|Home | About | Journals | Submit | Contact Us | Français|
Isoprenylcysteine carboxyl methyltransferase (Icmt) methylates the carboxyl-terminal isoprenylcysteine of CAAX proteins (e.g., Ras and Rho proteins). In the case of the Ras proteins, carboxyl methylation is important for targeting of the proteins to the plasma membrane. We hypothesized that a knockout of Icmt would reduce the ability of cells to be transformed by K-Ras. Fibroblasts harboring a floxed Icmt allele and expressing activated K-Ras (K-Ras-Icmtflx/flx) were treated with Cre-adenovirus, producing K-Ras-IcmtΔ/Δ fibroblasts. Inactivation of Icmt inhibited cell growth and K-Ras–induced oncogenic transformation, both in soft agar assays and in a nude mice model. The inactivation of Icmt did not affect growth factor–stimulated phosphorylation of Erk1/2 or Akt1. However, levels of RhoA were greatly reduced as a consequence of accelerated protein turnover. In addition, there was a large Ras/Erk1/2-dependent increase in p21Cip1, which was probably a consequence of the reduced levels of RhoA. Deletion of p21Cip1 restored the ability of K-Ras-IcmtΔ/Δ fibroblasts to grow in soft agar. The effect of inactivating Icmt was not limited to the inhibition of K-Ras–induced transformation: inactivation of Icmt blocked transformation by an oncogenic form of B-Raf (V599E). These studies identify Icmt as a potential target for reducing the growth of K-Ras– and B-Raf–induced malignancies.
Proteins that terminate with a carboxyl-terminal “CAAX” motif, such as the Ras and Rho proteins, undergo three sequential posttranslational processing events. First, the cysteine (i.e., the C of the CAAX sequence) is isoprenylated by protein farnesyltransferase (FTase) or geranylgeranyltransferase type I (GGTase I) (1). Second, the last three amino acids of the protein (i.e., the -AAX) are cleaved off by Rce1, an integral membrane protein of the ER (2). Third, the newly exposed isoprenylcysteine is methylated by an ER membrane–bound methyltransferase, isoprenylcysteine carboxyl methyltransferase (Icmt) (3). These modifications render the C terminus of CAAX proteins more hydrophobic, facilitating binding to membranes (4–6).
The posttranslational processing of CAAX proteins has attracted interest because of the central role of mutationally activated Ras proteins in the development of cancer (7, 8). The enzymes that carry out the posttranslational modifications of CAAX proteins (i.e., FTase, GGTase I, Rce1, and Icmt) have been considered as potential targets for modulating the activity of the Ras proteins and for blocking the growth of Ras-induced malignancies. Farnesylation is critical for Ras activity (9), and farnesyltransferase inhibitors (FTIs) have shown promise in treating tumors, both in experimental animals (10, 11) and in humans (12–17).
A potential drawback of the clinical use of FTIs is that K-Ras and N-Ras—the isoforms most often mutated in human tumors—can be efficiently geranylgeranylated in the setting of FTI therapy (18, 19). This alternate prenylation of the Ras proteins could limit the efficacy of FTIs in the treatment of Ras-induced tumors. The existence of an alternate means for prenylation has led several groups to focus on the postisoprenylation steps mediated by Rce1 and Icmt, since those steps are shared by farnesylated and geranylgeranylated CAAX proteins (6). We previously generated Rce1-knockout mice and found that they died quite late in gestation, after all of the major organ systems were formed (20). The Ras proteins in Rce1-deficient fibroblasts were mislocalized away from the plasma membrane, and the cells grew slightly slower than wild-type cells (20). More recently, we showed that the inactivation of Rce1 partially blocked transformation of cells by an activated form of H-Ras or K-Ras and sensitized transformed cells to the antiproliferative effects of an FTI (21).
The phenotype of Icmt deficiency in mice was more severe than Rce1 deficiency; an Icmt knockout caused grossly retarded growth during embryonic development and death at embryonic day 10.5–11.5 (22), possibly due to agenesis of the liver (23). Icmt deficiency causes mislocalization of the Ras proteins within cells, but virtually nothing is known about the effects of Icmt deficiency on cell growth and oncogenic transformation. To address these issues, we created a conditional (“floxed”) Icmt allele, generated fibroblast cell lines, and then analyzed the consequences of inactivating Icmt.
