Peptidomimetic Inhibitors of Ras Farnesylation and Function in Whole Cells*

The ras protooncogene is involved in regulation of cell growth. Mutations that activate the protein result in uncontrolled cell growth. Ras undergoes a series of post- translational processing events, the first of which, farnesylation, is crucial for the function of the protein. In- hibitors of the farnesyltransferase enzyme are therefore potential candidates for the development of anticancer drugs. Tetrapeptides have been reported to be good inhibitors of this enzyme in vitro. We have synthesized analogs of the tetrapeptide Cys-Val-Phe-Met by replace-ment of the amino-terminal amide bonds. One inhibitor, BS81, is permeable to the cell membrane. In the cell, it inhibits processing of two farnesylated proteins, H-ras and lamin A, but it does not inhibit processing of a ger- anylgeranylated protein, Rap lk Microiqjection of B581 into frog oocytes inhibits maturation induced by acti- vated, farnesylated H-ras but not maturation induced by activated, geranylgeranylated H-ras or by progesterone. These results demonstrate that this peptide mimic inhibits farnesylation selectively in the cell. The inhibi- tion of farnesylation results in inhibition of H-ras function. The a GTP-binding proteins that are posttranslationally modified by a pre-nyl group. Prenylation is followed by proteolytic digestion of the

* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Oncogenic ras has been linked to 50% of all cases of colon and 90% of cases of pancreas cancer (reviewed in Ref. 6). Single amino acid substitutions at position 12 or 61 are often responsible for Ras activation (i.e. its oncogenicity). By using mutants of the COOH-terminal cysteine (7) or inhibiting prenylation with inhibitors of isoprenoid biosynthesis (lovastatin) (8,91, it has been shown that the posttranslational modification of Ras by farnesyl is essential for its membrane association and transforming activity. Lack of farnesylation results in a cytoplasmic protein that has lost its growth regulatory function. Thus, inhibition of farnesylation has been postulated to be a good target for the development of anticancer drugs (10). In fact, because of the selectivity of FTase for its own substrate, it seems possible to develop a n anticancer drug with high specificity and low toxicity (10).
A number of studies have examined inhibitors of FTase i n vitro. These can be divided into two main groups: tetrapeptides with the CAAX motif (11-13) and analogs of FPP (14, 15). A third group of inhibitors with structures not resembling either tetrapeptides or FPP has been described (16)(17)(18)(19). None of the latter compounds has activities comparable to that of tetrapeptides or FPP analogs. With tetrapeptides and FPP analogs, potent and highly selective competitive inhibitors of the FTase have been described. However, both have some limitations. For example, peptides are not very membrane-permeable. In the case of FPP analogs, it is not known whether these compounds also affect other steps in cholesterol biosynthesis. FPP is a substrate not only for the farnesyltransferase that prenylates Ras, but also for other enzymes involved in the synthesis of squalene and other prenylation reactions (i.e. synthesis of ubiquinones, heme a) (10). In an effort to avoid non-selective effects related to cholesterol biosynthesis, we have developed inhibitors based on a tetrapeptide structure. This paper describes a tetrapeptide mimic, B581, based on the CVFM structure. This compound inhibits FTase i n vitro as well as Ras and lamin A farnesylation in whole cells. Furthermore, the inhibitor is specific for farnesylated versus geranylgeranylated proteins in mammalian cells and frog oocytes.

EXPERIMENTAL PROCEDURES
Proteins-Plasmids containing H-ras or H-ras(GlL,CAIL) in the bacterial expression vector PAT were obtained from Dr. C. Der (University of North Carolina). Recombinant protein was prepared from inclusion bodies by standard procedures; inclusion bodies were obtained by centrifugation after breaking bacterial cells by two cycles of freeze-thawing and clarification of the lysate by incubation in 33 pg/ml DNase. The inclusion body pellet was washed twice in 25% sucrose, 1% Triton X-100 and solubilized in 6 M guanidine hydrochloride, 1 r m EDTA, 50 rm HEPES buffer, pH 7.5, and 1 rm dithiothreitol. The soluble protein was dialyzed overnight against 200 rm NaCl, 5 rm MgC12, 25 rm HEPES buffer, pH 7.5,O.l m~ GDP, and 1 m M dithiothreitol. FTase and GGTase I were partially purified from bovine brain as described by Reiss et al. (2).

