Spectral sensitivity of guppy visual pigments reconstituted in vitro to resolve association of opsins with cone cell types

The guppy (Poecilia reticulata) shows remarkable variation of photoreceptor cells in the retina, especially those sensitive to middle-to-long wavelengths of light. Microspectrophotometry (MSP) has revealed varying "green", "green-yellow" and "yellow" cone cells among guppies in Trinidad and Venezuela (Cumana). In the guppy genome, there are four "long-wave" opsin loci (LWS-1, -2, -3 and -4). Two LWS-1 alleles have potentially differing spectral sensitivity (LWS-1/180Ser and LWS-1/180Ala). In addition, two "middle-wave" loci (RH2-1 and -2), two "short-wave" loci (SWS2-A and -B), and a single "ultraviolet" locus (SWS1) as well as a single "rhodopsin" locus (RH1) are present. However, the absorption spectra of these photopigments have not been measured directly and the association of cell types with these opsins remains speculative. In the present study, we reconstituted these opsin photopigments in vitro. The wavelengths of maximal absorbance (λmax) were 571nm (LWS-1/180Ser), 562nm (LWS-1/180Ala), 519nm (LWS-3), 516nm (LWS-2), 516nm (RH2-1), 476nm (RH2-2), 438nm (SWS2-A), 408nm (SWS2-B), 353nm (SWS1) and 503nm (RH1). The λmax of LWS-3 is much shorter than the value expected (560nm) from the "five-sites" rule. The two LWS-1 alleles could explain difference of the reported MSP λmax values for the yellow cone class between Trinidad and Cumana guppies. Absence of the short-wave-shifted LWS-3 and the green-yellow cone in the green swordtail supports the hypothesis that this cell class of the guppy co-expresses the LWS-1 and LWS-3. These results reveal the basis of variability in the guppy visual system and provide insight into the behavior and ecology of these tropical fishes.


