Heterologous Expression of Limulus Rhodopsin*

Invertebrates such as Drosophila or Limulus assemble their visual pigment into the specialized rhabdomeric membranes of photoreceptors where phototransduction occurs. We have investigated the biosynthesis of rhodopsin from the Limulus lateral eye with three cell culture expression systems: mammalian COS1 cells, insect Sf9 cells, and amphibian Xenopus oocytes. We extracted and affinity-purified epitope-tagged Limulus rhodopsin expressed from a cDNA or cRNA from these systems. We found that all three culture systems could efficiently synthesize the opsin polypeptide in quantities comparable with that found for bovine opsin. However, none of the systems expressed a protein that stably bound 11-cis-retinal. The protein expressed in COS1 and Sf9 cells appeared to be misfolded, improperly localized, and proteolytically degraded. Similarly, Xenopus oocytes injected with Limulus opsin cRNA did not evoke light-sensitive currents after incubation with 11-cis-retinal. However, injecting Xenopus oocytes with mRNA from Limulus lateral eyes yielded light-dependent conductance changes after incubation with 11-cis-retinal. Also, expressing Limulus opsin cDNA in the R1-R6 photoreceptors of transgenic Drosophila yielded a visual pigment that bound retinal, had normal spectral properties, and coupled to the endogenous phototransduction cascade. These results indicate that Limulus opsin may require one or more photoreceptor-specific proteins for correct folding and/or chromophore binding. This may be a general property of invertebrate opsins and may underlie some of the functional differences between invertebrate and vertebrate visual pigments.

In all animals, visual pigments consist of an apoprotein and bound chromophore, usually 11-cis-retinal (most animals), 3,4-didehydro-11-cis-retinal (some fish, amphibia, and reptiles), or 3-hydroxy-11-cis-retinal (some insects) (1)(2)(3). Typically, in both vertebrates and invertebrates the visual pigment is very stable in the dark, with thermal isomerization rates measured in years (4,5). This stability contributes significantly to the overall limit of visual sensitivity. In addition, the mechanism of visual pigment activation is similar in vertebrates and invertebrates, in that absorption of a photon causes the isomerization of the chromophore from the cis to the trans conformation. This event initiates a series of conformational changes in the chromophore-protein complex that lead to the formation of an active state that interacts with G-proteins to regulate membrane conductance (6 -8).
Despite these similarities, the invertebrate and vertebrate pigments differ in at least two significant ways. First, the active states have different thermal stabilities. In vertebrates, the active state, metarhodopsin II bleaches (or decomposes) rapidly into the apoprotein and free all-trans-retinal. In contrast, invertebrate metarhodopsin (M) is stable under physiological conditions (for a review, see Ref. 9). Second, the regeneration of pigment containing the 11-cis-chromophore following exposure to light differs between vertebrates and invertebrates. Vertebrate opsins can bind 11-cis-retinal, synthesized enzymatically in the adjacent pigment epithelium, and spontaneously form the Schiff's base linkage, apparently without any additional protein cofactors (2). In invertebrates, the pigment can be efficiently converted back to the dark-adapted state by light at ambient temperature, a process called photoregeneration, or by thermal isomerization of the chromophore (9). Photoregeneration is the main mechanism by which the rhodopsin concentration is maintained in bright light (10 -12).
Studies to investigate the molecular bases for the differences between vertebrate and invertebrate pigments have been limited by the lack of a suitable expression system for invertebrate opsins that would be comparable with those available for vertebrate opsins (13)(14)(15)(16)(17)(18)(19). Here, we have evaluated a number of these systems for their suitability to express Limulus rhodopsin. We find that none of the common systems used for the expression of vertebrate opsin produce a Limulus opsin capable of binding retinal. In fact, our findings suggest that one or more protein cofactors in invertebrate photoreceptors are necessary for the functional expression of the Limulus visual pigment.

EXPERIMENTAL PROCEDURES
Nucleic Acids COS1 Expression Vector-The Limulus rhodopsin cDNA, Lim Rh1 (20), was cloned into the mammalian expression vector, pMT3 (21). In order to facilitate detection and purification of the expressed protein, the carboxyl five amino acids of Limulus rhodopsin were replaced with 14 amino acids from bovine rhodopsin, thus introducing the epitope recognized by the 1D4 monoclonal antibody (22). The resulting cDNA, Lim Rh1-1D4, contained 180 base pairs of 5Ј-untranslated sequence, the coding region for amino acids 1-362 (of 367 total) from Lim Rh1 (including an alanine at position 105, which is a polymorphic site in Limulus (20)), and the cDNA for amino acids 330 -348 from bovine rhodopsin.
Sf9 Expression Vector-Lim Rh1-1D4 was transferred into the SmaI-NotI restriction sites of the baculovirus expression vector pVL1393 (Pharmingen). To form the infecting virus, this construct was recombined with Baculogold DNA (Pharmingen) in Sf9 cells according to the manufacturer's instructions.
Drosophila Expression Vector-An expression construct was generated in which either the Lim Rh1-1D4 cDNA (referred to as P{Rh1ϩLimMT} or wild type Lim Rh1 cDNA (referred to as P{Rh1ϩLim}) was cloned into an expression cassette containing 2.5 kb of Drosophila ninaE promoter sequence, the transcription initiation site, 33 bp of the 5Ј-untranslated region, a short polylinker, and 650 bp from the 3Ј-end of the ninaE gene that includes the polyadenylation signal (23). Both constructs were subcloned into the y ϩ marked Pelement vector "C4" (24).
