Structural homology among mammalian and Saccharomyces cerevisiae isoprenyl-protein transferases.

Farnesyl-protein transferase (FTase) purified from rat or bovine brain is an alpha/beta heterodimer, comprised of subunits having relative molecular masses of approximately 47 (alpha) and 45 kDa (beta). In the yeast Saccharomyces cerevisiae, two unlinked genes, RAM1/DPR1 (RAM1) and RAM2, are required for FTase activity. To explore the relationship between the mammalian and yeast enzymes, we initiated cloning and immunological analyses. cDNA clones encoding the 329-amino acid COOH-terminal domain of bovine FTase alpha-subunit were isolated. Comparison of the amino acid sequences deduced from the alpha-subunit cDNA and the RAM2 gene revealed 30% identity and 58% similarity, suggesting that the RAM2 gene product encodes a subunit for the yeast FTase analogous to the bovine FTase alpha-subunit. Antisera raised against the RAM1 gene product reacted specifically with the beta-subunit of bovine FTase, suggesting that the RAM1 gene product is analogous to the bovine FTase beta-subunit. Whereas a ram1 mutation specifically inhibits FTase, mutations in the CDC43 and BET2 genes, both of which are homologous to RAM1, specifically inhibit geranylgeranyl-protein transferase (GGTase) type I and GGTase-II, respectively. In contrast, a ram2 mutation impairs both FTase and GGTase-I, but has little effect on GGTase-II. Antisera that specifically recognized the bovine FTase alpha-subunit precipitated both bovine FTase and GGTase-I activity, but not GGTase-II activity. Together, these results indicate that for both yeast and mammalian cells, FTase, GGTase-I, and GGTase-II are comprised of different but homologous beta-subunits and that the alpha-subunits of FTase and GGTase-I share common features not shared by GGTase-II.

Farnesyl-protein transferase (FTase) purified from rat or bovine brain is an a/#? heterodimer, comprised of subunits having relative molecular masses of approximately 47 (a) and 46 kDa (#?). In the yeast Saccharomyces cerevisiae, two unlinked genes, RAMlIDPRl ( R A M I ) and RAM2, are required for FTase activity. To explore the relationship between the mammalian and yeast enzymes, we initiated cloning and immunological analyses. cDNA clones encoding the 329-amino acid COOH-terminal domain of bovine FTase a-subunit were isolated. Comparison of the amino acid sequences deduced from the a-subunit cDNA and the RAM2 gene revealed 30% identity and 58% similarity, suggesting that the RAM2 gene product encodes a subunit for the yeast FTase analogous to the bovine FTase a-subunit. Antisera raised against the RAMl gene product reacted specifically with the #?-subunit of bovine FTase, suggesting that the RAMl gene product is analogous to the bovine FTase #?-subunit. Whereas a raml mutation specifically inhibits FTase, mutations in the CDC43 and BET2 genes, both of which are homologous to R A M l , specifically inhibit geranylgeranyl-protein transferase (GGTase) type I and GGTase-11, respectively. In contrast, a ram2 mutation impairs both FTase and GGTase-I, but has little effect on GGTase-11. Antisera that specifically recognized the bovine FTase a-subunit precipitated both bovine FTase and GGTase-I activity, but not GGTase-I1 activity. Together, these results indicate that for both yeast and mammalian cells, FTase, GGTase-I, and GGTase-I1 are comprised of different but homologous #?-subunits and that the a-subunits of FTase and GGTase-I share common features not shared by GGTase-11.
Site-specific farnesylation or geranylgeranylation of cellular polypeptides a t a COOH-terminal cysteine residue is a functionally essential post-translational modification (see Ref. 1). Protein acceptor substrates in mammalian cells for farnesylation include the cell-transforming 20-kDa GTPase Ras, nuclear lamin B, and the y-subunit of retinal transducin.
