Role of cytoplasmic C-terminal amino acids of membrane proteins in ER export

Newly synthesized soluble and membrane-bound secretory proteins leave the ER at part-rough part-smooth transitional elements(Palade, 1975) and are transported to the Golgi apparatus via the ER-Golgi intermediate compartment(ERGIC) (Hauri et al., 2000a). ER export is mediated by vesicles(Schekman and Orci, 1996) that fuse either with pre-existing tubulovesicular clusters of the ERGIC (according to the stable compartment model) or with one another to form tubulovesicular clusters (according to the maturation model). The ERGIC clusters are the first station at which retrograde and anterograde cargo separate(Aridor et al., 1995;Klumperman et al., 1998). Subsequently, secretory cargo is delivered to the cis-Golgi by a highly dynamic process involving microtubules(Saraste and Svensson, 1991;Presley et al., 1997;Scales et al., 1997;Klumperman et al., 1998;Pepperkok et al., 1999;Stephens et al., 2000).

The mechanism of protein export from the ER remains controversial. According to the bulk-flow model (Wieland et al., 1987), secretory proteins contain no ER-export signals and are thus packaged into transport vesicles by default. The alternative mechanism of selective sorting predicts the presence of sorting signals and a mechanism of signal decoding (Lodish,1988; Kuehn and Schekman,1997). Evidence is mounting that ER export is selective, at least for some membrane proteins. These proteins are concentrated in vesicles during the budding process. They include VSV (vesicular stomatitis virus)-G protein,KDEL receptor and SNARE proteins (Balch et al., 1994; Bednarek et al., 1995;Klumperman et al., 1998;Martinez-Menarguez et al.,1999).

Two ER-export motifs present in the cytoplasmic domain of membrane proteins have been characterized. A diphenylalanine (FF) motif in the cytoplasmic domain of ERGIC-53 (Kappeler et al.,1997) and p24 proteins(Fiedler et al., 1996;Dominguez et al., 1998) was shown to be required for selective export from the ER. Similarly, a diacidic motif was found to be required for the efficient ER-to-Golgi transport of VSV-G protein (Nishimura and Balch,1997; Sevier et al.,2000).

Both budding of transport vesicles from the ER and selective recruitment of membrane proteins into these vesicles are mediated by COPII proteins. The mechanism of COPII vesicle budding has most completely been elucidated in yeast, but it appears to be similar in all eukaryotes(Barlowe et al., 1994;Schekman and Orci, 1996;Tang et al., 2000). COPII vesicle formation requires the recruitment of the small GTPase Sar1 to the ER by the guanine-nucleotide-exchange factor Sec12p, followed by the sequential recruitment of the heterodimeric coat complexes Sec23p-Sec24p and Sec13p-Sec31p (Matsuoka et al.,1998; Aridor et al.,2001). The polymerization of COPII drives vesicle budding in vitro. In higher eukaryotes, two Sec23p and four Sec24p isoforms have been identified, whereas there are three Sec24p homologues in yeast(Paccaud et al., 1996;Pagano et al., 1999;Roberg et al., 1999). Two of the Sec24p homologues are not essential for viability in yeast, but the precise role of the Sec24p isoforms is not known. In vivo, the budding process may require additional proteins, which includes the putative scaffolding protein Sec16p (Shaywitz et al.,1997) and the putative phospholipase p125(Tani et al., 1999;Mizoguchi et al., 2000).

Unlike membrane proteins, which may possess sorting signals recognizing COPII coat subunits (Kappeler et al.,1997; Dominguez et al.,1998; Kuehn et al.,1998; Springer and Schekman,1998; Aridor et al.,2001), the selection of soluble cargo requires receptors that interact with the COPII coat. The mannose lectin ERGIC-53(Hauri et al., 2000a) and p24 proteins (Muniz et al., 2000)are candidates for such cargo receptors. ERGIC-53 is a non-glycosylated type I glycoprotein that is most highly concentrated in the tubulovesicular clusters of the ERGIC near the Golgi apparatus and in the cell periphery. It forms disulfide-linked homodimers and homohexamers in the ER(Schweizer et al., 1988) and cycles in the early secretory pathway, a process that involves COPII coats for anterograde and COPI coats for retrograde transport(Kappeler et al., 1997). ERGIC-53 is required for efficient transport of cathepsin C(Vollenweider et al., 1998), a cathepsin-Z-related protein (Appenzeller et al., 1999), and coagulation factors V and VIII(Nichols et al., 1998;Neerman-Arbez et al., 1999;Nichols et al., 1999).

