Ubiquitinylation and ubiquitin-dependent proteolysis in vertebrate photoreceptors (rod outer segments). Evidence for ubiquitinylation of Gt and rhodopsin.

In corroboration of the hypothesized regulation of phototransduction proteins by the ubiquitin-dependent pathway, we identified free ubiquitin (8 kDa) and ubiquitin-protein conjugates (50 to >200 kDa; pI 5.3-6.8 by two-dimensional electrophoresis) in bovine rod outer segments (ROS). A 38-kDa ubiquitinylated protein and transducin (Gt) were eluted together from light-adapted ROS membranes with GTP. When ROS were dark-adapted, this 38-kDa ubiquitinylated species and Gt were readily solubilized in buffer lacking GTP. These data are consistent with ubiquitinylation of Gt and corroborate previous cell-free experiments identifying Gt as a substrate for ubiquitin-dependent proteolysis (Obin, M. S., Nowell, T., and Taylor, A. (1994) Biochem. Biophys. Res. Commun. 200, 1169-1176). Evidence for ubiquitinylation of rhodopsin (36 kDa), the (photo)receptor coupled to Gt, included (i) the presence in ROS membranes "stripped" of peripheral membrane proteins of numerous ubiquitin-protein conjugates, including two whose masses (44 and 50 kDa) are consistent with mono- and diubiquitinylated rhodopsin; (ii) catalysis by permeabilized ROS of 125I-labeled ubiquitin-protein conjugates whose masses (42, 50, and 58 kDa) suggest a "ladder" of mono-, di-, and triubiquitinylated rhodopsin; (iii) parallel mobility shifts on SDS-polyacrylamide gels of rhodopsin and these 125I-labeled ubiquitin-protein conjugates; and (iv) generation of enhanced levels of 125I-labeled ubiquitin-protein conjugates when stripped, detergent-solubilized ROS membranes (95% rhodopsin) were incubated with reticulocyte lysate. A functional ubiquitin-dependent pathway in ROS is demonstrated by the presence of (i) the ubiquitin-activating enzyme (E1); (ii) four ubiquitin carrier proteins (E214K, E220K, E225K, and E235K) and pronounced activity of E214K, an enzyme required for "N-end rule" proteolysis; (iii) ATP-dependent 26 S proteasome activity that rapidly degrades high mass 125I-labeled ubiquitin-ROS protein conjugates; and (iv) distinct ubiquitin C-terminal isopeptidase/hydrolase activities, including potent ubiquitin-aldehyde-insensitive activity directed at high mass ubiquitinylated moieties. Considered together, the data support a novel role for the ubiquitin-dependent pathway in the regulation of mammalian phototransduction protein levels and/or activities and provide the first identification of a non-calpain proteolytic system in photoreceptors.

A functional ubiquitin-dependent pathway in ROS is demonstrated by the presence of (i) the ubiquitin-activating enzyme (E1); (ii) four ubiquitin carrier proteins (E2 14K , E2 20K , E2 25K , and E2 35K ) and pronounced activity of E2 14K , an enzyme required for "N-end rule" proteolysis; (iii) ATP-dependent 26 S proteasome activity that rapidly degrades high mass 125 I-labeled ubiquitin-ROS protein conjugates; and (iv) distinct ubiquitin C-terminal isopeptidase/hydrolase activities, including potent ubiquitin-aldehyde-insensitive activity directed at high mass ubiquitinylated moieties. Considered together, the data support a novel role for the ubiquitin-dependent pathway in the regulation of mammalian phototrans-duction protein levels and/or activities and provide the first identification of a non-calpain proteolytic system in photoreceptors.
Phototransduction (the conversion of light into a nerve impulse) in the vertebrate eye occurs within the retina in highly specialized compartments, called outer segments, of rod and cone cells. The outer segments lack organelles. They are densely packed with photoreceptor membranes composed almost entirely of the phototransduction protein, rhodopsin, an important member of the superfamily of G protein-coupled receptors (1,2). In rod outer segments (ROS), 1 the rhodopsinrich membranes form individual photoreceptor discs, with up to 10 3 discs/ROS. Additional phototransduction proteins, including transducin, the heterotrimeric G protein (G t␣␤␥ ϭ G t ) coupled to rhodopsin, are associated with the discs or contained in the surrounding extraluminal cytosol (1)(2)(3)(4). Light initiates a conformational change in rhodopsin that promotes binding of G t to rhodopsin, thereby activating G t␣ to exchange GDP for GTP. Subsequent dissociation of G t␣ from rhodopsin and binding of G t␣ to its effector, a cGMP-specific phosphodiesterase, result in closure of cGMP-gated ion channels, rod cell plasma membrane hyperpolarization, and generation of the visual signal (1)(2)(3)(4). Restoration of rhodopsin to the dark state involves phosphorylation of light-activated rhodopsin by rhodopsin kinase and the subsequent binding of arrestin to phosphorylated rhodopsin. Both phosphorylation and arrestin binding inhibit the ability of rhodopsin to activate G t (1)(2)(3)(4).
Proteins targeted for degradation by the UPP are covalently ligated ("conjugated") to (multiple) ubiquitin (Ub), a 76-amino acid polypeptide. The carboxyl terminus (Gly 76 ) of Ub is linked by an isopeptide bond to a lysine residue of the target protein, and successive ubiquitins are added to each other, usually by formation of Gly 76 -Lys 48 isopeptide bonds. Ub-protein conjugation requires ATP, the Ub-activating enzyme (E1), Ub carrier proteins (E2 enzymes), and frequently, Ub-protein isopeptide ligases (E3 enzymes). The E2 and E3 enzymes are extremely selective. Substrates are frequently key cell regulators as mentioned above. Covalent modification (e.g. phosphorylation/dephosphorylation) or ligand-induced conformational change, possibly with exposure of inaccessible lysine residues, may account for the selective ubiquitinylation of such proteins (14,26). Ub-protein conjugates are degraded by the 26 S proteasome, a large ATP-dependent, multicatalytic protease complex (reviewed in Refs. 27 and 28), and monomeric Ub is regenerated for additional rounds of conjugation by the activity of Ub C-terminal isopeptidases/hydrolases (29). Conjugation of Ub to proteins can also mediate events independently of proteolysis, such as organelle biogenesis and activation-dependent receptor modification (15,26,30).
Using cell-free systems, our laboratory previously demonstrated that G t can be selectively ubiquitinylated and degraded by the UPP, thus suggesting potential interaction between the UPP and the visual transduction signaling pathway in mammals (19,20). The present study presents novel data documenting and characterizing UPP activity and rapidly degraded UPP substrates within bovine ROS. Most significantly, we present evidence for ubiquitinylation of both G t and rhodopsin by the ROS UPP.

EXPERIMENTAL PROCEDURES
Materials-Coomassie Plus protein assay reagent was purchased from Pierce. Materials for protein electrophoresis and nitrocellulose (0.2 m) were purchased from Bio-Rad. Immobilon P was obtained from Millipore Corp. (Bedford, MA), alkaline phosphatase-conjugated Protein A was from Organon Teknika-Cappell (West Chester, PA), 4-(2aminoethyl)benzenesulfonyl fluoride was from ICN Pharmaceuticals (Costa Mesa, CA), and Na 125 I and 125 I-labeled Protein A were from DuPont NEN. Materials for electron microscopy were obtained from Electron Microscopy Sciences (Fort Washington, PA). Rabbit antiserum recognizing the bovine proteasome activator PA28 (31)  . Rabbit reticulocytes were from Pel-Freez Biologicals (Rogers, AR). All other reagents were purchased from Sigma and were the highest grade available. Ubiquitin was iodinated by reaction with chloramine T as described (19).
