the visual pigments expressed in the retina within ecologically and phylogenetically diverse caecilian species and to determine the λ_max of such pigments using in vivo and in vitro methods.

Adult animals were used for all experiments as follows: one specimen of Rhinatrema bivittatum (Guérin-Méneville 1829) (voucher number MW 2390) was used for microspectrophotometry (MSP); six specimens of Typhlonectes natans (Fischer 1880) (MW 7346–MW 7351), with one specimen being used for MSP and five specimens for mRNA extraction; one specimen of Geotrypetes seraphini (Duméril 1859) (MW 3854) was used for MSP; two specimens of Ichthyophis cf. kohtaoensis (Taylor, 1960) (MW 4764 and MW 4765) for mRNA extraction. All caecilian species were obtained by fieldwork or from licensed commercial sources as approved by the local Animal Ethics Committee and were killed using an approved procedure. All vouchers will be accessioned into the collections of the Natural History Museum, London.

Physiological data were determined using MSP. A modified Liebman dual beam microspectrophotometer under computer control was used to determine the spectral sensitivities of the photoreceptors. With the help of an infrared converter, the measuring beam (normally 2 μm square cross section) was aligned to pass transversely through the photoreceptor outer segments or transparent pieces of skin covering the eye, while the reference beam passed through a clear space adjacent to the object examined. Spectra were scanned from 750 to 350 nm in 2 nm increments and back from 251 to 749 nm at the interleaved wavelengths. Only one absorption spectrum was usually obtained from a given outer segment to minimize the effects of bleaching. To verify the presence of a photolabile pigment, putative outer segments were bleached by exposure to white light from the monochromator passing through measuring beams of the microspectrophotometer. A standardized computer program was used to estimate the λ_max for each outer segment. The spectra obtained from the outer segments were averaged to obtain a mean curve. The curve was then fitted to a standard template curve in order to obtain an estimate of λ_max (Govardovskii et al., 2000). The standard template curve used was the Dartnall standard curve for rhodopsin placed with its λ_max at 502 nm and expressed on an abscissal scale of log frequency.

Complementary DNA (cDNA) was generated from mRNA extracted from the eyes of two species of caecilian, I. cf. kohtaoensis and T. natans, using three and ten eyes, respectively. The QuickPrep™ micro mRNA purification kit (GE Healthcare, Little Chalfont, UK) was used to purify polyadenylated mRNA. First strand cDNA was synthesized in a reaction using 500 ng of oligo(dT) or for 3′-RACE primer from the 5′-/3′-RACE Kit (Roche, Burgess Hill, UK) and 1–2 μg of mRNA, 1 i.u. Superscript III reverse transcriptase (Invitrogen, Paisley, UK), following the manufacturer's instructions.

Fragments of opsin coding sequence were amplified from cDNA using degenerate primers designed from an alignment of vertebrate opsins (Davies et al., 2007a; Davies et al., 2007b). Polymerase chain reaction (PCR) amplifications using genomic DNA from the same caecilian species and amphibian cone class-specific and degenerate primers were also performed (supplementary material Table S1). A nested protocol was used to detect sequences from exon 1 to exon 4. From the sequences obtained using standard nested PCR, specific primers (supplementary material Table S1) were designed for 5′- and 3′-RACE to isolate the untranslated regions (UTRs) of the rod opsin mRNA transcript. The reaction mixture typically contained 10 mmol l^–1 of each dNTP, 2 μmol l^–1 MgCl₂, 0.1 μmol l^–1 of each forward and reverse primer, 2.5 i.u. Biotaq DNA polymerase (Bioline, London, UK) and 20–100 ng of cDNA. Cycling conditions were, an initial denaturing step at 94°C for 5 min, then 40 cycles of denaturing at 94°C for 30 s, annealing at 55°C for 1 min and extension at 72°C for 1.5 min, with a final extension at 72°C for 10 min. Negative controls were run in parallel containing the same reagents but without the addition of cDNA. A second round amplification used 1/10 dilution of the first round PCR products, under identical conditions to the first round except for an annealing temperature of 60°C.

The 5′ and 3′ ends of the visual pigment cDNA were amplified using the 5′/3′-RACE kit (Roche), following the manufacturer's instructions. PCR products were cloned into pGEM T-Easy vector (Promega, Southampton, UK) prior to sequencing using either T7 or SP6 primers and the Big Dye Terminator v3.1 Cycle Sequencing kit on an ABI 3730 sequencer (Applied Biosystems, Warrington, UK).

