Cytoplasmic domain of rhodopsin is essential for post-Golgi vesicle formation in a retinal cell-free system.

In retinal photoreceptors, highly polarized organization of the light-sensitive organelle, the rod outer segment, is maintained by the sorting of rhodopsin and its associated proteins into distinct post-Golgi vesicles that bud from the trans-Golgi network (TGN) and by their vectorial transport toward the rod outer segment. We have developed an assay that reconstitutes the formation of these vesicles in a retinal cell-free system. Vesicle formation in this cell-free assay is ATP-, GTP-, and cytosol-dependent. In frog retinas vesicle budding also proceeds at 0 degrees C, both in vivo and in vitro. Vesicles formed in vitro are indistinguishable from the vesicles formed in vivo by their buoyant density, protein composition, topology, and morphology. In addition to the previously identified G-proteins, these vesicles also contain rab11. Concurrently with vesicle budding, resident proteins are retained in the TGN. Collectively these data suggest that rhodopsin and its associated proteins are sorted upon exit from the TGN in this cell-free system. Removal of membrane-bound GTP-binding proteins of the rab family by rab GDP dissociation inhibitor completely abolishes formation of these vesicles and results in the retention of rhodopsin in the Golgi. A monoclonal antibody to the cytoplasmic (carboxy-terminal) domain of rhodopsin and its Fab fragments strongly inhibit vesicle formation and arrest newly synthesized rhodopsin in the TGN rather than the Golgi. Therefore rhodopsin sorting at the exit from the TGN is mediated by the interaction of its cytoplasmic domain with the intracellular sorting machinery.

In retinal photoreceptors, highly polarized organization of the light-sensitive organelle, the rod outer segment, is maintained by the sorting of rhodopsin and its associated proteins into distinct post-Golgi vesicles that bud from the trans-Golgi network (TGN) and by their vectorial transport toward the rod outer segment. We have developed an assay that reconstitutes the formation of these vesicles in a retinal cell-free system. Vesicle formation in this cell-free assay is ATP-, GTP-, and cytosol-dependent. In frog retinas vesicle budding also proceeds at 0°C, both in vivo and in vitro. Vesicles formed in vitro are indistinguishable from the vesicles formed in vivo by their buoyant density, protein composition, topology, and morphology. In addition to the previously identified G-proteins, these vesicles also contain rab11. Concurrently with vesicle budding, resident proteins are retained in the TGN. Collectively these data suggest that rhodopsin and its associated proteins are sorted upon exit from the TGN in this cell-free system. Removal of membrane-bound GTP-binding proteins of the rab family by rab GDP dissociation inhibitor completely abolishes formation of these vesicles and results in the retention of rhodopsin in the Golgi. A monoclonal antibody to the cytoplasmic (carboxyl-terminal) domain of rhodopsin and its Fab fragments strongly inhibit vesicle formation and arrest newly synthesized rhodopsin in the TGN rather than the Golgi. Therefore rhodopsin sorting at the exit from the TGN is mediated by the interaction of its cytoplasmic domain with the intracellular sorting machinery.
Reconstitution of intracellular sorting events in cell-free systems has significantly contributed to our understanding of the molecular mechanisms that underlie membrane trafficking. Together with genetic studies in yeast and studies of regulated exocytosis of synaptic vesicles, in vitro assays reconstituting intra-Golgi transport have yielded invaluable information about common mechanisms involved in membrane budding and fusion from yeast to mammals (for review see Rothman and Warren (1994) and Rothman (1994)). Several cell-free assays have been employed to reconstitute early steps in exocytosis (endoplasmic reticulum to Golgi and intra-Golgi transport) and endocytosis. A limited number of studies have established in vitro conditions for cell-free post-Golgi vesicle formation from the TGN 1 (de Curtis and Simons, 1989;Salamero et al., 1990;Ohashi and Huttner, 1994). However, these events have not been studied in the primary tissues of neuronal origin.
Retinal photoreceptors are highly polarized sensory neurons with three distinct plasma membrane domains: rod outer segments (ROS) that capture light, the lateral plasma membrane of the inner segment, and the axon with the synaptic terminal. Photoreceptors are excellent models for cell biological studies because continuous ROS membrane renewal by polarized sorting of rhodopsin and its associated proteins on post-Golgi vesicles results in the addition of up to ϳ3 m 2 /min of ROS membranes (reviewed in Besharse (1986), Simons and Zerial (1993), and ). Rhodopsin-bearing post-Golgi vesicles isolated from frog retinas can be distinguished from all other subcellular compartments by several criteria, including kinetics of formation, buoyant density, and protein composition (Deretic and Papermaster, 1991). Several proteins of the small G-protein rab family and ␣-crystallins are vesicle-associated proteins and potential regulators of vesicle budding and sorting (Deretic andPapermaster, 1993a, 1995;Deretic et al., 1994. However, the mechanism of action of these associated proteins is still elusive in the absence of an assay that allows access to the intracellular compartments involved in vesicle formation. Inherited retinal diseases in humans and in animal models suggest that some of the mutations in the rhodopsin gene may lead to transport defects that ultimately result in retinal degeneration and blindness (Dryja et al., 1990;Olsson et al., 1992;Roof et al., 1994;Sung et al., 1991Sung et al., , 1994. The rhodopsin residues that are mutated in inherited retinal degenerations are conserved in the frog rhodopsin sequence, consistent with their important contribution to the stability of rhodopsin structure and its function in the photoreceptor cell (Pittler et al., 1992). In addition, retinal cells also appear to be very sensitive to the defects in isoprenylation of small G-proteins and rab proteins in particular (Seabra et al., 1993;Pittler et al., 1995). Therefore, identification of the proteins that interact with rhodopsin and guide it on its journey to the ROS, as well as the establishment of the role of rab proteins in these processes are of essential importance for the understanding of the establishment and maintenance of photoreceptor polarity and health.
