Identification of a lumenal sequence specifying the assembly of Emp24p into p24 complexes in the yeast secretory pathway.

The p24 proteins are transmembrane proteins of the endomembrane system that play a poorly defined role in vesicle traffic between the endoplasmic reticulum and the Golgi apparatus. Various lines of evidence indicate that p24 proteins fall into four subfamilies (alpha, beta, gamma, and delta) and that tetramers are assembled containing one representative from each subfamily; however, the nature of the protein-protein interactions within these hetero-oligomers is unknown. We have identified a lumenal segment of yeast p24beta (Emp24p) that is necessary for its assembly into p24 complexes. Replacement of 52 C-terminal residues of Emp24p with the corresponding sequence from Erv25p (p24delta) generates a chimeric protein able to replace Emp24p in p24 complexes that retain partial function in vivo, ruling out a role for the transmembrane and cytosolic domains in specifying p24 interactions. Substitution of a further 50 residues, encompassing a heptad repeat region, abolishes the ability of the chimera to replace Emp24p but instead creates a protein that resembles its Erv25p parent in its requirement for stabilization by Emp24p. These data point to a role for coiled-coil interactions in directing subfamily-specific assembly of p24 oligomers that project into the lumen of transport vesicles, where they may act to exclude secretory cargo from coat protein complex type I-coated retrograde transport vesicles.

cludes two conserved cysteine residues and, in the membrane proximal part of the domain, a region that includes a set of heptad repeats suggestive of a capability to participate in coiled-coil interactions.
Some of the first studies of p24 function (2,5) were taken to point to a role in selecting cargo molecules for packaging into COPII-coated vesicles leaving the ER (6). Although recent evidence indicates that the COPII proteins can form vesicles from liposomes without any involvement of integral membrane proteins (7), the formation of vesicles in vivo appears to be linked to a cargo selection process (8 -11). In yeast cells, the finding that p24 defects cause slowed kinetics of secretion of a subset of cargo proteins (2,5) led to the suggestion that p24 proteins might act as cargo receptors, the function of which is to ensure the concentration of cargo into COPII-coated vesicles. Consistent with this idea are the findings that p24 proteins are packaged into COPII vesicles (5) and that p24 defects allow the bypass of the requirement for one component of the COPII coat, Sec13p (12), which could reflect the existence of a regulatory mechanism linking cargo recruitment with the completion of COPII coat assembly. Although associations involving COPII components, p24 molecules, and cargo have been reported (13), direct evidence for a cargo selection role involving interactions between p24 molecules and specific cargo molecules has not yet emerged.
Other evidence points to a role for the p24s in COPI-mediated retrograde vesicle traffic from the Golgi to the ER. The proteins contain COPI binding motifs in their cytosolic tails that have been shown to function in vivo and in vitro (14 -16). Studies of the localization of the proteins in mammalian cells indicate that they cycle between the ER and Golgi but are to be found predominantly in the Golgi (15,17). A role for the proteins in retrograde traffic would explain the finding that, in yeast, p24 defects cause elevated secretion of the ER protein Kar2p (12), normally retrieved from the Golgi via retrograde trafficking of the Erd2 receptor (18). However, this phenotype could also stem from a loss of cargo control at the ER: previous studies indicate that the Erd2p-mediated retrieval mechanism is easily saturated in yeast (19), so any increased flow of proteins, such as Kar2p, out of the ER would be expected to lead to a significant overspill into the secretory pathway.
On the basis of sequence comparisons, p24s can be divided into four subfamilies, p24␣, p24␤, p24␥, and p24␦ (4,15). Several lines of evidence support the view that p24s from each of these subfamilies are assembled into hetero-oligomeric complexes. The observation that two yeast p24s, Emp24p and Erv25p, are interdependent for stability (5) raised the possibility that the proteins might interact in a complex, and this work has been extended by Marzioch et al. (4), who have concluded from the results of immunoprecipitation experiments that four yeast p24s (Erp1p (p24␣), Emp24p (p24␤), Erp2p (p24␥), and Erv25p (p24␦)) function together in a heteromeric complex.
We set out to investigate the association between Emp24p and other p24s. Our data point to the involvement of a region of the lumenal domain of the proteins that includes a heptad repeat region, suggesting that coiled-coil interactions between p24 proteins are at least in part responsible for the assembly of p24 complexes.