The full-length coding sequences of T. natans and I. cf. kohtaoensis rh1 were amplified using a proofreading KOD XL Taq polymerase (Novagen, Nottingham, UK) (supplementary material Fig. S1). The primer pairs were designed for the 5′ and 3′ ends with restriction sites for EcoRI and SalI. The digested PCR products were directionally cloned into a derivative of the mammalian expression vector pMT4 carrying the sequence of bovine rod opsin 1D4 epitope (including the stop codon) at the 3′ end of the coding sequence (Franke et al., 1988). HEK293T cells were transfected using a calcium phosphate precipitation technique with vector containing the opsin insert. Cells were harvested after 48 h and washed with 1× phosphate-buffered saline (PBS) buffer. Pigments were reconstituted by resuspending the harvested cells in PBS buffer and incubating with 20 μmol l^–1 11-cis retinal in the dark. The pigments were isolated by incubating with 1% (w/v) n-dodecyl-β-d-maltoside (DDM) and 20 mg ml^–1 phenylmethylsulphonylfluoride (PMSF) before passage over a cyanogen-bromide-activated (CNBr) Sepharose binding column coupled to an anti-1D4 monoclonal antibody.

The dark and the bleached (photobleached with broad spectrum white light for at least 30 min) absorption spectra were measured using a dual path spectrophotometer (Spectronic Unicam, Cambridge, UK). The λ_max was determined from the difference spectra (dark minus light bleached) by fitting a Govardovskii A1 template (Govardovskii et al., 2000) to the data using an Excel program.

An alignment of coding region amino acids was done manually in MacClade 4.08 (Maddison and Maddison, 2000), using the translated caecilian sequences and 68 vertebrate visual opsin sequences downloaded from the NCBI database (alignment and GenBank accession numbers are given in supplementary material Fig. S1), representing the five classes of visual opsins and including all amphibian representatives (typically up to three where available and including any with rudimentary eyes) of major vertebrate lineages. Bayesian inferences used Markov Chain Monte Carlo (MCMC) implemented in MRBAYES 3.1.2 (Huelsenbeck and Ronquist, 2001; Ronquist and Huelsenbeck, 2003) sampling every 1000 generations from two parallel runs (each of four chains) of 1,000,000 generations. Convergence was established by comparing the standard deviations of split frequencies and burnins determined by visual inspection of likelihoods using TRACER 1.4 (Rambaut and Drummond, 2007) (Tracer v1.4, Available from http://beast.bio.ed.ac.uk/Tracer). Model jumping was used within MRBAYES 3.1.2 to estimate the fixed-rate model for amino acid data during the tree-search, which chose the BLOSUM62 (BLOck SUbstitution Matrix) model of amino acid substitution that was produced from multiple alignments of evolutionarily divergent proteins (Henikoff and Henikoff, 1992). The analysis was rooted using LWS opsin protein sequences as outgroups. The sequences were identified using BLASTP 2.2.20 and BLASTN (Altschul et al., 1997) against published opsin sequences from the NCBI database.

Three species from ecologically diverse and phylogenetically distinct caecilian lineages were investigated. Rhinatrema bivittatum (a species that retains many ancestral features) and Geotrypetes seraphini, are terrestrial as adults, whereas Typhlonectes natans is aquatic. All have relatively well-developed eyes for caecilians, but those of T. natans and G. seraphini are possibly secondarily well developed. In all three species, only a single class of rod photoreceptors was found, and microspectrophotometry (MSP) demonstrated the presence in each species of a single visual pigment with similar λ_max values of 488.5±0.9 nm for T. natans (N=31), 487.5±1.0 nm for R. bivittatum (N=13), and 486.6±2.2 nm for G. seraphini (N=8; Fig. 1).

The light transmission of the skin immediately above the eye of T. natans was determined by MSP. Fig. 2 shows that this skin is almost transparent with only a small amount of absorbance attenuation of about 10% at wavelengths <500 nm.

