Targeted Inactivation of the Isoprenylcysteine Carboxyl Methyltransferase Gene Causes Mislocalization of K-Ras in Mammalian Cells*

After isoprenylation and endoproteolytic processing, the Ras proteins are methylated at the carboxyl-terminal isoprenylcysteine. The importance of isoprenylation for targeting of Ras proteins to the plasma membrane is well established, but the importance of carboxyl methylation, which is carried out by isoprenylcysteine carboxyl methyltransferase (Icmt), is less certain. We used gene targeting to produce homozygous Icmt knockout embryonic stem cells (Icmt−/−). Lysates fromIcmt−/− cells lacked the ability to methylate farnesyl-K-Ras4B or small-molecule Icmt substrates such asN-acetyl-S-geranylgeranyl-l-cysteine. To assess the impact of absent Icmt activity on the localization of K-Ras within cells, wild-type and Icmt−/− cells were transfected with a green fluorescent protein (GFP)-K-Ras fusion construct. As expected, virtually all of the GFP-K-Ras fusion in wild-type cells was localized along the plasma membrane. In contrast, a large fraction of the fusion in Icmt−/− cells was trapped within the cytoplasm, and fluorescence at the plasma membrane was reduced. Also, cell fractionation/Western blot studies revealed that a smaller fraction of the K-Ras in Icmt−/− cells was associated with the membranes. We conclude that carboxyl methylation of the isoprenylcysteine is important for proper K-Ras localization in mammalian cells.

Ras proteins, as well as other proteins that terminate with a so-called CaaX sequence, undergo several sequential posttranslational processing steps. First, a 15-carbon farnesyl or a 20-carbon geranylgeranyl isoprene lipid is added to the thiol group of the cysteine (C) residue by a cytosolic protein prenyltransferase (a process generally referred to as protein prenylation or isoprenylation) (1). After isoprenylation, the last three amino acids of the protein (i.e., the -aaX) are released by a specific endoprotease associated with the endoplasmic reticu-lum (ER) 1 (2). The newly exposed isoprenylcysteine is then methylated by an ER-associated methyltransferase (3). In addition, some forms of Ras undergo palmitoylation of cysteines upstream from the isoprenylcysteine residue (4).
The post-translational processing steps for the CaaX proteins have attracted considerable scrutiny, in large part because of the key role of activated forms of Ras in the pathogenesis of human cancers. The isoprenylation step is crucial for the proper targeting of the Ras proteins along the plasma membrane (1,5), and the growth of Ras-induced tumors in mice can be inhibited by pharmacological blockade of the isoprenylation step (6 -8). However, the physiologic importance of the "post-isoprenylation" protein processing steps has remained less certain (9); this is particularly the case for the carboxyl methylation step (9).
The enzyme responsible for catalyzing the formation of the methyl ester on the carboxyl-terminal isoprenylcysteine is protein-S-isoprenylcysteine O-methyltransferase (EC 2.1.1.100) (9). In Saccharomyces cerevisiae, the enzyme is encoded by STE14. The yeast enzyme (Ste14p) is predicted to have multiple transmembrane-spanning domains (10), and it resides in the ER (11). A human orthologue for STE14 was reported recently (12) following the appearance in the data bases of a mouse EST with sequence similarities to yeast STE14. Like the yeast enzyme, the human enzyme resides in the ER and is predicted to have multiple membrane-spanning domains (12). The human and mouse genes have been designated ICMT 2 and Icmt, respectively (for isoprenylcysteine carboxyl methyltransferase).
In this study, we sought to determine the influence of carboxyl methylation on the membrane localization of Ras proteins in mammalian cells. One potential means of approaching this issue would be to competitively inhibit the isoprenylcysteine carboxyl methyltransferase with small, cell-permeable, methyl-accepting isoprenylated substrates (e.g., N-acetyl-Sfarnesyl-L-cysteine or N-acetyl-S-geranylgeranyl-L-cysteine) (13)(14)(15)(16). The drawback of this approach, however, is that these types of small-molecule competitors probably have multiple effects on cells (see Ref. 9 for a recent review). For example, the fact that these molecules contain isoprenyl groups could lead them to displace isoprenylated proteins from their cellular binding sites (17). To avoid such problems, we used genetargeting techniques to produce a mammalian cell line lacking isoprenylcysteine carboxyl methyltransferase. In this report, we document the production of a mouse cell line that is homozygous for an Icmt knockout mutation, and we demonstrate that K-Ras localization is abnormal in those cells.

