key: cord-0894329-it5rm2ro
authors: Mayer, T; Tamura, T; Falk, M; Niemann, H
title: Membrane integration and intracellular transport of the coronavirus glycoprotein E1, a class III membrane glycoprotein.
date: 1988-10-15
journal: Journal of Biological Chemistry
DOI: 10.1016/s0021-9258(18)68131-1
sha: 05ac88d077d255e8a20a14bab1af967e0abbf99f
doc_id: 894329
cord_uid: it5rm2ro

The E1-glycoprotein (Mr = 26,014; 228 amino acids) of mouse hepatitis virus A59 is a class III membrane glycoprotein which has been used in this study as a model system in the study of membrane integration and protein transport. The protein lacks an NH2-terminal cleavable signal sequence and spans the viral membrane three times. Hydrophobic domains I and III could serve as signal sequences for cotranslational membrane integration. Domain I alone was sufficient to translocate the hydrophilic NH2 terminus of E1 across the membranes as evidenced by glycosylation of a newly introduced N-glycosylation site. The COOH-terminal part of E1 involving amino acids Leu124 to Thr228 was found to associate tightly with membranes at the post-translational level, although this part of the molecule lacks pronounced hydrophobic sequences. Membrane protection assays with proteinase K showed that a 2-kDa hydrophilic fragment was removed from the COOH terminus of E1 indicating that the protein is largely embedded into the membrane. Microinjection of in vitro transcribed capped and polyadenylated mRNA into CV-1 cells or into secretory AtT20 pituitary tumor cells showed that the E1-protein accumulated in the Golgi but was not detectable at the plasma membrane or in secretory granules. The 28 NH2-terminal hydrophilic amino acid residues play no role in membrane assembly or in intracellular targeting. Various NH2-terminal portions of E1 were fused to Ile145 of the cytoplasmic N-protein of mouse hepatitis virus. The resulting hybrid proteins were shown to assemble into membranes in vitro and were detected either in the rough endoplasmic reticulum or transient vesicles of microinjected cells.

The El-glycoprotein (Mr = 26,014; 228 amino acids) of mouse hepatitis virus A 5 9 is a class I11 membrane glycoprotein which has been used in this study as a model system in the study of membrane integration and protein transport. The protein lacks an NH2-terminal cleavable signal sequence and spans the viral membrane three times. Hydrophobic domains I and I11 could serve as signal sequences for cotranslational membrane integration. Domain I alone was sufficient to translocate the hydrophilic NH2 terminus of E l across the membranes as evidenced by glycosylation of a newly introduced N-glycosylation site. The COOHterminal part of E l involving amino acids to ThrZz8 was found to associate tightly with membranes at the post-translational level, although this part of the molecule lacks pronounced hydrophobic sequences. Membrane protection assays with proteinase K showed that a 2-kDa hydrophilic fragment was removed from the COOH terminus of E l indicating that the protein is largely embedded into the membrane. Microinjection of in vitro transcribed capped and polyadenylated mRNA into CV-1 cells or into secretory AtT2O pituitary tumor cells showed that the El-protein accumulated in the Golgi but was not detectable at the plasma membrane or in secretory granules. The 28 NH2-terminal hydrophilic amino acid residues play no role in membrane assembly or in intracellular targeting.

Various NHz-terminal portions of E l were fused to IleI4' of the cytoplasmic N-protein of mouse hepatitis virus. The resulting hybrid proteins were shown to assemble into membranes in vitro and were detected either in the rough endoplasmic reticulum or transient vesicles of microinjected cells.

