Selective inhibition of farnesyl-protein transferase blocks ras processing in vivo.

The ras oncogene product, Ras, is synthesized in vivo as a precursor protein that requires post-translational processing to become biologically active and to be capable of transforming mammalian cells. Farnesylation appears to be a critical modification of Ras, and thus inhibitors of the farnesyl-protein transferase (FPTase) that catalyzes this reaction may block ras-dependent tumorigenesis. Three structural classes of FPTase inhibitors were identified: (alpha-hydroxyfarnesyl)phosphonic acid, chaetomellic acids, and zaragozic acids. By comparison, these compounds were weaker inhibitors of geranylgeranyl-protein transferases. Each of these inhibitors was competitive with respect to farnesyl diphosphate in the FPTase reaction. All compounds were assayed for inhibition of Ras processing in Ha-ras-transformed NIH3T3 fibroblasts. Ras processing was inhibited by 1 microM (alpha-hydroxyfarnesyl)phosphonic acid. Neither chaetomellic acid nor zaragozic acid were active in this assay. These results are the first demonstration that a small organic chemical selected for inhibition of FPTase can inhibit Ras processing in vivo.

The ras oncogene is found mutated in approximately 25% of human cancers (1, 2). Many laboratories have focused on the biochemistry and biology of ras-induced cellular transformation in anticipation that this information might prove useful in the identification of novel anticancer therapeutics (1,2). One area of research that has recently attracted attention in this regard is the post-translational modifications of the ras gene product, Ras. These modifications are required for appropriate subcellular localization of Ras in the plasma membrane and for Ras to exhibit cell-transforming activity (3)(4)(5). The first and obligatory step in post-translational modification of the Ras precursor is farnesylation on the thiol group of the Cys residue located at the Ras COOH terminus. This Cys residue forms part of the prenylation recognition sequence, CAAX,' found in many mammalian proteins. Transfer of a farnesyl moiety from farnesyl diphosphate to * 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 U.S.C. Section 1734 solely to indicate this fact.
the Ras CAAX peptide is catalyzed by farnesyl-protein transferase (FPTase,. Several genetic studies have shown that inhibition of Ras farnesylation prevents Ras membrane localization and blocks Ras-induced cell transformation. For example, mutation of the Cys of CAAX to Ser renders Ras transformation-defective (9). Similarly, in the yeast Saccharomyces cereuisiae, mutation or disruption of the ram1 locus (which encodes the ( I subunit of FPTase) blocks the phenotypes normally observed in yeast with an activated [Val-lSIRASZ allele (10)(11)(12)(13)(14). These observations suggest that pharmacological inhibition of FPTase may also block Rasdependent biological properties in mammalian cells (3)(4)(5). By contrast, subsequent modifications of farnesylated Ras including removal of AAX residues by proteolytic cleavage, methyl esterification of the new carboxyl terminus, and palmitoylation apparently are not obligatory for cell-transforming activity (15)(16)(17).
Following the identification of FPTase, it was quickly recognized that the substrates utilized by this enzyme, farnesyl diphosphate and the CAAX peptides, could serve as models for designing selective FPTase inhibitors (6, 7). Farnesyl diphosphate preferentially serves as substrate for FPTase versus other mammalian prenyl-protein transferases including geranylgeranyl-protein transferase type I (GGPTase-I) and . Similarly, some CAAXsequences preferentially serve as substrates for FPTase while others function best as substrates of GGPTase-I or . Although the Cys residue appears to be essential for binding of CAAX tetrapeptides to FPTase (6, 7 ) , various amino acid changes can be introduced into the AAX positions which improve potency and/or confer nonsubstrate properties to the peptides (19,(26)(27)(28)(29). Most importantly, the COOH-terminal amino acid of CAAX tetrapeptides determines which protein-prenyl transferase will bind the peptide. FPTase prefers CAAX sequences ending in Ser or Met, whereas GGPTase-I prefers CAAX sequences ending in Leu and GGPTase-I1 or Rab-GGPTase modifies proteins containing XXCC or XCXC motifs. Since most mammalian proteins are geranylgeranylated rather than farnesylated (30, 31), selective inhibitors of FPTase should exhibit fewer toxic side effects than nonselective inhibitors of both FPTase and GGPTases. Here, we report the identification of potent and selective inhibitors of FPTase in vitro and show that one FPTase inhibitor is capable of blocking Ras processing in vivo.

