Dual Roles for Ste24p in Yeast a-Factor Maturation: NH₂-terminal Proteolysis and COOH-terminal CAAX Processing

The yeast strains used in this study are listed in Table I. The rce1-Δ1:: TRP1 deletion allele, referred to as rce1Δ, replaces codons 116 to the stop codon 315 with TRP1. Strains (SM3613, SM3614, SM3689, and SM3691) harboring the rce1Δ allele were constructed using one-step gene disruption by transforming SM1058, SM3103, SM2331, and SM3375, respectively, with a BamHI-XhoI fragment from pSM1285 bearing rce1-Δ1:: TRP1, and selecting for Trp+ transformants. The ste24Δ::LEU2 deletion allele, referred to as ste24Δ, is a γ disruption of ste24 that eliminates nearly the entire coding sequence (codons 1–444 of 453 total). Strains (SM3375 and others) harboring this allele were constructed by transformation of SM2331 with linearized pSM1072 and selection of Leu⁺ transformants, as described (Fujimura-Kamada et al., 1997). All deletion strains were confirmed by Southern analysis. MATα strains bearing the single (rce1Δ) and the double (ste24Δ rce1Δ) deletions were constructed by mating-type switching of MATa strains SM3613 and SM3614, respectively, using the HO endonuclease as described (Herskowitz and Jensen, 1991). Yeast transformations were performed by the Elble method (Elble, 1992). All strains were grown at 30°C in complete yeast extract/peptone/dextrose (YPD) media, synthetic complete drop-out (SC-URA, TRP, LEU), or synthetic minimal media (SD) (Michaelis and Herskowitz, 1988; Kaiser et al., 1994).

To assay mating, we used the semiquantitative patch mating test as described previously (Fujimura-Kamada et al., 1997). In this assay, master plates containing patches of strains to be tested are replica plated onto a mating tester lawn of the opposite mating type that has been spread on an SD plate. The lawn is prepared by resuspending the mating tester strain in YPD and spreading 0.3 ml of this suspension onto an SD plate. The stringency for the patch mating test was increased by resuspending the mating tester lawns in decreasing concentrations of YPD diluted in sterile water (100% YPD is permissive, 1% YPD is stringent) since limiting nutrients decrease plate mating efficiency, and thereby increase the mating test sensitivity. The strains tested are Trp⁻, Leu⁻, or His⁻ auxotrophs, whereas the mating testers (SM1067 and SM1068) are Lys⁻. Only cells that mate to generate prototrophic diploids are able to grow on SD minimal media after incubation for 2 d at 30°C.

To compare the level of extracellular a-factor for wild-type and mutant strains, we performed the a-factor spot halo assay as described previously (Nijbroek and Michaelis, 1998). Serial dilutions of concentrated a-factor (2 μl) were spotted onto a lawn of supersensitive halo tester cells (SM1086) spread on a YPD plate. Plates were incubated for 1 d at 30°C.

The plasmids used in this study are listed in Table II. The single-step gene disruption plasmid, pSM1285, used to generate rce1-Δ1::TRP1 alleles, was constructed as follows: A HindIII-MluI fragment containing the RCE1 open reading frame (ORF) from pHY01 (provided by A. Toh-e, University of Tokyo, Tokyo, Japan) (Yashiroda et al., 1996) was rendered blunt-ended with Klenow and subcloned into the EcoRV-SmaI sites of pBluescriptIISK (Stratagene, La Jolla, CA) to yield pSM1284. Plasmid pSM1284 was digested with EcoRI to remove a fragment corresponding to the last 200 codons of the RCE1 ORF, which was replaced with a TRP1 EcoRI fragment from pUC18-TRP1 (Sapperstein et al., 1994) to generate pSM1285.

Ubi–a-factor fusion constructs encode chimeric proteins consisting of ubiquitin (76 residues) fused either to the full-length a-factor precursor encoded by MFA1, to the NH₂ terminally truncated a-factor (P2) (see Fig. 1), or to mature a-factor (M). All of these fusions contain the intact COOH-terminal CAAX motif and are expressed under the control of the MFA1 promoter. These constructs are designated Ubi-P1, Ubi-P2, and Ubi-M, where the a-factor segments correspond to codons 1–36 (full-length), codons 8–36, and codons 21–36 of MFA1-encoded a-factor, respectively. The fusion constructs were generated by recombination-mediated PCR cloning, a method in which a linearized or gapped acceptor plasmid serves as the target for homologous recombination directed by a donor PCR fragment in yeast (Muhlrad et al., 1992; Oldenburg et al., 1997). The PCR fragments were amplified with oligonucleotides encoding precise fusion junctions between the COOH terminus of ubiquitin and the NH₂ terminus of the various species of a-factor.

The linear target vector and the donor PCR fragment were cotransformed into a strain deleted for the chromosomal a-factor genes mfa1Δ and mfa2Δ (SM2331). Subsequently, candidate plasmids were screened by yeast colony PCR. In brief, crude yeast extracts were prepared by incubating a small amount of yeast cells in 60 μl of lysis buffer (0.45% NP-40, 0.45% Tween 20, 50 mM KCl, 10 mM Tris, pH 8.3, 1.5 mM MgCl₂, 0.1% gelatin, 0.3 mg/ml zymolyase) for 90 min at 37°C. The cleared lysate (5 μl) was used as the template for the PCR screening. Plasmids were recovered from yeast as described (Robzyk and Kassir, 1992). Ubi-P1, Ubi-P2, and Ubi-M are encoded by pSM1368, pSM1369, and pSM1366, respectively. All Ubi-a-factor fusion plasmids were amplified in Escherichia coli strain DH5α, prepared by alkaline lysis, analyzed by restriction digests, and then confirmed by DNA sequencing.

