Solubilization of membrane-bound rod phosphodiesterase by the rod phosphodiesterase recombinant delta subunit.

Retinal rod and cone phosphodiesterases are oligomeric enzymes that consist of a dimeric catalytic core (alpha'2 in cones and alphabeta in rods) with inhibitory subunits (gamma) that regulate their activity. In addition, a 17-kDa protein referred to as the delta subunit co-purifies with the rod soluble phosphodiesterase and the cone phosphodiesterase. We report here partial protein sequencing of the rod delta subunit and isolation of a cDNA clone encoding it. The predicted amino acid sequence is unrelated to any other known protein. Of eight bovine tissue mRNA preparations examined by Northern analysis, the strongest delta subunit-specific signal was present in the retina. A less intense signal was seen in the brain and adrenal mRNA. In bovine retinal sections, rod delta subunit anti-peptide antibodies label rod but not cone outer segments. delta subunit, added back to washed outer segment membranes, solubilizes a large fraction of the membrane-bound phosphodiesterase, indicating that this subunit binds to the classical membrane associated phosphodiesterase. The subunit forms a tight complex with native, but not trypsin-released phosphodiesterase, suggesting that the isoprenylated carboxyl termini of the catalytic subunits may be involved in binding of the delta subunit to the phosphodiesterase holoenzyme.

For cyanogen bromide digestion, rod-soluble 17-kDa subunit protein was dried under vacuum. The sample was reconstituted in ammonium bicarbonate buffer and incubated in 77 mM triethylammonium acetate, 0.077% tributylphosphine and 1.5% 4-vinylpyridine for 2 h at a pH of 9.5 at 37°C under argon (14 -16). The sample was dried and reconstituted in 88% formic acid, diluted to 70% formic acid, and incubated with cyanogen bromide under argon at room temperature for 24 h. The proteolyzed sample was lyophilized twice and resuspended in 70% formic acid. After incubating for 2 h, the sample was diluted to 9% formic acid and injected onto a C-8 reverse phase column.
High Performance Liquid Chromatography Separation of ␣-Chymotryptic and Cyanogen Bromide Peptides-Chymotryptic fragments were separated by injecting the samples onto a 1.5 mm ϫ 100-mm octyl (C-8) reverse phase column (Brownlee) on an HP 1090 HPLC. The HPLC buffers were 0.08% trifluoroacetic acid (HPLC buffer A) and 0.08% trifluoroacetic acid, 80% acetonitrile (HPLC buffer B). The fragments were separated using a 0 -70% gradient of HPLC buffer B at a flow rate of 0.3 ml/min. Cyanogen bromide fragments were separated by the same procedure except the reverse-phase gradient was a 0 -80% gradient of HPLC buffer B.
Amino Acid Sequencing of the Rod-soluble 17-kDa Subunit-Edman degradation sequencing was performed on a Applied Biosystems model 470A gas-phase sequencer attached in line to a model 120A HPLC as described by Trong et al. (17).
Mass Spectral Analysis of the Expressed ␦ Subunit-Approximately 40 g of recombinant ␦ subunit protein was subjected to reverse-phase HPLC as described above. The subunit was concentrated to approximately 20% of the original volume in a Speedvac concentrator at room temperature, then used for mass spectrum analysis. The sample was injected into a matrix-assisted laser desorption ionization mass spectrometer, model Voyager Elite BioSpectrometry Research Station (Per-Septive Biosystems, Framingham, MA). This mass spectrometer employs a nitrogen laser at a wavelength of 337 nm. An ␣-cyano-4hydroxycinnamic acid matrix was used. The spectrum was acquired in linear mode with an acceleration voltage of 30,000, a grid voltage set at 90.0%, and a guide wire voltage set at 0.300%. A positive ion spectra was collected. The data were externally calibrated to an accuracy of 0.1%. The mass reported is an average of two sample injections measured as doubly charged (M 2ϩ ) and singly charged (M 1ϩ ) molecular ions. Each mass measurement is the result of 100 scans.
Nucleic Acid Sequencing-Most nucleic acid sequencing was performed as described elsewhere (21,23,24) with slight modifications. The procedure of Andersen et al. (21) was modified by annealing template DNA and sequencing oligonucleotide on ice for 5-10 min in place of the annealing step at 37°C. 2 One region of the full-length 17-kDa clones consistently resulted in band compressions using our standard sequencing procedure, therefore several other sequencing methods were employed (22)(23)(24)(25)(26) to sequence this "GC"-rich region.
RNA Isolation and Northern Analysis-Messenger RNA was isolated using Invitrogen's FastTrack mRNA isolation kit (version 3.1) and quantitated (20). Total RNA was isolated using the methods of Chomczynski and Sacchi (27). Messenger RNA was electrophoresed in a denaturing formaldehyde agarose gel (20). Nucleic acids were transferred to Hybond-N ϩ membrane overnight. Immediately after crosslinking, the membrane was stained (28), then incubated with random prime labeled cDNA probes (29,30). Blots were hybridized and then washed under increasingly stringent conditions (19). The most stringent wash conditions were at 55-60°C in 0.25 ϫ SSC, 0.1% SDS for at least 20 min.
To further select for a 17-kDa-specific DNA, a nested amplification reaction was performed on the RT-PCR products using the oligonucleotides, 17K-S5, 17K-AS8 (5Ј-CTCAGCATCCCGAAGGTTC-3Ј), and Thermus aquaticus (Taq) DNA polymerase. Control reactions contained either no DNA template, 17K-10 DNA (identical to clone 17K-11) in place of RT-PCR product, or substituted the oligonucleotide T7-24 (5Ј-CGACGGCCAGTGAATTGTAATACG-3Ј) for the oligonucleotide 17K-S5. A 220-bp DNA product was isolated and subcloned into the vector, pCR II, using standard techniques. Antibody Development and Western Analysis-The following peptides: EKFRLEQKVYFKGQV (17K-I) and RDAETGKILWQGTE-DLSVP (17K-II), were synthesized for producing 17-kDa subunitspecific anti-peptide antibodies. The sequence, RSQILMWSAN-KVFEELTDVE, was synthesized for producing a type 6 phosphodiesterase catalytic subunit-specific anti-peptide antibody (18,32). The peptides were conjugated to keyhole limpet hemocyanin in the presence of 20 mM glutaraldehyde in 100 mM sodium phosphate buffer, pH 7.5. After addition of diluted RIBI adjuvant, approximately 1.6 ml of the final conjugated peptide sample was injected into each rabbit. The rabbits were boosted 26 days after the first injection with freshly conjugated peptide. Serums were tested using an enzyme-linked immunosorbent assay plate (33). For Western analysis, proteins were electrophoresed in SDS-PAGE gels, transferred to nitrocellulose, and probed with antibodies (33,34).
