Modulation of Ras and a-Factor Function by Carboxyl-Terminal Proteolysis

See allHide authors and affiliations

Science  21 Mar 1997:
Vol. 275, Issue 5307, pp. 1796-1800
DOI: 10.1126/science.275.5307.1796


Prenylated proteins contain a covalently linked cholesterol intermediate near their carboxyl-termini. Maturation of most prenylated proteins involves proteolytic removal of the last three amino acids. Two genes in Saccharomyces cerevisiae, RCE1 and AFC1, were identified that appear to be responsible for this processing. The Afc1 protein is a zinc protease that participates in the processing of yeast a-factor mating pheromone. The Rce1 protein contributes to the processing of both Ras protein and a-factor. Deletion of both AFC1 and RCE1 resulted in the loss of proteolytic processing of prenylated proteins. Disruption of RCE1 led to defects in Ras localization and signaling and suppressed the activated phenotype associated with the allele RAS2val19.

Eukaryotic cells differ from prokaryotes in their use of specialized compartments and membranes to localize metabolic pathways, structural components, and regulatory processes. Some peripheral membrane proteins depend for proper membrane localization on co- or posttranslational covalent attachment of a lipid moiety to the protein. The most recently discovered such modification involves the covalent attachment of cholesterol biosynthetic intermediates to proteins and is referred to generically as protein prenylation (1).

Prenylation is catalyzed by one of three different prenyl-transferase enzymes, which attach either the 15-carbon farnesyl or the 20-carbon geranylgeranyl moieties to a cysteine near the COOH-terminus of the protein. Which enzyme prenylates which substrate is determined by the amino acid sequence at the COOH-terminus of the substrate. A pair of COOH-terminal cysteine residues, together with other determinants, direct geranylgeranylation by a type II protein geranylgeranyl transferase (2). In contrast, a CAAX motif, in which C is cysteine, A is any aliphatic amino acid, and X is any amino acid, is able to direct either geranylgeranylation or farnesylation by one of two different type I protein prenyl-transferases. If the X position is occupied by leucine or phenylalanine, the protein is geranylgeranylated (3). Other amino acids at this position cause farnesylation.

Prenylated proteins include the Ras oncogenes, fungal pheromones, the γ subunits of trimeric guanosine triphosphate (GTP)-binding proteins (G proteins), and a wide range of small GTP-binding proteins that participate in vesicular trafficking (4). Prenylation is critical to the activity of many of these proteins. Farnesylation of the Ras protein of yeast and humans is necessary for its membrane localization and transforming activity (5). Inhibitors of protein farnesyl transferases have been developed as potential treatments for Ras-based tumors (6). Farnesylation of the mating pheromone a-factor of the yeast Saccharomyces cerevisiae is the first step in a series of posttranslational modifications that convert the 36-amino acid precursor to the 12-amino acid farnesylated mature a-factor. Farnesylation appears to be essential for all subsequent modifications of a-factor, including proteolytic trimming by NH2- and COOH-terminal proteases and export by the Ste6 protein (1, 7).

In prenylated proteins with a CAAX motif, prenylation is almost invariably followed by two additional posttranslational modifications: The three COOH-terminal residues are proteolytically removed, and the carboxyl group of the resulting prenylcysteine is carboxyl-methyl esterified (8). The key unresolved issue in protein prenylation concerns the identity and role of the protease or proteases responsible for the COOH-terminal trimming. An endoproteolytic activity that cleaves the terminal tripeptide from prenylated substrates, which we refer to as RACE (Ras and a-factor converting enzyme), has been identified in membrane fractions of yeast, Xenopus, and mammalian cells. The processing of peptides by RACE is dependent on the prenylation of the substrate (9). Previously, the inability to identify the gene (or genes) encoding RACE made it impossible to establish unequivocally the relation between CAAX proteolysis and the function of prenylated proteins. We have now identified a pair of proteins required for proteolysis of prenylated proteins in S. cerevisiae and have established the role that each plays in the in vivo and the in vitro processing of Ras and a-factor precursors.