To examine a-factor biosynthetic intermediates, extracts were prepared from metabolically labeled cells and radiolabeled proteins were immunoprecipitated with anti–a-factor antiserum and analyzed by SDS-PAGE, essentially as previously described (Chen et al., 1997b; Fujimura-Kamada et al., 1997). In brief, 5 OD₆₀₀ U log phase cells were harvested and resuspended in 250 μl of SD media supplemented with the appropriate amino acids. Cells were pulse labeled with 150 μCi [³⁵S]cysteine for the indicated times, and the labeling was stopped on ice by the addition of an equal volume of 2× azide buffer (40 mM methionine, 40 mM cysteine, 20 mM NaN₃, 500 mg/ml BSA). For pulse-chase experiments, chases were initiated by addition of 10 μl of 1 M cysteine per time point. Intracellular (I) and extracellular (E) fractions were processed as described (Chen et al., 1997b; Fujimura-Kamada, et al., 1997). Radiolabeled a-factor was immunoprecipitated with anti–a-factor antiserum 9-137 or 9-497, and the immunoprecipitated material was subjected to SDS-PAGE and PhosphorImager (Molecular Dynamics, Inc., Sunnyvale, CA) analysis as described (Chen et al., 1997b). To maximize the resolution between partially processed a-factor precursors, we used 16% polyacrylamide separating gels that were 13 cm in length. The half-life of each a-factor precursor (P0* or P1) was determined by quantitation of the [³⁵S] counts corresponding to the a-factor signal using ImageQuant software (Molecular Dynamics, Inc.). The 0 min time point was used as the 100% reference value for each time course experiment.

The carboxyl methylation levels of the immunoprecipitated a-factor intermediates that had been cut out of a dried polyacrylamide gel were measured (Fujimura-Kamada et al., 1997). In general, the procedure involves generating double-labeled a-factor with [³H] at the carboxyl methyl group and with [³⁵S] at the prenyl cysteine. For each a-factor species (P0*, P1, P2, or M), the ratio of [³H]/[³⁵S] cpm reflects the absolute methylation level. The relative methylation level is then determined by dividing the absolute methylation level of the mutant strain by that of the wild-type strain and converting this number to a percentage value.

Specifically, this procedure was carried out in two steps. First, cells were radiolabeled, immunoprecipitated, and subjected to SDS-PAGE, as described in the previous section, with the following changes: cells were double-labeled with 50 μCi S-adenosyl-l-[³H-methyl]methionine and 150 μCi [³⁵S]cysteine for 6 min. For this assay, labeled cells were not lysed by base treatment since base hydrolysis could release the methyl esters which we wish to detect in this assay (see below). Instead, protein extracts were prepared by vortexing labeled cells at 4°C with zirconium beads in breaking buffer (50 mM potassium phosphate buffer, pH 7.4, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 1 mM 2-mercaptoethanol, 1 mM PMSF). After separation by SDS-PAGE, the radiolabeled and immunoprecipitated a-factor intermediates were visualized by autoradiography.

Second, to determine the extent of carboxyl methylation for each a-factor intermediate, we measured the levels of [³H] and of [³⁵S] incorporation by scintillation counting. The a-factor gel bands were excised from dried gels. The [³H] labeling was determined by the vapor-phase diffusion assay, which detects volatile [³H]methyl esters (i.e., carboxyl methyl) cleaved by base hydrolysis (Xie et al., 1990; Hrycyna et al., 1991). Gel slices (P0*, P1, P2, and M) were placed in an open microfuge tube containing 150 μl of 1 M NaOH. The microfuge tube was placed inside a tightly capped 20 ml scintillation vial that contained 5 ml of scintillation fluid such that the contents of the microfuge tube did not mix with the scintillation fluid. After incubation at 37°C for 24–36 h, the released volatile [³H]methanol was measured by scintillation counting. The total amount of [³⁵S] incorporated into each a-factor species was then determined. First, the contents of the microfuge tube (including the gel slice) were neutralized with 100 μl of glacial acetic acid and then dissolved in 1 ml of Solvable (New England Nuclear Research Products, Boston, MA), a tissue solubilizer, in capped microfuge tubes at 65°C for 6 h. The [³⁵S] counts of the entire sample were then measured by liquid scintillation counting. The ratio of [³H] cpm to [³⁵S] cpm was calculated for each a-factor species to give the absolute methylation level. The relative methylation level was determined by dividing the absolute methylation level of the mutant strain by the absolute methylation level of the corresponding wild-type a-factor species.

A BLAST search (Altschul et al., 1990) of the database of expressed sequence tags (ESTs) revealed several human ESTs with high amino acid similarity to S. cerevisiae (Sc) Ste24p. These ESTs represented cDNAs from many cell types. A 703-bp HindIII fragment from EST N76181 National Center for Biotechnology Information (NCBI) (no. 69928) was used to screen ∼3 × 10⁶ clones from a human B cell cDNA library (provided by S. Elledge, Baylor College of Medicine, Houston, TX) by standard methods (Ausubel, 1987). Nine positive clones were obtained and sequenced, each of which encoded partial ORFs corresponding to the 3′ end of human STE24. The most complete Homo sapiens (Hs) clone contained an ORF encoding 383 residues and included the poly A tail. To obtain the remaining 5′ Hs STE24 ORF, this clone was used in 5′ rapid amplification of cDNA ends (RACE) PCR procedure with a fetal brain library according to the Marathon-Ready cDNA kit (Clontech, Palo Alto, CA). The full-length STE24 ORF was subcloned by recombination-mediated PCR cloning into a yeast plasmid such that the coding sequence of Sc STE24 was precisely replaced with that of Hs STE24. The resulting plasmid (pSM1468) contains the 5′ upstream and 3′ downstream sequences of the yeast STE24 ORF flanking the human STE24 ORF.

To examine the roles of Ste24p and Rce1p in a-factor maturation, we first compared the extracellular a-factor produced from strains deleted for STE24 and RCE1. We used the spot halo dilution assay, which provides a semiquantitative measure of the amount of mature a-factor secreted by these strains. In this assay, extracellular a-factor causes the growth arrest of a lawn of supersensitive MATα sst2 cells, resulting in a clear zone (Michaelis and Herskowitz, 1988; Nijbroek and Michaelis, 1998). The relative amounts of a-factor produced by the different strains are compared by determining the highest dilution that yields a clear spot (Fig. 2,A). Whereas the ste24Δ mutant produces significantly decreased levels of extracellular a-factor (eightfold less than wild-type), the rce1Δ mutant is virtually indistinguishable from wild-type (Fig. 2 A). In contrast, we detect no extracellular a-factor for the double ste24Δ rce1Δ mutant. This result is consistent with the notion that Ste24p and Rce1p play redundant roles in the COOH-terminal AAXing step of a-factor maturation, such that only one gene must be intact for AAXing to occur, as proposed by Boyartchuk et al. (1997). However, the more severe phenotype of the ste24Δ single mutant versus the rce1Δ single mutant most likely reflects the additional role of Ste24p in the NH₂-terminal processing of a-factor.