Introduction
The guppy (Poecilia reticulata) shows remarkable polymorphism in male body coloration (Breden, 2006). Recent studies suggest that ambient light variation between microhabitats can influence female preference of different male signal combinations, thus explaining the evolution and persistence of color variation in males (Cole & Endler, 2015;Seehausen, 2015). However, the degree of variability in the visual system and its effect on female preference for male color pattern remains poorly understood.
A microspectrophotometry (MSP) study of the Trinidad guppy revealed cone photoreceptor cells in the retina with k max values of 389 nm (''ultraviolet"), 408 nm (''violet") and 464 nm (''blue") as well as rod cells of 501 nm (Archer & Lythgoe, 1990). Another recent study of the Venezuelan (Cumana) guppy reported a similar result [359 nm (ultraviolet), 406 nm (violet), 465 (blue) and 503 nm (rod)] . It has long been known through MSP studies that, in comparison to the short-wavelength cone classes and the rod cells, the cone classes in the middle-tolong wavelength range are highly variable in spectral sensitivity among individual guppies (Archer, Endler, Lythgoe, & Partridge, 1987;Archer & Lythgoe, 1990;Watson et al., 2011). When pooling k max values recorded from multiple cells in multiple individuals, k max distribution centers around three peaks, 533, 548 and 572 nm, in a study of the Trinidad guppy (Archer & Lythgoe, 1990). A study of the Cumana guppy reported a consistent result in which k max distributions are centered at 525 nm (''green"), 541 nm (''green-yellow") and 560 nm (''yellow"), slightly shorter than in the former study .
The 548-centered distribution of k max values found in the Trinidad guppy is remarkably broader than the 533-and 572-centered distributions, implying that the cells centered at the 548-nm peak represent a cell group that expresses both 533-and 572-nm opsin genes in the same cell, in varying ratios (Archer & Lythgoe, 1990). In the study of the Cumana guppy, the k max distribution of the 541-nm class cones ranges over 10 nm (minimum of 536 nm to maximum of 546 nm) and is broader than other classes (<5 nm), though authors note their small sample size . This spectral variation is not likely attributed to differences in chromophore (A1 vs A2) use among cells because the MSP absorption spectra of the guppy cones are well fitted by an A1 chromophore template (Archer & Lythgoe, 1990;Archer et al., 1987). A similar result is also reported for congeneric molly species (Poecilia mexicana, P. latipinna and P. formosa), for which longwavelength cone classes generally show relatively high variance (Korner, Schlupp, Plath, & Loew, 2006). It is noted, however, that the green-yellow cone class is absent in the green swordtail (Xiphophorus helleri), a species closely related to the guppy, while the green and the yellow cones are observed (Watson, Lubieniecki, Loew, Davidson, & Breden, 2010).
Quantitative polymerase chain reaction (qPCR) of retinal RNA has revealed that in adult guppy retinae the expression level of SWS2-A, LWS-2 and LWS-4 is relatively low (Laver & Taylor, 2011;. Study of in situ hybridization of adult guppies has revealed that LWS-4-positive cells are found confined to the peripheral dorsal-temporal retina (Rennison, Owens, Allison, & Taylor, 2011). In contrast, SWS2-A cells, as well as SWS-1, SWS2-B and RH2-2 cells, are observed throughout the retina. LWS-2 cells, as well as LWS-1 and LWS-3 cells, are distributed broadly in the central to dorsal area of the retina. RH2-1 cells are observed throughout the retina but with higher density in ventral area (Rennison et al., 2011).
Taking into consideration the inferred k max values of the opsins and their expression pattern, a possible association of the opsins to the MSP cell types is given in the third column of Table 1. One to one association of the cell with the opsin is reasonably inferable only for: 1) the violet cone with SWS2-B, 2) the ultraviolet cone with SWS1, and 3) the rod with RH1. In other cases, matching of the k max values between the inferred opsins and the observed cells is poor, or multiple opsins are inferred to have similar k max values. Although the assignment of LWS-1/180 Ala to the green-yellow cone is also suggested by previous studies based on the inferred k max and its high expression level (Laver & Taylor, 2011;Watson et al., 2011), this is opposed to the hypothesis that the green-yellow cones express two opsin types (Archer & Lythgoe, 1990). Archer and Lythgoe (1990) showed that the absorption curves of the yellow and green cones fit well to one-opsin templates while that of the green-yellow cone did not. They do, however, fit well to a mixture model. Assignment of an RH2 opsin to the green cone has also been suggested by previous studies (Hofmann & Carleton, 2009;Laver & Taylor, 2011;Watson et al., 2010). If the co-expression of the green-cone opsins and the yellow-cone opsins within the green-yellow cone type is true, it is unclear how an RH2 opsin gene (RH2-1) and an M/LWS gene are co-regulated, because the transcriptional regulation over two opsin classes (RH2 and LWS) would not be more likely than regulation within each opsin class (RH2 or LWS) (Tam et al., 2011;Tsujimura, Chinen, & Kawamura, 2007;Tsujimura, Hosoya, & Kawamura, 2010;Tsujimura, Masuda, Ashino, & Kawamura, 2015). Additionally, it is yet to be resolved why the green swordtail lacks the green-yellow cones, despite possessing RH2-1 and LWS-2 (P180) in the genome (Watson et al., 2010). To contribute to an improved understanding of the visual ecology of tropical fishes, we conducted in vitro reconstitution of the guppy visual opsins and measured their absorption spectra.

Animal protocols
The University of Tokyo Animal Care and Use Committee and Tohoku University approved all animal protocols in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki).

cDNA isolation
We isolated total RNA from eyes of an adult guppy individual (Poecilia reticulata) obtained from an introduced population in the Gabusoka River (Okinawa prefecture, Japan). The molecular phylogeny and genetic divergence of this population is well examined (Shoji, Yokoyama, & Kawata, 2007). We isolated cDNA containing the entire coding sequence for all the guppy opsin genes by reverse transcription (RT) PCR, using the forward and reverse primers complementary to the region immediately upstream of the initiation codon and immediately downstream of the stop codon, respectively ( Table 2). The nucleotide sequences of the primers were designed according to published sequences of the Cumana guppy opsin genes (SWS2-A, SWS2-B, LWS-1, LWS-2 and LWS-3: GenBank accession number HM540108; LWS-4: HM540107)  and to our unpublished screening of the k-phage-vectored genome libraries of guppies from the Gabusoka River and Shimoda (Shizuoka prefecture, Japan). These libraries were probed with medaka visual opsin sequences (Matsumoto et al., 2006). We cloned the PCR products into pBluescript II (SK-) plasmid vector (Stratagene, La Jolla, CA) and were subjected to nucleotide sequencing of both strands by an ABI PRISM 3130 Genetic Analyzer (Applied Biosystems Japan, Tokyo). The coding nucleotide sequences were confirmed by duplicate RT-PCRs.