To generate a conditional Icmt allele, exon 1 of Icmt along with upstream promoter sequences and parts of intron 1 were flanked with loxP sites. A 5′ arm of the gene-targeting vector (4 kb in length) was amplified from bacterial artificial chromosome DNA (24) with primers 5′-CTCTGTGCGGCCGCCTGTGTATAACTGTTTCCTTAGGTATG-3′ and 5′-ACGACGGCGGCCGCCCGGCGACGCCGGCTCGGGAAGGGC-3′ and cloned into the NotI site of pKSloxPNT (25). A 2.1-kb 3′ arm was amplified with primers 5′-ACGACGGCGGCCGCAGGGTAGGTGCACCAGGTACATTAGAACCG-3′ and 5′-ACGACGGAATTCCTTCAGTTCTGGCCAGAAGATGTTGTCGAGCG-3′ and cloned into the polylinker EcoRI site. A 717-bp fragment containing promoter sequences, exon 1, and parts of intron 1 was amplified with 5′-ACGACTATCGATTTGAATGCGCAGGCCGGCCTTCGGGTTTCCC-3′ and 5′-ACGACGGCGGCCGCATAACTTCGTATAATGTATGCTATACGAAGTTATGCCCCATCCCTGACTGATCGGTCCCTG-3′, which contained a loxP site. That fragment was inserted between the polylinker ClaI and BamHI sites. The vector was linearized with XhoI and electroporated into 129SV/Jae ES cells. ES cells were screened by Southern blotting; the genomic DNA was cleaved with BamHI, and the blots were hybridized with a 292-bp 5′-flanking probe (24). Two targeted clones, each with a single integration event, were used to generate germline-transmitting chimeric mice.
Primary fibroblasts were isolated from Icmtflx/flx embryos on day 15.5 post coitum (20, 26). The cells were immortalized with a modified 3T3 protocol (26) or by transfecting the cells with E1A and activated human K-Ras (with a G12V activating mutation) (21, 27, 28). For some experiments, the fibroblasts were also infected with a retrovirus encoding human K-Ras (G12V) and a puromycin-resistance cassette. Experiments were performed with two different K-Ras–transfected Icmtflx/flx cell lines (A and B), unless otherwise indicated. To produce B-Raf–transformed cell lines, spontaneously immortalized Icmtflx/flx fibroblasts were transfected with a retrovirus encoding an oncogenic form of human B-Raf (V599E) and a blasticidin-resistance cassette. Control cells were infected with a retrovirus encoding wild-type B-Raf. To inactivate Icmt (to produce IcmtΔ/Δ cells), Icmtflx/flx cells were infected with 108 PFU/ml of Cre-adenovirus (AdRSVCre). In parallel, control Icmtflx/flx cells were treated with 108 PFU/ml of a β-gal adenovirus (AdRSVnlacZ). The efficiency of Cre-mediated recombination was assessed with Southern blotting. All experiments with IcmtΔ/Δ cells were initiated three to four passages after the AdRSVCre infection. To produce mixed populations of Icmtflx/flx and IcmtΔ/Δ cells, Icmtflx/flx cells were infected with low titers of AdRSVCre (106 PFU/ml) and high titers (108 PFU/ml) of AdRSVnlacZ (so that all of the cells in the dish would be infected with adenovirus).
Icmt activity was measured with base-hydrolysis assays with S-adenosyl-L-[methyl-14C]methionine as the methyl donor and either N-acetyl-S-geranylgeranyl-L-cysteine or farnesyl-K-Ras as the substrate. The accumulation of “methylatable substrates” in cell extracts was quantified by measuring the methylation of proteins in cell extracts in the presence of recombinant Ste14p (the yeast ortholog of Icmt) and S-adenosyl-L-[methyl-14C]methionine (24).
Anchorage-dependent growth was assessed with a tetrazolium compound (MTS)–based cell proliferation assay (Cell Titer 96 AQueous One Solution Cell Proliferation Assay; Promega Corp., Madison, Wisconsin, USA). Icmtflx/flx and IcmtΔ/Δ cells were seeded on 96-well plates (500 or 1,000 cells/well, n = 12 wells/cell line, 1 plate per time point) and incubated at 37°C. At various time points, 20 μl of the MTS reagent ([3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)]-2H-tetrazolium, inner salt) was added to each well and incubated for 2 hours at 37°C. Cell density was quantified by analyzing absorbance at 490 nm. The relative growth rates of Icmtflx/flx and IcmtΔ/Δ cells also were assessed with Southern blots to compare the relative intensities of Icmtflx and IcmtΔ bands at different passages. Cells were passaged 1:10 every 4 days; the intensities of bands on Southern blots were quantified with a phosphorimager.
To fully inactivate Icmt in the liver (i.e., obtain nearly complete levels of recombination in the liver), Icmtflx/flx mice harboring the inducible Mx1-Cre transgene (29) were injected intraperitoneally with polyinosinic-polycytidylic ribonucleic acid (pI-pC; Sigma-Aldrich, St. Louis, Missouri, USA) (300 μg every other day for a total of four injections). To generate mice with partial recombination in the liver (i.e., mice whose livers contained a mixture of Icmtflx/flx and IcmtΔ/Δ hepatocytes), mice were given a single pI-pC injection (200–500 μg). In the latter mice, the relative rates of Icmtflx/flx and IcmtΔ/Δ cell proliferation were assessed after a partial hepatectomy (30). To gain insight into the relative growth rates of Icmtflx/flx and IcmtΔ/Δ hepatocytes, Southern blotting was used to compare the ratio of Icmtflx and IcmtΔ alleles at the time of partial hepatectomy and 10 days later, when liver regeneration was complete.