We have developed a cell-free assay that reconstitutes not only rhodopsin transport through the Golgi but also post-Golgi vesicle budding. Due to the complexity of the retinal tissue, biosynthetic organelles of the rod inner segments are accessible only upon removal of the ROS and the neural retina. Using the fractionation technique described before (Deretic and Paper-master, 1991), we have generated a frog retinal PNS that is highly enriched in the photoreceptor cell's biosynthetic membranes. When supplemented with ATP and an exogenous ATPregenerating system, this PNS is able to support rhodopsin transport and sorting in vitro. We used this cell-free assay to study the factors involved in the biogenesis of rhodopsin-bearing post-Golgi vesicles. We find that vesicle formation depends on the access of cytoplasmic proteins to the carboxyl-terminal domain of rhodopsin in the TGN membranes and on the presence of membrane-bound rab proteins.
Pulse Labeling of Frog Retinas and Preparation of Photoreceptorenriched PNS-All experiments were conducted under dim red light. Frogs were dark-adapted for 2 h before the experiment. Isolated frog retinas were incubated in oxygenated medium as described (Deretic andPapermaster 1991, 1993b). Seven retinas were incubated with [ 35 S]-Express protein labeling mixture (25 Ci/retina) in 15 ml of media at 22°C for 1 h.
To obtain PNS enriched in photoreceptor biosynthetic membranes, retinal fractionation was performed as described by Deretic andPapermaster (1991, 1993b). Retinas were sheared through a 14-gauge needle, and ROS were removed by flotation on 34% sucrose. Retinal pellets were rehomogenized in 0.25 M sucrose in 10 mM Tris acetate, pH 7.4, containing 1 mM MgCl 2 and spun at 4000 rpm (1250 g av ) (JA20 rotor, Beckman Instruments, Inc., Palo Alto, CA) for 4 min. Supernatant obtained after this centrifugation (PNS) contains biosynthetic organelles that participate in rhodopsin transport (Papermaster et al., 1975;Deretic and Papermaster, 1991), and it was used in subsequent in vitro incubations.
In Vitro Incubation of Photoreceptor-enriched PNS-The standard assay for cell-free post-Golgi vesicle formation was as follows: to 1 ml of PNS in 0.25 M sucrose (obtained from seven radiolabeled retinas), 100 l of 10ϫ concentrated buffer stock solution was added to give a final concentration of 25 mM Hepes-KOH, pH 7.0, 25 mM KCl, and 2.5 mM MgAc. The assay was initiated by the addition of 50 l of an ATPregenerating or an ATP-depleting system (Davey et al., 1985) and transfer to 22°C. The ATP-regenerating system contained equal volumes of 100 mM ATP (sodium salt), 800 mM creatine phosphate, and 4 mg/ml creatine phosphokinase in 50% glycerol. The ATP-depleting system contained 10 mg/ml hexokinase in 250 mM D-glucose. The assay was terminated by the addition of 0.2 M EDTA to a final concentration of 3 mM, and the assay mixture was loaded on linear sucrose gradients.
In some experiments PNS was subfractionated into membrane-enriched and cytosolic fractions. 1 ml of PNS in 0.25 M sucrose (obtained from 14 retinas) was overlaid on a step sucrose gradient containing 1 ml of 0.5 M sucrose and 3 ml of 49% sucrose. Gradients were spun for 1 h at 28,000 rpm (75,000 g av ) in an SW 50.1 rotor (Beckman). Soluble proteins that remained in 0.25 M sucrose were further centrifuged at 75,000 rpm (230,000 g av ) in an TLA 100.3 rotor (Beckman) to remove residual membranes, and supernatants containing cytosolic proteins were collected. Membrane proteins that entered the 0.5 M sucrose layer were collected and divided in half. To one half of the membranes 0.5 ml of 5 mg/ml bovine serum albumin in 0.25 M sucrose was added (Ϫcytosol), and to the other half 0.5 ml of cytosolic proteins was added (ϩcytosol), restoring photoreceptor-enriched PNS to the original ratio of cytosol:membranes. Reconstituted PNS (Ϯcytosol) was assayed for vesicle formation after addition of 10ϫ buffer stock, ATP, and an ATPregenerating system as described above.
