key: cord-022235-6ircruag
authors: Pugsley, Anthony P.
title: Later stages in the eukaryotic secretory pathway
date: 2012-12-02
journal: Protein Targeting
DOI: 10.1016/b978-0-12-566770-8.50009-4
sha: 
doc_id: 22235
cord_uid: 6ircruag

nan

The secretory pathway of eukaryotic cells comprises a succession of compartments, the secretory organelles, through which proteins pass en route to their final destinations. Although each secretory organelle has its own special characteristics, a number of basic features are common to all of them. Some proteins pass more or less unimpeded through the entire length of the secretory pathway, whereas others are retained in secretory organelles. This raises two of the fundamental questions which will be addressed in this chapter, namely, what determines whether a protein will be taken out of circulation at a particular step in the pathway, and how tight is the separation between secretory organelles?

Ultrastructure and cell fractionation studies indicate that there is little or no physical link between the RER and the Golgi. Furthermore, specific proteins are known to reside in individual compartments of the Golgi apparatus. As we shall see, the physical separation of Golgi enzymes is also indicated by the succession of posttranslational modifications to which secretory proteins are subjected, by in situ immunocytochemistry, and by the separation of Golgi-derived vesicles containing different en zymes. How then do secretory proteins move between these compart ments, and is the secretory pathway a continuous gradient of secretory organelles or are they functionally and structurally independent?

A feature common to all stages of the secretory route beyond the RER is that proteins move between secretory organelles in specific classes of transport vesicles. Consequently, soluble secretory proteins never come into direct contact with the outer face of the organelle to which they are being targeted and therefore can play no direct part in sorting. Integral membrane proteins usually have segments exposed on the cytoplasmic face of the transport vesicle which could be recognized by receptors on the target organelle. Microinjected antibodies recognizing the C-terminal, cytoplasmic tails of plasma membrane proteins can prevent their trans port to the cell surface (25, 590) , but this is probably due to antibodyinduced changes in protein conformation which make the protein incom petent for transport along the secretory pathway, rather than to inhibition of receptor-secretory protein interactions.

In this chapter, we will consider three different ways in which sorting of secretory proteins might occur:

(i) All secretory proteins have signals which target them successively through the secretory organelles and then on to their final target; some proteins remain in different organelles because they lack the signal necessary for targeting to the next organelle in the pathway. (ii) Secretory proteins only have sorting signals for the last stage in the secretory pathway; some proteins have retention signals which pre vent them joining the bulk flow through the secretory pathway, thereby causing them to be retained in specific organelles. (iii) As (ii), except that secretory organelle proteins pass through the secretory pathway in the bulk flow and are recycled via signaldependent counterflow.

If either of the last two models are correct, then molecules devoid of retention or sorting signals should travel through the secretory pathway as part of the bulk flow. The rate at which molecules are transported out of the cell will thus be determined by the rate at which they diffuse to sites within the RER and Golgi where transport vesicles are formed and depart en route to the next compartment in the chain. This flow rate has recently been measured with the N-glycosylation acceptor tripeptide Asn-Tyr-Thr (Section V.B.I). If this tripeptide contains a radioiodinated Tyr residue, its progress through the secretory pathway, which it seems to enter by diffusing across the RERM, can be followed by chromatography and autoradiography. Wieland et al. (1190) found that the tripeptide was gly cosylated in the ER (this prevented it from leaking back into the cytosol), then trimmed by mannosidases located in the Golgi (Section V.D), and finally secreted into the medium. Tripeptide was detected in the medium after 5-10 min, depending on the cell type, which is considerably faster than the time required for the transport of most secretory proteins through the secretory pathway. Thus, bulk flow is very efficient, implying that diffusion through the lumen of the ER and Golgi can occur relatively easily and that there is massive, vectorial movement of vesicles between the organelles and the cell surface. The way in which this result influences our understanding of protein sorting in the secretory pathway is discussed in following sections.

Soluble and integral membrane secretory proteins fold once they have crossed the RERM; they are also often covalently modified. The follow ing sections deal with the types of modifications which occur in the ER and their role, if any, in protein sorting. Palmitoylation and phosphatidyl inositol modification of secretory proteins are discussed in Section III.F.4.C; only their role in protein transit through the secretory route will be discussed here.

Most secretory proteins are nitrogen (N)-glycosylated in the RER. Be cause we are primarily concerned with the effects of glycosylation on protein targeting, only general details of the glycosylation reactions will be given here.

a.

The sequence of reactions shown in Fig. V .l, which is common to both yeasts and higher eukaryotes, results in the addition of a (Glucose) 3 -(Mannose) 9 -( J /V-acetylglucosamine) 2 complex onto Asn residues in the sequence Asn-X-Ser or Thr (the acceptor peptide, in which X can be any I  I  I  I amino acid except possibly proline or aspartate). Isolated tripeptides can act as acceptors only when the two extremities are blocked. Longer ac ceptor peptides are more rapidly glycosylated (577). As shown in Fig.  V .l, the entire complex is assembled before it is attached to the Asn residue in the lumen of the RER. At least some sugars cross the RERM as intermediates complexed with the long-chain lipid dolichol, whereas others may cross the RERM as nucleotide 5 '-diphosphate-sugar complexes. However, it is by no means certain that the complex is assembled entirely within the lumen of the RER. For example, there is no evidence that GDP-mannose can be trans ported across the RERM, leading to the proposal that the five mannose residues are added via GDP-mannose donors on the cytoplasmic face of the RERM, whereas later modifications, and possibly earlier modifica tions, occur in the lumen. This presupposes that the dolichol-PP-(NAGn) 2 complex can "flip" from the lumen to the cytoplasmic face and then back again once the five mannose residues have been added [see (452) for further discussion of the topology of glycosylation reactions in the RER]. The preferred donor lipid-carbohydrate complex is the com plete complex shown in Fig. V.l (1129) . However, truncated versions lacking glucose and even mannose residues can act as donors in vitro, and under-glycosylated complexes can act as donors in vivo in yeasts (598) and protozoa (577) . Variations in the mannose content of transferred oligosaccharides have also been reported (577).

The transfer of the first N-acetylglucosamine (NAGn) residue from the UDP complex to the dolichol, is inhibited by the antibiotic tunicamycin, which therefore blocks all N-glycosylation. This provides a valuable tech nique for studying the role of N-glycosylation in protein traffic. The yeast (S. cerevisiae) gene (ALG7) coding for the RERM-associated, tunicamycin-sensitive enzyme (UDP-NAGn : dolichol-P-transferase) was cloned by virtue of its ability to rescue tunicamycin-treated cells when present on multiple copy number plasmids (931). Null mutations in ALG7 are lethal. Mutants affected in other stages in the yeast glycosylation pathway have been isolated by mannose suicide selection (Section II.F.l.b). Only muta tions blocking the earliest stages in the pathway are lethal; incomplete dolichol-linked oligosaccharides containing a minimum of four mannose residues allow normal growth, presumably because they can be attached to secretory proteins if the full-length lipid-oligosaccharide complex is absent (see above) (598).

The complete oligosaccharide chain is transferred onto the Asn accep tor site in the lumen of the RER. N-glycosylation is generally thought to occur as the polypeptide is being threaded through the RERM, while it is still in its unfolded state. However, some acceptor sites are not glyco-sylated. This may be because they rapidly become inaccessible within the structure of the folded polypeptide, although other explanations are possi ble (577, 968) . An unexplained anomaly is that some ER membrane glyco proteins have glycosylated residues exposed on the cytoplasmic face of the ER (1). One explanation could be that additional NAGn transferases are present on the cytoplasmic face of the smooth ER, where these glyco proteins are predominantly located.

The initially homogeneous oligosaccharide is processed immediately fol lowing its attachment to the polypeptide chain, initially by the removal of glucose residues by one or more glucosidases. Glucose trimming of glyco sylated vesicular stomatitis virus (VSV) G protein has been reported to occur cotranslationally (27). Further processing steps are different in yeasts and in animal cells. In S. cerevisiae cells, a single al-2-linked mannose residue is replaced by αϊ-3 mannose. A mutation in the GLS1 gene coding for α 1-2 glucosidase does not affect removal of the mannose residue or subsequent chain elongation, whereas removal of the ninth al-2-linked mannose residue may be essential for outer chain elongation, which occurs in the Golgi (Section V.D.I) (598). Further processing of animal cell glycoproteins also occurs in the Golgi, although further man nose trimming of resident ER proteins and reglycosylation may both oc cur in the ER (577,847).

c.

Ser and Thr residues on secretory proteins of yeasts and possibly other fungi can be O-glycosylated in the ER (429). The process is less well characterized than N-glycosylation but seems to involve the direct trans fer of mannose residues from dolichol-1-mannose onto acceptor amino acids. Studies with acceptor oligopeptides suggest that no particular se quence is required around the modification site (54,621). Further process ing of the mannose residue occurs in the Golgi apparatus (429) (Section V.D.2). Yeast secretory proteins may be both N-and O-glycosylated.

Most studies indicate that O-glycosylation occurs exclusively in the Golgi apparatus (Section V.D.2).

Secretory proteins apparently fold spontaneously as they are extruded .across the RER membrane (194, 1222) . However, disulfide bridge forma tion is probably catalyzed by protein disulfide isomerase (PDI), which is loosely associated with the RERM (325,326). The formation of soluble protein complexes occurs shortly after synthesis (604), but trimerization of VSV G protein and influenza virus hemagglutinin (HA), both of which are integral membrane proteins, occurs only 7-10 min after synthesis (193, 194, 364, 589) and involves the selection of polypeptides from a ran dom pool of prefolded monomers rather than from a restricted pool of monomers synthesized on a single polysome (108). Kreis and Lodish (589) suggest that this delay may be due to the segregation of PDI in a late "compartment" of the ER. According to the papers discussed above, oligomerization of G and HA occurs just before they leave the ER, 5-10 min after synthesis. This view is challenged, however, by Yewdell et al. (1222) , who found that mono clonal antibodies specific for oligomerized HA reacted only with proteins in the Golgi apparatus, and not with HA in the ER (see next section). A possible explanation for this ambiguity is that HA monomers fold and trimerize in the ER and are then transported to the Golgi apparatus, where further modifications alter the conformation of the HA trimers to produce the antigenic sites recognized by the antibody used by Yewdell et al.

