Endoplasmic reticulum membrane localization of Rce1p and Ste24p, yeast proteases involved in carboxyl-terminal CAAX protein processing and amino-terminal a-factor cleavage

Abstract

Proteins terminating in the CAAX motif, for example Ras and the yeast a-factor mating pheromone, are prenylated, trimmed of their last three amino acids, and carboxyl-methylated. The enzymes that mediate these activities, collectively referred to as CAAX processing components, have been identified genetically in Saccharomyces cerevisiae. Whereas the Ram1p/Ram2p prenyltransferase is a cytosolic soluble enzyme, sequence analysis predicts that the other CAAX processing components, the Rce1p and Ste24p proteases and the Ste14p methyltransferase, contain multiple membrane spans. To determine the intracellular site(s) at which CAAX processing occurs, we have examined the localization of the CAAX proteases Rce1p and Ste24p by subcellular fractionation and indirect immunofluorescence. We find that both of these proteases are associated with the endoplasmic reticulum (ER) membrane. In addition to having a role in CAAX processing, the Ste24p protease catalyzes the first of two cleavage steps that remove the amino-terminal extension from the a-factor precursor, suggesting that the first amino-terminal processing step of a-factor maturation also occurs at the ER membrane. The ER localization of Ste24p is consistent with the presence of a carboxyl-terminal dilysine ER retrieval motif, although we find that mutation of this motif does not result in mislocalization of Ste24p. Because the ER is not the ultimate destination for a-factor or most CAAX proteins, our results imply that a mechanism must exist for the intracellular routing of CAAX proteins from the ER membrane to other cellular sites.

Many proteins are synthesized initially as precursors that undergo conversion to their mature form by post-translational processing activities and/or covalent modifications. Although protein maturation is exemplified best by the processing of secretory prohormones that are translocated into and transported through the luminal compartments of the vesicular secretory pathway [e.g., endoplasmic reticulum (ER), Golgi and trans-Golgi network], there are notable examples of protein maturation that occur in the cytosol or on the cytosolic face of membranes. Examples include the removal of initiator methionines by methionyl aminopeptidase, the maturation of the interleukin (IL)-1β precursor by the IL-1β converting enzyme, the liberation of monoubiquitin from ubiquitin precursors by ubiquitin-specific proteases, and the multiple modifications of proteins bearing a carboxyl-terminal CAAX motif (C = cysteine, A = an aliphatic amino acid, and X = one of several amino acids) by the components discussed below.

The carboxyl-terminal tetrapeptide CAAX motif is found in a number of eukaryotic proteins, including the nonclassically secreted Saccharomyces cerevisiae mating pheromone a-factor (1, 2). Proteins terminating in a CAAX motif are modified at their carboxyl-termini in a sequential three-step process consisting of isoprenylation (farnesylation or geranylgeranylation), proteolysis, and carboxylmethylation (1, 2); this three-step process will be referred to here as CAAX processing. The biogenesis of the a-factor precursor occurs by an ordered series of events involving first carboxyl-terminal CAAX processing, followed by two sequential amino-terminal proteolytic cleavages, and finally export of mature a-factor, which is an isoprenylated and carboxyl-methylated dodecapeptide (3). The export of a-factor is mediated by a nonclassical secretory mechanism involving the ATP binding cassette transporter Ste6p (4). The export of a-factor is unlike that of the α-factor mating pheromone, which traverses and undergoes proteolytic maturation within the classical vesicular secretory pathway.

The cellular components that mediate the CAAX processing of the a-factor precursor also mediate the processing of other CAAX proteins in yeast and have been identified genetically. The first step in CAAX processing, isoprenylation, requires the cytosolic Ram1p/Ram2p farnesyltransferase complex, which catalyzes the transfer of a C15 farnesyl isoprenoid to the cysteine of the a-factor CAAX motif (5, 6). The second step, proteolysis, is mediated by membrane-associated endoproteolytic activities that remove the terminal three amino acids (AAX) from the precursor (7, 8). Two yeast proteins predicted to be membrane-associated, Rce1p and Ste24p, are semiredundant for this function (9, 10). Whereas both Rce1p and Ste24p recognize the a-factor CAAX sequence (CVIA), each also recognizes its own specific subset of CAAX sequences (9). In addition, Ste24p can recognize non-CAAX sequences, as is apparent from its key role in cleaving the amino-terminal extension of a-factor (10, 11). In the third and final step of CAAX processing, membrane-associated Ste14p carboxylmethylates the newly exposed carboxyl-terminal farnesylated cysteine of the a-factor precursor (12, 13). Carboxylmethylation absolutely requires the prior removal of the carboxyl-terminal AAX residues.