ROS Purification-All procedures were performed at 0 -4°C either under fluorescent lights (light-adapted ROS) or under dim red light (dark-adapted ROS). Intact ROS were purified from fresh, lightadapted or dark-adapted bovine retinas on continuous sucrose gradients as described previously (20,37), except that the crude ROS fraction was obtained by floatation on 40% sucrose (38) rather than by filtration. Protein determination and assessment of ROS purity were as described (20). ROS purity was additionally corroborated by electron microscopy.
Briefly, ROS were fixed with 4% paraformaldehyde in 150 mM sodium phosphate (pH 7.3), 0.5 M NaCl, 100 mM sucrose and embedded in Epon 812. Inspection of at least four randomly chosen fields in each of four grids (magnification ϫ 2200 and 39,000; Phillips CM-10) revealed no rod inner segments or contaminating rod cell organelles. These observations were confirmed by two neuroanatomists, each of whom were unaware of the study objectives.
Preparation of ROS Supernatants and Membrane Preparations-All procedures were conducted at 0 -4°C. Washed, pelleted ROS were lysed in hypotonic buffer (5 mM Tris-Cl, 2.0 mM dithiothreitol, 0.5 mM MgCl 2 , 3 mM EGTA, 0.17 mM phenylmethylsulfonyl fluoride, 10 mM iodoacetamide (pH 7.6)) and passed once through a 22-gauge needle, and the supernatant (100,000 ϫ g, 20 min) was collected (39). This supernatant and two additional hypotonic washes were combined and concentrated in a Centricon-10 or Centricon-3 microconcentrator to Ϸ20 mg/ml. This is called the "hypotonic supernatant." The hypotonic supernatant concentrated in a Centricon-10 microconcentrator and prepared without iodoacetamide was used in UPP activity assays (see below). The pelleted membranes were washed in hypotonic buffer containing 50 M GTP (40), and this "GTP supernatant" was concentrated (Centricon-10) to 10 mg/ml. GTP-washed membranes were washed twice in Tris buffer (20 mM Tris-Cl, 2 mM dithiothreitol, 0.5 mM MgCl 2 , 3 mM EGTA, 0.17 mM phenylmethylsulfonyl fluoride (pH 7.6)) and then resuspended for 1 h in Tris buffer containing 5 M urea. At 0 and 30 min of incubation, the membranes were passed through a 22-gauge needle. These "stripped" membranes were washed twice more in Tris buffer and stored at Ϫ80°C. The "urea supernatant" was spin-dialyzed against Tris buffer and concentrated (Centricon-10). To obtain "stripped, detergent-solubilized membranes," stripped membranes were solubilized in 1.5% octyl glucoside (30 min), and insoluble material was removed by centrifugation (100,000 ϫ g, 30 min). When not used immediately, all preparations were aliquoted and stored at Ϫ80°C.
Preparation of Reticulocyte Lysate and Retinal Pigment Epithelial (RPE) Cell Supernatant-Reticulocyte lysate (Ϸ20 mg/ml non-hemoglobin protein) was prepared from rabbit reticulocytes by standard techniques (41), but without ATP depletion. The RPE supernatant was prepared from cultured cells as described previously (20). All preparations were stored at Ϫ80°C until use.
Preparation of Ubiquitin Antiserum-Antiserum to SDS-denatured Ub conjugated to ␥-globulin and preimmune serum were obtained from female New Zealand White rabbits as described (42). Solution-phase radioimmunoassay and Western blotting indicate that the antiserum recognizes both free (monomeric) and conjugated ubiquitins.
For immunoblotting, electrophoresed proteins were transferred to 0.2-m nitrocellulose or 0.45-m Immobilon with either a Bio-Rad Mini-Trans-Blot apparatus or a Pharmacia Biotech Novablot apparatus (semidry) as per the manufacturers' instructions. The Pharmacia protocol was optimized for the transfer and retention of lower molecular mass proteins (Ͻ30 kDa) and was used for the detection of free Ub. For immunological detection of free and conjugated ubiquitins and E2 enzymes, blots were incubated with appropriate antibodies or with pre-or nonimmune controls followed by 125 I-labeled Protein A. Rhodopsin was detected with an alkaline phosphatase-conjugated second antibody as described (20).
Conjugation Assays-Covalent ligation (conjugation) of exogenous 125 I-labeled Ub to endogenous ROS proteins was assessed in a conjugation assay modified from that of Hershko et al. (49). Assay conditions were as described above, except that the hypotonic supernatant was preincubated (15 min, 37°C) in the presence or absence of 1 l of the ATP-depleting reagent apyrase (10 units/ml final concentration) (50). Assays containing the ATP-depleted supernatant were incubated without ATP supplementation. Reactions were incubated for 30 min (37°C) and terminated by the addition of 2 ϫ Laemmli buffer followed by boiling for 3 min. 125 I-Labeled Ub-protein conjugates were detected as described above.
To assess substrates for ubiquitinylation in ROS, conjugation assays were conducted with whole ROS (350 g) that were permeabilized by freeze-thawing. In place of ATP, reactions contained 2.0 mM AMP-PNP (final concentration), a ␤,␥-nonhydrolyzable ATP analogue that supports Ub-protein conjugation, but does not support 26 S proteasomedependent degradation of ubiquitinylated proteins (51). Assays also contained the isopeptidase inhibitor Ubal (1 M final concentration) (52) and the proteasome inhibitor MG-132 (final concentration of 40 M in 0.4% dimethyl sulfoxide) (36). Reactions were preincubated (5 min), initiated by the addition of 125 I-labeled Ub, and incubated with occasional trituration for 30 min (37°C). Following the addition of 2 ϫ Laemmli buffer, nonboiled samples were analyzed by SDS-PAGE and autoradiography.
Covalent ligation of 125 I-labeled Ub to rhodopsin was assessed in ATP-depleted or ATP-supplemented rabbit reticulocyte lysate (150 g of non-hemoglobin protein). Assays also contained reaction buffer, 1 M Ubal, 40 M MG-132, 0.2 mM phenylmethylsulfonyl fluoride, 40 M E-64, and 10 mM EGTA. Dimethyl sulfoxide (0.4% final concentration) was added to all treatments not containing MG-132 to control for potential effects of the MG-132 vehicle. Reactions were preincubated (5 min, 37°C) before the addition of 10 g (1 l) of stripped, detergentsolubilized ROS membranes (95% rhodopsin) or vehicle (20 mM Tris-Cl containing 1.5% octyl glucoside). Thirty-minute reactions were terminated by boiling in Laemmli buffer, and formation of 125 I-labeled Ubrhodopsin conjugates was assessed by SDS-PAGE and autoradiography and by immunoblotting with mAb 4D2 (anti-rhodopsin) (34).