Immunocytochemistry-Polyclonal anti-peptide antibody 17K-I was affinity-purified against the ␦ subunit from 60 g of rod-soluble phosphodiesterase bound to nitrocellulose (33). The ␦-bound antibody was eluted from the nitrocellulose with 500 l of 100 mM glycine, pH 2.5, and neutralized by addition of 50 l of 1 M Tris-Cl, pH 8.0. Monoclonal antibody ROS1 was purified over a protein A column under high salt conditions at pH 9.0 (33). The concentration of the purified ROS1 antibody was approximately 6 mg/ml. Bovine eyes were obtained at a local slaughterhouse and stored on ice until dissection (approximately 1-2 h  (20,35). After rinsing in PBS, pH 7.4, the eye cups were floated in 30% sucrose, 130 mM sodium phosphate, pH 7.3, for 2-18 h, then embedded in O.C.T. compound on a dry ice block. Twenty-micrometer sections were cut on a Tissue-Tek II Cryostat set at Ϫ18 to Ϫ20°C (20).
Sections were incubated in blocking solution (1 mg/ml bovine serum albumin, 0.05% (v/v) Triton X-100, 2% (v/v) goat serum, 2% (v/v) bovine serum diluted into PBS, pH 7.4) for 1 h at room temperature. Protein A-purified monoclonal antibody ROS1 and affinity-purified 17K-I were diluted 1:500 and 1:10, respectively, in antibody diluent (1 mg/ml bovine serum albumin, 0.05% (v/v) Triton X-100, 1% (v/v) goat serum, 1% (v/v) bovine serum in PBS, pH 7.4) and incubated for longer than 1 h at room temperature. After three 5-min washes in rinse solution (PBS containing 0.05% (v/v) Tween 20, pH 7.3), sections were incubated in a 1:50 dilution of the appropriate secondary antibody (goat anti-rabbit antibody conjugated to fluorescein isothiocyanate or goat anti-mouse antibody conjugated to tetramethylrhodamine isothiocyanate). After rinsing three times for 5 min each, sections were mounted in Vecta-shield®. Signals were visualized and recorded on a Leitz Dialux 20 microscope equipped with a 100-watt high pressure mercury lamp and a Leitz vario-orthomat camera. Alternatively, data were collected on a Bio-Rad MRC-600 confocal laser scanning microscope utilizing a krypton/argon laser with emissions at 488 and 568 nm.
Sf9 Cell Culture and Viral Infection-Sf9 cells were grown on plates or in spinner flasks in TNM-FH medium supplemented with 100 units/ml penicillin G, 100 g/ml of streptomycin sulfate and 10% fetal calf serum (20,36,37). To infect cells for protein production, spinner flasks were infected at cell densities of 1-3 ϫ 10 6 cells/ml; plates were infected at cell densities that were 60 -80% confluent (36,37). Titers were conducted using the end point dilution procedure of O'Reilly et al. (37) and Reed and Muench (38). Most infections were done at a multiplicity of infection of 10 or greater.
Baculovirus Expression of the Rod 17-kDa Subunit cDNA-The EcoRV-BamHI fragment of the 17-kDa cDNA clone, 17K-11, was subcloned into the SmaI and BglII sites of the baculovirus vector, pVL1393 (19,20). Five g of the resulting transfer plasmid, pVL17-17, and 5 l of Bsu36I-digested BacPAK6 DNA (Clontech) were combined with 500 l of Sf9 cells at a cell density of 2 ϫ 10 6 cells/ml (39). After electroporation, 1.5 ml of complete TNM-FH medium was added, and cells were plated in a 35-mm dish. Following incubation for 4 days at 27°C, extracellular virus particles were collected and immediately subjected to agarose overlays (36). Viral plaques were screened (40) and the identity of the expressed protein was confirmed by Western analysis. 2 J. K. Bentley and C. H. Sherbert, unpublished observations. Viral stocks were amplified according to the procedures of Summers and Smith (36) and O'Reilly et al. (37).
Purification of the Expressed ␦ Subunit-The ␦ subunit was overexpressed in Sf9 cells (36,37). After 3 days, the cells were collected by centrifugation at 200 ϫ g and 22°C for 5 min. The cell pellet was washed with PBS, pH 7.4, containing 14 mM benzamidine, 0.5 g/ml leupeptin, 0.7 g/ml pepstatin A, and 1 mM dithiothreitol. After repelleting, the cells were homogenized in low salt buffer (20 mM Tris-Cl, 1 mM Na 2 EDTA, pH 7.4) supplemented with the same protease inhibitors and dithiothreitol. The homogenate was centrifuged at 16,000 ϫ g for 30 min at 4°C to separate the supernatant fraction from the pellet fraction. The supernatant was added to DEAE anion exchange resin (DE52) preequilibrated with low salt buffer, pH 7.4, at a concentration of 3-10 mg of protein/ml of resin. After incubating for 1-3 h at 4°C with constant rotation, the resin was loaded into a 1.5 cm ϫ 13.5-cm column (bed height was approximately 6 cm). The column was washed with 2-3 volumes of low salt buffer. A linear gradient was run from 20 mM Tris-Cl, 1 mM Na 2 EDTA, pH 7.4, to 300 mM NaCl, 20 mM Tris-Cl, 1 mM Na 2 EDTA, pH 7.4, at a flow rate of approximately 14 ml/h at 4°C. The total gradient volume was approximately six times the bed volume. Approximately 400-l fractions were collected and assayed for ␦ subunit using Coomassie-stained SDS-PAGE and Western analysis. The fractions containing ␦ subunit were pooled. For a larger scale procedure, the method was essentially the same except that a 2.5 cm ϫ 22-cm column was used, and approximately 1.1-ml fractions were collected.