A comparison of the MATa strains in a plate mating assay indicates that their mating efficiencies parallel the results of the a-factor spot dilution assay. As shown in Fig. 2,B, the rce1Δ mutant mates like wild-type as indicated by the confluent growth of diploids, whereas the ste24Δ mutant has a leaky mating defect as indicated by growth of fewer diploids. The double mutant fails to mate altogether, consistent with its complete lack of extracellular a-factor production (Fig. 2,B). Importantly, the mating defects exhibited by these mutants is MATa cell type-specific, since MATα strains bearing either the single (ste24Δ) or double (ste24Δ rce1Δ) mutations did not have any mating defects (Fig. 2 B). Because MATa and MATα cells differ mainly by pheromone and receptor expression, the MATa mating defects seen in the ste24Δ and ste24Δ rce1Δ mutants are likely to reflect specific effects on a-factor biogenesis.

To directly examine the roles of STE24 and RCE1 in a-factor maturation, we compared the a-factor biosynthetic intermediates produced in strains deleted for STE24 and RCE1. Cells were metabolically labeled, and the a-factor biosynthetic intermediates were immunoprecipitated and separated by SDS-PAGE. For the wild-type strain, the typical profile of the intracellular (I) forms of a-factor includes the partially processed precursor species (P1 and P2) and mature (M) a-factor (Fig. 3,A, lane 1; refer to Fig. 1 for a description of each band). The band migrating slightly faster than M is the a-factor–related peptide (AFRP), whose biogenesis involves mechanisms and machinery distinct from those used to generate mature a-factor (Chen et al., 1997a). The extracellular (E) fraction contains exported mature (M) a-factor (Fig. 3,A, lane 1). For the rce1Δ mutant, the biosynthetic profile is similar to wild-type (Fig. 3 A, lane 2). Because the stepwise removal of the NH₂-terminal extension appears to be normal in the rce1Δ mutant, Rce1p does not play a major role in NH₂-terminal processing of a-factor. Furthermore, this apparently normal processing in the rce1Δ mutant is expected if Ste24p and Rce1p play redundant functions in COOH-terminal AAXing of a-factor (see below), since Ste24p would compensate for the lack of Rce1p. In contrast, the ste24Δ mutant has a dramatic phenotype of P1 accumulation (Fig. 3,A, lane 3), as we have also shown previously (Fujimura-Kamada et al., 1997). In the ste24Δ mutant, COOH-terminal AAXing is complete but the cleavage of the NH₂-terminal extension is defective, indicating that Ste24p is necessary for the first NH₂-terminal cleavage step (P1→ P2) in a-factor maturation (see also Fig. 3,B). Notably, only a tiny amount of mature a-factor produced by the ste24Δ mutant is detected in the extracellular fraction by SDS-PAGE and autoradiography after a long exposure (Fujimura-Kamada et al., 1997); the more sensitive spot halo assay detects this low level of extracellular mature a-factor (refer to Fig. 2 A).

If Ste24p does indeed function redundantly in COOH-terminal AAXing with Rce1p, then we predict that the double mutant (ste24Δ rce1Δ) would be completely defective for a-factor AAXing. Such a processing block should produce a new intermediate (P0*) that is prenylated but retains its AAX and thus cannot be methylated (refer to Fig. 1). In the double mutant (ste24Δ rce1Δ), we observe a single band, designated P0*, that shows a subtle mobility shift compared with P1 (Fig. 3 A, compare lane 4 with 3). Since this slight mobility difference is difficult to detect consistently, we used a methylation assay, described below, to show that the intermediate generated by the double mutant is truly distinct from P1.

Methylation can serve as an indirect AAXing assay, since AAXing is required to expose the methylation substrate, the free carboxyl of prenyl cysteine (Hrycyna and Clarke, 1992; Ashby and Rine, 1995; Hrycyna et al., 1995). We measured the degree of carboxyl methylation for each intracellular a-factor species made in wild-type and mutant strains. Cells were metabolically double-labeled with S-adenosyl-l-[³H-methyl]methionine and [³⁵S]cysteine, which label the COOH-methyl and prenyl cysteine of a-factor, respectively. Methylation levels were first normalized to protein levels ([³H]/[³⁵S]) and then calculated as a fraction of the corresponding a-factor intermediate from the wild-type strain (refer to Materials and Methods). As we have previously shown, the P1 intermediate produced by the ste24Δ mutant is completely methylated relative to the methylation levels of the wild-type strain (Fig. 3,B) (Fujimura-Kamada et al., 1997), indicating that this strain has AAXing activity, presumably mediated by Rce1p. In the rce1Δ mutant, we also detect a significant amount of methylation of P2 and M, although the levels are somewhat reduced, and the slowest migrating species (Fig. 3,B, top band) is not detectably methylated (see below). The methylation that occurs in the rce1Δ mutant most likely reflects the AAXing activity mediated by Ste24p. In contrast to the single mutants, the double ste24Δ rce1Δ mutant produces a single intermediate that is completely unmethylated (Fig. 3,B), and thus represents P0*, a prenylated but unAAXed biosynthetic intermediate of a-factor. We previously have not detected P0* probably because it is converted to P1 very quickly. Unexpectedly, the rce1Δ mutant produces a precursor species (Fig. 3 A, lane 2, top band) that is likely to be P0* because it is unmethylated (Fig. 3 B). In this case, and in contrast to P0* of the double ste24Δ rce1Δ mutant, P0* in the rce1Δ mutant is converted to P2 and then to M. A likely explanation is that when Ste24p recognizes P0* in the rce1Δ mutant, the dual roles of Ste24p permit it to cleave P0* at both the NH₂ and COOH termini to generate a P2-like intermediate that is subsequently methylated. Our conclusions from this analysis of a-factor intermediates generated in vivo are consistent with AAXing roles for both Ste24p and Rce1p, and an additional NH₂-terminal processing role for Ste24p only.