Substrates for geranylgeranylation include some 20-kDa * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  GTPase proteins and the y-subunit of some heterotrimeric GTP-binding proteins (1,2). Three different enzymes in mammalian cells that catalyze the isoprenylation of these proteins have been identified and chromatographically resolved a farnesyl-protein transferase (FTase,' see Ref. 1) and two geranylgeranyl-protein transferases (GGTase-I and GGTase-11, Refs. 3-6). These enzymes each have different specificities for the protein substrate and the isoprenoid diphosphate utilized in the catalytic reaction. FTase preferentially farnesylates proteins having the COOHterminal sequence CauX, where a is usually an aliphatic amino acid and X is Ala, Ser, Gln, Cys, or Met (3)(4)(5)(6). GGTase-I geranylgeranylates proteins having the COOH-terminal sequence CaaX where X is Leu (3)(4)(5)(6). For FTase and GGTase-I, the major determinants for interaction between the enzyme and protein substrate reside within the CaaX sequence as demonstrated by selective isoprenylation of tetrapeptides (6). GGTase-I1 modifies proteins having the COOH-terminal sequence Gly-Gly-Cys-Cys (6). The interaction between GGTase-I1 and its substrates is complex as peptides of the modification site neither compete nor serve as substrates for geranylgeranylation.
FTase has been purified to homogeneity from rat brain by Reiss et al. (7). The enzyme is a heterodimer of approximately 100 kDa as determined by gel filtration chromatography on Superdex-75 (4). The a-subunit of approximately 49 kDa is immunologically distinct from the @-subunit of 46 kDa (4, 8). The @-subunit functionally binds the protein substrate as determined in cross-linking experiments (8). FTase and GGTase-I appear to be related because antisera raised against peptides from the FTase a-subunit recognize a similar molecular size polypeptide in GGTase-I and immunodeplete both activities (4).

Structural Homology among
Isoprenyl-protein Transferases 18885 30% amino acid sequence identity with each other and with another gene product Bet2/0rf2 (12)(13)(14). BET2 was described recently as a gene essential for membrane association of the secretory GTPase proteins Yptl and S e d having the COOHterminal sequence Gly-Gly-Cys-Cys, and the orf2-1 mutant was temperature-sensitive for this phenotype (13). In this paper, we show that Ram2 and Ram1 are homologous to the bovine FTase a and @-subunits, respectively. Mutations in the RAM1 homologues CDC43 and BET2 specifically impair GGTase-I and GGTase-I1 activity, respectively. These data suggest a general structure for isoprenyl-protein transferases as a/@ heterodimers consisting of a-subunits, which in some cases may be similar and distinct but homologous p-subunits.

MATERIALS AND METHODS
Amino Acid Sequence Determination-FTase was purified from bovine brain, and enzyme activity was assayed as described previously (6). The concentration of purified protein was estimated by silver staining (Enprotech, Hyde Park, MA) of an SDS-polyacrylamide gel in which known amounts of ovalbumin were used as a reference standard. Purified protein was dialyzed against water and approximately 9 pg was reduced and carboxymethylated essentially as described (15) and repurified on a 2 X 50-mm reversed-phase HPLC C, column (Separations Group) using a Pharmacia LKB microbore Smart System. Protein was eluted with a 30-min linear gradient of 0-67% acetonitrile in 10 mM trifluoroacetic acid at a flow rate of 0.1 ml/min collecting individual peaks identified by absorbance at 214 nm. Reduced and carboxymethylated protein (7 pg) was digested with endoproteinase Lys-C (Boehringer Mannheim) at a 201 mass ratio of substrate to enzyme for 16 h at 35 "C in 100 pl of 25 mM Tris-C1, pH 8.5, 1.0 mM EDTA. Polypeptide digestion products were loaded onto a 2 X 100-mm CIS reversed-phase column (Pharmacia LKB Biotechnology Inc.) equilibrated in a mixture of 98% solvent A (7.8 mM trifluoroacetic acid) and 2% solvent B (7.8 mM trifluoroacetic acid, 80% acetonitrile (v/v)) and eluted at a flow rate of 0.1 ml/min by sequential linear gradients of solvent B from 2 to 37% over 45 min, 37-75% over 30 min, and 75%-98% over 10 min using the microbore Smart System. Peaks were individually collected monitordescribed (16).