The pooled DEAE fractions were sequentially filtered through Amicon Centricon 100 and 50 concentrators prerinsed with ROS buffer (60 mM KCl, 30 mM NaCl, 20 mM MOPS, 2 mM MgCl 2 , pH 7.2), as follows. The Centricon 100 concentrators were centrifuged for 30 -60 min at 1000 ϫ g. The filtrates containing approximately 99% of the partially purified ␦ subunit were collected and loaded onto Amicon Centricon 50 concentrators. The Centricon 50 concentrators were centrifuged for 30 min at 5000 ϫ g. The filtrates were collected and concentrated on Amicon Centricon 10 concentrators for 60 -120 min at 5000 ϫ g. To increase recovery of the 17-kDa subunit protein, the Centricon 50 retentates were diluted with 2 ml of ROS buffer, pH 7.2, and centrifuged again. Purified expressed ␦ subunit was stored at 4°C. For larger scale purification, Amicon Centriprep concentrators were used.
Rod Outer Segment Membrane Isolation-For assay of ␦ subunit activity, purified rod outer segments were isolated from 50 frozen dark-adapted bovine retinas as described by Papermaster and Dreyer (41) with modification by Uhl et al. (42). This procedure releases nearly all of the endogenous ␦-bound rod and cone phosphodiesterase, but not the rod membrane-bound phosphodiesterase. The rod outer segment membranes were resuspended in ROS buffer, aliquoted, frozen in liquid nitrogen or on dry ice, and stored at Ϫ70°C. Protein concentration was determined either with the Bradford (43) assay or by measuring its absorbance at 280 nm. The concentration of the visual pigment, rhodopsin, was determined spectrophotometrically (44). To calculate phosphodiesterase concentration in rod outer segments, we assumed that 67% of the total protein was rhodopsin and that phosphodiesterase was present at Ϸ1/100 the rhodopsin concentration (45). The value obtained using this calculation is within 2-fold the value obtained by back calculating from phosphodiesterase activity assuming a cGMP hydrolytic rate of 2000 mol of cGMP hydrolyzed/min/mg of protein (data not shown).
␦ Subunit Activity Assay-Expressed ␦ subunit was combined with rod outer segment membranes diluted in ROS buffer, pH 7.2, and incubated for the times and temperatures indicated. The samples were centrifuged at 4°C for 30 min at 16,000 ϫ g to separate rod outer segment membranes and supernatants. The membranes were resuspended in ROS buffer, pH 7.2, or hypotonic buffer (5 mM Tris-Cl, pH 7.5) as indicated. Each fraction was assayed for phosphodiesterase activity using either the phosphate release assay (7,8) or the [ 3 H]cGMP assay (46). Phosphodiesterase was activated by preincubating samples with 0.02 mg/ml TPCK-trypsin for 8 -10 min on ice followed by addition of 0.125 mg/ml soybean trypsin inhibitor on ice.
Analysis of Dose-response Curves-The dose-response curves were analyzed using the program, SigmaPlot. They were fit with a rectangular hyperbole.
Gel Filtration Chromatography-Purified recombinant ␦ subunit protein was incubated with rod outer segment membranes as described above or with hypotonically extracted rod outer segment protein (47), or trypsin-released rod outer segment protein (5). Fifty M Pefabloc® was included in all experiments except those requiring trypsin treatment. Each sample was loaded onto a Sephadex G100 gel filtration column (1 cm ϫ 26 cm), then eluted with 200 mM NaCl buffer (200 mM NaCl, 20 mM Tris-Cl, 1 mM Na 2 EDTA, pH 7.4) at a flow rate of 8 -9 ml/h at 8°C.
Fractions were pooled and concentrated on Amicon Centricon 10 concentrators, and equal volumes were loaded onto an SDS-PAGE gel. The gels were stained with Coomassie (48) or silver (7), or were analyzed by Western analysis (33).
Protease Digestion of Membrane-bound Rod Outer Segment Phosphodiesterase-For trypsin-released rod outer segment protein, rod outer segment membranes were incubated in 0.02 mg/ml TPCK-trypsin for 1 min at 22°C, and the reaction was stopped with the addition of 0.125 mg/ml soybean trypsin inhibitor. Reactions were centrifuged for 30 min at 16,000 ϫ g and 4°C to separate bound proteins from soluble proteins. The trypsin conditions were optimized to allow for maximal release of PDE activity from the membrane without significant degradation of the inhibitory subunits (5).

Isolation of Peptide Fragments and Reverse-phase
HPLC-To obtain amino acid sequence for screening the retinal cDNA library, the rod-soluble ␦ subunit was purified by reverse-phase HPLC and cleaved with cyanogen bromide and ␣-chymotrypsin. After isolation on C-8 columns, the peptide fragments were analyzed using Edman degradation. The five peptide sequences identified for each of the protease treatments are listed in Table I, as well as the number of times that each sequence was observed. Most of the peptides were observed more than once, and two of the cyanogen bromide and chymotryptic peptide sequences overlapped by several amino acids.
Library Screening and Identification of Clones Encoding Rod-soluble 17-kDa Subunit Protein-Based on the sequences listed in Table I, redundant oligonucleotides were synthesized and used to screen the bovine retinal cDNA library of Li et al. (18). Two lifts were taken of each plate and probed with each of two redundant antisense oligonucleotides encoding the amino acid sequences, ESQMMPAH and AEQMEKF. Only viral plaques that were positive with both oligonucleotides were selected for further screening.
From the final screen, two groups of clones were identified. The first clone, 17K-14, contained approximately 90% of the open reading frame encoding the 17-kDa subunit and 35 bp of the 3Ј-noncoding sequence. However, this clone also contained a library cloning "artifact" or perhaps was derived from an incompletely processed RNA transcript. It contained an inframe stop codon 9 amino acids NH 2 -terminal to the ␦ subunit peptide sequence, LRDAE. There was no methionine between this in-frame stop codon and the ␦ subunit peptide sequence. The second group of clones contained two identical 1123-bp full-length clones (referred to as clone 17K-11 henceforth). A map of the sequencing strategies and the final cDNA sequence is shown in Fig. 1, A and B, respectively.