We have shown elsewhere that a defect in the carboxyl methylation of a-factor correlates with the metabolic instability of a-factor biosynthetic intermediates (Sapperstein et al., 1994). We compared the metabolic stability of P0* and P1 generated by the double (ste24Δ rce1Δ) and single (ste24Δ) mutants, respectively, by pulse-chase analysis (Fig. 4,A). This experiment reveals that the methylated P1 intermediate that accumulates in the ste24Δ mutant is metabolically stable with a half-life greater than 60 min (Fig. 4, A, lanes 7–9 and B). In contrast, the half-life of the unmethylated P0* species present in the double ste24Δ rce1Δ mutant is dramatically less (half-life of ∼9 min) (Fig. 4, A, lanes 10–12 and B). The biosynthetic profile and metabolic stability of the a-factor intermediates in the single rce1Δ mutant does not appear to be significantly different from that of the wild-type strain (Fig. 4 A, compare lanes 4–6 with 1–3).

Taken together, the experiments carried out in Figs. 3 and 4 indicate that although the gel mobility difference between P0* and P1 is small, they can be discriminated by a large difference in their methylation level and metabolic stability. P0* is unstable and unmethylated whereas P1 is metabolically stable and methylated. The CAAX processing of a-factor in the double mutant (ste24Δ rce1Δ) fails to proceed beyond P0*, and thus is consistent with redundant roles for Ste24p and Rce1p in COOH-terminal AAXing. In contrast, the single ste24Δ mutant accumulates the stable, methylated P1 intermediate, indicating that COOH-terminal AAXing can be carried out via Rce1p and that Ste24p is needed to mediate the first NH₂-terminal P1→ P2 processing step in a-factor biogenesis.

Since Ste24p appears to have dual roles in a-factor maturation involving both NH₂- and COOH-terminal steps, we sought a method to uncouple the two processing events. This goal necessitated a way to produce mature active a-factor independent of the normal NH₂-terminal processing steps. The expression of NH₂-terminally truncated MFA1 to directly generate the desired a-factor intermediates posed two problems: (a) translation requires an initiator methionine and thus would yield abnormal versions of the a-factor intermediates and (b) such MFA1 truncation mutants were expressed poorly (Nouvet, F., and S. Michaelis, unpublished data). Instead, we used a method for expressing polypeptides that does not require the initiator methionine (Bachmair et al., 1986). In this system, the ubiquitin structural gene is fused to the gene of interest, a-factor in our case. During translation, the ubiquitin moiety is recognized by ubiquitin-specific proteases (Ubp), which cleave precisely after the COOH-terminal residue of ubiquitin (Tobias and Varshavsky, 1991; Baker et al., 1992), and thus should yield the desired a-factor intermediates.

For this study, we made three ubiquitin–a-factor fusion constructs, designated here as Ubi–a-factor fusions (Fig. 5,A). We fused the ubiquitin gene to the full-length MFA1 gene (Ubi-P1), and to the truncated MFA1 corresponding either to the P2 intermediate (Ubi-P2) or to mature a-factor (Ubi-M). Ubi-P2 lacks the first seven residues of the MFA1 precursor and Ubi-M lacks the entire NH₂-terminal extension. As shown in Fig. 5,B, Ubi-P1, Ubi-P2, and Ubi-M directly generate the respective a-factor species that are depicted diagrammatically in Fig. 5 A.

These Ubi–a-factor fusions show normal production of mature exported a-factor, as determined by pulse-chase metabolic labeling, immunoprecipitation, and SDS-PAGE. As expected, the biogenesis profiles of Ubi-P1 and wild-type MFA1 are indistinguishable from one another (Fig. 5,B, compare lanes 5–8 with lanes 1–4). The ubiquitin moiety appears to be rapidly and effectively cleaved. Likewise, Ubi-P2 produces P2 directly, since it lacks the first seven residues of the NH₂-terminal extension. P2 is then processed correctly to M, which is exported (Fig. 5,B, lanes 9–12). The Ubi-M construct directly yields the predicted mature a-factor, which is exported normally (Fig. 5,B, lanes 13–16). However, the intracellular mature species derived from Ubi-M is metabolically unstable (Fig. 5,B, lanes 13–16). By the a-factor spot dilution assay and by mating assays carried out under stringent conditions, a strain bearing Ubi-M produces somewhat reduced extracellular a-factor activity compared with Ubi-P1 and Ubi-P2 (Fig. 5, C and D). Presumably this is due to the lower steady-state amount of exported a-factor, which in turn possibly reflects inefficient methylation or perhaps a protective role normally played by the NH₂-terminal extension (see Discussion). Since the Ubi-P1 and Ubi-P2 constructs produce mature a-factor that is stable and supports mating (Fig. 5), we used these two constructs to study specific processing steps in a-factor biogenesis.

We used the Ubi–a-factor fusion constructs to confirm the role of Ste24p in the first NH₂-terminal cleavage step of the a-factor precursor. To do so, we compared the processing of Ubi-P1 versus Ubi-P2 in the ste24Δ mutant (Fig. 6). As expected, MFA1 and Ubi-P1, which both possess the Ste24p NH₂-terminal cut site (P1→ P2), are not properly processed to mature a-factor in the ste24Δ mutant (Fig. 6, lanes 4 and 5). In contrast, mature a-factor expressed from the Ubi-P2 construct bypasses the requirement for STE24 in the ste24Δ mutant (Fig. 6, lane 6). The STE24 requirement in the production of mature a-factor by Ubi-P1, but not by Ubi-P2, provides strong independent evidence for the role of Ste24p in the first NH₂-terminal cleavage step of a-factor. Interestingly, these data also verify the obligatory sequential order of the two NH₂-terminal processing steps of the a-factor precursor, reinforcing the notion that the first seven residues of the NH₂-terminal extension must be removed before Axl1p can act in the second NH₂-terminal processing step.