On completion of carboxyl-terminal CAAX processing, the amino-terminal extension of the a-factor precursor undergoes two sequential processing steps. The Ste24p protease removes approximately the first third of the amino-terminal extension (i.e., 7 aa from the MFA1 encoded a-factor precursor), whereas the remaining portion is removed by the membrane-associated Axl1p protease (3, 11, 14). It is notable that Ste24p plays a dual role in the biogenesis of a-factor, as an amino-terminal protease, and also as a CAAX protease (10).

All of the a-factor biogenesis components are predicted or have been shown to be membrane-associated, with the exception of the Ram1p/Ram2p prenyltransferase. Yet, the precise localization of these components within the cell has not been examined previously. Therefore, an aim of this study is to establish the subcellular localization of the a-factor CAAX proteases, Rce1p and Ste24p. We demonstrate here that the CAAX proteases and the associated CAAX proteolytic activity are localized to a single intracellular membrane site, specifically the ER membrane. Elsewhere, we have determined that Ste14p methyltransferase activity is also localized to the ER membrane (15). Together, these findings indicate that CAAX processing occurs at the ER, presumably on its cytosolic face.

MATERIALS AND METHODS

Strains.

The yeast strains used in this study are all isogenic to SM1058 (MATa trp1 leu2 ura3 his4 can1), unless otherwise indicated (16). Strains were routinely grown at 30°C on complete or synthetic dropout media, as described (16). A strain in which chromosomally integrated HA∷STE24 (SM3365) replaces the wild-type STE24 locus was generated with a two-step gene disruption method by using XbaI linearized pSM1297 to transform SM1058, followed by selection on 5-fluoroorotic acid. SM3060 (ste24–1), SM3103 (ste24Δ∷LEU2), SM3613 (rce1Δ∷TRP1), and SM3614 (rce1Δ∷TRP1 ste24Δ∷LEU2) have been described (10, 11)

Plasmids.

The plasmids pOH (CEN URA3 OCH1∷HA), pSN218 (CEN URA3 KEX2∷HA), and p80 (CEN URA3 AXL1) were kindly provided by G. Waters (Princeton University, Princeton, NJ), T. Stevens (University of Oregon, Eugene) and C. Boone (Simon Fraser University, Burnaby, B.C., Canada), respectively. pSM1291 (CEN URA3 HA∷STE24-K450Q, K451Q) encodes a Ste24p protein that is tagged with hemagglutinin (HA) (triply iterated) at the amino terminus and terminates with QQKN and is referred to as Ste24p-QQ; the mutations were generated by PCR-directed mutagenesis. pSM1297 (URA3 HA∷STE24) contains the 2.7-kb BamHI–EcoRI STE24 fragment with a triply iterated HA-tag positioned immediately after the start codon in an integrating vector (pRS306; URA3) (17). pSM1275 (CEN URA3 RCE1) contains the HindIII–BglII fragment from pHY01 (18) subcloned into the same sites of pRS316 (17). pSM1314 (CEN URA3 RCE1∷HA) is the same as pSM1275, except that a triply iterated HA tag immediately precedes the stop codon of RCE1. pSM1153 (CEN TRP1 AXL1) was constructed by the recombinational cloning (19) of PvuI-digested p80 with PstI–XhoI-digested pRS314 (17). pSM962 (CEN LEU2 STE6∷myc), pSM1093 (CEN URA3 STE24), pSM1097 (CEN LEU2 HA∷STE24), and pSM1107 (CEN URA3 HA∷STE24) have been described (11, 20).

Antibodies, Immunoblotting, and Immunofluorescence.