The production of and response to mating pheromones is critical for mating of haploid yeast cells. Yeast mutants defective in either farnesylation of the a-factor precursor or prenylcysteine methylesterification cannot produce active pheromone and are sterile (10). Because these steps precede and follow COOH-terminal proteolysis of a-factor, respectively, a mutation in the COOH-terminal protease should also cause sterility. To isolate such sterile mutants, we adapted an autocrine arrest selection and conditionally expressed the a-factor receptor in cells of the a-mating type (11). Cells that both synthesize and respond to their own a-factor arrest in the G1 phase of the cell cycle and fail to divide. However, mutants defective in either the synthesis of or response to a-factor survive. To further sensitize the selection, we removed the two structural genes that encode a-factor in yeast and used a mutant form of the a-factor gene as the source of the pheromone (12). The three COOH-terminal amino acid codons, specifying isoleucine, valine, and alanine (CVIA) of a-factor, were replaced with codons for alanine, methionine, and glutamine (CAMQ), respectively (13). This CAMQ sequence is present at the COOH-terminus of the α subunit of rabbit muscle glycogen phosphorylase kinase, which is farnesylated but not proteolyzed (14). Thus, the CAMQ extension on a-factor could either block the proteolytic processing of a-factor or reduce its efficiency and thus provide a sensitized screen for mutants defective in CAAX proteolysis. A yeast strain lacking wild-type a-factor genes and expressing this mutant a-factor was able to mate and produced nearly wild-type amounts of the pheromone (Fig. 1A). The autocrine arrest selection performed with this strain identified a mutation in a previously unknown gene afc1-2 (a-factor converting enzyme) that is required for the processing of a-factor. The afc1-2 mutation reduced mating efficiency and production of a-factor in strains with the CAMQ form of the a-factor gene. An afc1-2 mutant with the wild-type a-factor gene had only a modest decrease in a-factor production (Fig. 1B). The residual a-factor produced in the afc1 mutant indicated either that afc1-2 was a leaky mutation or that a second protease capable of processing a-factor existed.

Fig. 1.

Amount of biologically active a-factor in wild-type and afc1-2 mutant strains depends on the amino acid sequence of the COOH-terminus of the a-factor gene. The wild-type CVIA sequence at the COOH-terminus of the MFA1 gene was changed to CAMQ (13). Both forms of the gene were introduced in the strain deleted for the MFA1 and MFA2 a-factor genes. Levels of a-factor production were assayed by spotting samples of the cell suspension on a lawn of a-factor supersensitive MATα sst2 cells (30). Released a-factor formed halos of growth-arrested MATα cells of sizes proportional to the amounts of a-factor processing. (A) Wild-type strain (JRY5312); (B) the afc1-2 mutant (JRY5313).

The sequence of the AFC1 gene, cloned by complementation from a recombinant library, revealed that the 453-amino acid Afc1 protein was likely to be a polytopic integral membrane protein. The COOH-terminal sequence of the protein (KKKN; K is Lys and N is Asn) matched the consensus KKXX sequence that mediates retrieval of endoplasmic reticulum proteins from the Golgi (15). Most importantly, Afc1p had a perfect match at position 297 to 301 to the HEXXH motif (H is His and E is Glu) of Zn-dependent metalloproteases and matched a larger consensus sequence characteristic of neutral Zn metalloproteases. To test the significance of this motif, we made two mutants by site-directed mutagenesis, each of which had one of the conserved histidines changed to alanine. Both mutations completely blocked the ability of the gene to complement the a-factor-processing defect of afc1-2 mutants, establishing that this protease motif was essential for Afc1p function. A null allele of AFC1 (afc1Δ) was made by replacing a part of the open reading frame by the selectable marker HIS3 (16). Disruption of the AFC1 gene showed that it was not required for viability, caused no mating defect in α cells, and affected a-factor production in a cells (Fig. 2).

Fig. 2.

Production of active a-factor depends on both Afc1p and Rce1p. Halo assays were performed on cell suspensions from isogenic strains differing as indicated. The MATa strains used in the test were as follows: wild type (JRY5314), afc1Δ::HIS3 (JRY5315), rce1Δ::TRP1 (JRY5316), and afc1Δ::HIS3 rce1Δ::TRP1 (JRY5317) (12).