In Vitro Transcription-Lim Rh1-1D4 was cloned into the EcoRI-NotI restriction sites of pALTER1 (Invitrogen). Capped RNA was prepared by run-off transcription using NotI-linearized plasmid and T7 polymerase (either RiboMax or RiboMax kit from Promega or Ambion, respectively). Transcripts were polyadenylated in vitro, purified using oligo(dT) chromatography (PolyATtract, Promega, Inc.), and quantified using UV spectroscopy. Final RNA samples were suspended in water.
mRNA-Retinas from adult Limulus polyphemus (200 animals), collected in summer and fall near Woods Hole, Massachusetts, were isolated and used either immediately for RNA extraction or frozen in dry ice and stored at Ϫ70°C until use. RNA was extracted using the TRIzol reagent (Invitrogen) and poly(A) ϩ RNA isolated using oligo(dT) chromatography (PolyATtract; Promega) and eluted in water at a concentration of ϳ0.5 mg/ml.

Expression in COS1 Cells
COS1 cells were grown and transfected using the DEAE-dextran protocol (25). All Limulus rhodopsin transfection experiments were performed in parallel with bovine rhodopsin transfections as a positive control. Transfected cells were allowed to incubate 48 -72 h post-transfection. Three methods were used for regeneration of visual pigment. In method 1 (26), reconstitution was carried out in the dark by incubation of cells suspended in Buffer Y (50 mM HEPES, pH 6.6, l40 mM NaCl, 3 mM MgCl 2 , and 1 g/ml protease inhibitors (aprotinin, pepstatin, leupeptin, and benzamidine) with 5 M 11-cis-retinal (in ethanol) for 3-18 h at 4°C. In method 2, cell membranes were first isolated (27) and then resuspended in 67 mM NaPO 4 , pH 7.0, containing protease inhibitors. Reconstitution was carried out with 2 M 11-cis-retinal at 4°C for 3 h. To remove retinal, membranes were washed with 4% BSA 1 in 67 mM NaPO 4 . In method 3, COS1 cells were grown in the dark in the presence of 5 M 11-cis-retinal, beginning 24 h after DNA addition. The medium was changed twice prior to harvest of the cells. Pigment formation was determined by taking UV-visible spectra of protein purified using immunoaffinity chromatography (26) or by measuring light-dark difference visible spectra on solubilized membranes. Membranes were solubilized using 2% digitonin or 1% DM on ice for 20 min with occasional homogenization with a Teflon-glass homogenizer. Insoluble material was removed by centrifugation in a JA-20 (Beckman) rotor at 19,000 rpm for 45 min. The top 9/10 of the supernatant was removed and transferred to a new tube for spectral analysis.

Expression in Sf9 Cells
The insect cell line, Sf9, from Invitrogen was cultured in Sf-900II SFM (Invitrogen) with 10 units/ml penicillin and 10 g/ml streptomycin. The titer of the amplified Baculo/LimRh1-1D4 was about 10 8 plaque-forming units as estimated by end-point assay. Sf9 cells at 5-6 ϫ 10 6 cells/ml were infected with the Baculo/LimRh1-1D4 by incubating cells, recombinant viruses and fresh medium, mixed in the ratio of 10:1:20 (v/v/v). Sf9 cells were harvested 1-4 days after infection. To reconstitute the opsin with chromophore, Sf9 cells were harvested 2 days after infection. The chromophore, 11-cis-retinal or all-trans-retinal, was added to a final concentration of 2 M 1 day after infection and again at each of the following steps: cell harvesting, membrane solubilization, and immunoaffinity purification. Infected cells were harvested as previously described (28). Cell membranes were solubilized in buffer A (5 mM PIPES, pH 6.8, 10 mM EDTA containing protease inhibitors) with 1% DM on ice for 20 min. The following 14 protease inhibitors were used in each step of the purification: aprotinin, benzamidine, calpain, chymostatin, leupeptin, pepstatin A, phenanthroline, phenylalanine, phenylmethylsulfonyl fluoride, tosyl arginine methyl ester, tosyl lysine chloromethyl ketone, tosyl phenylethyochloromethylketone, antipain, and p-chloromercuriphenyl sulfonic acid. All of these inhibitors were from Sigma and were used in a final concentration of 5 g/ml, with the exception of phenylmethylsulfonyl fluoride (0.1 mM). The unsolubilized material was removed by centrifugation at 43,000 ϫ g at 4°C for 30 min. The expressed opsin was purified by immunoaffinity chromatography and used for spectral analysis.

Expression in Drosophila
P{Rh1ϩLimRh1} and P{Rh1ϩLimRh1-1D4} were injected into y w; sr ninaE 17 mutant embryos, and multiple independent P-element-mediated germ line transformants were obtained, as previously described (24). Four homozygous lines containing P{Rh1ϩLim1} on the X (line 4), second (line 242), and third (lines 30 and 155) chromosomes were retained. Six lines of flies that contained P{Rh1ϩLim-1D4} on the X (lines 63 and 69), second (lines 133 and 147), and third (lines (93 and 179) chromosomes were retained. The strains used in this study were constructed using visible markers and balancer chromosomes. The following mutant alleles were used: ninaE 17 , norpA P24 , and G ␣q 1 .

