Protein prenyltransferases.

Prenylation is a type of lipid modification involving covalent addition of either farnesyl (15-carbon) or more commonly geranylgeranyl (20-carbon) isoprenoids via thioether linkages to cysteine residues at or near the C terminus of intracellular proteins. The attached lipid is required for proper function of the modified protein, either as a mediator of membrane association or a determinant for specific protein-protein interactions. Prenylated proteins play crucial roles in such vital cellular processes as signal transduction and intracellular trafficking pathways. This review focuses primarily on the enzymology of protein prenylation; the reader is directed to several other reviews for a more detailed description of related aspects not covered here (1–7). The enzymes responsible for isoprenoid addition to proteins have been identified and characterized at a molecular level both in mammalian systems and in lower eukaryotes. Three distinct protein prenyltransferases can be classified in two functional classes: the CAAX prenyltransferases, identified by their lipid substrate and termed protein farnesyltransferase (FTase) and protein geranylgeranyltransferase type I (GGTase-I); and the Rab geranylgeranyltransferase or protein geranylgeranyltransferase type II (GGTase-II) (Table I). FTase and GGTase-I are designated CAAX prenyltransferases since they act on proteins containing an invariant cysteine residue fourth from the C terminus in the context of a prenylation motif commonly referred to as the “CAAX box” (1, 3). Substrates for FTase include Ras GTPases, lamin B, several proteins involved in visual signal transduction, and fungal mating factors (1, 2, 5). Known targets of GGTase-I include most g subunits of heterotrimeric G proteins and Ras-related GTPases such as members of the Ras and Rac/Rho families (1, 6). GGTase-II attaches geranylgeranyl groups to two C-terminal cysteines in Ras-related GTPases of a single family, the Rab family (Ypt/Sec4 in lower eukaryotes) that terminates in Cys-Cys or Cys-X-Cys motifs (6, 8).

for enzyme recognition (9,(11)(12)(13). Both enzymes require that protein substrates contain a Cys residue fourth from the C terminus, while the C-terminal residue (i.e. the X of the CAAX motif) in general determines which of the two enzymes act on the protein. If X is Ser, Met, Ala, or Gln, the protein is processed by FTase while Leu at this positions directs modification by GGTase-I (1, 3), making it possible to predict with reasonable accuracy the type of prenyl modification on a protein by its primary sequence.
One important property of both FTase and GGTase-I is that they can recognize short peptides containing appropriate CAAX motifs as substrates (1,3). Specificity in recognition of CA 1 A 2 X sequences by these enzymes indicates that the A 1 position has a relaxed amino acid specificity, while variability at A 2 and X is more restricted (14). Moreover, substitution at the A 2 position by an aromatic residue in the context of a tetrapeptide creates peptides that are poor substrates for the enzymes but very potent competitive inhibitors (15). One such peptide, CVFM, has served as the basis for design of peptidomimetic inhibitors of FTase (see below).
The genes encoding both FTase and GGTase-I have been cloned from a number of mammalian and fungal species (2,3). The ␣ F/GGI and ␤ F subunits of mammalian FTase show about 30 and 37% identity with the proteins encoded by the Saccharomyces cerevisiae genes RAM2 and RAM1 (also known as DPR1), respectively (16 -18). These two genes were originally identified in a genetic screen of an RAS2 Val-19 , a mutationally activated RAS allele (2,4). RAM1 was also identified based on its involvement in a-factor processing and as a suppressor of G protein function (4). cDNA clones encoding ␤ GGI have been isolated from rat and human libraries (19). The cDNAs encode polypeptides of 377 residues, which share 30% identity with a yeast gene known as CDC43/CAL1. The CAL1 gene was isolated based on a Ca 2ϩ -dependent phenotype, while CDC43 was isolated based on a temperature-sensitive defect in cell polarity and also exhibits defects in localization of budding and secretion (2,4).