The C-terminal FF motif of ERGIC-53 is not only required for its efficient ER export but also mediates binding to COPII proteins as revealed by peptide binding studies (Kappeler et al.,1997). Since the FF export motif is not strictly conserved in eukaryotes we investigated to what extent it tolerates other amino acids. Here, we show that the FF motif can be functionally substituted by at least three types of motifs. These motifs also bind to COPII components in vitro. Most remarkably, a single C-terminal valine mediates efficient ER export and is transplantable to other reporter proteins. The results imply an important role for aromatic and hydrophobic C-terminal amino acids in mediating efficient transport of membrane proteins by interacting with COPII coat components.

Mouse monoclonal antibody (mAb) 9E10.2 against c-myc epitope, mouse mAb HP2/6.1 against human CD4 (kindly provided by F. Sanchez-Madrid), mouse mAb A1/182/5 against BAP31 (Klumperman et al.,1998), rabbit antibodies against human Sec23p, sec24Bp, sec24C(Paccaud et al., 1996) and p125 (Tani et al., 1999) were used.

Affinity-purified FITC-goat-anti-mouse (Cappel, USA) and IgG1-specific FITC-goat-anti mouse or IgG2a-specific TRITC-goat-anti-mouse antibodies(Southern Biotechnology, USA) were used. Cell culture media and reagents were from GibcoBRL (Basel, Switzerland).

Standard molecular biology protocols were adapted from Ausubel et al.(Ausubel et al., 1997) or Sambrook et al. (Sambrook et al.,1989). Oligonucleotides were from Microsynth (Switzerland). ERGIC-53 constructs have been described previously(Itin et al., 1995;Kappeler et al., 1997). Briefly, a c-myc epitope and an N-glycosylation site were introduced into ERGIC-53 cDNA by PCR-based splicing and mutagenesis. The construct was cloned into pECE vector, and PCR-generated mutants were introduced as AccI/XbaI fragments. To replace the transmembrane domain(TMD) of ERGIC-53 by 18 leucines, a BglII restriction site was introduced by silent mutagenesis changing the codon for R₄₉₉ to AGA. Equal amounts of two complementary oligonucleotides encoding VHL₁₈R with overlapping cohesive ends corresponding to AccI and BglII sites were annealed in 2× SSC buffer(30 mM sodium citrate and 300 mM NaCl, pH 7.0) by heating it to 95°C for 5 minutes and cooling ∼1°C per minute to room temperature (RT). The resulting dsDNA fragment was ligated with the large fragment of AccI/BglII digested ERGIC-53 construct. The CD4L₁₈R₂A₁₀ construct was prepared in two steps. First, a hybrid of the CD4 luminal domain with a stretch of 18 leucines and the cytoplasmic tail of ERGIC-53 was produced by PCR splicing using templates CD4KKFF (Andersson et al.,1999) and the ERGIC-53 construct with the L₁₈ TMD described above. The resulting fragment was cloned as a Bsu36I/XbaI fragment into CD4 cDNA in pcDNA3.1 vector(Andersson et al., 1999). The ERGIC-53 tail was then substituted with a polyalanine tail by using a PCR-generated cDNA fragment and Bsu36I/Xba sites. Additional PCR-generated mutants were similarly introduced as Bsu36I/XbaI fragments. Polyserine tails were fused to ERGIC-53 and CD4 constructs by PCR mutagenesis. Additionally, an AccIII restriction site was introduced adjacent to the TMD by silent mutagenesis, which changed the codon of the first R in the tail to CGG. This site allows insertion of fragments coding for cytoplasmic tails at AccIII and XbaI sites. cDNAs encoding cytoplasmic tails (seeFig. 6) were prepared by either annealing complementary oligonucleotides as described above or by gene synthesis. All mutations were confirmed by sequencing using standard methods and an ABI Prism 310 Genetic Analyser (PE Applied Biosystems, Rotkreuz,Switzerland).