Upon completion of cell-free vesicle formation, assay mixtures were overlaid on 10 ml-linear 20 -39% (w/w) sucrose gradients in 10 mM Tris acetate, pH 7.4, and 1 mM MgCl 2 above a 0.5-ml cushion of 49% (w/w) sucrose. After centrifugation at 28,000 rpm (100,000 g av ) in an SW 40 rotor (Beckman) for 15 h at 4°C, 0.9-ml fractions were collected from the top of the gradient. For each fraction the refractive index was determined from an aliquot, and the remainder was diluted with 10 mM Tris acetate, pH 7.4, and centrifuged at 50,000 rpm (230,000 g av ) for 40 min in a SW 50.1 rotor or at 75,000 rpm in an TLA 100.3 rotor. In some experiments subcellular fractions were pooled according to the kinetics of their acquisition of radiolabeled rhodopsin. Pellets were resuspended in 10 mM Tris acetate, pH 7.4, and aliquoted for analysis by SDS-PAGE or two-dimensional gel electrophoresis. Gel Electrophoresis, Autoradiography, Immunoblotting, and Detection of the GTP-binding Proteins-SDS-polyacrylamide and two-dimensional gel electrophoresis were performed as described previously (Deretic and Papermaster, 1993a;Deretic et al., 1994. To quantitate 35 S-labeled rhodopsin in retinal subcellular fractions, dried SDS gels were autoradiographed for various times at Ϫ70°C using Kodak BioMax MR film with intensifying screens. Dried gels were also exposed to storage phosphor screens, and the intensity of luminescence associated with the rhodopsin band was measured and analyzed by a Phos-phorImager densitometer (Molecular Dynamics).
Immunoblotting was performed as described  using the ECL Western blotting Detection System (Amersham Corp.). GTP-binding proteins were detected by [ 32 P]GTP overlays as described (Deretic and Papermaster, 1993a). Blots were autoradiographed at Ϫ70°C using Kodak BioMax MR film with intensifying screens. Autoradiograms and Hyperfilm ECL were scanned using the Image processing and analysis program (Wayne Rasband, NIH).
Miscellaneous Procedures-To determine the distribution of the TGN membranes sialyltransferase was assayed as described (Deretic and Papermaster, 1993a). Post-Golgi vesicle membranes were digested with thermolysin according to Deretic and Papermaster (1991).
Electron Microscopy-Membranes from the sucrose gradient fractions were pelleted for 1 h at 50,000 rpm in a Beckman SW 50.1 rotor with adaptors to obtain a small pellet. Pellets were fixed with 2% glutaraldehyde in 120 mM cacodylate, pH 7.4, containing 3% sucrose for 30 min on ice, postfixed with OsO 4 , stained with uranyl acetate, and embedded in 2% agarose as described (Deretic and Papermaster, 1991). Blocks of membranes in agarose were dehydrated in ethanol and embedded in Epon. Thin sections along the axis of sedimentation were examined in a Philips 301 electron microscope.

Post-Golgi Vesicles Carrying Newly Synthesized Rhodopsin
Are Formed in a Retinal Cell-free System in the Presence of ATP-Newly synthesized rhodopsin is incorporated by photoreceptor cells into post-Golgi vesicles when isolated frog retinas are pulse-labeled for 60 min and chased for 2 h (Fig. 1, top panel; Deretic and Papermaster (1991)). Rhodopsin-bearing post-Golgi vesicles formed in retinal cultures have buoyant density of 1.09 g/ml and sediment in fractions 4 -6 (maximum fraction 5) on the shallow linear equilibrium sucrose density gradients, which allows their separation away from the TGN as well as the Golgi, plasma membrane, synaptic vesicles, and endosomes (Deretic andPapermaster, 1991, 1993a;. Immunoisolation using specific antibodies has shown that fractions 4 -6 are highly enriched (Ͼ85%) in rhodopsin-bearing post-Golgi vesicles (Deretic and Papermaster, 1991).
To study the role of intrinsic and associated components that regulate vesicle formation, we have reconstituted this process in a cell-free system. For this purpose we modified the assay described by Huttner (1990, 1992). Retinal proteins are pulse-labeled for 60 min in the isolated frog retinas, the rod outer segments are removed, and the remainders of the retinas are homogenized. After low speed centrifugation to remove neural retina and nuclei, retinal PNS is supplemented with ATP and an ATP-regenerating system or with an ATP-depleting system (Davey et al., 1985) and incubated at 22°C for an additional 2 h. Fig. 1 shows that in PNS supplied with ATP, radiolabeled rhodopsin appears in vesicle-enriched fractions 4 -6 of the gradient. By contrast, ATP depletion arrests rho-dopsin in the heavy fractions (9 -12) enriched in Golgi and TGN membranes. The identity of these compartments was previously demonstrated by their galactosyltransferase and sialyltransferase activity, respectively (Deretic andPapermaster (1991, 1993a); see also Fig. 3). The distribution of newly synthesized rhodopsin in the gradient fractions after in vitro vesicle formation in the presence of ATP closely parallels its distribution after in vivo vesicle formation ( Fig. 1), suggesting that all the components necessary for this process are present in the photoreceptor-enriched PNS. In the presence of ATP, fractions 4 -6 contain ϳ32% of newly synthesized rhodopsin compared with 15% in the absence of ATP. The amount of the post-Golgi vesicles formed in the absence of ATP represents that which is initially formed in vivo during the 60-min pulse prior to breaking the cells (see Fig. 4).