The NH 2 -terminal-most peptide sequence recognized from The numbers indicate the number of times each peptide was observed during Edman degradation for each protease treatment listed.
b Capital letters indicate a strong signal during Edman degradation. Small letters indicate a weaker signal during Edman degradation.
c Dash indicates no specific amino acid could be assigned to this cycle during Edman degradation.
the Edman degradation analysis is "NLRDAE" (see underlined amino acid sequence, Fig. 1B). This left us with two possible starting methionines, the methionine immediately preceding this sequence (predicted molecular mass, 15,087 Da) and the methionine 20 amino acids NH 2 -terminal to this sequence (predicted molecular mass, 17,390 Da). It seems likely that the starting amino acid sequence must be the latter at the "MSA . . ." sequence for several reasons. 1) Neither ATG start site contains an ideal Kozak sequence at positions Ϫ3 and ϩ4; however, the ATG site at nucleotide 166 is clearly the preferred codon of the two (49). 2) The sequence immediately upstream from the ATG sequence at nucleotide 166 is highly "GC"-en-riched, often a signature of regions immediately preceding start "ATGs" (49). 3) There are no in-frame stop codons between ATG 166 and ATG 223 .
We knew that the expressed subunit aligned with the native rod soluble subunit in SDS-PAGE chromatography and Western analysis ( Fig. 8 and data not shown). However, we could not determine whether its molecular mass was 15 or 17 kDa in SDS-PAGE chromatography. In addition, the size of this subunit was reported as both 15 kDa (8) and 17 kDa (6). To determine its molecular mass, we performed mass spectral analysis on the recombinant ␦ subunit. The analysis indicated that the mass of the recombinant protein was 17,395 Ϯ 4 Da (data not shown) indicating that ATG 166 is the preferred start site for translation of the expressed subunit.
Clone 17K-11 contains 165 bp of a 5Ј-noncoding sequence and 505 bp of a 3Ј-noncoding sequence. The open reading frame consists of 453 bp with a deduced amino acid sequence of 150 amino acids, predicting a molecular mass of 17,390 Da. In addition, a polyadenylation signal is present near the end of the 3Ј-noncoding sequence in this clone. The predicted amino acid sequence is not related to any other known protein (Gen-Bank TM release no. 92). Since clone 17K-14 and clone 17K-11 overlapped in only 88% of the open reading frame (see Fig. 1A), we did a RT-PCR experiment to confirm that the 5Ј end of clone 17K-11 did not contain any "artifacts," as reported by Li et al. (18) for other cDNAs isolated from this library and as identified in our clone 17K-14. Oligonucleotides flanking the region of interest were used to reverse transcribe then amplify bovine retinal mRNA unrelated to the bovine retinal cDNA library. Two clones were identified whose sequence matched that of clone 17K-11 in the NH 2 -terminal portion of the open reading frame. The region of overlap between the RT-PCR product, 17K-15, and clone 17K-11 is indicated in Fig. 1A.
A Partial Human Deduced Amino Acid Sequence Is Nearly Identical to the Bovine Rod Phosphodiesterase ␦ Subunit-Recently, two sequences derived from a human placental cDNA library were submitted to GenBank TM as part of the Merck-Washington University EST Project (accession nos. N41734 and R81870). The cDNA, accession no. R81870, has a 78% nucleotide sequence similarity to part of the 3Ј-noncoding region of clone 17K-11 (data not shown). The cDNA, accession no. N41734, has a 90% nucleotide sequence identity, and a 96% amino acid sequence identity to the bovine ␦ sequence amino acids Met 1 to Val 80 , and contains no in-frame stop codon ( Fig.  1B and data not shown).
Tissue Distribution of 17-kDa Subunit-To initiate studies on the tissue distribution of the 17-kDa subunit and to confirm that our cDNA represented a message enriched in retina, we isolated messenger RNA from eight bovine tissues and analyzed them by Northern analysis. The Northern blot was probed with a cDNA probe that included the entire open reading frame of clone 17K-11. The results of this Northern analysis are shown in Fig. 2. The predicted size of the mRNA is approximately 1300 bases.
As expected, the probe produced an intense band in the retinal mRNA lane. The adrenal and brain mRNA lanes also contained a detectable signal even after a high stringency wash (Fig. 2, A and B). The bands in the adrenal gland and brain are approximately the same size as that in the retina ( Fig. 2A). Because of this unexpected result, several different Northern blots were probed using two different preparations of mRNA to confirm the presence of a signal in the bovine brain and adrenal preparations (Fig. 2, A and B). Analysis of the radiolabeled signals on the Northern blot using a PhosphorImager indicate that the adrenal and brain messages are present at approximately 25-30% of retinal message ( Fig. 2A, bottom). We con-  Table I that are encoded by this clone. The italicized cDNA sequence above poly(A) indicates the location of a polyadenylation site near the 3Ј end of the clone. clude that the ␦ subunit message, while highly enriched in the bovine retina, may also be present to a lesser extent in nonretinal tissues. Relatively little is known about the localization of other PDE6 subunits in nonretinal tissues (see "Discussion").
Western Analysis-In addition to cloning a cDNA encoding this subunit, we used the sequence data to synthesize two polyclonal anti-peptide antibodies. Although both antibodies recognized proteins in immunocytochemistry experiments (see below) and immunoabsorbent assays (data not shown), only the anti-peptide antibody, 17K-II, identified a 17-kDa band in Western analysis (Fig. 3, middle panel). As can be seen, the antibody identifies specifically a 17-kDa band in purified cone phosphodiesterase and rod-soluble phosphodiesterase. It does not recognize the 11-kDa subunit, 13-kDa subunit, or the larger 84-, 88-, and 92-kDa subunits of the rod and cone phosphodiesterases (compare silver-stained proteins to Western immunoreactivity, Fig. 3, left and middle panels).
In addition to the 17-kDa subunit anti-peptide antibodies, three anti-peptide antibodies were produced against peptides encoding portions of bovine cone ␣Ј subunit. One of the antibodies, PDE6 Cat pAb, identifies the large catalytic subunit of the rod and cone phosphodiesterases (Fig. 3, right panel) specifically. We used this antibody to identify the phosphodiesterase subunits in the gel filtration experiments described below.