Degradation of 125 I-Labeled Ub-ROS Protein Conjugates by ROS 26 S Proteasome-Pulse-chase conjugation assays were employed to determine whether the ROS supernatant degrades 125 I-labeled Ub-ROS protein conjugates by a 26 S proteasome-dependent mechanism. The hypotonic supernatant, preincubated (5 min, 25°C) with 1 M Ubal (final concentration), was diluted (3:4) in 50 mM Tris-Cl (pH 8.0), and 13.5-l aliquots (Ϸ250 g of protein) were preincubated for an additional 5 min (37°C) with 10 l of ATP-supplemented reaction buffer, AMP-PNPsupplemented buffer, ATP-supplemented buffer containing 40 M MG-132, or ATP-supplemented buffer containing 40 M E-64 and 10 mM EGTA. Dimethyl sulfoxide (0.4% final concentration) was added to all treatments not containing MG-132 to control for potential effects of the MG-132 vehicle. Reactions were initiated by the addition of 1.5 l of 125 I-labeled Ub (6.6 g; 32 M final concentration) and incubated for 30 min (37°C). Radiolabeled Ub was then "chased" by the addition of 50-fold excess nonradiolabeled Ub (in 1.5 l of reaction buffer), and aliquots (4.5 l) of each assay obtained immediately after the addition of nonradiolabeled Ub and after various times of additional incubation were boiled in Laemmli buffer and subjected to SDS-PAGE. Autoradiograms of dried gels were scanned with a Molecular Dynamics Model PD laser densitometer using 50 scanning, and data so obtained were quantified with ImageQuant software. Percent loss of high mass radioactive species was calculated from the time-dependent decrease in pixel volume. All pixel volumes obtained were within the linear dynamic range of the densitometer. To correct for within-treatment variation in gel loading (Ͻ12%), pixel volumes of high mass bands were normalized to total cpm in the gel lane containing the most counts. Total cpm/ sample was obtained by ␥-counting of excised gel lanes. Statistically significant differences between treatments were determined by Student's t test.
Characterization of ROS Ub C-terminal Isopeptidase/Hydrolase Activities-Pulse-chase conjugation assays were also employed to assess the presence of Ub isopeptidase ("deconjugating") activity. Assays containing the hypotonic or RPE supernatant were preincubated (5 min, 37°C) in the presence or absence of 1 M Ubal (final concentration) and then incubated with AMP-PNP-supplemented reaction buffer and 1 g of 125 I-labeled Ub for 30 min before a chase with 50-fold excess nonradiolabeled Ub. Samples were collected after 30, 60, and 90 min of chase and analyzed as described above.

Characterization of Ub Pools and Cellular Localization of Ub-Protein Conjugates in Light-adapted ROS
The fragility of the connection between the outer and inner segments of rod cells, the ability of the ROS plasma membrane to reseal following separation from rod inner segments, and the stability of ROS in isotonic and hypertonic solutions make it possible to obtain highly pure preparations of intact ROS using continuous sucrose gradients (53). The electrophoretic profile of such preparations is dominated by monomeric and multimeric forms of rhodopsin, the G protein-coupled receptor constituting Ն85% of ROS proteins (Fig. 1A). However, soluble and peripheral membrane phototransduction proteins such as transducin (G t␣␤␥ ), arrestin, rhodopsin kinase, and cGMP-specific phosphodiesterase ␣and ␤-subunits are observed as well (20,53). To optimize the detection of both free and conjugated ubiquitins in ROS, different techniques were used to prepare the Western blots shown in Fig Ligation of each molecule of Ub increases the apparent molecular mass of a protein on SDS-polyacrylamide gels by 6 -8 kDa and shifts its pI toward that of Ub (pI 6.8). Consequently, electrophoretic resolution of Ub-protein conjugates by both charge and mass can provide insight into the identity of ubiquitinylated substrates. Whole ROS were therefore subjected to two-dimensional NEPHGE/SDS-PAGE followed by immunoblotting with Ub antiserum. Most ubiquitinylated ROS proteins were visualized as smears of immunoreactivity, indicating the presence of heterogeneous Ub-protein conjugates, with a range of masses between 50 and Ͼ200 kDa (Fig. 1C, left panel). The majority of these Ub-protein conjugates had pI values between 5.3 and 6.8, with increased pI consistently associated with increased mass. Three regions of Ub-protein conjugates were noted on two-dimensional immunoblots. These were (i) Ϸ50 kDa (pI 5.3) to 70 kDa (pI 6.5), (ii) 80 kDa (pI 5.8) to 120 kDa (pI 6.5), and (iii) Ͼ130 kDa (pI 6.2-6.8) (Fig. 1C, left panel). Stained gels revealed that almost all of the protein contained within these regions of mass and pI was rhodopsin (monomer, 36 kDa; oligomers, Ն65 kDa), based on an obtained pI of 5.6 -6.0 (Fig. 1C, right panel) (54) and on immunoblotting with mAb 4D2 (data not shown). These observations suggest that some of the endogenous Ub-protein conjugates detected in ROS are ubiquitinylated rhodopsin and that the three regions of Ub immunoreactivity observed on two-dimensional immunoblots (Fig. 1C, left panel) are due (at least in part) to ubiquitinylated forms of rhodopsin monomer (36 kDa), dimer (65 kDa), and higher mass oligomers (Ն130 kDa), respectively (Fig. 1C, right panel). Because only a minor fraction of a protein is usually ubiquitinylated at steady state, the vertical ladder of rhodopsin and a diagonal ladder of putative Ub-rhodopsin conjugates are expected.
We subsequently investigated the localization of Ub-protein conjugates within ROS and assessed whether G t and other membrane-associated phototransduction proteins were ubiquitinylated ( Fig. 2A) under conditions of light and dark adaptation. First, ROS were lysed and washed in hypotonic buffer to release cytosolic and loosely associated peripheral membrane proteins ( Fig. 2A, lane 1). This procedure, by depleting ROS of GDP and GTP, stabilizes the avid binding of G t to photolyzed rhodopsin in ROS membranes (Ref. 40 Western blot analysis indicated that most of the soluble Ub-protein conjugates in light-adapted ROS were recovered in the hypotonic supernatant ( Fig. 2A, compare lane 4 with lanes 5 and 6). A majority of Ub-protein conjugates had high (Ͼ100 kDa) masses, and there appeared to be considerable heterogeneity in the masses. This suggests that the intense and broad band of immunoreactivity between 100 and 160 kDa is composed of several ubiquitinylated proteins. Interestingly, soluble Ub-protein conjugates were almost undetectable when the ROS supernatant was prepared in the absence of iodoacetamide, an inhibitor of Ub C-terminal isopeptidase/hydrolase activity (data not shown). The GTP-soluble fraction contained a Ϸ38-kDa ubiquitinylated species that migrated between the ␣and ␤-subunits of G t (Fig. 2A, lane 5). This Ub moiety was also observed in the hypotonic supernatant ( Fig. 2A, lane 4), but it was significantly more plentiful in the G t -rich GTP eluate. The urea-soluble fraction, containing arrestin and several unidentified membrane-associated ROS proteins, lacked detectable ubiquitinylated proteins ( Fig. 2A, lanes 3 and 6).