ATP-dependent Formation of Rhodopsin-bearing Vesicles in Frog Retinal PNS Also Proceeds at Low Temperature-To test the temperature dependence of the cell-free vesicle formation we have also incubated photoreceptor-enriched PNS in the presence or the absence of ATP on ice. The results are shown in Fig. 2. ATP-dependent vesicle formation proceeds in the chilled PNS with the same kinetics and to the same extent as when they are incubated at 22°C. Vesicle formation at 0°C is blocked by ATP depletion to a similar extent as at 22°C (compare Figs. 1 and 2). However, the distribution of radiolabeled rhodopsin in the gradient is different; whereas at 22°C rhodopsin predominantly accumulates in the Golgi-enriched fractions 11 and 12 (identified by their galactosyltransferase activity), at 0°C these fractions are partially depleted, and rhodopsin is arrested in fractions 7-10 that contain TGN mem-branes (Fig. 3). This suggests that rhodopsin transport proceeds until ATP is depleted by the frog photoreceptor ATPases rather than by the exogenous ATP-depleting system that is apparently inefficient at low temperature. Vesicle formation also proceeds in vivo in isolated frog retinas incubated at 0°C (Fig. 4).
Resident Membrane Proteins of the TGN Are Retained during Cell-free Vesicle Budding-We wanted to verify that our in vitro assay reconstitutes the physiological process of vesicle formation rather than the fragmentation of the TGN membranes, especially because this process does not appear to be slowed down at low temperature. Fig. 3 shows that the distribution of sialyltransferase, a resident membrane protein of the TGN, is not affected by vesicle budding in the cell-free assay at 22 or 0°C, and it parallels the distribution of the enzyme after in vivo vesicle formation. The activity of sialyltransferase in fractions 4 -6 is negligible. This suggests that our assay indeed monitors vesicle budding and not the fragmentation of the TGN.
Fractions 7-10 are enriched in sialyltransferase activity contributed by the TGN membranes. These are also the fractions that accumulate radiolabeled rhodopsin during the chase on ice in the absence of added ATP (see Fig. 2).
Cytosol and G-proteins Are Necessary for Vesicle Formation-To test the role of cytosolic proteins in vesicle formation, pulse-labeled retinal PNS is further centrifuged through a 0.5 M sucrose cushion, and membranes that enter the cushion are separated from the soluble proteins. After this fractionation step, which separates a significant proportion of cytosolic proteins from the membranes, vesicle formation, as measured by the accumulation of radiolabeled rhodopsin in pooled fractions 4 -6, is diminished to ϳ25% of control and is completely restored if the cytosol is added back (Fig. 4). Similarly, 1 M GTP␥S inhibits vesicle formation by ϳ50% after 30 min of preincubation. These data are consistent with the existence of a pool of membrane-bound cytosolic proteins and GTP-activated G-proteins that are interacting at the onset of incubation and appear to sustain initial rounds of vesicle budding. In addition, the observed inhibitory effect of GTP␥S is a consequence of its pleiotropic effect on heterotrimeric and small G-proteins.
Small GTP-binding Proteins Are Sorted to the Vesicles Formed in Vitro-We have compared the signature subset of GTP-binding proteins of post-Golgi vesicles formed in vivo (Deretic and Papermaster, 1993a; with the content of vesicles that are formed in vitro by high resolution two-dimensional gel electrophoresis and [ 32 P]GTP overlays. Fig. 5 shows that the GTP-binding protein composition of the vesicles formed in vivo or in vitro at 22 or 0°C are indistinguishable; they contain previously identified proteins rab6, rab8, and rab3, as well as another TGN and post-Golgi vesiclespecific protein rab11, now identified using the rab11 antiserum (Urbé et al., 1993). After vesicle formation, Golgi-enriched fractions retain a more complex small G-protein pattern, as observed after in vivo vesicle formation. This suggests that the post-Golgi vesicle budding in the retinal cell-free system is selective because only specific rab proteins, previously identified from retinal cultures, are properly sorted to the vesicles formed in vitro, and no additional GTP-binding proteins are found associated with their membranes.
High Resolution Two-dimensional Gel Electrophoresis Reveals the Greatly Simplified Protein Pattern of Post-Golgi Vesicles Formed in Vitro Compared with the Golgi-enriched Membranes-We have now extended our previous analysis of vesicle-associated GTP-binding proteins by two-dimensional gel electrophoresis to include all vesicle proteins. Silver-stained gels of post-Golgi vesicles formed in vitro at 22°C and its corresponding Golgi enriched fraction are shown in Fig. 6. The   FIG. 3. Membrane proteins of the TGN are retained during cell-free vesicle budding. The distribution of sialyltransferase, a resident membrane protein of the TGN, is not affected by vesicle budding in a cell-free assay at 22 or 0°C and resembles its distribution after in vivo vesicle formation (control). The bulk of sialyltransferase is associated with fractions 7-11, and very little is found in vesicle-enriched fractions 4 -6. After the completion of an in vitro or an in vivo chase, each subcellular fraction is assayed for sialyltransferase activity, and the distribution of the enzyme is expressed as a percentage of total activity recovered in retinal subcellular fractions.

FIG. 4. Cytosol and G-proteins are required for vesicle formation.
