Inhibition of Kirsten-ras Expression in Human Colorectal Cancer Using Rationally Selected Kirsten-ras Antisense Oligonucleotides¹

Introduction

The transformation of normal cells into malignant derivatives is a multistep process that requires a number of key events including the acquisition of self-sufficiency in growth signals (1). This can result from autocrine growth factor stimulation, overexpression of growth factor receptors, or overexpression or activation of the cytoplasmic components of these signaling pathways. One promising therapeutic strategy for treating cancer is to target these defects. Small molecule inhibitors of some of the key signaling molecules, such as Mek, receptor tyrosine kinases, PI3K,³ and the Ras family of guanine-nucleotide-binding proteins are currently under development (2).

Members of the Ras family serve as key effectors of growth factor-induced differentiation, proliferation, death, shape, and motility (3-5). The ras genes encode GTPases that function as molecular switches. Activity is regulated by many factors including distinct GEFs that promote the activation of GDP-bound Ras to GTP-bound Ras by exchanging GDP for GTP and by GTPase activating proteins that promote GTP hydrolysis and inactivation of Ras. Recruitment of Ras by GEFs after the binding of extracellular ligands to cell surface receptors results in the activation of signaling effector molecules such as Raf, PI3K, RalGDS, and AF6 (6).

Mutations of ras usually occur at codons 12, 13, and around codon 61, and are detected in ∼30% of all cancers (7). These mutations result in the constitutive activation of Ras by influencing affinity for GTP, the intrinsic GTPase activity, and the interaction of Ras with accessory proteins that regulate GTPase activity. Three isoforms of the ras oncogene, Kirsten- (Ki-), Harvey- (Ha-), and Neuroblastoma- (N-), have been identified; of these, Ki-ras is the most frequently mutated ras isoform in human cancers, although N-ras is sometimes mutated in hematological malignancies and mutated Ha-ras is occasionally detected in gastric and bladder carcinomas. (3, 7, 8). These isoforms are almost ubiquitously expressed; however, mutations of different ras isoforms are associated with different cancers, which suggests that each isoform could have a distinct biological role. Support for this hypothesis includes the fact that Ki-ras, but not N- or Ha-ras, is essential for mouse embryogenesis (9, 10) and that the different isoforms may interact with different GEFs or activate distinct effectors (11, 12). In addition, the ras homologues also differ in their ability to drive proliferation, to transform cells, to stimulate cell motility, and to regulate gene expression (13-16).

Ki-ras mutations occur early in colorectal cancer tumorigenesis (17), and we have previously demonstrated that acquisition of mutated Ki-ras significantly increases the risk of recurrence and reduces overall survival (8, 18). Evidence of increased signaling through the Ras→Erk pathway also comes from the observation that activated phosphorylated Erk1/2 is increased in a high proportion of tumor cell lines and primary cell cultures derived from the colon (19). This suggests that targeting Ki-ras activity or expression may be of benefit in the treatment of colorectal cancer.

Although the exact contribution of Ki-ras to colorectal tumorigenesis is still not entirely clear, several approaches have been adopted to elucidate the consequences of inhibiting Ki-ras function. One has been to produce Ki-ras overexpression by gene transfer. Studies of this type have generally been restricted to primary cell lines, such as fibroblasts, which do not accurately reflect cancer cell biology (20, 21), and the abnormally high levels of expression often obtained may obscure subtle functional differences by overwhelming normal regulatory mechanisms (13). An alternative approach is to inhibit Ki-ras expression or activity. Exogenously administered ribozymes specific to the valine-12 point mutation have been used effectively in vitro, but these molecules are large (of the order of 40 nucleotides), and their efficacy in vivo is unknown (22). Construction of stable cell lines expressing Ki-ras antisense genes or ribozymes or replacing the mutated gene with a wild-type gene by homologous recombination is another possibility. These cell lines can define the contribution of Ki-ras to tumor formation, but they cannot be used to assess the consequences of Ki-ras inhibition in established tumors because they fail to grow, or at best grow very slowly, in vivo (23-28).

A better way to explore the Ki-ras function in established tumors may be to block its activity with a small molecule inhibitor or to decrease its expression using antisense oligonucleotides. Inhibition of Ras function by blocking essential posttranslational modification with farnesyl (FTIs) and geranylgeranyl transferase inhibitors (GGTIs) is currently undergoing clinical evaluation (29). However, these agents also affect other prenylated proteins involved in malignant transformation (30, 31). A more appropriate strategy might be to inhibit Ki-ras expression using a specific antisense oligonucleotide strategy. There are a number of problems regarding the design and application of antisense oligonucleotides, but, if used appropriately, this strategy can be can effective both in vitro and in vivo. (32).