ing absorbance at 214 nm. Amino acid sequencing was performed as cDNA Cloning-A bovine brain oligo(dT)-primed cDNA library in cloning vector X g t l O (17) and a bovine liver 5' STRETCH cDNA library in SWAJ-2 (Clontech) were screened using the plaque hybridization method as described (18). Nitrocellulose filters containing denatured phage DNA were prehybridized and hybridized in 5 X standard saline citrate (SSC; 1 X SSC is 0.15 M NaCl, 0.015 M sodium citrate, pH 6.8), 10 X Denhart's solution (1 X Denhart's solution is 0.02% bovine serum albumin, 0.02% Ficoll, 0.02% polyvinylpyrrolidone), and 0.1% SDS at 50 "C. Oligonucleotides which correspond to the derived peptide sequences (see Fig. 1) and whose sequence was based on preferred codon usage (19) were used as probes. Overlapping complementary oligonucleotides were labeled using Klenow DNA polymerase and all four [32P]deoxynucleotide triphosphates. Filters were hybridized overnight at 50 "C, washed in 5 X SSC at 50 "C, dried, and autoradiographed (18). Inserts of positive clones were subcloned in pBluescript SK I1 (Stratagene) plasmid vector and were sequenced using the Sequenase I1 dideoxy sequencing kit (U. S. Biochemical Corp.).
Immunological Analyses-Antibody was raised against the synthetic peptide KNEIAFNYLE, which had been coupled to thyroglobulin with glutaraldehyde. A CheY-RAM1 fusion protein was constructed by digesting pUV-RAM1' with the restriction enzymes ScaI and SphI. The 1.6-kilobase fragment containing the RAMI gene beginning at the codon specifying amino acid residue 60 was ligated into a CheY fusion vector (a derivative of pJC264; Ref. 20). The insoluble CheY-RAM1 fusion protein, expressed in Escherichia coli strain RR1 lady (BRL), was purified by guanidine extraction as described (21). For both antigens, two New Zealand White rabbits were immunized with antigen in Freund's complete adjuvant. Rabbits were boosted with antigen in Freund's incomplete adjuvant at weeks A-Sepharose CL-4B (Pharmacia) followed by elution with 100 mM 5 and 7. IgG fractions were prepared by incubating sera with protein glycine, pH 3. Eluted antibody was neutralized with 200 mM Tris-C1, ' S. Powers, unpublished observation. pH 8, concentrated by centrifugation (Centricon-30, Amicon), and stored at 4 "C in phosphate-buffered saline.
For immunoblot analyses, either purified bovine brain FTase holoenzyme or resolved subunits were fractionated on 7.5% SDS-polyacrylamide gels. To separate the subunits, purified FTase from bovine brain was resolved on a 7.5% SDS-polyacrylamide gel. One portion of the gel was stained with silver and was used as a guide to separately excise the upper and lower bands of the FTase in the remainder of the gel. The separated subunits contained in gel slices were loaded directly and run on a second 7.5% SDS-polyacrylamide gel alongside an aliquot of holoenzyme and the proteins transferred to Immobilon-P (Millipore). Membranes were incubated with protein A-purified primary antibody at 4 "C overnight, washed, and incubated with horseradish peroxidase-labeled anti-rabbit antiserum (Amersham; 1:50,000 dilution). Proteins were visualized with the ECL detection system (Amersham).