Retinal subcellular fractions 4 -6, which are enriched in the post-Golgi vesicles, contain 16% of newly synthesized rhodopsin after a 60-min pulse (1) and 32% following the 2 h chase in vivo at 22°C (2). In the presence of ATP in the cell-free chase, a similar amount of radiolabeled rhodopsin is found in these fractions at 22 (5) and 0°C (7). ATP depletion completely inhibits additional vesicle formation during the chase at both 22 (4) and 0°C (6), i.e. there is no difference from the distribution during the pulse. Upon cytosol removal, vesicle formation during the chase is abolished by 75% (9), and when cytosol is added back it is completely restored (10). 1 M GTP␥S diminishes vesicle formation to ϳ50% of control (8). The data are expressed as the mean Ϯ S.E. of three to eight separate experiments.

FIG. 5. Post-Golgi vesicles formed in vivo and in vitro have identical content of membrane-associated small G-proteins.
rab11 is also a vesicle protein. After subcellular fractionation and high speed centrifugation, membrane-associated GTP-binding proteins from two frog retinas were detected by two-dimensional gel electrophoresis and [ 32 P]GTP overlays (A-D). Whereas the Golgi-enriched fraction has a more complex pattern (A) post-Golgi vesicles formed in vitro at 22 (B) or 0°C (C) are identical to the vesicles formed in vivo in intact retinal cultures (D). Previously identified proteins rab6 and rab8 are indicated by open and closed arrowheads, respectively. E, the post-Golgi vesicles also contain rab11. rab11 is identified in this panel by the anti rab11 antiserum (Urbé et al., 1993); it is also indicated with an arrow in D. The subcellular distribution of rab11 parallels that of rab6 in retinal photoreceptors (data not shown). protein composition of the vesicles formed in vivo (not shown) is identical to that of the vesicles formed in vitro, in keeping with already established identity of their GTP-binding proteins (shown in Fig. 5). Although the high resolution and very sensitive staining procedure reveal more vesicle-associated proteins than previously recognized, the protein composition of the vesicles is still simple when compared with the Golgi-enriched fraction. Anomalous isoelectric focusing of rhodopsin in these gels (probably due to its unusual carbohydrate chains) does not permit its identification as a well focused spot. Immunoblotting with anti-rhodopsin antibodies and the autoradiograms of the 35 S-labeled vesicles reveal a smear in the 35-kDa region (data not shown), which is not apparent by silver staining. Other previously characterized vesicle proteins are readily identified by this technique: the ␣ subunit of transducin, ␣A and ␣B crystallin, and membrane-bound actin (Deretic and Papermaster, 1991;Deretic et al., 1994. The number of membraneassociated and membrane proteins of the vesicles is significantly reduced compared with the Golgi-enriched fraction, whereas certain proteins like ␣-transducin and ␣A and ␣B crystallin are enriched in the vesicle fraction. Taken together these data indicate that selective budding of the vesicles results in retention of the TGN proteins and in vitro sorting of post-Golgi vesicle proteins away from the TGN.

In Vitro Formed Vesicles Are Morphologically Identical to the Vesicles Formed in Vivo and Have the Same Topology-Mor-
phological analysis of the vesicles formed in a cell-free assay by electron microscopy shows profiles indistinguishable from the vesicles formed in vivo (Fig. 7A). We have previously shown that fraction 5 contains small vesicles (ϳ300 nm diameter) without a visible coat and very little contamination from other morphologically distinguishable subcellular organelles (Deretic and Papermaster, 1991). This is also true for the vesicles formed in vitro, as shown in Fig. 7A. By contrast, the Golgienriched fraction 12 is very heterogeneous (Fig. 7B), but occasional profiles that may represent cross-sections of Golgi cisternae can be seen after in vitro vesicle budding.
Rhodopsin-bearing post-Golgi vesicles formed in vivo in retinal cultures contain rhodopsin in the membrane with its carboxyl-terminal domain exposed to the cytoplasm and aminoterminal domain inside the vesicle (Deretic and Papermaster, 1991). We have probed the orientation of rhodopsin in the vesicles that are formed in vitro at 22 or 0°C using mAb 11D5 to the carboxyl-terminal domain of rhodopsin and found that they contain rhodopsin inserted in the membrane with this domain exposed at the cytoplasmic side and susceptible to limited proteolysis with thermolysin (Fig. 8). This suggests that in vitro formed vesicles are topologicaly identical to the vesicles formed in vivo.
rab Proteins Are Essential Regulators of Rhodopsin Transport-To test the role of rab proteins in the transport of newly synthesized rhodopsin, we removed the membrane-associated rab proteins by incubating retinal PNS with purified rab GDI. This cytosolic protein regulates rab protein function by preventing GDP dissociation and subsequent nucleotide exchange onto rab proteins (Sasaki et al., 1990). We used recombinant bovine His 6 -tagged rab GDI, which has the ability to interact with wide range of rab proteins and extract GDP-bound rab proteins from the membrane (Ullrich et al., 1993(Ullrich et al., , 1994. We preincubated pulse-labeled retinal PNS with 200 M GDP and 2 M recombinant rab GDI and followed the appearance of newly synthesized rhodopsin in the post-Golgi vesicle fractions 4 -6 after the addition of ATP. Fig. 9A shows that under these FIG. 6. Two-dimensional gel pattern of post-Golgi vesicle proteins reveals significant reduction in complexity of these membranes (B) compared with the Golgi-enriched fraction (A). Silver-stained proteins separated by two-dimensional gel electrophoresis of vesicle membranes reveal several previously identified vesicle proteins: the ␣-subunit of transducin (arrow), ␣A and ␣B crystallin (open and closed arrowheads, respectively), and membrane-bound actin (asterisk), as well as a number of unknown proteins. The identity of the indicated proteins has been confirmed by immunoblotting. The ␤ subunit of transducin, detectable by immunoblotting, is not easily detectable here, probably due to its low staining with silver (Hamm and Bownds, 1986). The two proteins at the basic end of the gel in B, indicated by the notches on the right side, are 97-kDa and 30-kDa molecular mass standards.