Native Pigment
Limulus lateral eyes were dissected from 150 adult animals over a period of 3 days. Dissections of the retina from the cornea were done in ambient room light, and animals were not dark-adapted beforehand. Retinas were placed in saline solution (275 mM NaCl, 3.7 mM KCl, 8.0 mM CaCl 2 , 7.0 mM MgCl 2 , and 23 mM MgSO 4 ), covered in foil, and allowed to dark-adapt for 3 h. Retinas were divided into batches of 40 eyes, spun briefly to remove saline, flushed with N 2 , quick frozen on dry ice and stored at Ϫ70°C. All procedures were carried out in the dark. Limulus eyes (40) were allowed to thaw on ice, suspended in 67 mM NaPO 4 , pH 6.5, containing 5 g/ml protease inhibitors (benzamidine, pepstatin, leupeptin), and homogenized in a glass tissue grinder. This homogenate was layered on a 45% sucrose cushion and spun in an SW-27 rotor at 25,000 rpm at 4°C for 25 min. The membrane band was collected, and the pellet was rehomogenized. The membrane fractions were pooled and diluted 10-fold with phosphate buffer to lower the sucrose concentration, and membranes were collected by centrifugation in a Ti35 rotor at 30,000 rpm for 25 min. The pellet was solubilized with 0.3 ml of a solution containing 2% digitonin or 1% DM in phosphate buffer and incubated in ice for 1 h. The solubilized solution was centrifuged in a JA-20 rotor at 19,000 rpm for 30 min. The top portion of the supernatant was removed and used for spectral analysis. Bleached membranes were prepared by suspending in 10 ml of 0.1 M NH 2 OH in phosphate buffer. The membranes were put on ice and exposed to the light projector (Eastman Kodak Co. projector equipped with a 300-watt tungsten bulb and a high pass 515-nm cut-off filter) for 3 h with occasional mixing. Membranes were centrifuged in a 70Ti rotor at 50,000 rpm for 15 min. To remove the retinal oxime, the membranes were washed twice in phosphate buffer containing 2% BSA and resuspended in phosphate buffer.

Regeneration of Native Opsin
Bleached membranes (from 40 eyes) were reconstituted with retinal using two methods. In method 1, membranes were incubated with 5 M 11-cis-retinal at 4°C. The incubated membranes were collected by centrifugation, washed twice with 2% BSA in phosphate buffer, and each solubilized with 0.6 ml of 2% digitonin in phosphate buffer for 1 h. The insoluble material was removed by centrifugation, and the top portion of the supernatant was taken for spectral analysis. In method 2, a Limulus eye extract, containing soluble proteins including putative retinal photoisomerase activity, was prepared (30). Limulus retina (ϳ48) were homogenized in 5 ml of PI buffer (0.1 M NaPO 4 , pH 6.8). The homogenate was spun at 8,000 ϫ g at 4°C for 40 min. The supernatant was saved, and the pellet was re-extracted in another 5 ml of buffer PI and respun. The two supernatants were pooled (10 ml total) and used as a soluble Limulus eye extract. For reconstitution, membranes were prepared from Limulus retina as usual. The resuspended membranes were divided into two. One aliquot was bleached with NH 2 OH and subsequently washed with 2% BSA as described above. The other aliquot was not bleached. Membranes were resuspended in either 2.5 ml of buffer PI or 2.5 ml of eye extract, and a retinal solution containing a 2:1 molar ratio of all-trans-to 11-cis-retinal) was added to a final 5 M concentration. Samples were mixed at 4°C overnight, washed two times with 2% BSA in phosphate buffer, and finally solubilized in 0.7 ml of 2% digitonin in phosphate buffer for spectral analysis.

Western Blots and Immunofluorescence
Protein samples were digested with N-glycanase as described by Kaushal et al. (31). Solubilized protein samples were run on 10% SDS-PAGE gels and transferred to nitrocellulose membranes. LimRh1-1D4 was detected using 1D4 IgG and anti-mouse horseradish peroxidase secondary antibody. Horseradish peroxidase was visualized using DAB/ NiCl 2 . Blots were scanned, and figures were prepared using Photoshop (Adobe). COS1 cells were grown on gelatin-coated cover slips and transfected as usual. After 48 h, cells were fixed with cold (Ϫ20°C) methanol followed by acetone. Opsin was detected using 1D4 antibody and a fluorescein-conjugated secondary antibody.
UV-visible absorption spectra of purified visual pigments and solubilized membranes were recorded in the indicated buffers at 20°C with a Beckman DU 640 single beam spectrophotometer. Bleaching illumination was provided with light from a 300-watt projector (Kodak) at a distance of 50 -60 cm. Unfiltered light from the projector was termed white light, and a high pass colored glass cut-off filter (Ͼ515 nm; Edmund Scientific, Inc.) was used as indicated. Total radiant energy of the light stimuli was measured using an optical power meter (3M Photodyne) equipped with a calibrated silicon photodiode (EGG, Inc.). Unfiltered light delivered 4.0 milliwatts. Spectra (4 -6 scans) were obtained from the sample before and after illumination. All spectra were averaged and analyzed using SigmaPlot Software (Jandel).

Analysis of Lim Rh1-1D4 Expression in Drosophila
Microscopy-The immunohistochemistry using the 1D4 antibody and confocal imaging were performed as previously described (24). The confocal image was collected using a Zeiss LSM-310 microscope (Thornwood, NY).