Reconstitution of opsin photopigments
For the reconstitution of non-LWS class opsin photopigments, we followed a conventional procedure (Kawamura & Yokoyama, 1998;Matsumoto et al., 2006). In brief, the pMT5 clones were transfected into cultured COS-1 cells (RIKEN Cell Bank, Tsukuba, Japan). We incubated cells with 5 lM 11-cis retinal (Storm Eye Institute, Medical University of South Carolina, Charleston), and solubilized them with 1% dodecyl maltoside. We purified the reconstituted photopigments using immobilized 1D4 antibody (Molday & MacKenzie, 1983) (University of British Columbia, Vancouver). For LWS opsins, we shortened the incubation duration after transfection from 55-63 h to 50-51 h, in accordance with a study on mice reporting that a significant amount of mislocalized M opsin was degraded 55 h after transfection (Zhang, Zhang, Baehr, & Fu, 2011). The absorption spectra of the photopigments were recorded across 250-750 nm in 0.5 nm intervals using the U3010 dual beam spectrometer (Hitachi Ltd., Tokyo) at 20°C. Measurements were made a minimum of five times in dark and at least five more times after 3 min of light exposure from a 60-watt room lamp (in the case of RH1, RH2s and LWSs) with Kodak Wratten Gelatin Filter No.3, which cuts off wavelengths shorter than 440 nm, or exposure to 350-nm UV light (in the case of SWS1 and SWS2s). Recorded spectra were analyzed and the difference spectra between before and after irradiation were calculated with UV Solution Software (Hitachi Ltd., Tokyo).
We also reconstituted the LWS-2 and LWS-3 photopigments following the method described by Imai, Terakita, and Shichida (2000) with some modifications. Briefly, opsins were expressed in HEK293T cells and reconstituted with 11-cis retinal after harvesting the cells. Pigments were extracted with buffer E (50 mM HEPES, 0.75% CHAPS, 1 mg/ml phosphatidylcholine, 140 mM NaC1) under the red light (>670 nm) at 4°C. A spectrophotometer (Model UV-2450, Shimadzu Co., Ltd, Kyoto, Japan) was used to measure the UV-visible spectra. For these measurements, samples containing 10 mM hydroxylamine were irradiated for 5 min with light from a 1-kW tungsten halogen lamp (Rikagaku Seiki) that had been passed through a glass cutoff filter (omitting light 6520 nm). Difference spectra between before and after irradiations were calculated with Igor Pro software (Wavemetrics, Portland, OR).

Results
For eight opsins (LWS-1/180 Ser , LWS-1/180 Ala , RH2-1, RH2-2, SWS2-A, SWS2-B, SWS1 and RH1) we directly estimated the k max from absorption spectra measured in dark conditions (the fourth column of Table 1, Figs. 1 and 2). We validated this as follows: in the dark spectra, a peak at 280 nm appears due to absorbance of proteins. When we exposed the reconstituted pigments to light, a new absorption peak appeared at $380 nm, represented as a negative peak in the dark-light difference spectra (insets of Figs. 1 and  2). This indicates that light isomerized 11-cis retinal in the pigments and the resultant all-trans retinal was released, demonstrating that the reconstituted pigments were photo-reactive. We verified the ultraviolet sensitivity of SWS1 pigment by acid denaturation as in Chinen, Hamaoka, Yamada, and Kawamura (2003).
The difference spectrum is useful in the case of a low absorption peak in the dark spectrum. For two opsins (LWS-2 and LWS-3) we estimated the k max by using the difference spectra (the fourth col-umn of Table 1, Fig. 1). For LWS-4 opsin, we failed to obtain recognizable peaks even from difference spectrum and despite repeated trials.