Different proteins transit through the secretory pathway together, but the rates at which they are secreted may vary considerably (335, 343, 615, 1221) . The site at which the secretion lag is most prominent seems to be transit from the RER to the Golgi (646). It has been suggested that this delay might reflect the need for secretory proteins to interact with specific receptors in the RER for transport to the Golgi and that carbohydrates might form part of the recognition signal (320,647,648). Alternatively, proteins may be retained in the ER until they have folded correctly; different folding kinetics may result in different retention times in the ER (589,1192). Indeed, exit from the ER of one of the proteins studied by Fitting and Kabat (320) coincided with a partial proteolysis event, and polypeptides were "selected" for exit from a random pool of "new" and "old" proteins. Furthermore, studies discussed in the preced ing section showed that oligomerization of virus-encoded membrane pro teins occurs just before they are transported to the Golgi.

If secretory proteins are indeed retained in the ER until they are folded and oligomerized in the correct way, then exit-incompetent proteins should either remain indefinitely in the ER or be secreted at very much reduced rates. This could explain why genetically altered or hybrid pro teins are sometimes translocated into the RER without difficulty and yet do not transit further through the secretory pathway (104, 281, 364) , and why incomplete glycosylation or glucose trimming and failure to process signal peptides sometimes affect secretion kinetics (96, 577, 649, 1045) . Even minor sequence changes have been reported to affect protein con formation and exit from the ER (1192).

What happens to incorrectly folded proteins in the ER? Results from numerous studies indicate that they stay in the ER (604,681) or are de graded either in the ER (642a) or in the lysosome (723). These proteins do not seem to precipitate in large aggregates. Instead, a specific, major ER protein seems to bind to some incorrectly folded proteins, thereby pre venting their exit from the ER. Haas and Wabl (413) first detected this protein (BiP) complexed with immunoglobulin heavy chains synthesized in the absence of light chains (heavy and light chains are only transported to the Golgi as complexes) (97), and it was subsequently found to be complexed with incorrectly oligomerized or monomeric HA (194, 364) , nonglycosylated invertase (in vitro in dog pancreatic microsomes) and incorrectly folded prolactin (557), and recombinant human factor VII in Chinese hamster ovary cells (561). It was not detected in association with aggregated VSV G protein (250) or with incorrectly folded (mutant) class I histocompatibility antigen (1192). The dissociation of secretory protein-BiP complexes requires ATP (758), which might explain part of the ATP requirement for protein movement along the secretory pathway, but be cause ATP is usually present in cell ly sates, some BiP complexes (such as G-BiP) may dissociate during extraction. BiP has also been shown to associate with nascent polypeptides as they enter the lumen of the ER (557).

Two interesting features of BiP are that its synthesis is stimulated by glucose deprivation and that it is structurally related to a heat shock protein (758,857). Glucose starvation is likely to reduce glycosylation, thereby increasing the proportion of incorrectly folded secretory proteins in the ER. Studies have confirmed the idea that the extent of glycosylation can affect the association between secretory proteins and BiP, as well as their rate of secretion (251). The accumulation of misfolded (mutant) pro teins in the ER also increases BiP synthesis (583), possibly because the cell senses that its store of BiP has been sequestered into protein com plexes. Thus, synthesis of BiP may be increased according to require ments (857), but studies on the "proofreading" or quality control role of BiP are still at an early stage. A different, cytoplasmic quality control protein may be responsible for proofreading cytoplasmic domains in transmembrane proteins to ensure that they too are correctly folded and oligomerized (406a).

Protein conformation, rather than any specific structural feature such as glycosylated residues, is thus likely to determine whether a protein is competent to leave the ER. This feature is illustrated by studies on the effects of glycosylation on VSV G protein. Machamer et al. (672) used site-directed mutagenesis to replace the glycosylated Asn residues. Nonglycosylated G protein was transported from the ER to the Golgi but did not reach the cell surface. When only one of the Asn residues was de leted, the protein reached the cell surface; thus G protein with one glyco sylated Asn residue can be transported through the entire secretory path way. G protein export became temperature-sensitive when new glycosylation sites were added (671). Kotwal et al. (581) noted that some natural G protein variants have only one oligosaccharide whereas others have none at all and suggested that compensatory changes in primary structure may allow nonglycosylated protein to fold into a secretion-com petent conformation. Thus, glycosylation is probably needed for correct protein folding but is not directly implicated in the formation of secretion competence signals. At least one secreted protein, ovalbumin, is not gly cosylated.

The ER contains a large number of proteins, some of which it shares with the contiguous nuclear membrane (117). They include proteins involved in secretory protein translocation through the RERM, signal peptidase, ribophorins, BiP, enzymes involved in lipid synthesis and in protein glycosylation, and cytochrome P 4 5 0 . The characterized proteins are similar to other secretory proteins: They often have signal peptides (198, 630, 1089) , are glycosylated, and may be located in the lumen or in the RERM.

Rothman (965) has argued that it may be physically impossible for the ER to prevent the escape of endogenous proteins, and particularly mem brane proteins, to the cis Golgi. He cited two observations in support of the idea that these proteins could leave the RER as part of the bulk flow through the secretory pathway, and then recycle from the Golgi to the ER:

(i) Some ER proteins are detected in significant amounts in vesicles derived from the cis Golgi when cells are fractionated (102). (ii) Bulk lipid flow out of the ER necessitates efficient recycling, possi bly from the cis Golgi, which could provide "carriers" for the recy cling of ER proteins.

Most ER proteins are, however, almost completely excluded from the Golgi (663,697). Furthermore, ER proteins are not terminally glycosyla ted (117,1089), which means that they do not reach the medial or trans Golgi (Section V.E). Mannose residues on some ER proteins are not trimmed beyond the stage catalyzed by ER mannosidase (117,954,1089), but a lysosomal protein carrying an ER retention signal (see below) is phosphorylated by NAGn phosphotransferase, indicating that it reaches the early cis Golgi (858). Significantly, the phosphate groups are not modi fied by NAGn phosphodiesterase, which may be located in a different, ER-distal Golgi compartment (Table V.l). Warren (1166) has proposed that ER proteins are salvaged from an intermediate compartment, the socalled transitional element (Section V.C), located between the ER and the cis Golgi.

Recycling of "escaped" endogenous ER proteins, rather than receptormediated retention in the ER, is the currently favored model for the specific accumulation of proteins in the ER (151). It is not clear whether soluble and ER membrane proteins are both subject to the same retention mechanism, but studies on the rates of diffusion of membrane proteins in the ER indicate that RERM proteins are more restricted than BiP and that the mobility of BiP devoid of its ER retention-salvage signal (see below) cannot be distinguished from that of normal BiP (151).

Several studies have sought to determine the nature of the ER reten tion-salvage signal(s). The rotaviral type II ER membrane protein VP7 was found to be secreted when the two potential N-terminal transmem brane segments were deleted (884), suggesting that retention-salvage de pended on a membrane anchor domain. However, subsequent studies showed that the entire N-terminus was absent from mature VP7 (1084) and thus that some other feature must account for VP7 retention in the ER (883c). Although deletion of the C-terminal transmembrane and cyto plasmic tails of the type I ER membrane protein Ε19 of adenovirus also causes the protein to be secreted, deletion of the last eight residues (FIDEKKMP) of the C-terminal, cytoplasmic tail alone causes the protein to appear on the cell surface, suggesting that the C-terminus contains the ER retention-salvage signal. Furthermore, cell-surface interleukin 2 re ceptor protein β chain was converted into a resident ER protein following fusion of FIDEKKMP to its cytoplasmic C-terminus (832). The cytoplas- mically exposed signal could be recognized by "salvage receptors" in the transitional element. These studies on the Ε19 protein are reminiscent of earlier, seminal work on the soluble ER protein BiP which resulted from the observation that three soluble proteins in the lumen of the ER, including BiP, had the same four residues, Lys-Asp-Glu-Leu (KDEL), at their C-termini. Munro and Pelham (759) found that if this sequence was deleted or extended, BiP was secreted into the medium. This suggests that KDEL is the ER reten-tion signal and that it must be at the extreme C-terminus, which is pre sumably the last segment of the polypeptide to fold and may therefore be exposed on the surface of the protein. In a complementary study, DNA coding for KDEL was fused to the end of a cDNA clone coding for secreted lysozyme. The resulting hybrid protein was retained in a perinu clear region probably corresponding to the ER (759).

Further studies are required to determine whether KDEL is the "uni versal" ER retention signal. Preliminary studies suggest that some mam malian ER proteins have the sequence RDEL at the C-terminus, whereas yeast ER proteins have the sequence HDEL (858a). HDEL or KDEL is presumably recognized either by an endogenous ER membrane protein, which could anchor soluble proteins in the ER, or, more likely, by the recycling receptor located in the transition element or cis Golgi. Indeed, a putative HDEL receptor gene (ERD1) has recently been identified in yeast cells (858,858a). It is not clear what triggers the release of BiP from its receptor once it is recycled to the ER.

β-Glucanase is a typical lysosomal protein; small but significant amounts of it, however, are retained in the ER through a specific interac tion with an endogeneous ER protein, the esterase egasyn. Studies by Medda et al. (704) indicate that /3-glucanase-egasyn interaction is blocked by inhibitors of esterase activity, leading to β-glucanase secretion (rather than sorting to the lysosome). Whether this is a physiologically significant mechanism for retaining proteins in the ER remains to be de termined. Only about 10% of egasyn is complexed with /3-glucanase (666), so perhaps it also binds to other resident ER proteins. It would be inter esting to see whether egasyn itself has a C-terminal KDEL-like sequence.

As discussed above, secretory proteins do not move to the cis Golgi until they attain an exit-competent state, and some proteins are specifically retained in the ER, probably by recycling of escaped proteins. Both nor mal secretory proteins (1223) and exit-incompetent proteins have been reported to accumulate in a specific region of the smooth ER, the vacuolar transitional element (79,996), from which vesicles either migrate to and fuse with the cis Golgi (41) or coalesce to form new Golgi cisternae (797). Specific membrane proteins [e.g., the product of the SEC 12 gene in yeasts (767a)] may be required to form this specialized domain of the ER and hence directly or indirectly assist vesicle formation and protein transport to the Golgi.