MATERIALS AND METHODS
A Sequence-replacement Gene-targeting Vector-A mouse EST sequence (AA022288) that shared sequence similarities with STE14 (the yeast gene for protein-S-isoprenylcysteine O-methyltransferase) (18) was used to prepare a 262-base pair Icmt cDNA probe. The cDNA probe was amplified from mouse liver cDNA (CLONTECH, Palo Alto, CA) with Taq polymerase (Roche Molecular Biochemicals) and oligonucleotides 5Ј-TCAGCCTCGCTACATTCCTCC-3Ј and 5Ј-AACAGAGACAGCG-AGCACACGTA-3Ј. The cDNA probe was used to identify a bacterial artificial chromosome (BAC) clone spanning the mouse Icmt gene (Genome Systems Inc, St. Louis, MO). A 5-kb BamHI fragment of the BAC spanning the 5Ј portion of the Icmt gene was cloned into pBSSKII (Stratagene, La Jolla, CA). A sequence-replacement gene-targeting vector was constructed in pKSloxPNT (19), which contains a 3-phosphoglycerate kinase-neomycin resistance (PGK-neo) cassette and a thymidine kinase (tk) gene, along with appropriate polylinker cloning sites. The long arm of the vector (consisting of 4.5 kb of DNA 5Ј to the translational start site in exon 1) was amplified from BAC DNA with primers 5Ј-CTCTGT-GCGGCCGCCTGTGTATAACTGTTTCCTTAGGTATG-3Ј and 5Ј-ACGA-CGGCGGCCGCAAGCTTGGCGCAGGGCGGCGGAAGAGCCGGCGG-3Ј and then cloned into the polylinker NotI site of pKSloxPNT. Next, the short arm (2.2 kb in length, spanning sequences immediately 3Ј of the exon 1-intron 1 junction to coding sequences in a downstream exon) was enzymatically amplified from BAC DNA with the primers 5Ј-ACGACGA-TCGATAAGCTTGTGCGGGCTGAGTGCGGAGGGACCGGGACC-3Ј and 5Ј-ACGACGATCGATCTTCAGTTCTGGCCAGAAGATGTTGTCGAG-3Ј and cloned into the ClaI site. After verification of the integrity of the vector with DNA sequencing and restriction endonuclease mapping, the vector was linearized with XhoI and electroporated into mouse embryonic stem (ES) cells (strain 129/SvJae) (20). Targeted clones (Icmtϩ/Ϫ) were identified by Southern blotting of BamHI-digested genomic DNA with a 293-base pair genomic DNA probe located immediately 5Ј to the sequences contained in the long arm of the vector. This 5Ј flanking probe was amplified from BAC DNA with primers 5Ј-CCTACTGCTTGAGAAAGA-GGTCC-3Ј and 5Ј-GAGTGCATATGCCCCAAGGAACAG-3Ј. The probe detects a 5.0-kb BamHI fragment in wild-type (Icmtϩ/ϩ) cells. Targeted cells (Icmtϩ/Ϫ) were identified by the presence of a 6.8-kb band in addition to the 5.0-kb band.
Production of Homozygous Icmt Knockout Cells-To select homozygous Icmt knockout cells (IcmtϪ/Ϫ), two independent Icmtϩ/Ϫ clones were grown on gelatin-coated 100-mm dishes in medium containing high concentrations of G418 (6.0 mg/ml), as described by Mortensen et al. (23). Surviving colonies were picked and analyzed by Southern blotting with the 5Ј flanking probe. To define the growth characteristics of the IcmtϪ/Ϫ cells, two nontargeted clones (from the initial electroporation), and two IcmtϪ/Ϫ clones were plated onto wells of 96-well plates (n ϭ 12 wells/cell line/plate) and incubated for 24, 48, and 72 h. At each time point, the cell density was assessed with the Cell Titer 96 Aqueous One solution reagent (Promega, Madison, WI). In a separate experiment, 10 6 Icmtϩ/ϩ cells were mixed with 10 6 IcmtϪ/Ϫ cells and then plated onto multiple wells of 6-well plates. The cells were then passaged six times over a 20-day period. Southern blots (with the 5Ј flanking probe) were performed on DNA from cells harvested after the first passage and again after six passages, with the goal of determining whether the Icmtϩ/ϩ cells would outgrow the IcmtϪ/Ϫ cells.