The presence of a sequence motif suggestive of a Zn-dependent protease in Afc1p was surprising in that previous studies of the RACE activity of membrane fractions established that the activity was only slightly affected by the Zn chelator 1,10-phenanthroline. To determine whether AFC1 encoded the enzyme responsible for the RACE activity, we compared membrane fractions from wild-type and afc1Δ mutants for their ability to cleave the COOH-terminal three amino acids from a prenylated CAAX peptide by two different in vitro assays (17). In both assays, the afc1Δ mutation resulted in a slight but reproducible decrease in RACE activity (Fig. 3). The decrease in activity caused by the afc1Δ mutation was similar in magnitude to the decrease in activity of wild-type extracts caused by chelation of Zn. Moreover, removal of Zn caused no further decrease in activity in the afc1Δ extracts. A simple explanation for these observations is that RACE activity is a composite of two different CAAX proteases: a Zn-dependent activity encoded by AFC1 and a second protease encoded by a different gene. The active a-factor produced by an afc1 mutant would presumably result from the second protease.

Fig. 3.

Direct CAAX proteolysis assays. P100 membrane fractions were prepared from late logarithmic phase cultures of wild-type (JRY5314), afc1Δ::HIS3 (JRY5315), rce1Δ::TRP1 (JRY5316), and afc1Δ::HIS3 rce1Δ::TRP1 (JRY5317). The proteolytic activity was determined by measurement of the release of tritiated tripeptide from the farnesylated substrate (17). Zinc chelation experiments were performed in the presence of 4 mM 1,10-phenanthroline (1,10-Ø). The proteolytic activity was calculated relative to that of the wild-type strain in the absence of 1,10-Ø.

The AFC1 sequence has no homologs in the yeast genome, indicating that residual a-factor processing was performed by a structurally distinct protease. To search for additional CAAX prenyl proteases, we used an afc1 mutant strain to screen for mutations that would block the remaining a-factor production. In addition, multicopy libraries were used to screen for plasmids that, at high copy, could partially restore a-factor production in afc1 mutants with the CAMQ form of a-factor. Both routes led to the identification and cloning of the same gene, RCE1 for Ras and a-factor converting enzyme. The sequence of RCE1 revealed no sequence homologies adequate to imply a function for the protein. As with AFC1, RCE1 appeared to be a polytopic integral membrane protein, with a slight similarity to type IIb signal peptidases, which cleave signal sequences from proteins containing nearby lipid modifications. As a test of whether RCE1 encodes or regulates a CAAX prenyl protease, a null allele of RCE1 (rce1Δ) was constructed by replacing the entire open reading frame with the selectable marker TRP1. Disruption of the RCE1 gene revealed that, like AFC1, RCE1 was not essential for viability, had no effect on the mating ability of α cells, and caused only a modest decrease in a-factor production. The double mutant rce1Δ afc1Δ grew as well as wild type. Thus, CAAX proteolysis is not essential for viability. However, the double mutant was completely defective in the production of mature a-factor (Fig. 2). Therefore the functions of RCE1 and AFC1 are at least partially overlapping, and both proteins contribute to the production of mature a-factor.

As with afc1 mutants, the rce1 mutants showed reduced CAAX prenyl protease activity in two different in vitro assays (17) and was more defective than the afc1 mutant. Moreover, the activity that remained in an rce1Δ mutant was completely blocked by removal of Zn, as expected for the Zn-dependent protease encoded by AFC1 (Fig. 3). The double mutant had low levels of CAAX prenyl protease activity indistinguishable from background levels. Thus, AFC1 and RCE1 appeared to encode a pair of CAAX prenyl proteases that lack any sequence similarity but have overlapping substrate specificity.

These data implied that Rce1p was a CAAX protease; however, Rce1p could also be a regulator of the second CAAX protease. To determine whether RCE1 was actually the structural gene for the second CAAX protease, we expressed a tagged RCE1 gene from different strength promoters and tested whether different expression levels result in different levels of activity (13). Each promoter produced a different level of CAAX protease activity in an afc1Δ rce1Δ double mutant, and the amount of activity was linearly proportional with respect to the amount of protein, suggesting that Rce1p itself was the CAAX protease (18). Nevertheless, from these data one cannot exclude the formal possibility that Rce1p is a limiting noncatalytic subunit of a CAAX protease.

Yeast contain two Ras genes that are highly homologous to and functionally interchangeable with the Ras oncogenes of humans. The Ras genes of yeast are required for viability (19). Similarly, the role of Ras farnesylation is equally important in humans and yeast. We explored what role, if any, removal of the COOH-terminal amino acids from farnesylated Ras played in the function of the Ras protein by testing the phenotypes of Ras mutants in afc1 and rce1 mutant strains.