Yeast cells were grown in YEPD medium (1% Bacto-yeast extract, 2% Bacto-peptone, and 2% glucose) or selective medium (0.67% yeast nitrogen base without amino acids, 2% glucose, and supplements as required). In all experiments involving growth of transformed strains on minimal medium, the control strain contained the vector YCpU to allow growth of all strains in the same medium.
Plasmid Construction-Except where indicated, all plasmid constructs were created using DNA fragments generated by Pfu polymerase (Stratagene) and were verified by DNA sequencing of the coding region.
DNA fragments carrying either the ERV25 or the EMP24 region of the S. cerevisiae genome were amplified from AYS927 DNA. The EMP24 fragment starts at position -246 and ends at position 282 downstream of the EMP24 stop codon. The ERV25 fragment starts at position -276 and ends at position 324 downstream of the ERV25 stop codon. In both cases the PCR primers incorporated restriction sites (for BamHI and HindIII) allowing insertion into the plasmid YCpU (same as YCplac33; Ref. 22), generating the plasmids YCpEmp24 and YCpErv25.
A mutant allele of EMP24 in which the four residues 165 ESTN 168 were altered to AAAA was generated by use of a four primer PCR-based strategy using two mutagenic primers (ATAGAAATACTGCAGCAGCT-GCAGCAGATCGTGTTAAATGG and CCATTTAACACGATCTGCTG-CAGCTGCTGCAGTATTTCTAT); the flanking primers used were the same as those used in the cloning of the EMP24 gene. PCRs were carried out with YCpEMP24 DNA as template. The resulting PCR product was ligated into YCpU as described for the EMP24 and ERV25 fragments, generating the plasmid YCpEMP24 -165/168.
To create an N-terminally His-tagged form of Emp24p, a mutant allele of EMP24 was created in which an extra seven His codons were inserted between the codons for Ala-20 and His-21 of the Emp24p precursor. His-21 is the N-terminal residue of mature Emp24p; the strategy was the same as that outlined above for the position 165/168 allele, with the primers CACCATCACCACCACCATCATCATAATGTC-CTTCTTCCAGCT and ATGATGGTGGTGGTGATGGTGGGCGG-ACGCCGAGAAGAA.
Three EMP24-ERV25 fusions were constructed to encode chimeric proteins: Emv98p, Emv146p, and Emv164p, named for the last Emp24p-specific residue present. Emv98p consists of Emp24p1-98 ϩ CF ϩ Erv25p101-211; Emv146p consists of Emp24p 1-146 ϩ Erv25p155-211; and Emv164p consists of Emp24p1-164 ϩ ESTN ϩ Erv25p177-211. To create the the gene fusion encoding Emv146, we introduced a SalI site into the EMP24 sequence in the position corresponding to the single SalI site present in ERV25, using the four primer strategy with two mutagenic primers (TTAACCAGGGAAGTCGTCGA-CGAACAAAGTTATATTG and ATAACTTTGTTCGTCGACGACTTCC-CTGGTTAA), EMP24 flanking primers, and YCpEmp24 DNA as template. The mutagenesis resulted in the change K147V. The plasmid YCpEmv146 was created by ligating a BamHI-SalI fragment from the mutated EMP24 gene (corresponding to the EMP24 promoter and the N terminus of Emp24p) into YCpErv25, which had been similarly digested to remove the corresponding portion of ERV25. The gene fusions encoding Emv98p and Emv164p were created directly by the four primer method: for each fusion, in the first stage, two PCR fragments were generated, corresponding to the two portions of the hybrid gene. One of these fragments was derived from YCpEmp24 template using the upstream flanking EMP24 primer together with either ACGATA-CTGAGCCTGATTTTCGAAACAGTATTGGAAATGTCC (for Emv98) or CCCTTCTGTTGGTAGATTCTGCAGTATTTCTATGCGTCCT (for Emv164). The other was derived from YCpErv25 template using the downstream flanking ERV25 primer together with either GGACATTT-CCAATACTGTTTCGAAAATCAGGCTCAGTATCGT (for Emv98) or AGGACGCATAGAAATACTGCAGAATCTACCAACAGAAGGG (for Emv164). The products from the first stage reactions were then purified, mixed, and used as template in a second PCR with the upstream flanking EMP24 primer and the downstream flanking ERV25 primer. The resulting PCR product was then ligated into YCpU generating plasmids YCpEmv98 and YCpEmv164.
