key: cord-022354-aqtceqqo
authors: HUNTER, ERIC
title: Membrane Insertion and Transport of Viral Glycoproteins: A Mutational Analysis
date: 2012-12-02
journal: Protein Transfer and Organelle Biogenesis
DOI: 10.1016/b978-0-12-203460-2.50007-x
sha: 
doc_id: 22354
cord_uid: aqtceqqo

nan

The eukaryotic cell faces a continual problem of partitioning a variety of enzyme activities into different subcellular organelles where specific macromolecular reactions can occur. This subcellular organization requires specific intracellular targeting of macromolecules through the cytoplasm, by as yet ill-defined mechanisms, and through the secretory pathway via a more clearly defined vesicular transport mechanism (Sabatini et al. y 1982) . As obligate intracellular parasites with only limited genetic complexity, viruses must utilize the existing cellular transport mechanisms to colocalize their virion components in a part of the cell where assembly can take place. This problem of intracellular targeting is compounded for the enveloped viruses because they must transport capsid polypeptides through the cytoplasm and envelope components through the secretory pathway to a common point of assembly. Since these viruses depend on the preexisting host cell processes and possess lipid envelopes that are biochemically similar to cellular membranes, they can provide ideal systems for probing the cellular mechanisms involved in glycoprotein biosynthesis and transport.

The relatively simple structure of virions, the high level of expression of viral genes, the ease of molecularly cloning and manipulating those genes, together with the availability of conditional lethal mutants with defects in viral protein transport, confer several additional advantages for such studies. We have chosen an enveloped, RNA-containing virus, Rous sarcoma virus (RSV), for an analysis of viral glycoprotein biosynthesis and transport, because in addition to the aspects delineated above, the availability of molecularly cloned, infectious DNA copies of the genome of this retrovirus has also allowed us to pose questions about the role of the glycoproteins in virus assembly and virus infectivity.

Like most simple enveloped viruses, the retroviruses consist of a host membrane-derived, lipid bilayer that surrounds (envelopes) a protein capsid structure (Fig. 1) . The icosahedral capsid of RSV assembles as the virus particle buds from the plasma membrane of the cell, and so these two events are linked both temporally and spacially. Glycoprotein knobbed spikes extend from the surface of the virion, and it is generally thought that during virus assembly a specific interaction between the glycoproteins and capsid (and/or "matrix" protein) is required, since cellderived polypeptides are for the most part excluded from the budding structure.

The envelope glycoproteins span the lipid bilayer and are thereby divided into three distinct domains: an external, hydrophilic receptor-bind- ing domain that functions in virus-cell attachment; a hydrophobic membrane-spanning domain; and a hydrophilic cytoplasmic domain. In RSV two polypeptides, gp85 and gp37, make up this structure (Fig. 2) ; the external domain is primarily made up of the 341 amino acid long gp85 which contains regions that define the host-range and neutralization properties of the virus. The 198 amino acid long gp37 polypeptide, on the other hand, is a bitopic protein that anchors the envelope glycoprotein complex into the virion via disulfide linkages to gp85. It is less heavily glycosylated than gp85 (having only 2 versus 14 potential glycosylation sites) and contains two apolar regions in addition to the hydrophilic cytoplasmic domain. One apolar region is located near the amino terminus of gp37 and may be analogous to the fusion peptide of the hemagglutinin HA2 polypeptide of influenza virus that mediates viral entry into the cell. The Hatched regions represent the highly hydrophobic signal and anchor sequences found at the N and C terminus, respectively. The location of a nonpolar region that may be analogous to the fusion peptide of ortho-and paramyxoviruses is shown by the stippled box. The AUG at the start of the env orf was used to initiate translation of transcripts expressed in an SV40 vector (Wills et ai, 1984) , but during a virus infection the genomic length transcript is spliced such that the AUG and first 5 codons from gag (black box) are spliced into the env orf . In both cases the signal peptide is removed during translation, and cleavage of the polyprotein precursor to gp85 and gp37 occurs in the Golgi at the basic tetrapeptide, -Arg-Arg-Lys-Arg-. Branched structures denote potential N-linked oligosaccharide addition sites (Asn-X-Ser or Asn-X-Thr). (B) Orientations of gp85 and pg37 in the viral membrane are depicted schematically. In the electron microscope this structure is seen as a spiked knob, where gp37 is the spike and gp85 is the knob. second apolar region in gp 37 consists of a 27 amino acid long stretch of hydrophobic residues near the carboxy terminus that functions during translation to stop the movement of the protein into the lumen of the rough endoplasmic reticulum (RER) and to anchor the complex in the membrane. The general orientation and structure of the RSV glycoprotein is thus similar to that of the influenza virus hemagglutinin (HA) (Porter et al, 1979; Gething et al, 1980) , the vesicular stomatitis virus (VSV) G protein (Gallione and Rose, 1983) , and several membrane-spanning cellencoded glycoproteins.

The two viral glycoproteins of RSV are encoded by a single viral gene, env, and are translated in the form of a heavily glycosylated precursor polypeptide, Pr95 e/ii; . Since several processing and maturation events occur during the transport of the env gene products to the plasma membrane, they provide excellent markers for the subcellular compartments of the cell. A long (62 amino acid) amino-terminal signal peptide, which mediates translocation of the env gene product across the rough endoplasmic reticulum (RER), is removed cotranslationally from the precursor protein, and in the lumen of the RER 15-16 high-mannose core glycosylation units are added to the nascent Pr95. The marked increase in molecular weight (MW) (approximately 40K) that results from the addition of this endo-ß-TV-acetylglucosaminidase H (endo H) sensitive carbohydrate provides a clear indicator for the translocation event. Removal of glucose residues and some mannose moities appears to occur prior to transport of the protein to the Golgi, where further trimming of the mannose residues and addition of glucosamine, galactose, and fucose are observed. Cleavage of Pr95 to gp85 and gp37 takes place after galactose and prior to fucose addition; it thus provides an excellent marker for trans-Golgi locations. While the transit time of Pr95 from the RER to Golgi is quite long (t

 = 90 min) compared to other viral glycoproteins, movement through the Golgi appears rapid and precludes any dissection of individual compartments within this organelle. These major biochemical modifications to the env glycoprotein coupled with sensitive immunological probes allow a fairly accurate mapping of the cell's secretory pathway.

Several viruses assemble at points within the secretory pathway, and since in many cases the location of the viral glycoproteins defines the virus maturation point (Roth et al., 1983b; Jones et al., 1985; Gottlieb et al., 1986; Gahmberg, 1984; Kabcenell and Atkinson, 1985) these systems are proving useful in investigating the signals that target proteins to specific subcellular locations. Figure 3 is a schematic summary of the assembly points for the major groups of enveloped viruses. Herpes simplex virus (HSV), a complex enveloped DNA virus that encodes several glycoproteins (gB, gC, gD, gE) , buds into the nuclear envelope (Spear, 1985) . Virion glycoproteins synthesized on the RER appear to be transported to the nuclear membrane in an endo H-sensitive form where they are incorporated into nascent virions (Compton and Courtney, 1984) . It has been postulated that intact HSV virions traverse the secretory pathway thereby exposing the glycoproteins to the entire array of carbohydrate-modifying enzymes such that the mature virion contains glycoproteins with complex carbohydrate side chains (Compton and Courtney, 1984; Spear, 1985) . A majority of the HSV glycoproteins can be found on the surface of infected cells, suggesting that targeting to the nuclear membrane is not absolute. Expression of the cloned HSV gD gene in the absence of other viral components resulted in more rapid Alphaviruses Ortho-, Paramyxoviruses transport to the plasma membrane and reduced accumulation on the nuclear membrane, indicating that interactions with other viral-encoded proteins may be required for normal nuclear membrane localization (Johnson and Smiley, 1985) .

The corona-, flavi-, and rotaviruses have been reported to undergo assembly at the RER (Dubois-Dalcq et aL, 1984) , but the rotaviruses appear to target this organelle most specifically. While the mature rotavirus is not enveloped, it contains a glycosylated capsid protein in its outer shell that is derived from a transient membrane during the assembly process. The inner and outer protein shells of these viruses form sequentially and by very different mechanisms. The inner capsids, containing the genomic RNA segments, are the equivalent of nucleocapsids of other viruses and assemble in the cytoplasm at the edge of electron-dense inclusions called "viroplasm" (Petrie et aL, 1982) . They acquire a transient envelope, or pseudoenvelope, by budding at RER membranes adjacent to the viroplasm. Further maturation of rotaviruses occurs within the cisternae of the RER where enveloped particles are converted to mature double-shelled virions and the lipid bilayer is removed by a process that remains to be elucidated (Dubois-Dalcq et aL, 1984) . VP7, the glycosylated protein found in the outer capsid, has been shown to have carbohydrate structures consistent with its RER location and to target specifically to this organelle when expressed in the absence of other viral proteins from a recombinant expression vector (Kabcenell and Atkinson, 1985; Poruchynsky et aL, 1985) .

The Golgi body is the site of assembly for several viruses. Coronaviruses, for example, mature by budding into the lumina of RER or Golgi cisternae; virions form as the intracytoplasmic, helical neucleocapsids align under regions of intracellular membranes containing viral proteins. Two glycoproteins, El and E2, comprise these membrane-associated polypeptides. E2 forms the large peplomers or spikes characteristic of coronaviruses and is a multifunctional molecule, being responsible for virus-induced cell fusion, binding of the virion to receptors on the plasma membrane of susceptible cells, and for inducing neutralizing antibody (Dubois-Dalcq et aL, 1984) . El, in contrast, is an unusual polypeptide; it has only a short amino-terminal domain, which contains the glycosylation sites of the protein, and two long stretches of hydrophobic amino acids, suggesting that it may traverse the membrane more than once (Dubois-Dalcq et aL, 1984; Boursnell et aL, 1984) . During infection, E2 can be transported through the secretory pathway to the plasma membrane, whereas El is transported only as far as the Golgi apparatus, where it accumulates during the infection cycle (Sturman and Holmes, 1983) ; it has been suggested that this restricted intracellular movement of El ac-counts for the intracellular budding site of coronaviruses (Sturman and Holmes, 1983) .

Members of the bunyavirus family also mature intracellularly, by budding at the Golgi complex (Bishop and Shope, 1979; Dubois-Dalcq et al., 1984) . By immunofluorescent microscopy, Kuismanen and colleagues showed that in cells infected with Uukuniemi virus the Golgi region underwent an expansion and became vacuolized . Both glycoproteins, Gl and G2, accumulated in the Golgi region during virus infection; neither polypeptide could be chased out of the Golgi even after a 6-hr treatment with cycloheximide (Gahmberg et al., 1986) , conditions that would allow complete transport of the Semliki Forest virus membrane proteins from the Golgi (Green et al., 1981a) . Furthermore, the glycoproteins of a temperature-sensitive strain of Uukuniemi virus were retained in the Golgi even under conditions where no virus maturation took place and no nucleocapsids accumulated in the Golgi region (Gahmberg et al., 1986) . Thus intracellular targeting of these viral components appears to be independent of other viral components and of the assembly process itself. Moreover, it supports the concept that it is the glycoproteins themselves that dictate the cellular site of virus maturation.

For several virus groups, virion assembly does not occur until the envelope components have traversed the entire secretory pathway. Thus the ortho-and paramyxoviruses, rhabdoviruses, alphaviruses, and retroviruses mature at the plasma membrane. Even these proteins, however, possess additional membrane-targeting information such that in polarized epithelial cells, where different cell proteins are inserted in the apical and basolateral membranes, the different viruses assemble from distinct membranes; for example, the ortho-and paramyxoviruses bud from apical membranes, rhabdo-and retroviruses from basolateral membranes (Rodriguez-Boulan and Sabatini, 1978; Herrler et al., 1981; Roth et al., 1983a; Rindler et al., 1985) . As with those viruses that mature at points within the secretory pathway, it is the glycoproteins themselves that appear to specify the specific plasma membrane location for virus assembly, since glycoproteins expressed from recombinant expression vectors are transported in a polarized fashion (Roth et al., 1983b; Jones et al., 1985; Gottlieb et al., 1986; . The problem of polarized expression will be dealt with in more detail later in this chapter (see Section II, B, 4) .