Like oncogenic alleles of human Ras, mutant alleles of yeast RAS2 that block GTP hydrolysis cause extreme sensitivity to heat shock and to starvation (20). The heat-shock sensitivity of cells containing the RAS2val19 activated allele and null alleles of RCE1 and AFC1 was used to determine whether removal of the COOH-terminal tripeptide from farnesylated Ras2p affected the activity of Ras. The rce1Δ mutation suppressed the heat-shock sensitivity caused by activated Ras2p by a factor of about 100 (Fig. 4). Thus, a loss in CAAX proteolysis caused by rce1Δ reduced but did not eliminate Ras function. In contrast, the afc1Δ mutation caused only a modest effect on heat-shock sensitivity, and the afc1Δ rce1Δ double mutant was indistinguishable from the rce1 single mutant.

Fig. 4.

Heat-shock sensitivity assay of Ras2p function. Yeast cells (12) containing the activated form of RAS2 (RAS2val19, pJR1040) were grown for 45 hours at 30°C in YM medium plus 2% glucose and 2% casamino acids (Difco). The cultures were heat-shocked for 90 s at 55°C in a PCR thermocycler (PTC100, MJ Research, Watertown, Massachusetts), and plated on selective medium for viability. Values are presented as the percent of viable cells relative to the culture before heat shock and are the average of duplicate cultures from two independent transformants. The error bars indicate the standard deviation. Strains carrying a plasmid with the wild-type RAS2 gene (pJR1039) showed no decrease in viability in an analogous experiment.

In an independent test of the role of CAAX proteolysis, we evaluated whether loss of CAAX proteolysis would affect RAS2 mutations that reduce Ras function in cells. The ras2-23 allele is a partial loss-of-function allele that causes temperature-sensitive growth (21). The mutant grows well at 30°C, less well at 34°C, and cannot grow at all at 37°C. Deletion of RCE1 rendered ras2-23 mutants unable to grow at 34°C (Fig. 5). As in the previous assay, the afc1 mutation had no detectable effect. Thus, CAAX proteolysis, mediated by Rce1p, plays a modulatory role on Ras2p, decreasing the function of Ras but not eliminating it.

Fig. 5.

Effect of the deletions of RCE1 and AFC1 genes on the temperature sensitivity of cells with the ras2-23 allele (21). All strains are ras1Δ and have the ras2-23 temperature-sensitive allele (JRY5318) combined with the following deletion alleles: afc1Δ::HIS3 (JRY5319), rce1Δ::TRP1 (JRY5320), and afc1Δ::HIS3 rce1Δ::TRP1 (JRY5321). Strains were grown for 3 days on rich yeast medium plates at 30°C and 34°C.

Ras proteins are localized to the inner surface of the plasma membrane where they interface between signals from outside cells and the various response mechanisms inside cells. Agents or mutations that block prenylation of Ras block its membrane association and thereby block the function of Ras (22). Prenylation of other GTP-binding proteins also appears to be essential for their proper membrane localization and consequently their function (23). Therefore, we examined whether mutations in genes encoding CAAX prenyl proteases affected the subcellular localization of their substrates.

The green fluorescent protein (GFP) is a versatile protein module that can be fused to other proteins, enabling their localization by fluorescence microscopy (23). We used a fully bifunctional GFP-RAS2 fusion gene to monitor the localization of Ras2p in wild-type and mutant cells. In this fusion protein, the GFP sequences are fused to the NH2-terminus of the Ras protein, leaving the CAAX motif intact and available for posttranslational processing (13). In wild-type cells, the GFP-Ras2p was localized to the periphery of cells reflecting its association with the plasma membrane (Fig. 6), and cellular fractionation experiments confirmed that the fusion protein sediments with the P100 membrane fraction (18). There was still some GFP-Ras2p at the plasma membrane of the rce1Δ mutant (Fig. 6). However, a significant portion of the fusion protein was evident at internal locations, with the fluorescence outlining what appeared to be internal membranes. Thus, proteolytic processing of the CAAX motif of Ras2p contributed to the proper subcellular localization of Ras.

Fig. 6.

In vivo fluorescence of GFP-Ras2p fusions. Wild-type (JRY5314) and afc1Δ::HIS3; rce1Δ::TRP1 double mutant (JRY5317) strains carried a plasmid containing RAS2 coding sequence fused in frame to GFP under the control of the constitutive promoter (13). Cells were grown overnight in selective medium, and GFP-Ras2p localization was examined directly with a Zeiss Axioskop microscope at a magnification of ×100. The fields were captured by a charge-coupled device camera with NorthernExposure software (Phase3 Imaging Systems).