To explore the biological role of Ki-ras in colorectal cancer and the consequences of inhibiting its expression, we used an antisense strategy to inhibit Ki-ras expression. The formation of target mRNA/antisense oligonucleotide base-paired hybrid inhibits translation of the target mRNA through a number of mechanisms, including activation of RNase H, a ubiquitous intracellular enzyme. RNase H cleaves the RNA strand of an oligonucleotide/RNA duplex leaving the antisense oligonucleotide free to bind to additional complementary mRNA (33, 34). However, an antisense oligonucleotide approach can be difficult, with many factors including oligonucleotide stability, uptake into the correct intracellular compartment, and RNA structure influencing its efficacy and specificity (34).

There are a number of reports of Ki-ras antisense (16, 35-39); however, we decided initially to identify a reagent that effectively inhibited Ki-ras expression. Empirical screening of oligonucleotides has generally had more success than predictions derived from computer modeling (40). Therefore, we screened a library of oligonucleotides against Ki-ras mRNA in vitro. Having identified an effective and specific Ki-ras antisense oligonucleotide that targeted a site accessible to oligonucleotide and RNase H, we then used this oligonucleotide to investigate the role of Ki-ras in colorectal cancer cell growth, survival, and secretion of angiogenic factors. Ras isoforms have been demonstrated to activate selective signal transduction effectors in COS cells. Here we demonstrate for the first time in intact human tumor cells that specific inhibition of Ki-ras expression results in inhibition of Erk1/2, but not c-Akt, phosphorylation. Perhaps surprisingly, there were no major effects on proliferation or cell cycle distribution or death. However, significantly decreased VEGF-A₁₆₅ secretion was detected after Ki-ras antisense treatment, suggesting the potential for antiangiogenic effect. In addition, these molecules could also have potential as therapeutic agents. Finally, we have used the novel combination of Ki-ras antisense with a gene-array approach to profile gene expression changes in colon cancer cells after the inhibition of Ki-ras expression. The development of such a specific Ki-ras antisense oligonucleotide will allow us to ask additional questions about the biological effects of Ki-ras activation in tumors, and may help us to determine how potential Ki-ras inhibitors will exert their effects.

Materials and Methods

Oligonucleotides. A random sequence 17-mer phosphodiester oligonucleotide library (N₁₇) and a series of individual 17-mer phosphorothioate oligonucleotides were synthesized and reverse-phase high-performance liquid chromatography-purified (Sigma Chemical Co-Genosys Ltd., Cambridge, United Kingdom). Oligonucleotide purity and concentration were confirmed by spectrophotometry at A_260/280 and by 5′ end-labeling using T4 polynucleotide kinase and [γ-³²P]ATP (Amersham Pharmacia Biotech UK Ltd., Little Chalfont, Buckinghamshire, United Kingdom) followed by denaturing PAGE.

RNase H cleavage and Mapping. A Ki-ras exon 1–3 PCR product (39) was used as a template to add an upstream T7 bacteriophage RNA promoter by a further amplification using the following primers: 5′-cagtcagtaagcttcctaaacgactcactatagggagtataaggcctctgaaaatgactgaatataa-3′ and 5′-ctgtcttgtctttgctgatgtttcaataaa-3′. RNA was synthesized from this template by in vitro transcription with T7 bacteriophage RNA polymerase and radio-labeled at its 3′ terminus with [³²P]cytidine bisphosphate using T4 bacteriophage RNA ligase (Amersham Pharmacia Biotech UK Ltd.; Ref. 41).

For RNase H mapping experiments, a master mix of 3′ ³²P-labeled Ki-ras RNA, together with 2 μl/reaction of 5× buffer [700 mmol KCl, 125 mmol Tris (pH 7.6), and 50 mm MgCL₂], Escherichia coli RNase H (Amersham Pharmacia Biotech UK Ltd.) and 40 units/reaction rRNase inhibitor (Promega UK Ltd., Southampton, Hants, United Kingdom) was prepared. The master mix was added to a heat-quenched library of random sequence 17-mer oligonucleotides and incubated at 37°C. The reaction was terminated at 5 and at 15 min by the addition of 5 μl of gel loading buffer (10 m urea, 1.5× TBE [135 mm Tris-borate, 6 mm EDTA (pH 8.0)] 0.015% w/v bromphenol blue, and 0.015% w/v xylene cyanol) and incubated at 68°C for 10 min. Cleavage products were analyzed by electrophoresis on short (50-cm) and extra-long (100-cm) denaturing 5–10% polyacrylamide/7-m urea gels and detected by direct autoradiography. Cleavage sites were identified by comparison with the cleavage products of the same RNA produced by the sequence and structure-specific RNase T₁ under native and denaturing conditions (42). The secondary structure of Ki-ras mRNA was predicted using RNAstructure 3.21 software (43). A similar RNase H cleavage assay was used with individual oligonucleotides, except substrate and cleavage products were imaged and quantified by phosphorimaging using a Molecular Dynamics Storm phosphorimager (Amersham Pharmacia Biotech UK Ltd.).