MATERIALS AND METHODS
Inhibitors-Chaetomellic acids A and B were isolated from a fermentation extract of the Coelomycete Chaetomella acutiseta. The inhibitory activity was extracted with methyl ethyl ketone and subjected to chromatographic separation. Details of the fermentation, isolation, and structural elucidation of chaetomellic acids A and B will be described elsewhere.' Zaragozic acid A (32, 33) was provided by K. tions of FPTase, GGPTase-I, and GGPTase-I1 were isolated from bovine brain as described previously (19). Homogeneous bovine brain FPTase was obtained as detailed by Pompliano et al. (28). The cDNAs encoding the a and @ subunits of human FPTase were co-expressed in Escherichia coli, and the recombinant protein was purified to homogeneity.' Recombinant human FPTase is kinetically identical to FPTase purified from bovine brain. Ras-CAAX proteins were expressed in E. coli and purified as described previously ( Radiolabeling conditions were the same as those described by DeClue et al. (34). Inhibitors were prepared as concentrated solutions in 100% dimethyl sulfoxide and then were diluted 1000-fold into the culture media. Cells were lysed with buffer containing 20 mM NaHepes, pH 7.5, 1% Nonidet P-40 detergent, 5 mM MgC12, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 10 pg/ml aprotinin, 2 rg/ml leupeptin, and 2 pg/ml antipain. Ras protein was immunoprecipitated from cell lysates (15-25 x 10' acid-insoluble cpm/sample) as described (34) using a complex of protein A-Sepharose CL-4B (Pharmacia LKB Biotechnology Inc.), rabbit anti-rat immunoglobulin (Cappel), and rat monoclonal anti-Ras antibody Y13-259. The immunoprecipitates were analyzed by SDS-PAGE using 13% acrylamide gels. Radiolabeled Ras protein was visualized by autoradiography after intensification with Enlightning (Du Pont-New England Nuclear).

RESULTS AND DISCUSSION
Potency and Selectivity of FPTuse Inhibitors-To identify novel inhibitors of FPTase, we evaluated natural products produced by fermentation of microorganisms. Inhibition of FPTase was observed in a fermentation extract from the Coelomycete C. acutiseta. Two novel compounds were subsequently isolated and were named chaetomellic acids A and B.
The structures of these compounds are shown in Fig. 1. The anhydride of chaetomellic acid B was previously reported although no biological activity was ascribed to it (35). Another FPTase inhibitor was isolated from a different microorganism and shown to be zaragozic acid A (Fig. 1). This compound has been described previously as an inhibitor of squalene synthase (Ki of 78 pM), which is an enzyme that also utilizes farnesyl diphosphate (32,33). We characterized these new inhibitors of FPTase and also analyzed a semi-synthetic analog of zaragozic acid A and a previously described FPTase inhibitor, (a-hydroxyfarnesy1)phosphonic acid ( Fig. 1; Ref. 28).
These compounds were evaluated for inhibition of FPTase, GGPTase-I, and GGPTase-11 to assess their selectivity toward FPTase (Table I). The chaetomellic acids and (a-hy-droxyfarnesy1)phosphonic acid demonstrated the greatest selectivity for FPTase versus the GGPTases, exhibiting up to approximately 2200-fold preference for FPTase uersus GGPTase-11. The zaragozic acids exhibited less selectivity as inhibitors of FPTase versus GGPTase-I (2.8-142-fold), but were highly selective as inhibitors of FPTase versus GGPTase-I1 (305-1400-fold). Interestingly, substitution of a single side chain in zaragozic acid A yielded a zaragozic acid A analog with an l&fold improvement in potency toward FPTase and a 2.7-fold reduction in potency toward GGPTase-I. This result suggests that additional gains in potency and selectivity of these compounds may be possible through other structural changes. Selective inhibition of FPTase versus GGPTase-I has also been reported for CAAX-based tetrapeptides and farnesyl diphosphate (19); these values are shown in Table I for comparison.
Mechanism of FPTuse Inhibition-We previously showed that (a-hydroxyfarnesy1)phosphonic acid inhibited FPTase by competition with the farnesyl diphosphate substrate (28).
To determine whether the natural products identified here competed with farnesyl diphosphate or CAAX substrates, or both, a more detailed mechanistic analysis was performed. The more potent compounds from each class, chaetomellic acid A and the zaragozic acid A analog, were examined. As shown in Fig. 2, both chaetomellic acid A and the zaragozic acid A analog were competitive inhibitors with respect to farnesyl diphosphate and noncompetitive inhibitors with respect to Ras. The Ki values fit to these data are shown in Table 11. These results indicate that the chaetomellic and zaragozic acids analyzed here, although structurally distinct from one another and from farnesyl diphosphate, nonetheless inhibit FPTase by competing with farnesyl diphosphate. Presumably, the carboxylic acid groups and extended hydrophobic chains within these compounds behave as substitutes for the phosphates and isoprene moieties of farnesyl diphosphate.
Activity of FPTuse Inhibitors in Cells-Chaetomellic acid A, the zaragozic acids, and (a-hydroxyfarnesy1)phosphonic acid were further evaluated as inhibitors of Ras processing in rus-transformed NIH3T3 fibroblast cells. This cell line expresses viral Ha-Ras, a transforming Ras mutant that undergoes autophosphorylation due to a Thr for Ala substitution   TABLE I Pharmacological inhibition of prenyl-protein transferases The indicated compounds were incubated at various concentrations in reactions having prenyl-protein transferases partially purified from bovine brain with substrates at near-Km concentrations under conditions described under "Materials and Methods." Data are the average of 2-5 determinations. ND, not determined. GGPP, geranylger-any1 diphosphate. The peptide sequence is indicated using one letter amino acid abbreviations.  nM zaragozic acid A analog. FPP concentration was held constant at 20 nM. Assays were performed as described previously (28). In A and B, assays were done with recombinant human FPTase purified to homogeneity from E. coli (see footnote 4). Similar results were obtained with bovine FPTase (not shown). Assays in C and D were done with homogeneous FPTase purified from bovine brain (28).