The ability of K-Ras-IcmtΔ/Δ and parental K-Ras-Icmtflx/flx fibroblasts to form colonies in soft agar was analyzed in side-by-side experiments. For these studies, K-Ras-IcmtΔ/Δ fibroblasts were produced from K-Ras-Icmtflx/flx fibroblasts with AdRSVCre (108 PFU/ml). Control K-Ras-Icmtflx/flx fibroblasts were infected with 108 PFU/ml of AdRSVnlacZ. The K-Ras-IcmtΔ/Δ and K-Ras-Icmtflx/flx fibroblasts (500–10,000 cells) were placed in medium containing 0.35% agarose and then poured onto 12-well plates containing a 0.70% agarose base. After the plates had been incubated at 37°C for 21–30 days, the colonies were stained with 3-[4,5-dimethylthiazol-2-yl]2,5-diphenyltetrazolium bromide (thiazolyl blue, MTT; Sigma-Aldrich) and photographed. The number of colonies was quantified with an image-processing tool kit in Adobe Photoshop 6.0 (21).
To assess the relative capacities of K-Ras-IcmtΔ/Δ and parental K-Ras-Icmtflx/flx fibroblasts to form tumors in nude mice, a mixed population of K-Ras-Icmtflx/flx and K-Ras-IcmtΔ/Δ fibroblasts was generated by treating K-Ras-Icmtflx/flx fibroblasts with a low-titer AdRSVCre (106 PFU/ml) as already described. Those cells (5 × 106) were injected into the flanks of 10-week-old female nude mice (strain NU/J Foxn1nu; The Jackson Laboratory, Bar Harbor, Maine, USA). The relative capacities of the K-Ras-Icmtflx/flx and the derivative K-Ras-IcmtΔ/Δ fibroblasts to contribute to tumor growth were assessed by comparing the ratios of Icmtflx and IcmtΔ alleles in the genomic DNA of the injected fibroblasts and in the tumors (harvested 21 days after injection).
K-Ras-Icmtflx/flx fibroblasts were transfected with an expression vector (pcDNA6Myc/His-B; Invitrogen Corp., Carlsbad, California, USA) containing the human ICMT cDNA. Blasticidin-resistant colonies were picked and expanded. Two K-Ras-Icmtflx/flx:ICMT cell lines were infected with AdRSVCre, generating K-Ras-IcmtΔ/Δ:ICMT cell lines. The K-Ras-Icmtflx/flx:ICMT and the derivative K-Ras-IcmtΔ/Δ:ICMT cell lines expressed similar levels of Icmt enzymatic activity, indicating that most of the enzymatic activity in the cells was due to the human ICMT cDNA.
S100 (cytosolic) and P100 (membrane) fractions of K-Ras-IcmtΔ/Δ and K-Ras-Icmtflx/flx cells were prepared, and Ras proteins were immunoprecipitated as described (21). Western blots were performed with a K-Ras–specific mAb (Ab-1; Oncogene Research Products, San Diego, California, USA), a HRP-conjugated sheep anti-mouse IgG (Amersham Biosciences Corp., Piscataway, New Jersey, USA), and the Enhanced Chemiluminescence Kit (Amersham Biosciences Corp.).
Icmt+/+ and Icmt–/– fibroblasts were transfected with a plasmid coding for a fusion between EGFP and the C-terminal 20 amino acids of K-Ras (20). Transfections were carried out with SuperFect Reagent (QIAGEN Inc., Valencia, California, USA). Cells were fixed in 4% formalin 24–28 hours after transfection, mounted, and viewed with a Bio-Rad MRC-600 laser scanning confocal imaging system (Bio-Rad Laboratories Inc., Hercules, California, USA).
Cyclin A was immunoprecipitated from 1.0 mg of total cell protein from extracts of K-Ras-IcmtΔ/Δ and K-Ras-Icmtflx/flx cells with a polyclonal antibody (H-432; Santa Cruz Biotechnology Inc., Santa Cruz, California, USA) and resolved on 10–20% SDS-PAGE gels. To detect cyclin A–associated p21Cip1, immunoblots were then performed with an anti-p21Cip1 antibody (F-5; Santa Cruz Biotechnology Inc.).
To assess the impact of the Icmt gene inactivation on Ras and Rho turnover, K-Ras-Icmtflx/flx and the derivative K-Ras-IcmtΔ/Δ cells were seeded onto 60-mm plates, grown overnight, and then grown for 2 hours in methionine-free medium. Cells were metabolically labeled with 0.8 mCi L-[35S]methionine and L-[35S]cysteine (Pro-mix L-[35S] in vitro cell labeling mix; Amersham Biosciences Corp.) for 2 hours. The cells were washed and incubated with regular medium supplemented with 4 mM methionine and cysteine. Cells were harvested at different time points, and Rho and Ras proteins were immunoprecipitated from the extracts with pan-Rho [Anti-Rho (-A, -B, -C); Upstate Biotechnology Inc., Lake Placid, New York, USA] and pan-Ras (Ab-4; Oncogene Research Products) antibodies, respectively. The proteins were resolved on 10–20% SDS-PAGE gels that were subsequently treated with Amplify (Amersham Biosciences Corp.), dried, and exposed to film for 4–7 days.