Of the three species used for the MSP analysis, only eye tissue samples from T. natans were available for molecular studies. A fourth species, I. cf. kohtaoensis, was therefore added to the study at this stage. Opsin sequences were PCR amplified from cDNA; in all cases, the sequence of the amplified fragments corresponded to the rh1 gene. This was confirmed by phylogenetic analysis, with both caecilian sequences falling, as expected, within the amphibian rh1 group as a sister group of Batrachia (Fig. 3). The aligned sequences used are shown in supplementary material Fig. S1. No cone opsin sequences were obtained by PCR amplifications from genomic DNA or cDNA using either gene specific or degenerate primers.

The full rh1 coding and deduced amino acid sequences for T. natans and I. cf. kohtaoensis are shown in supplementary material Figs S1 and S2. The sequences show an 80–82% nucleotide and a 87–89% amino acid identity for T. natans and a 79–81% nucleotide and a 86–90% amino acid identity for I. cf. kohtaoensis to the rh1 orthologues of other amphibian species. There is a 96%amino acid identity between T. natans and I. cf. kohtaoensis sequences.

In order to confirm that the cloned gene sequences encoded the pigments identified in situ in photoreceptors by MSP, the T. natans pigment was generated in vitro by isolation of recombinant opsin from mammalian cells transfected with an expression vector containing the opsin coding sequence. After pigment reconstitution by the addition of 11-cis-retinal, the dark and bleached absorption spectra were measured by conventional UV–visible spectrophotometry. The difference spectrum yielded a λ_max at 493 nm (Fig. 4), which is sufficiently similar to the photoreceptor peak at 488.5±0.9 nm obtained in situ by MSP to confirm the identity of the photoreceptor pigment as rod opsin. The opsin coding sequence of I. cf. kohtaoensis was also expressed and regenerated in vitro to give an identical λ_max to that of T. natans at 493 nm.

Compared with the rod pigments of other Amphibia, which peak around 502–506 nm (Partridge et al., 1992; Chen et al., 1996; Fyhrquist et al., 1998; Ala-Laurila et al., 2002), the caecilian pigments measured in situ are short wavelength-shifted by between 13 and 19 nm. This shift must be the result of particular amino acid differences in the caecilian rh1 opsin pigments compared with other amphibian rh1 opsins. Key amino acid sites 83, 122, 207, 211, 265, 292, 295, which are known to be important in the spectral tuning of rh1 pigments (Davies et al., 2007a) are invariant across all the amphibian pigments (Fig. 5). However, the two caecilian pigments show consistent differences from other amphibians at sites 107, 158, 159, 213, 260, 289, 337 and 352 (Table 1) which may influence the spectral tuning of caecilian visual pigments.

Only a single spectral class of photoreceptor was found in three caecilian species, Rhinatrema bivittatum, Typhlonectes natans and Geotrypetes seraphini, with mean absorbance peaks between 487–489 nm. Consistent with this, the only expressed opsin detected in the eye of two caecilian species, T. natans and I. cf. kohtaoensis, belongs to the rod or Rh1 class of opsins that is expressed in rod photoreceptors throughout the vertebrate kingdom. No other classes of photoreceptors or expressed opsins were found and it is concluded that these caecilian species possess a rod-only retina, which agrees with previous research (Wake, 1985; Himstedt, 1995). Because cone opsins homologous with those of other vertebrates are found in members of the Urodela (salamanders) and Anura (frogs) (Rohlich and Szel, 2000; Takahashi et al., 2001; Sakikabara et al., 2002), it would appear that the loss of cone opsins, and cone photoreceptors, is a synapomorphy of Gymnophiona.

The peak sensitivities of rod visual pigments in Amphibia typically lie between 502–506 nm (Liebman and Entine, 1968; Harosi, 1975; Partridge et al., 1992; Chen et al., 1996; Fyhrquist et al., 1998; Ala-Laurila et al., 2002); peak sensitivities of caecilian rod pigments are therefore short wavelength-shifted by between 13 and 19 nm. These shifts are comparable to those seen in a number of deep-water vertebrates (Hunt et al., 2001b; Bowmaker et al., 1994; Hope et al., 1997) and marine mammals (Fasick et al., 1998; Fasick and Robinson, 2000) where they are thought to be an adaptation to the narrow bandwidth of down-welling light centred around 485 nm that penetrates deep water (Hunt et al., 2001b). These shifts involve different combinations of substitutions at a number of amino acid sites that include 83, 261, 292 and 299 (Hunt et al., 1996; Fasick and Robinson, 2000; Hunt et al., 2001b). None of these sites are, however, substituted in the caecilian pigments when compared with the rod pigments of other amphibians. Differences are, however, found at sites 107, 158, 159, 213, 260, 281, 337 and 352; these are therefore potential spectral tuning sites, although none has yet been implicated in the tuning of visual pigments. Thus, at present, it is not possible to conclude which of these sites is responsible for the short wavelength shift of caecilian rod pigments.