From the brief outline presented above it is clear that viral envelope components provide a plethora of systems for studying intracellular protein targeting. Recombinant DNA approaches, described below, are already providing information on the role of the different glycoprotein domains in this important aspect of viral and cell biology.

In addition to utilizing the secretory pathway of vertebrate cells for transporting viral components to the point of virus maturation, several enveloped viruses take advantage of a second vesicle-mediated transport system, the endocytic pathway, to gain entry into susceptible cells. These viruses, such as orthomyxoviruses, rhabdoviruses, and togaviruses, in contrast to paramyxoviruses, such as Sendai virus, which bind and fuse with the plasma membrane of the host cell, bind to the host cell surface and are subsequently internalized by endocytosis. This latter process serves an important role in the normal uptake of nutrients and in the internalization of receptor-bound ligands such as hormones, growth factors, lipoproteins, and antibodies (Mellman et al., 1986; Hopkins, 1983) . Bound virions are carried into clathrin-coated pits, which form continually on the surface of the cell, and which fold inward and pinch off into the cytoplasm to form "coated vesicles." As the coated vesicle moves into the cytoplasm, it loses its clathrin and fuses with an endosome, a large acidic vacuole with a smooth outer surface. For viruses entering by this pathway, membrane fusion occurs in the endosomal compartment (Marsh, 1984; Yoshimura and Ohnishi, 1984) . Fusion is triggered by the mildly acidic endosomal pH and is catalyzed by the virally encoded glycoproteins which undergo a low pH-dependent configurational change (Skehel et al., 1982; Kielian and Helenius, 1985) . The pH dependence of fusion varies among virus types, with the optimal pH for fusion generally falling within the range of pH 5.0-6.2 for endocytosed viruses (White et al., 1983) .

In order to obtain an understanding of the molecular mechanisms involved in these low pH-induced fusion reactions, virus mutants have been isolated which fuse with pH optima different from those of their respective parents. Through the use of an elegant selection scheme, in which mutagenized virus was allowed to fuse with nuclease-filled liposomes at a pH below 6.0, Kielian et al. (1984) isolated the first such fusion mutant of Semliki Forest virus. This virus, fus-1, fused at a pH optimum 0.7 pH units lower than that of the wild type (pH 5.5 versus 6.2). The mutant was, nevertheless, fully capable of infecting cells under standard infection conditions and even under conditions that prevent fusion of endosomes with lysosomes. On the other hand, the fus-1 mutant showed increased sensitivity to lysosomatropic agents that increase the pH in acidic vacuoles of the endocytic pathway. In addition to proving that alterations within viral structural components can significantly affect the pH at which virus-induced fusion can occur, these results showed that a pH below 5.5 exists within the endosomal compartment and thereby demonstrated the usefulness of mutant viruses as biological pH probes of this pathway.

In parallel studies, Rott and co-workers (1984) have shown that variants of the X31 strain of influenza virus, selected for their ability to undergo activation cleavage and growth in Madin-Darby canine kidney (MDCK) cells, have an elevated fusion pH threshold (approximately 0.7 pH units higher than the wild type). Similar virus variants have been selected by growth of influenza virus in the presence of amantadine, a compound that raises endosomal pH (Daniels et al., 1985) . In this latter study, viruses were obtained that fused at pH values 0.1-0.7 units higher than the parental strain. Analogous mutants have been reported to occur naturally within stocks of the X31 strain of influenza virus (Doms et al., 1986) . Such mutants should provide useful probes for elucidating the endocytic pathway.

The mechanisms by which cells send membrane-bound and secreted proteins to their proper subcellular locations remain a central problem in cell biology. It has been postulated to involve the specific interaction of "sorting signals," located within the structure of the newly synthesized proteins, with membrane-bound receptors in the RER and Golgi apparatus of the cell (for review, see Sabatini et al., 1982; Silhavy et al., 1983) . This concept is supported by the facts that cell, as well as viral, glycoproteins can be retained at or targeted to specific points in the secretory pathway and that cells can transport and secrete a variety of glycosylated and nonglycosylated proteins at distinctly different rates (Strous and Lodish, 1980; Fitting and Kabat, 1982; Gumbiner and Kelly, 1982; Ledford and Davis, 1983; , Kelly, 1985 .

Very little is known about the composition(s) or indeed the exact role of sorting signals, but it is generally thought that they are composed of protein. Clearly, the initial step that introduces polypeptides into the secretory pathway is mediated by the interaction of a sequence of amino acids (the signal sequence) within the polypeptide and the signal recognition particle (SRP)/docking protein (DP) complex (Blobel and Dobberstein, 1975; Blobel, 1980, 1981a,b; Walter et al., 1981; Meyer and Dobberstein, 1980; Meyer et al., 1982; Gilmore et al., 1982a,b; . Mutants of secreted proteins that are defective in later stages of transport have been identified that differ from the wild-type forms by one (Mosmann and Williamson, 1980; Wu et al., 1983; Shida and Matsumoto, 1983) or two (Yoshida et al., 1976; Hercz et al., 1978) amino acid substitutions, supporting the concept that sorting signals are composed of protein. Also, several conditional transport-defective mutants of membrane-bound viral glycoproteins have been identified (for example, Knipe et al, 1977b; Zilberstein et al, 1980; Pesonen et al, 1981) .

Studies using tunicamycin, an inhibitor of glycosylation, suggest that carbohydrate moieties are not recognized directly by the sorting machinery but may be important for maintaining the proper secondary or tertiary structures of (protein-composed) sorting signals (Struck et al., 1978; Gibson et al, 1978 Gibson et al, , 1979 Leavitt et al, 1977; Hickman et al, 1977; Roth et al, 1979; Strous et al, 1983; Green etal, 1981b) . The lack of a direct role for carbohydrate moieties in the sorting process is perhaps to be expected in view of the fact that many secreted proteins are not glycosylated at all (for example, Strous and Lodish, 1980; Underdown et al, 1971) . Nevertheless, addition of carbohydrate to molecules that are unable to be transported through the secretory pathway can release the block to their transport (Guan et al, 1985; Machamer et al, 1985) . In addition, the transport of certain hydrolases to the lysosome (and away from the secretory pathway) does appear to require the addition of a carbohydrate moiety (mannose 6-phosphate) (Hasilik and Neufeld, 1980; Sly and Fisher, 1982; Creek and Sly, 1984) , but these additions in turn must require the recognition of signals within the polypeptide chains.

Even less is known about the intramolecular location(s) of sorting signals. In the case of the membrane-spanning glycoproteins, three protein domains exist which together or separately may harbor sorting signals: (1) the internal or cytoplasmic domain, (2) the hydrophobic or transmembrane domain, and (3) the extracytoplasmic or external domain. Since most secreted proteins (which may be cotransported with membranebound glycoproteins, Strous et al, 1983) possess only external domains, it might be reasonable to expect the transmembrane and cytoplasmic domains to be unimportant to the sorting process. We have tested this hypothesis by introducing genetic lesions into the gene encoding the envelope glycoproteins of RSV (Wills et al, 1983 (Wills et al, , 1984 Hardwick et al, 1986; , as have others for the VSV G protein Bergmann, 1982, 1983; Rose et al, 1984; Adams and Rose, 1985a,b) , the influenza virus hemagglutinin protein (Sveda et al, 1982 Gething and Sambrook, 1982; Doyle et al, 1985 Doyle et al, , 1986 Gething et al, 1986) , and the glycoproteins of Semliki Forest virus (Garoff et al, 1983; Garoff, 1985; . Genetic analyses of protein transport in prokaryotic systems have provided both support for the role of the signal peptide in protein translocation and valuable insights into the polypeptide interactions that are required for the intracellular targeting of bacterial secreted and membrane proteins (reviewed by Michaelis and Beckwith, 1982; Silhavy et al, 1983; Benson et al, 1985; Oliver, 1985) . While similar experiments are more difficult to perform in eukaryotic cells with the enveloped virus systems described here, both the classic and molecular genetic approaches outlined below are providing information on the role of different protein domains in the transport process.

During the genetic analysis of enveloped virus replication through the isolation and biochemical characterization of spontaneous and mutageninduced variants, complementation groups were established for several viruses that contained mutants defective in normal transport of the viral glycoproteins (Knipe et al., 1977a,b; Zilberstein et al., 1980; Pesonen et al, 1981; Gahmberg, 1984; Ueda and Kilbourne, 1976) . The existence of conditional lethal mutants that were blocked at different stages of virus glycoprotein maturation suggested that the viral polypeptides themselves might contain the signals necessary for normal sorting by the cells' transport machinery and raised the possibility that such mutants could be used tq dissect the maturation pathway of a glycoprotein. Since it is impossible in this chapter to provide a detailed review of the characterization of mutants in each of these systems, and since the transport of the influenza virus glycoproteins has recently been discussed in detail by , this section will concentrate primarily on mutants of the VSV G protein gene as an example of these approaches.

Temperature-sensitive (ts) mutants in complementation group V of this VSV have defects in the structural gene for the viral glycoprotein, G, and cells infected at the nonpermissive temperature with such mutants produce markedly reduced yields of virus like particles which are noninfectious and specifically deficient in G protein. At the nonpermissive temperature the mutant G polypeptide is synthesized normally; however, it does not accumulate on the cell surface, nor is it incorporated into virions (Knipe et al., 1977b; Zilberstein et al., 1980) . The ts(V) mutants can be subdivided into two subclasses with respect to the stage of posttranslational processing at which the block occurs (Zilberstein et al., 1980) . Three mutants, tsL5\3, tsM50l, and ts045, encode G proteins that at the nonpermissive temperature are blocked at an early, pre-Golgi step of the secretory pathway. While insertion into the ER membrane, removal of the amino-terminal hydrophobic signal sequence, and addition of the two N-linked high-mannose core oligosaccharides occur in a way that is indistinguishable from the wild type, all subsequent Golgi-mediated carbohydrate processing reactions are blocked (Zilberstein et al., 1980) . These results are consistent with subcellular fractionation and immunoelectron microscopy studies which indicated that the G protein in taM501 or ta045-infected cells was arrested in its transport from the RER to the Golgi complex at the nonpermissive temperature (Zilberstein et al., 1980; Bergmann et al., 1981; Bergmann and Singer, 1983) . The defect in transport in these mutants is a reversible phenomenon, thereby excluding irreversible denaturation as the basis for lack of movement; proteins synthesized at the nonpermissive temperature rapidly move by stages to the plasma membrane upon shift to the permissive temperature Bergmann and Singer, 1983) . Within 3 min after shift to 32°C, G protein of tsOA5 could be seen by immunoelectron microscopy at high density in saccules at one face of the Golgi complex and by 3 min later was uniformly distributed through the complex (Bergmann and Singer, 1983) . Movement of the mutant proteins to the cell surface occurred rapidly and was accompanied by incorporation into virions .

The second class of ts mutants of VSV is represented by taL511. G protein encoded by this mutant is transported normally through most of the Golgi-mediated functions involved in the processing of carbohydrate side chains, including addition of the terminal sialic acid residues. However, this molecule does not undergo two posttranslation modification reactions that take place with the wild-type G protein. The first is the addition of fucose to the teL511 oligosaccharide chains, which is reduced at both permissive and nonpermissive temperatures (Zilberstein et al., 1980) . In the second, a molecule of palmitic acid (a 12-carbon fatty acid) is covalently attached to the wild-type G polypeptide near the membranespanning region Rose et al., 1984) . Attachment occurs at a late stage of maturation, just before oligosaccharide processing is completed (Schmidt and Schlesinger, 1980) , probably in the eis compartment of the Golgi complex (Dunphy et al., 1981) . This modification of the G protein does not occur at the nonpermissive temperature in cells infected with the tsL5ll mutant. Taken together with the almost complete processing of the mutant's oligosaccharide side chains, this suggests that at the nonpermissive temperature the AsL511 glycoprotein accumulates at a specific region within the Golgi complex.