Discussion
We reconstituted the visual opsins of the guppy in vitro and measured their absorption spectra. Difficulties in functional reconstitution of opsins have been reported for heterologous gene expression assays, possibly due to malfunction of cellular processes such as post-translational modifications, trafficking and localization (Chinen, Matsumoto, & Kawamura, 2005;Reeves, Callewaert, Contreras, & Khorana, 2002;Reeves, Kim, & Khorana, 2002;Zhang et al., 2011). In this study, reconstitution of the LWS opsins was particularly challenging. By testing several cell lines and culture conditions, the absorption peak of LWS-1/180 Ser and LWS-1/180 Ala became prominent in the dark spectra and that of LWS-2 and LWS-3 became detectable in the difference spectra. Reconstitution of LWS-4 remained unsuccessful and may require co-introduction of assisting genes such as those encoding receptor-transporting protein family as seen in the case of olfactory receptors (Zhuang & Matsunami, 2008).
Regarding SWS1, SWS2, RH2 and RH1 genes, the observed k max values were mostly concordant with those of the medaka orthologs (see the second and the fourth columns of Table 1). For LWS-2, the observed k max value (516 nm) was concordant with prediction by the five-sites rule and consideration of Ser 180 Pro effect (515 nm). On the contrary, the observed k max value of LWS-3 (519 nm) deviated largely from prediction by the five-sites rule (560 nm). Deviation was also notable in LWS-1/180 Ser (571 nm observation vs 560 nm prediction) and LWS-1/180 Ala (561 nm vs 555 nm).
Amino acid replacements responsible for these deviations from the five-sites rule should be clarified in future studies by sitedirected mutagenesis as in Matsumoto et al. (2014). Here we report that the guppy LWS-3 contains an amino acid replacement at a highly conserved site, Tyr 178 Phe (the site number corresponds to that in bovine rhodopsin). This site falls within a track of amino-acid sequence, 174Gly-Trp-Ser-Arg-Tyr178, in the second intradiscal/extracellular loop that is highly conserved in vertebrate visual opsins, and is essential to opsin structure (Doi, Molday, & Khorana, 1990). Thus, this site would be an ideal candidate for the mutagenesis experiment. Interestingly, LWS-3 of the green swordtail (Xiphophorus helleri) (S180-2) has Tyr at this site and differs from LWS-1 (S180-1) at only one amino acid site (Val in S180-1 and Ile in S180-2 at their 62nd residue located in the first transmembrane domain) (GenBank GQ999832). The k max value of the green swordtail LWS-3 is hence predicted be identical to that of its LWS-1.
Previously, on the basis of inferred k max values, LWS-3 and/or LWS-4 (k max both inferred to be at $560 nm) were suspected to be expressed in yellow cones, LWS-1/180 Ala ($555 nm) in greenyellow cones and LWS-2 ($515 nm) or RH2-1 (492-516 nm) in green cones (Hofmann & Carleton, 2009;Laver & Taylor, 2011;Watson et al., 2010Watson et al., , 2011. This assignment appears consistent with an observation that the green swordtail is reported to lack the green-yellow cones and to have 180 Ser fixed in LWS-1 (i.e. LWS-1/180 Ala is not found) (Watson et al., 2010. However, the assignment of a single gene (LWS-1/180 Ala ) to the greenyellow cones appears not compatible with the broad distribution of k max of this cone class (Archer & Lythgoe, 1990;Korner et al., 2006;Watson et al., 2011) and with the poor match between the absorption curve of the cone class and the single-opsin template (Archer & Lythgoe, 1990). Archer and Lythgoe (1990) proposed the co-expression model of the yellow and the green cone opsins in the green-yellow cones to account for the large variance.
Given the novel finding of guppy visual opsin k max values presented in this study, and of the recent finding of 180 Ser as an allele of LWS-1 (Tezuka et al., 2014) and of varying allelic frequencies of LWS-1/180 Ser and LWS-1/180 Ala among localities Sandkam, Young, Breden, Bourne, & Breden, 2015;Tezuka et al., 2014), our understanding of association of opsins with cone types in guppy can be greatly promoted (the fifth column of Table 1). Archer and Lythgoe (1990) collected guppies from rivers in Trinidad including Aripo and reported that yellow cone cells have k max at 572 nm by the MSP method. The LWS-1/180 Ser allele is found common in northern Trinidad including the Aripo river (Tezuka et al., 2014). On the other hand, the Cumana guppy is reported to have yellow cones with k max at 560 nm by the MSP method and to have the LWS-1/180 Ala allele . The present study suggests that the two alleles of LWS-1 (having k max at 571 and 562 nm, respectively) explain the difference of k max of yellow cones between the two MSP studies.
Our unexpected finding of the short k max (519 nm) for the guppy LWS-3 could explain why the green-yellow cone is present in the guppy but is absent in the green swordtail (Watson et al., 2010), if we postulate that co-expression of LWS-1 and LWS-3 occurs in a subset of cones in closely related Xiphophorus and Poecilia fishes. This satisfies the Archer and Lythgoe (1990) co-expression model for the green-yellow cones. The difference of green-yellow cones in k max between Trinidad (centered at 548 nm) and Cumana (541 nm) guppies could also be explained by co-expression model because of their difference of the yellow cones.
The difference of the present co-expression model from the original Archer and Lythgoe model (1990) is that the co-expressed opsin with the yellow cone opsin (LWS-1) is not the green cone opsin. The green cone is present in both the green swordtail (Watson et al., 2010) and the guppy and cannot be compatible with LWS-3 because of the long-wave absorption predicted for the green swordtail. Although k max values of the LWS-2 and RH2-1 photopigments of the green swordtail need to be confirmed, in the guppy the two opsins have similarly short k max values ($516 nm) and differ in major expression area in the retina (RH2-1 in ventral retina and LWS-2 in central and dorsal retina: Rennison et al. (2011)). Thus, the green cones can be assigned to either RH2-1 or LWS-2 in different part of the retina. This assignment precludes the necessity of postulating co-expression of different classes of opsins in the green-yellow cones: an RH2 opsin (RH2-1) and an M/LWS opsin (LWS-1). However, the problem still remains: the k max values of the two opsins are shorter than the MSP values of the green cones in both Trinidad and Cumana guppies (the fourth and fifth columns of Table 1). Further survey of allelic polymorphism of LWS-2 and RH2-1 and improvement of their in vitro reconstruction could finally solve this problem.
We still have no clear association of LWS-4 and SWS2-A with cell types (the fourth and fifth columns of Table 1). LWS-4 is expressed only in the peripheral dorsal-temporal retina (Rennison et al.,  (Laver & Taylor, 2011;. Considering this expression pattern, the role of LWS-4 in vision may be negligible, if any. SWS2-A is expressed uniformly throughout the retina but its expression is much lower than SWS2-B (Laver & Taylor, 2011;Rennison et al., 2011;. SWS2-B is also expressed uniformly throughout the retina (Rennison et al., 2011). SWS2-A (k max at 438 nm) may be co-expressed with SWS2-B (408 nm); Archer and Lythgoe (1990) found that the measured absorption curve of the violet cones is slightly long-wave shifted relative to the best fitting template. Thus, we need further endeavor to decipher visual system of the guppy.