For immunoprecipitations, bovine isoprenyl-protein transferases were partially purified and resolved as described (6) with an additional HPLC step using a Bio-Gel Phenyl-5-PW column (Bio-Rad). Each enzyme was separately incubated with protein A-purified antibody for 3 h at 0 "C in phosphate-buffered saline. Protein A-Sepharose CL-4B beads (100 p1 of a 50% slurry) were added for 1 h at 4 "C and subsequently washed with phosphate-buffered saline. Assay reaction mixture (70 11) was added to the packed beads and incubated for 20 min at 30 "C. The clarified supernatant was then analyzed for isoprenylated substrate (6). The following strains were a gift from James Friesen (University of Toronto): YF1593, strain YS34 carrying the plasmid pYS104 which contains a wild-type ORF2 gene; and YF1594, strain YS34 carrying the plasmid pYS135 containing a temperature-sensitive allele orj2-1 as the sole copy of the gene (see Refs. 13 and 14).

RESULTS
We have previously reported the purification of FTase from bovine brain (6). The purified bovine enzyme consists of two subunits which co-migrate with rat FTase (purified as in Ref. 7) and have apparent molecular weights of 47,100 ( a ) and 45,600 ((3) on denaturing gels; the holoenzyme is approximately 100 kDa by gel filtration on Superdex-75 (data not shown). To learn more about the structure of FTase, we initiated efforts to clone the cDNAs for both subunits of the bovine enzyme.
Following reduction and carboxymethylation, the purified holoenzyme was digested with Lys-C, and peptides were isolated by reversed-phase HPLC. The amino acid sequences obtained from analysis of the resulting peptides are shown in Fig. 1. Oligonucleotides complementary to the FTase messenger RNA sequence predicted from the peptide sequence data were synthesized and used to screen amplified bovine brain and bovine liver cDNA libraries. All of the five hybridization positive clones from the brain library and the one clone from the liver library cross-hybridized. Sequence analysis of four of the brain clones and the liver clone revealed an open reading frame of 987 base pairs, bounded on the 5' end by an in-frame initiation codon and on the 3' end by an in-frame termination codon (Fig. 1). This open reading frame would encode a 329-amino acid polypeptide in which the sequences of all but one of the peptide fragments derived from the purified enzyme were found. The predicted M , of this polypeptide is 39,502, which is less than the apparent M , observed for either of the subunits on SDS-PAGE. Although this difference might be explained by post-translational modification of the protein isolated from bovine brain, it seems likely that the clone does not contain the complete coding region. The proposed initiation codon is preceded by only 35

CCACATMTG~GTGCGTGGMTTATTTGMAGGGATTTTGCAGGATCGTGGTCTTTCC P H N E S A W N Y L K G I L Q D R G L S
""-" " _" " "   (22). The parameters for gap weight and gap length weight were 3.00 and 0.10, respectively. nificant sequence homology was observed between the bovine protein and Raml, significant homology was observed between the bovine protein and the recently cloned full-length RAM2 gene product (316 amino acids) which has a size of 38 kDa by SDS-PAGE (Fig. Z). 4 The two proteins have 30% amino acid identity and 58% amino acid similarity. As shown in Fig. 2, the homology extends over the entire Ram2 protein and nearly all of the bovine protein.

M T T T G G G G G A G G G A G A G G~C A G A A A G T C C C A T~G G M C T T T T G T A G T C T T A T C M
We used immunoblot analysis to determine whether the protein product of the cloned cDNA corresponded to the a or p-subunit of the purified bovine brain FTase. Antiserum (rabbit 793) was raised against a synthetic peptide corresponding to the sequence of one of the products of the Lys-C digestion of purified FTase and which was found in the amino acid sequence deduced from the cDNA (see Materials and Methods). Anticipating that RAMl might encode a structural polypeptide of FTase (10) which would share homology with bovine FTase, we also prepared antiserum (rabbit 114) to a bacterially-expressed CheY-Ram1 fusion protein. Fig. 3A shows immunoblots of purified bovine FTase. Antibody 793 reacted only with the a-subunit, and antibody 114 reacted only with the @-subunit. To confirm this assignment, the two subunits of the purified enzyme were resolved on an SDSpolyacrylamide gel, separately excised from the gel, and fractionated on a second SDS-polyacrylamide gel. As shown in Fig. 3B, antibody 793 specifically reacted with the isolated asubunit and antibody 114 with the isolated P-subunit. These results demonstrate that the cloned cDNA corresponds to the a-subunit and suggest that the bovine P-subunit shares homology with the RAMl gene product.