Secretory pathway "shuttle" vesicles are difficult to isolate because (533) also showed that ATP depletion caused secretory pro teins to accumulate in transitional elements, but this study did not distin guish between energy requirements for protein folding and for vesicular transport from the transitional element to the Golgi. A different in vitro assay was developed by Haselbeck and Schekman (428) to study protein movement from the ER to the Golgi in yeast cell extracts. Their assay uses donor ER vesicles derived from a strain carry ing the secl8ts mutation, which blocks protein movement from the ER, and the mnnl mutation, which prevents terminal (Golgi) mannosylation of secretory proteins. Invertase accumulated in the ER grown at the restric tive temperature was mannosylated after transfer to recipient Golgi from wild-type cells. The transfer efficiency was low, however, possibly be cause invertase accumulated at the restrictive temperature did not be come exit-competent at the permissive temperature in vitro, or because recipient Golgi were saturated. As in the mammalian system, protein transfer was ATP-dependent and required soluble cofactors including SEC 18 protein (269b) as well as proteins on the surface of the recipient Golgi. A similar but more efficient reconstitution system using gently lysed yeast-cells has recently been developed (35a). Results obtained with this system indicate the probable requirement for GTP in ER to Golgi traffic.

Secretory proteins are subjected to further chemical modification as they transit through the Golgi. These processes are discussed in the following sections (except palmitoylation, which was covered in section III.F.4.c); their significance with regards to compartmentalization of the Golgi appa ratus and their role in protein targeting will be discussed in later sections of this chapter.

In Section V.B.I, we saw that the basic oligosaccharide core on Asn residues is trimmed by glucose-and mannosidases before secretory pro teins leave the ER. Following their arrival in the Golgi, mannose residues on secretory proteins may be processed in one of two ways, depending on whether they are targeted to the lysosome.

Soluble lysosomal proteins carry phosphorylated mannose residues which function as lysosomal sorting signals (Section V.G.5). They un dergo specific mannose phosphorylation catalyzed by one or two Nacetylglucosamylphosphotransferases and 7V-acetylglucosamine-1 -phosphodiester-a-N-acetylglucosaminidase (phosphodiesterase) in different cis Golgi compartments (614a). Sequential action of these enzymes results in the transfer of TV-acetyl glucosamine-1-phosphate from UDP-NAGn to any one of five mannose residues in the oligosaccharide core, followed shortly afterwards by the removal of the jV-acetylglucosamine to expose the phosphomannosyl group (577). Goldberg and Kornfeld (379) found three partially phosphorylated peptides in β-glucuronidase, indicating that lysosomal enzymes may not be uniformly phosphorylated. Lysosomal enzymes presumably contain signals which are recognized by the phosphorylating enzymes (919). Deglycosylated (endoglycosidase Η-treated) lysosomal cathepsin D is not a substrate for the phosphoryl ation reaction in vitro, but it inhibits phosphorylation of intact lysosomal enzymes. Proteolytic fragments of glycosylated cathepsin D are also not phosphorylated, and do not inhibit phosphorylation when they are dephosphorylated, indicating that the "phosphorylation signal" is probably not a linear sequence of amino acids (612). Frog oocytes are also able to recognize the phosphorylation signal of human cathepsin D (301). Renin is closely related to cathepsin D and yet is secreted by mammalian cells, presumably because it lacks the phosphorylation signal. However, renin produced in oocytes is phosphorylated, remains intracellular, and is de graded (presumably in the lysosome) (302). This suggests that renin has a phosphorylation signal which is recognized by amphibian but not by mam malian phosphorylating enzymes.

Outer chains on lysosomal and nonlysosomal secretory proteins of com plex eukaryotes may be further trimmed by Golgi mannosidase I to leave five mannose residues on the core oligosaccharide. Further modifications involve the addition of an N-acetylglucosamine residue by N-acetylglucosaminyltransferase I, further removal of two mannose residues by Golgi mannosidase II, fucosylation of the innermost TV-acetylglucosamine by fucosyltransferase, and the addition of galactose, N-acetylglucosamine, and sialic acid residues by appropriate transferases (Fig. V. 2). Oligosaccharides of both lysosomal and nonlysosomal proteins may be further modified by sulfation of mannose and N-acetylglucosamine resi dues and O-acetylation of sialic acid residues (577). Outer chain modification in yeasts is markedly different. Golgi mannosyltransferases extend the basic (man) 8 core oligosaccharide to produce large mannan structures typical of many yeast mannoproteins, which may carry as many as 150 mannose residues (598).

The single O-linked mannose residue on yeast glycoproteins (Section V.B.l.c) may be further modified by the addition of up to four more mannoses transferred from GDP-mannose (598,1107 blocked in sec mutants, which prevent secretory proteins from reaching the Golgi (429). In complex eukaryotes, hydroxyl groups of Ser or Thr residues are Oglycosylated by enzymes thought to be located exclusively in the Golgi (418,786). Acceptor octapeptides can be O-glycosylated. 7V-acetylgalactosamine is the primary sugar in O-linked oligosaccharides; further resi dues of galactose, sialic acid, fucose, ^-acetylgalactosamine, and Nacetylglucosamine can then be added. Lipid intermediates are not in volved in O-glycosylation, and the reaction is not inhibited by tunicamy cin (418), but it is not known whether other sugar transferases are in volved in both N-and O-glycosylation.

Glycosylated and unglycosylated secretory proteins are major substrates for tyrosine sulfation (501). Tyrosynylprotein sulfotransferase is enriched in Golgi-derived membrane fractions and has its active site oriented to ward the Golgi lumen (619). The sulfate donor is 3 '-phosphoadenosine 5 phosphosulfate. Almost all tyrosinated proteins have an acidic residue, a glycine or proline residue, and no cysteines, basic residues, extended secondary structure, or N-glycosylation sites close to the modified tyro sine, but there does not appear to be a strict consensus sequence around the modification site (474,619). Inhibition of tyrosine sulfation is reported to retard the exit of a secretary protein from the TGN (331b).

Many secretory proteins undergo secondary proteolytic processing fol lowing removal of the signal peptide. The best characterized examples of this class of processed secretory proteins are α factor and killer toxin of yeasts. The α-factor peptide is present four times in the pro-a-factor protein which reaches the Golgi apparatus. This probably represents a way of reducing "shipping costs" since α factor itself may be too small to be efficiently transported to the RERM and it would be inefficient to pad the polypeptide out with redundant sequences (602). Julius et al. (545) found that pro-a-factor processing was blocked by sec mutations, which prevented protein movement through the Golgi, and that mature α factor was normally present in secretory vesicles en route to the cell surface. Thus, processing occurs in the Golgi. Some viral coat proteins such as influenza virus HA are also proteolytically processed in the trans-Golgi network (TGN) or trans Golgi. Processing of mammalian prohormones, which is similar to that of pro-α factor, occurs in secretory granules budding from the trans Golgi [see Section V.G.4.C and (1003)]).

The four 13-residue-long α-factor peptides in pro-α factor were found to be separated by 6-8 residues (Lys-Arg-Glu-Ala-Asp-Ala-Glu-Asp) (602). This suggests that at least one processing protease has trypsin-or chymotrypsin-like activity. Killer toxins are processed at similar sites (245,1074). In fact, the enzyme which performs this initial processing step, the product of the KEX2 gene, is a Ca 2 + -thiol protease which cleaves between basic residues. The enzyme is inhibited by anti-αΐ-trypsin and can correctly process mammalian proalbumin (53). Strains mutated at KEX2 secrete unprocessed pro-α factor (546). The product of the STE13 gene, a membrane-associated aminopeptidase (544), processes the Nterminal tetrapeptide, and further processing of the C-terminus is per formed by KEX1 -encoded carboxypeptidase (245). Other secretory pro teins produced by different species of yeasts may also be processed by one or more of these enzymes (695b).

Results from a number of experimental approaches (summarized below) show quite conclusively that individual Golgi cisternae are separate, bio chemically and functionally distinct entities organized according to a very strict pattern and that secretory proteins progress in a synchronized wave from one end of the stack of cisternae to the other en route to their final destinations. According to these data, Golgi stacks must contain at least three distinct cisternae (generally referred to as cis, medial, and trans according to their orientation with respect to the ER). Heterogeneity within these "domains" indicates that the actual number of cisternae is probably higher than three (see Table V .l, Fig. V .3) (299). Cells normally have one or a very limited number of Golgi stacks. The stacks break up into clusters of many vesicles during mitosis, apparently to ensure equal partitioning of Golgi components to daughter cells, although the number of Golgi clusters produced is far in excess of that required for this pur pose. Golgi breakdown presumably involves membrane fission, so each cluster of vesicles contains cis, medial, and trans components (662). Pro tein secretion is usually shut down during mitosis or meiosis, but the yeast S. cerevisiae continues to process and secrete invertase during mitosis, implying that the Golgi fragments remain active (685).

In general, results from different analyses give a coherent picture of the cis to trans organization of Golgi cisternae (see Table V . 1), on the basis of which a simple map can be drawn (Fig. V.3) . The segregation of modify ing enzymes presumably allows prosthetic groups to be added or removed according to a strict sequence and prevents competition between process ing enzymes which could act on the same substrate (266). Table V .l is five, but the actual number of Golgi cisternae may be higher than that shown. NAGn, iV-acetylglucosamine; TGN, trans-Golgi network.

Early studies showed that one or two cis (ER-proximal) Golgi cisternae were preferentially stained during prolonged exposure to osmium [pre sumably due to strong reducing conditions in these cisternae (333)]. Sub sequent histological and cytochemical tests showed that many enzymes involved in protein glycosylation and other reactions, as well as some proteins with unknown function, were not evenly distributed through the Golgi stock but were compartmentalized in one or two cisternae (Table  V .l). The techniques employed include immunocytological detection of proteins using specific antibodies, enzymatic cytochemical reagents for specific enzymes, and the detection of lectin-specific sugar residues on terminally modified glycoproteins in transit through the Golgi (Table V .l). As noted by Farquhar (299), some studies with different cell lines give conflicting results, suggesting that the organization of the Golgi stack may vary depending on cell type and function.