COS-1 cells were grown in 35-mm dishes and transfected by the DEAE-Dextran method (Kappeler et al.,1997). 42 hours later the cells were pulsed with[³⁵S]-methionine for 10 minutes, chased in the presence of 10 mM of unlabeled methionine and subjected to immunoprecipitation with anti-myc and anti-CD4. Endo H digestion of ERGIC-53 constructs was performed as described in Kappeler et al. (Kappeler et al.,1997). Proteins were separated on 7-10% gradient SDS-polyacrylamide gels and visualized by fluorography. Fluorograms were quantified with a ChemImager™ device and AlphaEase™ software(Alpha Inotech Corporation, USA).

COS-1 cells cultured in poly-L-lysine-coated 8-well glass chamber slides were transfected with 100 ng DNA per well by the DEAE-Dextran method and processed for immunofluorescence after 24 hours. For surface labeling, cells were incubated with anti-myc antibodies for 30 minutes on ice. The cells were washed three times with cold medium and twice with cold PBS containing 0.9 mM CaCl₂ and 0.45 mM MgCl₂ (D-PBS). They were fixed with 3%paraformaldehyde for 30 minutes in D-PBS. After washing twice with PBS, the reaction was quenched twice with 20 mM glycine in PBS for 5 minutes in total. Non-specific binding was blocked by three washes of PBS/1% BSA for 10 minutes. The secondary FITC-goat-anti-mouse antibody was incubated in PBS/1% BSA for 30 minutes. The cells were washed four times with PBS/1% BSA, twice with PBS and embedded in 90% glycerol/10% PBS/0.1% 1,2-phenylendiamine. For double staining, cells were washed twice with D-PBS and fixed. After quenching, the cells were permeabilized with PBS/0.1% saponin/20 mM glycine for 20 minutes at RT and incubated with anti-myc and anti-BAP31 antibodies for 30 minutes. Cells were washed four times with PBS/0.1% saponin and incubated with FITC-goat-anti mouse IgG1 and TRITC-goat-anti-mouse IgG2a for 30 minutes. After four washes with PBS/0.1% saponin and two washes with PBS, the cells were embedded and analyzed with a Reichert Polyvar immunofluorescence microscope.

This assay was performed as previously described(Kappeler et al., 1997). Briefly, synthesized peptides of defined sequence were coupled to thiol Sepharose 4B. Trition X-100 extracts of HepG2 cells were incubated with the peptide-beads for 2 hours at 4°C under low salt conditions (50 mM HEPES,90 mM KCl, 2.5 mM MgOAc, 1% Triton X-100; pH 7.3). Bound proteins were eluted and analyzed by SDS-PAGE followed by immunoblotting.

The FF ER-export motif is not strictly conserved in ERGIC-53 orthologs(Fig. 1A). Although present in ERGIC-53 of man, monkey, rat and frog, it is replaced by a phenylalanine-tyrosine motif in Drosophila and Caenorhabditis elegans and by two leucines in yeast.

We wondered if other amino acids can functionally substitute for the FF motif and focused particularly on aromatic and hydrophobic residues. To this end, we generated mutants of a glycosylated variant of human ERGIC-53 in which the dilysine signal was replaced by alanines to prevent recycling(Fig. 1B)(Kappeler et al., 1997). These mutants carry doublets of aromatic or hydrophobic amino acids at positions -1 and -2. The mutants were expressed in COS cells, and their transport from ER to Golgi was probed by endo H. None of the mutations interfered with oligomerization, which indicates that the proteins are correctly folded (not shown). As shown in Fig. 1C,D,the FF motif mediated efficient transport of the reporter, whereas two alanines in place of the phenylalanines were inefficient. Previous studies using sucellular fractionation demonstrated that the FF motif enhances ER exit rather than transport between ERGIC and Golgi(Kappeler et al., 1997). A dityrosine was only slightly less efficient than the FF motif in mediating transport (Fig. 1C,D). The hydrophobic motifs dileucine, diisoleucine and divaline mediated transport at least as efficiently as an FF motif (Fig. 1C,D). Dimethionine (Fig. 1D) or dicysteine (data not shown) were inefficient compared with dialanine.

Next we investigated whether a single aromatic or hydrophobic amino acid in position -1 or -2 suffices for efficient transport.Fig. 1E shows that a single phenylalanine or tyrosine at position -2 accelerated transport as efficiently as two aromatic residues. Only partial acceleration was obtained with a phenylalanine at position -1. Conversely, a single valine at position -1 was sufficient to accelerate transport as effectively as divaline, but a valine in position -2 was ineffective. By contrast, a single leucine or isoleucine at either position was unable to mediate efficient transport. Overall, this analysis revealed three types of ER export motifs: (1) a single phenylalanine or tyrosine at position -2, (2) two leucines or isoleucines at positions -1 and -2 and (3) a single valine at position -1.