Treatment of SW480 Colorectal Cancer Cells in Culture with Antisense Oligonucleotides. SW480 cells, derived from a human primary colorectal carcinoma mutant (mut) p53, wild-type (wt) Rb and val-12 Ki-ras), were maintained in nonhumidified, 5% CO₂ atmosphere incubator at 37°C and cultured in DMEM, supplemented with 50 units/ml penicillin, 50 mg/ml streptomycin, 2 mml-glutamine (Invitrogen Ltd, Paisley, United Kingdom) and 10% heat-inactivated FCS (PAA laboratories Ltd., Teddington, Middlesex, United Kingdom). Cells were seeded at 10⁴/ml to achieve 50% confluency on the day of treatment. Cells were treated with oligonucleotides in the presence of cationic lipid (cellFECTIN reagent; Invitrogen Ltd.). A 10-μg/ml solution of cellFECTIN reagent, was made up in serum-free OptiMEM medium (Invitrogen Ltd.), with l-glutamine and 2 mm CaCl₂. A solution of oligonucleotide was prepared in an equal volume of OptiMEM. The two stock solutions were mixed and incubated at room temperature, and the cells were washed twice in PBS. The transfection solution was added to each tissue culture flask, and cells were incubated for 5 h at 37°C; the transfection solution was then removed, and serum-containing medium was added.

Viability of cells was assessed by trypan blue exclusion. Cell cycle distribution was analyzed using propidium iodide staining of DNA content. Cells were harvested and resuspended in 300 μl of PBS, fixed by the addition of 700 μl of ethanol, and stored at 4°C. Fixed cells were recovered by centrifugation at 14,000 rpm for 30 s in a microfuge, and the ethanol was aspirated. The cell pellet was resuspended in PBS containing propidium iodide (40 μg/ml) and RNase A (100 μg/ml) and then incubated at 37°C for 30 min. DNA content of cells was analyzed using a Cytaron Absolute cell sorter at 488 nm and measuring red fluorescence.

RNase Protection Assay and Western Blot Analysis. Single- stranded, high specific-activity antisense RNA probes for human glyceraldehyde-3-phosphate dehydrogenase (gapdh) (exons 5 to 8) and Ki-ras (exons 1 to 3) were prepared by in vitro transcription in the presence of [α-³²P]UTP (41). RNase protection analysis of mRNA was performed directly on cell lysates, 4.5 × 10⁵ cells/assay, using a Direct Protect assay (Ambion, Houston, TX). Protected fragments were separated by electrophoresis on denaturing 5% polyacrylamide/7-m urea gels. RNA was quantified by phosphorimaging.

For determination of protein levels by Western blot, cellular extracts were prepared as described previously (39, 44). Protein and rainbow molecular weight markers (Amersham Pharmacia Biotech UK Ltd.) were separated by electrophoresis on polyacrylamide gels and electrotransferred to Hybond-c nitrocellulose (Amersham Pharmacia Biotech UK Ltd.). Immunoblots were blocked with 5% nonfat milk in TBST [10 mm Tris-HCl (pH7.6), 142 mm NaCl, and 0.1% Tween 20] and then incubated with 4 μg/ml anti-N-ras mouse monoclonal antibody, 4 μg/ml anti-Ki-ras mouse monoclonal antibody (Oncogene Research, Cambridge, MA), 1 μg/ml anti-phospho-erk1/2 rabbit polyclonal antibody, 1 μg/ml anti-c-Akt rabbit polyclonal antibody, 1 μg/ml anti-phospho-c-Akt ser-473 mouse polyclonal antibody (New England Biolabs, Beverley, MA) and 0.5 μg/ml anti-Erk1/2 mouse monoclonal antibody (Sigma Chemical Co-Aldridge, St. Louis, MO). Specific antigen-antibody interaction was detected with a horseradish peroxidase-conjugated antimouse IgG using enhanced chemiluminescence Western blotting detection reagents (Amersham Pharmacia Biotech UK Ltd.). When indicated, blots were stripped using the Re-Blot Western Recycling kit (Chemicon International, Inc., Temecula, CA). Membranes were rinsed in TBST, 5 ml of 1× antibody stripping solution was added, and the membranes were incubated with gentle mixing for 10 min at room temperature, reblocked with 5% nonfat milk in TBST, and then incubated as described above.