Two ERV25-EMP24-ERV25 gene fusions were constructed to encode chimeric proteins: Rmr146p and Rmr164p. These chimeras are named for their Emv parent. Rmr146p consists of Erv25p 1-98 ϩ Emv146p 99 -203; Rmr164p consists of Erv25p 1-98 ϩ Emv164p 99 -203 The genes encoding the Rmr98p and Rmr146p chimeras were generated using the primers ACGGCATTCGACGTTTGTTTTTTGAACGAGAAC-ACCGG and CCGGTGTTCTCGTTCAAAAAACAAACGTCGAATGCC-GT. The former primer was used together with the downstream ERV25 primer and template DNA (YCpEmv146 or YCpEmv164 as appropriate), whereas the latter primer was used together with the upstream ERV25 primer and YCpErv25 template. A second PCR was then carried out with the two ERV25 primers to generate the new DNA fragment for cloning into YCpU. The two plasmids obtained were YCpRmr146 and YCpRmr164.
Fusion Protein and Antibody Production-Polyclonal antisera directed against the lumenal portions of Emp24p and Erv25p were raised in rabbits using protein A fusion proteins as antigens at the Scottish Antibody Production Unit. For Emp24p, the primers (GAATTCCATA-TGCATAATGTCCTTCTTCCAGC and CTCGAGGCTCTTCCGCATTT-AACACGATCGTTGGTCGAC) were used to amplify a 486-base pair product using AYS927 DNA as template. The PCR product was ligated directly into SmaI-digested pK19 (23) and sequenced; this revealed that part of the EcoRI site in the primer was lost during cloning. Flanking polylinker sites were used to transfer an EcoRI-SphI DNA fragment into the protein A fusion vector pAX11 (24). The resulting plasmid encodes a protein A fusion containing amino acids His-21 to Lys-172 of Emp24p (His-21 is the N-terminal residue of the mature protein; Lys-172 is likely to be the last residue of the lumenal domain).
For Erv25p, the primers GAATTCCATATGTTACATTTCGACATTG-CAGC and CTCGAGGCTCTTCCGCATCTTACCCTTCTGTTGGTAG were used to amplify a 510-base pair product using AYS927 DNA as template. The PCR product was ligated directly into SmaI-digested pK19 and sequenced; this revealed that parts of both primers were missing, but flanking polylinker sites were used to transfer an EcoRI-SphI DNA fragment into pAX11. The resulting plasmid encodes a protein A fusion containing amino acids Leu-21 to Arg-180 of Erv25p (Leu-21 is the N-terminal residue of the mature protein; Arg-180 is likely to be the last residue of the lumenal domain). In each case, an EcoRI-SphI fragment was then transferred into the plasmid pEX11 (25) for production of a corresponding ␤-galactosidase fusion protein for use in antibody affinity purification (24).
The synthesis of protein A fusion proteins was induced in mid-log bacterial cultures by the addition of 0.3 mM isopropyl ␤-D-thiogalactoside. The fusion proteins were purified by affinity chromatography using IgG-Sepharose (Amersham Pharmacia Biotech).
Measurement of Kar2p Secretion-Cultures of yeast transformants grown to early stationary phase in selective medium were used to inoculate YEPD medium to an A 600 of 0.5 and then grown for 6 h to early stationary phase (A 600 of approximately 2.5). Two methods were used to analyze proteins in the culture medium: in some experiments, proteins were precipitated from 20 ml of culture supernatant using trichloroacetic acid (final concentration, 10%) with 0.015% sodium deoxycholate as carrier, and pellets were dissolved directly in SDS sample buffer for analysis by SDS-PAGE. In other experiments, samples of culture supernatant were mixed directly with SDS sample buffer. The amount of material loaded was adjusted according to the A 600 of the culture so that p24 Interactions directly comparable amounts of secreted material were analyzed. Kar2p was detected by immunoblotting using rabbit anti-Kar2p antiserum provided by Dr Jeremy Brown, University of Edinburgh.