Preparation of Whole-cell Lysates and Cellular Fractions for Icmt Activity Assays-ES cells were grown to near confluency on gelatincoated 100-mm dishes (without feeder cells), washed twice with phosphate-buffered saline (PBS), and scraped into 0.5 ml of a buffer containing 10 mM Tris-HCl (pH 7.5), 100 mM NaCl, 5 mM MgCl 2 , 0.1 mM phenylmethylsulfonyl fluoride (PMSF), and a protease inhibitor mixture (cat. no. 1836170, Roche Molecular Biochemicals). To prepare whole-cell lysates, the cells were freeze-thawed three times and sonicated. For isolation of membrane and cytosolic fractions, the cells were grown in gelatin-coated T175 flasks, washed with ice-cold PBS, and harvested by scraping into 1 ml of PBS. After a brief centrifugation, the cells were resuspended in 1.225 ml of a hypotonic lysis buffer containing 10 mM Tris-HCl (pH 7.5), 1 mM MgCl 2 , 1 M dithiothreitol, 10 g/ml chymostatin, 10 g/ml leupeptin, 10 g/ml pepstatin, 2 g/ml aprotinin, and 1 mM PMSF. After incubation on ice for 10 min, the cell suspension was homogenized by 20 strokes with a 7-ml Dounce homogenizer, adjusted to 155 mM NaCl to a total volume of 1.450 ml, and subjected to ultracentrifugation at 100,000 ϫ g for 30 min at 4°C. The supernatant fluid (S100, cytosolic fraction) was transferred to a new tube, and the pellet (P100, membrane fraction) was resuspended in 800 l of a buffer containing 50 mM Tris-HCl (pH 7.5), 0.2 M sorbitol, 5 mM EDTA, 0.02% sodium azide, 10 g/ml chymostatin, 10 g/ml leupeptin, 10 g/ml pepstatin, 2 g/ml aprotinin, and 1 mM PMSF.
Icmt Activity Assay-To measure Icmt activity, whole-cell lysates or cellular fractions (40 - (20), was also tested as a methyl-accepting substrate (this substrate can be methylated by whole-cell mammalian lysates because the lysates contain Rce1, which cleaves the last three amino acids of the protein, as well as Icmt). The total reaction volume for all of the methylation reactions was 50 l. After a 30-min incubation at 37°C, the methylation reaction was stopped by adding 50 l of a 1.0 M NaOH solution. The majority of the mixture (90 l) was immediately spotted onto a pleated 2 ϫ 8-cm filter paper wedged in the neck of a 20-ml scintillation vial that contained 5 ml of scintillation fluid (ScintiSafe Econo 1, Fisher). The vials were capped and incubated at room temperature for 5 h to allow the [ 14 C]methanol (formed by base hydrolysis of carboxyl methyl esters) to diffuse into the scintillation fluid (3). The filter papers were then removed, and the vials were counted for radioactivity. Control reactions were also set up in which the S-adenosyl-L-[methyl-14 C]methionine was added to the cell lysates in the absence of a methyl-accepting substrate (e.g. AGGC or AFC). Methyltransferase activity (pmol/mg total cell protein/min) was calculated after subtracting the background level of methylation in the control reactions.
Quantification of Substrate Accumulation in IcmtϪ/Ϫ Cells-To assess the level of methylation substrates within cells, whole-cell lysates (Icmtϩ/ϩ, IcmtϮ-, and IcmtϪ/Ϫ) were incubated with S-adenosyl-L-[methyl-14 C]methionine and recombinant Ste14p (10 g of membrane protein from Sf9 cells that overexpressed yeast STE14) (24). The amount of base-labile methylation was quantified as described above. Methylation of cell lysates was also quantified in the presence of S-adenosyl-L-[methyl-14 C]methionine but without recombinant Ste14p.