2011) and its expression level is low
Despite a few unresolved associations, we emphasize the importance of direct measurement of opsin absorption spectra to provide a solid basis for understanding the evolution of the visual system. Comparison of opsin absorption spectra between populations (e.g. LWS-1 between Trinidad and Cumana guppies) and between species (e.g. LWS-3 between guppy and swordtail) is also key to understanding color vision variation among them. With results we generate in this study, M-L cone spectral variation of guppies described by Archer and Lythgoe (1990) can be interpreted with a new view: the relative expression level among LWS-1, LWS-2 and LWS-3 may differ substantially among individuals, resulting in the broad k max range of the green-yellow cones and the various ratio of the yellow, green-yellow and green cones in the guppy retina. The expression plasticity of opsins in evolution and in ontogeny is a focus of recent studies on guppies under varying predation and lighting conditions (Sakai, Ohtsuki, Kasagi, Kawamura, & Kawata, 2016;. Deciphering genetic mechanism behind the plasticity of opsin expression will further promote our understanding of the color vision, behavior and ecology of these tropical fishes.

Conflict of interest
The authors declare no conflict of interests.

Author contributions
1: Conception of the research, 2: Design of the experiments, 3: Collection of data, 4: Analysis of data, 5: Interpretation of data, 6: Drafting the manuscript, 7: Revising the manuscript critically for important intellectual content.

Funding
Supported by a Grant-in-Aid for Scientific Research A (15H02421), a Grant-in-Aid for Challenging Exploratory Research (23657164) from JSPS and Grants-in-Aid for Scientific Research on Priority Areas 'Comparative Genomics' (20017008) and 'Cellular Sensor' (21026007) from MEXT to SK and by Grants-in-Aid for Scientific Research to MK (15H04419) and to HI (15H05242) from JSPS.

Acknowledgments
We thank Dr. Amanda D. Melin for proofreading and commenting on the manuscript. We are grateful to anonymous reviewers for their advices to improve the manuscript. We are also grateful to Drs. Felix Breden and Corey T. Watson for helping us retrieving opsin sequences from their guppy and swordtail BAC clones as references during this study. Fig. 2. Absorption spectra of RH2, RH1, SWS2 and SWS1 opsins reconstituted in vitro. The dark spectra and their k max values are shown with difference spectra in insets. The discontinuity of absorbance curve at around 350 nm appeared in the charts except of SWS2-A. This is due to an electronic noise when light sources changed between ultraviolet and visible light.