Serum- and EGF-stimulated phosphorylation of the downstream Ras effectors Erk1/2 and Akt1 was assessed by seeding 105 nontransfected Icmtflx/flx and derivative IcmtΔ/Δ cells (i.e., no activated K-Ras) on 60-mm dishes and then subjecting them to overnight serum starvation (in 0.5% serum). The next morning, medium containing 10% serum or 50 ng/ml human EGF (Sigma-Aldrich) was added to the cells. At various times after EGF stimulation, the cells were harvested and total cell extracts were analyzed by immunoblotting with antibodies recognizing phospho-Erk1/2 (phospho-p44/42 MAP kinase E10 mAb; Cell Signaling Technology, Beverly, Massachusetts, USA), phospho-Akt1 [(Ser 473)-R, polyclonal; Santa Cruz Biotechnology Inc.], and total Erk1/2 (p44/42 MAP kinase, polyclonal; Cell Signaling Technology).
Northern blots were performed with standard techniques (31) on total cellular RNA extracted with the TRI reagent (Sigma-Aldrich). A mouse Kras2 cDNA probe was amplified from a mouse liver cDNA library (CLONTECH Laboratories Inc., Palo Alto, California, USA) with forward primer 5′-AAAGCGGAGAGAGGCCTGCTGAAAATG-3′ and reverse primer 5′-ACCATAGGCACATCTTCAGAGTCC-3′. A mouse RhoA (Ahra2) cDNA probe was amplified from the mouse liver cDNA library with forward primer 5′-ACTGGTGATTGTTGGTGATGGAGC-3′ and reverse primer 5′-TCATTCCGAAGGTCCTTCTTGTTCCC-3′. A mouse p21Cip1 (Cdkn1a) cDNA probe was amplified from the mouse liver cDNA library with forward primer 5′-AAAAGCACCTGCAAGACCAGAGGG-3′ and reverse primer 5′-TTAAGTTTGGAGACTGGGAGAGGG-3′.
Male p21Cip1-deficient mice (Cdkn1a–/–) (32) (provided by Allan Balmain, University of California, San Francisco, California, USA) were bred with Icmtflx/flx mice. Icmt+/flx Cdkn1a+/– mice were intercrossed and embryonic fibroblasts were harvested 15.5 days post coitum. Icmtflx/flxCdkn1a–/– spontaneously immortalized fibroblasts were infected with a retrovirus expressing the activated human K-Ras. K-Ras-Icmtflx/flxCdkn1a–/– fibroblasts were treated with AdRSVCre (108 PFU/ml) to generate K-Ras-IcmtΔ/ΔCdkn1a–/– cells. Control K-Ras-Icmtflx/flxCdkn1a–/– fibroblasts were treated with AdRSVnlacZ (108 PFU/ml).
To assess the importance of Icmt for cell growth and Ras transformation, we created a conditional Icmt-knockout allele in which exon 1 and upstream promoter sequences were flanked by loxP sites (Figure (Figure1a).1a). Two independent ES cell lines (each with a single neo integration) were used to create chimeric mice, which were bred to produce heterozygous mice (Icmt+/flx). When Icmt+/flx mice were intercrossed, homozygous mice (Icmtflx/flx) were born at the predicted mendelian frequency and were fertile and healthy. Fibroblasts from Icmtflx/flx embryos were immortalized as described in the Methods section. Cre-mediated excision of the floxed segment of Icmt resulted in the production of IcmtΔ/Δ cells (Figure (Figure1b).1b). Not surprisingly, IcmtΔ/Δ cells exhibited both a complete absence of Icmt enzymatic activity (Figure (Figure1c)1c) and a substantial accumulation of Icmt protein substrates (Figure (Figure11d).
The inactivation of Icmt in immortalized fibroblasts reduced cell growth by 40–60% under routine cell culture conditions (anchorage-dependent growth) (Figure (Figure2a).2a). To further examine the consequences of “switching off” Icmt, we produced mixed populations of Icmtflx/flx and IcmtΔ/Δ cells by treating Icmtflx/flx cells with low titers (106 PFU/ml) of AdRSVCre and high titers (108 PFU/ml) of AdRSVnlacZ (so that all of the cells were infected with an adenovirus). We then initiated competitive growth experiments to define the relative growth rates for Icmtflx/flx and IcmtΔ/Δ cells. For this, Southern blots were used to assess the ratio of the 7.0-kb Icmtflx band and the 4.8-kb IcmtΔ band (Figure (Figure2b).2b). If the Icmt inactivation had no effect on cell growth (i.e., if there were no differences in the competitive fitness of Icmtflx/flx and IcmtΔ/Δ cells), one would expect that the Southern blots would reveal stability—at different passages—in the relative intensities of the Icmtflx and IcmtΔ bands. This was not the case. In each experiment, the ratio of the Icmtflx band to the IcmtΔ band increased in later passages, indicating a competitive growth advantage of the parental Icmtflx/flx cells over the IcmtΔ/Δ cells (Figure (Figure22b).