The urea-insoluble fraction composed of stripped membranes contained at least 12 discrete bands of Ub immunoreactivity between 40 and Ͼ200 kDa, visible against a background of smeared immunoreactive material (Fig. 2B, lane 2). The relative amount of Ub-protein conjugates in this ROS membrane pool is significant because blots of ROS membranes contain an order of magnitude less ROS protein equivalents than are contained on blots of the ROS supernatant ( Fig. 2A, lane 1). Membrane Ub-protein conjugates with apparent masses of 50, 78, 90, and Ն200 kDa seen in Fig. 2B (lane 2, arrowheads) were also predominant species in whole ROS (Fig. 1B, lane 2). In addition, Ub immunoreactivity in stripped membranes was distributed in three regions of increasing molecular mass (i.e. Ͻ70, 80 -120, and Ͼ120 kDa), a pattern detected previously in whole ROS (Fig. 1C, left panel). Considered together, these data indicate that a significant proportion of ubiquitinylated ROS protein is distributed in the membrane pool. The fact that FIG. 1. ROS protein profile and detection of free and conjugated ubiquitins in ROS. A, ROS proteins separated on 12% SDSpolyacrylamide gels in the presence of 2-mercaptoethanol, but without sample boiling. Predominant phototransduction proteins were identified by immunoblotting (20) and by previously documented apparent molecular mass (53). Mobilities of molecular mass standards are indicated to the left. rho 1, rhodopsin monomer; rho 2, rhodopsin dimer; kinase, rhodopsin kinase; PDE ␣␤ , ␣and ␤-subunits of cGMP-specific phosphodiesterase. B, immunological detection of Ub and Ub-protein conjugates. Following SDS-PAGE, proteins in the ROS hypotonic supernatant Ͻ30 kDa (lane 1) and total ROS protein Ͼ30 kDa (lanes 2 and 3) were transferred to nitrocellulose (see "Experimental Procedures") and incubated with Ub antiserum (lanes 1 and 2) or with preimmune serum (lane 3), both containing 10 g/ml IgG. Blots were subsequently incubated with 125 I-labeled Protein A (2 ϫ 10 5 counts/ml) and visualized by autoradiography. Free Ub was detected at the dye front (df; lane 1) and migrated with an apparent mass of 7.5 kDa on higher percentage gels (not shown). Mobilities of molecular mass standards are indicated to the left. C, NEPHGE/SDS-PAGE analysis of Ub-protein conjugates in whole ROS. Two-dimensional immunoblots of 100 g (0.15 retina equivalent) of total ROS protein (Ͼ30 kDa) were incubated with Ub antiserum followed by 125 I-labeled Protein A (left panel). Ub-protein conjugates are in three mass and pI groups indicated by brackets. The arrows denote the pI of rhodopsin (5.6) as determined on Coomassie Bluestained two-dimensional gels (right panel). Mobilities of prestained molecular mass standards are indicated between the panels. these ROS membranes are stripped of peripheral proteins makes it likely that at least some of the Ub-protein conjugates detected in these preparations are intrinsic membrane proteins. Moreover, the overwhelming predominance of rhodopsin (36 kDa) in stripped membranes (Fig. 2B, lane 1) suggests that the 44-and 50-kDa conjugates (*) detected on immunoblots (Fig. 1B, lane 2,) constitute a "ladder" of mono-and diubiquitinylated rhodopsin monomer, respectively. Finally, with the exception of a 60-kDa band detected in all three supernatant fractions (but not in ROS membranes), no proteins were recognized by preimmune serum (data not shown).

Effect of Dark Adaptation on Membrane Binding of 38-kDa Ub-Protein Conjugate Eluted from ROS Membranes with GTP
The presence of a specific Ub-protein conjugate in the G protein-rich GTP eluate from light-adapted ROS (Fig. 2A, lane  5), together with the previous demonstration of G t ubiquitinylation in vitro (19), led us to speculate that the 38-kDa ubiquitinylated species contained either a G t subunit or a G t -associated protein. If so, we hypothesized that the 38-kDa Ubprotein conjugate would, similar to G t , bind loosely to membranes from dark-adapted ROS and could thus be recovered in hypotonic buffer lacking GTP (Ref. 40; reviewed in Ref. 3). Accordingly, dark-adapted ROS were washed sequentially in hypotonic and GTP buffers, and the resulting supernatants were electrophoresed and immunoblotted as described above. As expected, the hypotonic supernatant from dark-adapted ROS contained most of the soluble and peripheral membrane proteins, including G t (Fig. 2C, lane 1). The GTP supernatant contained small quantities of two unidentified proteins, including one with an apparent mass of 38 kDa (Fig. 2C, lane 2). However, Western blots revealed that the 38-kDa Ub-protein conjugate was present exclusively in the hypotonic supernatant from dark-adapted ROS (Fig. 2C, compare lanes 3 and 4), i.e. GTP was not required for solubilization when ROS were darkadapted, but it was required to elute the 38-kDa Ub-protein conjugate from light-adapted ROS. Thus, light and GTP modulate the association of the 38-kDa conjugate with the ROS membrane in a manner similar to that observed for G t . It is therefore likely that the 38-kDa Ub-protein conjugate contains a G t subunit (see "Discussion").

ROS-catalyzed Conjugation of 125 I-Labeled Ub to Soluble ROS Proteins
To confirm that endogenous ROS Ub-protein conjugates could be generated de novo within the outer segment, we assessed the ability of enzymes in ROS cytosol to catalyze the formation of conjugates of endogenous protein(s) and exogenous 125 I-labeled Ub. Reaction mixtures containing the hypotonic supernatant from light-adapted ROS were either supplemented with ATP or depleted of ATP with apyrase. Reactions were terminated after 30 min, by which time levels of 125 Ilabeled Ub-protein conjugates had attained steady state (data not shown). Autoradiography following SDS-PAGE demonstrated that the ROS supernatant catalyzed the ATP-dependent ligation of 125 I-labeled Ub to soluble ROS proteins, as indicated by the significant incorporation of radiolabel into were analyzed on stained 10% SDS-polyacrylamide gels (lanes 1 and 2) and on immunoblots probed with Ub antiserum and 125 I-labeled Protein A (lanes 3 and 4) as described for A. The area of the immunoblot between 37 and 38 kDa is enlarged to more clearly visualize the 38-kDa Ub-protein conjugate (lane 3, arrow) and the proximity of G t␤ (determined by Ponceau S staining). For all experiments, the conditions used for protein transfer preferentially retained Ͼ30-kDa proteins.  1-3) and on immunoblots probed with Ub antiserum (lanes 4 -6). Supernatants were obtained by sequential solubilization of ROS in hypotonic buffer (Hypo) (lanes 1 and 4; 1 retina equivalent), 50 M GTP (lanes 2 and 5; 1 retina equivalent), and 5 M urea (lanes 3 and 6; 2 retina equivalents). Mobilities of molecular mass standards are indicated to the right. PDE ␣␤ , ␣and ␤-subunits of cGMP-specific phosphodiesterase. B, protein profile and Ub-protein conjugates in the residual stripped photoreceptor membranes following sequential treatment of ROS with hypotonic buffer, GTP, and urea (as described for A). Stripped membranes (Ϸ0.1 retina equivalent) were electrophoresed without boiling (lane 1) and immunoblotted (lane 2) as in described for A. Arrowheads indicate conjugates whose apparent masses are identical to conjugates detected in whole ROS (see Fig. 1B,  lane 2); brackets denote the three regions of molecular mass in which discrete conjugates are distributed (see also Fig. 1C, left panel). *, presumed Ub-rhodopsin conjugates based on apparent molecular mass (see also Fig. 4B); rho 1, rhodopsin monomer; rho 2, rhodopsin dimer. C, protein profile and Ub-protein conjugates in dark-adapted ROS. Proteins from dark-adapted ROS were solubilized in hypotonic buffer (lanes 1 and 3) or in buffer containing 50 M GTP (lanes 2 and 4) and Ͼ8-kDa species in the ATP-supplemented assay (Fig. 3, compare lanes 1 and 2). Radiolabeled adducts generated in the presence of ATP included predominant species with masses of Ͼ200, 83, 70, 55, and Ϸ30 kDa. Significant quantities of radiolabel were incorporated into high mass species (Ն200 kDa) in ATP-supplemented reactions (Fig. 3, lane 1). The high mass of these Ub-protein conjugates reflects the ligation of multiple ubiquitins to target proteins, although we cannot rule out the possibility that some high mass radiolabeled species may be chains of 125 I-labeled Ub (see below). The faint radiolabeled species in the apyrase-treated reaction (Fig. 3, lane 2) are Ub-protein conjugates presumably generated in the presence of low levels of residual ATP. These results indicate that the ROS supernatant contains the enzymes (E1, E2 enzymes, and perhaps E3 enzymes) required to ligate Ub to protein substrates as well as soluble protein substrates of ROS conjugating activity (see also Ref. 22). Despite the evidence for Ub-protein conjugates in ROS membranes (see above), no conjugating activity was detected in the residual membrane fraction following hypotonic washes (data not shown).