To test if the novel transport motifs also bind to COPII proteins we coupled peptides comprising the motifs to thiol-Sepharose and incubated the beads with pre-cleared homogenates of HepG2 cells. Bound proteins were separated by SDS-PAGE, and COPII binding was probed by blotting with antibodies against Sec23p, the Sec24p isoforms Sec24Bp and Sec24Cp(Pagano et al., 1999) and the Sec23p-interacting protein p125 (Tani et al., 1999). This assay is specific for COPII binding(Kappeler et al., 1997). Sec23p bound to the FF motif as expected, but no significant binding to a peptide bearing two C-terminal alanines was observed(Fig. 2A). Sec23p binding was also observed for the dityrosine motif. Likewise, the divaline and diisoleucine motifs showed efficient Sec23p binding, whereas the dileucine motif bound somewhat less efficiently but consistently stronger than dialanine.

Interestingly, some selectivity was observed for the binding of the two Sec24p isoforms. FF and dityrosine motifs bound more efficiently to Sec24Bp than Sec24Cp, whereas the divaline motif showed an inverse binding preference. Binding of the dileucine and diisoleucine motifs to Sec24Bp and Sec24Cp was low and comparable to Sec23p binding. Binding to p125 was comparable to Sec24Bp binding in that the aromatic motifs bound more efficiently than the hydrophobic motifs.

Next we tested the binding of single amino-acid motifs.Fig. 2B shows that aromatic residues mediated efficient binding to Sec23p and Sec24Bp when located at position -2. By contrast, the valine required a -1 position for best interaction with Sec23p and Sec24Cp. Single leucines or isoleucines did not exhibit any binding preference in either position (data not shown). Binding of the single amino-acid motifs to p125 was similar to the binding of Sec23p and the Sec24p isoforms to p123.

Collectively, the results of the in vitro COPII binding assay are consistent with the data from the transport assay. Motifs active in the transport assay also displayed binding to COPII, albeit with different preferences.

To test for transplantability, we appended the motifs to the type I plasma membrane glycoprotein CD4 (Fig. 3A). The cytoplasmic domain of CD4 was replaced by two arginines followed by 10 alanines, and the TMD was replaced by 18 leucines, because it contains targeting information (Kappeler et al., 1997). This change slowed the ER-to-Golgi transport of the reporter when compared to wild-type CD4 (not shown). The amino acids at position -1 and -2 were then replaced by the transport motifs described above. Correct topology of the constructs was confirmed by protease protection experiments (not shown). Transport was assessed by pulse-chase/endo H experiments. CD4 bears two N-glycosylation sites, one of which acquires complex glycans in the Golgi. Fig. 3 shows that a single valine in position -1 or two valines in positions -1 and -2 significantly accelerated transport of the reporter. The other motifs were as inefficient as two alanines. These results suggest that a single valine acts as a transport signal.

To further characterize the valine signal, we studied the steady-state distribution of reporter constructs by immunofluorescence microscopy. In pilot experiments, we noticed that both the ERGIC-53 and the CD4 reporters were transported to the cell surface irrespective of the presence or absence of a valine. This may be due to additional transport determinants in these proteins. In an attempt to obtain more efficiently retained reporters, we further modified ERGIC-53 (Fig. 4). These constructs possess the luminal domain of ERGIC-53 in addition to a myc-tag and a glycosylation site(Itin et al., 1995), a TMD of 18 leucines and a cytoplasmic domain of two arginines followed by alanines or serines. None of these changes affected the topology and folding of the reporter as indicated by unaltered oligomerization(Fig. 4A). Substituting the TMD of ERGIC-53 by leucines reduced transport dramatically despite the presence of the FF motif (Fig. 4B,construct 2 versus 4). This reduction is due to the elimination of additional transport determinants within the TMD of ERGIC-53 (O.N. and H.-P.H.,unpublished). By contrast, a single C-terminal valine accelerated transport(Fig.4B, construct 6 versus 4). The presence of a polyalanine tail (construct 5) was still strong enough to locate this construct to the cell surface (not shown). Transport could be further reduced by a polyserine instead of a polyalanine tail(Fig. 4B, construct 7), which maintained the transport signal function of the valine (construct 8).