Sucrose Gradient Fractionation of Yeast Membranes-The procedure used was essentially that described (26). Cells were grown in selective medium to an A 600 of approximately 0.4, whereupon 400 A 600 units of cells were harvested and converted to spheroplasts. Sucrose gradients consisted of (w/w) 18,22,26,30,34,38,42,46, 50, and 54% sucrose. Following centrifugation for 2.5 h in an SW41 rotor at 38,000 rpm (174,000 ϫ g), 330-l fractions were collected and analyzed. GDPase was assayed using published methods (27). For immunoblot analysis, samples were pooled in pairs and analyzed directly by SDS-PAGE.
Purification of Protein Complexes Using Ni-NTA Agarose-Cultures were grown in selective medium to mid logarithmic phase (A 600 ϭ 0.4). Cells were converted to spheroplasts and lysed by homogenization. Following a low speed centrifugation to remove unbroken spheroplasts, membranes were recovered by centrifugation for 30 min at 158,000 ϫ g. Membranes (equivalent to 200 A 600 units of cells) were resuspended in Buffer A (150 mM sodium phosphate, pH 8.0, 2% Triton X-100) containing 10 mM imidazole and incubated on ice for 30 min. Solubilized membrane proteins were recovered in the supernatant following centrifugation at 158,000 ϫ g for 15 min. This supernatant was incubated with 20 l of Ni-NTA agarose (Qiagen) for 1 h at 4°C. Unbound material was recovered, and the agarose was then washed with 400 l of Buffer A ϩ 20 mM imidazole. Bound proteins were then eluted with Buffer A containing 250 mM imidazole. Equivalent samples of total solubilized material, unbound material, and eluate were analyzed by SDS-PAGE and immunoblotting.

RESULTS
We set out to define the region of Emp24p that is responsible for its assembly into p24 hetero-oligomeric complexes, using the stabilization of Erv25p as a convenient indicator of assembly. The approach taken was to create Emp24p-Erv25p chimeras. Although Emp24p and Erv25p are both members of the p24 family of proteins, they are only distantly related to one another: as shown in Fig. 1A, apart from two conserved cysteine residues in their N-terminal portions, the main region of close identity of these proteins lies in their C-terminal regions. We constructed three gene fusions encoding Emv proteins (Emp-Erv chimera), in which C-terminal portions of Emp24p were replaced by the corresponding sequence from Erv25p (Fig.  1B). In Emv164p, the substituted sequence of 34 residues encompasses the C-terminal transmembrane domain and cytosolic tail, together with a short conserved region from the lumenal domain. In Emv146p, a further 18 residues of the lumenal sequence of Emp24p was replaced with the corresponding Erv25p sequence. Beyond this point in the two sequences, the alignment breaks down; the position of the junction in Emv98p corresponds to the second conserved cysteine residue, lying ϳ50 residues further upstream. There was no obvious basis upon which to construct Emv chimeras intermediate to Emv146p and Emv98p.
Cells expressing one of the three Emv chimeras in place of Emp24p were analyzed for the presence both of the chimeric proteins and of Erv25p (Fig. 1C). As expected, deletion of the EMP24 gene caused a marked reduction in the level of Erv25p, and the reintroduction of EMP24 on a centromeric plasmid restored Emp24p and Erv25p to normal levels. The chimeric proteins Emv164p and Emv146p were detected at levels similar to Emp24p, and their presence was accompanied by restored levels of Erv25p. However, Emv98p was only weakly detected, and in this strain, Erv25p remained at the same low level seen in the absence of an Emp24p-related protein. Because Emv chimeras are detected by both antisera direct comparisons of the levels of the different Emv proteins are not reliable.
Because Emv146p is able to stabilize Erv25p, we conclude that Emv146p can replace Emp24p in p24 complexes, indicating that a portion of Emp24p sequence lying to the N-terminal side of Lys-147 is sufficient to specify the assembly of Emp24p into heteromeric complexes; Emp24p sequences downstream of this point may participate in the inferred interaction with other p24 proteins, but these are not essential for Emp24p-specific interactions. The failure of Emv98p to stabilize Erv25p suggests that the region of Emp24 between residues Leu-101 and Lys-147 is, however, essential for this property of Emp24p.
The absence of Emp24p results in the secretion of lumenal ER proteins, such as Kar2p (12). We tested the Emv chimeras p24 Interactions for their ability to reverse this effect (Fig. 1C): Emv164p was able to reduce secretion of Kar2p as effectively as Emp24p, Emv146p gave a partial restoration of Kar2p localization, and Emv98p had no effect. This result suggests that the p24 complex formed by Emv146p may be partially defective, even though Erv25p levels are fully restored, pointing to an important functional role for the portion of Emp24p that lies near the lumenal surface of the ER membrane. Further support for this idea comes from the analysis of an Emp24p mutant in which residues 165-168 (ESTN) have been replaced with alanines. This form of Emp24p is also fully able to restore Erv25p but does not restore normal localization of Kar2p (Fig. 1C).