Since Ubi-P2 can bypass the need for STE24 in NH₂-terminal processing, we could study the involvement of Ste24p in the COOH-terminal AAXing of a-factor in a system where its NH₂-terminal processing activity was not required to generate mature a-factor. Thus, we were able to probe the individual contributions of Ste24p and Rce1p solely with respect to the AAXing of a-factor. We examined the amount of extracellular a-factor derived from Ubi-P2 in wild-type and mutant strains by the spot halo assay. As shown in Fig. 7,A, compared with the wild-type, the single mutants harboring Ubi-P2 show reduced but equivalent levels of a-factor activity. This result with Ubi-P2 contrasts with our findings with MFA1 where the single ste24Δ and rce1Δ mutants differ markedly in extracellular a-factor levels (refer to Fig. 2 A). Apparently this difference is due to the additional role of Ste24p in NH₂-terminal processing of the full-length MFA1-encoded a-factor. The use of Ubi-P2 thus allows us to dispense with the need for the NH₂-terminal processing activity and permits us to conclude that Ste24p and Rce1p can play roughly equivalent roles in contributing to the AAXing of a-factor. As expected, the double mutant fails to produce any extracellular a-factor from Ubi-P2, confirming the redundant roles for Ste24p and Rce1p in a-factor AAXing.

We also examined the production of mature a-factor from Ubi-P2 by pulse-chase analysis in wild-type and mutant strains (Fig. 7,B). As predicted by the spot dilution test, extracellular a-factor is generated by the single mutants at a reduced level and is absent altogether for the double mutant (Fig. 7,B). Notably, in the intracellular fractions of all strains including the double mutant, we observe the conversion of P2 to M (from Ubi-P2), indicating that Axl1p processing is intact. A reasonable explanation for the presence of intracellular but not extracellular a-factor in the double mutant is that AAXing fails to occur and consequently methylation is blocked. The lack of methylation would in turn prevent recognition by the Ste6p transporter and therefore block export (Sapperstein et al., 1994). To ascertain if AAXing is defective for Ubi-P2 in the double mutant, we compared the methylation levels of a-factor when the wild-type and mutant strains are expressing Ubi-P2 (Fig. 8). For the double mutant, we detect no methylation for the slowest migrating band and thus conclude that this band is P2*, not P2 (Fig. 7,C). This verifies the overlapping roles of Ste24p and Rce1p in a-factor AAXing. The single mutants bearing Ubi-P2 show slightly different patterns of methylation from one another. The methylation profile for Ubi-P2 in the ste24Δ mutant is similar to that of the wild-type strain (Fig. 8), indicating that Rce1p can mediate AAXing very efficiently. The rce1Δ mutant strain also shows a methylation profile similar to wild-type, but only 30 min after labeling. At the early time point, the level of methylation is about half that of wild-type (Fig. 8). Thus, the AAXing mediated by Ste24p in the rce1Δ mutant may be somewhat less efficient than that mediated by Rce1p in the ste24Δ mutant. Nevertheless, either Ste24p or Rce1p can mediate sufficient AAXing to produce a substantial amount of steady-state mature a-factor.

CAAX processing is evolutionarily conserved. Multicellular organisms as well as yeast express and process CAAX proteins, such as Ras. Of the trio of human CAAX-processing enzymes, only the CAAX prenyl protease(s) has not been cloned to date. Towards this end, we searched for homologues of Ste24p in the database of expressed sequence tags and identified multiple ESTs with amino acid similarity to S. cerevisiae Ste24p (Boyartchuk et al., 1997, Fujimura-Kamada et al., 1997). Using a fragment from one of these ESTs, N76181, we screened a human B cell cDNA library and identified several positive clones encoding the 3′ partial ORFs with sequence similarity to S. cerevisiae Ste24p. The most complete clone encoded an ORF of 383 residues and contained the poly A tail. We obtained the remaining 5′ ORF sequence using the longest clone for 5′ RACE PCR with a human fetal brain library.

The human clone, which we designate Hs STE24, encodes a protein of 475 amino acids with 36% identity and 51% similarity to the S. cerevisiae Ste24p (Fig. 9,A). Like Ste24p of S. cerevisiae and S. pombe, Hs Ste24p has a characteristic zinc metalloprotease motif (HEXXH: H, histidine; E, glutamate; X, any amino acid), several corresponding transmembrane spans as predicted by hydropathy analysis (Fig. 9,B), and regions I, II, and III that are conserved among the homologues of this subfamily of zinc metalloproteases (Fig. 9,A) (Fleischman et al., 1995; Fujimura-Kamada et al., 1997; Kornitzer et al., 1991). Interestingly, Hs STE24 has a short stretch of residues within region I that is not present in either yeast homologue (Fig. 9,A, dashed lines). Hs Ste24p does not possess a consensus dilysine ER localization signal at its COOH terminus but does contain lysines at positions −3 and −6 from the COOH terminus (Fig. 9 A), which may reflect a degenerate ER localization signal.

We expressed the human STE24 clone in S. cerevisiae strains deleted for STE24 and RCE1 to examine whether it can complement their mating defects. We find that Hs STE24 partially corrects the mating defect of the double ste24Δ rce1Δ mutant strain (Fig. 9 C) and fully complements the modest mating defect of the single ste24Δ mutant under stringent mating conditions (data not shown). These data suggest that Hs Ste24p can carry out both the NH₂-terminal processing and the COOH-terminal AAXing steps in a-factor maturation.

The analysis of biosynthetic intermediates in a-factor biogenesis has facilitated the characterization of processing activities required for several posttranslational modifications, including the CAAX farnesyltransferase (Ram1p/ Ram2p), the prenyl cysteine carboxyl methyltransferase (Ste14p), and the a-factor P2→ M protease (Axl1p) (Hrycyna and Clarke, 1990; Schafer et al., 1990; He et al., 1991; Adames et al., 1995). The present study focuses on the most recently discovered a-factor processing component, Ste24p, a transmembrane protease that has been previously assigned distinct, and seemingly disparate roles in a-factor processing, by us and others. We showed that Ste24p is required for the first NH₂-terminal cleavage step in a-factor biogenesis, the P1→ P2 cleavage, by characterizing the a-factor intermediate that accumulates in a ste24 mutant (Fujimura-Kamada et al., 1997). In contrast, Boyartchuk et al. (1997) provided evidence that Ste24p and a dissimilar, but functionally redundant gene product (Rce1p) are involved at a different step, COOH-terminal CAAX processing of a-factor, using a biochemical assay. Here we investigate and present compelling evidence for the unifying hypothesis that Ste24p mediates dual steps in the maturation of the a-factor precursor: COOH-terminal CAAX processing and NH₂-terminal (P1→ P2) processing. We show that Ste24p functionally overlaps with Rce1p only for CAAX processing. Our evidence for the roles of Ste24p and Rce1p in a-factor maturation is summarized below.