Inactivating Icmt also interfered with the growth of hepatocytes in vivo. We bred Icmtflx/flx mice with mice harboring the inducible Mx1-Cre transgene and then administered a single intraperitoneal injection of pI-pC. A single pI-pC injection causes incomplete recombination in the liver; thus, the livers of these mice contained a mixed population of Icmtflx/flx and IcmtΔ/Δ hepatocytes. Two weeks later, we performed a partial hepatectomy, which triggers hepatocyte proliferation. To gauge the relative capacities of Icmtflx/flx and IcmtΔ/Δ hepatocytes to contribute to liver regeneration, we used Southern blots to measure the ratio of the 7-kb Icmtflx band to the 4.8-kb IcmtΔ band in liver genomic DNA, both at baseline (at the time of the partial hepatectomy) and 10 days later when liver regeneration was complete. After liver regeneration, the ratio of the Icmtflx band to the IcmtΔ band invariably increased, indicating that Icmtflx/flx hepatocytes contributed more than the IcmtΔ/Δ hepatocytes to liver regrowth (Figure (Figure2c;2c; P < 0.01, n = 5). This difference could not be attributed to hepatocyte death due to the Icmt knockout. By treating Icmtflx/flx mice with four intraperitoneal injections of pI-pC, we readily generated mice in which Cre-mediated recombination in the liver was virtually complete. In those mice, plasma transaminases remained normal, and there was no histological evidence of hepatocyte necrosis (not shown).
Icmtflx/flx fibroblasts that had been transfected with an activated form of K-Ras (K-Ras-Icmtflx/flx) exhibited a transformed phenotype, forming numerous colonies in soft agar (Figure (Figure3a).3a). To assess the consequences of Icmt deficiency on K-Ras transformation, we then used AdRSVCre to inactivate Icmt in the K-Ras-Icmtflx/flx cells, producing K-Ras-IcmtΔ/Δ cell lines. In six independent side-by-side experiments, the K-Ras-IcmtΔ/Δ cells yielded 90–95% fewer colonies in soft agar than the parental K-Ras-Icmtflx/flx cells (Figure (Figure3,3, a and b, P < 0.0001). When these experiments were repeated with K-Ras-Icmtflx/flx cells that expressed a human ICMT cDNA, AdRSVCre inactivation of the mouse Icmt gene had no detectable effect on colony growth (Figure (Figure3,3, a and b, n = three independent experiments, P = 0.63). Thus, inactivation of Icmt strikingly inhibited the ability of K-Ras–transfected cells to grow in soft agar — except when the cells had been transfected with human ICMT. Similar data were obtained in experiments with fibroblasts from conventional Icmt knockout embryos (Icmt–/–) and littermate Icmt+/+ controls. After immortalization of the fibroblasts and K-Ras transfection, the Icmt+/+ fibroblasts, but not the Icmt–/– fibroblasts, formed colonies in soft agar (n = five independent Icmt–/– and n = three independent Icmt+/+ cell lines) (not shown).
To assess the consequences of Icmt deficiency on tumor cell growth in vivo, mixed populations of K-Ras-Icmtflx/flx and K-Ras-IcmtΔ/Δ cells (produced by treating K-Ras-Icmtflx/flx cells with a low titer of AdRSVCre and a high titer of AdRSVnlacZ) were injected subcutaneously into nude mice, and tumors were harvested 21 days later. The ratio of the Icmtflx band to the IcmtΔ band in the injected cell mixture was 1.1 ± 0.2, reflecting nearly equal numbers of Icmtflx/flx and IcmtΔ/Δ cells (Figure (Figure4).4). In the tumors, the ratio of the Icmtflx band to the IcmtΔ band was 15.2 ± 2.2 (Figure (Figure4,4, n = 10 tumors), indicating that it was mainly the Icmtflx/flx cells that contributed to tumor growth.
A GFP–K-Ras fusion protein was mislocalized in Icmt-deficient fibroblasts (Figure (Figure5a).5a). Also, a large portion of the K-Ras in K-Ras-IcmtΔ/Δ fibroblasts was present in the soluble (S100) fraction of cells, whereas nearly all of the K-Ras in the parental K-Ras-Icmtflx/flx fibroblasts was associated with the membrane (P100) fraction (Figure (Figure5b).5b). Notably, inactivation of Icmt in the K-Ras-Icmtflx/flx fibroblasts invariably caused about a twofold increase in steady-state levels of total Ras proteins and GTP-bound Ras proteins within cells (Figure (Figure5,5, b and c), while the ratio of total to GTP-bound Ras remained constant (not shown). The higher levels of the Ras proteins could not be explained by higher levels of mouse K-Ras or human K-Ras mRNA (Figure (Figure5d)5d) and are apparently due to a decreased turnover of Ras in these cells (see later).