Our results provided most of the missing pieces in understanding the trio of posttranslational processing steps shared by prenylated proteins. Afc1p is an integral membrane prenyl protein protease with a metalloprotease sequence motif essential for its activity. Afc1p was responsible for some, but not all, processing of a-factor precursor in vivo, and in vitro it contributed about 35% of the prenyl-protein protease activity detectable with the peptide substrates used here. Rce1p also appears to be an integral membrane prenyl protein protease and to be responsible for processing the remaining a-factor precursor on the basis of the complete absence of a-factor processing in the double mutant. In vitro, Rce1p accounted for about 60% of the detectable activity. Rce1p and Afc1p are adequate to account for all a-factor proteolysis in vivo and all detectable prenyl protein protease activity in vitro. Thus, there may be only two prenyl protein proteases in yeast. However, the absence of any sequence similarity between these two proteins and the possibility of substrate specificity tempers this conclusion. There appears to be some difference in the substrate specificity of Afc1p and Rce1p. Both proteins contribute to the processing of a-factor precursor. However, Ras2p processing required Rce1p but did not require Afc1p.

The farnesyl transferases from all organisms examined are soluble cytoplasmic enzymes (25). In contrast, the carboxyl-methyltransferase Ste14p is membrane-associated and is likely to be an integral membrane protein (26). The sequences of Afc1p and Rce1p reported here indicate that they both are integral membrane proteins, consistent with the fractionation data. Thus, it appears that prenylation of precursor proteins leads to their association with a membrane where they are subsequently trimmed of their AAX extensions and methylated.

Previously, it was assumed that prenylation led to membrane association of the protein at the plasma membrane where the last two processing steps occur. However, our data suggest a different hypothesis with interesting consequences. In the rce1Δ mutant cells, a portion of the GFP-Ras2p protein was associated with internal cellular membranes, indicating either that this fraction of the GFP-Ras2p was mislocalized, or that Ras2p associates first with internal membranes and then is transported to the plasma membrane. At present the second alternative seems the more likely one on the basis of the experiments indicating that the methyl-transferase of mammalian cells is in the endoplasmic reticulum-microsome fraction and that the microsome fraction contains a high-affinity binding site for prenylated peptides (27).

If prenylated Ras2p initially associates with internal membranes, then there must be some mechanism for moving it to the plasma membrane. To date there is no evidence that, once processed, Ras2p leaves the membrane fraction, although a short-lived, soluble processed form cannot be formally excluded. Thus, we are led to consider a model in which Ras2p may be transported from internal membranes to the plasma membrane by membrane transport vesicles. If this model is correct, then unlike secreted proteins, which are packaged in the lumen of transport vesicles, Ras2p would have to be associated with the external, cytoplasmic face of transport vesicles.

At face value, this model would seem to conflict with published reports that secretion of a-factor in yeast is independent of the secretory pathway (28). In these experiments a-factor secretion was not blocked upon a shift to the restrictive temperature in temperature-sensitive mutants that had mutations of the secretory pathway. The a-factor precursor undergoes the same COOH-terminal processing steps as Ras2p, and one would expect that both proteins reach the plasma membrane the same way, with a-factor subsequently released by the Ste6p transporter protein. However, because the Ste6p transporter itself requires the secretory pathway to reach the plasma membrane, there may be technical issues with the published work that merit reevaluation.

Because mutations in prenyl-transferases suppress the activated or transformed phenotype caused by mutant Ras proteins, a substantial effort has gone into developing inhibitors of prenyl-transferases as potential anticancer compounds (6). Some of these have proven effective in mice, but their value in humans has not been established (29).

The Rce1p family of prenyl protein proteases offer a potentially better target for therapeutic intervention than the prenyl-transferases. Like the prenyl-transferases, blocking of Rce1p function reduces but does not eliminate Ras function. Moreover, null mutations in either RCE1 or AFC1 or in both genes cause no obvious growth or viability defects, whereas mutations in prenyl-transferases cause cells to be either growth defective or dead, depending on the mutation. Thus, if these results in yeast translate to human cells, inhibitors of CAAX proteases may be more valuable therapeutic agents than inhibitors of prenyl-transferases.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
View Abstract

Navigate This Article