To demonstrate that the cloned cDNA encoded a polypeptide required for FTase activity, we used antibody 793 to precipitate FTase activity from a partially purified bovine enzyme preparation. In addition, to determine if this antibody would identify a common epitope in either GGTase-I or B. He, P. Chen, S.-Y. Chen  purified bovine FTase (about 400 ng) was subjected to 7.5% SDS-PAGE and transferred to Immobilon-P. The filters were incubated with antibody 793 derived from a peptide sequence within the cDNA (Anti-alpha, 1:500 dilution), antibody 114 (Anti-RAMI, 1:50 dilution), or both antibodies (Both). B, approximately 400 ng of holoenzyme (total) and 400 ng of each of the gel-isolated subunits (alpha, beta) were fractionated in separate lanes on a 7.5% SDS polyacrylamide gel and transferred to Immobilon-P. One portion of the membrane was incubated with antibody 793 (Anti-alpha) and another portion with antibody 114 (Anti-RAMI). Aberrant banding patterns in the alpha and beta lanes are due to nonideal protein electroelution from the gel slices in the loading wells. Immunoreactive proteins were visualized using the ECL detection system following incubation with horseradish peroxidase-labeled anti-rabbit antiserum. FTase subunit bands were not detectable with preimmune sera (not shown).
The migration of prestained ovalbumin (Bio-Rad) is indicated.
GGTase-11, we performed the immunoprecipitation using pools of the partially purified bovine GGTases. Following incubation with either preimmune antibody or antibody 793, protein A-agarose beads were added and the isoprenyl-protein transferase activity present in the pellet was determined. Whereas no isoprenyl-protein transferase activity was detected in pellets incubated with preimmune antibody, a significant amount of FTase and GGTase-I but not GGTase-I1 activity was present in pellets incubated with 793 (Table I). The amount of activity increased with increasing amounts of antibody 793, suggesting that the activity was due to precipitation by the antibody. This result indicates that the polypeptide encoded by the cDNA is required for FTase activity and demonstrates that the a-subunits of bovine FTase and GGTase-I share a common epitope for immunoprecipitation that is not shared by GGTase-11.
Since the cloning and immunological analyses suggested that RAM1 and RAM2 encode the subunits of S. cerevisiae FTase, we evaluated the activities of isoprenyl-protein transferases in S. cerevisiae strains having defects in the genes cdc43 and bet2 which are homologous to but distinct from RAMI. As has been shown previously, RAMI is not required for GGTase-I or GGTase-I1 activity in yeast (Table 11; Refs. 6 and 11). In contrast, a mutation in ram2 inhibited not only FTase but also GGTase-I, but did not significantly inhibit GGTase-I1 activity (Table 11; Refs. 6 and 11). Recently, the protein product of the S. cerevisiae CDC43 gene was shown to be required for GGTase-I activity in yeast (11). T o determine if CDC43 was also required for GGTase-I1 activity, we analyzed crude soluble extract from yeast having a mutation in this gene for the ability to geranylgeranylate Yptl. As

TABLE I Immunoprecipitation of bovine isoprenyl-protein transferase activities
Partially purified bovine FTase, GGTase-I or GGTase-11, completely resolved from one another, were incubated with the indicated amount of antibody and then precipitated with protein A-agarose beads. The respective isoprenyl-protein transferase activity in the pellet was determined (6) with 0.25 P M tritium-labeled isoprenoid diphosphate and 1-1.5 PM protein substrate (FTase, farnesyl diphosphate, and Ras-CVLS; GGTase-I, geranylgeranyl diphosphate, and Ras-CAIL; GGTase-11, geranylgeranyl diphosphate, and Yptl-GGCC). The data represent the average of two separate experiments. Antibody 793 was not neutralizing in solution, and the activity recovered in the pellet was proportional to the activity depleted from the supernatant (not shown). the peptides obtained from the Lys-C digest should have derived from both subunits because the two subunits of the enzyme appear to be present in equimolar amounts. However, 10 of the 11 peptide sequences obtained were from the asubunit. A possible explanation may lie in our observation that the @-subunit degrades readily. The remaining peptide derives from the p-subunit5 (24).
