The Emp24 Complex Recruits a Specific Cargo Molecule into Endoplasmic Reticulum–Derived Vesicles

In eukaryotic cells, protein transport between the organelles of the secretory pathway is mediated by vesicles that bud from a donor compartment and fuse with an appropriate acceptor compartment (Palade 1975). The starting point of the exocytic route is the ER. There, correctly folded and assembled cargo molecules can enter COPII-coated vesicles for transport to the cis-Golgi compartment (Schekman and Orci 1996). Many cytosolic and transmembrane proteins have been identified that mediate this vesicular transport step (Rothman and Wieland 1996; Schekman and Orci 1996).

p24 proteins are present in heteromeric complexes that have been proposed to cycle between ER and Golgi compartments because they are found in both COPII and COPI vesicles in addition to ER and Golgi membranes (Schimmöller et al. 1995; Stamnes et al. 1995; Belden and Barlowe 1996; Sohn et al. 1996; Rojo et al. 1997; Dominguez et al. 1998; Füllekrug et al. 1999; Marzioch et al. 1999). In mammals, functional data suggest a role for these proteins in the formation of ER exit sites (Lavoie et al. 1999). In yeast, eight genes encode p24 family members: EMP24, ERV25, and ERP1–ERP6 (Marzioch et al. 1999). Mutants in several of these genes show selective protein transport defects. Strains with a single deletion of EMP24, ERV25, ERP1, or ERP2 show a delay in the ER to Golgi transport of Gas1p (Schimmöller et al. 1995; Belden and Barlowe 1996; Marzioch et al. 1999), a glycosylphosphatidylinositol (GPI)-anchored protein. The emp24 and erv25 mutants also show a transport defect for invertase, a soluble secreted protein. Several other cargo molecules, including α factor, a secreted pheromone, do not require p24 proteins for efficient transport. Mutation of several yeast p24 genes results in secretion of the ER lumenal protein, Kar2p (Elrod-Erickson and Kaiser 1996; Marzioch et al. 1999).

The Emp24 complex in yeast comprises Emp24p, Erv25p, and most likely Erp1p and Erp2p. Deletion of EMP24 causes a strong reduction in the levels of the other three proteins of the complex (Belden and Barlowe 1996; Marzioch et al. 1999). In this study we provide evidence that the Emp24 complex is directly required for efficient packaging of Gas1p into ER-derived vesicles. Two subunits of this complex can be directly cross-linked to the cargo protein in purified ER-derived vesicles, consistent with the hypothesis that the Emp24 complex plays a role as a cargo receptor in ER to Golgi transport.

An myc epitope was introduced at the NH₂ terminus of mature Emp24p. The myc-tagged emp24-E178A mutant was constructed by substituting an appropriate fragment of EMP24 with the sequenced mutant version obtained by PCR techniques. RH4443 (MATα, ura3, leu2, his4, bar1, emp24:: KanMx) was obtained by replacing the entire EMP24 coding sequence of RH1959 (MATα, ura3, leu2, his4, bar1) with a KanMx cassette. EMP24 alleles were cloned into a YCplac111 (CEN/ARS) plasmid. The S. cerevisiae strain RH696-2B (MATα, sec18-20 gap1Δ::LEU2 ura3 ade2 leu2 lys2Δ201 pPL269) was obtained by crossing PLY129 (pPL269) (MATα gap1Δ::LEU2 ura3 ade2 leu2 lys2Δ201 pPL269) (Kuehn et al. 1996), with RH478 (MATα sec18-20 leu2 his4). Cytosols were prepared from RH732 (MATα his4 leu2 ura3 lys2 pep4::URA3 bar1) and RH2043 (MATα sec18-20 his4 leu2 ura3 pep4::URA3 bar1).