Evidence for Conjugation of 125 I-Labeled Ub to Rhodopsin
Rhodopsin Electrophoretic Mobility Shift-The preponderance of Ub-protein conjugates in ROS with pI values between 5.8 and 6.8 (Fig. 1C) and the demonstration of significant levels of ubiquitinylated protein in stripped ROS membranes (Fig.  2B, lane 2) suggested that the intrinsic membrane protein rhodopsin is ubiquitinylated by the ROS UPP. To investigate this possibility, we conducted conjugation assays with whole permeabilized ROS. Analysis of these conjugation assays exploited a unique but characteristic oligomerization of rhodopsin that is induced by boiling even in the presence of SDS. This phenomenon is illustrated in Fig. 4A. Following incubation with 125 I-labeled Ub, samples were solubilized in Laemmli buffer, subjected to SDS-PAGE either without boiling (Fig. 4A,  lanes 1 and 3) or after boiling for 3 min (lanes 2 and 4), and immunoblotted for rhodopsin with mAb 4D2 (lanes 1 and 2). When samples were not boiled, rhodopsin migrated at 36, 65, 130, and Ͼ200 kDa (Fig. 4A, lane 1). (Incubation at 37°C during the conjugation assay induced some oligomerization, as evidenced by enhanced levels of rhodopsin dimer versus monomer.) Major 125 I-labeled Ub-protein conjugates detected in nonboiled samples had masses of 42, 50, 58, 78, 90, 160, and Ͼ200 kDa, and radiolabel was equally distributed between the 42-and Ͼ200-kDa species (Fig. 4A, lane 3). The masses of these 125 I-labeled Ub-protein conjugates are, with the exception of the 160-kDa species, comparable to the masses of endogenous conjugates detected in nonboiled, stripped ROS membranes (Fig. 2B, lane 2). Notably, because the covalent ligation of one Ub increases the apparent molecular mass of a substrate protein by 6 -8 kDa, the molecular masses of the major 42-, 50-, and 58-kDa species suggest a ladder of mono-, di-, and triubiquitinylated rhodopsin monomer, respectively. The 78-and 90-kDa conjugates and the 160-kDa radiolabeled species (Fig. 4A, lane 3) may be ubiquitinylated rhodopsin monomer with more ubiquitins attached or higher mass forms of rhodopsin (65 and 130 kDa, respectively) (Fig. 4, lane 1) to which fewer ubiquitins are attached. As predicted, when samples were solubilized and then boiled before electrophoresis, lower mass forms of rhodopsin (36 and 65 kDa) were not apparent on the immunoblots, and the proportion of very high mass rhodopsin oligomers (Ͼ200 kDa) increased (Fig. 4A, compare lanes 1 and 2). The induced mobility shift of rhodopsin was coincident with the loss of radiolabeled conjugates (Ͻ160 kDa) observed in nonboiled assays and by a corresponding increase in very high mass (Ͼ200 kDa) radiolabeled species (Fig. 4A, compare lanes 3 and  4). Greater than 90% of 125 I-labeled Ub was detected at the top of the gel when ROS were boiled in SDS-PAGE sample buffer. This coordinated migration of radiolabeled Ub and rhodopsin before and after the induction of rhodopsin mobility shifts is consistent with ubiquitinylation of rhodopsin in these assays.

Conjugation of 125 I-Labeled Ub to Stripped, Detergent-solubilized ROS Membranes by Reticulocyte
Lysate-To corroborate data (above) suggesting rhodopsin ubiquitinylation in vitro, stripped, detergent-solubilized ROS membranes were incubated with ATP-supplemented or ATP-depleted rabbit reticulocyte lysate, a cell-free system noted for a potent and well characterized UPP (Fig. 4B) (47)(48)(49). A salient feature of this assay is that the ROS membrane preparation is Ͼ95% rhodopsin; the predominant residual proteins, arrestin isoforms (data not shown), appear not to be UPP substrates ( Fig. 2A, lanes 3  and 6) (19,20). Reactions were initiated by the addition of 125 I-labeled Ub, incubated for 30 min, terminated by the addition of Laemmli buffer, and electrophoresed (Fig. 4B, lanes  1-4). SDS-PAGE samples were boiled before electrophoresis in order to concentrate 125 I-labeled Ub-rhodopsin conjugates of heterogeneous mass at the top of the gel, thereby enhancing chances for their autoradiographic detection amidst a background of radiolabeled reticulocyte Ub-protein conjugates. Autoradiograms demonstrated that 125 I-labeled Ub-protein adducts were generated only in the presence of ATP (Fig. 4B,  compare lanes 5 and 6 with lanes 7 and 8), indicating that such adducts are Ub-protein conjugates. Assays containing exogenous rhodopsin generated elevated levels of high mass (Ͼ200 kDa) conjugates relative to assays incubated without rhodopsin (Fig. 4B, compare lanes 5 and 6 and, in a less exposed autoradiogram, lanes 9 and 10). The presence of exogenous rhodopsin was also associated with the de novo formation of a novel 125 I-labeled Ub-protein conjugate with an apparent mass of 120 kDa (Fig. 4B, compare lanes 5 and 6). This 120-kDa conjugate can be rationalized as being derived from the lower mass rhodopsin species (70 kDa) detected on immunoblots probed with mAb 4D2 (Fig. 4B, lane 11). Thus, ROS-specific Ub-protein conjugates are generated in a Ub-conjugating system from which almost all ROS proteins other than rhodopsin have been excluded.
In summary, these data suggest that the (photo)receptor rhodopsin is ubiquitinylated by the UPP of ROS and reticulocyte lysate. The ligation of exogenous 125 I-labeled Ub to rho- dopsin in vitro, in conjunction with the detection of putative endogenous Ub-rhodopsin conjugates in purified ROS (Fig. 1C) and in stripped ROS membranes (Fig. 2B, lane 2), suggests a dynamic equilibrium of Ub conjugation to rhodopsin. We estimate that the percent of rhodopsin conjugated to Ub as detected in these studies is small (Ͻ1.0%; data not shown) and propose that only a subset of rhodopsin is ubiquitinylated under the conditions of light adaptation reported here.

Demonstration and Characterization of Functional E1 and E2 Enzymes in ROS
The catalytic function of E1 and E2 enzyme families involves the formation of thiol esters with one molecule of Ub (reviewed in Refs. 15 and 16). Both the E1ϳUb and E2ϳUb thiol esters are sensitive to reducing agents such as 2-mercaptoethanol, a fact that can be exploited to identify them on SDS-polyacrylamide gels (47,48). Brief incubations of the ATP-supplemented ROS supernatant with 125 I-labeled Ub generated distinct radiolabeled proteins on SDS-polyacrylamide gels run in the absence of 2-mercaptoethanol (Fig. 5A, lane 1). Consistent with these radiolabeled moieties being thiol esters is the observation that each radiolabeled species was not observed in the presence of 2-mercaptoethanol (Fig. 5A, lane 2). The apparent molecular masses of the thiol esters calculated from SDS-polyacrylamide gels were 120, 41, 36, 33, 28, 25, 24, and 23 kDa. The mass of the 120-kDa species is almost identical to that reported for thiol esters of Ub and mammalian E1 enzymes (Ϸ110 kDa) (48). The electrophoretic mobilities of the 41-, 33-, 28-, and 25-kDa 2-mercaptoethanol-sensitive 125 I-labeled Ub-protein adducts were almost identical to those of some (but not all) E2ϳUb thiol esters detected in rabbit reticulocyte lysate (data not shown), suggesting that bovine ROS contain homologues of the previously identified rabbit E2 35K , E2 25K , E2 20K , and E2 14K (in actuality, a 17-kDa protein) (42,48). This conclusion is supported by immunological and biochemical data obtained from nonretinal bovine tissues. 3 Unlike reticulocytes, ROS did not contain E2 230K (55), and the activities of ROS E2 35K , E2 25K , and E2 20K relative to E2 14K were markedly less than in reticulocyte lysate (data not shown). The 36-kDa thiol ester (Fig. 5A,  lane 1) is likely to be a previously reported higher mass thiol ester of E2 20K containing two Ub molecules (42). Activity assays thus demonstrate that ROS contain Ub-activating enzyme(s) (E1) and Ub carrier proteins (E2 enzymes) capable of forming thiol esters with Ub, i.e. requisite catalytic intermediates in the formation of Ub-protein conjugates.