The inactivation of Icmt in nontransfected cells (i.e., no activated K-Ras) did not affect serum growth factor–stimulated or EGF-stimulated phosphorylation of the downstream Ras effectors, Erk1/2 and Akt1 (Figure (Figure6,6, a and b). These results indicate that the absence of carboxyl methylation of CAAX proteins in Icmt-deficient cells has prominent effects on Ras targeting to membranes and on the steady-state levels of Ras proteins in cells, but apparently little or no effect on the growth factor–stimulated activation of Erk1/2 and Akt.
The Rho GTPases play important roles in cell-cycle progression and cell transformation (33–35). Rho proteins are CAAX proteins that, like the Ras proteins, undergo Icmt-mediated methylation. We hypothesized that the slow growth of IcmtΔ/Δ fibroblasts and their resistance to K-Ras transformation might be accompanied by a defect in Rho metabolism and signaling. Indeed, our data suggest that this was the case. The steady-state levels of total RhoA and GTP-bound RhoA in K-Ras-IcmtΔ/Δ fibroblasts were only about 5–10% of those in the parental K-Ras-Icmtflx/flx cells (Figure (Figure7,7, a and b, P < 0.001). This finding was clearly due to absence of Icmt activity, since K-Ras-IcmtΔ/Δ fibroblasts expressing the human ICMT cDNA exhibited normal levels of GTP-bound RhoA (Figure (Figure7a).7a). The low levels of RhoA protein in K-Ras-IcmtΔ/Δ fibroblasts were not accompanied by a comparable reduction in the levels of RhoA mRNA (Figure (Figure77c).
To determine if accelerated Rho turnover was responsible for the low levels of RhoA in the K-Ras-IcmtΔ/Δ fibroblasts, we used metabolic labeling/pulse-chase experiments to assess the half-life of Rho proteins in cells. In four separate experiments, Rho proteins disappeared more quickly in the K-Ras-IcmtΔ/Δ cells than in the parental K-Ras-Icmtflx/flx cells (Figure (Figure8,8, a and b) (Rho half-life, 22.0 ± 9.4 hours in K-Ras-Icmtflx/flx cells and 2.8 ± 0.4 hours in the K-Ras-IcmtΔ/Δ cells). Notably, the absence of methylation retarded Ras turnover within cells (Figure (Figure8c)8c) (Ras half-life, 13.9 ± 6.1 hours in K-Ras-Icmtflx/flx cells and 32.5 ± 6.0 hours in K-Ras-IcmtΔ/Δ cells), probably explaining the differences in steady-state levels of Ras proteins in Figure Figure5,5, b and c. Thus, the absence of carboxyl methylation is associated with accelerated turnover of the Rho proteins but retarded turnover of the Ras proteins.
The regulatory protein p21Cip1 binds to cyclins and inhibits cell-cycle progression (35, 36). p21Cip1 levels are upregulated by overexpression of activated Ras, but this Ras-induced upregulation can be antagonized by RhoA (32, 34). In light of the low levels of RhoA in K-Ras-IcmtΔ/Δ fibroblasts, we hypothesized that those cells might exhibit higher levels of p21Cip1. Indeed, p21Cip1 levels in K-Ras-IcmtΔ/Δ cells were more than fourfold higher in K-Ras-IcmtΔ/Δ cells than in the parental K-Ras-Icmtflx/flx cells (Figure (Figure9a,9a, P < 0.01). The increase in p21Cip1 protein levels in K-Ras-IcmtΔ/Δ fibroblasts was accompanied by a 1.4-fold increase in p21Cip1 mRNA levels (Figure (Figure9b,9b, P < 0.001). The increase in p21Cip1 protein levels was not observed in K-Ras-IcmtΔ/Δ cells expressing a human ICMT cDNA (Figure (Figure9c).9c). The increased level of p21Cip1 in K-Ras-IcmtΔ/Δ fibroblasts was accompanied by an increased association between p21Cip1 and cyclin A (Figure (Figure9a,9a, middle panel). To test whether signaling through the Erk1/2 pathway was involved in the upregulation of p21Cip1, we treated K-Ras-IcmtΔ/Δ fibroblasts and the parental K-Ras-Icmtflx/flx cells with the MEK inhibitor PD98059. PD98059 decreased p21Cip1 levels in the K-Ras-IcmtΔ/Δ fibroblasts, but it did not affect p21Cip1 levels in the K-Ras-Icmtflx/flx fibroblasts (Figure (Figure9d).9d). This result does not prove that Ras signaling is involved in the upregulation of p21Cip1. However, in nontransfected IcmtΔ/Δ fibroblasts, the levels of p21Cip1 were similar to those in the parental Icmtflx/flx fibroblasts (not shown). These data suggest that increased Ras signaling through the Erk1/2 pathway could be responsible for the upregulation of p21Cip1 in K-Ras-IcmtΔ/Δ fibroblasts.