Electrophysiology-Electroretinograms (ERGs) were recorded from immobilized white-eyed flies using glass microelectrodes filled with normal saline (0.9% NaCl, w/v) as previously described (23,24). Spectral sensitivity was measured using a modification of the voltage clamp method of Franceschini (32, 33), which we have described in detail elsewhere (23). Briefly, the amplitude of the ERG response to a flickering (10 Hz) monochromatic stimulus was maintained at a criterion level while the wavelength of stimulating light was varied during a scan. Throughout the scan, the criterion response was maintained by constantly adjusting the light intensity using a proportional integralderivative algorithm (34). Spectral sensitivity was defined as the inverse of the light flux required to produce the criterion response, taking into account the wavelength and intensity of the stimulating light (i.e. spectral sensitivity ϰ 1/(light intensity ϫ wavelength)). Sensitivity data were normalized to an amplitude of 1.0 at the wavelength of maximal sensitivity, averaged, and smoothed with a window of 10 nm.
Rhodopsin Nomogram Modeling-Spectral sensitivity measured in vivo in the fly's eye was curve-fit to the theoretical rhodopsin absorption using the exponential function described by Stavenga et al. (35). Briefly, the spectral shape of the rhodopsin ␣-band absorption can be described by the following log normal function: , where x ϭ 10 log(/ max ), A ϭ 1, a 0 ϭ 380, a 1 ϭ 6.09, and a 2 ϭ 3a 1 2 /8. The curve-fitting routine was implemented in KaleidaGraph™ (version 3.08d, Synergy Software, Reading, PA) using the Levenberg-Marquardt (nonlinear least-squares) algorithm. The computer solved for the max and amplitude of the rhodopsin absorption spectrum and calculated the S.D. value for each variable and the correlation coefficient (Pearson's r). The effect of ␤-band absorption ( max ϭ 340 nm) was ignored, because this is likely to be minimal in the measured region (450 -600 nm).

Analysis of Lim Rh1-1D4 Expression in Oocytes
Immunoprecipitation-Oocytes were injected with ϳ50 ng of LimRh1-1D4 cRNA and cultured in labeling medium (MBS containing 10% heat-inactivated newborn calf serum and 10 Ci/l [ 35 S] TRANS label (ICN, Inc.)) for 4 days. The medium was replaced each day with fresh labeling medium. Oocytes were washed in modified Barth's saline and homogenized in 150 mM NaCl, 10 mM NaPO 4 , pH 7.0, and 0.5% dodecyl maltoside (10 l/oocyte). Insoluble material was removed by centrifugation at 14,000 rpm for 15 min at 4°C. Opsin was immunoprecipitated using 1D4-Sepharose and eluted from the resin with competing peptide. 35 S-Labeled proteins were resolved on 12% acrylamide gels and exposed to x-ray film.
Electrophysiology-The procedure for recording light responses has been described previously (29). Briefly, oocytes were selected 4 -7 days postinjection and analyzed using a two-electrode (2Ϫ5 megaohms) voltage clamp. Pigment generation was performed following impalement of the oocytes by switching the bath solution to one containing 5-10 M 11-cis-retinal. Light responses were elicited by exposing the cell to a stimulus delivered through an optical light guide. Responses were digitized and collected using pCLAMP.

RESULTS
Native Pigment-As a preliminary study for heterologous expression of a Limulus opsin cDNA, we characterized the photobleaching and regeneration properties of detergent extracts from lateral eye membranes (Fig. 1). In digitonin, there was a broad absorbance between 400 and 500 nm that arose from screening pigments, thus preventing direct observation of the visual pigment. However, exposure of the extract to light caused a loss of absorbance at ϳ540 nm and an increase at ϳ370 nm (Fig. 1B). The difference spectrum fit the profile expected for an 11-cis-retinal-based rhodopsin. The visual pigment was quite stable in detergent extracts when kept in the dark, with only a minor loss of absorbance after 18 h (Fig. 1B). The same results were obtained with membranes that were solubilized in dodecyl maltoside (data not shown). Rhodopsin could be isolated from dark-adapted and light-adapted animals as long as the retina was incubated in the dark prior to extraction (Fig. 1C). This shows that the M form of the visual pigment efficiently thermally converts back to R in the dark.
The chromophore could be completely removed from the visual pigment by exposing to light in the presence of hydroxylamine, resulting in no light-sensitive spectra. We attempted to regenerate the visual pigment by two methods. In the first, membranes were incubated with 11-cis-retinal in the dark before solubilization with digitonin. There were no light-sensitive changes detectable (Fig. 1C). In the second method, membranes were incubated with an extract of soluble proteins from Limulus retina and either 11-cis-or all-trans-retinal. Again, no detectable visual pigment was regenerated. The reason that chromophore did not bind apoprotein is not clear. One possibility is that the chromophore must bind to the opsin co-translationally, prior to conformational changes that form the final pigment. Alternatively, there might be accessory proteins not extracted or stable under our conditions (e.g. integral or peripheral membrane proteins) required for insertion of chromophore into the apoprotein.