The mutants of VSV, together with equivalent mutants in other viral systems that were blocked at different stages of the secretory pathway (Pesonen et al., 1981; Gahmberg, 1984; Kuismanen et al., 1984) , raised the possibility of identifying and understanding the nature of sorting signals in secreted polypeptides. With the advent of recombinant DNA and rapid nucleotide sequence techniques, it has been possible to determine at the amino acid level the basis for these defects (Gallione and Rose, 1985; Arias et al., 1983) , but the interpretation of this information with regard to protein transport has been less than straightforward. Gallione and Rose (1985) determined the nucleotide sequence of the ts045 mutant of VSV and compared it to that of the parent and a wildtype revertant. The mutant and revertant differed in three amino acid residues, and through the construction and expression of hybrid genes it was possible for these investigators to demonstrate that the basis of the temperature-sensitive phenotype was a single amino acid change of phenylalanine to serine. Since this polar substitution occurred within a very hydrophobic region of the G protein, it was suggested that it might significantly affect protein folding in this region such that reversible denaturation of the protein might occur at the nonpermissive temperature. This denaturation could prevent further transport to the Golgi apparatus and the cell surface. Alternatively, Gallione and Rose (1985) pointed out that the conformational change at the nonpermissive temperature might be more subtle, perhaps preventing recognition by a component of the protein transport machinery. However, since hydrophobic residues are generally buried within a protein (Kyte and Doolittle, 1982) , it is unlikely that the mutated sequence itself would play a direct role in such an interaction. Nothing is known about the 3-dimensional structure of the G protein or whether accessory proteins are involved at this stage of protein transport, thus both suggestions remains viable possibilities. A similar analysis has been carried out by Arias et al. (1983) , who sequenced the genes encoding the viral glycoproteins of tslO and ts23, mutants of Sindbis virus defective in the intracellular transport of their glycoproteins, and of revertants of these mutants. These investigators found ts23 to have a double mutation in glycoprotein El, while ts was a single mutant in the same glycoprotein. In each case reversion to temperature insensitivity occurred by changes at the same site as the mutation, in two cases restoring the original amino acid and in the third case substituting a homologous amino acid (arginine in place of lysine). Since the three mutations were far apart from each other in the protein, these authors concluded that the 3-dimensional conformation of El was very important for the correct migration of the glycoproteins from the ER to the plasma membrane. Similarly, two ts mutants in the HA gene of influenza virus that result in HA protein transport being arrested in the RER are also caused by single point mutations that probably disrupt the tertiary structure of the molecule (Nakajima et al., 1986) .

In summary, several conditional mutants have been isolated by classic genetic approaches that at the nonpermissive temperature disrupt the normal transport of viral glycoproteins through the secretory pathway. While these mutants carried the promise of defining specific protein domains that might interact with components of the transport machinery, the evidence from the nucleotide sequencing experiments outlined above suggests that many or all of the mutations may exert their phenotype through distortion of the 3-dimensional shape of the molecule. While this mechanism does not preclude a role for specific protein-protein interactions in the secretory pathway, it provides no direct evidence for it at this time.

The recent development of cDNA cloning, gene sequence manipulation, and gene expression technologies has opened up new approaches for localizing and characterizing those structural features of a protein that act as sorting signals. These new methodologies have allowed investigators to delete or modify potentially important structural regions of a protein at the nucleotide level and then to determine the effect of such changes by expressing modified genes in suitable eukaryotic expression vectors. Furthermore, in some instances it has been possible to test directly the functional role of a particular peptide region by fusing it to another protein and analyzing the behavior of the chimera. Since these general approaches to protein sorting have been reviewed recently (Garoff, 1985; Gething, 1985) , this chapter will describe our recent recombinant DNA analyses of the biosynthesis and transport of the RSV envelope glycoproteins, within the context of similar analyses in other viral systems.

As we have discussed earlier, the RSV env gene encodes two viral glycoproteins, gp85 and gp37, that mediate recognition of, attachment to, and penetration of the susceptible target cell. These proteins are synthesized as a glycosylated precursor protein, Pr95, that is proteolytically cleaved in the Golgi complex. The coding sequences for gp85 and gp37 have been placed in an open reading frame that extends from nucleotide 5054 to nucleotide 6863, and predict sizes of 341 amino acids (40,000 MW) for gp85 and 198 amino acids (21,500 MW) for gp37 ( Fig. 2) . Carbohydrate makes up a significant contribution to the observed molecular weights of these polypeptides-the predicted amino acid sequence contains 14 potential glycosylation sites (Asn-X-Ser/Thr) in gp85 and 2 in gp37. Experiments aimed at determining the number of carbohydrate side chains yielded results consistent with most or all of the sites being occupied .

Although an initiation codon is located early (codon 4) in the open reading frame, during a viral infection splicing yields an mRNA on which translation initiates at the same AUG as the gag gene to produce a nascent polypeptide in which gp85 is preceded by a 62 amino acid long leader (signal) peptide (Fig. 2) . This peptide contains a hydrophobic sequence that we have shown (see below) is necessary for translocation across the RER and is completely removed from the env gene product during translation . It represented one of the longest signal peptides described to date, and we were therefore interested in determining the signal peptide requirements for normal biosynthesis of gp85 and gp37. For these studies the env open reading frame was excised from the RSV genome and inserted into an SV40 expression vector under the control of the late-region promoter (Wills et al., 1983; . In this construction translation is initiated at the AUG present at the start of the open reading frame (at nucleotide 5054 of the RSV genome) and results in the synthesis of an even longer (64 amino acid) signal peptide; nevertheless, biosynthesis of Pr95 and signal peptide cleavage occur normally. Furthermore, expression of the RSV env gene in African Green monkey (CV-1) cells parallels that seen in a normal virus infection in avian cells (Wills et al., 1984; Hardwick et al., 1986) , making it an excellent system for the analysis of mutant env genes.

a. Deletion/Substitution of the Signal Peptide. In order to examine the role of the signal peptide in RSV glycoprotein biosynthesis we constructed a series of deletion mutations within the 5' coding region of the env gene using the double-stranded exonuclease BaBl. Oligonucleotide linkers of the sequence CATCGATG were ligated to the ends of the truncated molecules to introduce a unique restriction endonuclease cleavage site and to replace the deleted in-frame AUG. The mutants were then sized and their nucleotide sequence determined to find those with a suitable deletion and an in-frame AUG. One such mutant, Al, contained a deletion of 171 nucleotides within the env coding sequences and encoded an env product that completely lacked an aminoterminal hydrophobic sequence (Fig. 4) . Expression of this gene from the SV40 vector resulted in the synthesis of a nonglycosylated, 58 kDa cytoplasmic protein that was similar in size to the nonglycosylated wild-type env gene product produced in the presence of the glycosylation inhibitor, tunicamycin. In contrast to the tunicamycin product, however, the Al protein was not associated with membrane vesicles and was rapidly degraded (half-life < 5 min; E. Hunter, K. Shaw, and J. Wills, unpublished) . Thus the signals for initiating translocation of the RSV env gene product must reside within the cotranslationally removed amino-terminal sequence, and in their absence the molecule is synthesized as an unstable cytoplasmic protein. This result is similar to those obtained by Gething and Sambrook (1982) and by Sekikawa and Lai (1983) with the influenza virus HA gene product. Influenza hemagglutinin sequences are depicted by italicized text. Underlined text denotes amino acid residues encoded by oligonucleotide linkers. Arrows depict the signal peptidase cleavage site at which the signal is removed cotranslationally from each of the constructs. Plus symbols indicate that translocation, glycosylation, or transport to a plasma membrane location is observed; minus symbols mean that the properties above are not observed.

Since the signal peptide of the RSV env gene product is exceptionally long, it was of interest to determine whether another signal peptide could substitute for it. For these experiments we have utilized the signal sequences of the influenza virus HA gene (A/Jap/305/57; Gething et al., 1980) . Two constructions were made: in the first of these the Al deletion mutant coding sequence was fused in-frame to the HA signal coding sequence at the signal peptidase cleavage site of the latter (Fig. 4) and in the second, we made use of a Sail restriction enzyme site in the HA gene and an Xhol site in the RSV env gene, so that the signal sequence and 16 amino acids of HA1 were fused with env 6 amino acids into gp85 (Fig. 4) . Expression of these hybrid genes in CV-1 cells resulted in the biosynthesis of a glycosylated Pr95 protein that was transported to the Golgi complex, cleaved to gp85 and gp37, and displayed on the cell surface (E. Hunter, K. Shaw, and J. Wills, unpublished) .

To demonstrate that the Pr95 molecules expressed from the HA-A1 fusion gene had undergone signal peptide cleavage, Pr95 was immunoprecipitated from [ 3 H]leucine, pulse-labeled cells and analyzed by sequential Edman degradation, in order to determine the amino-terminal sequence. To our surprise the signal peptidase had cleaved at the HA cleavage site, despite the fact that according to the analyses of von Heijne (1983) only the RSV cleavage sequence should have been recognized. Thus the H A -Al fusion protein contains two potential signal peptidase cleavage sites (that from the HA and that remaining in the env sequences), but only the first of these is utilized. Both gene fusions, therefore, result in the synthesis of aberrant gp85 proteins-that from the HA-A1 fusion having an 8 amino acid amino-terminal extension, and that from the SallXho fusion having lost 6 amino-terminal amino acids and gained 16 from HA1which nevertheless can be transported to the plasma membrane (Wills et al., unpublished data) .

Reciprocal gene fusions, in which the env gene signal peptide was fused to the structural sequences of HA (Fig. 4) , also resulted in translocation of the HA molecule across the RER membrane, supporting the concept that this transient sorting sequence is not polypeptide specific. However, only in the construction where the signal sequence of env was precisely fused to the amino terminus of HA1 was transport beyond the RER observed (Wills et al., unpublished data) . In constructions where the amino-terminal sequence of HA1 was perturbed, the recombinant protein was apparently prevented from assembling into trimers and its transport was blocked in the RER . Thus, while signal peptides may be capable of mediating the translocation of foreign polypeptides across the RER, other sorting "signals" must be active for transport of the molecule to continue.

Site. The experiments described above have been extended to determine the following: (a) whether the hydrophobic region of the signal peptide carries all the information required for transfer of the env gene product into the RER; (b) what the structural specificities of the signal peptide are; and (c) where the specificity for signal peptidase cleavage is located. More than 200 prokaryotic and eukaryotic signal peptides have been sequenced (Watson, 1984) . Comparison shows that most extensions comprise 20-40 amino acid residues; one of the longest being that of the RSV envelope glycoprotein. There is no homology between sequences, but a characteristic distribution of amino acid chains is observed. Three structurally distinct regions have been observed so far: a positively charged aminoterminal region, a central region of 9 or more hydrophobic residues, and a more polar carboxy-terminal region that appears to define the cleavage site (von Heijne, 1983 (von Heijne, , 1984 (von Heijne, , 1985 Perlman and Halvorson, 1983) . The importance of these general features has been supported by the genetic studies in prokaryotic systems (reviewed by Silhavy et al., 1983; Benson et al, 1985) .

To investigate these questions we initially constructed a series of internal deletion mutants that initiated within the amino-terminus of gp85 and extended into the signal peptide. The deletion mutations were introduced into the coding region for the envelope glycoprotein by digestion of a plasmid containing the env gene at a unique Xhol site located 13 base pairs (bp) from the 5' end of the coding sequence for gp85, followed by digestion with the double-stranded exonuclease Bal3l. Potential mutants were identified by restriction enzyme analysis and DNA sequencing, and those of interest were engineered into the SV40 expression vector. These are depicted in Figs. 5 and 6. Mutants X4-A -B, and -C were derived from a single out-of-frame parent, and they represent a nested set of mutants in which the hydrophobic sequence varies from the wild-type length of 11 to only 9 amino acids. Expression of these mutant genes in CV-1 cells gave the results summarized in Fig. 5 . Mutant polypeptides with the shortest hydrophobic domain (X4-C) resembled the Al mutant polypeptides in that they had a cytoplasmic location, were nonglycosylated, and were rapidly degraded. They differed from the Al mutant in length (65 versus 58 kDa) confirming that the mutated signal was not removed. Mutant X4-A polypeptides, on the other hand, were translocated and glycosylated with an efficiency equivalent to wild type, despite the substitution of serine and isoleucine for leucine and cysteine residues within the hydrophobic domain. Mutant X4-B expressed a phenotype intermediate between that of X4-A and X4-C, approximately 50% of the polypeptides being translocated and glycosylated. None of the mutants contain the sequences that specify the signal peptidase cleavage site, and molecules of X4-A/X4-B that were translocated across the RER retained an uncleaved signal peptide.