The anti-HA antibodies were from Boehringer Mannheim (3F-10, rat anti-HA) or Babco (Richmond, CA) (12CA5, mouse anti-HA). The rabbit polyclonal Kar2p, Pma1p, and Gda1p antibodies were provided by M. Rose (Princeton University, Princeton, NJ), C. Slayman (Yale University, New Haven, CT), and C. Hirschberg (Boston University, Boston), respectively. The Rce1p antibody was generated in rabbits (Covance, Denver, PA) against a keyhole limpet hemocyanin-coupled (Pierce, Rockford, IL) synthetic peptide derived from Rce1p (amino acids 24–42 plus an added cysteine at the carboxyl terminus for coupling purposes). Rabbit Ste14p polyclonal antibodies have been described (15). For immunoblotting, SDS/PAGE separated samples were transferred to nitrocellulose, probed with appropriate primary and secondary antibodies under empirically optimized conditions, and detected by chemiluminescence (Boehringer Mannheim). Secondary antibodies used were horseradish peroxidase (HRP)-donkey anti-rabbit, HRP-sheep anti-mouse, and HRP-sheep anti-rat (Amersham). The indirect immunofluorescence detection of proteins was performed as described (21) by using primary antibodies, as noted in figure legends, and appropriate secondary antibodies. Secondary antibodies used were rhodamine-conjugated goat anti-mouse and fluorescein isothiocyanate-conjugated goat anti-rabbit (Boehringer Mannheim) and Cy3-conjugated goat anti-mouse (Jackson ImmunoResearch).

Sucrose Gradient Fractionation.

The fractionation of subcellular organelles was based on sedimentation through a sucrose step gradient essentially as described, with the following changes (15, 22). Cells were recovered by centrifugation subsequent to treatment with 50 mM Tris/10 mM DTT, pH 9.6. Spheroplasts were harvested and lysed in buffer A (10 mM Mops/0.8 M sorbitol/1 mM EDTA/0.02% NaN₃, pH 7.2) containing 1 μg/ml leupeptin and chymostatin, 2 μg/ml pepstatin and aprotinin, and 1 mM phenylmethylsulfonyl fluoride. Cleared lysates were loaded on a nine-step sucrose gradient composed of 4-ml layers of sucrose (12–54% wt/vol in 6% increments) layered over a 2-ml 60% wt/vol sucrose pad (all sucrose layers were prepared in buffer A) and centrifuged at 100,000 × g (27,000 rpm) for 3 h at 4°C in an SW28 rotor (Beckman). Equivalent fractions (4.2 ml) were collected from the bottom of the gradient and assayed for protein concentration by using the Bio-Rad Protein Assay reagent, for marker proteins by immunoblotting, and for the distribution of AAXing activity.

AAXing Assays.

AAXing activity was determined with a coupled proteolysis–methylation assay (23) (J. Otto and P. Casey, personal communication) by using 100 μg of yeast membranes (≈5 μl) or 12.5 μl of gradient fractions. Samples were assayed for 1 h at 30°C in 50 μl of AAXing reaction buffer (100 mM Hepes, pH 7.5/100 mM NaCl/5 mM MgCl₂/4 mM 1/10-phenanthroline, and protease inhibitors, as above) in the presence or absence of 2 μM farnesylated human Ki-Ras-4B. Gradient fraction samples were normalized for sucrose (final sucrose approximately 30% wt/vol) before being assayed. Proteolysis was halted and methylation was initiated with the addition of 25 μl of 3 × Methylation buffer [75 mM EDTA/60 μM AdoMet (0.2 Ci of [³H]AdoMet/mmol; New England Nuclear), 5 mM NaHPO₄, pH 7] containing 10 μg of membranes derived from Ste14p-expressing Sf9 cells. After a 10-min incubation at 30°C, each reaction was terminated with the addition of 0.5 ml of 4% SDS containing 100 μg of carrier protein (see below), incubated for 10–20 min at room temperature, and precipitated for 20 min with the addition of trichloroacetic acid (TCA) (15% final vol/vol). TCA precipitates were captured on GF/A (Whatman) glass fiber filters by vacuum filtration and were washed four times with 2% SDS/6% TCA and twice with 6% TCA. Dried filters were placed in scintillation fluid and ³H counts were determined. Counts from samples assayed with the Ki-Ras substrate were corrected for nonspecific methylation (typically 100–400 cpm) by subtraction of counts determined from a parallel sample assayed without the substrate; the average of two reaction sets is reported, with the error bars representing the difference between the mean and the sample points.