Measurement of VEGF-A₁₆₅ Secretion by ELISA. Five × 10⁴ cells/ml were seeded in a 6-well plate and treated with 1 μm oligonucleotide in accordance with the transfection protocol on each of 3 successive days. One day after treatment, the growth media was collected and VEGF-A₁₆₅ levels were determined by ELISA (Quantikine VEGF-A₁₆₅ Immunoassay; R&D Systems, Minneapolis, MN). Each individual sample was measured in duplicate in parallel, with standards ranging from 15.6 to 2000 pg/ml. VEGF-A₁₆₅ levels were plotted as pg/ml media/10⁶ cells.

Gene Arrays. Gene expression was profiled as described by Clarke et al. (44). Briefly, ³³P-labeled single-strand cDNA probes were prepared by reverse transcription of polyadenylated mRNA prepared from cells treated as described above. Radiolabeled cDNA was hybridized to commercially available gene arrays carrying 4132 I.M.A.G.E/LLNL cDNA clones spotted on nylon membranes (GF211; Invitrogen Ltd.). Hybridization signals were detected by phosphorimaging (Storm PhosphorImager, Molecular Dynamics), and an intensity output was produced using Pathways 2.01 image analysis software (Invitrogen Ltd.). All of the array experiments were performed in duplicate. Signals less than 1.5 × the background signal were omitted and the quality of signal for each spot was confirmed by visual inspection.

Results

Identification of Sites Accessible to RNase H and Oligonucleotides. A 462-nucleotide transcript encompassing exons 1–3 of the Ki-ras mRNA was used for initial screening of the random oligonucleotide library containing 4¹⁷ different sequences. Exon 4 was omitted because this exon is subject to alternative splicing that results in Ki-ras mRNAs with an exon 4A or 4B. There is evidence that the 4A and 4B Ki-ras isoforms have different properties and can be differentially expressed, although it is not known which particular Ki-ras splice variant, if any, is important in colon cancer biology (11, 14, 45). The RNA product was small enough to allow identification of cut sites by end-labeling and electrophoresis on extra-long 100-cm gels. The amounts of oligonucleotide library and E. coli RNase H were titrated to achieve cleavage of less than 10% of the input RNA. This ensured that sites of cutting were likely to be primary sites rather than secondary cuts arising from altered RNA structure after an initial cut (41). The final conditions were 0.675 nmol of 17-mer randomers (∼4e⁻⁸ pmol of each possible unique sequence) incubated with 1.6 pmol Ki-ras mRNA and 0.01 units/μl E. coli RNase H.

Fig. 1, A and B, shows an example of the results obtained using the above conditions after 5- and 15-min incubations. There was no cleavage of the target RNA after incubation with the oligonucleotide library in the absence of RNase H. However, after the addition of RNase H, highly accessible regions of RNase H cleavage were apparent between nucleotides 379–385, 392–401, and 413–421, respectively (Fig. 1, A-C). In addition, minor cleavage sites were detected at nucleotides 319, 321, 335, and 338. The predicted secondary structure of Ki-ras mRNA was compact with few single-stranded stretches that might be accessible to oligonucleotides; however, none of the cleavage sites localized to these single-stranded stretches (Fig. 1C). The major cut sites at 379–385 were located around a single nucleotide bulge at a bifurcation of a number of base-paired stems, whereas the sites between 392–401 and 413–421 were localized to a small branched stem-loop, its junction with the main stem structure, and the area surrounding it. A number of minor cuts were also detected adjacent or close to bulge regions toward the end of this major stem.

Oligonucleotide Design and Activity. We had previously noted that, in an in vitro assay screen of 13–18-mers, 17-mer oligonucleotides exhibited the greatest activity but maintained the ability to distinguish between wild-type and point mutant (39). Therefore, to continue the current study, a series of 17-mer phosphorothioate oligonucleotides that were complementary to the accessible sites described above were designed (Fig. 1D). To avoid the potential for nonspecific toxicity, we excluded regions with runs of four or more of the same nucleotide, in this case G_415–418. The minor accessible sites between nucleotides 319–338 were targeted with two oligonucleotides (KR1 and KR2). The major sites of cleavage were targeted with a series of oligonucleotides positioned either centrally or skewed toward the 5′ end or 3′ end of the major cleavage sites (KR3–11). One random sequence oligonucleotide (KRSCR) was used as a control in all subsequent studies. This oligonucleotide had the same base composition as oligonucleotides KR3, KR4, KR6, KR7, and KR8. Even at a single oligonucleotide concentration of 1 pmol, incubated with 0.4 pmol of substrate, it was apparent that the various oligonucleotides showed differences in their efficiency of substrate cleavage (Fig. 1E). The control oligonucleotide, KRSCR, was inactive in this assay, whereas the 11 antisense oligonucleotides had a range of activities from poor cleavage mediated by KR11 to apparently efficient cleavage by KR2, KR3, KR4, and KR8.