The data from these mutations suggest the following: (1) that the length, rather than the amino acid composition, of the hydrophobic domain of the env signal peptide is critical for translocation across the RER and (2) that signal peptide cleavage is not a requirement for translocation. The first of these conclusions is supported by genetic experiments in prokaryotic systems, where a requirement for secondary structure in the signal peptide was suggested Bankaitis et al., 1984) . The second is consistent with the presence of permanent insertion sequences in secreted and membrane-spanning proteins that are translocated across the RER membrane without removal of an amino-terminal signal peptide (Palmiter et al., 1978; Bos et al., 1984; Markoff et al., 1984; Zerial et al., 1986; Spiess and Lodish, 1986) . Several membrane-spanning proteins are anchored in the membrane by an amino-terminal anchor/signal domain and display what is termed group II protein topology (Garoff, 1985; Wickner and Lodish, 1985) , where the amino terminus of the protein is cytoplasmic and the carboxy terminus is luminal.

After translocation of a nascent chain across the endoplasmic reticulum has been initiated, the signal peptide is removed. This cleavage is carried out by signal peptidase, a cellular gene product. Two classes of signal peptidases have been described. A signal peptidase of Escherichia coli (SPase I) has been cloned into pBR322 (Date and Wickner, 1981) , and has been shown to accurately cleave eukaryotic precursor proteins as well as bacterial protein precursors (Talmadge et al., 1980) . Conversely, the eukaryotic signal peptidase will accurately cleave prokaryotic proteins (Watts et al., 1983) . The latter enzyme has been studied, using detergentsolubilized dog pancreas signal peptidase (Jackson and White, 1981) and hen oviduct signal peptidase (Lively and Walsh, 1983) , and been demonstrated to be an integral membrane protein that can be solubilized only when the lipid bilayer is dissolved. A second prokaryotic signal peptidase (E. coli SPase II) has been described that is specific for prolipoproteins (Hussain et al., 1982; Tokunaga et al., 1982) and membrane-bound penicillinases (Nielsen and Lampen, 1982) . This enzyme maps to a different locus on the E. coli genome and requires a glyceride-modified cysteine for cleavage. Perlman and Halvorson (1983) and von Heijne (1983) have examined sequences of a number of membrane proteins and have described amino acid sequence patterns that allow prediction of signal peptidase cleavage sites with greater than 90% accuracy. The most striking feature of signal peptidase cleavage sites is the presence of an amino acid with a small, uncharged side chain at the carboxy terminus of the signal peptide. The most common amino acids found at this position are alanine and glycine. From statistical analyses, the peptidase cleavage site appears to be determined by sequences within the signal peptide and not by sequences beyond the cleavage site. This is in contrast to the observations that mutations within the structural protein itself prevent signal peptidase cleavage of the lamB gene product and the M13 coat protein (Emr and Bassford, 1982; Benson and Silhavy, 1983; Rüssel and Model, 1981) .

To investigate this question with regard to the RSV env gene product, the deletion mutants shown in Fig. 6 were constructed as described above. Mutant XI has a 14 amino acid deletion encompassing residues 4-17 of gp85, which results in the loss of one potential glycosylation site. This deletion resulted in the synthesis of a slightly smaller precursor polypeptide that lacked one carbohydrate side chain but was otherwise glycosylated normally. Based on size estimations in pulse-labeling experiments, in the presence and absence of tunicamycin, the precursor polypeptide lacked the long, signal-containing leader peptide. Thus, although the mutation in XI significantly alters the sequences near the signal peptidase site, the signal peptidase still recognized and removed the signal peptide. In mutants X2 and X3 the amino-terminal nine and six amino acids, respectively, of gp85 are deleted. They therefore encode gp85 poly- Xhol site in DNA r\-

 5 | 10 15 °| asp val his leu leu glu gin pro gly asn leu trp ile thr trp ala asn arg . 1 Fig. 6 . Amino acid sequence deduced from DNA sequence of mutants XI, X2, and X3. The RSV glycoprotein is schematically represented: the location of the hydrophobic signal sequence within the long (64 amino acid) leader peptide is denoted by a stippled bar, and the mature gp85 glycoprotein by a hatched bar. The signal peptidase cleavage site in both the cartoon and the numbered amino acid sequence is denoted by a long arrow. The potential glycosylation site in the amino terminus of gp85 is shown as a CHO. The amino acid sequence of the last 7 amino acids of the signal peptide and the amino-terminal 18 amino acids of gp85 are shown for the wild-type gene product. The solid black bars show the lengths and positions of the deletions in mutants XI, X2, and X3. peptides with novel amino termini that alter the signal peptidase cleavage site from Ala/Asp-Val-His to Ala/Asn-Leu-Trp and Ala/Gln-Pro-Gly, respectively. Thus both the charge and secondary structure of the cleavage site would be predicted to be altered by the loss of Asp and His (X2) and by the relocation of proline near the cutting site (X3). Nevertheless, the signal peptidase efficiently cleaved the leader peptide from the nascent polypeptide, and its specificity of cleavage was unaffected by these alterations (Hardwick et aL, 1986) . These experiments then support the generalized conclusion from statistical analysis, that the sequence to the right of the signal peptidase site is not critical for signal peptidase cleavage.

None of the deletions generated in the RSV glycoprotein gene resulted in a loss of recognition and cleavage by the signal peptidase. This result contrasts with previously described prokaryotic mutants. A 12 amino acid deletion starting at the fifth residue beyond the signal peptidase site of the lamB gene product blocked cleavage of the signal (Emr and Bassford, 1982; Emr et al, 1981) , and a deletion of 130 amino acids beginning 70 amino acids downstream from the signal abolished signal cleavage although the shortened protein was localized correctly . In addition, the substitution of a leucine, in place of the glutamic acid, at residue 2 of the mature M13 coat protein also inhibited signal peptidase cleavage; however, in this latter instance the procoat protein was transported inefficiently across the inner membrane (Boeke et al., 1980; Rüssel and Model, 1981) . Although more mutants will be required to properly define these systems, the prokaryotic cleavage site appears to be more sensitive to manipulation than that of eukaryotes. There is accumulating evidence that transported prokaryotic proteins, unlike those of eukaryotes, may not be transferred across membranes in a strictly cotranslational manner (Randall and Hardy, 1984) . Thus, altered regions within the structural protein portion of a molecule would have the opportunity to interact and interfere with signal peptidase cleavage; such as interaction would not be possible in the cotranslational system described for eukaryotes.

Although the mutant XI polypeptides were translocated across the RER membrane in a normal fashion, immunofluorescence experiments and posttranslational modification probes indicated that the transport and maturation of the XI glycoprotein was halted shortly after exiting the RER, perhaps within pre-or cis-Golgi vesicles. Cells synthesizing the mutant protein showed no surface immunofluorescence, no cleavage of the Pr95 to gp85/gp37, and no terminal sugar additions (Hardwich et al., 1986) . The basis for this block appears to be the altered amino acid sequence rather than the loss of the carbohydrate side chain since by using a mutagenic oligonucleotide we have modified the amino terminus of the XI gp85 from Asp-Val-His-Arg-Thr-to Asp-Val-Asn-Arg-Thr-, thereby reinserting the glycosylation site missing from this mutant. The derivative mutant, X1A, is glycosylated at this site but remains blocked at the same intracellular point in the secretory pathway (K. Shaw, K. Kervin, and E. Hunter, unpublished) . A second deletion mutant of the RSV env gene is also blocked in intracellular transport. This mutant, C3, has an engineered deletion at the carboxy terminus of gp37 that removed the cytoplasmic tail and transmembrane region (see below and Wills et al., 1984) . Its transport is clearly blocked at an earlier stage than that of the XI mutant since it was localized to the ER and never reached the Golgi apparatus, whereas by immunofluorescent staining of fixed cells the XI protein appeared to colocalize with the Golgi complex (see Fig. 7 ; Hardwick et al., 1986) .

Although the XI and C3 mutants contain alterations at opposite ends of the env gene product, they both appear to lack an element that normally signals their transport to and beyond the cis-Golgi. While there may be a specific amino acid sequence (analogous to the amino-terminal signal sequence) that is required for these transport steps, it is as likely that a correctly aligned tertiary structure is the critical factor. Just as small changes as the amino terminus of HA1 can disrupt assembly and transport UN Rhodamine Fig. 7 . Intracellular immunofluorescence of cells expressing wild-type and mutant polypeptides. Fixed infected CV-1 cells were stained to detect the intracellular localization of wild-type and mutant glycoproteins. Rabbit anti-glycoprotein antibodies were tagged with fluorescein-conjugated goat anti-rabbit antibodies which in wild-type and XI-infected cells could be localized on the nuclear membrane (NM), endoplasmic reticulum (ER), and Golgi apparatus (G). The Golgi was localized by staining the same cells with rhodamine-conjugated wheat germ agglutinin. In mutant C3-infected cells neither the nuclear membrane nor the Golgi apparatus stained with the antiglycoprotein antiserum. Magnification is 500. 131 of the HA trimer, the deleted amino acids unique to the XI mutation may similarly play a critical role in the tertiary structure of the env complex, such that sorting signals required for transport through the Golgi complex are lost. Mutants X2 and X3 have deletions that begin at the amino terminus of gp85 and extend nine and six amino acids into this structural protein; thus these mutations overlap with the deletion in XI. Nevertheless, both mutant proteins were transported to the cell surface and were indistinguishable from the wild type. Mutants X2 and X3 thus indicate that the terminal nine amino acids of Pr95 are not required for normal intracellular transport and define the critical region in gp85 as the seven amino acids that are uniquely deleted in the XI mutant.

a. Truncation of the Cytoplasmic Sequences. DNA and protein sequence studies demonstrated the presence of a 27 amino acid long hydrophobic (and presumably membrane-spanning) domain and a 22 amino acid long cytoplasmic domain at the carboxy terminus of gp37 ; see Fig. 1 ). Comparison of these domains with those of other exogenous and endogenous strains of RSV has revealed that the sequence within the hydrophobic domain is highly conserved and that within the cytoplasmic domain the sequence of the first 18 amino acids (adjacent to the hydrophobic domain) is also highly conserved while those at the carboxy termini diverge greatly (Hughes, 1982; Hunter et al., 1983) . These results raised the possibility that the conserved region of the cytoplasmic domain might play a functional role in either transport of the env gene product through the secretory pathway or in virus assembly. To investigate this question we initially altered the cytoplasmic domain by introducing deletion mutations into the molecularly cloned sequences of the proviral env gene and examined the effects of the mutations on transport and subcellular localization in CV-1 cells. We found that replacement of the nonconserved region of the cytoplasmic domain with a longer unrelated sequence of amino acids from SV40 vector sequences (mutant Cl) did not alter the rate of transport to the Golgi apparatus nor the appearance of the glycoprotein on the cell surface. Larger deletions, extending into the conserved region of the cytoplasmic domain (mutant C2), however, resulted in a 3-fold slower rate of transport to the Golgi complex, but did not prevent transport to the cell surface (Wills et al., 1983 (Wills et al., , 1984 . These results were thus consistent with the cytoplasmic domain of the RSV env gene product playing some role in transport to the Golgi complex.

Similar results were obtained by Rose and Bergmann (1983) who introduced into the cDNA clone encoding the VSV G protein a series of deletions that affected the cytoplasmic domain. These mutants fell into two classes; the first was completely arrested in their transport at a stage prior to the addition of complex oligosaccharides (presumably the RER) and the second showed severely reduced rates of transport to the Golgi complex although the proteins were ultimately transported to and expressed on the cell surface. The method by which these mutants were constructed (as with the RSV env mutants) meant that the truncated G proteins terminated in SV40 sequences, and in at least one case the block to transport could be alleviated by substitution of a termination codon for these "poison" sequences. Even in these constructions, however, three foreign amino acids were translated prior to termination (Rose and Bergmann, 1983) .