Protein levels were determined by extraction of log phase cultures and Western blotting (Sütterlin et al. 1997) using antibodies raised against the Emp24p cytosolic tail or the myc epitope. Analysis of Gas1p transport in vivo was by a pulse-chase protocol (Sütterlin et al. 1997). Gas1p maturation was analyzed using a 4-min pulse and subsequent chase at 30°C, followed by immunoprecipitation, SDS-PAGE, and fluorography. Immature (105 kD) and mature (125 kD) forms of Gas1p were quantified using a densitometer. The percentage of mature Gas1p was used as an indication of transport to the Golgi apparatus. Determination of partitioning of Gas1p between aqueous and detergent phases of Triton X-114 was performed as described (Nuoffer et al. 1993), except that the media fractions were not analyzed. After separation into detergent and aqueous phases, Gas1p was denatured, immunoprecipitated, resolved by SDS-PAGE, visualized, and quantified using a PhosphorImager.

The permeabilized, cell-based, ER cargo packaging assay was performed as described (Kuehn et al. 1996) with a few modifications. Strains RH1959, RH4438, and RH696-2B were transformed with pCNYG1 to overexpress Gas1p (approximately fivefold) (Nuoffer et al. 1991). Overexpression did not affect Gas1p transport kinetics or dependence on EMP24. 10 mM 2-mercaptoethanesulfonic acid replaced 10 mM dithiothreitol. Budding reactions contained 30 μl of membranes (from 12 × 10⁷ cells), 300 μg crude cytosol, 3 μg Sar1p, 1× ATP mix, and 0.2 mM GTP in a 150-μl volume. After 2 h at 25°C, a portion of the sample was removed for analysis (total), the remaining aliquot was sedimented (14,000 rpm, 2 min, 4°C), and 125 μl of the supernatant collected and subjected to flotation on a Nycodenz^® step gradient (Barlowe et al. 1994). 100 μl from the top was discarded and the next 800 μl transferred to a new tube. 400 μl was diluted threefold with B88 and centrifuged (100,000 g, 1 h, 4°C). The membrane pellet was dissolved in 1% SDS in TEPI buffer (Sütterlin et al. 1997) for 10 min at 55°C, subjected to immunoprecipitation, and analyzed by SDS-PAGE with subsequent exposure and quantitation using a PhosphorImager.

Vesicles were produced in a budding reaction using sec18 membranes and cytosol. The sec18 membranes were prepared as above, except that the last 5 min of depletion and the pulse-labeling were at 32°C. The sec18 cytosol was preincubated (32°C, 10 min) before use. Vesicles were immunoisolated with or without Emp24p anti-tail antibody and processed according to Kuehn et al. 1996.

Nycodenz-purified vesicles produced in a budding reaction were adjusted to 2.5 M urea in B88 and incubated with 1 mM dithiobis(succinimidylpropionate) (DSP) or various amounts of disuccinimidyl glutarate (DSG; Pierce) (20°C, 20 min). The cross-linking reaction was quenched by addition of glycine (50 mM final, 5 min, 20°C). Vesicles were sedimented at 100,000 g (1 h, 4°C), dissolved with 1% SDS in TEPI (5 min, 95°C for Gas1p and glycosylated pro α factor [gpαF], or 55°C for Gap1p), and immunoprecipitated with Emp24p anti-tail antibody or Erv25 antibody (Belden and Barlowe 1996) and protein A–Sepharose. Precipitated material was eluted from the Sepharose beads by incubation with 1% SDS in TEPI (5 min, 95 or 55°C) and reimmunoprecipitated with anti-Gas1p or anti-Gap1p antibody.

To investigate whether Emp24p plays a role in cargo exit from the ER we quantified the packaging of different secretory proteins into vesicles that were generated from wild-type and emp24 mutant ER membranes in vitro. Budding of Gas1p was much less efficient (>70% less) from emp24Δ membranes than from wild-type membranes, whereas the budding efficiencies of gpαF and an amino acid permease (Gap1p) were similar from the two membrane sources (Fig. 1). gpαF is an established control for selectivity, because its transport from the ER to the Golgi is not affected in emp24Δ cells (Schimmöller et al. 1995). Gap1p was also investigated because it has been shown to be efficiently incorporated into COPII-coated vesicles in vitro (Kuehn et al. 1996). Therefore, Emp24p is required for efficient packaging of at least one secretory protein into COPII-coated vesicles. The magnitude of the defect in Gas1p budding from the ER is consistent with the ER to Golgi transport defect seen in vivo (Schimmöller et al. 1995) and could, therefore, be the entire explanation for the transport delay.