18415
Inhibitors-CVFM was prepared by standard solution phase peptide synthesis methods using suitably protected, commercially available amino acids. Reduced amide bonds were prepared via reductive amination of N-t-butoxycarbonyl amino aldehydes and the requisite amine HCl salt, using Na(CN)BH, in dimethylformamide/EtOH (21). N-t-Butoxycarbonyl amino aldehydes were prepared by diisobutyl aluminum hydride reduction of the corresponding methyl esters. All final compounds were purified by reverse phase high performance liquid chromatography on a C18 column. Stock solutions were prepared in dimethyl sulfoxide.
Preparation of Cell Lysate-A cell lysate was prepared by hypotonic shock of K-ras(l2V)-transformed NIH3T3 fibroblasts. Cells from a confluent 10-em plate were harvested by trypsinization and centrifugation. The cell pellet was resuspended in 400 pl of water and incubated on ice for 15 min, followed by homogenization with a Dounce homogenizer. The suspension was centrifuged at 300,000 x g for 20 min, and the supernatant was used as cell lysate.
Frog Oocytes-Groups of 20 Xenopus laeuis oocytes were microinjected with 50 nl of solution containing protein and vehicle (0.5% Me2SO) or protein and inhibitor. The oocytes were incubated in 96 m~ NaCl, 2 m M KCl, 1.8 m~ CaCl,, 1 m~ MgC12, and 5 m~ Hepes, pH 7.6 (ND-96) supplemented with 5% horse serum, with or without 6.4 1.1~ progesterone. Appearance of the white spot corresponding to the breakdown of the germinal vesicle (GVBD) was scored at the indicated times.