Membrane Association of the Ras Proteins-The membrane association of K-Ras within ES cells was assessed by subcellular fractionation followed by Western blotting with a K-Ras-specific antibody. Confluent 100-mm dishes of ES cells were washed with ice-cold PBS, collected in 1.0 ml of PBS, split into two equal fractions, and centrifuged at 500 ϫ g for 10 min. Cells were incubated for 10 min on ice in 1225 l of hypotonic buffer (10 mM Tris-HCL, pH 7.5, 1.0 mM MgCl 2 , 0.5 mM PMSF, 10 g/ml leupeptin, 10 g/ml aprotinin, and 1 M dithiothreitol). The cells were then disrupted with an ice-cold Dounce tissue homogenizer, after which 225 l of 1 M NaCl was added. A total of 450 l of this solution (total cellular lysate) was transferred to a microcentrifuge tube and set aside. The remaining 1 ml was transferred to a polycarbonate ultracentrifuge tube and spun at 100,000 ϫ g for 30 min at 4°C. The supernatant fluid (S100) was transferred to a new microcentrifuge tube; the pellet (P100) was resuspended in 850 l of hypotonic buffer and 150 l of a 1 M NaCl solution. Next, 50 l of a solution containing 10% sodium deoxycholate, 10% Nonidet P-40 (v/v), and 5% SDS was added to the total lysate sample, and 110 l was added to the S100 and P100 fractions. After incubation on ice for 10 min, the lysates were clarified by centrifugation at 25,000 ϫ g for 30 min at 4°C. The supernatant fluids were transferred to a new microcentrifuge tube, pre-cleared by incubation with protein G-agarose (Roche Molecular Biochemicals) for 30 min at 4°C, and then incubated with 5 g of Y13-259 (Santa Cruz Biotechnology, Santa Cruz, CA) overnight at 4°C. The immune complexes were then incubated with protein G-agarose beads for 2 h and centrifuged for 5 min at 12,500 ϫ g. After three washes with radioimmune precipitation buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5 mM MgCl 2 , 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 0.5 mM PMSF, 10 g/ml leupeptin, 10 g/ml aprotinin), the pellets were resuspended in 20 l of sample buffer, and the proteins were fractionated on a 10 -20% gradient SDS-polyacrylamide gel. After electrophoretic transfer of the proteins to a nitrocellulose membrane, Western blots were performed with a K-Ras-specific monoclonal antibody (Ab-1, Oncogene Science, Uniondale, NY) and a horseradish peroxidase-conjugated sheep anti-mouse IgG (Amersham Pharmacia Biotech). Antibody binding was detected with the enhanced chemiluminescence kit (ECL, Amersham Pharmacia Biotech). Band intensities were quantified by densitometry with Quantity One quantitation software (Bio-Rad Laboratories, Hercules, CA).
Transfection of Cells with a Green Fluorescent Protein-K-Ras4B Fusion Construct-Icmtϩ/ϩ and IcmtϪ/Ϫ cells were plated onto 2 ϫ 4-cm chamber slides (1-well Permanox, Nalge Nunc Int., Naperville, IL) at a density of 10 5 cells/slide. Once the cells had achieved 70% confluence, they were transfected with an enhanced green fluorescent protein (GFP)-K-Ras4B fusion construct (containing the entire GFP coding sequence fused in-frame to the carboxyl-terminal 18 amino acids of mouse K-Ras4B) (25). In wild-type mouse fibroblasts, this fusion protein (like the endogenous Ras proteins) is located almost entirely along the plasma membrane (25). The cell transfections were carried out with 1 g of plasmid DNA and SuperFect reagent (Qiagen, Valencia, CA) according to the manufacturer's instructions. 36 h after the transfection, the cells were fixed with 4% paraformaldehyde, and the fluorescence in the transfected cells was visualized by a Bio-Rad MRC-600 laser scanning confocal imaging system.

RESULTS
Production of Icmt-deficient Cells-We sought to use gene targeting to produce a mammalian cell line lacking Icmt expression. We approached this objective with mouse ES cells because of the numerous reports of high efficiency homologous recombination in those cells (26,27). A sequence-replacement gene-targeting vector (Fig. 1A) was constructed to delete exon 1 sequences containing the translational start site (ATG) and the first 65 amino acids of the protein (specifying the first two membrane-spanning domains). After electroporation of the vector into ES cells and 10 days of selection, 160 ES cell colonies were picked. Southern blots revealed that eight clones were correctly targeted (i.e. heterozygous for the knockout mutation (Icmtϩ/Ϫ)) (Fig. 1B).