Preparation of Extracts and Carrier Protein.

Total cell extracts (2 OD₆₀₀ equivalents) were prepared by the base hydrolysis/TCA precipitation method, as described (11). Total lysates for membrane preparations were prepared by bead beating 100 OD₆₀₀ of cells in 200 μl of buffer A containing protease inhibitors (as above) after conversion of cells to spheroplasts. The lysates were cleared by centrifugation (500 × g, 10 min), and membranes were recovered by centrifugation for 1 h at 100,000 × g. The membranes were washed with buffer A, recovered by centrifugation as above, resuspended in buffer A, and assayed for protein concentration (typically 17–22 mg/ml). The supernatant from the first spin was reserved for use as carrier protein in AAXing assays.

RESULTS

RCE1∷HA and HA∷STE24 Complement a Δrce1Δste24 Strain for Mating Activity.

Proteolytic trimming of CAAX proteins in S. cerevisiae depends on the Rce1p and Ste24p proteases. These proteins are predicted to have multiple membrane spans, as determined by hydropathy analysis of their coding sequences (Fig. 1) (9, 11). Neither Rce1p nor Ste24p has been localized previously to specific intracellular membranes. Sequence analysis reveals that Ste24p possesses two clearly identifiable motifs: (i) an HEXXH sequence (where X is any amino acid) that defines Ste24p as a zinc metalloprotease (11) and (ii) a carboxyl-terminal dilysine sequence (KKXX) that has the potential to act as an ER retrieval signal in yeast (24). In contrast to Ste24p, Rce1p lacks identifiable functional or localization motifs.

Figure 1.

The predicted transmembrane structure and motifs of Rce1p and Ste24p. Kyte and Doolittle hydropathy plots of Rce1p and Ste24p (Lower) are rendered as schematic models (Upper). Predicted membrane spans are indicated [filled bars (Upper) and filled peaks (Lower)], as are the characteristic metalloprotease motif (HEIGH) and ER retrieval motif (KKKN) of Ste24p. At the top of each diagram and along the x-axis of the hydropathy plots, a scale for amino acid position is shown.

To facilitate the localization of Rce1p and Ste24p, we HA-epitope-tagged these proteases. The tagged proteins function comparably with their untagged counterparts, as determined in a yeast patch mating assay (Fig. 2A). The mating defect of the Δste24 Δrce1 double mutant (Fig. 2Aa) can be fully rescued to wild-type mating levels (Fig. 2Ab) by transformation with low-copy plasmids carrying untagged and tagged STE24 (Fig. 2Ac and 2Ad, respectively) or partially rescued with untagged and tagged RCE1 (Fig. 2Ae and 2Af, respectively). The partial rescue by untagged and tagged RCE1 is expected, since STE24 is absent in the double mutant and Ste24p has an additional role in the amino-terminal processing of the a-factor precursor that is not carried out by Rce1p (10).

Figure 2.

Complementation of the CAAX protease-deficient Δrce1 Δste24 mutant by RCE1∷HA and HA∷STE24 and immunodetection of the epitope-tagged proteins. (A) Strains were tested for their mating efficiency under stringent conditions by a patch mating assay, with the SM1068 (MATα lys1) mating partner, as described (11). The growth of prototrophic diploids is indicative of mating. MATa strains tested are: SM3614, SM1058/pRS316, SM3614/pSM1093, SM3614/pSM1107, SM3614/pSM1275, SM3614/pSM1314 for a-f), respectively. (B) Blots were probed with the anti-HA antibody (lanes 1–4) or the Rce1p antibody (lanes 5–8), as described in Materials and Methods. The approximate molecular mass (kDa) of a set of protein standards is indicated for each blot. Strains used are: SM3614/pRS316, SM3614/pSM1107, SM3614/pSM1314, SM1058/pSM1097 and pSM1314, SM3613, SM1058/pRS314, SM3613/pSM1275, and SM1058/pHY01 for lanes 1–8, respectively.