Degradation of 125 I-Labeled Ub-ROS Protein Conjugates by ROS Supernatant
Cellular proteins marked for selective ATP-dependent proteolysis by conjugation to Ub are degraded by the 26 S proteasome, a large (Ϸ1500 kDa) enzyme complex. The mammalian 26 S proteasome is composed of a subset of Ͼ10 proteins (Յ32 kDa) of the 20 S proteasome and additional subunits with masses between 42 and 110 kDa (27,28,31). We have determined by immunoblotting that bovine ROS contain subunits common to both the 20 S and 26 S proteasomes (e.g. PA28) (31) as well as at least one subunit (MSS1/S7) (27,28) specific to the 26 S proteasome. 4 We assessed 26 S proteolytic activity in ROS by testing whether two inhibitors of the 26 S proteasome (i.e. the nonhydrolyzable ATP analogue AMP-PNP and the peptide aldehyde MG-132) inhibited the time-dependent loss of Ubprotein conjugates in the ROS supernatant. 125 I-Labeled Ubprotein conjugates were generated by the ROS supernatant during an initial pulse with radiolabeled Ub. In the presence of AMP-PNP (Fig. 6A, lane 2), steady-state levels of 125 I-labeled Ub conjugates were 3-4-fold greater than in the presence of ATP (lane 1). This observation suggests that 125 I-labeled Ub-ROS protein conjugates are stabilized when the 26 S proteasome is inhibited.
The time-dependent loss of high mass (Ͼ135 kDa) radioactive species was monitored by SDS-PAGE and autoradiography following a chase with nonradiolabeled Ub. After 10 min of chase, 38% of high mass 125 I-labeled Ub-protein conjugates were lost from assays containing ATP. In contrast, only 10% of high mass conjugates were lost in assays containing AMP-PNP (p Ͻ 0.05) (Fig. 6B). The percent loss of short-lived, radiolabeled high mass species measured at 30 min into the chase was also markedly reduced in the presence of AMP-PNP (Fig. 6B).
MG-132 also inhibited the loss of Ub-protein conjugates generated by the ATP-supplemented ROS supernatant. After 5 min of chase, 25% of high mass radiolabeled species were lost from (control) assays not containing MG-132, whereas only 16% of high mass conjugates were lost from assays containing MG-132 (p Ͻ 0.01) (Fig. 6C). The percent loss of short-lived, radiolabeled high mass species measured at 20 min into the chase was significantly reduced in the presence of MG-132 as well (Fig. 6B). Because MG-132 also inhibits lysosomal cysteine proteases (e.g. cathepsin B) and calpains (36), we evaluated the effects of the cysteine protease inhibitor E-64 and the calcium chelator EGTA on degradation of Ub-protein conjugates by the ROS supernatant. Neither the individual (data not shown) nor the combined effects of E-64 and EGTA inhibited rates of conjugate loss relative to untreated controls (Fig. 6D). This result indicates that the inhibition of 125 I-labeled Ub-protein conjugate loss obtained with MG-132 (Fig. 6C) is a specific consequence of 26 S proteasome inhibition. Taken together, the data from inhibitor studies support the conclusion that ROS contain an active 26 S proteasome capable of degrading ubiquitinylated ROS proteins. The data further indicate that these ubiquitinylated ROS proteins are in a dynamic state. These results are consistent with relatively high E2 14K and/or E2 17KB activity in ROS (Fig. 5A) because these E2 enzymes are acknowledged to mediate degradation of short-lived proteins by the UPP (56,57).
Differences in the percent loss of high mass radiolabeled species observed in control versus either AMP-PNP-or MG-132-treated assays became less apparent after 45 min of chase (data not shown). This loss of high mass conjugates in the presence of inhibitors of the 26 S proteasome suggests enzymatic removal of ubiquitin(s) from conjugated ROS proteins (i.e. deconjugation) by Ub C-terminal isopeptidase(s)/hydrolase(s) present in ROS (see below).

Characterization of ROS Ub C-terminal Isopeptidase/Hydrolase Activities
Ubiquitin monomers can be deconjugated from Ub-protein conjugates or from poly-Ub chains by the action of a variety of Ub C-terminal isopeptidases/hydrolases (29, 58 -60). To demonstrate isopeptidase activity in ROS, we employed pulse- chase conjugation assays (described above) in the presence or absence of Ubal, a specific inhibitor of many isopeptidases (52,59). ROS 125 I-labeled Ub-protein conjugates were generated in the presence of AMP-PNP and analyzed by SDS-PAGE and autoradiography after 30, 60, and 90 min of "chase" with nonradiolabeled Ub. Because AMP-PNP precludes conjugate degradation by the 26 S proteasome (see above), the loss of conjugates over time is due primarily to the action of Ub isopeptidase(s) and perhaps nonspecific proteolysis. After a 30-min pulse with 125 I-labeled Ub, heterogeneous 125 I-labeled Ub-protein conjugates were detected over a full range of masses (Fig. 7, lane 1), the quantity and pattern of which were comparable in the presence or absence of Ubal (data not shown). However, after 30 min of chase, almost all 125 I-labeled Ub adducts were lost from assays lacking Ubal (Fig. 7, lane 2). In contrast, distinct 24 -85-kDa 125 I-labeled Ub adducts were present in assays containing the isopeptidase inhibitor (Fig. 7,  compare lanes 5 and 2), and these radiolabeled species were stabilized for up to 90 min in the presence of inhibitor (Fig. 7,  compare lanes 6 and 3 and lanes 7 and 4). These results confirm Ub C-terminal isopeptidase/hydrolase activity in ROS. However, we noted that it was primarily (some) lower mass (Յ85 kDa) Ub adducts that were stabilized in the presence of Ubal. Both absolute and relative levels of high mass (Ն140 kDa) radiolabeled species detected after 30 min of chase were significantly reduced relative to steady-state levels measured at the initiation of the chase (Fig. 7, compare lanes 2 and 5 with lane  1) and were less affected by the presence of inhibitor (compare lanes 2 and 5). The reduced ability of Ubal to stabilize high mass Ub-protein conjugates in ROS contrasts with the severalfold enhancement of levels of high mass 125 I-labeled Ubprotein conjugates observed in the Ubal-treated supernatant from RPE cells (Fig. 7, compare lanes 8 and 9). These observations suggest that ROS contain potent isopeptidase activity that is resistant to inhibition by Ubal (see "Discussion"). This isopeptidase activity probably accounts for the loss of high mass Ub-protein conjugates observed in the presence of Ubal and inhibitors of the 26 S proteasome (Fig. 6, B and C). DISCUSSION As part of the daily process of photoreceptor renewal, the tips of photoreceptor cell outer segments are phagocytosed and nonselectively degraded in the lysosomes of the overlying RPE (61,62). Because phagocytosis is under circadian control and because only the tips of the outer segments are ingested, this mechanism can provide neither rapid nor localized control of protein levels within outer segments. Accordingly, we investigated intracellular proteolytic mechanisms that may provide relatively rapid and selective in situ regulation of levels of vertebrate photoreceptor proteins, including phototransduction proteins. Because the UPP is nonlysosomal, selectively degrades highly regulated proteins, and has been shown to degrade the plant regulatory photoreceptor protein, phytochrome (63), it was proposed as a potential regulator of phototransduction protein levels and/or activities within ROS (19 -22). We presented evidence in support of this hypothesis by demonstrating that rod G t was a preferred substrate for UPPdependent proteolysis in non-ROS cell-free systems (19,20). However, until the present study, no published report had verified that photoreceptor outer segments possess a functional UPP, that phototransduction proteins are endogenous substrates of this pathway in ROS, and that localization of these conjugates is affected by light levels.