FIG. 7. Post-Golgi vesicles formed in the cell-free system are morphologically identical to the vesicles formed in vivo and differ from Golgi-enriched fraction. Electron micrographs of the thin sections through the pellet obtained after in vitro vesicle formation, sedimentation of fraction 5 or 12 membranes, and embedding in Epon are shown. A, a population of small vesicles is a major component of fraction 5. They are of the same size (ϳ300 nm) and appearance as vesicles formed in vivo, which we have previously characterized. B, fraction 12 is very heterogeneous. Occasional profiles that may represent cross-sections of the Golgi cisternae can be seen (open arrow). In addition this fraction contains mitochondria, smooth membranes, and multivesicular bodies. Bar, 0.3 m.
conditions the amount of membrane bound rab6 and rab8 in the vesicle fraction is reduced to Ͻ40% of the original content. The amount of soluble GTP-binding proteins, and rab6 in particular, increases 3-fold (Fig. 9B). Extraction of membranebound rab proteins with 2 M rab GDI results in the accumulation of more slowly migrating (untrimmed) forms of rhodopsin in the Golgi (fractions 11 and 12) and complete inhibition of post-Golgi vesicle formation (Fig. 9C). This GDI effect is concentration-dependent (IC 50 ϭ 0.3 M), and the reduction in the radiolabeled rhodopsin content in vesicle fractions 4 -6 is accompanied by an equivalent increase in Golgi fractions 11 and 12 (Fig. 10).
The Cytoplasmic Domain of Rhodopsin Is Involved in Vesicle Budding from the TGN-Finally, we wanted to test if the cytoplasmic domain of rhodopsin plays a role in the budding of post-Golgi vesicles from the TGN and in intracellular sorting. We preincubated pulse-labeled retinal PNS for 30 min in the presence of 100 g of anti-rhodopsin COOH-terminal antibody mAb 11D5, previously used to immunoisolate rhodopsin-bearing post-Golgi membranes (Deretic and Papermaster, 1991). Binding of mAb 11D5 to the cytoplasmic domain of rhodopsin results in the profound (Ͼ85%) inhibition of vesicle formation during the cell-free chase, whereas control antibody of the same IgG subclass has no effect (Fig. 11). Completely processed radiolabeled rhodopsin accumulates in the TGN-enriched fractions 7-10. Golgi-enriched fractions 11 and 12 have a similar content of newly synthesized rhodopsin as the control. The distribution of other vesicle-associated radiolabeled proteins, as exemplified by 45-kDa protein, resembles the distribution of rhodopsin. Anti-rhodopsin antibody arrests post-Golgi vesicle formation to nearly the same extent as removal of ATP (Figs. 1 and 4) or rab proteins (Fig. 10).
To test if potential antibody cross-linking changes the sedimentation of the vesicles rather than inhibits their budding, we also tested the mAb 11D5 Fab fragments. 11D5 Fab fragments inhibit vesicle formation in the same fashion as the whole antibody (Fig. 11B), whereas control Fab fragments have no effect (data not shown). Because Fab fragments can not cause membrane cross-linking, this implies that mAb 11D5 inhibits rhodopsin transport. The presence of the antibody in the cellfree assay does not appear to interfere with the intra-Golgi transport because radiolabeled rhodopsin is found in lighter fractions (7-10) compared with its arrest in fractions 11 and 12 by GDI. We conclude that mAb 11D5 exhibits its inhibitory effect on the vesicle formation from the TGN. This suggests that the intracellular sorting machinery recognizes the cytoplasmic domain of rhodopsin upon its appearance in the TGN, where it is sorted to the appropriate vesicular carriers.

DISCUSSION
Establishment of a cell-free system that reconstitutes rhodopsin-bearing post-Golgi vesicle formation represents the first step toward understanding the molecular mechanisms of polarized sorting in rod retinal photoreceptor cells. Studies of intracellular trafficking of newly synthesized rhodopsin may also provide a more general insight, because rhodopsin is a member of the large family of seven-helix transmembrane receptors. We find that rhodopsin-bearing post-Golgi vesicles can form in the frog retinal cell-free system under physiological conditions and that vesicle budding is accompanied by the retention of resident membrane proteins within the TGN. Similar conditions for cell-free post-Golgi vesicle formation in other cells have been reported previously Salamero et al., 1990;Jones et al., 1993). However, in contrast to the cells or cell-lines that originate from the warm blooded animals, frog-derived tissues also appear to be able to maintain membrane transport at very low temperatures.