It has been proposed that Emp24p may be a negative regulator of vesicle formation (Elrod-Erickson and Kaiser 1996) and that its absence would result in increased Kar2p exit from the ER. To address this point, we compared the incorporation of Kar2p into vesicles using wild-type and emp24Δ membranes. Kar2p was poorly packaged into ER-derived vesicles using either membrane source (Fig. 1). Therefore, it does not seem that Emp24p is required for excluding Kar2p from ER-derived vesicles. This suggests, but does not prove, that the reason for Kar2p secretion in the emp24 mutant is a defect in retrograde transport.

Since Emp24p could play a role in retrograde transport it is possible that the inefficient budding of Gas1p from emp24Δ membranes is an indirect consequence of a retrograde transport defect. To address this point, we used antibodies against the cytosolic tail of Emp24p to inhibit the function of Emp24p in wild-type membranes. Preincubation of membranes with antibodies on ice reduced the budding efficiency of Gas1p greater than threefold (Fig. 2 A). This inhibition was specific for Gas1p, because the packaging of Gap1p and gpαF was not affected. As a control for the specificity of our anti-tail antibodies, we used an emp24 mutant with an E178A substitution in the cytosolic tail. This point mutation does not affect the function of Emp24p (Fig. 2 B), but the anti-tail antibodies no longer recognize the mutant protein (Fig. 2 B). Preincubation of membranes derived from emp24-E178A cells with anti-tail antibodies had no effect on budding efficiencies of Gas1p, gpαF, or Gap1p (Fig. 2 A). Therefore, we conclude that Emp24p is required directly for the efficient incorporation of Gas1p into ER-derived vesicles.

One requirement for Gas1p exit from the ER is the attachment of the GPI anchor (Doering and Schekman 1996). To test whether the GPI anchor is efficiently attached to Gas1p in the emp24Δ mutant, we performed a pulse-chase analysis followed by separation of the protein extracts into detergent and aqueous phases using Triton X-114. GPI-anchored Gas1p partitions efficiently into the detergent phase, whereas the unanchored protein partitions approximately equally between detergent and aqueous phases (Nuoffer et al. 1993). After the 4-min pulse-labeling, ∼90% of the Gas1p was already found in the detergent phase in both wild-type and emp24Δ mutant strains (Fig. 2 C), demonstrating that the GPI anchor is efficiently attached to Gas1p in the emp24Δ mutant. Therefore, a defect in GPI anchoring cannot explain the transport defect of Gas1p in the emp24Δ mutant.

Since Emp24p is efficiently incorporated into ER-derived vesicles (Schimmöller et al. 1995) and is directly required for Gas1p recruitment, we investigated if Gas1p could be found in a protein complex with Emp24p in ER-derived vesicles. ER-derived vesicles were generated in vitro and uncoated vesicles were exposed to the cleavable membrane-permeable cross-linker DSP. After lysis and denaturation, the samples were immunoprecipitated using anti-Emp24p tail antibodies. The precipitates were denatured and subjected to a second immunoprecipitation using antibodies against Gas1p. The cross-linker was cleaved and the samples were analyzed by SDS-PAGE. Gas1p was cross-linked to Emp24p with good efficiency (Fig. 3). No Gas1p was recovered if cross-linker was omitted. Moreover, Gas1p could not be detected when the cross-linking was performed on vesicles generated from emp24-E178A membranes (Fig. 3), proving that Gas1p was recovered as part of a complex containing Emp24p. This complex is specific because another cargo molecule, Gap1p, whose budding was not reduced from emp24 mutant membranes, was not cross-linked to Emp24p. These data clearly demonstrate that Emp24p and Gas1p can be found in the same protein complex in ER-derived vesicles and suggest that this complex is specific.