The skin above the eye of T. natans shows only minimal absorption of light across the spectrum from 400 to 750 nm. Absorbance by the skin would not be expected therefore to alter the peak spectral sensitivity of the eye of this caecilian species. This differs from the situation in I. cf. kohtaoensis where absorbance of the skin covering the eye approaches 100% at 410 nm, reducing to around 40% at 500 nm and to almost 0% at 550 nm (Himstedt, 1995). The effect of this is to shift the peak sensitivity of the eye to around 520 nm, and this remains the case even when the true peak of 493 nm obtained from the present data is substituted for the peak of 502 nm assumed by Himstedt (Himstedt, 1995).

The cave salamander Proteus anguinus shows a similar morphological diversity and rudimentation of the eye to that in caecilians (Kos et al., 2001). One population from the Otovec doline in south-west Slovenia appears to parallel caecilians in the loss of cone opsins while retaining the rod pigment. The eyes of salamanders from this population have regressed and photoreceptors have no discernable outer segments (Kos et al., 2001). This contrasts with another population from the Planina cave which have retained both a long wavelength-sensitive (red) cone opsin and a rod opsin, as determined by immunocytochemistry. The estimated time of divergence of surface and cave-dwelling salamanders is 2–5 million years (Trontelj et al., 2007). Therefore, loss of cone pigments in cave-dwelling species has occurred over a relatively short evolutionary time and their loss in caecilians, which diverged from other amphibians at a much earlier time, is unsurprising.

Caecilians are found in both aquatic (swamps, rivers) and terrestrial (forest, open bush and agricultural landscapes) habitats in the wet tropics (e.g. Gower et al., 2004; Jones et al., 2006). Terrestrial species live mostly in soil but may also exploit epigeic microhabitats such as under leaf litter and rotting vegetation (e.g. Burger et al., 2007). Caecilians (at least those species that are partially surface active) are mostly considered nocturnal (Himstedt, 1995; Kupfer et al., 2004; Burger et al., 2007), so it is likely that the peak absorbance of caecilian rod photoreceptor pigments are spectrally tuned to maximize the absorbance of light and the wavelengths available in either a forest floor, soil or night environment. Tropical forests are heterogeneous in the spectral composition of ambient light (Endler, 1993), where the light varies from white in large gaps between vegetation to yellowish-red in small gaps. In forest shade, essentially all of the light has been transmitted through or reflected from leaves to give a range of wavelengths of 470–490 nm (Endler, 1993), similar therefore to the maximum absorbance of caecilian rod visual pigments. By contrast, at night the wavelengths of light are long wavelength-shifted, but there is no evidence for a shift towards longer wavelengths in any nocturnal animals (Peichl, 2005), including caecilians. The light at twilight is predominately around 480 nm, which is similar to the λ_max of caecilian pigments, and after the sun sinks below the horizon, the loss of middle wavelengths becomes more pronounced (Endler, 1993). Therefore, it is possible that the main role of the rod visual pigment in caecilians is to detect twilight, a critical period during which diurnal and nocturnal species change behaviour patterns and locations and when prey detection and predator avoidance becomes highly significant (Munz and Mcfarland, 1977).

We are grateful to Dr Rosalie Crouch for the generous gift of 11-cis retinal. We thank Alex Kupfer (Universität Jena), Jeannot and Odette (Camp Patawa), Guy Tiego (Direction regionale de l'environment Guyane), Céline Dupuy (Direction des Services, Vétérinaires de la Guyane), and especially Philippe Gaucher (Centre National de la Recherche Scientifique) for facilitating our collection of Rhinatrema bivittatum in French Guiana. We would also like to thank Dr Livia Carvalho, Dr Susan Wilkie and Dr Jill Cowing for technical assistence with the in vitro regeneration of caecilian visual pigments.