Sustained intracellular transport at 0°C may be due to the preservation of membrane fluidity at low temperatures in frog FIG. 8. Topological analysis of in vitro formed post-Golgi vesicles probed with thermolysin and an mAb 11D5 to the carboxylterminal cytoplasmic domain of rhodopsin. Blots of the untreated and thermolysin-treated vesicles were probed with the mAb 11D5, whose antigenic site is within nine amino acids from the COOH-terminus of rhodopsin (Deretic and Papermaster, 1991). Proteolytic cleavage of the carboxyl-terminal domain of rhodopsin by thermolysin destroys the mAb 11D5 binding site on the vesicles formed in the cell-free assay at 22 or 0°C.
FIG. 9. Extraction of rab proteins by rab GDI arrests unprocessed rhodopsin in the Golgi. To remove membrane-bound rab proteins, radiolabeled retinal PNS was preincubated with 200 M GDP for 1 h at 22°C, and then 2 M GDI was added. the preincubation was continued for 30 min at 30°C, followed by 2 h of cell-free chase at 22°C. A, [ 32 P]GTP overlays of the post-Golgi vesicle membrane proteins separated by two-dimensional gels show Ͼ60% reduction in membrane associated rab6 and rab8 (open and closed arrowheads, respectively). The more basic of the two ubiquitous unidentified proteins that comigrates with rab11 (Urbé et al., 1993) remains unchanged (arrow). B, [ 32 P]GTP overlays of soluble proteins separated by SDS-PAGE show a significant increase of small G-proteins in the cytosol after GDI extraction. Immunoblotting with anti-rab6 antibody and quantitative analysis of the blots developed using the ECL system reveals a 3-fold increase in soluble rab6. C, removal of rab proteins by GDI inhibits vesicle formation (fractions 4 -6) and arrests slowly migrating rhodopsin in the Golgi (fractions 11 and 12). Subcellular fractions were pooled according to the kinetics of their acquisition of newly synthesized rhodopsin, and membrane proteins were separated by SDS-PAGE. Autoradiograms of the gels are shown. R, rhodopsin. The arrows indicate rhodopsin dimers and trimers. The gels were exposed for a short period so that the autoradiograms reveal only rhodopsin and its multimers as described previously (Deretic and Papermaster, 1991). photoreceptors. We have recently found that the polyunsaturated docosahexaenoic acid, a major contributor to the exceptionally high ROS membrane fluidity, incorporates into transport membrane phospholipids at the site of vesicle formation at 22°C. 2 It will be of interest to determine the temperature dependence of this process to define the role of lipid remodeling in intracellular transport and the post-Golgi vesicle budding.
A potential chaperone activity of ␣A and ␣B-crystallins may also contribute to the stability of post-Golgi vesicle formation at low temperature. These heat shock proteins associate uniquely with rhodopsin-bearing vesicles and may provide protection against temperature-induced stress (Deretic et al., 1994).
The presence of excess (2 M) rab GDI in the retinal cell-free assay probably inhibits intra-Golgi transport by depleting a rab protein pool present on the Golgi membranes and inhibiting recycling and recruitment of soluble rabs. Similar effects of rab GDI on the intra-Golgi transport in other cells have been reported (Elazar et al., 1994;Peter et al., 1994). The pattern and the distribution of newly synthesized rhodopsin upon removal of rab proteins in retinal cells greatly resembles the effect of short (30 min) exposure to brefeldin A which results in the disruption of the Golgi and inhibition of rhodopsin processing (Deretic and Papermaster, 1991). The similarity between GDIand brefeldin A-induced arrest of transport suggests that removal of rab proteins interferes with the glycoprotein processing in the Golgi (and, by inference, possibly its structure as well), which ultimately leads to reduced vesicle formation. rab6 is extracted from the photoreceptor membranes upon treatment with GDI (Fig. 9). Because rab6, as well as rab1, appears to be critical for the integrity of the Golgi complex (Martinez et al., 1994;Peter et al., 1994), their collective removal from the photoreceptor Golgi by rab GDI may contribute to the arrest of transport there. Additional experiments will be necessary to determine the role of rab6 and other post-Golgi vesicle-associated rab proteins in the budding and targeting of rhodopsinbearing post-Golgi vesicles.
All rhodopsin-bearing vesicle-associated rab proteins are properly sorted to the post-Golgi vesicles formed in vitro. How-ever, the amount of rab8 detectable by GTP-overlays appears to be diminished compared with the vesicles formed in vivo. rab6 remains unchanged. A possible explanation for this finding is that the cell-free assay reconstitutes vesicle budding but may not reconstitute vesicle targeting and fusion. In this case the difference between the two rab proteins would be consistent with our previous conclusion that rab8 acts at the later step of rhodopsin transport than rab6 .
The principal unique finding presented in this study is the important role of rhodopsin in photoreceptor post-Golgi vesicle formation. In the cell-free assay, mAb 11D5 to the COOHterminal cytoplasmic domain of rhodopsin and its Fab fragments appear to block transport of rhodopsin and its associated post-Golgi vesicle proteins at the exit from the TGN. This finding is consistent with our previous proposal that rhodopsin FIG. 10. rab GDI affects rhodopsin transport in a concentration-dependent manner. Retinal PNS was preincubated with 200 M GDP for 1 h at 22°C followed by 30 min at 30°C with increasing concentrations of GDI. After SDS-PAGE the distribution of radiolabeled rhodopsin was determined by a PhosphorImager densitometer. The experiment was repeated three times. The GDI-induced decrease in radiolabeled rhodopsin content of the vesicular pool in fractions 4 -6 (IC 50 ϭ 0.3 M GDI) is accompanied by an equivalent increase in the Golgi-enriched fractions 11 and 12.