The concept that the cytoplasmic domain might influence or govern the rate at which membrane-spanning proteins were transported to the Golgi complex was supported by similar studies of Doyle et al. (1985) on the HA polypeptide. While the HA cytoplasmic domain could be replaced by the equivalent region from the RSV env gene product without affecting the rate of transport of the hybrid HA from the RER to the Golgi complex, truncation of the HA cytoplasmic domain or addition of the 21 amino acid long cytoplasmic domain from gp37 slowed transport significantly, and addition of 16 amino acids encoded by pBR322 sequences blocked transport of the HA from the RER. On the other hand, studies on the class I histocompatibility antigens (Zuniga et al., 1983; Murre et al., 1984) , the p62 of Semliki Forest virus (Garoff et al., 1983) , and additional studies on the HA of influenza indicated that the cytoplasmic domains of these proteins could be truncated without affecting transport to the cell surface, although the kinetics of transport were not determined in every case.

b. Substitution Mutations. It should be noted that in most of the mutant constructions described above, the carboxy-terminal region contained one or several aberrant amino acids as a result of the recombinant DNA approach. Thus in order to determine more directly the role of the cytoplasmic domain of gp37, we have used oligonucleotide-directed mutagenesis to introduce an early termination codon in the coding sequences of gp37 such that the arginine residue that represents the first amino acid of the cytoplasmic domain is changed to an opal terminator. This mutation creates a truncated viral glycoprotein lacking specifically the cytoplasmic domain of gp37. The biosynthesis and transport of the products of this mutant viral glycoprotein gene were analyzed by expression from an SV40 late-region replacement vector, and its ability to be active in viral assembly was investigated by substitution of the mutated gene for the wild-type gene in an infectious avian retrovirus vector. In contrast to our previous results, deletion of the entire cytoplasmic domain alone had no effect on the biosynthesis or rate of intracellular transport of the env glycoprotein. Thus it seems unlikely that the conserved amino acids present in this region play a role in intracellular transport. Although the cytoplasmic domain contains several charged, hydrophilic residues, it does not appear, by itself, to be required for anchoring the complex in the membrane, since molecules lacking the cytoplasmic domain were expressed stably on the plasma membrane and were not shed into the cell culture medium .

A recent study by Gething et al. (1986) has demonstrated that mutations within the cytoplasmic domain of the influenza virus HA can affect the conformation of the extracellular domain by preventing assembly and trimerization of the HA molecule, thereby resulting in a failure of those mutants to be efficiently transported. A similar requirement for the assembly of oligomeric forms of the VS V G protein prior to its transport to the Golgi complex has also been reported (Kreis and Lodish, 1986) . The inconsistency of our previous results with the ones obtained with the opal mutant could be explained in a similar way. We cannot rule out the possibility that in our earlier experiments the extra amino acids, added as a consequence of the loss of the env termination codon, created a conformational change in the extracellular domain of Pr95 and slowed its transport from the RER. Our present results indicate that the cytoplasmic domain of gp37 is neither a recognition signal for transport to the plasma membrane nor a requirement for anchoring the molecule to it. These findings also support the idea that the charged amino acids present in most of the cytoplasmic domains of many transmembrane proteins (Garoff et al., 1983; Sabatini et al., 1982) are dispensable for anchor function (Davis etal., 1984) .

This latter question has also been addressed by Cutler and co-workers , who mutated the cytoplasmic domain of the p62 polypeptide of Semliki Forest virus. This region, which normally contains a charge cluster (Arg-Ser-Lys) flanking the hydrophobic domain, was changed to a neutral (Met-Ser-Gly) or an acidic (Met-Ser-Glu) one using oligonucleotide mutagenesis. Expression analyses of these mutant proteins confirmed that the basic amino acids were not required for cell surface transport since they reached the surface in a biologically active form. Nevertheless, both mutant polypeptides showed reduced stability when membranes containing them were extracted with high-pH buffer . Charged residues within the cytoplasmic domain may thus provide an additional measure of stability to the membrane-bound complex.

Since the conserved residues in the cytoplasmic domain of gp37 were not required for protein transport it seemed possible that this region might play a role in the process of infectious virus assembly. The fact that the mutant protein was efficiently transported to the cell surface allowed us to analyze this potential role for the cytoplasmic domain in the process of virus budding. Chemical cross-linking experiments have demonstrated an interaction between gp37 and pl9, one of the gag gene products that structure the viral core of RSV (Gebhardt et aL, 1984) . While it is clear that virus assembly can occur in the absence of glycoproteins, it was suggested that the pl9/gp37 interaction may be part of the driving force for the process of viral assembly and budding. Furthermore, since host membrane glycoproteins are excluded from the viral membrane there must be some positive signal for inclusion of the viral env gene products in the budding virion. To determine whether the cytoplasmic domain is involved in this interaction and required for infectious virus assembly, we reconstructed a retrovirus genome carrying the "tail(-)" env gene mutation. Surprisingly, such mutant viruses were infectious on avian cells and spread through the culture with similar efficiency to those containing a native env glycoprotein complex. Furthermore, this truncated env gene complex was incorporated as efficiently into virus particles as the wildtype complex . This fact suggests that if an interaction between gp37 and pl9 is required to mediate the incorporation of the glycoproteins into the envelope of the budding viral particle, it must occur within the lipid bilayer, presumably with the hydrophobic anchor domain. It is thus unlikely that interactions between viral capsid proteins and the cytoplasmic domain of the env complex constitute a driving force for preferential incorporation of the viral glycoproteins in the avian retroviral envelope.

What then is the function of the cytoplasmic domain of the env glycoprotein? Since this segment of the viral polypeptide does show a region of conserved sequence, it is possible that it has evolved to facilitate transport to the plasma membrane without being a requirement for it; clearly, randomly inserted alterations within this domain can exert a negative effect on the transport process. While we have observed normal assembly and infection by virus encoding a "tail(-)" env product, it will be of interest to determine whether continued growth of the virus results in the dominant appearance of revertants that encode a functional cytoplasmic domain.

Many cell surface and membrane proteins of animal viruses are bound to the lipid bilayer by a membrane-spanning hydrophobic peptide close to the carboxy terminus of the polypeptide (reviewed by Warren, 1981; Armstrong et aL, 1981) . Experimental evidence for this first came from deletion mutants of the influenza virus HA (Gething and Sambrook, 1982; Sveda et al, 1982) , the VSV G protein (Rose and Bergmann, 1982) , and the minor coat protein of phage fl (Boeke and Model, 1982) in which removal of sequences that encoded the cytoplasmic and membrane-spanning domains resulted in secretion of the protein.

The hydrophobic membrane-spanning peptide of these polypeptides is thought to be an essential component of the cotranslational signal that results in the arrest of chain transfer across the RER membrane during synthesis. These stop-translocation sequences have been proposed to be a region of the nascent protein molecule which halts insertion through the membrane by disassembling the translocation apparatus and thereby creates proteins with three topological domains (Blobel, 1980) . They appear to be inseparable from the anchor sequences (Yost et al, 1983; Rettenmier et al., 1985) since the transfer of intact transmembrane domains to normally secreted proteins has caused translocation of the constructed hybrid molecules to stop at the added sequences (Yost et al., 1983; Guan and Rose, 1984) . However, the precise structural and physical properties of the stop-translocation sequences have not been defined. Wold et al. (1985) have suggested that the cytoplasmic domain of membrane-spanning proteins might act to interrupt translocation; however, this seems unlikely since deletion mutants lacking this domain are found to be associated with the membrane in a normal manner (see above; Garoff et al., 1983; Zuniga et al, 1983; Murre et al, 1984; Doyle et al, 1986; .

While length and sequence vary widely among regions described as transmembrane anchors, they do have characteristics in common. Most often, they are long stretches (19-30 residues) of predominantly nonpolar and hydrophobic amino acids, bounded by charged residues, at the carboxy terminus of membrane proteins. Membrane-spanning sequences have also been described, however, at the amino terminus of some viral proteins (Blok etal.,A9S2; Palmiter et al, 1978; Bos et al, 1984; Markoff et al, 1984; Zerial et al., 1986; Spiess and Lodish, 1986) and in the middle of other proteins (Rettenmier et al, 1985; Kopito and Lodish, 1985; Finer-Moore and Stroud, 1984) . We have investigated the structural requirements for a functional anchor/stop-translocation sequence in the RSV env system by constructing both deletion and point mutations in this region.

a. Deletion of the Anchor Domain. During our studies on the role of the cytoplasmic domain in env product transport, we characterized a mutant (C3) in which the entire cytoplasmic and transmembrane domains were deleted. This mutant, in contrast to those described for the influenza virus HA and the VSV G protein, was arrested in its transport at the RER and thus was not secreted from the cell (Wills et al., 1984) . Pulse-chase experiments coupled with oligosaccharide precursor labeling experiments showed that the C3 polypeptide was not transported to the Golgi complex, even though it accumulated in a soluble, nonanchored form in the lumen of the RER; the mutant thus appeared to lack a functional sorting signal. Surprisingly, immunofluorescent labeling studies showed that the C3 protein (unlike the wild type) did not accumulate on the nuclear membrane but rather in vesicles distributed throughout the cytoplasm (Fig. 7) , suggesting that movement to the nuclear membrane, blocked in C3, may require a specific transport event, even though the RER and nuclear membranes appear to be continuous. This hypothesis is supported by studies on the VSV G protein that indicate that transitional vesicles (for the transport of glycoproteins to the Golgi apparatus) may be derived from "blebs" in the nuclear membrane (Bergmann and Singer, 1983) .

Although these studies raised the possibility that sorting signals might exist within the deleted region of C3, the 95 amino acid deletion in this mutant, which extends into the external domain of gp37, would be expected to prevent normal folding of the glycoprotein. To determine whether the transmembrane domain was required for intracellular transport, we have modified the env gene by oligonucleotide-directed mutagenesis, changing the lysine (AAA) codon, which precedes the hydrophobic domain of gp37, to an ochre nonsense codon (TAA). This modified gene thus encodes a protein consisting of the entire external domain of Pr95 and lacking precisely the hydrophobic membrane-spanning and hydrophilic cytoplasmic domain. The biosynthesis and intracellular transport of the truncated protein in CV-1 cells was not significantly different from that of the wild-type glycoprotein, suggesting that any protein signals for biosynthesis and intracellular transport of this viral glycoprotein complex must reside in its extracellular domain. In contrast to the case of the C3 mutant, this complex lacking just the transmembrane and cytoplasmic domains is secreted as a soluble molecule into the culture medium . Since the glycoprotein complex lacking only the cytoplasmic domain of gp37 is stably expressed on the cell surface, in a manner similar to the wild-type complex, it can be concluded that the transmembrane domain alone is required for anchoring the RSV env complex in the cell membrane.

b. Requirements for a Functional Stop-Translocation/Anchor Sequence. We have approached the question of the compositional requirements for membrane anchoring and orientation (stop-translocation) of a membrane-spanning protein by substituting an arginine for a centrally positioned leucine in the hydrophobic anchor region of the RSV env gene product (Fig. 8) . The arginine substitution is one of the most drastic ^ Membrane-spanning Mutant « domain * Wt his leu leu lysjgly leu leu leu gly leu val val ile leu leu leu leu val cys leu pro cys leu leu gin phe val ser ser ser ile|arg lys met μΆ^ his leu leu lys|gly leu leu leu gly leu val val ile leu leu leu leu val cys|arg|pro cys leu leu gin phe val ser ser ser ile) arg lys met T24 his leu leu lys|gly leu leu leu gly leu val val ile leu leu leu leu val cysj ;;;-::;;;: : :;|i|leu leu gin phe val ser ser ser ile|arg lys met T18 his leu leu lys|gly leu leu leu gly leu val val ile leu leu leu leu val| |ser ser se7Üe|arg lys met T16 his leu leu lys|gly leu leu leu gly leu val val ile leu leu[ jval ser ser ser ile| arg lys met T11 his leu leu lys|gly leu leu leu gly leu val| |ser ser ser ile|arg lys met T5 his leu leu lys|gJ7leu||| |ser ser ile|arg lys met compositional point mutations that could be made since it is only rarely found buried in hydrophobic environments (Kyte and Doolittle, 1982) and has a high predicted potential for terminating membrane-buried helices (Rao and Argos, 1986) . The substitutions we have made fall within the conserved leucine-rich " I C " region proposed by Patarca and Haseltine (1984) and near the two cysteine residues where palmitate may be covalently added (Gebhardt et aL, 1984; Kaufman et al., 1984) .