Both the level of free Ub in ROS (18 -24 pmol/mg of ROS protein) and the observation that most ROS Ub exists in the conjugated form typify other tissues with demonstrated UPP activity (64,65). These observations, in conjunction with immunogold localization of Ub moieties within ROS, 4 suggested that ROS contained an active UPP. However, because all ROS protein is synthesized in the rod cell body and translocated into the outer segment, it was plausible that Ub-protein conjugates detected in ROS were ubiquitinylated outside of ROS and translocated into the outer segment. Activity assays subsequently confirmed for the first time that ROS contain the three hallmark activities of the UPP, i.e. ATP-dependent Ub conjugating activity (Figs. [3][4][5], ATP-and 26 S proteasome-dependent protein degrading activity (Fig. 6), and Ub C-terminal isopeptidase/hydrolase activity (Fig. 7). Verification that the rod cell microenvironment specialized for phototransduction contains both a functional UPP and rapidly degraded endogenous UPP substrates represents the first evidence obtained from photoreceptors consistent with the hypothesized regulation of phototransduction proteins by the UPP. As such, these data extend prior reports of UPP activity or moieties in whole retina (21, 22, 66 -68).
Demonstration of UPP Activity in ROS-Ub conjugating activity requires sequential enzymatic reactions catalyzed by the Ub-activating enzyme(s) E1 and one of a variety of Ub carrier proteins (E2 enzymes), often in association with a Ub-protein isopeptide ligase (E3) (reviewed in Refs. 15 and 16). Thiol ester assays documented the presence within ROS of E1 and at least four E2 enzymes (i.e. E2 14K , E2 20K , E2 25K , and E2 35K ). Of the four E2 enzymes detected in ROS, E2 14K is most clearly implicated in Ub-dependent proteolysis in other tissues. In association with one of two E3 enzymes (E3␣ and E3␤), E2 14K selectively polyubiquitinylates so-called "N-end rule" substrates i.e. proteins whose interaction with E3 enzymes and subsequent half-lives are determined by their N-terminal residue (Ref. 57;reviewed in Ref. 14). The preponderance of E2 14K ϳUb thiol esters generated in the ROS supernatant argues that at least some soluble Ub-protein conjugates in ROS are catalyzed by an E2 14K /E3 mechanism and that these conjugates are substrates for degradation by the 26 S proteasome (see below). Moreover, because UPP-dependent degradation of Saccharomyces cerevisiae G ␣ is catalyzed by the yeast E2 14K homologue (UBC2 (ubiquitin carrier enzyme)) (17), we speculate that E2 14K may mediate the UPP-dependent conjugation and degradation of G t previously reported by this laboratory (Refs. 19 and 20; see below). E2 35K may interact with E2 14K to support protein degradation, as suggested by studies of their respective yeast homologues (UBC3 and UBC2) (69). The role of E2 25K in protein degradation is less clear. E2 25K catalyzes the synthesis of Gly 76 -Lys 48 -linked poly-Ub chains from free ubiquitins (33). These chains are competent intermediates in the UPP (33,70) and may act as negative regulators of 26 S proteasome activity (65).
Analyses of endogenous ROS conjugates and conjugates generated by the ROS supernatant in vitro indicate that the ROS Ub-conjugating enzyme system catalyzes the formation of numerous high mass (Ͼ200 kDa), presumably polyubiquitinylated soluble ROS protein(s) (Fig. 3). High mass conjugates are believed to be especially susceptible to proteolysis by the 26 S proteasome (Ref. 71; see also Ref. 35). Consistent with this observation, we demonstrated in pulse-chase assays that high mass adducts of cytosolic ROS protein(s) and Ub are rapidly degraded by the ROS 26 S proteasome (Fig. 6). Ubiquitinylated moieties with masses of Ͻ150 kDa were not degraded (data not shown). The rapid turnover of these high mass ubiquitinylated species is consistent with the demonstrated role of the UPP in the turnover of short-lived (i.e. regulatory) proteins (14,16). These results corroborate and extend previous studies by this laboratory in which soluble ROS proteins were selectively polyubiquitinylated and degraded to acid-soluble species by the UPP of reticulocyte lysate and the RPE supernatant (19,20).
Soluble Ub-protein conjugates in ROS were also substrates for endogenous Ub C-terminal isopeptidase/hydrolase activity, as indicated by increased yields of endogenous soluble conju- Isopeptidase activity was assessed in pulse-chase conjugation assays (as described in legend for Fig. 6) in the presence or absence of the isopeptidase inhibitor Ubal (1 M final concentration). All assays contained AMP-PNP and MG-132 in order to support conjugation but inhibit conjugate proteolysis. Steady-state levels of 125 I-labeled Ubprotein conjugates catalyzed in the presence (lane 1) or absence (not shown) of Ubal were indistinguishable. Following the addition of excess nonradiolabeled Ub, the time-dependent loss of conjugated 125 I-labeled Ub was determined (as described for Fig. 6) in the absence (lanes 2-4) and presence (lanes 5-7) of Ubal. Efficacy of Ubal was apparent in the RPE supernatant, in which Ub C-terminal isopeptidase activity directed toward high mass forms of conjugated 125 I-labeled Ub was significantly inhibited (compare lanes 8 and 9). Mobilities of molecular mass standards are indicated between lanes 1 and 2. min, minutes of chase.
gates when ROS purifications included the isopeptidase inhibitor iodoacetamide (data not shown) and by stabilization of 125 I-labeled Ub-protein conjugates in the presence of the isopeptidase inhibitor Ubal (Fig. 7). Ub C-terminal isopeptidases/hydrolases regulate steady-state levels of free Ub in cells by cleaving Ub-protein or Ub-Ub linkages. The deconjugating activities of these enzymes can regulate rates of ubiquitin-dependent proteolysis as well as modulate the interaction of ubiquitinylated proteins with other cellular components (29, 58 -60). Our data suggest that ROS contain at least two distinct Ub C-terminal isopeptidase/hydrolase activities. One acts preferentially on high mass conjugates, is relatively insensitive to Ubal (Fig. 7), and was probably responsible for the loss of high mass conjugates observed in degradation assays containing 26 S proteasome inhibitors (Fig. 6, B and C). This activity may be due to isopeptidase T, a 100-kDa Ubal-resistant enzyme that preferentially cleaves Ub-Lys 48 -Ub linkages in poly-Ub chains of high mass conjugates (59). The other isopeptidase activity is clearly inhibitable by Ubal, appears to act on smaller (Ͻ85 kDa) ubiquitinylated moieties (Fig. 7), and may be attributable to several isopeptidase species or isoforms reported in other tissues (59,60). Preliminary evidence suggests, however, that ROS contain characteristic Ub C-terminal isopeptidase/hydrolase activities distinct from those in other retinal cells. We propose that ROS are distinguished by (i) relatively high levels of Ubal-insensitive isopeptidase(s) (see above); (ii) the demonstrated absence of Ub C-terminal hydrolase PGP 9.5, present in high concentrations elsewhere in the retina (67,68); and because ROS are not sites of protein synthesis, (iii) the absence of Ub C-terminal ␣-NH 2 -protein hydrolase activity required for the processing of Ub gene products (29). The particular Ub C-terminal isopeptidase/hydrolase activities in ROS presumably reflect the highly specialized function of photoreceptor cell outer segments.