FIG. 11. A monoclonal antibody 11D5 to the carboxyl-terminal cytoplasmic domain of frog rhodopsin inhibits post-Golgi vesicle formation and arrests rhodopsin in the TGN. A, radiolabeled retinal PNS was preincubated with 100 g of mAb 11D5 or control antibody for 30 min at 0°C before the 2-h cell-free chase. Subcellular fractions were pooled as described in the legend to Fig. 9. Autoradiograms of the SDS-PAGE are shown. Although mAb 11D5 blocks vesicle formation, control antibody has no effect. R, rhodopsin (ϳ35 kDa). The arrow indicates rhodopsin dimer. A 45-kDa radiolabeled protein visible in this autoradiogram and other less abundant (not shown) vesicleassociated proteins are also arrested in fractions 7-10 by the action of anti-rhodopsin antibody. B, the distribution of radiolabeled rhodopsin was measured by the PhosphorImager densitometer in four separate experiments. The data are presented as the means Ϯ S.E. In the control (with ATP at 22°C) in addition to the vesicles formed during the pulse, ϳ16% of total radiolabeled rhodopsin accumulates in the vesicle fraction during the cell-free chase. mAb 11D5 and its Fab fragments diminish this additional vesicle formation to ϳ2%. Therefore, vesicle formation during the chase in the presence of mAb 11D5 is inhibited Ͼ85%. Concurrently, fractions 7-10 enriched in the TGN membranes (see Fig.  3) accumulate radiolabeled rhodopsin. Two additional antibodies to other photoreceptor proteins and the Fab fragments from the control antibody have no effect on vesicle formation (not shown). not only regulates its polarized sorting but also the sorting of ROS proteins transducin and cGMP phosphodiesterase, its partners in signal transduction (Deretic and Papermaster, 1991). Moreover, additional evidence for the important regulatory role of rhodopsin has been provided by Roof et al. (1994), who showed that in transgenic mice mis-sorting of mutant rhodopsin also leads to delocalization of other nonmutant ROS proteins.
Sorting into specific post-Golgi vesicles in photoreceptors may be accompanied by the binding of the cytosolic coat proteins that are presumed to be essential to deform the membrane and drive the budding process. This coat could be assembled at or near the cytoplasmic surface of rhodopsin, or perhaps rhodopsin molecules may bind integral TGN membrane protein(s) responsible for the recruitment of coat proteins and/or regulation of sorting to the appropriate vesicular carriers. In either case, antibody bound to the cytoplasmic surface of rhodopsin could prevent coat binding to its membrane receptor(s). Preliminary experiments show that protein p200, a candidate for a coat protein that regulates vesicle budding from the TGN (Narula et al., 1992;Ladinsky et al., 1994) associates with photoreceptor TGN membranes but not the post-Golgi vesicles and could participate in vesicle budding. 3 It is possible that some of the coat proteins are photoreceptor-specific. This is consistent with reports of nonpolarized expression of rhodopsin in polarized kidney-derived cell lines, which suggest that these cells lack appropriate machinery to sort rhodopsin to one domain of the plasma membrane (Oprian et al., 1987;Nathans et al., 1989). Although we find that proteins that regulate transport to the basolateral plasma membrane in epithelial cells, such as rab8, may be involved in rhodopsin transport in rods, , it is possible that photoreceptors use additional molecular information to sort rhodopsin.
It is difficult to assess which part of the cytoplasmic domain of rhodopsin is responsible for the sorting function because bound mAb 11D5 or its Fab fragments probably render the entire surface unavailable for interactions with other proteins. However, it is possible that the extreme COOH-terminal that binds the antibody also contains the sorting signal. The five COOH-terminal amino acids QVA(S)PA are highly conserved among different species. This domain of rhodopsin is in close proximity to the domains that participate in light-dependent G-protein activation or subsequent phosphorylation and inactivation by rhodopsin kinase; however; it has not been implicated in any light-dependent function (Weiss et al., 1994;Chen et al., 1995;Sung et al., 1994;Hargrave and McDowell, 1992). Mutations in the COOH-terminal domain severely impair rhodopsin function and lead to rapid retinal degenerations and blindness in patients with autosomal dominant retinitis pigmentosa (Nathans, 1994). Recently, Sung et al. (1994) have shown that transgenic mice expressing mouse rhodopsin that lacks five COOH-terminal amino acids accumulate this but not the endogenous rhodopsin in the lateral photoreceptor plasma membrane, which supports the role of the COOH-terminal in sorting to or retention in the ROS.
Our data indicate that the cytoplasmic domain of rhodopsin interacts with the intracellular sorting machinery at the exit from the TGN. We are currently attempting to map the region of rhodopsin responsible for this interaction using synthetic peptides derived from the cytoplasmic domain of bovine and frog rhodopsin. 4 Our future efforts will also be focused on identification of the membrane and/or coat proteins that inter-act with rhodopsin upon exit from the TGN. These factors should further define the mechanism of development of photoreceptor polarity, which may involve the modification of the sorting machinery common to other polarized cells, or the development of additional sorting stations to fulfill the maintenance of three highly polarized domains in this specialized neuron.