We demonstrate that Ste24p and Rce1p are functionally redundant for COOH-terminal CAAX processing of a-factor by the characterization of biosynthetic intermediates generated in both the ste24 and rce1 single and double mutants. A significant amount of AAXing occurs (indicated by methylation) in each of the single mutants (ste24 or rce1) but completely fails to occur in the double mutant (ste24 rce1) (refer to Fig. 3 B) (see below for a discussion of P0*). Since either STE24 or RCE1 is sufficient for AAXing, and deletion of both genes results in a complete defect in AAXing, we conclude that Ste24p and Rce1p share redundant roles in a-factor AAXing and appear to be the only CAAX proteases for a-factor in yeast.

A key finding here is the identification of the novel a-factor biosynthetic intermediate, P0*, in the double mutant (ste24 rce1). P0* is an early a-factor intermediate that is incompletely COOH terminally processed (prenylated but not AAXed or methylated). P0* is formed before P1, from which it shows only a subtle difference by gel mobility (refer to Figs. 1 and 3). The lack of methylation and metabolic instability of P0* distinguish it from P1, which is fully methylated and metabolically stable. These methylation differences between P1 and P0* stem from the absence or presence of the AAX tripeptide, respectively, which must be absent for the methyltransferase to act. The P0* species of a-factor is transient, and thus is not readily apparent in a wild-type strain. Presumably this is because each step of CAAX processing occurs in rapid succession, resulting in the fast conversion of P0* to P1; we can detect fully processed a-factor (M) even after a short 5 min of pulse labeling. The accumulation of the unmethylated (and unAAXed) P0* in the double mutant (ste24 rce1) provides the basis for our conclusion that Ste24p and Rce1p have redundant roles for the COOH-terminal AAXing of a-factor.

A role for Ste24p in NH₂-terminal processing of a-factor was implicated by a block in P1→ P2 processing that results in P1 accumulation (neither P2 nor M is formed) in our initial study of the ste24 mutant (Fujimura-Kamada et al., 1997). In the present study, we provide independent confirmation that Ste24p is in fact necessary for this step by use of Ubi–a-factor fusions. The observation that production of mature a-factor from a construct that lacks the P1→ P2 cut site (Ubi-P2) can bypass the Ste24p requirement for NH₂-terminal processing is consistent with the role of Ste24p in the first NH₂-terminal processing step of a-factor maturation (Fig. 6). Interestingly, although Ste24p and Rce1p function redundantly for CAAX processing of a-factor, Rce1p is unlikely to contribute to NH₂-terminal processing, based on the finding that P1 fails to be converted to P2 in the ste24 mutant where Rce1p is present (refer to Fig. 3 A).

As described above, Ste24p has dual roles in a-factor processing and functionally overlaps with Rce1p in one of these steps. This predicts that Rce1p should be dispensable for a-factor processing. Indeed, we observe that in the rce1 single mutant, the a-factor biogenesis profile is similar to wild-type; P2 and M are generated at similar rates in the rce1 mutant as in a wild-type strain and export of M is also indistinguishable in these two strains (refer to Fig. 3,A and Fig. 2).

Having established here that Ste24p participates in two distinct proteolytic processing steps for a-factor, we can explain the previous discrepant findings for the role of Ste24p (Boyartchuk et al., 1997; Fujimura-Kamada et al., 1997). These were due to (a) the use of two different assays and (b) the complexity arising from the situation that one protein, Ste24p, carries out two functions and that two proteins, Rce1p and Ste24p, mediate a common step. Our combined approach of examining the NH₂- and COOH-terminal processing steps of a-factor in strains deleted for STE24 and RCE1 has now resolved this issue.

In light of our current view of dual functions for Ste24p (COOH-terminal AAXing and NH₂-terminal cleavage) and a single function for Rce1p (COOH-terminal AAXing), we can reevaluate the mature a-factor production and mating phenotypes of the ste24 and rce1 single and double mutants (Fig. 2). The double mutant (ste24 rce1) is completely sterile because it can carry out neither COOH- nor NH₂-terminal processing, and thus does not produce any mature a-factor. As expected, the rce1 mutant has a wild-type mating efficiency because Ste24p mediates both NH₂- and COOH-terminal processing of a-factor in this strain. In contrast, the ste24 mutant has a dramatically decreased mating efficiency (5% of wild-type) (Fujimura-Kamada et al., 1997), that is due solely to a block in NH₂-terminal processing, since COOH-terminal AAXing can be carried out by Rce1p in the ste24 mutant.

The residual mating exhibited by the ste24 mutant is significant since it reflects a low level of mature a-factor production. What gene product is responsible for the residual a-factor processing in the ste24 mutant? An unknown enzyme, or possibly even Rce1p, could mediate a very low level of P1→ P2 cleavage. Alternatively, the P1→ P2 processing step might be bypassed at a low level by a one-step removal of the entire NH₂-terminal extension from P1. A candidate for this latter possibility is Axl1p, whose normal role is to mediate the second NH₂-terminal (P2→ M) cleavage step (between N21 and Y22) in a-factor processing. Axl1p may be able to generate mature a-factor directly from P1 somewhat inefficiently (P1→ M cleavage). We favor the second hypothesis because overexpression of AXL1 can significantly suppress the ste24 mating defect (Fujimura- Kamada, K., and S. Michaelis, unpublished results).

As discussed above, the SDS-PAGE and phenotypic analyses of the rce1 and ste24 single and double mutants indicate that either Ste24p or Rce1p is sufficient for the COOH-terminal AAXing of a-factor. Using the Ubi-P2 construct to circumvent the Ste24p requirement for NH₂-terminal processing, we could investigate the individual contributions of Ste24p and Rce1p solely to the COOH-terminal AAXing of a-factor. Based upon a-factor spot dilution assays and SDS-PAGE analysis (refer to Fig. 7), we concluded that Ste24p and Rce1p each can carry out a substantial amount of the AAXing of a-factor, since the single mutants produced approximately equivalent amounts of the bioactive mature species. Although the levels of P2 methylation in the single mutants expressing Ubi-P2 are indistinguishable from one another at a later time point, at an early time point their methylation levels differ by ∼50% (i.e., a twofold lower level of methylation in the rce1 versus the ste24 strains for P2 [refer to Fig. 8]). This is consistent with findings for the somewhat asymmetric roles of Ste24p and Rce1p in a-factor AAXing (35 versus 60%, respectively) using an in vitro AAXing assay with a synthetic peptide substrate and membrane extracts from mutant strains (Boyartchuk et al., 1997).