The deduced amino acid sequence of the bovine FTase asubunit shares significant sequence homology with the predicted protein product of the full-length s. cereuisiae RAM2 gene. This homology extends over the entire Ram2 polypeptide and over most of the cloned FTase a-subunit, suggesting that the two proteins might be functionally similar. Studies to determine whether this hypothesis is correct are underway. If this hypothesis were correct, it would indicate that the functional domain of the mammalian FTase a-subunit lies within the cloned region.
Mutation of RAM2 affects not only FTase activity but also GGTase-I activity, indicating that the two yeast enzymes share an identical a-subunit. Since the ram2 mutation completely inhibited S. cereuisiae FTase, partially inhibited GGTase-I by 67%, and did not significantly inhibit GGTase-I1 activity (Table 11), it is possible that at least one other polypeptide having functional properties of an a-subunit might be a component of S. cereuisiae GGTases. However, we cannot exclude the possibility that the partial inhibition of GGTase-I is specific for the mutant allele of ram2 tested. RAM24 and CDC43 (12), but not RAMl (Refs. 9 and lo), are essential genes in S. cereuisiae. If the yeast cell requires either FTase or GGTase-I activity for viability, it is not surprising that the isolated ram2 allele allows some GGTase-I activity.
We have shown that an antibody raised against a peptide from the bovine FTase a-subunit is able to precipitate bovine FTase and GGTase-I activity but not GGTase-I1 activity, suggesting that at least one additional a-subunit polypeptide might be similarly present in mammalian cells. These results corroborate the immunological data of Seabra et al. (4) who showed that antisera to rat FTase a-subunit cross-reacts with a similar size polypeptide in a GGTase-I preparation resolved from FTase and quantitatively immunodepletes both activities. Together, these data suggest that a common a-subunit may be a general property of eukaryotic FTase and GGTase-I. These three isoprenyl-protein transferases are all likely to be a/@ heterodimers because mammalian FTase, GGTase-I, and GGTase-I1 have a similar size by gel filtration chromatography (data not shown and Refs. 4 and 23).
We have also observed that the @-subunit of bovine FTase is immunologically similar to Raml (Fig. 3), suggesting that Raml is a functionally equivalent P-subunit for S. cereuisiae FTase. Reiss et al. (8) have shown that the @-subunit binds protein substrate. By analogy, it seems likely that Cdc43 and Bet2, which are 30% homologous to Raml, comprise the protein substrate-binding @-subunit of GGTase-I and GGTase-11, respectively. This hypothesis would be consistent with the absolute specificity that mutations in RAMl, CDC43, ' R. E. Diehl, D. D. Soderman, K. A. Thomas, J. B. Gibbs, and N. E. Kohl, unpublished results. and BET2 have for FTase, GGTase-I, and GGTase-11, respectively (Table 11). Indeed, the temperature sensitive phenotype observed with the orf2-1 allele indicates that BETZ/ ORF2 encodes a structural polypeptide of GGTase-11. These data suggest, then, that mammalian isoprenyl-protein transferases will have different but homologous p-subunits. Structure-function studies should identify the regions of these proteins which confer the high substrate specificity observed for the enzymatic reactions.