By changing the anchor's hydrophobic integrity through the insertion of point mutations we hoped to define better what constituted a functional anchor sequence. Placing a highly charged basic side chain into the hydrophobic core of the membrane might be expected to either (1) terminate the membrane-spanning helix, thereby partitioning the charged residue to one side or the other of the membrane, or (2) destroy the stop-translocation signal, causing the protein to be secreted. The results of these experiments showed that a single amino acid substitution in the transmembrane anchor did not affect membrane association or its orientation in the membrane; unexpectedly, however, it affected targeting of the protein at a stage late in the transport pathway, such that the mutant protein was rapidly degraded in lysosmes . The early translation products of both the arginine-mutant and wild-type genes behaved normally: they were synthesized with equal efficiency, had normal bitopic symmetry, and were glycosylated. The kinetics for the turnover of these precursors were nearly identical to those previously reported in infected chicken embryo fibroblasts (Bosch and Schwarz, 1984) and in SV40 expression vectors (Wills et al., 1984) . Furthermore, in the Golgi, palmitate was added to the precursors, they were cleaved to gp85-gp37, and they received terminal sugars. Only after this last stage did the presence of the charged side chain of the substituted arginine alter expression. At the level of the trans Golgi, a post-Golgi compartment (Saraste and Kuismanen, 1984) , or at the cell surface, the gp85-gp37 complex was rapidly shunted to lysosomes and degraded-as shown by the protection afforded the terminally glycosylated env proteins by the lysosomatropic agent, chloroquine . The exact pathway that the molecules take to the lysosome is not known. They may be transported directly from the trans Golgi or first to the surface where they are rapidly endocytosed. Discriminating between the alternate pathways has not been possible from current data.

Why the insertion of an arginine into the anchor should result in targeting to lysosomes is not obvious. A possible explanation is that we have introduced a specific sorting signal into the molecule; however, this is unlikely since other polypeptides with charged residues in the membranespanning domain are not so targeted (Kabcenell and Atkinson, 1985; Saito et al., 1984; Hayday et al., 1985) . The arginine's charge is incompatible with the hydrophobic environment of the lipid bilayer; to achieve stability, the charged guanidinium group needs to be neutralized, and how this is done inside the bilayer is not clear. Parsegian (1969) has postulated that a lone charge sequestered in a membrane must form a pore or tunnel along with localized membrane thinning to acheive the lowest energy state. If the charged residue in the gp37 anchor causes the mutant molecules to aggregrate and form channels in the membrane in an analogous manner, it would likely kill the cell unless there was a mechanism to remove it rapidly. Alternatively, since the env protein is not an isolated entity in the membrane, it is conceivable that it aggregates with other components of the membrane to reduce net charge cooperatively, and thereby triggers the endocytotic machinery (Mellman and Plutner, 1984) .

Charged residues are found in several proposed membrane-spanning helices (Kabcenell and Atkinson, 1985; Saito et al., 1984; Hayday et al., 1985; reviewed by Rao and Argos, 1986) . The charged residues in bacteriorhodopsin membrane-spanning a helices have been suggested to be neutralized by forming ion pairs (Engelman et al., 1980) . This is likely a special case, however, since the energy required to bury an ion pair in the membrane is not much different from that required to bury the free charged group itself (Parsegian, 1969) . Neutralization of strong charges, particularly of lysine and arginine, may occur through the formation of strong hydrogen bonds with tyrosine (Kyte and Doolittle, 1982) ; however, no tyrosine residues are present in the anchor domain of the env gene product which could participate with the arginine. The T cell a, ß, and y gene products and the rotavirus VP7 protein have a putative structure similar to that of the arginine mutant, with a lysine centered within the transmembrane anchor; however, unlike the molecule we created, they invariably have tyrosine residues adjacent to the lysine (Saito et al., 1984; Kabcenell and Atkinson, 1985; Hayday et al., 1985) which could stabilize the charge through hydrogen bonds (Kyte and Doolittle, 1982) . Adams and Rose (1985a) have described the similar insertion of an arginine (and glutamine) residue at the center of the transmembrane domain of the VSV G protein. Their mutant protein, like the env mutant we have characterized, was bitopic, could be seen localizing in the Golgi, and did not accumulate on the cell surface. Since these investigators observed a "lower level of protein expression" with their mutant G protein, it is possible that it also was rapidly degraded in lysosomes following terminal glycosylation. In contrast, observed no alteration in the biosynthesis and transport of a mutant p62 polypeptide of Semliki Forest virus in which the hydrophobic domain was interrupted by insertion of a glutamic acid residue. In this protein, however, the outer boundary of the hydrophobic domain is not delineated by a charged residue and so it is possible that additional uncharged residues from the external domain were pulled into the membrane.

Since the insertion of a charged polar residue into the transmembrane region of the RSV env gene product did not interfere with its anchor/stoptranslocation function, we have investigated the requirement for the long (27 amino acid) hydrophobic domain in arresting translocation and anchoring the env complex. A series of deletion mutations was generated by progressively removing base pairs to either side of a unique Sph\ restriction site that had been previously engineered into the center of the anchor coding region. This produced env proteins with truncated transmembrane anchors that ranged in length from 24 (T24) to a single apolar amino acid (Tl) summarized in Fig. 8 ; .

While the effects of the deletions on the transport and subcellular localization of the env gene product appeared to be a complex function of the length and composition of the remaining anchor, the mutants appeared to fall into three broad phenotypic classes (summarized in Table I ). Even the smallest deletion (T24), which removed only three amino acids, greatly reduced the surface expression of the mature env proteins. T24 and mutants T18, T17, and T16 had a normal bitopic orientation in the membrane but appeared to be cleared from the cell surface and degraded in lysosomes, since they accumulated only in the presence of choroquine, an inhibitor of lysosomal degradation. The reduced surface expression of the 

As determined by the distance between charged residues. b Hydrophobicity score as determined using the hydropathicity values of Kyte and Doolittle (1982) . Mean hydrophobicity is hydrophobicity score divided by the apparent anchor length. c Symbols: + denotes either a wild-type or positive response; ± denotes an intermediate response; -denotes a negative response. d SEC, Secreted. e ND/na, Not determined/not applicable. f With chloroquine treatment, Pr95 appears in the medium with gp85 and gp37. « Anchorless and tailess deletion mutant .

largest deletion mutant (T24) was surprising since the effective hydrophobic peptide remaining in this construct was as long or longer than functional anchors reportedly present in other integral membrane proteins [e.g., 19 amino acids: M 2 protein of influenza (Lamb, 1985) ; 20 amino acids: VSV G protein (Rose et al., 1980) and adenovirus E3 protein (Wold et al., 1985) ]. Deletions which reduced the size of the transmembrane anchor to seven amino acids (T7) or less resulted in the secretion of mature glycoproteins into the medium. A third class of mutants with hydrophobic regions of 14 (T14) and 11 (Til) amino acids, respectively, while remaining membrane associated, no longer appeared to span the RER as bitopic proteins. Neither mutant could be found at the surface of cells, nor could their degradation be arrested by chloroquine treatment.

From the sum of the data obtained with these mutants, it would appear that bitopic insertion of the env gene product is possible with effective anchor domains of at least 16 amino acids; if additional amino acids are removed from the domain, the protein can no longer exist bitopically, and it either partitions monotopically to the luminal side of the RER membrane or withdraws amino acids from the cytoplasmic side of the membrane into the bilayer. Nevertheless, such sequences from the cytoplasmic domain are not able to stabilize the shortest anchors (Tl, T5, and T7) since these are secreted from the cell. Davis and Model (1985) have investigated the requirements of a functional anchor domain by inserting artificial hydrophobic peptides of varying length into the membrane-associated pill protein of the bacteriophage fl. Their results show that 17 hydrophobic amino acids are sufficient to maintain the protein in a bitopic configuration; however, the 17 amino acid anchor was "deleterious to the cell" presumably because it was too short to assume a stable conformation compatible with existence in the bilayer and thereby destabilized the membrane. A construct with an anchor of 12 hydrophobic amino acids was membrane associated but showed an intermediate phenotype in its sensitivity to solubilization by alkali. In contrast, a construct with only 8 hydrophobic amino acids-also membrane associated-was completely released into the supernatant at high pH. Adams and Rose (1985b) reduced the anchor domain of VSV G by precisely deleting amino acids from within the hydrophobic core. When the length of the anchor was reduced from 20 amino acids to as few as 14, the protein was normally membrane associated and expressed in a bitopic fashion on the cell surface. On the other hand, proteins with an anchor domain of 12 or 8 amino acids, while spanning the membrane, appeared to be transported only as far as the Golgi where they accumulated; the surface expression of these proteins was greatly reduced (12 amino acid anchor) or undetectable (8 amino acids). Doyle et al. (1986) have characterized a series of carboxy-terminal deletion mutants of the HA polypeptide in which the 27 amino acid anchor domain was truncated to 17, 14, and 9 amino acids, respectively. In this case, molecules with a 17 amino acid long transmembrane domain were stably anchored but were transported less efficiently to the plasma membrane. Truncation of the hydrophobic anchor to 9 or 14 residues resulted in HA proteins that were unstable and whose transport appeared to be blocked in the RER or in a pre-Golgi compartment-resembling the Til mutant of the RSV env gene described above. From the results of these different systems and approaches it would seem that in strictly physical terms, anchors may be significantly reduced in their length without any consequence to the membrane association. The limits for this length ap-pear to be about 8-12 amino acids, but it must be appreciated that the mere presence of a stretch of hydrophobic amino acids within a protein does not serve to constitute an anchor. Several mammalian virus envelope proteins contain in their external domain long hydrophobic amino acid regions that are equivalent to the truncated bitopic anchors have described here (Gething et al, 1978; White et al, 1981) . Indeed, gp85 contains a strongly hydrophobic 11 amino acid region that clearly does not act as a stop-translocation sequence or play a role in membrane association . The work of Davis and Model (1985) , on the other hand, implies that the length of a hydrophobic region is the major determinant as to whether or not it will confer membrane association properties to a protein, although they point out that the position of such sequences within the molecule may play a role.

Most eukaryotic membrane-spanning polypeptides have a complex tertiary structure that is stabilized by multiple disulfide linkages, and it is possible that the entropy of a correctly folded molecule is sufficient to pull potential stop-transfer regions through the membrane. A corollary of this hypothesis, therefore, is that, once folding is complete, short hydrophobic regions that can potentially span the membrane as an a helix would stop translocation. The length requirements for such a region could be shorter than the 20 residues predicted from a-helix dimensions if the region were flanked by arginines or lysines. Since the latter have long side chains, equivalent in length to a single turn of an a helix, a stretch of hydrophobic amino acids 13-14 amino acids long might be sufficient. Such a prediction fits well with the data we have obtained and with those of Adams and Rose (1985b) and Davis and Model (1985) . It will be of interest to determine what effect inserting the truncated anchors of mutants T18 and T16 into the middle of the env precursor has on translocation; if the above speculations are correct they should be extruded into the external domain.

Finally, it should be reemphasized that merely providing a bitopic membrane anchor/stop-translocation is not sufficient to confer wild-type biological activity on a polypeptide. Mutant T24 of the RSV env gene has a hydrophobic domain which might be expected to be sufficiently long and hydrophobic in character to span the membrane stably; it is modified normally by palmitic acid and can clearly be transported to its targeted cellular location. Nevertheless, it is degraded rapidly by the cell. These results imply that hydrophobic transmembrane domains contain additional (and perhaps subtle) signals that remain to be deciphered, a conclusion that is supported by the finding that deletion of the anchor domain of the rotavirus VP7 protein abolishes its specific targeting and retention in the RER (Poruchynsky et al, 1985) .

Polarized epithelial cells exhibit apical and basolateral membrane domains that are separated by well-defined tight junctions. Each membrane domain has a unique protein composition (Louvard, 1980; Reggio et al., 1982) , indicating that mechanisms must exist to specifically target membrane proteins to different surfaces. The sorting process occurs during or shortly after passage of the glycoproteins through the Golgi complex Pfeiffer et al., 1985; Rindler et al., 1984 Rindler et al., , 1985 . However, the mechanisms determining this directed transport to either the apical or basolateral membranes are not understood. Their study has been facilitated by the use of cultured epithelial cell lines, such as the MDCK cell line, and by the observation that certain RNA viruses bud exclusively from apical or basolateral domains of these polarized cells in culture (Rodriguez-Boulan and Sabatini, 1978; Herrler et al., 1981; Roth et al., 1983a; Rindler et al., 1985) . Avian and mammalian retroviruses together with rhabdoviruses such as VSV mature from the basolateral surface, while ortho-and paramyxoviruses bud from the apical surface. The carbohydrate residues present on the different proteins do not appear to play a role in this sorting process, since tunicamycin does not interfere with the polarized release of the viruses (Roth et al., 1979; Green et al., 1981b) .