Immunocytochemical Localization of ␦ Subunit to Rod Outer Segments-Gillespie and Beavo (7) and Gillespie et al. (8) isolated cone and rod phosphodiesterase isozymes that contained a 17-kDa subunit. To confirm that the ␦ subunit was localized to cone and rod outer segments and to confirm that our peptide sequence coded for a photoreceptor cell-specific subunit, we performed immunocytochemistry experiments on bovine retinal sections using the anti-peptide antibodies. The anti-peptide antibody, 17K-I, appeared to identify only rod outer segments (Fig. 4A). The signal was absent when 17K-I antibody was preincubated with rod-soluble phosphodiesterase (Fig. 4B). Preimmune sera or the absence of primary antibody from the histochemical reactions yielded similar results to the competition experiment (Fig. 4, C and D). Labeling by the other antipeptide antibody, 17K-II, yielded similar results (data not shown).
To further examine these results, double-labeling experiments were performed using a mouse monoclonal antibody, ROS1, and affinity-purified 17K-I with visualization on a confocal microscope. The results of the 17K-I and ROS1 labeling are shown in Fig. 5, A and B, respectively. Hurwitz et al. (50) previously demonstrated that ROS1 labels both cone outer segments and rod outer segments in several species including bovine. Fig. 5A demonstrates the labeling of bovine retina sections with the 17K-I antibody in green. In Fig. 5B, the labeling of rod and cone outer segments by the ROS1 antibody is shown in red. In Fig. 5C, the images from Fig. 5, A and B, were merged. The yellow color indicates areas of signal overlap, demonstrating that 17K-I signal and ROS1 signals overlap in rod outer segments but not in cone inner or outer segments (Fig. 5C, arrows). The apparant discrepancy between the im- munocytochemistry results (Fig. 5) and the Western analysis (Fig. 3) probably results from an exchange of the rod ␦ subunit into the cone phosphodiesterase preparation during purification (see "Discussion").
Expression of ␦ Subunit in Baculovirus-The cDNA containing the entire open reading frame and most of the noncoding regions of the rod-soluble ␦ subunit was subcloned into a baculovirus expression vector, pVL1393, and a recombinant baculovirus, BCV11, was isolated. Fig. 6A demonstrates the presence of a 17-kDa protein expressed specifically in cells infected with the recombinant virus and identified by the anti-peptide antibody, 17K-II, in Western analysis.
Purification of Expressed ␦ Subunit from Baculovirus Extracts-Since the purification of the ␦ subunit using the methods of Gillespie and co-workers (7,8) results in an insoluble ␦ subunit, we devised a new method of purification for the expressed subunit protein. The expressed protein binds to anion exchange (DE52) resin under low salt conditions (20 mM Tris-Cl, 1 mM Na 2 EDTA, pH 7.4) and elutes early in a 0 to 300 mM sodium chloride gradient (see Fig. 6B, the heavy line below the graph and the inset) with the midpoint of the elution at approximately 60 mM NaCl. The expressed protein in the pooled fractions represents approximately 10% of the total protein present in the pool as determined by densitometry scans of Coomassie-stained gels (see Fig. 6C, first lane). In the pooled fractions, most of the Sf9 and viral proteins were much larger than the expressed 17-kDa protein. We therefore reasoned that we might be able to use a sizing step to further purify the protein.
The results of filtering the protein through a series of Amicon filters are shown in Fig. 6C. Approximately 99% of the ␦ subunit filtered through an Amicon 100 concentrator resulting in a 3-fold purification of the anion exchange pool and elimination of many high molecular mass contaminants (Fig. 6C, compare   lanes 100R, 100F, and 50R). Nearly all of the remaining contaminants were eliminated from the preparation by filtering the ␦ subunit through an Amicon 50 concentrator. Since a fraction of the ␦ subunit was retained by this filter, the retentate was diluted and refiltered to increase the yield of ␦ protein (Fig. 6C, compare lanes 50R and 50 -2R). The purified subunit was concentrated using an Amicon 10 concentrator (Fig. 6C,  lane 10R). Analysis of Coomassie-stained SDS-PAGE gels using densitometry scanning indicates that the recombinant ␦ subunit protein was approximately 95% of the total protein (Fig. 6C, lane 10R). In general, we can purify 1-2 mg of recombinant ␦ per liter of Sf9 culture (1.5 ϫ 10 9 cells).
Purified Recombinant ␦ Solubilize Membrane-bound Rod Phosphodiesterase Activity-Gillespie et al. (8) suggested that the 17-kDa subunit may solubilize rod membrane phosphodiesterase since it associated with the isotonically soluble isozymes, but they were unable to test this hypothesis because the buffers required to keep the 17-kDa preparation available in solution were highly denaturing. Therefore we developed an assay to determine if the recombinant subunit protein is capable of removing membrane-bound rod outer segment phosphodiesterase. Purified recombinant ␦ was incubated with rod outer segments under isotonic salt conditions. The membrane fraction and soluble fraction were isolated by centrifugation, and each fraction was analyzed for phosphodiesterase activity. Fig. 6D demonstrates the ability of the recombinant protein to release phosphodiesterase activity into the supernatant fraction, whereas isotonic buffer alone released little membranebound phosphodiesterase activity. Fig. 7A demonstrates a dose-dependent solubilization of phosphodiesterase activity by increasing concentrations of purified expressed ␦ subunit protein. The apparent EC 50 for ␦ activity at 5 nM phosphodiesterase was 14 Ϯ 2 nM (assuming that active ␦ is a 17-kDa monomer and that 100% of the expressed ␦ subunit is active). Since equilibrium could not be reached at concentrations of phosphodiesterase lower than 5 nM (data not shown), the dose-response curve might represent a titration event and not a measure of affinity.

Solubilization of the Rod Membrane Phosphodiesterase Depends on the Concentration of Purified ␦ Subunit-
The Expressed ␦ Subunit Does Not Alter the Activity of ROS Phosphodiesterase-Gillespie et al. (8) determined that the purified rod-soluble and rod membrane-bound phosphodiesterase had similar activation and kinetic properties. However, Gillespie et al. (8) worked with purified enzymes and did not have the ability to test for activity in the presence of added ␦ subunit protein. It seemed possible that the ␦ subunit protein, in addition to solubilizing membrane bound phosphodiesterase activity, could directly alter its activity. Therefore we assayed rod membrane phosphodiesterase activity in the presence and absence of expressed ␦ subunit (Fig. 7B). The ROS membranes were preincubated with ␦ subunit prior to dilution and trypsin activation of the phosphodiesterase. Fig. 7B demonstrates that addition of purified expressed ␦ subunit protein does not appear to alter the K m or V max activities of trypsin-activated ROS phosphodiesterase activity in the presence of membranes.