To produce IcmtϪ/Ϫ cells, two Icmtϩ/Ϫ clones were subjected to selection in high concentrations of G418. Sixty colonies were picked, and 40 were homozygous for the Icmt knockout mutation. Fig. 1B shows a Southern blot of two nontargeted Icmtϩ/ϩ clones, two Icmtϩ/Ϫ clones, and two IcmtϪ/Ϫclones. Interestingly, the IcmtϪ/Ϫ cells appeared to grow more slowly than the Icmtϩ/ϩ cells. This observation was supported by quantitative cell density measurements. IcmtϪ/Ϫ cell densities were reduced by 5-30% at 24, 48, and 72 h, compared with the Icmtϩ/ϩ cells (p Ͻ 0.01, data not shown). Moreover, when equal numbers of Icmtϩ/ϩ and IcmtϪ/Ϫ cells were mixed and passaged six times over 20 days, the Icmtϩ/ϩ cells outgrew the IcmtϪ/Ϫ cells. After six passages, the 6.8-kb BamHI fragment (corresponding to the IcmtϪ/Ϫ cells) was undetectable by Southern blot analysis (data not shown).
Absence of Isoprenylcysteine Carboxyl Methyltransferase Activity in IcmtϪ/Ϫ Cells-Previous studies have documented that Icmt has a high degree of specificity for isoprenylcysteine residues (reviewed by Young et al. (9)) and that the enzyme is capable of methylating small molecules such as AGGC and AFC (albeit at a higher K m than with isoprenylated proteins) (28). Indeed, we found that whole-cell lysates from Icmtϩ/ϩ cells were capable of using S-adenosyl-L-[methyl-14 C]methionine to methylate both AGGC and AFC (Fig. 2). Consistent with previous studies, (11,12), all of this enzymatic activity was confined to the P100 fraction, and none was in the S100 fraction (data not shown). In contrast, whole-cell lysates from the IcmtϪ/Ϫ cells were incapable of methylating AGGC or AFC (Fig. 2). The Icmtϩ/Ϫ clones contained half-normal levels of enzymatic activity (Fig. 2). The results were similar when farnesyl-K-Ras was used as the substrate (this protein can be processed in this in vitro system because the carboxyl-terminal three amino acids of the protein are removed by the endoproteolytic processing activity in the cell lysates) (25). The methylation of K-Ras was robust with Icmtϩ/ϩ lysates, absent with the IcmtϪ/Ϫlysates, and occurred at half-normal levels with the Icmtϩ/Ϫ lysates (Fig. 2).
Methylation Substrates Accumulate in Icmt-deficient In some cases, the standard deviations were low so that the error bar is not visible above the bar. Similar results were obtained five times for AGGC and AFC and four times for K-Ras. Similar results were observed for three independent Icmtϩ/Ϫ clones and two independent IcmtϪ/Ϫ clones. Cells-To quantify the level of substrates for carboxyl methylation within cells, we incubated whole-cell lysates of Icmtϩ/ϩ, Icmtϩ/Ϫ, and IcmtϪ/Ϫ cells with S-adenosyl-L-[methyl-14 C]methionine and 10 g of Sf9 membranes expressing yeast STE14 (24). Low levels of base-labile methylation were observed with the Icmtϩ/ϩ lysates (Fig. 3A). We suspect that these low levels of methylation were largely the result of other methyltransferase activities (e.g. Pcmt1 (29,30)), but some of this methylation was probably because of Ste14p-mediated methylation of isoprenylated cellular proteins that had not yet been methylated by the endogenous Icmt activity. In contrast, a substantial increase in the level of methylation was observed in lysates from the IcmtϪ/Ϫ cells, indicating that these cells had accumulated Ste14p substrates (nonmethylated isoprenylated proteins) (Fig. 3A). In addition, there was a modest but significant accumulation of Ste14p substrates in Icmtϩ/Ϫ lysates, compared with the Icmtϩ/ϩ lysates (Fig. 3A).