Reports that p24 proteins are found in both the ER and the Golgi (15,17) are consistent with the suggestion that the proteins are involved in the formation of COPII vesicles at the ER membrane and of COPI vesicles at the Golgi membrane. To further characterize cells containing Emv146p, we compared the subcellular distribution of Erv25p in this strain with that seen in a wild-type strain in which Emp24p and Erv25p were found broadly distributed across a sucrose density gradient with peaks co-localizing with the Golgi marker enzyme GDPase and with the ER membrane protein Sec61p (Fig. 2). In cells containing Emv164p in place of Emp24p, the chimera and Erv25p showed a relative distribution similar to that seen in wild-type cells, indicating that the Erp complexes incorporating the chimera behave normally. We do not believe the apparent shift in distribution of Sec61p in the experiment shown to be significant because measurements of sucrose concentration in the gradient fractions indicate that the peaks of Sec61p occurred at similar densities on the two gradients.
If residues Leu-101 to Lys-147 of Emp24p are responsible for the assembly of the protein into a hetero-oligomeric complex with other p24 proteins, then perhaps Emv98p, containing the corresponding region from Erv25p, might show Erv25p-like behavior and be stabilized by Emp24p. To test this idea, we compared the levels of Emv98p in cells lacking either Emp24p or Erv25p. As shown in Fig. 3A, the instability of Erv25p in a strain lacking Emp24p is accentuated in stationary phase cells, with Erv25p being undetectable in this strain but fully stabilized by the presence of Emv146p. To assess the influence of Emp24p upon the stability of Emv98p, we therefore compared levels of the latter protein in cells containing either Emp24p or Erv25p, in both growing and stationary cultures, with Emv146p as a control (Fig. 3B). In the presence of Erv25p, Emv146p persisted into stationary phase (Fig. 3B, left lane). In contrast, in this same background, Emv98p disappeared from stationary phase cells along with Erv25p, consistent with the conclusion that Emv98p cannot replace Emp24p in p24 complexes and thus cannot stabilize Erv25p. In cells containing Emp24p but lacking Erv25p, Emv146p persisted into stationary phase in the same way that Emp24p itself does (Fig. 3B, second lane from right; also compare the right and second from right lanes of Fig. 3A), and, in marked contrast to the result seen in cells lacking Emp24p, so does Emv98p (Fig. 3B, right  lane). Thus, Emv98p is stabilized in the presence of Emp24p, just like its Erv25p parent, suggesting that this chimera can mimic Erv25p by assembling into p24 complexes: in the transition from Emv146p to Emv98p, the replacement of ϳ50 residues downstream of the second conserved cysteine converts the protein from one that is able to act like Emp24p to one that, like Erv25p, depends upon Emp24p for stability.
We also tested the possibility that the region of Emp24p defined by this analysis as necessary for the stabilization of Erv25p could, if transplanted into Erv25p, convert that protein to one that was able to replace Emp24p in p24 complexes. Gene fusions were constructed to encode chimeras Rmr146p and Rmr164p in which an internal segment of Erv25p was replaced with the corresponding portion of Emp24p. The ability of these proteins to stabilize Erv25p in cells lacking Emp24p was then assessed (Fig. 4). In growing cells, both proteins had a small but significant stabilizing effect (Fig. 4A), and in stationary phase cells, a trace of Erv25p was still detected under condi-FIG. 2. Emv146p restores normal subcellular distribution of Erv25p. A, subcellular fractionation of wild-type (YLC200(YCpU)) cells. B, subcellular fractionation of YLC233(YCpEmv146) cells. Cultures of the two transformant strains were grown in selective medium to midlog phase. Cells were converted to spheroplasts and lysed by homogenization, the homogenate was analyzed on a 10-ml sucrose gradient (in steps from 18 to 54% (w/w)), and 660-l fractions were collected. SDS-PAGE and immunoblotting were used to locate Sec61p (an ER membrane protein), Erv25p, Emp24p, and related chimeras. An assay for GDPase was used to locate fractions enriched for Golgi membranes.