Intermediates in the secretion pathway can be detected by pulse-chase experiments in which oligosaccharides or other prosthetic groups are la beled by the metabolic incorporation of radioactive precursors. The se quence in which processing occurs can thus be determined and correlated with the location of processing enzymes in the Golgi cisternae.

Golgi enzymes involved in N-and O-glycosylation, sulfation, and palmi toylation of secretory proteins can be separated by density gradient cen trifugation of Golgi-derived vesicles, which apparently differ in density due to differences in cardiolipin content (817). Enzymes identified in cis Golgi by kinetic and histological or cytochemical tests are generally found in heavier Golgi membrane fractions, and density is now widely used to define the approximate location of Golgi proteins in the stack (Table V .l).

The ionophore monesin slows or arrests intra-Golgi transport and inhibits late (trans) Golgi functions, and can thus be used to distinguish between early and late Golgi processing events. Monesin also causes secretory proteins to accumulate in medial or late Golgi compartments, which be come distended and vacuolated, aiding their identification by cytochemi cal methods (400,905,1108). However, as discussed by Dunphy and Rothman (266), studies on the effects of monesin in different cell types often give conflicting results and do not necessarily indicate the precise site of secretory protein accumulation or processing.

Oligosaccharides on secretory proteins are only cleaved by endoglycosi dase Η before they are processed by NAGn transferase I and mannosi dase II. These enzymes are probably located in medial Golgi cisternae. Thus, endoglycosidase Η-sensitivity provides a simple test for determin ing whether secretory proteins have reached these cisternae (267).

Rothman et al. (969, 970) demonstrated that pulse-labeled VSV G protein present in the Golgi of a cell lacking a particular modification enzyme can be modified upon fusion with a cell producing the modifying enzyme. The Golgi cisternae from donor and recipient cells did not fuse, and the G protein, rather than the modifying enzyme, was transferred from one cisterna to the other. These observations have implications for the mecha nisms of protein movement between cisternae (Section V.F) but, like in vivo kinetic experiments, may also indicate the sequence in which pros thetic groups are added or removed ( 

Having established that secretory proteins are progressively processed as they migrate through the Golgi stacks, we now turn our attention to how the proteins themselves migrate between cisternae, and how Golgi pro teins can be specifically retained within specific cisternae.

Of all of the models explaining intra-Golgi movement of secretory pro teins (300), only that invoking shuttle vesicles fits the experimental data showing that Golgi cisternae are distinct entities. Numerous proteincoated vesicles are produced by Golgi stacks in vitro under conditions which favor the intra-Golgi movement of secretory proteins (see below) (40,820). These uniformly sized vesicles were shown to contain at least one secretory protein, VSV G protein, indicating that they could be bona fide intra-Golgi shuttle vesicles (820). Similar vesicles were also found around the Golgi apparatus in situ (996, 1088) . Furthermore, different se cretory proteins destined to different locations and exported or secreted at different rates were present in the same vesicles (1088), which agrees with the idea that secretory proteins are not segregated from each other during intermediate stages in the secretory pathway. The protein coat on these vesicles is distinct from the clathrin-containing coats present on endocytic and lysosomal vesicles and on budding secretory granules (see Section V.G.6). The only regions of the Golgi apparatus which have pro tein coats are those from which the Golgi shuttle vesicles bud.

A number of different conditions prevent intra-Golgi movement of secre tory proteins in vivo and in vitro. The ionophore monesin (Section V.E.4) probably prevents movement though the trans cisternae by disrupting a proton gradient maintained by an ATP-dependent protein pump present in Golgi membranes (48), thus causing the pH of the normally acidic trans Golgi cisternae to rise. Saraste and Hedman (994) demonstrated that mi gration between different Golgi cisternae was blocked at different critical temperatures and suggested that this might be caused either by ATP depletion or by changes in membrane fluidity. ATP may be required to maintain the acidic pH of trans cisternae, and vesicle fission and fusion, which must occur at the cisternal membranes, can probably only occur when membrane lipids are in a "fluid" state above the phase transition temperature. ATP and soluble cytosolic factors including a 74-kDa, N-ethylmaleimide-sensitive protein (93a) are required in both early and late stages of intra-Golgi movement in vitro (39,40,685a,966) and in permeabilized cells (62). Intriguingly, soluble yeast cell extracts can replace endogenous cy toplasmic components in a mammalian cell-derived assay system for in tra-Golgi movement of proteins (268), raising the possibility of using ex tracts from yeast sec mutants blocked at different stages in the secretory pathway to define the role of the corresponding wild-type gene products in intra-Golgi transport. Indirect evidence based on the effects of GTP analogs suggests that GTP is also needed for intra-Golgi movement of secretory proteins (710), and studies with antibodies against a yeast GTP binding protein indicate that a similar protein is apparently located in the Golgi apparatus in multicellular eukaryotes (1030). The significance of this observation is not understood, but GTP and GTP binding proteins (G proteins) may be involved in maintaining vectorial movement of shuttle vesicles or in vesicle recycling (see below). Surface components on Golgi membranes are also required for intra-Golgi transport (3,39,40). A recep tor may recognize the protein coat on the Golgi transport vesicles.

Balch et al. (40) considered that the migration of secretory proteins from one cisterna to another depends on three separate events:

(i) Secretory proteins are primed to make them competent or available for transport. Priming probably involves the migration of secretory proteins to regions of the cisternae where budding occurs, princi-pally at the outer rims. This step might depend on interactions be tween the secretory protein and receptors migrating to the budding areas or may rely on free diffusion within the cisternal membrane or lumen. Segregation may depend on signals generated by processing enzymes in individual Golgi cisternae, but most experiments in which processing inhibitors have been used do not support this idea and show instead that processing is nonessential for secretory pro tein targeting. Some mutationally altered secretory proteins, how ever, transit normally through the early stages of the secretory path way and yet are not transported through the Golgi (344,1040). Perhaps secretory proteins must fold into a particular conformation to be competent for movement between Golgi cisternae. The accu mulation of such abnormal proteins in the Golgi may cause it to become distended, but this does not drastically affect intra-Golgi movement and secretion of other secretory proteins (344). (ii) Vesicles move from one cisterna to another. Vesicle migration be tween Golgi cisternae may be either vectorial or random. Vectorial movement is more compatible with traditional views on the strict sequence of events in secretory protein processing and is supported by the observation that the likelihood of forward transfer is at least five times greater than that of lateral movement in Golgi of fused cells (969, 970) . This implies that there are receptors on the surfaces of Golgi cisternae which recognize vesicles budding from the pre ceding cisterna in the chain. Nonetheless, the topology in the Golgi complex is well maintained, possibly because the cytoskeleton pre vents cisternae from coming into direct contact. Another puzzling feature of vesicular intra-Golgi transport is that small transport vesi cles would be expected to diffuse away from the Golgi complex. The cytoskeleton may restrict the movement of Golgi vesicles and may even play a more positive role in directing vesicles between cister nae. However, although movement along microtubules has been well documented for large organelles (1136) and may be important in the sorting of some proteins leaving the Golgi complex (Section V.G.3), there is no evidence that it could play more than a minor role in the movement of vesicles over the very short distances which separate Golgi cisternae (139). Furthermore, cell fusion studies by Rothman et al. (969) show that inter-Golgi movement of VSV G protein can occur, indicating either fusion of the two Golgi com plexes or, more likely, that movement between Golgi cisternae is dissociative; i.e., the vesicles do indeed diffuse into the cytoplasm. (iii) Vesicle fusion with the membrane of the acceptor cisterna and re lease of vesicle contents into the lumen of the cisterna occurs.

This model for intra-Golgi transport fits many experimental observa tions, but further studies are required to define clearly the steps involved. For example, ATP may be required for vesicle fission and fusion, as well as for reducing the pH in trans Golgi compartments, but this has not been proven, and the nature and role of some of the cytosolic components required in in vitro assays have yet to be determined. Further work is needed to define how proteins find their way to budding regions of the Golgi membrane and what triggers vesicle formation and vesicle fusion with acceptor membranes.

How are proteins specifically retained in individual Golgi cisternae? We saw earlier that receptor-dependent recycling of ER proteins from the cis Golgi or an intermediate compartment could explain how these proteins remain almost entirely in the ER (Section V.B.4). Golgi residents proba bly also have signals which are recognized by some kind of receptor. Indeed, coronavirus El membrane glycoprotein appears to have a Golgi retention signal in one of its transmembrane domains (670). However, recycling of escaped proteins is a far less attractive model for explaining how Golgi proteins are retained than it is for the case of ER proteins. One possibility, proposed by Pfeffer and Rothman (872) , is that the membranes of Golgi cisternae have two domains: a fluid domain, close to the budding rims, and an immobile phase, in which endogenous membrane proteins are anchored. A protein would require a signal to associate with a recep tor in the immobile phase or to become anchored to it, whereas all other proteins would enter the mobile membrane phase or the bulk phase of the cisternal lumen. Pfeffer and Rothman point out that this model reduces the number of proteins which need to have signals for routing through or retention in the Golgi, because fewer proteins are retained in the Golgi than transit through it. Cytoskeletal structures, including possibly the cytoplasmic matrix, which "glues" the cisternae together, were proposed to limit movement in the immobile phase. An alternative idea, also con sidered by Pfeffer and Rothman, is that endogenous Golgi proteins inter act to form patches which, by virtue of their size, are too small to fit into transport vesicles.

Secretory proteins transit through the Golgi apparatus and arrive in the trans cisternae together. The trans Golgi compartment is therefore the point at which the different branches of the secretory pathway diverge.

The following sections deal with the site at which sorting occurs, the ways in which proteins are sorted, and what happens when vesicles carrying secretory proteins arrive at their destinations.