An alternative explanation for the absences of a role of Emp24p in Gap1p budding and of Gap1p from the cross-linked Emp24 complex would be if Emp24p and Gap1p were found in different ER-derived vesicles. To test this we produced vesicles using membranes and cytosol from a sec18 mutant. SEC18 encodes yeast N-ethylmaleimide–sensitive fusion protein, and its inactivation under the conditions of this experiment blocks fusion of ER-derived vesicles with the Golgi apparatus (Muñiz, M., and H. Riezman, manuscript in preparation). These conditions ensure that we are analyzing primary ER-derived vesicles. The vesicles were immunoisolated using antibody against the cytosolic tail of Emp24p. 83% of the Gas1p and 96% of the Gap1p copurified with the immunoisolated vesicles (Fig. 4). Virtually no cargo proteins were isolated when Emp24p tail antibody was omitted. Therefore, Emp24p is found in the same vesicles as Gap1p and Gas1p. The lack of effect of the emp24 mutation on Gap1p budding and Gap1p absence from the cross-linked Emp24 complex cannot be due their presence in distinct vesicles.

The cross-linking results presented above suggest that Emp24p may be part of a cargo receptor that improves the efficiency of cargo entry into ER-derived vesicles. However, they do not show that Emp24p binds the cargo molecule directly. To address this issue we used a noncleavable cross-linker, DSG, and performed similar cross-linking studies on ER-derived vesicles. Emp24p was cross-linked to Gas1p as seen by a smear starting at around 25,000 higher apparent molecular mass than the ER form of Gas1p (Fig. 5 A). Additional bands were seen at higher molecular weights. The band at 130 kD must represent a directly cross-linked product between Emp24p (24 kD) and Gas1p (105 kD). All of the precipitated material was specific because it was not seen when the vesicles were generated from emp24-E178A mutant membranes.

Next, we immunoprecipitated the cross-linked material with antibodies against Erv25p, followed by antibodies against Gas1p. A strong band appeared at the position expected for a directly cross-linked product between the two molecules. Therefore, we conclude that Erv25p can also be directly and efficiently cross-linked to Gas1p in isolated ER-derived vesicles. We could also detect this cross-linked product in vesicles derived from emp24-E178A mutant (data not shown), confirming that vesicle budding occurred using these membranes.

Finally, we performed a titration of DSG on the isolated vesicles. When cross-linker was omitted, no Erv25p-Gas1p product could be detected (Fig. 5 B). At low concentrations of cross-linker the directly cross-linked product was by far the main product seen. At higher concentrations of cross-linker more cross-linked products were seen and the bands at higher molecular weight became more apparent. The cross-linking studies shown above demonstrate that Gas1p can be specifically and directly cross-linked to the Emp24 complex. These results are consistent with the hypothesis that the Emp24 complex is a cargo receptor required for the efficient incorporation of Gas1p into ER-derived vesicles.

A major finding of this study is that Emp24p is directly required for efficient incorporation of Gas1p into ER-derived vesicles in vitro. Two independent assays were used to show this. One used mutant membranes devoid of Emp24p. The other used antibodies to inhibit Emp24p function directly in wild-type membranes. Both assays showed that Gas1p budding depended strongly on Emp24p, whereas budding of other proteins did not. This striking specificity suggests that this complex has a highly specific role in packaging of Gas1p into ER-derived vesicles.

This direct role of Emp24p in selective budding from the ER can be explained in three ways. First, Emp24p could be required for efficient anchor attachment to Gas1p. We have shown that this is not the case. Second, Emp24p could be required for the formation of ER-derived vesicles. It has been suggested recently that α₂p24, a mammalian member of the p24 family, plays a role in the formation of ER cargo exit sites (Lavoie et al. 1999). However, in yeast, the absence of Emp24p or the inhibition of its function by the tail antibody does not lead to a general defect in vesicle formation because budding of other cargo proteins was not affected. Also, mutant membranes efficiently package the vSNARE Sec22p (Belden and Barlowe 1996). Moreover, since Emp24p is found in vesicles with Gas1p as well as with Gap1p, it is highly unlikely that Emp24p is required only for formation of vesicles containing Gas1p.