The finding that p21Cip1 levels were elevated in K-Ras-IcmtΔ/Δ fibroblasts suggested that p21Cip1 might be important for the markedly reduced oncogenic capacity of these cells. To test this hypothesis, we generated p21Cip1-deficient K-Ras-Icmtflx/flx fibroblasts and then used AdRSVCre to produce p21Cip1-deficient K-Ras-IcmtΔ/Δ fibroblasts (Figure (Figure10a).10a). In the absence of p21Cip1, the inactivation of Icmt no longer blocked the ability of K-Ras–transfected cells to grow in soft agar (Figure (Figure10b).10b). Similarly, in the absence of p21Cip1, there was no difference in the growth rate of K-Ras-IcmtΔ/Δ and K-Ras-Icmtflx/flx fibroblasts under routine cell-culture conditions (anchorage-dependent growth) (not shown).
To determine if the effect of the Icmt inactivation was limited to inhibiting the growth of K-Ras–transformed fibroblasts, we analyzed the effect of inactivating Icmt in Icmtflx/flx cells that had been infected with a retrovirus encoding a mutationally activated form of human B-Raf (V599E). B-Raf-Icmtflx/flx fibroblasts grew faster than nontransfected cells, lost contact inhibition, and were capable of forming colonies in soft agar (Figure (Figure11).11). AdRSVCre-mediated deletion of Icmt from B-Raf-Icmtflx/flx fibroblasts (to produce B-Raf-IcmtΔ/Δ fibroblasts) restored contact inhibition and eliminated the ability of the cells to grow in soft agar (Figure (Figure11).11). In addition, B-Raf–transfected Icmt–/– fibroblasts did not form colonies in soft agar, whereas B-Raf–transfected Icmt+/+ fibroblasts did (not shown).
In the current study, we show that inactivation of Icmt in mouse fibroblasts results in slow growth and a striking reduction in oncogenic transformation induced by K-Ras and B-Raf. For the K-Ras–transfected fibroblasts we also provide evidence that the absence of carboxyl methylation results in a rapid turnover of RhoA and an upregulation of the cyclin-dependent kinase inhibitor p21Cip1. These phenotypes were abrogated by the ectopic expression of a human ICMT cDNA. In addition, the finding that deletion of Cdkn1a (the gene encoding p21Cip1) abolished the effects of the Icmt inactivation on cell growth and K-Ras–induced colony formation in soft agar, suggests that the Icmt knockout–mediated inhibition of cell growth and K-Ras transformation requires p21Cip1.
Cre-mediated inactivation of Icmt caused a 90–95% reduction in K-Ras–induced colony growth in soft agar and a reduced capacity to contribute to tumor growth in nude mice. This phenotype is almost certainly due to inactivation of Icmt rather than to an artifact related to adenoviral infection—for several reasons. First, all of the control cells for these experiments were treated with an adenovirus encoding β-gal. Second, the adenovirus treatment was temporally remote from the actual experiments; all of the transformation experiments were performed at least three or four passages—more than 20 cell divisions—after adenoviral infection. Third, Cre-mediated inactivation of Icmt had no detectable effects in cells that had been previously transfected with a cDNA encoding human ICMT. Fourth, we obtained very similar cell growth and transformation results with conventional Icmt-knockout fibroblasts (Icmt–/– cells), which were not treated with adenovirus. Finally, we previously showed that Cre-adenovirus infection, under the conditions used here, did not change the growth rate of fibroblasts (21).
It is widely believed that targeting of Ras proteins to the plasma membrane is very important for downstream signaling cascades. Icmt deficiency caused a striking mislocalization of GFP–K-Ras away from the plasma membrane, as well as a redistribution of Ras proteins to the cytosolic fraction. However, the mislocalization of K-Ras was not accompanied by detectable decreases in growth factor–stimulated activation of Erk1/2 and Akt1. We have made identical observations in Rce1-deficient cells—a striking mislocalization of Ras proteins (21) but no apparent effect on Erk1/2 signaling (E. Kim et al., unpublished observations). We do not yet fully understand why the mislocalization of Ras proteins is not accompanied by detectable changes in signaling, but several possibilities exist. First, growth factor stimulation of Erk1/2 and Akt1 in Icmt-deficient cells may simply be mediated by non-Ras pathways. Second, it is possible that a small percentage of the Ras proteins make it to the plasma membrane in Icmt-deficient cells and that this small amount of normally targeted Ras is sufficient for the signaling “readouts”. Another possibility is that some of the Ras proteins in Icmt-deficient cells are mislocalized to the Golgi or ER membranes and that signaling occurs fairly normally from those locations. This explanation is not farfetched. It was recently shown that Ras targeted to the Golgi or ER was able to recruit Raf-1, activate the Erk pathway, and transform fibroblasts (37).