Surprisingly, a C-terminal valine could even mediate transport when the ERGIC-53 reporter was prevented from forming disulfide-linked oligomers by substituting the cysteines at position 466 and 475(Fig. 4B, construct 10 versus 9). These cysteines are important for oligomerization but do not contribute to folding of the luminal domain as indicated by the fact that monomeric ERGIC-53 has lectin activity (Appenzeller et al.,1999; Lahtinen et al.,1999). By contrast, FF-dependent transport requires disulfide-bond-mediated oligomerization of ERGIC-53 (F.K. and H.-P.H.,unpublished).

COS cells transfected with either the construct L53L₁₈R₂S₁₀ or its monomeric form mL53L₁₈R₂S₁₀ exhibited no myc-specific cell surface labeling in non-permeabilized cells, very much in contrast to L53L₁₈R₂S₉V and mL53L₁₈R₂S₉V, which showed strong cell-surface staining (Fig. 5). A few cells expressing very high levels of the constructs without the valine showed weak cell-surface staining. Permeabilized cells expressing the constructs with a C-terminal valine showed strong cell-surface fluorescence,whereas the constructs without the valine gave a reticular pattern including a nuclear ring that costained with an ER marker. This indicates that the constructs without a C-terminal valine were largely retained in the ER and hence can be considered signalless. These morphological findings are consistent with those of the transport assays and suggest that a C-terminal valine is an ER-export signal that can provide ER export of a signalless reporter protein.

To more systematically investigate the position dependence of the valine signal, we tested the effect of additional tail mutants. A single valine was introduced at different positions of polyserine tails attached to L53L₁₈R₂(Fig. 6A). Shifting the valine from the -1 to the -2 or -3 position of the 12-residue tail abolished transport signal activity (Fig. 6A, compare constructs 2, 3 and 4). Lengthening the L53L₁₈R₂S₉V construct by two or four serines also inactivated the signal (Fig. 6A, constructs 5 and 6). Likewise, a C-terminal valine of shortened tails did not accelerate transport(Fig. 6A, constructs 7 and 8). These experiments indicate that the valine requires a C-terminal position and a minimal tail length. To test whether the nature of the -2 position affects the valine signal we introduced positively (i.e. arginine and lysine) and negatively charged (i.e. aspartic acid and glutamic acid) amino acids. They were found to have no effect on valine-mediated transport(Fig. 6B). Tryptophan only slightly reduced the level of transport. Although we have not tested all the amino acids, these results suggest that the penultimate amino-acid position has little effect on the signal function of a C-terminal valine.

How common are the C-terminal transport motifs we have identified among type I proteins? We screened non-redundant databases from Swiss-Prot (release 39.8) and trEMBL (release 15.4) for `type I membrane proteins' of homo sapiens or mammalia and determined the frequency of C-terminal transport motifs. We first screened for motifs that were functional in our transport assays: a valine at C-terminal position -1, a phenylalanine or tyrosine at position -2 and two leucines or isoleucines at positions -1 and -2. Combinations of two functional motifs were excluded. A single C-terminal valine was found in 9.8% of 488 human type I membrane proteins(Table 1). This is 1.7-fold higher than theoretically expected from the frequency of valine in proteins. Moreover, a valine was found 2.7 times more often in position -1 than -2. Similarly, a single phenylalanine or tyrosine at -2 position appeared more frequently than at the -1 position. An aromatic residue at position -2 and a valine at position -1 (a combination not tested experimentally in our study)were 1.9 times more prevalent than expected. Overall, 18.2% of the analyzed human type I proteins had at least one of the functional motifs. By contrast,non-functional motifs, such as C-terminal tryptophans in the absence of any transport motif, made up only 0.4%. This frequency is much lower than theoretically expected from the prevalence of tryptophan. The results for 1255 mammalian type I membrane proteins were rather similar. Functional transport motifs at the C-terminus made up 19%. A valine at position -1 or a phenylalanine or tyrosine at position -2 appeared up to 2.9 times more often than in the non-functional position. Remarkably, none of the positive hits was an ER-resident protein.

Our observation that many non-ER type I membrane proteins possess C-terminal transport motifs is consistent with the notion that selective ER export is a common mechanism.