RESULTS AND DISCUSSION
The most active tetrapeptide inhibitor of farnesyltransferase reported to date corresponds to the sequence CVFM, which has an ICso = 60 n M in vitro (11). CVFM is not modified by the enzyme, apparently due to the presence of both an aromatic residue and a free amino terminus (11). Therefore, CVFM is not a substrate but a true inhibitor. These properties make it a good candidate for in vivo inhibition of farnesylation. However, in general, peptides are neither stable nor permeable enough through cell membranes to be used in uiuo. In a n effort to increase both stability and permeability, we have modified the CVFM molecule by replacing one or two NH2-terminal peptide bonds to produce compounds B515 and B581, respectively (Fig.  1). These modifications resulted in a 2-fold increase in activity as inhibitors of farnesyltransferase in comparison to CVFM ( Table I). While their selectivity against FTase is not as great as that of CVFM, these inhibitors are still more than 30-fold more active against FTase than GGTase. Since B581 has a slightly better selectivity for FTase than B515, B581 was used in all subsequent studies.
One important characteristic of a cell-based inhibitor is its stability. To test resistance to proteolysis, B581 at 10 p~ concentration was incubated in a cell lysate expected to contain many proteolytic enzymes (see "Experimental Procedures"). After incubation, the compound in the cell lysate was diluted and HCI CVFM B515 $m. evaluated as inhibitor in a farnesyltransferase assay. The results ( Fig. 2) indicate that B581 is not only stable to proteolysis, but its activity is not affected by the presence of cell components. Under the same conditions, CVFM becomes completely inactive (Fig. 2). Analysis of prenylated products by TLC indicates that B581 is not modified by either FTase or GGTase (data not shown). Thus, B581 is a true inhibitor of farnesyltransferase in vitro.
Although tetrapeptides are not very permeable through plasma membranes, the structural modifications introduced in B581 might increase its cell permeability with respect to CVFM. To determine permeability and stability, NIH3T3 fibroblasts transformed with c-H-ras(61L) were incubated for 24 h in the presence or absence of 10-500 w B581. A cell extract was analyzed by Western blotting for posttranslational protein processing. The effect of B581 was compared to the effect of the known prenylation inhibitor lovastatin. B581 inhibits H-ras processing in a dose-dependent manner (Fig. 3A). This is indicated by the appearance of a slower migrating band corresponding to non-prenylated, unprocessed H-ras. The IC5o for inhibition of processing is approximately 50 p~, and the effect is comparable to that induced by treatment with lovastatin.
The fact that the concentration necessary to inhibit processing by 50% is more than 1000-fold higher than the concentration needed to inhibit farnesylation in vitro is probably the result of low or limited permeability. However, although the in vitro assay indicated proteolytic stability (Fig. 2), other factors such as binding or compartmentalization might contribute to lower activity in the cell.
In contrast to the processing of Ras, analysis of posttranslational processing of an endogenous geranylgeranylated protein, Rap lA, showed that processing of this protein was not affected by B581 (Fig. 3B). On the other hand, treatment with lovastatin did inhibit Rap 1A processing, as indicated by the appearance of a more slowly migrating protein. These results demonstrate that the inhibitor B581 selectively inhibits farnesylation and does not affect geranylgeranylation. Since the majority of isoprenylated proteins in the cell are modified by geranylgeranylation, these results support the idea that a specific farnesylation inhibitor like B581 could be developed as a potential drug against Ras-dependent cancer.
Studies on another farnesylated protein, nuclear lamin A (231, also indicated that posttranslational processing was inhibited by B581 (Fig. 3C). The inhibition was concentrationdependent, with an ICso higher than 100 w. These results indicate that B581 is capable of blocking farnesylation of other non-Ras substrates by FTase. It has been suggested that lamin A is prenylated in the nucleus (241, which indicates that B581 is capable of permeating not only the cell membrane, but also the nuclear membrane. It is possible that the higher concentration of inhibitor needed to inhibit lamin A farnesylation (as opposed to inhibition of H-ras farnesylation) is due to the need to permeate the nuclear membrane.
Frog oocytes were used to determine whether inhibition of Ras farnesylation affects Ras function. Microinjection of activated, recombinant H-ras into X. laevis oocytes induces meiotic maturation (25). This process is characterized by the appearance of a white spot in the animal pole, corresponding to the breakdown of the germinal vesicle. As in mammalian cells, farnesylation is necessary for Ras function in oocytes: microinjection of oocytes with lovastatin or CAAX tetrapeptides inhibits GVBD (8,261. Thus, the oocytes possess the adequate posttranslational processing machinery. On the other hand, oocytes undergo maturation through at least two different pathways: Ras-dependent and Ras-independent, progesterone-mediated (27). The Ras-dependent pathway has been compared to mammalian cell transformation (28). This provides us with a system in which we can show selectivity as well as possible side effects of the inhibitor.
Oocytes were microinjected with 74 ng of H-ras(61L). The appearance of GVBD was observed 3-4 h after microinjection, with 50% of the oocytes showing GVBD in 4.5 h (Fig. 4A, closed  circles). When H-ras(61L) and B581 were coinjected, the rate of appearance of GVBD was decreased in a dose-dependent manner: 2.5 or 5 pmol of B581 delayed the appearance of GVBD by about 3 h (Fig. 4A, crosses and triangles). In the presence of 25 pmol of B581, maturation was not apparent until 7 h after microinjection, i.e. 4 h later than in the absence of inhibitor (Fig. 4A, open circles). Control, mock-injected oocytes did not undergo maturation, and oocytes injected with B581 alone had the same appearance as mock-injected oocytes. To demonstrate that no other functions are affected, oocytes injected with B581 or vehicle were incubated in 2 pg/ml progesterone. The result shown in Fig. 4B indicates that B581 had no effect on progesterone-induced oocyte maturation.
To test the selectivity of the inhibitor, activated H-ras with the 3 COOH-terminal amino acids mutated to Ala-Ile-Leu was injected into oocytes. In this case, the protein is expected to be modified by GG instead of farnesyl. Oocytes injected with H-ras(GlL,CAIL) undergo maturation in 4 h (Fig. 4 0 , similar to oocytes injected with H-ras(61L). However, in this case coinjection of Ras with B581 did not result in inhibition of the oocyte maturation process. Inhibition of geranylgeranylation A processes associated with oocyte maturation are affected.
In summary, the data presented in this paper demonstrate that peptidomimetic inhibitors of farnesyltransferase, such as hibitors should provide tools to answer questions regarding the role of prenylated proteins in cell function. Important  inhibits maturation induced by H-~~s(G~L,CAIL).~ The lack of inhibition of GG-dependent oocyte maturation by B581 is in agreement with the data presented above (Fig. 3B). In other words, B581 does not inhibit geranylgeranylation in mammalian cells or in oocytes. These results demonstrate that the effect of B581 is specific for farnesylation and that no other