The fact that growth factor–stimulated Erk1/2 and Akt signaling was normal in the setting of Icmt deficiency prompted us to seek other explanations for the retarded cell growth and the reduced capacity of K-Ras–transfected cells to form colonies in soft agar. The Rho proteins are required for Ras transformation and are sometimes found at increased concentrations in Ras-transformed cells (35, 38). Because the Rho proteins are also substrates for Icmt, we asked whether Icmt deficiency might perturb Rho metabolism. Indeed, this was the case. The steady-state levels of GTP-bound RhoA and total RhoA were lower in the IcmtΔ/Δ fibroblasts than in the parental Icmtflx/flx cells, as a result of accelerated protein turnover. These findings were consistent with earlier results from Backlund (39). He used a combination of pulse-chase studies and two-dimensional gel electrophoresis to show that nonmethylated RhoA is less stable than the methylated form. Notably, the effect of methylation on protein stability can be very different for other CAAX proteins; the Ras proteins in IcmtΔ/Δ cells were actually more stable than the Ras proteins in Icmtflx/flx cells. In the case of the Ras proteins, one might reasonably hypothesize that normal Ras turnover occurs only at certain membrane surfaces, and that changes in Ras partitioning within the cell would have secondary effects on protein degradation. The finding of increased Ras levels is consistent with data from Holstein et al. (40). They found that Ras protein levels increase after mevalonate depletion, which inhibits isoprenylation of CAAX proteins. However, in their experiments, RhoA protein levels also increased. Thus, isoprenylation and carboxylmethylation could have opposite effects on the steady-state levels of RhoA.
Data from Olson et al. (36) suggest that constitutive Ras signaling in the absence of Rho increases p21Cip1 levels, leading to cell-cycle arrest. Microinjection of activated Ras into fibroblasts increased DNA synthesis, whereas the simultaneous microinjection of Ras and a specific Rho inhibitor (C3 exoenzyme) increased p21Cip1 and inhibited DNA synthesis (36). In our studies, the inactivation of Icmt in K-Ras–transfected fibroblasts was associated with a striking reduction in RhoA protein levels and a 4.2-fold upregulation of p21Cip1. We predicted that the upregulation of p21Cip1 might be an important influence in the reduced capacity of Icmt-deficient fibroblasts to grow in soft agar. Indeed, this seemed to be the case. When we performed experiments with K-Ras-Icmtflx/flx cells lacking p21Cip1 (K-Ras-Icmtflx/flxCdkn1a–/–), we found that Cre-mediated inactivation of Icmt had no measurable effect on cell growth on plastic plates or on the growth of colonies in soft agar. Notably, both FTase inhibitors and GGTase I inhibitors have been reported to increase p21Cip1 levels in human tumor cells (41, 42).
Our results indicate that the effect of the Icmt inactivation on K-Ras transformation may have been a consequence of its effects on Rho and p21Cip1 rather than being due to a direct effect on the intrinsic properties of K-Ras itself. To explore this issue further, we examined the consequences of inactivating Icmt in cells that had been transformed with mutationally activated B-Raf, a downstream Ras effector. Activating mutations in B-Raf have been identified in 66% of malignant melanomas (43). B-Raf is not a CAAX protein and hence not a substrate for Icmt, yet the Cre-mediated inactivation of Icmt reversed the transformed phenotypes in B-Raf–transfected cells. This finding is in keeping with the notion that the reversion of K-Ras transformation may not have been due to a direct effect on K-Ras.
The current study supports the idea that Icmt could be a therapeutic target for treating cancer. Perhaps more importantly, however, the current study has created a tool for addressing this issue. With the creation of mice harboring a conditional Icmt allele, it is now possible to test whether a deficiency in Icmt would retard tumor growth in vivo. Several groups have developed “latent oncogenes” that can be activated in vivo by expressing Cre, resulting in the formation of tumors (44, 45). In those mice, it would be of value to determine if the Cre-induced tumor development could be prevented by the simultaneous inactivation of Icmt.
We thank A. Balmain for the Cdkn1a–/– mice, J. Otto for recombinant K-Ras, and G. Howard and S. Ordway for comments on the manuscript. We also thank P.A. Ambroziak, A. George, and C. Gregory for technical assistance during the early stages of this work. This work was supported in part by NIH grants HL41633, RO1 CA099506, and RO1 AR050200 (to S.G. Young), grants from the University of California Tobacco-Related Disease Research Program (to M.O. Bergo and S.G. Young), and the Swedish STINT Foundation (to M.O. Bergo).
See the related Commentary beginning on page 513.
Conflict of interest: The authors have declared that no conflict of interest exists.
Nonstandard abbreviations used: isoprenylcysteine carboxyl methyltransferase (Icmt); fibroblasts harboring a “floxed” Icmt allele and expressing activated K-Ras (K-Ras-Icmtflx/flx); farnesyltransferase (FTase); geranylgeranyltransferase type I (GGTase I); farnesyltransferase inhibitor (FTI); Cre-adenovirus (AdRSVCre); β-gal-adenovirus (AdRSVnlacZ); enhanced GFP (EGFP); mitogen-activated protein (MAP); polyinosinic-polycytidylic ribonucleic acid (pI-pC).