For strains expressing the epitope-tagged proteases alone or together, bands corresponding to HA-Ste24p and Rce1p-HA were detected by immunoblotting with the HA antibody (Fig. 2B, lanes 2–4). In a strain lacking these constructs (Fig. 2B, lane 1), no other bands were detected, thereby making these constructs suitable for use in immunolocalization experiments. The size of the band corresponding to HA-Ste24p (53 kDa) is close to its predicted size of 57 kDa (Fig. 2B, lanes 2 and 4). Rce1p-HA, however, migrates approximately 8–10 kDa smaller than its expected 41 kDa size (Fig. 2B, lanes 3 and 4). Similarly, using a polyclonal antibody that we generated against an amino-terminal peptide of Rce1p (amino acids 24–42), we found that untagged Rce1p also runs approximately 8–10 kDa smaller than its predicted size of 36 kDa (Fig. 2B). The polyclonal antibody detects a 27-kDa band at increasing intensity in strains expressing chromosomal (RCE1), low (CEN RCE1), and high (2 μ RCE1) copy RCE1 plasmids, respectively (Fig. 2B, lanes 6–8); this band is absent in the Δrce1 strain (Fig. 2B, lane 5). The aberrant mobility of Rce1p is unlikely to occur because of a cleavage event, since Rce1p is detected by antibodies specific for either the amino or carboxyl terminus of the protein. Instead, the aberrant mobility of Rce1p may occur because of its multispanning nature, as has been reported for other such membrane proteins (15, 25).

Components Involved in the Modification of CAAX Proteins Are ER-Localized.

Here we use cell fractionation and immunofluorescence microscopy to localize Rce1p and Ste24p. A cellular homogenate derived from a yeast strain expressing Rce1p and HA-Ste24p was fractionated on a sucrose step gradient (15, 26), and fractions were examined by immunoblotting for the distributions of Rce1p and HA-Ste24p in relation to marker proteins of various subcellular organelles. As shown in Fig. 3A, light membranes containing the vacuolar marker Vph1p, the trans-Golgi network marker Kex2p-HA, and the Golgi marker Gda1p are distributed near the top of the gradient (fractions 1–3). Heavy membranes defined by ER and plasma membrane markers (Kar2p and Pma1p, respectively) are distributed mainly in the middle of the gradient (fractions 4–6). Whereas these latter marker proteins are both present in fraction 6, the broader fractionation pattern and double-peak profile of the ER marker Kar2p is easily distinguishable from that of the plasma membrane marker Pma1p. Cytosolic proteins do not enter the gradient, as is evident by the distribution of total protein and the localization of hexokinase at the top of the gradient (fraction 1).

Figure 3.

Distribution of CAAX proteases and associated AAXing activity by subcellular fractionation. (A) A total yeast lysate derived from a strain expressing HA-Ste24p and Kex2p-HA was fractionated on a sucrose step gradient. Equivalent volumes from each fraction (50 μl) were solubilized in Laemmli’s sample buffer, subjected to SDS/15% PAGE, transferred to nitrocellulose, and probed with antibodies for Rce1p, HA (HA-Ste24p and Kex2p-HA), and organellar markers. Gradient fractions were also assayed for protein concentration and AAXing activity, as described in Materials and Methods. The broken line represents an arbitrary division between light and heavy membranes. The strain used is SM3365/pSM962/pSN218/pSM1153. (B) Total membranes prepared from yeast cells of the indicated genotypes were assayed for AAXing activity, as above. Strains used are (left to right) SM1058/pRS316, SM3103/pRS316, SM3613/pRS316, and SM3614/pRS316.

The cells used for fractionation express the Ste14p methyltransferase. Consistent with our previous report that Ste14p activity is localized to the ER (15), we find that Ste14p cofractionates with Kar2p (Fig. 3A, fractions 4–6). Likewise, the distributions of Rce1p and HA-Ste24p are most similar to that of Kar2p, suggesting that these proteases are also localized to the ER. Although unlikely, the partially overlapping distribution of ER and plasma membranes in fraction 6 allows for the possibility that a small fraction of Rce1p and Ste24p may be localized to the plasma membrane.