It is rather surprising that Ste24p and Rce1p function redundantly in the CAAX processing of a-factor, since they are neither homologues nor do they share common consensus sequences. Whereas Ste24p possesses the HEXXH zinc metalloprotease motif, Rce1p lacks any known protease motifs. The sole resemblance between Ste24p and Rce1p is that both are extremely hydrophobic with several predicted membrane spans. This situation contrasts with another redundant pair of a-factor processing components, Axl1p and Ste23p, that are structurally similar to each other and contain the same protease motif (Adames et al., 1995). Both Axl1p and Ste23p can mediate the final cleavage step (P2→ M) in a-factor maturation, accounting for >90% and <10% of processing, respectively (Adames et al., 1995).

Given the dissimilarity of Ste24p and Rce1p and that their roles do not overlap completely, it is possible that one (or both) of these proteins acts indirectly in a-factor maturation. Ste24p and/or Rce1p may regulate another protease(s) that cleaves a-factor. Because Rce1p lacks any currently known protease motifs, it may either be a cofactor involved in AAXing or it may represent a novel protease. The development of an in vitro assay with purified Ste24p (or Rce1p) and its substrate will resolve this issue. Since both candidate proteases, Ste24p and Rce1p, as well as the potential substrate, prenylated a-factor, are quite hydrophobic and likely membrane bound, the purification and assay development will be challenging.

STE24 likely encodes a protease, however, because the conserved protease motif of Ste24p is necessary to complement the mating and a-factor production defects of a ste24 mutant (Boyartchuk et al., 1997, Fujimura-Kamada et al., 1997). Furthermore, mutations in the HEXXH motif are sufficient to abolish AAXing as well as NH₂-terminal processing (data not shown). Aside from the Ste24p subfamily of unusually hydrophobic zinc metalloproteases (Fujimura-Kamada et al., 1997), only one other subclass of HEXXH multispanning membrane proteases is known, the unrelated S2P protease involved in cholesterol homeostasis (Rawson et al., 1997). Both the S2P protease and its substrate are very hydrophobic (Rawson et al., 1997). S2P releases the sterol regulatory element binding protein transcription factor by cleaving within a transmembrane span (Hua et al., 1996; Sakai et al., 1996). It will be interesting to see how such a hydrophobic protease recognizes and processes its similarly hydrophobic substrate and whether Ste24p and S2P use analogous processing mechanisms.

If Ste24p directly cleaves a-factor, then how does Ste24p cut its substrate at two distinct sites? The COOH-terminal AAXing site (between residues C32, which is prenylated, and V33) bears no resemblance to the NH₂-terminal P1→ P2 processing site (between residues T7 and A8). One explanation is that the prenyl cysteine serves as a landmark or recognition site for recruitment of Ste24p and that cleavage occurs at certain nearby sites. Alternatively, Ste24p could have broad substrate specificity, like the signal peptidases that recognize hydrophobic residues (for review see Dalbey et al., 1997). Ferreting out the rules for Ste24p cleavage will require assaying a large number of defined substrates.

Since Ste24p and Rce1p are not homologues but do function redundantly in a-factor processing, it is reasonable to postulate that each enzyme recognizes subsets of overlapping substrates. Various CAAX-containing proteins possess different CAAX sequences. Specific CAAX sequences can direct farnesylation, geranylgeranylation, or even both (Moores et al., 1991; Trueblood et al., 1993, 1997; Caplin et al., 1994, 1998), which could also provide another level of substrate discrimination for Ste24p and Rce1p. Although Ste24p and Rce1p mediate processing of the a-factor CAAX box, each is also likely to promote AAXing of other yeast CAAX proteins, such as Ras or the γ subunit (Ste18p) of the heterotrimeric G protein that transduces the pheromone response. For example, it has been suggested that Rce1p is involved in AAXing of Ras2p, whereas Ste24p is not (Boyartchuk et al., 1997; Schmidt et al., 1998). The function of another CAAX protein, Ste18p, does not require Ste24p and Rce1p since the MATα ste24 rce1 mutant does not have a mating defect (Fig. 2 B). This suggests that (a) Ste18p does not require AAXing by Ste24p and Rce1p for its function, (b) AAXing is not necessary for Ste18p function, or (c) another unidentified CAAX protease processes Ste18p. The substrate specificities of Ste24p and Rce1p remain to be elucidated.

In addition to reaffirming the role of Ste24p in NH₂-terminal processing, the Ubi–a-factor fusions provided a method to verify the order of a-factor processing events and also to examine the role of the NH₂-terminal extension in the production and export of mature a-factor. First, the NH₂-terminal extension must be removed in two successive steps for production of mature a-factor. This sequence of events is based on the finding that the second NH₂-terminal processing step by Axl1p can proceed efficiently only after completion of the first NH₂-terminal cleavage, either via Ste24p or as encoded by Ubi-P2. The P1→ P2 cut could expose the substrate recognition site for Axl1p. Second, since the NH₂-terminal extension is ultimately removed from a-factor before its export, a reasonable hypothesis is that the NH₂-terminal extension could play a transient role, such as targeting a-factor to its transporter, Ste6p. However, our results with Ubi-M suggest that this is not the case because Ubi-M, which lacks the NH₂-terminal extension, can still produce mature a-factor that is properly exported (refer to Fig. 5 B). Instead, the major difference between mature a-factor generated conventionally from MFA1 or unconventionally from Ubi-M is that the latter is highly metabolically unstable. An interesting possibility is that the NH₂-terminal extension plays a role in protecting a-factor from degradation. A candidate for a degradation pathway is the N-end rule pathway since the first residue (tyrosine) of mature a-factor is a destabilizing residue. Another possibility is that the NH₂-terminal extension acts as a chaperone to stabilize mature a-factor. Alternatively, the mature species derived from Ubi-M may not be methylated as efficiently as wild-type a-factor, and thus could be degraded by the same machinery that degrades P0*.