FIG. 3. Emv98p is stabilized by Emp24p.
Cultures of YLC233 (⌬emp24) and YLC244 (⌬erv25) harboring various plasmids were grown overnight in selective medium into stationary phase. A, the instability of Erv25p in the absence of Emp24p is accentuated in stationary phase cells. Protein extracts from stationary phase cells were analyzed by SDS-PAGE, blotted, and probed with a mixture of antibodies for Emp24p and Erv25p. B, the presence of Emp24p stabilizes Emv98p. Protein extracts from both midlog and stationary phase cells were analyzed. The blot was probed with antibodies for Erv25p, which also recognize Emv chimeras. wt, wild-type.
p24 Interactions tions in which it is normally undetectable. Thus, the Rmr chimeras have a weak ability to stabilize Erv25p, although it is clear that in the context of the Erv25p sequence, this region cannot confer full stabilizing activity.
The interpretation of the data presented so far has relied upon inferences about p24 complex formation drawn from the stabilization of Erv25p. We wished to obtain direct biochemical evidence that Emv146p is present in stable protein complexes that contain Erv25p. Others have detected the existence of p24 protein complexes from the behavior of the proteins in rate zonal centrifugation of cholate-solubilized material, and directly by co-immunoprecipitation. In our experiments, we could not use immunoprecipitation because our polyclonal antisera for Emp24p and Erv25p were both able to recognize the Emv chimeras. We therefore turned to a tagging approach, exploiting the fact that the proteolytic cleavage that removes the Emp24p signal sequence occurs between residues Ala-20 and His-21. We introduced seven extra histidine codons into the EMP24 sequence so that signal cleavage should generate a form of Emp24p carrying eight N-terminal histidine residues. As shown in Fig. 5A, cells containing this modified gene contained an Emp24p immunoreactive protein with reduced electrophoretic mobility, as expected if the protein carries the Nterminal extension. Furthermore this His-tagged Emp24p protein restored normal Erv25p levels (Fig. 5A) and also fully complemented the Kar2p localization defect in a strain lacking Emp24p (not shown). We were able to purify the His-tagged Emp24p from these cells following a protocol employing complete denaturation (6 M urea and 1% Triton X-100), whereupon the material recovered on Ni-NTA-agarose contained no detectable Erv25p, as expected (data not shown). In order to recover protein complexes containing His-tagged Emp24p, we used Ni-NTA chromatography to analyze a Triton X-100 solubilized extract prepared from yeast membranes. As shown in Fig. 5B, when the detergent extract from cells containing His-tagged Emp24p was treated with Ni-NTA agarose, there was significant depletion of the protein from the extract, and the bound protein survived extensive washing with buffer containing 20 mM imidazole but was eluted from the column by exposure to 250 mM imidazole. Furthermore, immunoblot analysis revealed that Erv25p was also depleted from the detergent extract by Ni-NTA agarose and was present in the eluted fraction. Two controls illustrate the specificity of this result. First, a control membrane protein, the ER protein Sec61p, was not present in the eluted material. Second, when the detergent extract was prepared from wild-type cells, neither Emp24p nor Erv25p was recovered in the affinity-purified material, and indeed neither protein was detectably depleted by the initial binding reaction; thus, the binding of both proteins to Ni-NTA agarose depends upon the presence of the N-terminal His tag on Emp24p. We conclude that p24 complexes containing His-tagged Emp24p and Erv25p can be recovered by nickel affinity chromatography. We repeated this experiment with extracts prepared from cells containing an analogous His-tagged form of Emv146p and found that this also led to the recovery of Erv25p. This result provides clear confirmation that the Emv146p chimera can replace Emp24p in p24 complexes. DISCUSSION We have presented evidence that a region of the lumenal domain of Emp24p consisting of some 50 amino acids is necessary to direct its assembly into hetero-oligomeric complexes with other p24 family members. We investigated which part of Emp24p is responsible for its assembly into p24 complexes by FIG. 4. Replacement of internal sequences of Erv25p with the corresponding sequence from Emp24p generates chimeric proteins able to provide weak Erv25p stabilizing activity. Cultures of the transformed strains indicated were grown in selective medium; samples were taken in midlog phase and after overnight growth (stationary phase). Cell extracts were prepared and analyzed as described in the legend to Fig. 1. Blots were probed with a mixture of anti-Emp24p and anti-Erv25p antibodies. The portion of the blot showing some higher molecular weight nonspecific bands has been shown to emphasize the control of sample loading. wt, wild-type.