The third explanation for the direct role of Emp24p in selective budding from the ER is to consider Emp24p as part of a cargo receptor that is required for the packaging of a very small subset of cargo proteins. This cargo receptor would recruit specific proteins into the ER-derived vesicles, increasing the efficiency of packaging. A second major finding of our study, that Emp24p and Erv25p can be directly cross-linked to Gas1p, supports this hypothesis. This model raises two issues. Why does mutation of EMP24 or other genes encoding members of the p24 family affect the transport of so few proteins? And why is transport of these proteins not completely defective in emp24 mutant cells?

One possibility to explain the small number of proteins affected is that there could be a large and heterogeneous family of cargo receptors with overlapping function. Evidence exists for another type of cargo receptor acting in ER to Golgi transport, namely ERGIC 53, a lectin that also apparently facilitates transport of some glycoproteins (Appenzeller et al. 1999). It is interesting to note that mutation of either receptor only affects a small number of cargo molecules in a highly specific manner. There is a large number of proteins that cycle between ER and Golgi compartments. Many of these could have cargo receptor functions in direct or indirect ways.

Another way to explain the small number of proteins affected in a cargo receptor mutant would be that only a small subset of proteins requires a receptor at this step. It seems clear that some very abundant proteins in specialized secretory cells are not concentrated in COPII-coated vesicles (Martínez-Menárguez et al. 1999). Therefore, not all secretory proteins require a receptor for efficient ER exit. So why would some proteins require receptors for this transport step? Some proteins may bind with low affinity to other proteins that are retained in the ER lumen. These cargo molecules would require a higher affinity interaction to liberate them from the ER. Alternatively, some proteins may require high rates of secretion, for example, a protein with an activity that would be detrimental in the ER. Gas1p probably has a function in cross-linking β 1,3 glucans with β 1,6 glucans and chitin in the cell wall (Popolo and Vai 1999). Such an activity may have negative consequences in the ER.

Previously, a bulk flow mechanism (Wieland et al. 1987) where proteins leave the ER simply by being available for entry into the forming vesicle, similar to the process of fluid-phase pinocytosis, has been proposed. This could explain why mutation of genes encoding members of the p24 family affects transport of some cargo proteins strongly, but not completely. Without a receptor, the cargo protein would still be able to enter ER-derived vesicles by bulk flow. When the function of other cargo receptors, such as the mannose-6-phosphate receptor (Kornfeld 1992), or yeast Vps10p (Marcusson et al. 1994) is blocked, the lysosomal or vacuolar proteins that are affected still exit the Golgi. However, they are secreted rather than targeted to the lysosome or vacuole. Since, as far as we know, there is only one destination for ER-derived vesicles, namely the Golgi apparatus, protein missorting would not occur in the absence of receptors in the ER. This may also help to explain why only a small subset of proteins actually needs a receptor. As long as they can exit the ER they will go to the Golgi compartment.

A role in specific cargo recruitment is not likely to be the only role of the Emp24 complex. Secretion of Kar2p caused by mutations in subunits of this complex implies a role in the retrograde transport from the Golgi compartment to the ER (Elrod-Erickson and Kaiser 1996; Marzioch et al. 1999). Consistent with this, p24 tails from animal cells can bind COPI proteins (Fiedler et al. 1996; Sohn et al. 1996; Dominguez et al. 1998) and when displayed on liposomes they can stimulate the formation of COPI-coated vesicles (Bremser et al. 1999). Quality control in the ER involves recycling from the Golgi compartment to the ER of misfolded proteins (Hammond and Helenius 1994). Interestingly, a mutation in a p24 protein from C. elegans, SEL-9, allows some misfolded proteins to escape the quality control mechanisms in the early secretory pathway (Wen and Greenwald 1999) which could perhaps be explained by a retrograde defect. Studies that can directly address the retrograde or other putative functions of p24 proteins in vesicular traffic would be desirable.

We thank C. Barlowe, M.J. Kuehn, and R. Schekman for materials and advice; B. André for antibody; Riezman laboratory members and M.I. Geli for critical reading of the manuscript; and T. Aust and T. Eberle for technical assistance.

This work was supported by Swiss National Foundation grants (to H. Riezman and H.-P. Hauri), Federation of European Biochemical Societies and Human Frontiers Science Program Organization fellowships (to M. Muñiz).