As with the maturation of viruses that assemble at intracellular locations within the secretory pathway, polarized budding of enveloped viruses is dependent on the site to which viral glycoproteins are transported. Expression of cloned viral glycoprotein genes from both SV40-based and vaccinia expression vectors in polarized cells has demonstrated that the HA and neuraminidase polypeptides of influenza virus are targeted to the apical surface (Roth et al., 1983b; Jones et al., 1985; Gottlieb et al., 1986) while the G protein of VSV and the gp70/pl5E complex of murine leukemia virus (MuLV) are transported exclusively to the basolateral membranes . In an attempt to locate the signals which direct these glycoproteins to the apical or basolateral domains, recombinant DNA techniques have been employed to construct chimeric proteins and express these in polarized cells in culture. In experiments where sequences encoding the external domain of HA were fused to those encoding the transmembrane and cytoplasmic domains of the VSV G protein, the hybrid glycoprotein behaved in the same manner as wild-type HA and was transported to the apical domain of polarized cells (McQueen et al., 1986; Roth et al., 1986) . Conversely, fusing the external domain of G protein to the anchor/cytoplasmic domain of HA results in basolateral transport (McQueen et al., 1987) . These experiments thus suggest that the ectodomains of HA and G protein contain signals for apical and basolateral transport, respectively. While expression of a secreted form of the HA glycoprotein in an apical polarized manner supports this conclusion (Roth et al. y 1986) , the unanchored ectodomain of the MuLV gp70/pl5E complex, which is normally targeted to the basolateral membrane, is secreted in a nonpolarized fashion . It is possible that this soluble protein is improperly folded and thus is unable to interact with the sorting machinery, alternatively it also raises the possibility that targeting signals may be located in more than one domain of these molecules. Further analyses should shed light on this problem.

The studies described in this chapter demonstrate the breadth of information that has been and can be obtained from studies on enveloped virus glycoprotein biosynthesis. Many of the studies were performed at a time when the cloned genes and molecular probes for cellular glycoproteins were unavailable and thus provided valuable insights into the manner in which cells compartmentalized and transported membrane proteins. The exciting possibility of utilizing viral glycoprotein genes for genetic analyses of the transport pathway, in a way analogous to that pursued in prokaryotic systems (Michaelis and Beckwith, 1982; Silhavy et al. y 1983; Oliver, 1985) , has led to a plethora of studies that utilized both classic and recombinant DNA genetic approaches. These investigations have resulted in great progress in our understanding of the general processes involved in intracellular transport of proteins through the secretory pathway but at the same time have raised difficult questions about the molecular interactions required for protein sorting.

The initial observation that proteins destined for secretion contain an amino-terminal sorting sequence provided a precedent on which to build models for protein targeting based on topogenic sequences (Blobel, 1980) . To a large extent the identification of signal sequences was facilitated by their transient nature, not by a conserved primary sequence. Indeed, while some common characteristics of signal peptides can be recognized (von Heijne, 1985) , the identification of signal peptides in proteins where they are not removed has proved difficult and has required the use of sophisticated recombinant DNA technology [e.g., ovalbumin (Tabe et ah, 1984) ]. The requirement for a signal sequence to initiate translocation across the ER membrane has been clearly confirmed through the isolation and construction of mutants which lack this functional region (as discussed above) and by fusion of this sequence to proteins that are not normally translocated (Lingappa et al., 1984) . These approaches have also been facilitated by the transient, nonstructural nature of many aminoterminal, translocation signals. Recent experiments by Friedlander and Blobel (1985) and Kaiser et al. (1987) , however, raise questions about the informational content of signal sequences. In particular Kaiser and colleagues showed that several random amino acid sequences derived from human genomic DNA fragments could act as signal sequences for translocation of the yeast invertase enzyme. Thus even in this well-defined situation, where an amino acid sequence is known to play a functional role in the sorting process, it can be impossible to predict with confidence its location in the protein; how then might we expect to identify additional sorting sequences that may or may not exist within the structural domain of a transported protein?

The possibility that additional sorting sequences might be involved in steering the transport of a membrane-spanning or secreted protein through the vesicular maze of the secretory pathway remains open. At the present time the necessity for a native tertiary structure cannot be separated from the possibility of such additional sorting sequences. It is clear that disruption of a poly peptide's normal folding can completely prevent its transport from the ER Kreis and Lodish, 1986) , and the simplest explanation for the phenotypes of a variety of conditional and nonconditional transport mutants would be that they alter the tertiary structure of the mature protein (see Section II, A, above). A 3-dimensional structure has been determined for only a few molecules that traverse the secretory pathway, and even with these proteins the current, predictive algorithms are insufficiently accurate to model potential changes in molecular shape in response to mutations. The question of a direct role for tertiary structure in protein transport thus represents a major challenge to molecular biologists. Furthermore, one might argue that a change in protein shape could also mask or distort a necessary (peptide) sorting sequence. This possibility is supported by the observation that for a majority of eukaryotic proteins the amino-terminal signal peptide is unable to initiate translocation across the ER if translation is allowed to proceed to completion, presumably because the tertiary structure of the nascent polypeptide precludes the interaction of the signal peptide with the translocation machinery.

It is quite feasible that sorting signals and targeting signals could be represented by different entities within a single polypeptide, particularly if the latter were required to fix the intracellular location of a protein. For example, the rotavirus VP7 polypeptide accumulates within the ER un-less its amino-terminal hydrophobic anchor region is deleted, whereupon it is transported to the cell surface and secreted (Poruchynsky et al., 1985) ; in this instance the deleted region presumably contains a sequence that can fix the intracellular location of the protein despite the fact that the molecule has the potential to be exported from the cell.

Since most translocated polypeptides appear to follow a common pathway to a late compartment of the Golgi (Kelly, 1985) , it might be argued that a native conformation is the sole requirement for transport to this organelle and that the observed differences in the transport rates of proteins to the Golgi merely reflect the time necessary for completion of the folding process. Nevertheless, proteins leaving the Golgi appear to be sorted into specific pathways; for example, in secretory cells proteins may follow either the constitutive pathway or be sequestered in secretory granules (Moore and Kelly, 1985; reviewed by Kelly, 1985) , and in epithelial cells specific proteins appear to be transported directly to either the apical or basolateral membranes (see above). Thus, it would seem likely that some form of sorting signal must be present in the polypeptide at this point in the secretory pathway in order to correctly direct its transport; initial results from viral glycoprotein expression studies indicate that at least in polarized cells the ectodomain of the sorted protein plays a dominant role (McQueen et al., 1986; . Additional studies should provide a clearer picture of this complex process.

In summary, studies on the biosynthesis and transport of enveloped virus glycoproteins have provided important insights into the general processes involved in the intracellular movement of these membrane-associated molecules. The specific questions that remain to be answered are many and difficult, but it is likely that these viral systems will continue to play a vital role by providing clues and direction in this important area of cell biology. 

Incorporation of a charged amino acid into the membrane spanning domain blocks cell surface transport but not membrane anchoring of a viral protein

Structural requirements of a membrane-spanning domain for protein anchoring and cell surface transport

Sequence analysis of two mutants of Sindbis virus defective in the intracellular transport of their glycoproteins

Domain structure of bacteriophage fed adsorption protein

Intragenic suppressor mutations that restore export of maltose binding protein with a truncated signal peptide

Information within the mature LamB protein necessary for localization to the outer membrane of Eseherichia coli K-12

Genetic analysis of protein export in Escherichia coli K-12

Immunoelectron microscopic studies of the intracellular transport of the membrane glycoprotein (G) of vesicular stomatitis virus in infected Chinese hamster ovary cells

Passage of an integral membrane protein, the vesicular stomatitis virus glycoprotein, through the Golgi apparatus en route to the plasma membrane

Bunyaviridae. In "Comprehensive Virology

Intracellular protein topogenesis

Transfer of proteins across membranes. I. Presence of proteolytically processed and unprocessed nascent immunoglobulin light chains on membrane-bound ribosomes of murine myeloma

Studies on the size, chemical composition and partial sequence of the neuraminidase (NA) from type A influenza viruses show the Nterminal region of NA is not processed and serves to anchor NA in the viral membrane

A prokaryotic membrane anchor sequence: Carboxyl terminus of bacteriophage fl gene III protein retains it in the membrane

Processing of filamentous phage precoat protein: Effect of sequence variations near the signal peptidase cleavage site

NH 2 -terminal hydrophobic region of influenza virus neuraminidase provides the signal function in translocation

Processing of gPr92env, the precursor to the glycoproteins of Rous sarcoma virus: Use of inhibitors of oligosaccharide trimming and glycoprotein transport

Sequence of the membrane protein gene from avian coronavirus IBV

Virus-specific glycoproteins associated with the nuclear fraction of herpes simplex virus type 1-infected cells

Mutants of the membrane-binding region of Semliki Forest virus E2 protein. I. Cell surface transport and fusgenic activity

Mutants of the membrane-binding region of Semliki Forest virus E2 protein. II. Topology and membrane binding

Fusion mutants of the influenza virus hemagglutinin glycoprotein

Isolation of the Escherichia coli leader peptidase gene and effects of leader peptidase overproduction in vivo

A charged amino acid substitution within the transmembrane anchor of the Rous sarcoma virus envelope glycoprotein affects surface expression but not intracellular transport

Altered surface expression, membrane association and intracellular transport result from deletions within the transmembrane anchor of the Rous sarcoma virus envelope glycoprotein

An artificial anchor domain: Hydrophobicity suffices to stop transfer

Fine structure of a membrane anchor domain

Variant influenza virus hemagglutinin that induces fusion at elevated pH

Mutations in the cytoplasmic domain of influenza virus hemagglutinin affect different stages of intracellular transport

Analysis of progressive deletions of the transmembrane and cytoplasmic domains of influenza hemagglutinin

Assembly of Enveloped RNA Viruses

Early and late functions associated with the Golgi apparatus reside in distinct compartments

Localization and processing of outer membrane and periplasmic proteins in Escherichia coli strains harboring export-specific suppressor mutations

Suppressor mutations that restore export of a protein with a defective signal sequence

Importance of secondary structure in the signal sequence for protein secretion

Path of the polypeptide in bacteriohodopsin

Amphipathic analysis and possible formation of the ion channel in an acetylcholine receptor

Evidence for a glycoprotein "signal" involved in transport between subcellular organelles

Bovine opsin has more than one signal sequence

Characterization of two recombinant-complementation groups of Uukuniemi virus temperature-sensitive mutants

Uukuniemi virus glycoproteins accumulate in and cause morphological changes of the Golgi complex in the absence of virus maturation

Nucleotide sequence of a cDNA clone encoding the entire glycoprotein from the New Jersey serotype of vesicular stomatitis virus

A single amino acid substitution in a hydrophobic domain causes temperature-sensitive cell-surface transport of a mutant viral glycoprotein

Using recombinant DNA techniques to study protein targeting in the eucaryotic cell

Expression of Semliki Forest virus proteins from cloned complementary DNA. II. The membrane-spanning glycoprotein E2 is transported to the cell surface without its normal cytoplasmic domain

Rous sarcoma virus pl9 and gp35 can be chemically crosslinked to high molecular weight complexes. An insight into viral association

Cold Spring Harbor Laboratory, Cold Spring Habor

Construction of influenza haemagglutinin genes that code for intracellular and secreted forms of the protein

Purification of the fusion protein of Sendai virus: Analysis of the NH 2 -terminal sequence generated during precursor activation

Cloning and DNA sequence of double-stranded copies of haemagglutinin genes from H2 and H3 strains elucidates antigenic shift and drift in human influenza virus

Expression of wild-type and mutant forms of influenza hamagglutinin: The role of folding in intracellular transport

Synthesis and infectivity of vesicular stomatitis viruses containing nonglycosylated G protein

The nonglycosylated glycoprotein of vesicular stomatitis virus is temperature-sensitive and undergoes intracellular aggregation at elevated temperatures

Protein translocation across the endoplasmic reticulum. I. Detection in the microsomal membrane of a receptor for the signal recognition particle

Protein translocation across the endoplasmic reticulum. II. Isolation and characterization of the signal recognition particle receptor