Expression in COS1 Cells-COS1 cells were transfected with an epitope-tagged Limulus opsin expression plasmid. Western analysis ( Fig. 2A) of the purified protein showed that the expressed Limulus opsin was found in a number of different sized bands: a major broad band at 39 kDa, aggregated species of the size of dimers and trimers, and a proteolytic fragment at 21 kDa. Treatment of Limulus opsin with N-glycanase caused a mobility increase of monomer and higher aggregates, indicative of the removal of N-linked carbohydrate. Transfected COS1 cells were incubated with 11-cis-retinal, and the protein was solubilized in 1% dodecyl maltoside and then purified using immunoaffinity chromatography. The recovered protein did not have any associated retinal as measured by absorbance spectroscopy (Fig. 2B) or by dark-light difference. In order to determine whether the purification protocol caused the protein to release retinal, 11-cis-retinal was added directly to the purified protein sample. No light-sensitive pigment was formed. Transfected COS1 cell membranes were solubilized (conditions in which the native Limulus rhodopsin was stable) (Fig. 1) and analyzed by dark-light difference spectroscopy. Again, no detectable visual pigment was formed (Fig. 2D), in contrast to the situation with bovine rhodopsin (Fig. 2C). In order to determine whether 11-cis-retinal was required co-translationally, COS1 cells were transfected and then grown in the dark with 11-cisretinal for 60 h. After isolation of membranes and solubilization, no detectable visual pigment was found in dark-light difference spectra. These results indicate that Limulus opsin is not correctly processed in COS1 cells. As such, we expect that it would not be transported to the plasma membrane. Immunohistochemistry on transfected COS1 cells (Fig. 2E, right panel) confirmed that Limulus opsin was localized predominantly in the endoplasmic reticulum and Golgi, and very little was found in the plasma membrane. This finding is in contrast to the normal membrane localization of bovine opsin expressed in these cells (Fig. 2E, left panel). From these results, we conclude that mammalian COS1 cells lack the ability to properly process and target the Limulus opsin.
Expression in Sf9 Cells-Sf9 cells, derived from the invertebrate fall armyworm, have been used to express a number of proteins in functional form, including vertebrate rhodopsin. We prepared a baculovirus vector encoding the epitope-tagged Limulus opsin and infected Sf9 cells. After 2 days, a dose-dependent expression of opsin was detected by Western blot (Fig.  3A, lanes 1 and 2). The expressed Limulus opsin was found in a number of bands, whose pattern was very similar to that observed in transfected COS1 cells: monomer, aggregates, and a proteolytic fragment. In Sf9 cells, the monomer band was actually a doublet of glycosylated and unglycosylated forms of the protein. Sf9-expressed Limulus opsin was not stable in detergent extracts and underwent rapid proteolysis at 4°C. Inclusion of an extensive protease inhibitor mixture (see "Experimental Procedures") slowed degradation significantly (data not shown). Next, Sf9 cells were infected with viral supernatants (four individual cultures inoculated under identical conditions) and harvested at 1, 2, 3, and 4 days postinfection. A Western blot of the solubilized samples revealed that under these conditions only the cells harvested on day 2 gave a substantial fraction of Limulus opsin monomer (Fig. 3, A (lanes  3-6) and B). Samples from the other harvest times were com-pletely degraded. Sf9 cells were inoculated with viral supernatant encoding either bovine or Limulus opsin cDNA, either 11-cis-or all-trans-retinal was added to the culture medium, and the cultures were incubated in the dark for 2 days. No effect on the profile of expressed protein from infected Sf9 cells in Western blots was detected (Fig. 3, lanes 7-10). Cells grown in the presence of 11-cis-retinal were harvested, and membranes were prepared. Additional 11-cis-retinal was added, the membranes were solubilized with 1% dodecyl maltoside, and Limulus opsin was purified by immunoaffinity chromatography. Neither the absorbance spectra (Fig. 3C) nor dark-light difference spectra (Fig. 3D) had any detectable visual pigment. These results suggest, as in COS1 cells, that one or more components required for processing of Limulus opsin are missing in invertebrate Sf9 cells and that a component specifically found in invertebrate photoreceptors is required for correct formation of Limulus visual pigment.
Expression in the R1-R6 Photoreceptor Cells of Drosophila-To demonstrate that the Limulus lateral eye opsin cDNA encodes a functional visual pigment, we expressed the native and epitope-tagged forms of the Limulus opsin gene in the R1-R6 photoreceptor cells of Drosophila under the control of the ninaE opsin gene promoter. The R1-R6 photoreceptor cells are a suitable environment for the expression of novel opsins and have been used extensively for studying the spectral and physiological properties of novel or modified opsin genes in vivo, including those from other invertebrates such as the honeybee (23, 29, 36 -38).
Transgenic flies expressing the epitope tagged Limulus opsin construct [P{Rh1ϩLimMT}] show proper localization of the tagged protein to the rhabdomeres of the R1-R6 cells of the compound eye (Fig. 4D). This demonstrates that the transgene is expressed in the proper cell types and that the protein is stable and appropriately targeted within the photoreceptor cells. To assess the functional activity of the Limulus opsin, we examined the ERGs of transgenic flies. As mentioned previously, ninaE flies were used as a host strain because they lack the visual pigment normally expressed in the R1-R6 photoreceptor cells. To avoid any interfering signal from the R7 and R8 cells, which are unaffected by the ninaE mutation and produce a small depolarization in the ERG of ninaE flies, we made use  (2) at 0°C in the dark. The samples were exposed to light ( Ͼ515 nm), and the dark-light difference spectrum was calculated. Both samples had a positive peak max of 540 nm, whereas the negative peak had a minimum at 370 nm. Sample 2 contained 84% of the initial visual pigment (1). C, one set of retinas were collected from dark-adapted animals, membranes isolated and solubilized under dim red illumination (D). A second set of membranes were prepared from light-adapted animals in normal room light (L). The membranes were incubated in the dark at 4°C for 3 h. Both samples were exposed to light; the dark-light difference