As its name suggests, the trans-Golgi network (TGN, also called Golgi endoplasmic reticular lysosomes or GERL) is the most distal compart ment of the Golgi apparatus (relative to the RER). It differs from the Golgi cisternae in that it has a distended, reticular appearance rather than that of a flattened dish. Early morphological studies suggested that the TGN was a reticular adjunct of the Golgi specifically involved in lysosome biosynthesis [lysosomal enzymes were originally thought to bypass the Golgi (398)]. The TGN is now known to be distinct from endosomes or lysosomes. Endocytosed horseradish peroxidase does not accumulate in the TGN (401), although some endocytosed proteins may be recycled to the cell surface via the trans Golgi and especially via the TGN (319) (see Chapter VIII). The TGN probably contains tyrosinyl sulfotransferase (33), acid phosphatase, sialotransferase, and galactosyltransferase (365,957) (Table V .l, Fig. V.3) , although TGN-derived vesicles are diffi cult to distinguish from those derived from the trans cisternae. The TGN also marks the site at which assembled clathrin, one of the proteins which coat some secretory and endocytic vesicles, appears along the secretory pathway (see Section V.G.6).

The TGN is the most acidic Golgi compartment, although the pH is almost certainly not as low as in secretory granules (823) or in lysosomal sorting vesicles, in which low pH causes the dissociation of lysosomal proteins from the mannose-6-phosphate receptor (577) (see Section V.G.5). Furthermore, influenza virus hemagglutinin, which is activated at low pH, reaches the cell surface as a nonactivated form, indicating that it does not spend an appreciable period (more than 2 min) in a compartment with a pH of less than 6 (107).

Different secretory proteins accumulate together in the TGN (1088,1123), which can become further distended when the load of secre tory or lysosomal proteins increases (398). Transport of secretory pro teins from the TGN is blocked at 20°C, and numerous protein-coated and naked vesicles accumulate as buds on the surface of the TGN (401). Different types of vesicles seem to be involved in sorting secretory pro teins into different branches of the secretory pathway. Thus, different classes of soluble secretory proteins may be segregated into different domains of the TGN according to their interaction with specific receptors (Fig. V.4 ). Although this model does not explain how receptor proteins Note that constitutive default sorting is not receptor-mediated and that sorting into the regulated secretory branch of the pathway may result from protein aggregation and granule formation rather than receptor interactions. TGN, trans-Golgi network; L, lysosome; PL, prelysosome. segregate into these domains, it does serve to illustrate the possible mech anisms involved in receptor-dependent segregation and packaging secre tory proteins discussed in the following sections.

Default sorting is the final stage in the secretion of proteins which lack specific sorting (lysosomal, vacuolar, or polarity) signals and which are not accumulated within specific secretory storage granules of the regu lated branch of the secretory pathway, i.e., those which are constitutively secreted (Fig. V.4) . (Note, however, that some constitutively secreted proteins carry sorting signals.) Immunocytochemical studies show that proteins secreted by the default pathway accumulate in secretory vesicles which do not have protein coats (401,821). Plasma membrane proteins are transported to the cell surface in the same vesicles as secreted proteins (114, 462, 1088) . (1204) (see Section II. F. 3) required ATP and was sensitive to trypsin, implying that at least one cytoplasmic or vesicle-cytoplasmic membrane surface protein is involved. ATP might be needed for fusion between the vesicle and cytoplasmic mem brane.

Saccharomyces cerevisiae strains carrying a temperature-sensitive mu tation in the SEC4 gene accumulate large numbers of secretory vesicles. These were purified by Walworth and Novick (1165), who found them to contain three dominant proteins (110 kDa, 40-45 kDa, and 18 kDa) to gether with invertase, the secretory marker protein. These three proteins were made during the period in which the vesicles accumulated, i.e., after secretion had been shut down at the nonpermissive temperature. The 110-kDa protein was the most abundant protein in the lumen of the secretory vesicles, whereas the other two proteins were membrane-associated and had cytoplasmic domains which could interact with the cytoskeleton, the cytoplasmic membrane, or cytoplasmic components (1165). It should be noted, however, that exocytosis in S. cerevisiae might be considered as polarized rather than default sorting because secretory vesicles are di rected towards the growing bud rather than being randomly distributed over the entire cell surface (Section V.G.2 and Fig. II.5) .

The SEC4 gene was cloned and sequenced by Salminen and Novick (987), who found it to be homologous to GTP-binding RAS regulatory proteins of higher eukaryotes. Whether this is significant for the role of SEC4 in secretion is unclear, especially since it is not known whether GTP is required for exocytosis in yeasts. Recent studies show that SEC4 protein binds GTP (391). Overexpression of SEC4 suppresses the effects of mutations in three other SEC genes, but strains carrying sec4ts muta tions rapidly become secretion-defective at the nonpermissive tempera ture, suggesting a direct role in secretion. The cytoplasmic and secretory vesicle membrane-associated SEC4 product does not itself appear to be a secretory protein because its predicted primary sequence does not in clude a potential secretory routing signal. One possibility raised by Salminen and Novick is that the C-terminal cysteine residue of SEC4 is acylated, as are the GTP-binding RAS proteins and that the acyl groups anchor the polypeptide in the membrane, but this could not be confirmed directly. Another GTP binding protein, YPT1 (which is 50% homologous to SEC4) was detected close to the bud as well as in ill-defined structures (possibly the ER and Golgi) of S. cerevisiae cells (1030). Mutations in the YPT1 gene affect several stages in the secretory pathway, but YPT1 pro tein probably plays an indirect role in protein transport as a result of its role in Ca 2 + regulation (1014).

The plasma membrane of epithelial cells is divided into two distinct do mains. The apical surface, which may have microvilli, is oriented toward the outside (e.g., lumen of the intestine), whereas the basolateral surface is on the inside, facing the basolateral surface of other cells or resting on an extracellular matrix of basal lamina (Fig. V.5) . The two membrane domains have distinctly different lipid contents in their outer leaflets [the apical membrane has a higher glycolipid and cholesterol content, and a lower phosphatidylcholine content than the basolateral membrane (1050)], although the lipid contents of their inner leaflets may be identical (inset to Fig. V.5 ). This implies that only the outer leaflets of the two membranes are separated by tight junctions, morphologically distinct structures rich in nonbilayer lipids and containing unique proteins which form the junction between apical and basolateral surfaces and probably prevent the movement of outer leaflet lipids and proteins between them, as well as acting as ion gates between adjacent cells (Fig. V .5) (177, 260, 409, (706) (707) (708) 1050) . There may also be differences in basal and lateral membrane composition. Cell-cell interactions are required for effi cient formation of the basolateral surface of polarized cells, but not for sorting to the apical zone (1143). Adjacent cells may be held together by desmosomes.

The major breakthrough in studies on protein sorting in polarized cells (878). In general, the steady-state distribution of polarized membrane proteins indicates that sorting of basolateral pro tein is highly efficient (>97% fidelity) (873), whereas apical proteins may be found in significant amounts in the basolateral membrane. However, the surface area of the apical membrane is usually much smaller than that of the basolateral membrane, which means that the fidelity of apical tar geting is actually quite high (about 88%) (873). Basolateral and apical sorting seem to represent two distinct pathways, but most polarized cells also secrete some proteins from both surfaces, implying that neither basolateral nor apical sorting are default pathways (390,574). Furthermore, lysosomal proteins are secreted by the default pathway from both apical and basolateral surfaces when lysosomal sort ing is blocked (140). This may not be the case in Caco-2 cells in which high-level basolateral secretion of normally nonpolarized lipoproteins im plies that the basolateral sorting predominates over all other sorting path ways, including the default pathway (136,930a).

Do polarized cells sort and direct proteins directly to their target mem branes, or are they all first randomly targeted to both domains, or specifi cally to one or other domain, and then transcytosed? Although some studies designed to answer this question have been conducted with en dogenous proteins, the most detailed results come from studies on the basolaterally targeted VSV G protein and the apically targeted influenza virus hemagglutinin (HA) in MDCK cells. Furthermore, it is through derivatives of these proteins that most recent studies have attempted to identify apical and basolateral targeting signals (see next section).

The results of pulse-chase experiments combined with a trypsin sensi tivity assay and tests with anti-HA antibodies applied to the basal surface of MDCK cells led Matlin and Simons (695) to conclude that HA was targeted directly to the apical surface. Conversely, Pfeffer et al. (873) showed that pulse-labeled G protein went directly to the basolateral mem brane, where it could be detected with specific monoclonal antibodies. Another approach to studying the sorting of HA and G proteins is to ''freeze" them in the Golgi apparatus by cooling the cells to 20°C (which specifically blocks protein exit from the TGN) or to use cells infected with viruses carrying temperature-sensitive mutations in the HA or G genes at the restrictive temperature. The bulk movement of these proteins can thus be followed by immunocytochemistry when the cells are restored to the permissive temperature. Rindler et al. (929) found that Gts protein was transported directly from the Golgi to the adjacent lateral cell sur face, whereas wild-type HA accumulated at 20°C was sorted directly to the region of the apical surface closest to the TGN. Similarly, Pfeiffer et al. (873) found that G protein accumulated in the TGN went to the baso lateral membrane 67 times faster than to the apical surface when the cells were warmed to 37°C. These studies are compatible with the idea that apical and basolateral proteins go directly to their respective target mem branes. Different results were reported by Bartles et al. (50) , who found that endogenous apical proteins in hepatocytes appeared first in the baso lateral membrane and were subsequently redistributed to the apical sur face by transcytosis (Section VIII.A.5.b). Whether this result indicates the existence of totally different mechanisms for sorting apical proteins in kidney and liver cells could be tested by expressing hepatocyte genes for apical proteins in MDCK cells.