Membrane proteins have been divided into three groups based on their specific orientation in the membrane (Wickner and Lodish, 1985; Garoff, 1985) . According to this classification of El-glycoprotein of MHV' A59 belongs to the group I11 proteins which span a membrane several times (Armstrong et al., 1984; Rottier et al., 1986) . The El-protein has three functional domains. The ectodomain representing the 28 NHn-terminal amino acids is hydrophilic and carries exclu-' The abbreviations used are: MHV, mouse hepatitis virus; RER, rough endoplasmic reticulum; ACTH, adrenocorticotropic hormone; WGA, wheat germ agglutinin; SDS, sodium dodecyl sulfate.

sively O-linked oligosaccharides which exhibit, in conjunction with the terminal amino acid sequence Ser-Ser-Thr-Thr-, blood group M activity (Niemann et al., 1984b) . A hydrophilicity analysis of E l according to Kyte and Doolittle (1982) reveals four internal hydrophobic stretches ( Fig. 1 ) that span the viral membrane three times and presumably contribute to the rigidity of the viral membrane. The carboxyl-terminal part of E l interacts with the viral nucleocapsid and thus plays an important role in the stages of virus formation (Sturman et al., 1980) .

Cell fractionation studies of MHV A59-infected cells indicated that the E l protein was synthesized on membraneassociated polysomes (Niemann et al., 1982) . In contrast to most other viral glycoproteins the E l protein could not be detected at the plasma membrane of infected cells other than in the form of virus particles. The intracellular distribution of E l was restricted to perinuclear regions (Doller and Holmes, 1980) and thus paralleled the sites at which budding of coronavirus particles was observed at early stages of infection (Becker et al., 1967; Holmes et al., 1981; Tooze et al., 1984) . Recent studies showed that this intracellular accumulation of the El-protein in smooth vesicles is not due to an interaction of E l with other coronavirus proteins but is an integral feature of the El-protein itself (Machamer and Rose, 1987; Rottier and Rose, 1987; Niemann et al., 1987) .

In this study we have used in vitro transcription/translation and microinjection techniques in combination with indirect immunofluorescence to study the membrane assembly process and the transport properties of the El-polypeptide in more detail. We show that the El-protein accumulates in perinuclear regions of fibroblasts and secretory cells. Based on the expression of various El-mutants we show that deletions or additional N-glycosylation of the amino-terminal domain of E l do not effect the Golgi-specific transport block. Internal hydrophobic domains I and I11 could mediate cotranslational integration of the polypeptide into microsomal membranes. An El-mutant lacking all three hydrophobic domains associates with membranes also post-translationally. We show that fusion proteins between various parts of the E l and a cytoplasmic protein integrate into membranes cotranslationally and accumulate in membranes of the RER and perinuclear vesicles. Kyte and Doolittle (1982) . The positions of restriction sites used for the construction of mutants are indicated. Restriction sites marked with a star were introduced by site-directed mutagenesis. Panel B, amino acid sequence of the El-polypeptide shown in the single letter code. Amino acid changes resulting from the generation of restriction sites are indicated. Charged residues are indicated by + orunderneath the sequence. Open boxes show sequences with a-helical probability according to Eisenberg et al. (1984) . Dots indicate the location of 8bends determined by the programs of Chou and Fasman (1978) and

Cid et al. (1982) . Panel C, construction of E l deletion mutants and process of the El-protein of MHV A59 we used in vitro synthesis of capped and polyadenylated El-specific mRNA from pSP65 vectors (Krieg and Melton, 1984) and its subsequent translation in the presence of translocation-competent microsomal membranes. To obtain polyadenylated transcripts, an oligo(dA-dT) fragment derived from pSVa970 (Min Jou et al., 1980) was inserted downstream from the E l coding sequences (Niemann et al., 1984a) as detailed in the Miniprint Section. Membrane translocation was assessed (i) by protection of the translocated domains from attack of exogenous proteinase K; (ii) by cosedimentation of the translated products with the microsomal fraction at neutral or alkaline pH; (iii) by glycosylation of a newly introduced N-glycosylation site at the NHZ terminus of the El-protein. Based on predictions of the secondary structure of the El-protein ( Fig. 1B ; Rottier et al., 1986 ) and on the hydrophobicity ( Fig. LA ; Kyte and Doolittle, 1982) we introduced additional restriction sites into the El-gene by site-directed mutagenesis. These sites were used to construct a set of deletion mutants and fusion proteins as indicated in Fig. 1C . To analyze the intracellular distribution of the individual proteins, the corresponding mRNA was microinjected into various cell types and the proteins were visualized by indirect immunofluorescence.