FIG. 5. Biochemical evidence for assembly of Emv146p into p24 complexes. A, a His-tagged form of Emp24p restores Erv25p levels. YCpEmp24-NHis encodes a version of Emp24p with an extra seven histidine residues at its N terminus. Cultures of the transformed strains indicated were grown overnight in selective medium into stationary phase. Cell extracts were prepared and analyzed as described in the legend to Fig. 1, with separate gels blotted and probed for Erv25pand Emp24p-immunoreactive proteins. B, affinity purification of protein complexes containing His-tagged forms of Emp24p and Emv146p. Emv146-NHis is a form of Emv146p carrying the same histidine tag as described above. Solubilized membrane proteins were incubated with Ni-NTA agarose, unbound material was recovered, and the agarose was then washed with buffer containing 20 mM imidazole. Bound proteins were then eluted with buffer containing 250 mM imidazole. Equivalent samples of total solubilized material (t), unbound material (u), and eluate (e) were analyzed by SDS-PAGE and immunoblotting as indicated. In the samples derived from the strain containing Emv146-NHis, the antibodies for Erv25p detect both Erv25p and the chimeric protein.
p24 Interactions exploiting the observation that the absence of Emp24p leads to the degradation of Erv25p, presumably because it fails to be stabilized by being assembled into a complex. Our initial hypothesis was that the assembly of p24 heteromers might be driven by interactions between the transmembrane domains of the proteins that contain polar residues suggestive of interactions within the plane of the lipid bilayer. We tested this idea by investigating the properties of chimeric p24 proteins.
In the chimeras Emv164p and Emv146p, the C-terminal portions of Emp24p encompassing the transmembrane domain, comprising 34 and 52 residues, respectively, were replaced with the corresponding sequences from Erv25p. Both of these chimeras were able to function as replacements for Emp24p: in cells containing either chimera, we found that Erv25p levels were restored to normal, suggesting that the chimeric proteins were able to enter p24 complexes in place of Emp24p. We used a His-tagged form of Emp24p to isolate complexes containing Erv25p, and because Erv25p was also isolated in association with a His-tagged form of Emv146p, we conclude that sequences downstream of Emp24p Lys-147 are not essential for its assembly into p24 complexes.
Replacement of a further 47 residues of Emp24p resulted in the chimera, Emv98p, which was unable to replace Emp24p, indicating that sequences necessary for the protein to assemble as Emp24p into protein complexes have been lost. This suggests that a region of Emp24p between Lys-147 and the second conserved cysteine residue (Cys-99) specifies the assembly property of Emp24p. A corollary of this conclusion would be that Emv98p, containing the corresponding region of Erv25p, might be assembled into p24 complexes in place of Erv25p. Evidence in support of this came from the finding that Emv98p was more stable in cells containing Emp24p, just like its Erv25p parent. We conclude from these data that the region of p24 proteins corresponding to the Emp24p region Lys-147 to Cys-99 specifies their assembly into p24 complexes.
The finding that the cytosolic domain and transmembrane domain of Emp24p can be replaced with the corresponding region from Erv25p also indicates that the unusual C-terminal sequence of Emp24p is not essential for function. The cytosolic portion of Emp24p is unique among p24 family members in that it lacks a C-terminal KKXX motif, implicated in COPI binding, although it does contain the FF motif that is also thought to be important for COPI association (14,15). The absence of the C-terminal motif from Emp24p is apparently of no functional significance because the p24 complexes that are formed in the presence of Emv146p, which carries the cytosolic domain of Erv25p, were found to be distributed normally be-tween the ER and Golgi and provided normal function as assessed by restoration of Kar2p localization. Studies of the interactions between p24 C-terminal peptides and ␥-COP have revealed a range of affinities, pointing to possible differences in the tendency of the various p24 proteins to partition between the ER and Golgi and to a possible dynamic equilibrium between unassembled p24 molecules and heteromeric oligomers (20). However, the apparent lack of functional significance of the unusual C terminus of Emp24p suggests rather that the interactions of p24s with COPI occur predominantly when the proteins are assembled, explaining how the loss of a COPI signal from a single component can be accommodated.