Gel Filtration Chromatography Indicates That the ␦ Subunit Protein Binds to PDE Catalytic Subunits-To further demonstrate that the ␦ subunit was interacting with the membrane associated rod phosphodiesterase, we used gel filtration in combination with Western analysis and phosphodiesterase activity assays to confirm that the ␦ subunit protein interacts with phosphodiesterase. Purified ␦ subunit was incubated with rod outer segments for 30 min at 30°C, and the supernatant and pellet fractions were separated by centrifugation. The supernatant fraction was then fractionated on a G100 gel filtration column. Purified recombinant ␦ eluted near the included volume of the column (Fig. 8A). When recombinant ␦ subunit was first preincubated with rod outer segment membranes, a fraction of recombinant ␦ co-migrated in the void volume with phosphodiesterase activity and catalytic subunit immunoreactivity (Figs. 8B and 9A). In addition, the PDE6-specific ROS1 antibody co-immunoprecipitated recombinant ␦ and catalytic subunit immunoreactivity and phosphodiesterase activity (data not shown). The ␦-solubilization activity assay, the size exclusion chromatography, and the immunoprecipitation results strongly suggest that the ␦ subunit binds to the membrane-bound phosphodiesterase.
Expressed 17-kDa Subunit Protein Does Not Bind to Trypsinreleased Phosphodiesterase-To determine if the 17-kDa subunit protein could bind to soluble phosphodiesterase, the purified ␦ subunit protein was incubated with hypotonically extracted rod outer segment protein in the presence of a serine protease inhibitor prior to size exclusion chromatography. As shown in Fig. 9B, the 17-kDa subunit protein migrated in the void volume when incubated with hypotonically extracted phosphodiesterase.
To determine if the phosphodiesterase catalytic subunit COOH termini containing the isoprenyl and carboxymethyl modifications might be involved in the solubilization of phosphodiesterase by the 17-kDa subunit protein, we repeated the size exclusion experiments in the presence of trypsin-released rod outer segment proteins. To determine optimum conditions for release of phosphodiesterase activity from the membrane, the trypsin release experiments described by Wensel and Stryer (5) were repeated. The purified expressed ␦ subunit protein was incubated with trypsin-released phosphodiesterase and applied to the G100 column. The results of this experiment, shown in Fig. 9C, demonstrate that the recombinant ␦ no longer migrates in the void volume. The phosphodiesterase activity profiles for each of these experiments is shown below the Western analysis indicating the location of the phosphodiesterase activity in the column profile.
In Fig. 9D, an equal amount of phosphodiesterase activity present in the void volumes of the experiments shown in Fig. 9, A-C, were analyzed with the 17K-II anti-peptide antibody and catalytic subunit-specific anti-peptide antibody in Western analysis. This experiment demonstrates the inability of the ␦ subunit to co-migrate with the trypsin-released phosphodiesterase, suggesting that the catalytic subunits' COOH termini may contain a ␦ subunit binding site.
FIG. 5. Double labeling bovine retinal sections with ROS1 (a rod ؉ cone cell antibody) and the ␦ anti-peptide antibody, 17K-I, demonstrates that 17K-I labels rod outer segments only. Immunocytochemical analysis was performed as described in Fig. 4 except that fluorescence staining was visualized using a confocal microscope. A, labeling with the ␦ anti-peptide antibody, 17K-I; B, labeling with the mouse monoclonal antibody, ROS1, in the same section as A; and C, visualization of both antibodies simultaneously. Green indicates areas of staining with the 17K-I antibody complex, red indicates areas of staining with ROS1 antibody complex, and yellow indicates overlapping regions of immunoreactivity. Arrows point to cone outer segments.

DISCUSSION
This is the first report demonstrating a function for the rod ␦ subunit of retinal phosphodiesterase. It appears to be one of only a few proteins thus far characterized that are capable of solubilizing a membrane-bound enzyme. It is, to our knowledge, the only protein described to date that solubilizes a protein modified by both a farnesyl and geranylgeranyl group. In addition, we demonstrate that it may interact directly with one or both isoprenylated COOH termini of the phosphodiesterase catalytic subunits. The localization of the ␦ subunit's message in other tissues suggests that it also may interact with phosphodiesterases or other proteins in nonretinal tissues.
To initiate studies examining the biochemical function of the ␦ subunit of the soluble rod phosphodiesterase, we obtained a full-length cDNA clone encoding the soluble rod phosphodiesterase ␦ subunit. Confirmation that the clone encodes the rod phosphodiesterase ␦ subunit is based on the following criteria. 1) The deduced amino acid sequence contains 4 of the 5 peptides isolated and sequenced from the soluble rod phosphodiesterase ␦ subunit and codes for a 17-kDa protein. 2) A probe made from the ␦ subunit cDNA clone binds to bovine retina mRNA in Northern analysis. 3) The expressed ␦ subunit comigrates with native ␦ subunit in SDS-PAGE analysis (Fig. 8). 4) A ␦ subunit-specific anti-peptide antibody identifies a 17-FIG. 6. Recombinant ␦ subunit Western analysis, purification using anion exchange chromatography and size filtration and its activity in ROS membranes. Recombinant virus was obtained as described under "Experimental Procedures." A, 50 g of supernatant or pellet fraction from infected and uninfected cells were electrophoresed in a 15% SDS-PAGE gel and analyzed using Western analysis. The rabbit anti-peptide polyclonal antibody, 17K-II, was diluted 1:2000. UI, uninfected Sf9 cell supernatant or pellet fractions; ␦I, supernatant or pellet fractions from cells infected with virus expressing the rod soluble ␦ subunit. B, recombinant ␦ subunit protein was overexpressed in a 100-ml culture of Sf9 cells as described under "Experimental Procedures." The cell extract was loaded, washed, and eluted from a DEAE column essentially as described under "Experimental Procedures." Fractions 50 -69 were combined into three pools. The location of ␦ subunit in the gradient was determined by Coomassie-stained SDS-PAGE analysis (as shown in the inset) and Western analysis (not shown). C, the ␦ subunit was purified using Amicon Centricons essentially as described under "Experimental Procedures." Each fraction was electrophoresed in a 15% SDS-PAGE gel and stained with Coomassie Brilliant Blue. DE-52 pool, 5.7 g of pooled DEAE fractions 50 -55 from the gradient shown in B; 100R, 10 g of Amicon Centricon 100 retentate; 100F, 1.6 g of Amicon Centricon 100 filtrate; 50R, 10 g of the first Amicon Centricon 50 retentate; 50F, 0.3 g of the first Amicon Centricon 50 filtrate; 50 -2R, 8.2 g of the second Amicon Centricon 50 retentate; 10R, 3 g of Amicon Centricon 10 retentate. D, ROS membranes were incubated with purified ␦ in ROS buffer or incubated with buffer alone for 30 min at 30°C. Membranes were pelleted, and the supernatant and pellet fractions were diluted in ROS buffer supplemented with 1 mg/ml bovine serum albumin such that 20 -30% hydrolysis of substrate was achieved in a 10-min [ 3 H]cGMP phosphodiesterase assay at a substrate concentration of 1 mM. Phosphodiesterase was activated by trypsin. ROS only, 1.5 mg/ml ROS membranes (Ϸ300 nM phosphodiesterase) incubated in ROS buffer only; ROS ϩ ␦, 1.5 mg/ml ROS membranes incubated with 2 M purified ␦ in ROS buffer. kDa protein expressed by our recombinant baculovirus. 5) A ␦ subunit-specific anti-peptide antibody labels bovine rod outer segments specifically.