To ensure that the steady-state distribution of the Rce1p and Ste24p proteases reflects the distribution of their activities, we carried out an in vitro coupled proteolysis-methylation assay (AAXing assay) to measure AAXing activity across the gradient (23) (J. Otto and P. Casey, personal communication). In total yeast membranes, this assay is specific for Rce1p-dependent AAXing activity (Fig. 3B). Significant methylation is detected in wild-type and Δste24-derived membranes, whereas very low levels of methylation are detected for Δrce1- and Δrce1 Δste24-derived membranes. The lack of methylation in the Δrce1 and Δrce1 Δste24 strains can be attributed to a lack of proteolysis, which is absolutely required prior to methylation (23). Even though the Δrce1 strain should have CAAX proteolytic activity because of the presence of Ste24p, it has been reported that the yeast Ras2p CAAX sequence (CIIS) is not recognized by the Ste24p protease (9). By analogy, it may be that the human Ki-Ras CAAX sequence (CVIM) used in this assay also is not recognized by yeast Ste24p. We are currently designing a suitable substrate to assess both Rce1p and Ste24p activities in yeast. Assaying gradient fractions for AAXing activity (Fig. 3A), we determined that the fractions containing peak levels of Rce1p by immunoblot (fractions 4 and 6) also contain AAXing activity peaks, thus confirming that ER-localized Rce1p is enzymatically active.

To further define the localization of the CAAX proteases, we examined Rce1p-HA and HA-Ste24p by indirect immunofluorescence. With this technique, perinuclear ER staining in yeast is easily distinguished from plasma membrane rim staining and punctate Golgi staining. For both Rce1p-HA and HA-Ste24p, we observe perinuclear staining patterns that completely colocalize with that of the ER marker Kar2p (Fig. 4 A and C). Neither of the CAAX proteases colocalizes with the rim-staining plasma membrane marker Pma1p (Fig. 4B and data not shown). The staining patterns of the CAAX protease are also distinct from that of the punctate-staining Golgi marker Och1p-HA (Fig. 4E); this panel also illustrates that the HA antibody does not always generate a perinuclear staining pattern. Furthermore, in cells lacking epitope-tagged proteins, no staining is observable with the HA antibody (Fig. 4F).

Figure 4.

Immunofluorescence localization of Rce1p-HA and HA-Ste24p to the ER compartment. Indirect immunofluorescence detection of proteins was performed to examine the localization of CAAX components in relation to ER (Kar2p), plasma membrane (Pma1p), and Golgi (Och1p-HA) markers. Coimmunofluorescence was carried out by using the anti-HA antibody (A–E, Left), the anti-Kar2p antibody (A, C and D, Right) or the anti-Pma1p antibody (B, Right). Nonspecific staining for the anti-HA antibody is shown in F. Primary and appropriate fluorescent secondary antibodies were used at empirically determined dilutions. Images were captured at ×100 magnification by using a Zeiss microscope equipped with fluorescence optics and a charge-coupled device camera. Fluorescence filter sets were optimized such that no bleed-through was observed between fluorescent channels. Strains used are A–B: SM1058/pSM1314; C–F: SM3060/pSM1107, SM3103/pSM1291, SM1058/pOH and SM1058, respectively.

Taken together, the fractionation profiles and immunofluorescence staining patterns of Rce1p and Ste24p indicate that the CAAX proteases are indeed localized to the ER. This result is consistent with reports that AAXing activity is associated with microsomal fractions in yeast (7, 8), with the observation that Ste24p contains a canonical dilysine ER retrieval motif at its carboxyl terminus (11), and with our recent findings that the Ste14p carboxyl methyltransferase is also ER-localized (15). Our results thus indicate that all the membrane-associated CAAX processing components reside in the ER membrane.

The Ste24p Carboxyl-Terminal Dilysine Motif Is Not Essential for ER Localization.