The Hydrophilic NHz-terminal Domain of the El-protein 

Aliquots were treated with proteinase K (P) in the absence or (-) tion products obtained in the presence of membranes yielded a truncated 24,000-dalton form. Rottier et al. (1985) have shown that this species represents the El-protein lacking a 2,000-dalton fragment from the carboxyl-terminal end. To assess lumenal exposure of the NHp-terminal domain, an Nglycosylation site ( -Am3-Thr-Thr-) was introduced into this region by site-directed mutagenesis. The resulting polypeptide, designated ElAsn, was indeed glycosylated in the presence of membranes, as indicated by the formation of a 29,000dalton species. The proteolytic cleavage product from this glycosylated species was larger (Mr 26,500) than that of Elwt, again demonstrating that in the absence of detergent the proteolytic attack occurred exclusively within the carboxylterminal part of the El-molecule. In the presence of detergent the ElAsn-species was degraded to a 15,600-dalton fragment as also obtained from Elwt, indicating that the N-glycosylation site was removed (data not shown). Consistent with the size of the deletions, the two mutants ElA4-23 and ElA4-28 generated integral membrane proteins that were about 2,500 or 3,000 daltons smaller than the Elwtpeptide. Both peptides were efficiently integrated into the membranes. Proteolysis gave products that were again about 2,000 daltons smaller than the original peptides indicating that their overall structure in the lipid bilayer remained unaltered. In the presence of detergent all El-mutants were degraded to the 15,600-dalton species indicating that the NH, terminus was removed under such conditions. As indicated by the size of this fragment and further evidence below, additional cleavage in detergent also removed parts of the COOHterminal tail.

To analyze which of the internal hydrophobic domains was essential for membrane integration, we produced mutants in which one or more of these domains were deleted. The results are summarized in Fig. 3 . A deletion of the first hydrophobic domain, as present in ElA4/50 (MI 20,100), neither prevented membrane integration nor did it alter the orientation of the protein in the membrane, as indicated by the proteolytic removal of the typical 2,000-dalton fragment. Analysis of ElA45-132 (Mr 15,000), retaining solely the first hydrophobic sequence, did not yield any detectable protected fragment. The results obtained with preprolactin control mRNA ( Fig.  3C ) indicated that the membrane preparation was not leaky for the protease. The ElA45-132-Asn molecule, carrying the newly created N-glycosylation site, yielded a glycosylated 18,500-dalton species. Treatment with endo-8-N-acetylglucosaminidase H created a third molecular species which was somewhat larger than the nonglycosylated form. The cotranslational addition of increasing amounts of an acceptor peptide for N-glycosylation (benzoyl-Asn-Leu-Thr-methylamide; Bause, 1983 ) revealed that only one of the two sites was glycosylated (data not shown).

The ElA23-123-Asn (Fig. 3A ) lacked all three hydrophobic domains and provided the NHp-terminal glycosylation site as a reporter group for lumenal exposure. This peptide was not glycosylated and was completely degraded by the protease even in the absence of detergent. This finding excludes the possibility that smaller El-peptides could diffuse through the membrane and provides further evidence that the COOHterminal hydrophilic part of E l was not intrinsically resistant to the protease.