The predicted molecular mass of the 17-kDa subunit protein is 17,390 Da, which corresponds closely to the observed mobility of the expressed protein in SDS-PAGE electrophoresis and is the same mass obtained in mass spectral analysis. Reversetranscription and polymerase chain reaction analysis demonstrated that the 5Ј end of the open reading frame of clone 17K-11 does not contain a cloning artifact. From the Northern FIG. 7. Dose-dependent solubilization of ROS phosphodiesterase activity by ␦ and trypsin-activated ROS membrane phosphodiesterase activity in the presence and absence of ␦ subunit. A, dose-dependent solubilization of phosphodiesterase activity in ROS membranes by the expressed ␦ subunit. Phosphodiesterase assays were conducted using the phosphate release assay at a concentration of 1 mM cGMP and an assay time of 20 min. Purified expressed ␦ protein was incubated with ROS membranes for 15 h at 4°C prior to separation of supernatant and membrane fractions. Phosphodiesterase was trypsin activated immediately prior to the phosphodiesterase assay. Purified ␦ subunit protein was corrected for percent purity (Ͼ88%) based on densitometric scans of Coomassie-stained SDS-PAGE gels. B, 1.6 mg/ml ROS membranes (Ϸ300 nM phosphodiesterase) were preincubated in ROS buffer or ROS buffer plus 2.4 M purified ␦ subunit at 30°C for 30 min. Samples were diluted immediately prior to trypsin activation and phosphodiesterase assay. Phosphodiesterase activity was measured in duplicate using the 3 H-assay as described under "Experimental Procedures." The experiment was repeated with essentially identical results.
FIG. 8. Expressed ␦ subunit shifts its mobility in G100 size exclusion chromatography when preincubated with ROS membranes. G100 size exclusion chromatography and Western analysis was performed as described under "Experimental Procedures." In A, 20 l of 32 M expressed purified ␦ subunit were diluted to 120 l with elution buffer and loaded without further treatment onto the G100 column. Numbers below the Western results represent pooled fractions eluted from the column. Each pool was concentrated and analyzed using Western analysis as described under "Experimental Procedures." The Western was incubated concurrently in the rabbit polyclonal anti-peptide antibodies, PDE6 Cat pAb and 17K-II, each diluted 1:1000. The goat anti-rabbit horseradish peroxidase-conjugated secondary antibody was diluted 1:3000. In B, 20 l of 32 M expressed purified ␦ subunit were incubated with 180 l of Ϸ8 mg/ml ROS for 30 min at 30°C. After centrifugation to separate the membrane from the supernatant fraction, the supernatant fraction was loaded onto the G100 column. Fractions were pooled, concentrated, and analyzed as in A. The void fraction in each experiment is indicated above the Western analysis result. PDE, 1.8 g of purified rod-soluble phosphodiesterase; 17K, 0.54 g of purified expressed rod-soluble ␦ subunit.
analysis, we conclude that clone 17K-11 contains approximately 88% of the full-length transcript and all of the coding sequence.
We found evidence in the Northern analysis to indicate that that human PDE6 ␤ subunit message is present in brain. Carcamo et al. (52) recently demonstrated the presence of a cone-type phosphodiesterase in mammalian pineal gland. The rat pineal extract contained a 15-kDa protein that may or may not be the same as our ␦ subunit. The identification of a highly conserved "␦" or "␦"-like cDNA sequence from a human placental library supports the conclusion that ␦'s message is present in nonretinal tissues. The distribution of this polypeptide's message in nonretinal tissues may indicate that it interacts with other phosphodiesterases or proteins. The unusually high degree of amino acid similarity between the human placental ␦ and rod bovine ␦-deduced amino acid sequences indicates that at least a portion of the cDNA is highly conserved within mammals and suggests a conserved function for the retinal PDE6 ␦ subunit.
The immunocytochemistry results shown here indicate that the ␦ subunit protein in the retina is localized only to the rod outer segments in bovine retinal sections, data apparently in conflict with previous results (7,8) and the data presented in Fig. 3. There are several possible explanations for these results. 1) It is possible that free or excess ␦ subunit exists in the rods and exchanges with cone phosphodiesterase during purification of cone phosphodiesterase. Hamilton and Hurley (4) indicated that the rod phosphodiesterase ␥ subunit may bind to the cone enzyme during purification. 2) The rod and cone ␦ subunits may share partial, but not complete, sequence homology; therefore, the polyclonal antibodies may identify different epitopes in Western analysis and immunocytochemical analysis. 3) It is also possible that, in the immunocytochemistry experiments, the ␦ subunit protein was not accessible to the antibody in cone cells, although capable of labeling rod outer segments. Whether the 17-kDa subunit protein is present in a complex with cone phosphodiesterase in situ remains uncertain.