We observed only low levels of base-labile methylation in Icmtϩ/ϩ and IcmtϪ/Ϫ lysates when S-adenosyl-L-[methyl-14 C]methionine was added in the absence of Ste14p (Fig. 3B). Once again, we suspect that this low level of methylation was because of other methyltransferase activities (although some of the methylation in the Icmtϩ/ϩ lysates may have been due to the action of the endogenous Icmt activity on "not yet methylated" isoprenylated proteins). Interestingly, methylation was far greater when S-adenosyl-L-[methyl-14 C]methionine was added to an equal mixture of Icmtϩ/ϩ and IcmtϪ/Ϫ lysates than when it was added to either alone, demonstrating that that Icmt substrates that accumulated in IcmtϪ/Ϫ cells were readily methylated by the endogenous Icmt activity from Icmtϩ/ϩ lysates (Fig. 3B).
K-Ras Is Mislocalized in Icmt-deficient Cells-To determine whether the absence of Icmt activity altered the subcellular distribution of K-Ras, we compared the relative levels of K-Ras in the membrane (P100) and cytosolic (S100) fractions of Icmtϩ/ϩ and IcmtϪ/Ϫ ES cells by immunoblotting. In Icmtϩ/ϩ cells, the vast majority of K-Ras was in the P100 fraction with very little in the S100 fraction (Fig. 4). In contrast, a significant proportion of the K-Ras in IcmtϪ/Ϫ cells was present in the S100 fraction. To further assess the effect of Icmt deficiency on the membrane association of K-Ras, Icmtϩ/ϩ and IcmtϪ/Ϫ cells were transiently transfected with a GFP-K-Ras4B fusion plasmid. As expected, the GFP-K-Ras fusion protein was localized along the plasma membrane in Icmtϩ/ϩ cells (Fig. 5A). In contrast, a large fraction of the fusion protein produced in IcmtϪ/Ϫ cells was trapped within the cytoplasm, and the amount at the plasma membrane was reduced (Fig. 5B).

DISCUSSION
In this study, we inactivated both Icmt alleles in mouse ES cells. One allele was inactivated with a conventional sequencereplacement gene-targeting strategy. The second allele was inactivated by a second round of selection in a high concentration of G418 (6 mg/ml). This procedure, first described by Mortensen et al. (23), probably works by selecting for the survival of rare cells in which an interchromosomal gene conversion event has "repaired" the wild-type chromosome with the mutant chromosome, thereby providing it with an extra copy of the neo selection marker. In the case of the Icmt locus, the "high G418" selection procedure was extremely efficient and proved to be far simpler than knocking out the other allele with a second gene-targeting vector containing a different drug selection marker (31).
The homozygous knockout cells lacked all Icmt activity against a protein substrate (farnesyl-K-Ras4B) or against farnesylated and geranylgeranylated small molecules (AFC and AGGC, respectively). Thus, the IcmtϪ/Ϫ cells did not appear to contain any redundant biochemical activity capable of methylating isoprenylated substrates, at least under the conditions of our assays. Interestingly, we had no difficulty in using methylation assays to document an accumulation of Icmt substrates in the IcmtϪ/Ϫ cells. In the future, it should be possible to methylate the substrates in IcmtϪ/Ϫ cells with S-adenosyl-L-[methyl-14 C]methionine, and then use two-dimensional gel electrophoresis to identify the most abundant Icmt substrates.
Our principal objective in creating an Icmt-deficient mammalian cell line was to assess the importance of the carboxyl FIG. 4. P100/S100 distribution of K-Ras in Icmt؉/؉ and Icmt؊/؊ ES cells. Membrane (P100) and cytosolic (S100) fractions from Icmtϩ/ϩ and IcmtϪ/Ϫ ES cells were isolated by ultracentrifugation. Ras proteins were immunoprecipitated from the P100 and S100 fractions (as well as from the total cell lysates) with antibody Y13-259. The immune complexes were then analyzed by Western blotting with the K-Ras-specific antibody Ab-1. As judged by densitometry, the P100/ S100 band density ratio was 11 for the Icmtϩ/ϩ cells and 3 for the IcmtϪ/Ϫ cells. Similar results were obtained in three independent experiments. methylation step for the intracellular localization of the Ras proteins. To address this issue, we prepared P100 (membrane) and S100 (cytosolic) fractions from IcmtϪ/Ϫand Icmtϩ/ϩ cells and then analyzed the membrane association of K-Ras by Western blotting. We also transfected IcmtϪ/Ϫand Icmtϩ/ϩ cells with a GFP-K-Ras fusion plasmid in which the GFP coding sequences were ligated in-frame to the carboxyl-terminal 18 amino acids of K-Ras 4B. The results of both lines of investigation were clear. Compared with the Icmtϩ/ϩ cells, more of the K-Ras in the IcmtϪ/Ϫ cells was in the S100 fraction. In addition, the K-Ras fusion protein in the wild-type cells was localized almost exclusively along the plasma membrane, whereas a large fraction of the fusion in the IcmtϪ/Ϫ cells was trapped in the cytoplasm and less was at the plasma membrane.