TABLE I1 Inhibition patterm and constants for inhibitors of FPTase
The Ki value is derived from a nonlinear least squares fit of the initial velocity data to either a purely competitive or purely noncompetitive inhibition model. Data for chaetomellic acid A and zaragozic acid A analog are from Fig. 2 appears as a doublet with the phosphorylated form migrating more slowly than the nonphosphorylated form (see Fig. 3, lanes 1 and 8). Approximately 25% of viral Ha-Ras is phosphorylated in these cells (36). This banding pattern is useful because it generates an internal standard for assessing relative band mobilities and intensities. In the processing assay, the mobility of Ras on SDS-PAGE is a function of farnesylation with processed Ras migrating more rapidly than unprocessed Ras. Lovastatin, an inhibitor of the rate-limiting step in the isoprenoid biosynthetic pathway, has previously been shown to block Ras processing in vivo and was used here as a positive control (Fig. 3, lanes 2 and 9 ) (17, 37,38). Chaetomellic acid A and the zaragozic acids did not inhibit Ras processing when tested up to 100 ~L M for 24 h (Fig. 3, lanes 3-5). Shorter incubations of 4 h with these compounds in serum-free medium also did not reveal any inhibitory activity (not shown).
T o confirm that the Ras mobility shifts observed in the compound-treated cells were due to inhibition of Ras processing, we demonstrated that the bands assigned as unprocessed Ras were soluble upon cell fractionation (not shown) and were not associated with the cell membrane fractions as is normally the case with farnesylated Ras (17,38). We also attempted but were unable to radiolabel the unprocessed Ras in the compound-treated cells with [3H]mevalonic acid.
[3H]Mevalonic acid is converted to [3H]farnesyl diphosphate in vivo and is then utilized by FPTase to farnesylate Ras under normal conditions (17,38). As an additional control, we evaluated (a-hydroxyfarnesy1)phosphonic acid in RAT1 cells expressing a form of viral Ha-ras that has a mutation in the CAAX sequence (Ras-CVLS mutated to Ras-CVLL). This substitution changes Ras from a substrate for FPTase to a substrate for GGPTase-I. (a-Hydroxyfarnesy1)phosphonic acid at 100 p~ did not inhibit the processing of Ras-CVLL, whereas 15 p~ lovastatin effectively inhibited the processing of this geranylgeranylated protein (not shown). These results are consistent with our interpretation that the inhibition of Ras-CVLS processing by (a-hydroxyfarnesy1)phosphonic acid of FPTase Blocks Ras Processing in Vivo is due to inhibition of Ras farnesylation. The inhibition of Ras processing observed with 100 p~ (a-hydroxyfarnesy1)phosphonic acid was not as extensive as that observed with 15 p~ lovastatin (Fig. 3). We were unable to determine whether (a-hydroxyfarnesy1)phosphonic acid could produce further inhibition of Ras processing if used at higher concentrations because of cellular toxicity observed at compound levels >lo0 PM. This toxicity was evidenced by inhibition of viable staining with 3-(4,5-dimethylthiazol-2-y1)-2,5diphenyltetrazolium bromide (39) and may be due to nonspecific effects such as detergent-like action of (a-hydroxyfar-nesy1)phosphonic acid or due to inhibition of other enzymes that are required for cell viability. For example, (a-hydroxy-farnesy1)phosphonic acid also inhibits squalene synthase activity in vitro with an ICso of 630 nM? Chaetomellic acid A and the zaragozic acids are highly charged compounds, and they did not appear to inhibit Ras processing in vivo. These compounds may require masking of the charged groups via a prodrug strategy in order to promote better cell penetration and thereby enhance their ability to inhibit Ras processing in vivo.
The results with (a-hydroxyfarnesy1)phosphonic acid demonstrate that a compound selected for inhibition of FPTase can block Ras processing in vivo. Our work has also identified novel natural products and a semi-synthetic derivative of zaragozic acid A which potently inhibit FPTase. One other natural product, 10'-desmethoxystreptonigrin, has also been reported to be an inhibitor (ICm = 21 p~) of FPTase (40).
The ultimate utility of these compounds will depend upon their selectivity for inhibition of FPTase versw other enzymes and their ability to penetrate and function in mammalian cells. The availability of more potent compounds or derivatives of these compounds with greater cell penetration properties should allow biological evaluation of farnesylation inhibitors as antagonists of cellular transformation by the ras oncogene.