Sorting of polarized secreted and membrane proteins is assumed to de pend on sorting signals (probably signal patches) [see Section I.D.2 and (88)] present in the sorted proteins and on their interaction with receptors. Sorting presumably occurs in the TGN, where at least one set of cognate receptors are probably located. To date, the primary approach used to locate sorting signals has been to introduce major sequence alterations including the deletion of cytoplasmic or transmembrane domains from the HA, G, or other polarized viral glycoproteins, or to create hybrids be tween them. Even these relatively unsophisticated studies give conflicting results. For example, McQueen et al. (702) and Roth et al. (959) found that either deleting the transmembranous and cytoplasmic tail of HA (to make a soluble, secreted protein) or replacing them with the correspond ing regions of G did not affect apical targeting and concluded that the apical sorting signal was located in the N-terminal, extracytoplasmic do main of HA. Gonzalez et al. (387) reported, however, that truncated HA was secreted by the default pathway and concluded that the sorting signal was in the C-terminal transmembranous or 10-amino-acid cytoplasmic domains. There is also disagreement concerning the location of the G protein basolateral sorting signal, which, according to the deletion and HA gene fusion studies of Paddington et al. (835) and Gonzalez et al. (387) is located in the 29-amino-acid, C-terminal cytoplasmic domain. Stephens and Compans (1080) McQueen et al. (703) have suggested that the origin of these conflicting results may be the fact that recombinant genes are not stably expressed in MDCK cells. Their studies (702,703) were carried out shortly after infec tion by recombinant viruses and are, they claim, more likely to reflect the true sorting pathway of the recombinant gene product. Other groups may in fact be studying the sorting of hybrid or truncated proteins which have been further modified by genetic selection for more stable cell lines. An other argument in favor of the idea that sorting signals are in the extracy toplasmic domain is that this part of the signal is initially localized in the lumen of the TGN, where sorting is presumably initiated, and could there fore bind to hypothetical sorting receptors which segregate sorted pro teins from those destined for default export or secretion. These receptors may themselves have cytoplasmic domains recognized by receptors on the basolateral or apical surfaces. Furthermore, such a system could han dle both secreted and exported (membrane) proteins. However, there is some evidence, based on the effect of NH 4 C1 (or pH) on basolateral sorting of laminin and heparin sulfate proteoglycan, that soluble and membrane proteins may be sorted by different mechanisms (142).

Clearly, more work remains to be done before we can identify polarity sorting signals with any precision. One of the difficulties may be that these signals are probably not linear sequences of amino acids and are therefore less amenable to gene fusion studies, which have been so useful in identi fying routing signals. Even very subtle conformational changes may alter sorting signals and render them nonfunctional; this could explain why prevention of glycosylation with tunicamycin may cause normally polar ity-sorted proteins to enter the default pathway (1135).

Another aspect of the work on polarity sorting signals which needs to be pursued is the identity of receptors and the stage at which sorted proteins dissociate from them. Sorting receptors are presumably re cycled. Little is known about how the vesicles are targeted to different regions of the plasma membrane. One possibility is that sorting of lipids ( Fig. V.5) is involved (709,1050a) . Alternatively, sorting vesicles could interact with components of the cytoskeleton, along which they are driven by ATP-dependent motors attached to microfilaments; further receptor-ligand interactions might complete the sorting process when vesi cles arrive at the plasma membrane (1136). Although there is no evidence that microfilaments are required for default sorting of secretory proteins (1136), Rindler et al. (930) have reported that colchicine and other microtubule-disorganizing agents abolished specific apical sorting of HA and caused influenza virus to bud from both cell surfaces in polarized cells. Salas et al. (986) obtained the opposite result, however, and neither group observed any effect on basolateral targeting of VSV G protein. Rogalski (944), however, found that agents which caused microtubule disassembly caused random sorting of G protein. Thus, the role of the cytoskeleton in the targeting of polarity-sorted secretory proteins in complex eukaryotes' proteins remains unclear. A temperature-sensitive mutation in the S. cerevisiae actin gene results in abnormal exocytosis and bud formation at the nonpermissive temperature, suggesting that actin filaments may direct secretory vesicles to the growing bud in yeast cells (793).

There is evidence to show that the cytoskeleton may be important in maintaining polarized distribution in complex eukaryotic cells (410). A polarized ATPase has recently been shown to bind directly to ankyrin, a protein which is known to link an integral erythrocyte membrane protein to spectrin and actin of the erythrocyte cytoskeleton (770). Thus, ankyrin might link the ATPase to microfilaments and thereby maintain its polar ized distribution.

Two features distinguish regulated exocytosis from other branches of the secretory pathway:

(i) Protein release only occurs when the cells are stimulated by secretagogues (e.g., cyclic AMP or Ca 2 + ). (ii) Proteins accumulate within the cell before their release.

Proteins accumulate in a special class of secretory vesicles called secre tory granules, wherein protein concentrations reach such high levels that they become electron-dense and can be clearly identified by electron mi croscopy. The secretory pathway is restricted to certain cell types (e.g., endocrine and exocrine cells, mast cells), and only certain types of pro teins (examples include hormones, albumin, and some degradative en zymes including proteases and lipases) enter into the pathway. Another feature of the regulated secretory pathway is that proteins may be pro-teolytically processed before release or as they are released from the cell, although this feature is also found in some constitutively secreted pro teins.

Regulated and constitutive branches of the secretory pathway can coexist in the same cell (395, 567, 1109) , implying that proteins destined for storage in secretory granules must be sorted from other secretory proteins. Elec tron microscopic studies show that both classes of secreted proteins tran sit through the Golgi cisternae together and are segregated in the TGN, where proteins destined for the regulated branch of the pathway condense into specific areas coated with the protein complex clathrin (see Figs. V.4 and 6; see also Section V.G.6) (824,1122). The clathrin coat is subse quently removed as the condensing granules bud from the TGN and ma ture (818) (see Fig. V.6 ). These areas probably represent the sites at which secretory granules mature and are released from the TGN. It should be noted, however, that some proteins which are normally se creted by the regulated pathway may 4 'escape" and be secreted 4 'consti tutively" (31). It remains to be determined whether this is due to incorrect sorting or to low-level, secretagogue-independent secretion from storage granules, as suggested by studies by von Zastrow and Castle (1231). Bur gess and Kelly (136) propose that this "spillover" secretion may be due to inefficient sorting in the specific cell lines tested, since Rhodes and Halban (921) observed much more efficient sorting into the regulated pathway.

Efficient segregation of proteins into regulated and constitutive branches of the secretory pathway implies that the former have sorting signals (see Section V.G.4.b for an alternative explanation). The exis tence of these signals was suggested by Moore and Kelly (731), who transfected a pituitary tumor cell line with a hybrid gene comprising the 5 ' end of the VSV G protein and the 3 ' end of the gene for human growth hormone. The hybrid protein was diverted into the regulated secretory pathway, indicating that the constitutive pathway had been bypassed due to sorting signals present in the growth hormone part of the hybrid. There are no obvious sequence similarities between proteins secreted by the regulated pathway, implying that the sorting signal is probably a patch signal rather than any identifiable linear stretch of amino acids. Several proteins secreted via the regulated pathway are proteolytically processed prior to their release from the cell (see below). This raises the possibility that the sorting signal could reside in that part of the secretory polypep tide which is eventually cleaved off. This possibility was tested by Bur gess et al. (135) , who found that deleting DNA coding for the propeptide part of trypsinogen had only a minor effect of the targeting of the enzyme into secretory granules. Thus, they concluded, there must be at least one sorting signal in the mature part of the trypsinogen.

One surprising feature of regulated pathway sorting is that the putative sorting signal seems to be universal. Moore et al. (732) , for example, found that human proinsulin was packaged into secretory granules in mouse adrenocorticotropic hormone (ACTH)-secreting cells, that it was correctly processed (see below), and that its release from these cells was stimulated by the same secretagogues that stimulated ACTH release. Similar results were obtained in studies on the expression of human kid ney renin DNA in the same cell line (337). Fibroblast L cells, however, which do not have the regulated pathway, secreted (unprocessed) proin sulin via the constitutive pathway (732).

Receptors localized in specific domains of the TGN membrane may segregate proteins into the regulated branch of the secretory pathway, but this has not been proven, and other factors may be important. For exam ple, treatment with the weak base chloroquinone causes ACTH to be secreted by the constitutive pathway, indicating that low pH is required for sorting into the regulated pathway (733). Low pH in condensing secre tory granules may dissociate proteins from their receptors, which can then be reused (but see Section V.G.4.b). Another feature of secretory granules which appears to have been largely overlooked is that proteases involved in post-TGN processing of secretory proteins must also carry sorting signals. It remains to be seen whether the membrane content of secretory granules differs significantly from that of vesicles of the consti tutive branch of the secretory pathway.

In certain exocrine cells, the concentration gradient of a secretory protein between the RER and secretory granules can be as high as 200 (988). The dense core of aggregated protein is sometimes seen to be separated from the membrane of the secretory granule (136) and may remain intact when the membrane is removed (1230) or upon exocytosis (19, 1122) . Although receptor-mediated sorting into specific regions of the TGN may assist the condensation process, proteins secreted by the regulated pathway may aggregate spontaneously and be packaged into secretory granules when they reach a critical size. Thus, the packaging of G protein-growth hor mone hybrids into secretory granules discussed above (731) may be due to the presence of "aggregation" sequences in the growth hormone segment of the polypeptide, rather than to the presence of a specific sorting signal. The low pH of the condensing granule (19,824) may be important for protein aggregation, since aggregates dissociate at high pH (534). Pfeffer and Rothman (872) suggest that this could explain the failure of chloroquinone-treated cells to package secretory proteins into secretory gran ules (see above). Secretory granules of exocrine (hormone-secreting) cells remain acidic during maturation, but those of endocrine (enzyme-secret ing) cells return to neutral pH as they mature (534). ATP is also needed for secretory granule formation.

There are conflicting views as to whether different secretory proteins cosegregate and coaggregate into the same secretory granule. Detailed studies by Fumagalli and Zanini (341) revealed that bovine growth hor mone and prolactin could be present either in different aggregates in the same secretory granule or in mixed aggregates in the same granule or in pure aggregates in different granules. The ratios of the three type of granules varied from animal to animal. Similar results were reported by Mroz and Lechene (744) , who showed that the enzyme content of individ ual secretory granules derived from single cells from the same gland can vary enormously. The simplest interpretation of these data seems to be that segregation is a random process and that the formation of mixed or pure aggregates depends on the local concentration of the respective pro teins and on their preference for forming homo-rather than heteroaggregates. Packaging of proteins into different secretory granules, however, might permit their release to be stimulated by different secretagogues, but there is only limited evidence for such a phenomenon at the level of individual cells (11,136,317a) .