Mutant ElA4-80 (Fig. 3C) , retaining hydrophobic domain 111, was inserted into the membranes. Treatment with the protease revealed that it was not secreted but remained anchored in the membranes. This domain seemed to be sufficient to stabilize the carboxyl-terminal part of the molecule within the membrane, since protease treatment removed only the COOH-terminal2,000-dalton fragment from ElA4-80. In the presence of detergent, however, the ElA4-80 molecule was degraded to an 8,500-dalton species. The size of this product in comparison to that obtained from ElA154-194 under detergent conditions (15,500 daltons) indicates that in both instances the resistant fragments contained hydrophobic sequences and parts from the COOH-terminal part of the Elmolecule. The deletion of amino acids 154-194 made the COOH-terminal region susceptible to proteinase K even in the absence of detergent, as evidenced by the release of a 4,500-dalton fragment yielding a peptide of almost the same size as the product obtained in the presence of detergent. When part of the hydrophobic domain I1 was deleted, as shown in Fig. 3C for ElA65-80-Asn, the overall topology of the mutant protein remained unaltered. Protease cleavage removed a 6500-dalton fragment and thus did not occur at the original site around amino acid 205, but about 40 amino acids displaced toward the NH, terminus yielding a protected fragment of about 21 kilodaltons. We interpret these findings to mean that part of the domain I1 helps to stabilize the COOH-terminal tail of E l in the membranes.

Co-and Post-translational Interaction of the El-mutants with Microsomal Membranes-To analyze whether membrane integration of the individual mutants was coupled to translation, we examined peptides, to which membranes had been added before or after their synthesis, for cosedimentation with the membranes at neutral or alkaline pH. The results of Fig. 4 show that all the molecular species retaining one of the hydrophobic domains integrated exclusively at the cotranslational level and were present in the pellet fraction. The finding that the peptides ElA45-132 and ElA4-80 were not released at alkaline pH further supports our conclusion that the hydrophobic domains I and I11 function simultaneously as signal and stop transfer sequences.

In contrast, peptide ElA23-123, although lacking all three internal hydrophobic domains, clearly associated with the membranes at the co-and post-translational level at either pH.

Membrane Assembly of El-N Fusion Proteins-We have constructed four El-N fusion proteins containing NHz-terminal El-specific sequences fused via the amino acid indi- cated to Ile145 of the nucleoprotein of MHV JHM (Fig. 1C) .

The results summarized in Fig. 5 revealed that all peptides with the exception of El-N(3-145) were integrated and anchored in the membranes. As demonstrated by the analyses of El-N(64-145) and El-N(80-145), the second hydrophobic domain or the remainder of it was also embedded into the membranes and thus protected against proteolytic attack yielding products of 10,400 and 11,200 daltons, respectively. Fragments of this size could not be derived from the nucleoprotein, since no proteolytic degradation products could be identified from El-N(3-145). El-N(207-145) yielded fragments in the protease protection assay that were indistinguishable from the corresponding fragment derived from Elwt, indicating the identical membrane topology of the fusion protein. The topology of the NH2 termini was verified by analyzing the corresponding variants carrying the newly created N-glycosylation site (data not shown).

Intracellular Transport Properties of the El-protein and Its Mutants-The in vitro synthesized mRNA was capped and polyadenylated in order to increase its half-life after microinjection into eucaryotic cells (Huez et al., 1981; Drummond et al., 1985) . The intracellular targeting of the El-proteins was studied by indirect immunofluorescence as detailed under "Experimental Procedures."

In agreement with published data (Machamer and Rose, 1987; Niemann et al., 1987; Rottier and Rose, 1987) , the Elprotein accumulated in perinuclear regions of the injected cells (Fig. 6B) . In double-labeling experiments these regions could not be distinguished from those recognized by the Golgispecific rhodamine-labeled wheat germ agglutinin (Fig. 6A) . The specific distribution of E l was observed in about 50% of the injected cells while the remaining cells did not respond with any synthesis of El-protein.

No E l could be detected on the surface of injected cells as judged by the failure of staining with polyclonal El-specific antibodies against virus particles and purified by elution from Western blots. In addition, no staining was obtained with antibodies directed against a synthetic peptide consisting of the eight NHz-terminal amino acids (Ser-Ser-Thr-Thr-Gln-Ala-Pro-Glu) of E l (data not shown). Even at late stages after injection or when 3-fold larger amounts of RNA (3 pg/ml) were injected, E l was absent from the plasma membrane. In such instances also the nuclear membrane and the RER of sylation of the amino-terminal domain of E l does not alter its intracellular transport properties.