Two carboxyl-terminal motifs are known to facilitate the retrieval of ER resident proteins from post-ER compartments in yeast: the HDEL sequence for the retrieval of ER luminal proteins such as Kar2p and the dilysine motif (KKXX or KXKXX) for the retrieval of ER membrane proteins (24, 27). Of the three membrane-associated components involved in CAAX processing, only Ste24p possesses the consensus dilysine ER retrieval motif. To determine whether this motif is essential for keeping Ste24p in the ER membrane, we performed the following experiments. First, the dilysine motif of Ste24p was mutated such that the consensus lysine residues were converted to glutamines (i.e., KKKN → QQKN). The mutation of consensus lysine residues has been reported to affect the localization of some, but not all, dilysine containing ER membrane proteins (24, 28). For the HA-tagged mutant Ste24p (Ste24p-QQ), a wild-type ER staining pattern was observed by immunofluorescence (Fig. 4D, Left); for comparison, the staining pattern for the ER marker Kar2p in the same cell is also shown (Fig. 4D, Right). Next, we examined the localization of wild-type HA-Ste24p in the temperature-sensitive ret1 mutant background, in which proteins bearing consensus KKXX motifs can escape the ER because of a defect in the ER retrieval system (29). Under restrictive conditions in the ret1 strain, HA-Ste24p remains localized to the ER (data not shown). From the above experiments, we conclude that the dilysine motif is not essential for the ER localization of Ste24p under normal growth conditions. In addition, we have determined that Ste24p-QQ is as metabolically stable as wild-type Ste24p (both half-lives are greater than 12 h), indicating that Ste24p-QQ is not being degraded at an appreciable rate. Thus, by all indications Ste24p-QQ behaves as the wild-type protein. It is therefore not surprising that Ste24p-QQ also retains normal function in a mating assay (data not shown).

DISCUSSION

Previous findings have indicated that proteolytic activities that remove the terminal three amino acids (AAX) from yeast and mammalian CAAX precursors are localized to crude membranes or microsomal fractions (7, 12, 30, 31). Our present study extends these findings, demonstrating that the Rce1p and Ste24p proteases, which mediate the proteolytic trimming of the CAAX proteins, are ER-localized in yeast. We therefore propose that, subsequent to isoprenylation, which is carried out in the cytosol, the processing of yeast CAAX proteins is an ER membrane-associated event (Fig. 5). Consistent with this view, we have shown elsewhere that the CAAX methyltransferase, Ste14p, is also ER-localized (15).

Figure 5.

Model for the organization of a-factor biogenesis components. The a-factor precursor is modified by several components. Isoprenylation is carried out by the cytosolic Ram1p/Ram2p complex. The Rce1p and Ste24p proteases and the Ste14p methyltransferase are localized to the ER membrane, presumably with cytosolically disposed active sites. Axl1p is localized to a different, as yet unidentified, compartment. The Ste6p transporter is localized to the plasma membrane. Routing through these processing stations is required for the maturation of the a-factor precursor and for the export of mature a-factor.

Because Ste24p has an additional role in the initial processing of the amino-terminal extension of the a-factor precursor, we infer from our results that the first amino-terminal processing of a-factor is also carried out at the ER membrane (Fig. 5). The second and final amino-terminal proteolytic step in a-factor biogenesis is mediated by the Axl1p protease, which has been shown to have an additional role in the axial budding of yeast (14, 32). Our preliminary studies suggest that Axl1p is localized not to ER, but rather to a different intracellular membrane compartment. We are currently further characterizing Axl1p localization. Mature a-factor is transported from the cell by Ste6p, which likely functions at the plasma membrane (33, 34).

The localization of resident ER proteins in yeast may involve two distinct signals: retention signals, of which only a small number are known, and retrieval signals, which facilitate the return of proteins that have escaped to the Golgi (24, 27, 35). In yeast, retrieval signals include the HDEL sequence of soluble proteins and the dilysine motif (KKXX or KXKXX) of membrane proteins, both of which are located at the carboxyl termini of certain ER resident proteins. Interestingly, we have shown here that the Ste24p dilysine motif is not needed for its ER localization (Fig. 4). A similar finding has been reported for Wbp1p, a subunit of the multicomponent glycosyltransferase complex in yeast (24). Therefore, we infer from our results that an as yet unidentified retrieval signal or an additional mechanism other than retrieval may be involved in the proper ER localization of Ste24p. It is possible that Ste24p can physically interact with other ER proteins and that this interaction is sufficient to maintain ER localization. Such a mechanism has been postulated for maintaining the ER localization of Wbp1p (24). Other CAAX processing components, such as Ste14p and Rce1p, are attractive candidates for cellular components with which Ste24p might interact, although neither of these components has an identifiable ER localization motif. Alternatively, Ste24p may be independently retained in the ER by an as yet uncharacterized ER retention signal.