FTase and GGTase-I are zinc metalloenzymes, and each contains a single zinc atom required for activity (11, 20 -22). It is likely that the role of zinc is the same in both enzymes. The zinc atom is not required for isoprenoid substrate binding but is required for protein and peptide substrate binding by both FTase (20) and GGTase-I. 2 It is not yet known whether the zinc plays a structural role or whether it is directly involved in catalysis. One possibility for a catalytic role of the zinc is that the metal could activate the sulfhydryl of the substrate protein cysteine residue and make it more nucleophilic. There is evidence for such a mechanism in a DNA repair enzyme termed Ada, which catalyzes a reaction chemically similar to that of FTase (23). Metal substitution studies combined with spectroscopic analysis could provide evidence for such a mechanism in the CAAX prenyltransferases. In this regard, the zinc in GGTase-I can be replaced by Cd 2ϩ with retention of enzymatic activity (24), although the Cd 2ϩ -substituted enzyme exhibits somewhat altered specificity for substrates. 2 Neither Zn 2ϩ nor Mg 2ϩ alone restores the activity of metal-depleted FTase, but addition of both Zn 2ϩ and Mg 2ϩ fully restores activity (11,20). The dependence on millimolar levels of Mg 2ϩ for full activity indicates that this metal is probably not an integral component of FTase (11,20). Somewhat surprisingly, the activity of metal-depleted GGTase-I can be restored with Zn 2ϩ alone, in contrast to the strict Mg 2ϩ requirement for FTase activity. 2 While the mechanistic significance of this observation is not yet clear, the lack of a requirement for Mg 2ϩ exhibited by GGTase-I highlights the importance of determining the precise role(s) for the metal ions in the function of both these enzymes.
FTase and GGTase-I can bind either substrate independently (12,25). Binding of peptide substrate to FTase has been examined by NMR, revealing that the CAAX sequence of a peptide substrate adopts a Type I ␤-turn conformation when bound to the enzyme (26). A similar study conducted with a peptidomimetic inhibitor of FTase revealed a slightly different conformation most closely approximating a Type III ␤-turn (27). Binding of isoprenoid substrate by either enzyme is of such high affinity that the complex can be isolated by gel filtration. No covalent adduct is involved, however, because the isoprenoid diphosphate can be released intact upon enzyme denaturation (12,25). The binding sites for both substrates of these enzymes are thought to predominately reside on their ␤ subunits. Photoactivatable isoprenoid analogs cross-link to the ␤ subunits of both FTase and GGTase-I (24,28). Both protein and peptide substrates can be cross-linked to ␤ F (25,29), while divalent affinity-labeled short peptide substrates are cross-linked to both the ␣ and ␤ subunits of FTase upon photoactivation (29); the latter result suggests that the binding site for the peptide substrate may be near the interface of the two subunits.
Steady-state kinetic data on both enzymes initially indicated that their reactions proceed via a random sequential mechanism, an interpretation also consistent with the substrate binding studies (22,25,30). However, isotope partitioning studies have indicated that the preferred catalytic pathway is through the enzyme-isoprenoid binary complex, while the pathway through the enzyme-peptide binary complex is much slower (24,31). A more detailed kinetic study that included presteady-state analysis confirmed that, at least in the case of FTase, the random sequential mechanism is only an approximate description of a true ordered sequential path (32). Additionally, these studies established that FTase binds FPP in a two-step process to form FTase⅐FPP*, with the second step presumably involving a conformational change in the enzyme-substrate complex. FTase⅐FPP* then rapidly reacts with the peptide substrate to form a product, and product release is the rate-limiting step in catalysis (32). A scheme of the reaction catalyzed by FTase is shown in Fig. 1.