spectra exhibited positive peaks with max of 529 nm, and a negative peak had a minimum at 370 nm. The variation in max in B and C reflects the different photostationary state compositions formed in the two samples. Pigment recovery in the light-adapted samples was 82%. A second aliquot of light-adapted membranes was exposed to light in the presence of 0.1 M hydroxylamine to bleach the pigment of chromophore. After removing the retinal oximes, membranes were incubated with 11-cis-retinal (5 M) in the dark. Dark-light difference spectra were taken (B-R). No pigment regeneration was detectable. of a special host strain, norpA. In this strain, the phospholipase C required in all photoreceptor cells of the compound eye is removed genetically by the norpA mutation (39). The norpA cDNA is expressed under the control of the ninaE (Rh1) promoter so that only the R1-R6 photoreceptor cells are competent for transducing the light response (40). Fig. 4A shows that norpA mutant flies that express norpA in the R1-R6 cells and also carry a single wild type copy of ninaE (denoted ninaE/ϩ) have a robust response to light at both 350 and 520 nm (Fig.  4B, top traces). Removing the wild type copy of ninaE to generate the host strain (that lacks the opsin in the R1-R6 cells and lacks the PLC in the R7 and R8 cells) yields no response to either 350-or 520-nm pulses of light (Fig. 4A, second row of  traces). Introducing the epitope tagged or native forms of the Limulus lateral eye opsin cDNA into this background (Fig. 4A, third and fourth rows of traces, respectively) restored the electrophysiological response to light. This shows that both the native and epitope-tagged forms of the pigment are biologically active and functional when expressed in the Drosophila R1-6 photoreceptor cells, providing definitive proof that the Limulus cDNA encodes a functional visual pigment. The fact that these flies must be stimulated with very intense light in order to evoke a response (100-fold higher than those expressing Rh1) suggests that although the Limulus opsin is biologically active when expressed in fly photoreceptor cells, it may not be expressed as highly or couple as efficiently to the phototransduction cascade as the native Drosophila opsin. Interestingly, although there is a detectable response in the UV (350 nm) for both types of flies, the response at 520 nm is much greater. This indicated that the Limulus opsin does not effectively couple to the sensitizing pigment in the R1-R6 cells, which is responsible for their sensitivity to UV light (41,42).
To test whether the Limulus opsins utilize the same signal transduction pathway as the native Drosophila rhodopsins, we examined two genetic mutants that lack the phospholipase C (norpA, as described above) or lack the G-protein G␣ q that mediate the light response in Drosophila photoreceptors and are required from normal phototransduction (39,43). As shown in Fig. 4B, flies expressing the native Drosophila rhodopsin require G␣ q for phototransduction, because mutating this gene abolishes the light response (Fig. 4B, top trace). Similarly, expressing either of the Limulus opsin constructs in G␣ q mutant flies fails to produce a light response in the ERG (Fig. 4B,  second and third traces from the top). Drosophila norpA mu- 2 and 3) or Limulus (lanes 4 and 5) opsin cDNAs, and protein was purified by immunoaffinity chromatography and analyzed by Western blots using 1D4 antibody. Limulus opsin (lane 4) was expressed to levels similar to those obtained with bovine opsin (lane 2) but exhibited significantly more protein in aggregated form. In addition, there was a proteolytic product at ϳ21 kDa, comprising the Cterminal fragment. Treatment of the Limulus (lane 5) or bovine (lane 3) samples with N-glycanase caused the major bands to increase mobility, reflecting the removal of N-linked carbohydrate moieties. Bovine rhodopsin from retina (lane 1) is shown for reference. Major bands observed were monomer (m), dimer (d), trimer (t), and proteolytic degradation product (p). B, transfected COS cells were incubated with 11-cis-retinal, and protein was solubilized in 1% dodecyl maltoside and purified as before. The absorbance spectrum of Limulus opsin exhibited no chromophore bound in contrast to bovine opsin (inset), which had the typical max of 500 nm. C and D, membranes from COS cells transfected with bovine (C) or Limulus (D) opsin cDNA were isolated, incubated with 11-cis-retinal, and solubilized with 2% digitonin. Samples were illuminated with Ͼ515 nm light, and darklight difference spectra were calculated. No pigment formation was seen in the Limulus opsin sample. E, immunohistochemistry. Transfected COS cells were grown for 48 h, fixed, and stained with 1D4 antibody and fluorescent secondary antibody. In cells transfected with bovine opsin (left), cross-reactivity with antibody is observed predominantly in plasma membrane with some staining of organelles. In cells transfected with Limulus opsin (right), only intracellular staining is found, indicating that Limulus opsin is not correctly transported in COS cells.

FIG. 2. Expression of an epitopetagged Limulus opsin cDNA in COS1 cells. A, COS cells were transfected with bovine (lanes
tants, lacking the photoreceptor cell-specific PLC do not respond to light (Fig. 4B, fourth trace from the top). Likewise, expressing the Limulus opsin constructs in norpA mutant flies also fails to produce a detectable ERG (Fig. 4B, lower two  traces). These results indicate that the Limulus opsin constructs require both PLC and G␣ q in order to generate a response and suggest that the Limulus opsin couples to these effector molecules during phototransduction.
To determine the precise spectral sensitivity of the Limulus lateral eye rhodopsin, we measured spectral sensitivity physiologically in transgenic flies that expressed the native and epitope-tagged forms of the pigment. We carried out these experiments in the same norpA; ninaE host strain, as described above for the ERG recordings, that removes the effect of the R7 and R8 cells. As shown in Fig. 4C, the flies are sensitive to light in the blue-green region of the spectrum. The sensitivity curve is well fit by a rhodopsin absorption nomogram having a max at 513 nm. This measurement is in good agreement with physiological recordings from the Limulus lateral eye and with absorption spectra from rhodopsin purified from the lateral eye (44). We found no sensitivity in the UV region, consistent with the inability of the Limulus rhodopsin to couple to the R1-R6 photoreceptor sensitizing pigment (as discussed above).