In addition, a deletion of most of the hydrophilic NHzterminal domain had no influences on the intracellular targeting as shown for the ElA4-28-protein in Fig. 6 , G and H. Intracellular Transport of El-N Fusion Proteins- Fig. 7 shows the intracellular distribution of newly synthesized El-N-proteins. A monoclonal antibody directed against the nucleoprotein was used to detect the fusion proteins. In agreement with the observation that El-N(3-145) did not integrate into the membranes in uitro (Fig. 5) , the polypeptide was found dispersed throughout the cytoplasm of the injected cell (Fig. 7A) . In contrast, El-N(64-145) containing the first and part of the second hydrophobic domain accumulated in membranes of the RER (Fig. 7C) as indicated by double-labeling with a polyclonal antibody binding to the carboxyl-terminal domain of canine ribophorin I (Fig. 70) . Therefore, both antibodies bound to epitopes that were located at the cytoplasmic face of the RER. The El-N(80-145)-protein containing the first two membrane-spanning domains accumulated the injected cells contained El-protein (data not shown). This observation indicates that the El-protein is accumulating rapidly in membranes of the Golgi and piles up in the RER only after the former membranes are saturated (Tooze et al., 1984) .

To determine whether the perinuclear accumulation of E lprotein was a phenomenon restricted to fibroblasts, we injected mRNA into AtT20-cells, a transformed mouse pituitary gland cell line secreting ACTH. Again, the El-protein was present in the Golgi region of the injected cells (Fig. 60) . No E l was detectable at the cell surface (not shown), and no E l was present in peripheral secretory granules that were labeled with antibodies against ACTH (Fig. 6C) . in perinuclear membranes (Fig. 7E) which were not labeled with the ribophorin-specific antibody (not shown). Some of the El-N containing compartments were stained by the Golgispecific lectin (Fig. 7F) . The intracellular distribution E l -N(207-145) followed basically the pattern specific for the RER. The labeled structures, however, seemed to have a more vesicular character. By using WGA in similar double-labeling experiments it became obvious that these vesicles were not closely associated with Golgi compartments. It is feasible to assume that these vesicular structures represent transient vesicles which are derived from the RER and constitute the primary sites of virus maturation in the infected cell (Becker et al., 1967; Tooze et al., 1984) . DISCUSSION We have analyzed the topogenic signals and the intracellular transport properties of the glycoprotein E l of MHV A59, a class I11 membrane glycoprotein.

One of the models for the biosynthesis of polytopic membrane proteins suggests that these multispanning proteins are translocated into the endoplasmic reticulum membrane by alternating signal and stop transfer sequences (Friedlander and Blobel, 1985; Kopito and Lodish, 1985) . Recently Zerial et al. (1987) have demonstrated that foreign peptides could replace the internal signal and anchor sequence of the human transferrin receptor. These studies suggested that the hydrophobic character and the position in the molecule rather than the actual amino acid composition determine the character of a transmembrane sequence. In light of these findings we did not attempt to take the internal hydrophobic domains of the El-protein of MHV A59 completely out of their context by transferring them into different proteins. Instead, we have constructed deletion mutants and fusion proteins which retained authentic El-sequences either from the NH2 terminus or from the COOH terminus.

We show here that the domains I and I11 could function as signal and stop transfer sequences determining the topology of the El-molecule (Fig. 8) .

(i). The NHn-terminal hydrophilic domain of E l does not play a role in the membrane integration process or in determining the topology of the El-protein. No cleavable signal sequence is uncovered by the removal of this part of the Elmolecule which notably shows the largest degree of heterogeneity among different strains of coronaviruses (Lapps et al., 1987; Rasschaert et al., 1987; Boursnell et al., 1984) .