The precise chemical mechanism of the reactions catalyzed by FTase and GGTase-I is still undefined. A recent study using fluorosubstituted FPP analogs has been interpreted to indicate that the mechanism is electrophilic in nature (33), whereas analysis of the stereochemical course of the reaction using FPP with chiral deuterium-for-hydrogen substitutions at the C-1 carbon indicated that the reaction proceeds with inversion, suggesting a more nucleophilic mechanism. 3 Additional studies will be required to firmly establish the chemical mechanism.
Structure-function analyses of FTase and GGTase-I are beginning to provide some information on these enzymes. The ␤ subunits of GGTase-I and FTase share about 30% identity, with the highest homology in the central regions (19). In vitro mutagenesis has identified a mutation in the yeast ␤ F subunit (S159N) based on suppression of the cal1 (GGTase-I null) phenotype. The mutant enzyme shows an increased ability to farnesylate a GGTase-I substrate while its ability to modify a standard FTase substrate is reduced (34), suggesting that this residue may be in the peptide substrate binding site. Deletion of 51 residues from the N terminus of the rat ␣ F/GGI subunit does not affect enzyme activity, but removal of 106 residues from the N terminus or 5 residues from the C terminus of the subunit abolished FTase activity when co-expressed with ␤ F (35). Mutation of Lys-164 to Asn in ␣ F/GGI produced a polypeptide that still dimerized with ␤ F , and the resulting FTase produced retains its ability to bind substrates, but the mutant enzyme had no activity (35). These data are probably the best evidence to date that the ␣ subunit has a direct role in the catalysis by FTase.
CAAX prenyltransferases are generally quite selective for their respective protein substrates. However, cross-specificity has been observed (13,36,37), and such capacity to modify alternate substrates may be of biological significance. In this regard, yeast lacking RAM1 exhibit growth defects that can be partially suppressed by overexpression of CDC43 (the ␤ GGI subunit), suggesting that GGTase-I can at least partially modify substrates of FTase (38). CDC43 null mutants are not viable, but overexpression of two essential substrates of this enzyme, Rho1 and Cdc42, allows growth in a RAM1-dependent manner; presumably FTase prenylates these substrates of GGTase-I when they are overproduced (38,39). A specific form of mammalian Ras, K-RasB, can serve as a relatively efficient substrate for both FTase and GGTase-I (36); this dual ability is also seen for a related protein termed R-Ras2 (40). Additionally, a Ras-related GTPase implicated in organization of the actin cytoskeleton termed RhoB can be modified by either farnesyl or geranylgeranyl, and both isoprenoids can be transferred to the protein by GGTase-I (37).

Farnesyltransferase Inhibitor Studies
Development of protein prenylation inhibitors is an active area of research. The primary driving force for such efforts came from the finding that oncogenic forms of Ras proteins require farnesylation for their ability to transform cells. Inhibitors of FTase that have been synthesized or identified include analogs of both substrates (41)(42)(43)(44), fused forms of the two substrates (45), and a number of natural products and other compounds identified in screening programs (46,47). Many of these inhibitors block Ras processing and inhibit the growth of Ras-transformed cells (41). The potential for use of FTase inhibitors in cancer chemotherapy is highlighted by a recent study in which administration of an FTase inhibitor to mice bearing tumors resulting from expression of an oncogenic Ha-Ras transgene led to almost complete tumor regression without visible toxicity to the animal (48).
An open question in this field is why FTase inhibitors exhibit such low toxicity even though they can apparently completely block processing of Ras and other crucial farnesylated proteins in cells. Furthermore, the presence of oncogenic Ras in cancer cells is not an absolute predictor of whether the cells will respond to FTase inhibitors. Many types of cancer cells not containing oncogenic forms of 3   Ras respond to inhibitors, and non-responsive cells that nonetheless contain activated Ras alleles have been identified (49,50).
Possibilities to explain the current data include: (a) certain forms of Ras such as K-RasB may be resistant to FTase inhibitors (36, 51), a critical issue since K-RasB is the form of Ras most commonly found mutated in human cancers (52); (b) the aforementioned potential for GGTase-I to modify FTase substrates (36); and (c) the finding that farnesylated proteins other than Ras may be important in maintaining the transformed phenotype of cancer cells, and activities of these proteins are more critically dependent on maintenance of FTase activity (53).