Expression in Xenopus Oocytes-Transgenic experiments are not feasible in Limulus, so we utilized the Xenopus oocyte expression system, which has been used to express bovine rhodopsin as well as light-sensitive conductance from squid retinal mRNA. When Xenopus oocytes were injected with Limulus mRNA, light-sensitive chloride currents were generated after incubation with 11-cis-retinal (45). In these cells, the rhodopsin is able to couple to an endogenous, G-protein-coupled signal transduction pathway that regulates the membrane potential of the oocytes through a calcium-dependent chloride channel (46). In order to determine the kinetics of retinal binding to the apoprotein, we investigated the time course of regeneration of the light response. Oocytes were injected with mRNA from Limulus lateral eyes and incubated 4 -7 days in culture medium without retinal. Oocytes were voltage-clamped and stimulated with light flashes. No responses were seen in the absence of 11-cis-retinal or in the presence of all-transretinal. However, the light-sensitive currents appeared 5 min after the addition of 11-cis-retinal (Fig. 5). These results indicate that 11-cis-retinal can bind to Limulus opsin and that a portion of it is properly folded. Although longer incubation times gave slightly larger currents, most visual pigment formation is rapid. Retinal binding appears to occur posttranslationally.
Next, we examined whether oocytes alone contained the components necessary for correct biosynthesis of Limulus opsin. Oocytes from a single donor were injected with either  1 and 2). Cells were harvested 2 days later, solubilized in 1% dodecyl maltoside, and analyzed by Western blot. Three major bands were observed: monomer (m), dimer (d), and proteolytic degradation product (p). Time of infection was as follows. Sf9 cells were infected with viral supernatant and harvested at 1, 2, 3, and 4 days postinfection (lanes 3-6). Only day 2 cells contained a significant proportion of monomer. Effect of retinal was as follows. Sf9 cells were infected with viral supernatants containing baculovirus encoding bovine opsin (lanes 7 and 8) or Limulus opsin (lanes 9 and 10) and grown in the dark in medium containing 2 M all-trans-retinal (lanes 7 and 9) or 11-cis-retinal (lanes 8 and 10). There was no effect of retinal on expression levels or bands recovered. Bovine rhodopsin from retina (lane 11) is shown for reference. B, quantification of the distribution of various Limulus opsin forms from Sf9 cells at different times postinfection was performed using densitometry on lanes 3-6 in A. The m band contains two very closely spaced bands, one that is sensitive to N-glycanse (ϩCHO) and the other not (ϪCHO). The amount of protein in each portion of the m band was quantified separately. C and D, absorbance properties of rhodopsin purified from Sf9 cells infected with Limulus encoding baculovirus. The cells were grown in the presence of 11-cis-retinal, which was present in all solutions except the final one. The absorbance spectrum (C) was dominated by light scattering generated by aggregated protein, and no light-sensitive pigment was detectable in a difference spectrum (D).
cRNA encoding the epitope-tagged Limulus opsin or mRNA from Limulus lateral eye retina. The oocytes were tested for light-sensitive conductance changes following incubation with 11-cis-retinal (Fig. 6, A and B). None of the cRNA oocytes exhibited light-dependent currents, although they were syn-thesizing levels of polypeptide (Fig. 6C) comparable with that obtained with oocytes injected with cRNA encoding bovine opsin (29,46). On the other hand, 11 of 13 oocytes injected with retinal mRNA produced light-sensitive currents. These results support the idea that Limulus retinal mRNA encodes one or  (47). At night, an oscillator in the brain transmits efferent impulses to the photoreceptors reducing the rate of their generation of discrete "dark" electrical events, so-called quantum bumps. Both theoretical and experimental studies indicate that the circadian input reduces photoreceptor noise by reducing the small population of relatively unstable rhodopsin molecules in which Schiff base linkages between the chromophore and opsin are unprotonated (4,5,48). Deprotonation would destabilize rhodopsin by lowering the energy barrier for isomerization and thus allow activation of transduction mechanisms in the absence of light. We investigated expression systems for Limulus rhodopsin in order to test this molecular mechanism for photoreceptor noise.
The results presented here show that Limulus opsin can be expressed in COS1, Sf9 cells, Drosophila R1-6 photoreceptors, and Xenopus oocytes. In the heterologous cell culture systems, the polypeptide is made efficiently and is glycoslyated. However, a functional visual pigment was formed only in Drosophila photoreceptors and Xenopus oocytes that co-express additional retinal proteins. We performed experiments on purified preparations and directly on solubilized membranes with equivalent results. Experiments using native Limulus mem-branes showed that rhodopsin was stable in either digitonin or DM in the dark, indicating that the failure to detect bound retinal was not a result of instability under our conditions. In the dark, invertebrate visual pigments appear to have similar stability in DM as vertebrate pigments, as shown in our previous study that characterized purified Drosophila Rh1 (49). However, once Limulus rhodopsin was bleached, it was not possible to regenerate the pigment.
In both COS1 and Sf9 cells, there is substantial degradation of the protein, suggesting that at least a portion is misfolded. Immunofluorescence of the transfected COS cells showed the mislocalization of the protein during biosynthesis. In our expression constructs, the C terminus of Limulus opsin was altered to that of bovine opsin. This sequence has been shown to be important in the proper localization of vertebrate rhodopsin The oocyte was placed under two-electrode voltage clamp (Ϫ60 mV) and stimulated with light to verify that there were no light-sensitive changes in conductance. Then, 11-cis-retinal (5 M) was introduced into the bathing solution, and the oocyte was exposed to flashes of white light (arrows) at subsequent times following the addition of retina (times of 11-cis-retinal exposure are indicated in minutes to the right of each trace). The traces shown are excerpts from one continuous recording, with most of the interflash record omitted.