In the current study, we have uncovered an unanticipated role for the last two C-terminal amino acids of mammalian type I membrane proteins in ER export. We found that the previously defined C-terminal FF ER-export motif of ERGIC-53 can be functionally substituted by two tyrosines, leucines, isoleucines or valines. Further analysis revealed three different minimal transport motifs: a single phenylalanine or tyrosine at position -2, two leucines or isoleucines at positions -1 and -2 or a single valine at position -1. Thus, although not identical, the C-terminal ER-export motif is functionally conserved in all known ERGIC-53 orthologs. A C-terminal valine motif has not been observed yet in any ERGIC-53 protein, but the number of known ERGIC-53 ortholgs is still rather limited.

Although the three motifs exhibited similar efficiencies in mediating ER export, they have different physicochemical properties. Phenylalanine and tyrosine are aromatic, whereas leucine, isoleucine and valine are apolar. Interestingly, neutral residues, such as methionine or cysteine, or the polar residue serine did not accelerate transport, although they are comparable in size to leucine, isoleucine or valine. The apolar character of valine, leucine or isoleucine is unlikely, however, to be the main determinant of efficiency of transport as the apolar alanine was non-functional. It is more likely that a branched methyl group or a non-polar side chain (which has to be a certain size) are important. Alanine contains only a single methyl group, whereas the side chains of the other apolar amino acids are more complex. With the aromatic motifs, size may be a critical factor. Phenylalanine was more efficient than the slightly larger tyrosine, whereas the bulky tryptophan,although comparable in polarity to tyrosine, was nonfunctional.

Another difference between the transport motifs concerns position dependence. Aromatic residues were found to function best in the -2 position,whereas valine must be in position -1 to be effective. There is no obvious explanation for this difference. Our database searches for the different transport motifs in mammalian type I membrane proteins revealed a preference for aromatic residues in the -2 and of valine in the -1 position, which supported the notion that the characteristic position of these amino acids is critical for function. Most notably, none of these proteins turned out to be ER-resident proteins. While a single aromatic residue is clearly more efficient in the -2 than the -1 position, we have not tested whether such a residue would also operate in other positions. However, previous studies with ERGIC-53 showed that two phenylalanines in positions -4 and -5 mediated transport as efficiently as in positions -1 and -2 [seefig. 6 of Itin et al.(Itin et al., 1995)]. Likewise, p24 proteins possess a conserved diphyenylalanine motif within their cytoplasmic domain (Stamnes et al.,1995; Dominguez et al.,1998). The p24 proteins constitute a complex of homologous type I membrane proteins that cycle in the early secretory pathway and are implicated in cargo selection in the ER (Schimmoller et al., 1995; Kaiser,2000; Muniz et al.,2000). When this internal diphenylalanine motif was mutated to alanine, the p24 complex was redistributed from a post-ER to a more ER-like location, and the interaction with COPII coat proteins in vitro was abolished(Dominguez et al., 1998). In reporter-p24 tail chimeras this internal diphenylalanine motif was found to affect transport in conjunction with other targeting determinants, although its precise role is controversial (Fiedler et al., 1996; Nakamura et al.,1998). Collectively, these findings suggest that the diphenylalanine motif can function at various positions within the C-terminal domain of type I membrane proteins. Whether or not the dileucine and the diisoleucine motifs must be in the C-terminal position remains to be shown.

The C-terminal valine was the only motif that accelerated ER export of a modified, slowly transported CD4 reporter. Hence, a C-terminal valine is a signal in the true sense. The other transport motifs depend on additional transport information. This is best characterized for the FF motif of ERGIC-53, which acts in concert with other transport determinants(Itin et al., 1995;Kappeler et al., 1997;Hauri et al., 2000b). By contrast, the valine signal appears to have no such requirement. It mediated efficient transport of ERGIC-53 constructs, having either a polyalanine (or a polyserine) tail or its TMD substituted by polyleucine. Moreover, endogenous transport determinants of ERGIC-53 are dependent on covalently linked oligomerization (F.K. and H.-P.H., unpublished observations), whereas the valine signal operated equally well in disulfide-bond-mediated oligomeric and apparently monomeric ERGIC-53 constructs.