Prenylation of Rab Proteins, Rab Geranylgeranyltransferase or Protein
Geranylgeranyltransferase Type-II A family of Ras-related GTP-binding proteins designated Rab in mammals and Ypt/Sec4 in yeast requires modification by geranylgeranyl isoprenoids for their action as molecular switches regulating vesicular transport in exocytic and endocytic pathways (54,55). However, the majority of Rab proteins do not contain a CAAX sequence at the C terminus but rather contain a so-called CC or CXC prenylation motif (6). An enzymatic activity that catalyzes geranylgeranyl addition to Rabs was purified from rat brain and shown to consist of two chromatographically separable components, initially designated components A and B (8,56). Structural and functional data described below led to the assignment of component B as the catalytic component, now referred to as Rab geranylgeranyltransferase or protein geranylgeranyltransferase type-II (GGTase-II), and component A as a Rab escort protein (REP).
GGTase-II from rat brain is a heterodimeric enzyme composed of a 60-kDa ␣ subunit (␣ GGII ) and a 38-kDa ␤ (␤ GGII ) subunit that are homologous to the ␣ and ␤ subunits of the CAAX prenyltransferases (57). GGTase-II shares many characteristics with the CAAX prenyltransferases. Association of both subunits of the heterodimer is required for catalytic activity (57,58). The reaction requires millimolar levels of Mg 2ϩ , but Rab prenylation is inhibited by micromolar concentrations of Zn 2ϩ for reasons that are at present unclear (8).
GGTase-II has a strict lipid and protein substrate specificity. GGTase-II binds GGPP with submicromolar affinity to form a stable complex and does not recognize FPP or geranyldiphosphate. 4 A striking difference between GGTase-II and the CAAX prenyltransferases concerns the recognition of the peptide substrate. Peptides corresponding to C-terminal amino acids of Rab proteins do not bind to GGTase-II nor compete for prenylation of full-length Rab3A (8,59). These observations are explained by the finding that the actual substrate for GGTase-II is a complex of Rab and REP. REP binds newly synthesized Rab substrates and presents them to GGTase-II so that prenyl transfer to Rabs occurs (60). GGTase-II itself does not stably bind Rabs (61), and REP is required for even one round of catalysis. 4 All Rabs tested to date are substrates for GGTase-II (58,62). While no other protein substrates have been identified, it is possible that a protein kinase in yeast is processed by this enzyme (63).
Both cysteine residues at the C terminus of Rabs are modified by geranylgeranyl groups (64). Recent studies with mutant Rab1A substrates that could only accept one prenyl group suggest that there is not a strict order of addition of the isoprenoid to the acceptor residues in either the Cys-Cys-or Cys-X-Cys-containing substrates but that the N-terminal cysteine is somewhat preferred (61). Additionally, the stable monoprenyl-Rab⅐REP complex could be isolated, suggesting that each isoprenoid transfer is an independent reaction (61).
The ␣ and ␤ subunits of GGTase-II in S. cerevisiae have been identified as the products of the BET4 (previously known as MAD2) and BET2 genes, respectively, that share 24 and 52% identity with the ␣ and ␤ subunits of the mammalian enzyme (65,66). Yeast GGTase-II requires the presence of both BET4 and BET2 gene products for activity, and mutations in either gene lead to defects in geranylgeranylation and membrane association of Ypt1 and Sec4 (65,66). A mutation termed bet2-1 has been identified that results in an enzyme with a reduced affinity for GGPP. This mutation can be suppressed by the overexpression of BTS1, which encodes a GGPP synthase (67), suggesting that the ␤ GGII subunit is directly involved in GGPP binding.