Most secreted molecules, such as the other S. cerevisiae mating pheromone a-factor, undergo posttranslational processing in the lumenal compartments of the secretory pathway. In contrast, a-factor and other CAAX proteins are thought to be COOH terminally processed in the cytosol or on the cytosolic face of membranes. For example, it has been suggested that most of unprocessed Ras2p accumulates on intracellular membranes in yeast strains deleted for the CAAX proteases (ste24 rce1) (Boyartchuk et al., 1997). We have sought elsewhere to define the intracellular site where a-factor CAAX processing and NH₂-terminal processing take place (Schmidt et al., manuscript submitted for publication). In those studies, we demonstrated by subcellular fractionation and indirect immunofluorescence that the CAAX proteases (Ste24p and Rce1p) and the methyltransferase (Ste14p) are localized to the membrane of ER, presumably with their active sites facing the cytosol (Romano et al., 1998, Schmidt et al., manuscript submitted for publication). Those findings, together with the results presented here that show dual roles for Ste24p, lead to our current view of the intracellular trafficking and processing of a-factor (refer to Fig. 1). The newly synthesized a-factor precursor (P0) is prenylated by the cytosolic Ram1p/Ram2p complex, permitting its association with the ER membrane. At the cytosolic face of the ER membrane, prenylated a-factor (P0*) completes COOH-terminal CAAX processing, including AAXing by Rce1p or Ste24p, followed by carboxyl methylation by Ste14p to form P1. Next, the first NH₂-terminal cleavage step is mediated by Ste24p to yield P2. P2 is subsequently shuttled to another compartment where Axl1p (or Ste23p) (Schmidt, W.K., and S. Michaelis, unpublished data) performs the final cleavage step to generate M, which is then exported by Ste6p.

This model raises two intriguing issues. First, since Ste24p, Rce1p, and Ste14p are localized to the same intracellular membrane and act sequentially in CAAX processing, these three components could form a complex, a notion that we are presently investigating. Second, how does a prenylated protein, and a-factor in particular, traffic from the ER to its final destination (the plasma membrane)? We propose that a-factor could diffuse through the cytosol, use a carrier, or hitchhike on the outside of vesicles using the classical secretory pathway (Schmidt et al. 1998).

The COOH-terminal CAAX protein motif and its processing enzymes are conserved evolutionarily among eukaryotes. Here we report the cloning of the first mammalian CAAX prenyl protease, human (Hs) STE24. We have shown elsewhere that S. cerevisiae (Sc) STE24 defines a novel subfamily of zinc metalloproteases containing homologues from bacteria (E. coli, H. influenza) and other fungi (S. pombe) (Fujimura-Kamada et al., 1997). Additionally, ESTs that are similar to STE24 exist in many multicellular organisms (A. thaliana, C. elegans, D. melanogaster, M. musculus, H. sapiens). Members of the Ste24p subfamily are characterized by multiple membrane spans, a zinc metalloprotease motif (HEXXH), several highly conserved regions designated I, II, and III, and a COOH-terminal dilysine motif potentially involved in ER retrieval. Hs Ste24p and Sc Ste24p are 36% identical and 51% similar; Hs Ste24p shares the common features of this subfamily. However, instead of the canonical dilysine motif for ER retrieval (lysines at positions −3 and −4, or −3 and −5 from the COOH terminus), Hs Ste24p possesses what may be a degenerate dilysine motif (lysines at positions −3 and −6). It will be interesting to determine if human Ste24p, like its S. cerevisiae counterpart, is localized to the ER membrane. We expect that this will be the case, since another CAAX processing component (Ste14p methyltransferase) localizes to the ER membrane in yeast and mammalian cells (Dai et al., 1998; Romano et al., 1998).

We show by complementation that Hs Ste24p can function in yeast (Fig. 9,C). Notably, Hs Ste24p complements the mating defect of the double mutant (ste24 rce1), suggesting that it is capable of mediating not only COOH-terminal AAXing, but also NH₂-terminal processing of a-factor (Fig. 9 C). What is the physiological substrate for Hs Ste24p; is there a human homologue for a-factor that undergoes similar processing steps? Clearly, no such human substrate has been discovered to date. A potential substrate is the prelamin A precursor, which undergoes a series of maturation steps that includes COOH-terminal CAAX processing, followed by a subsequent proteolytic cleavage located 14 residues NH₂-terminal to the prenyl cysteine (Weber et al., 1989). This latter cleavage may be analogous to the NH₂-terminal processing of a-factor, which occurs at sites that are 26 and 12 residues away from the prenyl cysteine (Chen et al., 1997b). The endoprotease that processes the prelamin A precursor has yet to be cloned, although an enzymatic activity has been detected in nuclear extracts (Kilic et al., 1997). In addition to a possible role in NH₂-terminal proteolytic cleavage similar to the P1→ P2 cut of the a-factor precursor, Hs Ste24p is also likely to be involved in the processing of multiple CAAX proteins found in diverse human cell types. ESTs of Hs STE24 have been identified in cDNA libraries derived from a variety of tissues (fetal brain, bone, prostate, fetal lung, pancreas tumor, human tonsillar cells enriched for germinal center B cells, retina, and heart) indicating a wide expression pattern. We anticipate that future studies will reveal the mammalian substrates of Ste24p, some of which may use a processing pathway similar to that of yeast a-factor.

We thank L. Roman, R. Jensen, and members of the Michaelis laboratory for critical reading of the manuscript (all from Johns Hopkins University School of Medicine, Baltimore, MD). We are grateful to the members of the Michaelis laboratory for valuable discussions. We thank A. Toh-e and A. Varshavsky (California Institute of Technology, Pasadena, CA) for plasmids; S. Elledge for the human B cell cDNA library; L. Roman, P. Bhanot, and D. Cabin (all three from Johns Hopkins University School of Medicine) for technical advice.

This study was supported by a grant from the National Institutes of Health (GM41223) to S. Michaelis.

E

extracellular

EST

expressed sequence tag

Hs

Homo sapiens

I

intracellular

M

mature

ORF

open reading frame

Sc

Saccharomyces cerevisiae

SD

synthetic minimal medium

Sp

Schizosaccharomyces pombe

Ubi

ubiquitin

Ubp

ubiquitin-specific protease

YPD

yeast extract/peptone/dextrose medium