(ii). Hydrophobic domain I alone was sufficient to translocate the amino-terminal part of the El-molecule to the lumenal side as demonstrated by the glycosylation of the newly created N-glycosylation site in ElA45-132-Asn. No glycosylation was observed when membranes were added posttranslationally. The orientation of the ElA45-132 molecules is identical to that of the M2-protein of influenza virus (Lamb et al., 1985) but differs from that of other glycoproteins with internal uncleavable signal sequences such as the asialoglycoprotein receptor (Spiess and Lodish, 1986) , the human transferrin receptor (Zerial et al., 1986) , or the human glucose transporter (Mueckler and Lodish, 1986) . At present we do not know whether domain I can translocate only NHz-terminal sequences of a limited size. While the El-proteins from the bovine and the avian coronaviruses have hydrophilic ectodomains containing 28 and 22 amino acids, respectively, the corresponding ectodomain of the El-preprotein from transmissible gastroenteritis virus is 46 amino acids in length. Interestingly, this polypeptide is synthesized with an additional NH2-terminal cleavable signal sequence of 17 amino acid residues (Laude et al., 1987) .

(iii). The transmembrane domain I functioned as a stop transfer sequence, even though basic amino acid residues present in the cytoplasmic loop between domains I and I1 were removed together with domains I1 and 111. Clearly, ElA45-312 was not secreted into the lumen since the native glycosylation site (AsnZ7-Phe-Ser) was not glycosylated in this deletion mutant or in a corresponding El-N fusion protein.

(iv). The presence of a second signal sequence within the third hydrophobic domain was demonstrated by the analysis of ElA4-80. This protein was inserted into the membrane exclusively at the cotranslational level, and the peptide exhibited the authentic orientation (Fig. 8) . It has been shown previously that signal recognition particles exert a translational block as late as up to a point in the translation when two-thirds of the El-molecule (150 amino acids) have been synthesized (Rottier et al., 1985) . These data are in agreement with our observation that the third domain indeed functions as a signal sequence.

Our conclusion that the hydrophobic domain I1 of the Elprotein is not actively involved in the membrane insertion process is based on indirect evidence. First, the two polypeptides containing either a combination of domains I and I1 ) or I1 and I11 (present in ElA4-50) assembled in the membrane in the original orientation. Second, ElA65-80 which lacked the first half of domain I1 was integrated efficiently into membranes with the authentic topology, as indicated by N-glycosylation of the NH2 terminus. We interpret these findings to mean that membrane integration and orientation of domain I1 are predetermined by the presence of domains I and 111. However, our data do not exclude the possibility that domain I1 could function independently as a signal sequence.

The capability of the COOH-terminal tail of E l to associate with membranes post-translationally was unexpected. This behavior may reflect the natural role of E l as a matrix protein guiding the viral nucleocapsid to the place of virus budding (Sturman et al., 1980) .

Our microinjection experiments indicated that the El-protein has an intrinsic signal for a retention in Golgi-like compartments in fibroblasts and secretory AtT2O cells. This retention signal of the El-protein is functional in the absence of other viral proteins. Similar results have been obtained previously for the El-protein of avian infectious bronchitis virus (IBV) (Machamer and Rose, 1987) and for the Elprotein of MHV A59 using DNA expression vectors (Niemann et al., 1987; Rottier and Rose, 1987) . In MHV A59-infected AtT2O cells virus particles were shown to bud into pre-Golgi compartments and then share the secretory pathway with the secretory protein ACTH through the same Golgi stacks into the trans-Golgi network. At this site the constitutive secretory pathway for the virus and the regulated secretory pathway for the hormone diverged (Tooze et al., 1987) . We show here that this transport property was also shared by the isolated E lprotein, since it was not detected in secretory post-Golgi vesicles filled with ATCH.

Studies by Machamer and Rose (1987) demonstrated that the first transmembrane domain of the protein from the infectious bronchitis virus was responsible for its retention in the Golgi while a protein retaining only the third transmembrane domain was transported to the plasma membrane. Unfortunately, we were unable to detect El-peptides after microinjection of mRNA encoding ElA45-132 and ElA4-80. At present we do not know whether this is due to an instability of the corresponding mRNA, whether the protein synthesized in vivo was degraded, or whether it was too dispersed throughout the cells to be detected with the antibodies.