FIG. 6. Characterization of Xenopus oocytes injected with
Limulus opsin cRNA. A and B, oocytes from a single female donor were injected with either Limulus mRNA (A) or cRNA encoding epitopetagged Limulus opsin (B). Oocytes were cultured for 4 -7 days, and then those with resting potentials ϽϪ45 mV were tested for the presence of 11-cis-retinal-dependent, light-sensitive conductance changes. Recordings were obtained by alternating oocytes from each group. Light flashes were delivered at the arrow, and the scale for both sets is given at the bottom of A. Only 2 of the 13 retinal mRNA injected failed to respond to light, whereas none of the Limulus opsin cDNA-injected oocytes gave a response. C, two batches of oocytes (ϳ20 in each set) were injected with Limulus opsin cRNA (lanes 2 and 4), and another two batches were injected with water (lanes 3 and 5). The oocytes were cultured in medium containing 35 S-labeled amino acids for 4 days. Oocyte membranes were prepared and solubilized in 1% dodecyl maltoside, and then Limulus opsin was isolated by immunoaffinity chromatography. The fluorograph of the resulting proteins showed that oocytes produce one major and one minor band (arrows) when injected with Limulus opsin cRNA, whereas control oocytes only display nonspecific, background bands. in rods (50); however, an opsin mutant lacking the carboxyl terminus does bind retinal and exhibit light-dependent activation of the phototransduction cascade (51). Thus, it does not appear that this change in the Limulus opsin sequence would be expected to have a deleterious effect on pigment generation. We have tested this directly in transgenic Drosophila and found that constructs with either carboxyl terminus were equivalent in pigment formation and in light-induced ERG response (Fig. 4A). In our previous study of Drosophila Rh1 (49), we introduced a 1D4-epitope tagged Rh1 gene into transgenic Drosophila and found that this protein was comparable with the wild type protein in function and fully capable of generating a PDA. The use of the 1D4 epitope to purify other G-protein-coupled receptors in a functional state has also been described (e.g. see Refs. [52][53][54][55]. Thus, the modification of the carboxyl terminus of Limulus rhodopsin does not account for the inability of Limulus opsin to bind retinal in COS1 or Sf9 cells. Bovine rhodopsin mutants that have folding defects exhibit similar behavior in COS1 cells. Since both vertebrate COS1 and invertebrate Sf9 cells give similar results, there does not seem to be a general invertebrate-specific protein required for Limulus opsin biosynthesis. Rather, it appears most likely that there are one or more proteins specifically expressed in invertebrate photoreceptors that are required for forming visual pigment. Although no protein cofactor has been reported for rhodopsin, the C-terminal fragment of RAN BP2 (RBD4) was shown to act as a chaperone for human red opsin in COS cells (56,57) and increase the yield of recovered pigment.
Interestingly, RAN BP2 is a cyclophilin-related protein homologous to Drosophila ninaA, which is required for the proper processing and targeting of a subset of opsins. Drosophila opsin is transiently glycosylated and then deglycosylated as it moves through the endoplasmic reticulum and Golgi. This transport process requires the chaperone ninaA (58,59). It has been shown that different Drosophila opsins have different requirements for ninaA (60), suggesting that there are significant structural differences between different visual pigments within the organism. Our finding that transgenic flies expressing the Limulus opsin are 100-fold less sensitive to light than wild type flies also shows that within invertebrate photoreceptors the processing machinery is likely to differ significantly between different organisms and is specific for the natively expressed pigments.
Vertebrate opsins spontaneously bind 11-cis-retinal from solution and form a Schiff base linkage. This interaction forms part of the visual cycle. Invertebrate opsins also bind retinal, but it is not clear when this occurs or whether retinal is released from the protein at any time. The question remains as to how invertebrate rhodopsin is generated. Experiments to regenerate the native pigment in bleached (i.e. with hydroxylamine-treated) membranes were unsuccessful, even when soluble retinal proteins were added to the membranes. One possibility is that hydroxylamine denatures the protein and thus prevents retinal binding. A second possibility is that one or more proteins are necessary to permit retinal binding, a possibility supported by the experiments with mRNA injected oocytes. These proteins may be the same proteins involved in the proper folding of Limulus rhodopsin, or there may be additional proteins required for post-translational interaction with ligand. If there are specific proteins necessary for regeneration, their activity was apparently not preserved during the preparation of the membranes or soluble extract (or both). The issue concerning a requirement for the accessory protein(s) at the time of opsin translation in order for proper pigment formation remains. The rapid regeneration of mRNA-injected oo-cytes suggests that retinal binding can occur subsequent to biosynthesis.
Over the past few years, a large number of vertebrate opsins have been identified that are expressed in cells other than the rods and cones. Among these are pineal opsin (61,62), parapineal opsin (63), peropsin (64), melanopsin (65), VA opsin (66), encephalopsin (67), multitissue opsin (68), and retinal G-protein-coupled receptor opsin (69). In general, these proteins are either not expressed or are poorly expressed in functional form outside their native cell type. The deficiency may be similar to one observed here for Limulus opsin and could reflect a general property of these opsins, namely a requirement for specific chaperones and/or retinal shuttle proteins to form a functional pigment. Further work on Limulus folding and regeneration may help address these broader questions.