Our data implicate a relatively short region of p24 proteins as being responsible for defining the assembly properties of the protein. There are two features of this region that could play a role in determining the specificity and stability of complex formation: the presence of a heptad repeat sequence, and the presence of sequence motifs that are specific for p24 subfamilies.
As pointed out by others (4,15), p24 proteins contain a region with a heptad repeat pattern, consistent with the possibility that coiled-coil interactions are involved in oligomerization. As shown in Fig. 6, this region of Emp24p and Erv25p spans, but lies mainly to the N-terminal side of, the junction point in Emv146p, a position consistent with its involvement in subfamily-specific interactions. Are there sequences in this region that are specific to a particular p24 subfamily, over and above any heptad repeat sequence? To address this question, we examined patterns of subfamily-specific sequence conservation. For Emp24p, this analysis was not particularly fruitful: the sequence of Emp24p is only distantly related to other members of the p24␤ subfamily, and the region in question is almost devoid of conserved residues, although the conservation of amino acids corresponding to Emp24p positions Leu-132, Leu-139, and Val-147 (the spacing of which corresponds to the heptad repeat) is noteworthy. For the p24␦ subfamily (which includes Erv25p), the most obvious feature in the region is a conserved sequence, LKXXEVELRR, which lies within the heptad repeat. This ␦-specific sequence may be important for the assembly of p24␦ into hetero-oligomeric p24 complexes.
Marzioch et al. (4) have suggested that Emp24p, Erv25p, Erp1p, and Erp2p form a tetramer, and one possibility is that four p24 molecules, one from each subfamily, assemble into a tetramer that is formed, at least in part, through a fourstranded parallel helical bundle. However, the pattern of interdependence seen in yeast raises the possibility that the postulated p24 tetramer is a dimer of dimers. Thus, the loss of FIG. 6. Alignment of p24␤ and p24␦ sequences showing heptad repeats and conserved sequence motifs in p24 subfamilies. The sequences of six p24 proteins are shown: C. elegans p24␤, human p24␤, Emp24p, Erv25p, human p24␦, and C. elegans p24␦ (GenBank TM protein sequence identification numbers are 2275635, 1212965, 1322830, 854478, 1359886, and 3877300, respectively). All sequences have been truncated to run from the second conserved cysteine residue. The gaps in the p24␤ sequences were introduced manually to facilitate annotation of the short conserved sequence KXVXF. The transmembrane domains are underlined in the yeast sequences. The boxed regions correspond to the heptad repeat region; these are only weakly predicted (e.g. by the COILS algorithm (31) as implemented in the program MacStripe version 2.0), and the positions of the boxes are for illustrative purposes only. Residues conserved in all three p24␤ sequences are marked above the alignment; residues conserved in all three p24␦ sequences are marked below the alignment. p24 Interactions either Erp1p or Erp2p causes only a modest reduction in the levels of Emp24p and Erv25p, and in the absence of all three yeast members of the p24␣ subfamily (Erp1p, Erp5p, and Erp6p), the Emp24p and Erv25p molecules that remain behave as a dimer in gel filtration. Perhaps coiled-coil interactions are responsible for p24␤-p24␦ associations and for p24␣-p24␥ associations, with the formation of the tetrameric complex resulting from a different type of interaction between the dimers.
The picture of p24 oligomer structure that begins to emerge is one of a quasi-structural role for these proteins in vesicle trafficking. Recent findings indicate that secretory cargo is selected for anterograde transport in vesicular-tubular clusters lying between the ER and Golgi, where COPI-coated buds form in a manner that excludes secretory cargo (28,29). The major transmembrane components of these COPI-coated buds are likely to be the p24 proteins (17). Interactions between the polar faces of p24 transmembrane helical domains are consistent with the existence in COPI-coated vesicles of an inner lining of p24 molecules projecting well into the vesicle lumen, as is the existence of a lumenal rod-like structure formed as a result of the subfamily-specific interactions that we have detected in our experiments. This raises the possibility that the p24 proteins may act to prevent secretory cargo returning in COPI vesicles by masking the inner charged surface of the vesicles and by passively filling much of the space within them. In this way, only membrane proteins with retrieval motifs (e.g. Erd2p) and their ligands would be able to make the return trip to the ER: there would be no room for anything else. Further studies of structure-function relationships in p24 molecules will help to define their role in vesicle traffic.