Prenylation of Rab Proteins, Rab Escort Proteins
In order to undergo prenylation, newly synthesized Rabs must bind and form a stable complex with REP, which is then recognized by GGTase-II (60). After prenylation, the modified Rab remains associated with REP, presumably because it is too hydrophobic to be released into aqueous solution. The REP⅐Rab complex is then competent for membrane delivery of prenylated Rabs (68), possibly via a membrane-bound Rab receptor. Free REP is then released and can support another round of Rab prenylation.
The REP-Rab interaction may be mediated by multiple binding sites in Rab proteins, one involving the C-terminal region and the other one or more regions of upstream Rab sequences. Rab3A has a lower V max than wild-type Rab1A, and exchanging the last 10 amino acids between the two proteins reverses their kinetics of prenylation (58). This V max effect may reflect differences in binding of the prenylated C terminus to REP, because V max appears to be largely determined by the stability of the REP⅐Rab complex (60). There is evidence for additional REP recognition sites on Rab proteins. Rab1A that contains two Ser-for-Cys substitutions at the C terminus cannot accept prenyl groups yet still binds to REP and competes for prenylation of the wild-type protein (56,61). Two putative upstream regions in Rabs have been identified. One corresponds to the L3/␤3 region in Ras (62), and another is an Nterminal region containing a conserved lysine residue (69).
Two REPs that share 75% identity have been identified in mammalian cells (58). REP-1 is the product of the choroideremia (CHM) gene on the X chromosome, and REP-2 (or CHM-like) is encoded by an intronless gene on chromosome 1 (58,70). CHM is a retinal degeneration disease, classified under the broad group of retinitis pigmentosa, characterized by slowly progressive peripheral retinal degeneration leading to complete degeneration and blindness by middle age (71). CHM is caused by deletions in REP-1 (72). The fact that patients with deletions in the ubiquitously expressed CHM gene have no other clinical abnormality other than the retinal lesion and that CHM lymphoblasts retain Rab prenylation activity suggested that REP-1 function could be partially compensated by REP-2. In agreement with this hypothesis, REP-2 is as effective as REP-1 in assisting in the geranylgeranylation of most Rabs (58). However, one Rab protein, Ram/Rab27, was found selectively unprenylated in CHM lymphoblasts, suggesting that REP-2 cannot assist effectively in its prenylation (73). These results raise the possibility that the selective dysfunction of Rab27 may lead to retinal degeneration in CHM.
A gene encoding the yeast homologue of REP has been identified as MSI4/MRS6 (74,75). MSI4/MRS6 is an essential gene required for geranylgeranylation of yeast Rabs, suggesting that there is only one REP in yeast. Loss of the MSI4/MRS6-encoded protein leads to defects in prenylation and membrane association of both Ypt1 and Sec4 (75,76).

Concluding Remarks
There has been major progress on the enzymology of protein prenylation since the isolation of the first enzyme in 1990. Efforts to design selective cell-active inhibitors of FTase in particular have been quite successful and have been greatly aided by the acquisition of mechanistic information on the enzymes. There still is much to be learned about these enzymes, however, including the question of whether additional enzymes exist (77), and it is likely that many of these secrets will yield to the enzymologist and structural biologist in the next 5 years. The ever increasing evidence that inhibitors of protein prenylation could be effective therapeutic agents in the treatment of many human cancers will continue to drive much of these efforts. Beyond the enzymology, the challenge is to understand the role of prenyl groups in mediating the function of the modified proteins, in particular to unravel the mechanism by which prenyl groups mediate protein-membrane and protein-protein interactions (78). Another important implication resulting from the research in this field was the identification of a prenylation defect as the molecular defect in one human hereditary disease, choroi-deremia. Understanding the role of REP-1 in the geranylgeranylation of Rab proteins and the pathogenesis of CHM will contribute to a better understanding of the role of prenylated Rab GTPases in vital cellular processes as well as the development of rational therapies for the disease.