The mechanism of protein translocation across the endoplasmic reticulum membrane

Sorting and endocytosis of viral glycoproteins in transfected polarized epithelial cells

Passage of viral membrane proteins through the Golgi complex

Glycosylation does not determine segregation of viral envelope proteins in the plasma membrane of epithelial cells

Conversion of a secretory protein into a transmembrane protein results in its transport to the Golgi complex but not to the cell surface

Glycosylation allows cell-surface transport of an anchored secretory protein

Two distinct intracellular pathways transport secretory and membrane glycoproteins to the surface of pituitary tumor cells

Amino-terminal deletion mutants of the Rous sarcoma virus glycoprotein do not block signal peptide cleavage but block intracellular transport

Biosynthesis of lysosomal enzymes in fibroblasts: Phosphorylation of mannose residues

Structure, organization, and somatic rearrangement of T cell a genes

Antitrypsin: The presence of excess mannose in the Z variant isolated from liver

Isolation and structural analysis of influenza virus C virion glycoproteins

Studies on the mechanisms of tunicamycin inhibition of IgA and IgE secretion by plasma cells

The importance of the endosome in intracellular traffic

Sequence of the long terminal repeat and adjacent segments of the endogenous avian virus Rous-associated virus

Complete sequence of the Rous sarcoma virus env gene: Identification of structural and functional regions of its product

Mechanism of signal peptide cleavage in the biosynthesis of the major lipoprotein of the Escherichia coli outer membrane

Phospholipid is required for the processing of presecretory proteins by detergent-solubilized canine pancreatic signal peptidase

Intracellular transport of herpes simplex virus gD occurs more rapidly in uninfected cells than in infected cells

Surface expression of influenza virus neuraminidase an amino-terminally anchored viral membrane glycoprotein, in polarized epithelial cells

Processing of the rough endoplasmic reticulum membrane glycoproteins of rotavirus SAH

Many random sequences functionally replace the secretion signal sequence of yeast invertase

Cysteines in the transmembrane region of major histocompatability complex antigens are fatty acylated via thioester bonds

Pathway of protein secretion in eukaryotes

pH-induced alterations in the fusogenic spike protein of Semlilci Forest virus

Membrane fusion mutants of Semliki Forest virus

Separate pathways of maturation of the major structural proteins of vesicular stomatitis virus

Maturation of viral proteins in cells infected with temperature-sensitive mutants of vesicular stomatitis virus

Primary structure and transmembrane orientation of the murine anion exchange protein

Oligomerization is essential for transport of the vesicular stomatitis virus glycoprotein to the cell surface

Uukuniemi virus maturation: An immune fluorescence microscopy study using monoclonal glycoprotein-specific antibodies

A simple method for displaying the hydropathic character of a protein

Influenza M2 protein is an integral membrane protein expressed on the infected-cell surface

Impaired intracelluiar migration and altered solubility of nonglycosylated glycoproteins of vesicular stomatitis virus and Sindbis virus

Kinetics of serum protein secretion by cultured hepatoma cells: Evidence for multiple secretory pathways

Determinants for protein localization: /3-Lactamase signal sequence directs globin across microsomal membranes

Hen oviduct signal peptidase is an integral membrane protein

Reversible block in intracellular transport and budding of mutant vesicular stomatitis virus glycoprotein

Hepatoma secretory proteins migrate from rough endoplasmic reticulum to Golgi at characteristic rates

Apical membrane aminopeptidase appears at site of cell-cell contact in cultured kidney epithelial cells

L985). A single N-linked oligosaccharide at either of the two normal sites is sufficient for transport of vesicular stomatitis virus G protein to the cell surface

Polarized expression of a chimeric protein in which the transmembrane and cytoplasmic domains of the influenza virus hemagglutinin have been replaced by those of the vesicular stomatitis virus G protein

Basolateral expression of a chimeric protein in which the transmembrane and cytoplasmic domains of vesicular stomatitis virus G protein have been replaced by those of the influenza virus hemagglutinin

Glycosylation and surface expression of the influenza virus neuraminidase requires the N-terminal hydrophobic region

The entry of enveloped viruses into cells by endocytosis

Internalization and degradation of macrophage Fc receptors bound to polyvalent immune complexes

Acidification of the endocytic and exocytic pathways

Identification and characterization of a membrane component essential for the translocation of nascent proteins across the membrane of the endoplasmic reticulum

Secretory protein translocation across membranes-the role of the "docking protein

Mechanism of incorporation of cell envelope proteins in Escherichia coli

Secretory protein targeting in a pituitary cell line: Differential transport of foreign secretory proteins to distinct secretory pathways

Structural mutations in a mouse immunoglobulin light chain resulting in failure to be secreted

Construction, expression and recognition of an H-2 molecule lacking its carboxyl terminus

Identification of the defects in the hemagglutinin gene of two temperature-sensitive mutants of A/WSN/33 influenza virus

Membrane-bound penicillinases in grampositive bacteria

Carbohydrate moieties of glycoproteins, a reevaluation of their function

Protein secretion in Escherichia coli

Ovalbumin: A secreted protein without a transient hydrophobic leader sequence

Energy of an ion crossing a low dielectric membrane: Solutions to four relevant problems

Similarities among retro virus proteins

Mutations within the proteolytic cleavage site of the Rous sarcoma virus glycoprotein precursor block processing to gp85 and gp37

Mutants of the Rous sarcoma virus envelope glycoprotein that lack the transmembrane anchor and/or cytoplasmic domains: Analysis of intracellular transport and assembly into virions

A putative signal peptidase recognition site and sequence in eucaryotic and procaryotic signal peptides

Reversible defect in the glycosylation of the membrane proteins of Semliki Forest virus tsl mutant

Localization of rotavirus antigens in infected cells by ultrastructural immunocytochemistry

Intracellular sorting and basolateral appearance of the G protein of vesicular stomatitis virus in MDCK cells

Complete nucleotide sequence of an influenza virus haemagglutinin gene from cloned DNA

Deletions into an NH 2 -terminal hydrophobic domain result in secretion of rotavirus VP7, a resident endoplasmic reticulum glycoprotein

Export of protein in bacteria

A conformation preference parameter to predict helices in integral membrane proteins

Surface and cytoplasmic domains in polarized epithelial cells

Transmembrane orientation of glycoproteins encoded by the w-fms oncogene

Viral glycoproteins destined for apical or basolateral plasma membrane domains traverse the same Golgi apparatus during their intracellular transport in doubly infected Madin-Darby canine kidney cells

Polarized delivery of viral glycoproteins to the apical and basolateral plasma membranes of Madin-Darby canine kidney cells infected with temperature-sensitive viruses

Asymmetric budding of viruses in epithelial monolayers: A model system for study of epithelial polarity

Intracellular transport of influenza virus hemagglutinin to the apical surface of Madin-Darby canine kidney cells

Expression from cloned cDNA of cell-surface secreted forms of the glycoprotein of vesicular stomatitis virus in eukaryotic cells

Altered cytoplasmic domains affect intracellular transport of the vesicular stomatitis virus glycoprotein

Vesicular stomatitis virus is anchored in the viral membrane by a hydrophobic domain near the COOH terminus

The presence of cysteine in the cytoplasmic domain of the vesicular stomatitis virus glycoprotein is required for palmitate addition

Polarity of influenza and vesicular stomatitis virus in MDCK cells; lack of a requirement for glycosylation of viral glycoproteins

Influenza virus hemagglutinin expression is polarized in cells infected with recombinant SV40 viruses carrying cloned hemagglutinin DNA

Basolateral maturation of retroviruses in polarized epithelial cells

Heterologous transmembrane and cytoplasmic domains direct functional chimeric influenza virus hemagglutinins into the endocytic pathway

Membrane insertion and intracellular transport of influenza virus glycoproteins

The large external domain is sufficient for the correct sorting of secreted or chimeric influenza virus hemagglutinins in polarized monkey kidney cells

Studies on the adaption of influenza viruses to MDCK cells

A mutation downstream from the signal peptidase cleavage site affects cleavage but not membrane insertion of phage coat protein

Mechanisms for the incorporation of proteins in membranes and organelles

Complete primary structure of a heterodimeric T-cell receptor deduced from cDNA sequences

Pre-and post-Golgi vacuoles operate in the transport of Semliki Forest virsus membrane glycoproteins to the cell surface

Fatty acid binding to vesicular stomatitis virus glycoprotein: A new type of posttranslational modification of the viral glycoprotein

Relation of fatty acid attachment to the translation and maturation of vesicular stomatitis and Sindbis virus membrane glycoproteins

Evidence for covalent attachment of fatty acids to Sindbis virus glycoproteins

Defects in functional expression of an influenza virus hemagglutinin lacking the signal peptide sequences

Analysis of the hemagglutinin glycoprotein from mutants of vaccinia virus that accumulates on the nuclear envelope

Mechanisms of protein localization

Changes in the conformation of influenza virus hemagglutinin at the pH optimum of virus-mediated membrane fusion

The phosphomannosyl recognition system for intracellular and intercellular transport of lysosomal enzymes

Glycoproteins specified by herpes simplex viruses

An internal signal sequence: The asialoglycoprotein receptor membrane anchor

Nonpolarized expression of a secreted murine leukemia virus glycoprotein in polarized epithelial cells

Polarized transport of the VSV G surface expression of viral glycoproteins is polarized in epithelial cells infected with recombinant vaccinia viral vectors

Intracellular transport of secretory and membrane proteins in hepatoma cells infected by vesicular stomatitis virus

Vesicular stomatis virus glycoprotein, albumin, and transferrin are transported to the cell surface via the same Golgi vesicles

Effect of tunicamycin on the secretion of serum proteins by primary cultures of rat and chicken hepatocytes

The molecular biology of coronaviruses

Cell surface expression of the influenza virus hemagglutinin requires the hydrophobic carboxy-terminal sequences

Influenza virus hemagglutinin containing an altered hydrophobic carboxy terminus accumulates intracellularly

Segregation of mutant ovalbumins and ovalbumin-globin fusion proteins in Xenopus oocytes: Identification of an ovalbumin signal sequence

Eukaryotic signal sequence transports insulin antigen in Escherichia coli

Prolipoprotein signal peptidase in Escherichia coli is distinct from the M13 procoat protein signal peptidase

Temperature-sensitive mutants of influenza virus: A mutation in the hemagglutinin gene

Structural studies of IgA myeloma proteins having anti-DNP antibody activity

Patterns of amino acids near signal-sequence cleavage sites

How signal sequences maintain cleavage specificity

Signal sequences, the limits of variation

Purification of a membrane-associated protein complex required for protein translocation across the endoplasmic reticulum

Signal recognition protein (SRP) mediates the selective binding to microsomal membranes of m-i>//ro-assembled poly somes synthesizing secretory protein

Translocation of proteins across the endoplasmic reticulum III. Signal recognition protein (SRP) causes signal sequence-dependent and sitespecific arrest of chain elongation that is released by microsomal membranes

Translocation of proteins across the endoplasmic reticulum I. Signal recognition protein (SRP) binds to m-wVro-assembled polysomes synthesizing secretory protein

Membrane proteins: Structure and assembly

Compilation of published signal sequences

M13 procoat and a preimmunoglobulin share processing specificity but use different membrane receptor mechanisms

Cell fusion by Semliki Forest, influenza, and vesicular stomatitis viruses

Membrane fusion proteins of enveloped animal viruses

Multiple mechanisms of protein insertion into and across membranes

Alterations in the transport and processing of Rous sarcoma virus envelope glycoproteins mutuated in the signal and anchor regions

Mutations of the Rous sarcoma virus env gene that affect the transport and subcellular location of the glycoprotein products

The 19-kDa glycoprotein precursor coded by region E3 of adenovirus

Secretion of a X2 immunoglobulin chain is prevented by a single amino acid substitution in its variable region

Molecular abnormality of human a r antitrypsin variant (Pi-ZZ) associated with plasma activity deficiency

Uncoating of influenza virus in endosomes

A stop transfer sequence confers predictable transmembrane orientation to a previously secreted protein in cell-free systems

The transmembrane segment of the human transferrin receptor functions as a signal peptide

Mutants of vesicular stomatitis virus blocked at different stages in maturation of the viral glycoprotein

Expression and function of transplantation antigens with altered or deleted cytoplasmic domains