The described modifications of the ectodomain of the E lmolecule had no influence on the El-specific transport properties. To assess the applicability of parts of the El-molecule to direct fusion proteins into the Golgi, we have microinjected mRNA encoding various parts of the El-protein fused in frame to a carboxyl-terminal part of the cytoplasmic Nprotein of MHV JHM. Each of the fusion proteins containing one or more of the hydrophobic domains of E l was detected in perinuclear membranes. The fusion proteins El-N(64-145) and El-N(80-145) were not transported into the Golgi indicating that particular nucleoprotein-specific sequences added to the cytoplasmic COOH terminus prevented release from the RER. Only in very few cells the intracellular distribution of El-N(207-145) overlapped with the Golgi pattern as stained by WGA, and it was also different than the pattern obtained with RER-specific antibodies. We suggest that the compartments harboring the El-N(207-145) are transient vesicles which in the virus-infected cells are the sites of particle formation. Experiments involving immunoelectron microscopy on cells infected with recombinant vaccinia virus are currently in progress. was obtained by treafinq ScaI-dlgeated ElH13mpl9-DNA with 8.131 and subsequent liqation of the truncated El-sequences into E1113mp19 that was Cleaved with ScaI and XindIII. C l~n e pSP65Elb65180 was obtained by insertion of the 750 bp SspI-HindIIl-traqmenr from pSP65EllIle811 into psP65E1 previously cleaved with Ball and HindIII. To delete codinq sequences correapondlng to the first and second membrane *Panning domain. pSP65E11IleS11

was llnearlrsd with Ssp1 and partially dlqerted xlth Scal.

The 3199 bp fragment was isolated and religated to Yield pSP65Eld4/80. The deletion mutant pSP65ElA451132. lackinq the sequences Corresponding to the second and third hydrophobic domain.

was produced by cleavinq pSP65ElIThr451 with KpnI, deletion o f the 244 bp fraqnmnt and reliqatlon. For the Conltructlon of pSP65ElA231123. encodinq an El-protein leckinq all the lnternal hydrophoblc domains. pSP65E1lLys1261 wan treated with AflII. the 3766 bp-fragment was laolated and reliqated. PSP65ElA1541194 was constructed in the folloYln9 manner: pSP65ElA45/132 wan cleaved rich Ball and ACCI. the 5"PrOtrUdinq ends were fllled with Xlsnov polymerase and the vector-fraqment was relimated to yleid pSP65ElA45/132'61541194. A 367 bp fraqment carrying the desired deletlon was isolated affer'dlgestion with Kpnl and HindIII and was used to replace the corresponding fraqment in the wild type El-gene. Mutants pSP65ElAsnb451132 and pSP65ElAsnA.23/123 were obtained by inserting the Af1II-HlndlII-fraqments from pSP65E16451132 and pSP65ElA231123. re.pectivsly. into pSP65ElAan digested with Af1II and HindIII. Fig. 9 

A 5 0 1 8 1 1 to Ilel811

5"CTTTATCAGGACT&GTAGCTGGTGG-l 1SPeII EllLys1261 Met11261 to Lysl1261

IAflIII

Proc. Natl. Acad. Sci

Molecular Cloning, A Laboratory Manual

5045-Spiess

Acknowledgments-We thank David Meyer (UCLA, Los Angeles) and John Tooze (EMBL, Heidelberg) for antibodies directed against ribophorin and ACTH, respectively. We are indebted to Bernhard Dobberstein (EMBL, Heidelberg) for providing dog pancreatic membranes and for fruitful discussions. We thank Juan Ortin (Universitad Autbnoma, Madrid) and Carl Blobel (University of California, San Francisco) for plasmids pSVa970 and pB4.