These data indicating mislocalization of K-Ras in the absence of carboxyl methylation require several comments. First, these results are clearly consistent with the finding of abnormal Ras2p processing and abnormal Ras2p membrane localization in ste14⌬ S. cerevisiae (18). However, despite the concordance of the current data and the yeast data, we are reluctant to extrapolate our results to all Ras isoforms in mammals. In our study, we examined the localization of a K-Ras, which relies in part on a polybasic sequence for plasma membrane localization (32)(33)(34). It is possible that the methylation step could be less important for other mammalian Ras isoforms (e.g., H-Ras) that undergo palmitoylation of cysteine residues within the carboxyl-terminal part of the molecule (35)(36)(37). Second, the degree of the mislocalization of the K-Ras fusion in IcmtϪ/Ϫ cells, although quite marked, did not appear to be as striking as that observed in Rce1-deficient fibroblasts (where the Ras proteins are neither endoproteolytically processed nor methylated) (25). The simplest explanation for this finding is that both the endoproteolysis step and the carboxyl methylation step contribute independently to Ras localization and that eliminating both steps, as occurs in the Rce1 knockout, has a greater impact than eliminating only the methylation step. Once again, however, caution is warranted, because the Rce1 experiments were performed in fibroblasts, whereas the Icmt experiments were performed in ES cells. Third, we emphasize that the mechanism(s) for the K-Ras mislocalization in IcmtϪ/Ϫ cells is not fully understood. The Icmt-mediated conversion of a carboxylate ion to an ␣-carboxyl methyl ester would be predicted to render the carboxyl terminus of the protein more hydrophobic, facilitating its interaction with the plasma membrane (9). However, that may not be the only explanation for our results. The subcellular trafficking of K-Ras appears to go through a route distinct from that of H-Ras and N-Ras (38,39), and microtubules may be involved in this process for K-Ras (38). It is conceivable that the absence of methylation adversely affects the transport of K-Ras, thereby reducing the level of this protein at the plasma membrane.
The IcmtϪ/Ϫ cell line will be a valuable reagent for investi-gators who are interested in the metabolism of isoprenylated proteins. As noted above, two-dimensional gel experiments with these cells could help to identify Icmt substrates. Perhaps more importantly, the cells could yield clues as to why the isoprenylcysteine methylation step has been so conserved in eukaryotic biology. In yeast, the STE14-mediated "capping" of a-factor (a farnesylated CaaX protein) with an ␣-methyl ester improves a-factor stability by preventing rapid intracellular degradation of the protein (18). Based on that precedent, we hypothesized that the absence of Icmt might adversely affect the stability of mammalian isoprenylated proteins (9). With the production of IcmtϪ/Ϫ cells, it will be possible to test that hypothesis in a direct fashion. In addition, because many of the substrates for Icmt are signaling molecules (9), it will be possible to assess the influence of methylation on various signal transduction pathways. Finally, we find it difficult to resist speculating that Icmt might function in mammalian cells to facilitate a receptor-mediated transport of an isoprenylated peptide across a membrane. In yeast, methylation of a-factor is crucial for the STE6-mediated transport of a-factor out of the cell (10). Although no a-factor orthologue has been identified in higher organisms, it is nevertheless conceivable that transmembrane transport of an as yet unidentified isoprenylated protein occurs in mammals and that the methylation of the isoprenylcysteine is important for this process. The existence of IcmtϪ/Ϫ cells could be quite helpful in examining that possibility.