Endogenous Substrates for Ubiquitinylation in Light-or Dark-adapted ROS-We have tentatively identified two phototransduction proteins that are substrates of the ROS UPP. First, we have identified a 38-kDa ubiquitinylated protein whose binding to ROS disc membranes exhibits characteristics similar to those of G t . Specifically, the conjugate is somewhat loosely associated with dark-adapted ROS membranes, from which it can be eluted with hypotonic buffer, but binds tightly to light-adapted ROS membranes, from which it can be selectively eluted with GTP (Fig. 2). These data strongly suggest that the 38-kDa conjugate contains a subunit of G t . Because the ligation of each Ub molecule increases the apparent molecular mass of a substrate by 6 -8 kDa, the 38-kDa conjugate must contain G t␥ (6.5 kDa) rather than G t␤ (37 kDa) or G t␣ (40 kDa). The ligation of four Ub molecules to G t␥ would produce an adduct with an apparent mass of Ϸ38 kDa.
Evidence for ubiquitinylation of G t␥ by the ROS UPP corroborates previous in vitro studies in which radiolabeled G t␥ was degraded by the UPP of human RPE and rabbit reticulocytes (19,20). Those studies also indicated that G t␣ and G t␤ were preferred UPP substrates as well; however, in this study, ubiquitinylated intermediates of either of these two subunits were not apparent in the G protein-rich (GTP) supernatant. We speculate that ubiquitinylated forms of G t␣ and G t␤ may be degraded sufficiently rapidly to preclude their detection by immunoblotting. Alternatively, only G t␥ is conjugated to Ub, and this ubiquitinylation is sufficient to target the intact heterotrimer (G t␣␤␥ ) for degradation by the 26 S proteasome. In yeast, the G protein ␣-subunit is degraded by the UPP via the N-end rule (17,57). If degradation of G t also proceeds via the N-end rule, ubiquitinylation of any or all of the three G t subunits must involve either cleavage of the ␣-NH 2 group or the presence of an accessory protein that binds to the G protein and is itself recognized by an N-end rule Ub-protein isopeptide ligase (E3␣ and E3␤) (reviewed in Ref. 14). These requirements follow from the fact that each G t subunit possesses either an acetylated or "stabilizing" N-terminal residue that does not support efficient N-end rule ubiquitinylation catalyzed by E2 14K /(E3␣/E3␤). Irrespective of the conjugating enzymes involved, ubiquitinylation of G t could be coupled to the visual transduction cascade through the light-dependent conformational alterations in G t or G t subunits that are manifest during phototransduction (1)(2)(3)(4)72). These conformational changes may expose or mask lysine residues on G t that are targets for Ub ligation.
This study also presents several lines of evidence that rhodopsin is ubiquitinylated in vivo and can be ubiquitinylated in vitro. These include (i) the predominance in whole ROS of endogenous ubiquitinylated proteins whose apparent molecular masses (50 to Ն200 kDa) and pI values (5.8 -6.8) are consistent with those predicted for ubiquitinylated rhodopsin and rhodopsin oligomers (Fig. 1, B and C); (ii) the presence in membranes, from which extrinsic proteins were removed, of 44and 50-kDa Ub-protein conjugates whose masses suggest a ladder of mono-and diubiquitinylated rhodopsin, respectively (Fig. 2B, lane 2); (iii) the ATP-dependent generation in whole ROS of 125 I-labeled Ub-protein conjugates whose masses (42,50, and 58 kDa) also suggest a ladder of mono-, di-, and triubiquitinylated rhodopsin (Fig. 4A, lane 3); (iv) parallel electrophoretic mobility shifts of rhodopsin and the predominant 125 Ilabeled Ub-protein conjugates catalyzed by whole ROS when samples were boiled before electrophoresis (Fig. 4A); (v) catalysis of the formation of novel 125 I-labeled Ub-protein conjugates by cell-free systems from which almost all ROS proteins other than rhodopsin were excluded (Fig. 4B); and (vi) immunolocalization of Ub moieties to ROS disc membranes at the electron microscope level. 4 These findings and recently described work in yeast (see below) are to our knowledge the first reports of ubiquitinylation of a G protein-coupled receptor.
Our data further indicate that an E2 present in both ROS and reticulocyte lysate can catalyze the ligation of Ub to rhodopsin. However, it is unlikely that ubiquitinylation of rhodopsin in situ is mediated by the E2 14K /(E3␣/E3␤)-dependent mechanism that ubiquitinylates N-end rule substrates (see above) since the N terminus of rhodopsin, like that of other intrinsic membrane proteins, extends into the extracytosolic (intraluminal) space and is consequently inaccessible to conjugating enzymes that require recognition of a substrate's Nterminal residue. Thus, rhodopsin ubiquitinylation either does not appear to require a Ub-protein isopeptide ligase (E3) or else involves an as yet unspecified E3 other than E3␣ or E3␤.
Levels of endogenous and in vitro catalyzed ubiquitinylated moieties in stripped ROS membranes appeared relatively stable in the absence of inhibitors of 26 S proteasome activity (data not shown). This observation is consistent with a prior report that rhodopsin is not degraded in vitro by an energyrequiring neutral protease (73). In yeast, ubiquitinylation targets intrinsic membrane proteins, including the G proteincoupled mating pheromone receptor, 5 for degradation in the yeast vacuole (74), the organelle analogous to the lysosome. Although ubiquitinylation could conceivably assist the delivery of rhodopsin to the RPE lysosome following phagocytosis of shed ROS tips (61,62), the absence of lysosomes in ROS suggests that rhodopsin ubiquitinylation serves a function other than lysosomal targeting. Ubiquitinylation could modulate the receptor's signal transducing properties because seven of rho-dopsin's nine cytoplasmic ("ubiquitinylatable") lysines are contained within regions of the molecule that participate in G protein binding and activation, receptor folding and stabilization, or receptor phosphorylation and arrestin binding (1,2,75).
Considered together, the results of this study provide a new perspective regarding photoreceptor protein regulation. They demonstrate that photoreceptor proteins can be ubiquitinylated and degraded in situ by an energy-dependent mechanism distinct from the nonselective, lysosomal proteolysis that occurs in RPE cells. Evidence for ubiquitinylation of G t and rhodopsin by the ROS UPP supports a novel role for the UPP in the regulation of phototransduction protein levels or activities. These findings, in conjunction with the demonstration that truncation of arrestin by calpain can modulate rhodopsin phosphorylation (76), suggest that selective proteolytic pathways may be intimately involved in the essential phototransduction function of photoreceptors. Conformational and functional conservation of heterotrimeric G proteins and their coupled receptors (1,77) renders it plausible that the UPP is involved in other vertebrate G protein-linked signaling pathways as well (e.g. Ref. 78). Finally, we suggest that this study, in conjunction with the acknowledged involvement of the UPP in programed cell death (23), implicates the UPP in the pathophysiology of various inherited retinal degenerations characterized by the selective post-translational loss of phototransduction proteins (79) and/or by photoreceptor cell apoptosis (12,80).