Previous studies have reported a requirement for a C-terminal valine for efficient transport of TGFα, MT1-MMP and CD8, but in none of these cases was valine tested to discover whether it was sufficient for transport(Briley et al., 1997;Urena et al., 1999;Iodice et al., 2001). Nevertheless, these findings are consistent with our observation that a C-terminal valine operates as an ER-exit signal. A C-terminal valine might also play a role in the transport of p24 proteins. If the tail of yeast Emp24p was fused to invertase and the C-terminal leucine-valine motif was substituted by two alanines, transport to the Golgi was reduced(Nakamura et al., 1998). Both residues were required for maximal transport efficiency of the chimera, but the C-terminal valine was more important than the leucine. Some of the p24 proteins carry a diisoleucine, a leucine-isoleucine or a isoleucine-leucine motif in place of the leucine-valine motif, again suggesting that the transport motif is functionally conserved. However, it remains to be shown whether these C-terminal motifs are indeed functional in intact p24 proteins.

Cytoplasmic C-terminal transport motifs may also contribute to efficient trafficking of polytopic membrane proteins. Evidence for this suggestion is our finding that the C-terminal aromatic/hydrophobic amino acids of presenilin 1 and presenilin 2 are required for efficient transport (O.N. and H.-P.H.,unpublished). Some other polytopic membrane proteins carry a C-terminal putative valine signal. This group includes aquaporins, plasma membrane calcium ATPase, band 3 anion exchanger, various potassium channels, claudins,glucose transporters and multidrug resistant proteins.

VSV-G protein and some other membrane proteins carry a diacidic (DXE)ER-export motif in the cytoplasmic domain(Nishimura and Balch, 1997). This motif is part of a larger six-residue signal that also includes a YxxØ endocytosis motif (Sevier et al., 2000). Obviously, the anterograde transport signal of VSV-G is based on a complex interplay of at least two distinct determinants. By contrast, the valine signal can act independently of any other determinants and features.

How are the C-terminal ER-export motifs decoded? Our in vitro binding studies point to a mechanism involving COPII proteins. All the transport motifs we have defined mediate the binding of COPII components from HepG2 lysates. The individual double motifs exhibited different binding patterns. Strongest binding of COPII was observed for FF and dityrosine, whereas binding to di-isoleucine and particularly dileucine was weak, although it was consistently more prominent than for dialanine. The bead assay we have used is difficult to accurately quantify and must be considered semi-quantitative. Future experiments will be required to determine binding affinities for the different signals.

An interesting difference was noted for the two Sec24p isoforms. Sec24Bp(Iss1p in yeast) showed stronger binding to the aromatic motifs than to the valine motif, whereas Sec24Cp (Lst1p in yeast) exhibited an inverse binding preference. The single amino-acid minimal transport motifs showed binding features that were largely similar to those of the double amino-acid motifs. The binding preference of the transport motifs is consistent with the notion that individual isoforms of Sec24p may provide some selectivity in the recruitment of cargo during COPII vesicle budding(Pagano et al., 1999;Roberg et al., 1999;Kurihara et al., 2000). That Sec24p isoforms may determine some selectivity of cargo recruitment is supported by a recent study in yeast(Shimoni et al., 2000). These authors found that cytosol from a 1st1-null strain in yeast supported packaging of alpha-factor into COPII vesicles but was deficient in the packaging of the ATPase Pma1p. Packaging of Pma1p was most efficient with a mixture of Sec23-Sec24p and Sec23-Lst1p. The combinatorial subunit composition of the coat might serve to expand the range of cargo to be packaged into COPII vesicles. Our findings are in line with and extend this concept. They suggest that the nature of the ER-export motif may determine the COPII subunit composition that recruits the cargo into budding vesicles.

In conclusion, our results provide strong evidence that a considerable subset of mammalian membrane proteins contains short C-terminal transport motifs that facilitate ER export by interaction with COPII coat components. We propose that the motifs directly interact with COPII coat subunits. Binding experiments with purified COPII proteins will be required to test this notion. The findings reported here, that C-terminal motifs accelerate protein export from the ER, may provide solutions to problems of insufficient ER export of membrane proteins often encountered in biotechnology and biomedicine.

We thank Käthy Bucher for excellent technical assistance and Francisco Sanchez-Madrid for providing antibodies against CD4. The authors were supported by the Swiss National Science Foundation and the University of Basel. This work was reported in part at the 40^thAmerican Society for Cell Biology Annual Meeting and published as an abstract(Mol. Biol. Cell 2000 11, 208a).