It is likely that membrane-associated a-factor biogenesis components are oriented in the membrane such that their active sites are cytosolically disposed. This assumption is based in part on the observations that defects in the classical secretory pathway do not affect the biogenesis of the a-factor mating pheromone (i.e., a-factor does not require translocation into the luminal compartments of the secretory pathway for its processing and export) (4). In addition, the predicted topology for Ste24p places both the HEXXH protease motif and carboxyl-terminal dilysine motif on the same face of the membrane. Because dilysine motifs are normally cytosolically disposed, this potentially places the protease motif on the cytosolic face of the ER as well.

Yeast CAAX proteins can be associated with the cytoplasmic membrane face of various organelles, including the plasma membrane (Ras1p, Ras2p, and Ste18p), ER to Golgi vesicles (Ykt6p), and the ER and mitochondria (Ydj1p). Although a-factor is eventually exported from the cell, the intracellular precursors of this pheromone are associated with as yet uncharacterized membranes (3). Given our findings that several CAAX processing components (Rce1p, Ste24p, and Ste14p) are ER-localized, it is likely that a-factor precursors will be found associated with the cytosolic face of the ER membrane as well. In fact, Ras is proposed to be partially associated with internal yeast membranes in a strain that lacks the CAAX proteases (9).

The diverse intracellular distribution of CAAX proteins necessarily implies that prenylated proteins must be routed between the site of CAAX processing on the cytoplasmic face of the ER and various intracellular membrane target sites. For example, mature a-factor must reach the plasma membrane where Ste6p likely functions (33, 34). Little is known, however, about the intracellular routing of CAAX proteins. By using a plasma membrane-localized prenylated protein as an example (e.g., Ras), three intracellular trafficking mechanisms can be proposed (Fig. 6). In the first, the prenylated protein could reach the plasma membrane by dissociating from the ER membrane and diffusing across the cytosol. This mechanism may be thermodynamically unlikely, given the tight membrane association imparted by the attachment of prenyl groups to CAAX proteins (36). In a second mechanism, a plasma membrane-localized prenylated molecule may use the cytosolic face of secretory organelles as stepping stones to the plasma membrane. Whereas this mechanism may be suitable for Ras trafficking, it is unlikely to be used by a-factor, because the export of this pheromone appears to be independent of a functional secretory pathway (4). Finally, a novel protein, lipid, or vesicular carrier might mediate the trafficking of plasma membrane-localized prenylated molecules. In the case of geranylgeranylated Rab proteins, Rab escort proteins (REP1 and REP2) have been proposed to interact with and facilitate the proper trafficking of Rab proteins in the cell (37, 38). Escort proteins for farnesylated proteins such as Ras and a-factor have not been identified. We cannot exclude, of course, the possibility that the intracellular trafficking of CAAX proteins depends on more than one of the above mechanisms. The identification of possible CAAX trafficking components may come from future studies of this phenomenon.

Figure 6.

The biogenesis of CAAX proteins requires routing from the ER membrane. Whereas CAAX processing occurs at the ER membrane, CAAX proteins are destined to other membrane sites, for instance the plasma membrane. Transport of CAAX proteins such as Ras from the ER membrane to the plasma membrane could involve trafficking by one of the three possible mechanisms shown: (1) diffusional translocation across the cytoplasm, (2) transport along the cytoplasmic face of known organelles such as those of the secretory pathway, or (3) transport via an as yet unidentified protein, lipid, or vesicular carrier.

ABBREVIATIONS

ER

endoplasmic reticulum

IL

interleukin

HA

hemagglutinin

HRP

horseradish peroxidase

TCA

trichloroacetic acid