To further confirm the identity of the cDNA sequence, to initiate studies examining the biochemical mechanism of ␦ subunit protein activity and to initiate physiological studies determining the role of the soluble phosphodiesterase in phototransduction, the cDNA encoding the ␦ subunit protein was expressed in a baculovirus expression system. The recombinant baculovirus expressed a protein of the expected molecular mass that is recognized by the anti-peptide antibody, 17K-II. The protein appears to be expressed at relatively high levels for a phosphodiesterase subunit (approximately 1-4% of total protein in cell extracts). 3 Purification of this recombinant subunit proved to be relatively simple due to the lack of low molecular mass proteins in baculovirus infected extracts. Purification was obtained by a 2-step procedure, anion exchange chromatography and size filtration. Although the fold purification varies somewhat depending on expression levels, with most preparations Ͼ85% purity is reached after approximately 50-fold purification. The fact that most of the expressed protein filtered through Amicon 100 and 50 concentrators indicates that the recombinant protein is not highly aggregated.
A functional assay was developed to test the hypothesis that the ␦ subunit protein is capable of solubilizing the membranebound rod phosphodiesterase under physiological (e.g. isotonic) salt concentrations. The results of the functional assay indicate that a majority of the phosphodiesterase can be released into the supernatant due to the presence of the 17-kDa subunit protein. However, in all the experiments conducted thus far we have yet to see more than 60% of the phosphodiesterase solubilized. It is unclear why 40% of the phosphodiesterase activity in rod outer segment cannot be solubilized by the ␦ subunit. It is possible that ␦ is capable of binding to the phosphodiesterase only when it is in a certain membrane compartment (i.e. disc membranes versus plasma membranes), when it is in a specific tertiary or quaternary structure, when the active dimer is composed of certain subunits (i.e. ␣/␣ versus ␣/␤ versus ␤/␤), when the phosphodiesterase is in association with other molecules, or when the phosphodiesterase is present in a specific lipid environment (e.g. positively charged lipid head groups). It is also possible that only a portion of the phosphodiesterase is accessible in our assay, that is, the disc membranes may be inside-out or otherwise not accessible or the membranes may not be completely permeable to small protein molecules.
The release of phosphodiesterase activity into the supernatant fraction is dependent on the concentration of ␦ subunit. The apparent EC 50 for the solubilization is approximately 14 nM (assuming a monomer molecular mass and 100% activity of the protein). Since we could not perform the assay under more dilute phosphodiesterase concentrations in our experimental system, we could not determine whether the EC 50 is a measure of titratable phosphodiesterase activity or whether it approximates the affinity of ␦ subunit for membrane-bound phosphodiesterase. The EC 50 value obtained may therefore represent a lower limit of the affinity of ␦ subunit for membrane-bound phosphodiesterase. A high affinity of ␦ for phosphodiesterase is indicated by the tight association of ␦ in the purified rodsoluble phosphodiesterase (8), the nM EC 50 obtained here, and the lack of dissociation of ␦ during size exclusion chromatography or immunoprecipitation. Consistent with the experiments of Gillespie et al. (8), who compared the catalytic activities of the rod soluble and rod membrane phosphodiesterases, recombinant ␦ subunit had no apparent effect on trypsin-activated ROS phosphodiesterase activity.
Since the results of the solubilization assay implied that the expressed ␦ subunit protein was binding directly to the phosphodiesterase catalytic subunits and solubilizing them, we chose to examine this event using size-exclusion chromatography. Rod outer segment proteins extracted under isotonic conditions by the ␦ subunit caused a shift in the ␦ subunit protein mobility, indicating that the ␦ subunit protein had combined with one or more proteins to form a larger complex. Western analysis as well as phosphodiesterase activity assays demonstrated the co-elution of the ␦ subunit protein and phosphodiesterase catalytic subunits in the void volume of the column. The observation that rod membrane phosphodiesterase activity is solubilized in the presence but not absence of expressed ␦ subunit, the shift in molecular mass of the ␦ subunit in the presence but not absence of rod outer segment membranes, and the immunoprecipitation of ␦-bound rod membrane phosphodiesterase strongly implies that the ␦ subunit is binding the phosphodiesterase directly to solubilize its activity.
After the isolation and purification of a rod-soluble phosphodiesterase, there was some question as to the relationship of the soluble rod and membrane-bound rod phosphodiesterase catalytic subunits. The results presented here imply that the critical difference between the membrane-bound rod and soluble rod phosphodiesterases is the presence of the ␦ subunit and that the catalytic subunits are not likely to be different gene products.
Several groups have recently reported the presence of isoprenyl groups on the carboxyl termini of the ␣ and ␤ subunits of the rod membrane phosphodiesterase (9,10). It is generally concluded that one or both of the isoprenyl groups serve to anchor the phosphodiesterase subunits on the membrane. We hypothesized that the ␦ protein could cause solubilization of the phosphodiesterase by binding to the isoprenylated catalytic subunits at their carboxyl termini. When phosphodiesterase activity was released from the membranes under conditions that proteolyzed the catalytic subunits' COOH termini but kept ␥ intact, the ␦ subunit protein no longer migrated in the void volume of the G100 size exclusion column. These experiments suggest that the ␦ subunit protein is interacting with the COOH-terminal region of either or both of the large catalytic subunits and that either the COOH-terminal sequence, the isoprenyl groups, or both are necessary for binding.
Only a few other proteins have been characterized that interact with lipid groups on post-translationally modified proteins. The protein, guanine-dissociation inhibitor, has been shown to solubilize geranylgeranylated membrane-bound proteins (53)(54)(55)(56). In addition, a family of proteins known as the fatty acid-binding proteins are capable of binding to fatty acid moieties. The chick retina fatty acid protein displays little expression in post-embryonic chick retinas (57)(58)(59). Neither of these proteins share any sequence identity to the rod soluble ␦ subunit cDNA reported here.
The function of the soluble phosphodiesterase in visual transduction is unclear. Solubilization of the enzyme by the ␦ subunit protein may serve to uncouple the phosphodiesterase from its activator molecule, transducin. The ␦ subunit solubilization activity may be responsive to one or several mechanisms of light adaptation in vertebrate photoreceptors. The availability of an active, purified expressed ␦ subunit protein should allow us to elucidate the physiological role of the soluble phosphodiesterase in the process of phototransduction.