Many of the proteins secreted by the regulated branch of the secretory pathway are proteolytically processed and activated, usually in secretory granules. Processing often involves the proteolytic removal of an N-terminal propeptide and can be mimicked by exogenous proteases such as trypsin (336). Alternatively, short spacer peptides may be removed from polyprotein precursors (187). In the latter cases, cells in different tissues can process the precursors to give different "mature" forms. This is the case for prosomatostatin, which is processed to a 28-amino-acid form by cells in the gut and to a 14-amino-acid form by brain and pancreatic cells (790). Islet tissue from angler fish pancreas contains at least two proteases which process prosomatostatin to give products of different lengths. One of these proteases can also process proinsulin (674) . Other examples of processing of heterologous secretory proteins (188, 444, 732) indicate the existence of only a limited number of processing proteases, which may also be found in some cell types which do not have a regulated secretory pathway (444, 1168) . When secretory proteins which would normally use the regulated pathway are produced in cells which do not have the regu-lated pathway, however, they are constitutively secreted as unprocessed (pro-) forms (889). The site of proinsulin processing was determined cytochemically by Orci et al. (816, 825) , using monoclonal antibodies specific for mature insulin. Fully processed insulin was first detected in clathrin-coated vesi cles budding from the TGN, and subsequently in naked granules. Process ing was coincident with condensation and acidification and was inhibited at higher pH, indicating that the two proinsulin-processing proteases have low pH optima (821). Therefore, this and other proteolytic processing steps probably occur in a late "compartment" of the TGN.

Although propeptides may not play a role in protein sorting, they may prevent enzymes such as proteases from folding into active conforma tions prior to their release, thereby protecting secretory granules from endoproteolytic attack. Proinsulin and other prohormones may be loosely membrane-associated (789,819). In these cases, bridge sequences may contribute to the patch signal which shunts proteins into budding secre tory granules.

Relatively little appears to be known about the events which accompany protein release from secretory granules. The general view seems to be that secretagogues directly, or more likely indirectly, stimulate fusion between the granule membrane and the cytoplasmic membrane, resulting in the release of granule contents to the outside of the cell. Secretagogues bind to specific cell-surface receptors and promote Ca 2 + influx, which seems to be intimately related to the fusion event (756). GTP binding proteins also seem to be involved in generating the signal which stimu lates secretion (138), and microtubules may play a minor role in directing storage granules to the cell surface (112). Inhibitors of metalloprotease and dipeptide protease substrates inhibit exocytosis, suggesting that pro teolytic cleavage of a membrane protein may be essential for exocytosis. Mundy and Strittmatter (756), who found that metalloprotease activity was highest in the plasma membrane, propose that proteolysis may un mask the active site on a fusogenic membrane protein.

Breckenbridge and Aimers (121) have recently studied exocytosis-associated changes in membrane capacitance in a mouse mast cell mutant with enlarged secretory granules. Small fluctuations in capacitance preceded larger increases, which themselves preceded granule swelling and the release of a fluorescent tracer dye from the lumen of the granule. The large increase in capacitance probably results from productive fusion be tween the plasma and granule membranes (312), whereas capacitance "flutters" may represent nonproductive membrane association. The most plausible interpretation of these data is that release of secretory granule contents is preceded by the formation of a narrow channel, the fusion pore, (121) between the two membranes, leading eventually to the open ing of the granule membrane to the outside of the cell and the subsequent swelling and dissociation of the granule contents and their release as soluble proteins.

Lysosomes (in animal cells) and vacuoles (in plant and fungal cells) con tain most of the cells' degradative enzymes, which function not only in general "housekeeping" but also in the degradation of endocytosed mate rial (Chapter VIII). The following sections review the evidence for lysoso mal and vacuolar sorting signals, special features of the sorting pathways, and differences between the lysosomal and vacuolar routes.

Specific sorting of soluble lysosomal enzymes is determined by mannose-6-phosphate (M6P) residues on N-linked core oligosaccharides (Section V.D.I.a). Two receptors have been identified. The major M6P receptor (275 kDa, formerly called the 215-kDa receptor) was detected predomi nantly in the cis Golgi compartment (131), leading to the proposal that the lysosomal pathway diverged from the main secretory pathway at the cis end of the Golgi stack rather than in the TGN. This idea is incompatible with the observation that some lysosomal proteins are terminally pro cessed by enzymes located in medial and trans Golgi compartments.

Other cytological studies indicate that M6P receptors are located in the TGN, as well as throughout the Golgi stack (208), in coated vesicles and the plasma membrane (365,366) (see below for explanation), and in a Golgi-proximal vesicular structure (402), but not in lysosomes. Thus, ly sosomal enzymes probably do transit through the TGN, possibly already complexed with their receptor, although Farquhar (299) argues strongly in favor of multiple lysosomal sorting pathways and in particular for sorting from the cis Golgi in certain cell lines. Thus, the site of accumulation of M6P receptors along the secretory pathway may have little relevance for lysosomal enzyme sorting. Two M6P receptors (275 kDa and 46 kDa) have been identified and characterized. The receptor activity of the 275-kDa protein, but not that of the 46-kDa protein is cation-independent, and the 46-kDa receptor recognizes only phosphomonoesters whereas the 275-kDa protein also binds methylphosphomannosyl residues (461). The 275-kDa protein ap pears to be present in most mammalian cell lines tested so far (982), but the distribution of the 46-kDa receptor has not been determined. Some mutant cell lines lack the 275-kDa M6P receptor yet still target lysosomal enzymes normally, which suggests that both receptors are involved in lysosomal sorting (461). Studies with such cell lines revealed another difference between the two receptors, however. Although part of the cellular pool of both receptors is located on the cell surface, only cells with the 275-kDa protein can endocytose secreted lysosomal proteins (1077). This "defect" in the 46-kDa protein appears to be due to a failure to bind the ligands, since antibodies against the 46-kDa protein are endocytosed normally (see Chapter VIII for more details on endocytosis). Thus, the 46-kDa protein seems to be specifically involved in the sorting of endogenous lysosomal enzymes.

The genes for both M6P receptors have been cloned and sequenced. Although the predicted sequence of the two gene products are generally different, the two proteins have a region of moderate sequence similarity in their lumenal domains (645,826) which Dahms et al. (208) propose could be the M6P binding domain.

The sorting of the lysosomal enzymes cathepsin C and cathepsin D was studied directly by Lemansky et al. (626) , who found that lysosomal enzyme precursors occurred only in coated vesicles. Proteolytically ma tured forms were found in lysosomes. Schulze-Lohoff et al. (1022) also observed the transient accumulation of one of these enzymes in coated vesicles, which, they proposed, are specifically involved in the sorting of lysosomal proteins from the secretory pathway. Lemansky et al. (626) devised procedures which allowed vesicles derived from the secretory pathway to be separated from those derived from the endocytotic path way. They found that both classes of vesicles contained precursor forms of cathepsin. These studies have two profound implications:

(i) The fact that vesicles involved in direct sorting of cathepsins to the lysosome contain clathrin implies that they had passed through the TGN, the first site along the secretory pathway at which clathrin is detected (Section V.G.I). (ii) The fact that some lysosomal enzymes are "fished" out of the sur rounding medium and retargeted to the lysosome implies that some lysosomal enzymes are incorrectly sorted, probably into the consti tutive secretory pathway (see also Section V.D. 1 .a). Higher levels of incorrect sorting occurs in NH 4 -Cl-treated cells, probably because low pH is required to dissociate M6P from its receptor in the prelysosome ( Another interesting observation concerning the M6P receptor is that it specifically binds to one of the components (the 100-kDa accessory pro tein) of the clathrin cage which coats the sorting vesicles (854) (see be low). This may have particular relevance for the sorting of lysosomal enzymes because different classes of clathrin-coated vesicles appear to have different types of accessory proteins (855).

Coated vesicles almost certainly do not transport proteins directly into lysosomes. Instead, the vesicles are targeted to endosome-like reticular organelles (prelysosomes or secondary endosomes), where the receptor probably dissociates and recycles back to the Golgi cisternae. This organ elle is also the site to which endocytosed lysosomal proteins are targeted (134,402) ( Fig. V.6 and Section VIII. A.4.b). A different class of vesicles may complete the transport of lysosomal enzymes once they have dissoci ated from their receptor, but von Figura and Hasilik (313) consider it more likely that there is a gradual transition from tubular prelysosomes to lyso somes proper.

Although M6P is undoubtedly the major, and in some cells the only, sorting signal on lysosomal enzymes, some lysosomal proteins do not have M6P residues. Owada and Neufeld (829) found that a human liver cell line devoid of iV-acetylglucosamine-l-phosphotransferase [and there fore unable to phosphorylate mannose residues (see Section V.D.I.a)], still targeted some lysosomal enzymes correctly with, at most, only slightly reduced efficiency. These cells may have a completely M6P-independent system for sorting lysosomal enzymes. All lysosomal membrane proteins are also devoid of M6P residues. Therefore, some lysosomal enzymes may have membrane-associated intermediates which are sorted to the lysosome together with authentic lysosomal membrane proteins. Barriocanal et al. (49) used immunocytochemistry to follow the fate of three lysosomal membrane proteins which they detected in lysosomes, the Golgi apparatus, and coated and uncoated vesicles in the region of the TGN. They found that oligosaccharide modifications were not required for lysosomal targeting (although they may be required to protect against proteolysis). Thus the sorting pathway for these proteins remains to be determined; they could be sorted completely independently of lysosomal enzymes or could be colocalized to the same coated vesicles by an M6Pindependent receptor and then segregated from recycling vesicle mem brane components in the prelysosome. Green et al. (397) have recently found that newly synthesized lysosomal membrane proteins appear in lysosomes with the same kinetics as newly synthesized plasma membrane proteins appear at the cell surface, making it unlikely that the former pass via the plasma membrane en route to the lysosome.

Although vacuoles are the functional equivalents of lysosomes in animal cells, the sorting of vacuolar enzymes is completely independent of M6P receptors. This was demonstrated most simply by the fact that tunicamy cin treatment did not affect vacuolar protein targeting (572,1025). How ever, at least one vacuolar protein does have phosphorylated mannose residues.

Most of the work on the sorting of vacuolar proteins has concentrated on the identification of sorting signals in yeast vacuolar proteins. Early studies showed that vacuolar proteases were proteolytically processed in two distinct stages, the first of which corresponded to the removal of a signal peptide (711). The second processing step is catalyzed by a vacuo lar protease, proteinase A, which removes an additional N-terminal poly peptide segment, the propeptide, from other vacuolar enzymes. Protein ase A is also autoactivated by the same mechanism (15,1206). The second processing step is blocked by certain sec mutations, which cause secre tory proteins to accumulate in the RER or Golgi apparatus, whereas mutations which affect the final stage of the secretory pathway from the Golgi to the cell surface do not affect vacuolar protein sorting or process ing (1082). This implies that the Golgi is the site of sorting of vacuolar and secreted or plasma membrane proteins in yeast cells.

Gene fusion studies conclusively demonstrated that propeptides are vacuolar sorting signals. Bankaitis et al. (44) and Johnson et al. (541) found that at most 50 N-terminal residues of preprocarboxypeptidase Y (CPY), including the 20-residue propeptide, could target the normally secreted enzyme invertase into the vacuole. Similar results were obtained with proteinase A-invertase hybrids (572). Vails et al. (1137) subse quently found that mutations affecting the sequence of the proCPY pro peptide caused the enzyme to be secreted in an inactive form. There appears to be no sequence similarity between the propeptides of different yeast vacuolar enzymes, even though genetic studies described below suggest that they are sorted into the vacuole via a common pathway. Part of the propeptide may be required to maintain vacuolar enzymes in an inactive form until they reach the vacuole and may also maintain the precursors in a competent conformation for transport through the secre tory pathway.

The overproduction of vacuolar proteinase A (972) and of CPY-inver-tase hybrids (44) causes them to be secreted into the medium, suggesting that some component of the vacuolar sorting pathway (e.g., a receptor) had been saturated. These observations led to the development of tech niques for selecting mutants which secreted CPY or CPY-invertase with out overproduction (Section II.F.l.b). Mutations in over 50 different genes (called VPL or VPT) have been identified (44,939a,971) . The extent of the sorting defect varied in different mutants: Some of them, for exam ple, did not affect the targeting of proteinase A, and none of them affected the sorting of the vacuolar membrane protein α-mannosidase, which is presumably sorted to the vacuole by an alternate pathway (971). It is unlikely that any of the mutations affected protein retention in the vacuole because vacuolar enzymes were not terminally processed, and the kinet ics of CPY secretion were comparable to those of a normally secreted protein. The characterization of the VPT or VPL gene products and their localization in the cell could provide revealing insights into the mecha nisms of vacuolar protein targeting, but at present we can only speculate on their roles. Obvious candidates are the propeptide receptor, vacuolar or Golgi ATPases which might produce a low pH environment necessary for sorting or receptor dissociation, or proteins involved in vesicle fission and fusion. Much less work has been done on plant vacuolar proteins. Tague and Chrispeels (1100) found that the plant vacuolar storage protein phytohemagglutinin was targeted mainly to the vacuole when its structural gene was expressed in yeast cells. This protein does not have a cleavable propeptide (it does have a signal peptide, which was at least partially processed in yeast cells), which means that the vacuolar plant sorting signal which is recognized by the yeast vacuolar sorting pathway is lo cated in the mature part of the phytohemagglutinin polypeptide. mRNA coding for a second plant storage protein, globulin P, has been microinjected into frog oocytes, which secreted the protein into the medium (51). This confirms that plant vacuolar proteins are bona fide secretory proteins and that plant cells have a special branch of the secretory pathway which shunts storage proteins into vacuoles.

As we have seen, protein-coated vesicles are involved in transporting secretory proteins at various stages of the secretory pathway. Some vesi cles (e.g., lysosomal sorting vesicles and immature secretory granules), are coated with a protein complex called clathrin, which also coats endocytotic vesicles (Chapter VIII). Other secretory vesicles (e.g., those me diating protein transport through the Golgi) have a different type of pro tein coat (797).

Clathrin, which is composed of equimolar amounts of heavy and light chains, forms a three-layered cage which envelops vesicles in a shell with fibrous interconnections, which give it mechanical strength and stability. The vesicle membrane is thought to have receptors which anchor the clathrin cage to the surface via a number of ancillary assembly proteins which probably act as bridges (1148, 1149) . Different assembly proteins are found in different classes of clathrin-coated, TGN-derived, and endocytic vesicles, suggesting that they might contribute to their respective specificity for particular membrane targets (7).

The clathrin coat probably prevents intimate contact between fusing membranes and must thus be removed to allow fusion to occur. This may explain, for example, the disappearance of the clathrin coat from matur ing secretory granules (Section V.G.4.b) and the presence of lysosomal protein precursors in naked as well as coated vesicles. Vesicles coated with other proteins are also presumably uncoated to allow fusion to oc cur. The uncoating of clathrin-coated vesicles is mediated by the ATPdependent cytosolic "uncoating" protein, which remains attached to the released clathrin (1010). Soluble clathrin retains its typical triskelion con formation. Uncoating protein is a member of a highly conserved group of stress proteins and may be related to HSP70 heat shock proteins involved in other stages in secretory protein transport and in mitochondrial protein import (Sections III.C.3 and VLB.5). The significance of this observation remains to be determined (855,967), but one possibility is that the ATPase (HSP70) is required to activate an uncoating enzyme which is already present in the clathrin complex.

The action of uncoating protein must be triggered in some way to pre vent it from destroying protein coats on immature vesicles or on coated buds, but the nature of the signal remains to be determined. The require ment for ATP for uncoating activity could in part explain the observed requirement for the same nucleotide during the transport of proteins be tween different stages of the secretory pathway if a similar activity is required to remove other, nonclathrin coats.

Attempts to determine the role of clathrin in protein targeting in yeast cells have given ambiguous results. Yeasts are known to have clathrin, and coated vesicles have been observed, but it is not known whether clathrin coats vesicles involved in secretory protein targeting (1165). The gene for the clathrin heavy chain was independently cloned by two groups who used it to inactivate the chromosomal gene to study the effect of the absence of clathrin on cell growth and protein secretion. Payne and Schekman (852) and Payne et al. (853,853a) reported that their mutants grew somewhat more slowly than wild-type cells, secreted invertase at a slightly reduced rate, and were partially defective in prepro-a-factor pro-cessing. The mutants accumulated unusual vacuoles, vesicles, and Golgiderived structures. These results suggested that the absence of clathrin did not completely impair plasma membrane growth and protein secretion but that there was nevertheless a reduced rate of transit of secretory proteins through the later stages of the secretory pathway. Different results were obtained using exactly the same approach by Lemmon and Jones (627). They found that cells lacking the clathrin heavy chain were not viable unless they also carried a suppressor mutation. Even with this mutation, the cells grew slowly, were larger and rounder, had an unusual granular appearance, tended to aggregate in liquid culture, and were poly ploid. These results suggest that the absence of clathrin is highly detri mental to yeast cells, making it difficult to determine whether clathrin plays a specific role in protein secretion in this organism.

The secretory pathway is almost certainly the route by which the vast majority of secreted and plasma membrane proteins are exported by eukaryotic cells. However, there is increasing evidence that some plasma membrane and secretory organelle proteins may reach their final destina tions directly rather than via the secretory pathway. Examples of this class of proteins include RAS-like GTP binding proteins such as SEC4 (987) and RAS2 (232) of the yeast S. cerevisiae and similar proteins from complex eukaryotes (1194), mating pheromones in yeast and some fungi (726, 886, 983) , capsid proteins of picornaviruses (851), and src proteins of Rous sarcoma and other transducing viruses (1029). Most if not all of these proteins are fatty acylated. Two different amino acids seem to be modified: N-terminal glycines, which are substrates for myristoyl CoA protein 7V-myristoyltransferase (1124), and C-terminal cysteines in RASlike proteins (232). These Cys residues are reported to be palmitoylated, but a farnecyl residue has been found in basidiomycete pheromones (983). The absence of the fatty-acylated amino acid disrupts membrane associa tion of RAS and src proteins (232, 555, 1029, 1194) , indicating that fatty acids probably anchor these proteins in their respective membranes. It remains to be determined how fatty-acylated proteins actually cross the plasma membrane (as in the case of the fungal pheromones) or what determines their specificity for certain membranes. A further example of a secreted protein which does not have a secretory routing signal is interleukin 1, but very little appears to be known about how this protein crosses the plasma membrane.

There is general agreement concerning the events which lead to the sort ing of secretory proteins into different terminal branches of the secretory pathway, as illustrated in Fig. V. 6, but we are clearly a long way from understanding exactly what directs proteins to their specific targets. Patch signals are undoubtedly necessary for sorting soluble proteins (other than lysosomal enzymes) into the various branches of the secretory pathway, but these will be difficult to identify by gene fusion techniques. At present, we have no clear idea of the extent to which the sorting vesicles have different membrane contents, but it seems probable that specific groups of membrane proteins (lysosomal, secretory granule, apical, and basolateral) accumulate at different sites in the membrane of the TGN from which sorting vesicles bud. This specialization is presumably also determined by protein-protein interactions, but the possibility that other interactions (e.g., protein-lipid) might be involved should not be over looked.

Another interesting observation is that ATP seems to be required at almost every stage in the secretory pathway. ATP has been proposed to act in a variety of ways, including acidification of secretory organelles and vesicles, phosphorylation of receptors or ligands, protein folding and "proofreading," activation of cytoskeletal motors, and vesicle uncoating activity. There is increasing evidence that GTP and GTP binding proteins (G proteins) are also involved at several stages in the secretory pathway. GTP binding proteins are also known to be involved in the generation of other intracellular signals, such as the activation or inactivation of adeny late cyclase, the activation of cyclic GMP phosphodiesterase, and the control of phospholipase C action (769). By analogy, GTP binding pro teins may act as signals or to activate receptors or ligands to ensure vectorial transport through the secretory pathway (109).

Constitutive and regulated secretion of proteins

Progress in unravelling pathways, of Golgi traffic

Assembly of asparagine-linked oligosaccharides

The sorting of proteins to the plasma membrane in epithelial cells

Biosynthetic protein transport and sorting by the endoplasmic reticulum and Golgi

Protein localization and membrane traffic in yeast

Lysosomal enzymes and their receptors