To test the in vitro activity of the Ki-ras antisense oligonucleotides, we used SW480 colon adenocarcinoma cells that are homozygous for an activating glycine-to-valine mutation of Ki-ras at codon 12 (39). Uptake of oligonucleotide was enhanced by transfection with a cationic lipid and confirmed by monitoring uptake of a FITC-labeled oligonucleotide. Conditions were optimized to allow maximum uptake while maintaining viability; under these conditions, fluorescence was detected in the nucleus of the majority of cells after treatment with 400 nm labeled oligonucleotide (data not shown; (39). SW480 cells were treated with a single 400-nm concentration of each oligonucleotide, and the expression of Ki-ras and gapdh were measured 24 and 48 h after treatment. Decreased expression of Ki-ras, compared with gapdh control, was detected after 24-h treatment with the antisense oligonucleotides (Fig. 2A) but not after treatment with the control KRSCR oligonucleotide. Forty-eight h after treatment, Ki-ras expression was decreased compared with gapdh, but not to the same extent as at 24 h. Fig. 2B summarizes data for the antisense effect of oligonucleotides KR1–11 from three separate experiments after 24-h treatment. There were significant differences in the degree of activity exhibited by the different oligonucleotides. Oligonucleotides KR1, KR3, KR4, KR5, KR6, and KR8 reduced Ki-ras expression by greater than 50%. The greatest and most consistent reduction in Ki-ras expression was observed with KR4 (decrease of 93% ± 4.6; n = 3; P < 0.01). The site targeted by KR4 encompassed nucleotides 379–385, one of the highly accessible regions. The other oligonucleotides complimentary to this region, KR3, KR5, and KR6, also significantly reduced Ki-ras mRNA levels (P < 0.05; Fig. 2, A and B), suggesting that this region is particularly accessible to antisense oligonucleotides and RNase H.

The ability of KR4, the most active oligonucleotide, to inhibit expression of Ki-ras protein was also tested. The half-life of Ki-ras mRNA is only 4 h (46), whereas the protein half-life is in excess of 24 h (39). Therefore, we measured Ki-ras protein 48 h after oligonucleotide transfection. N-ras expression was measured as a control for nonspecific effects on protein levels. A single treatment with 400 nm KR4 only partially inhibited Ki-ras protein expression compared with N-Ras. This may be related to the observation that Ki-ras mRNA levels start to recover at 48 h (Fig. 2A) and to oligonucleotide degradation, because phosphorothioates have a 12–24-h t_1/2 in tissue culture (32), as well as the long (24-h) half-life of the Ki-ras protein. To combat this, we retreated cells with an additional dose of oligonucleotide 24 h after the first dose. In this case, the retreatment protocol resulted in complete inhibition of Ki-ras expression compared with N-ras. Ki-ras expression was not inhibited in cells treated once or twice with the control KRSCR oligonucleotide (Fig. 2C).

We had previously screened a series of 29 13–17-mers targeted at the homozygous valine-12 point mutation of Ki-ras carried by SW480 cells (39). Although we identified a 17-mer (17.2) that appeared to be selectively active in a cell-free RNase H assay, this oligonucleotide was unable to inhibit Ki-ras gene expression in SW480 cells. The lack of cleavage sites around the mutation site after incubation with the 17-mer randomer oligonucleotide library implied that the mutation site was not particularly accessible to oligonucleotides or RNase H. The cell-free data with a single concentration of oligonucleotide also suggested that 17.2 was not as active as KR4 (data not shown). Direct comparison of the dose- and time-dependence of KR4 and 17.2 cleavage of Ki-ras RNA in the cell-free RNase H assay demonstrated that KR4 was 1.5–2 orders of magnitude more active than was 17.2 (data not shown).

Biological Effects of Ki-ras Inhibition. Having isolated an antisense reagent that could specifically inhibit Ki-ras expression, we analyzed the biological consequences of inhibiting Ki-ras expression in the SW480 human colon adenocarcinoma cell line. SW480 cells were treated with three doses of oligonucleotide over 72 h to ensure complete inhibition of Ki-ras expression [confirmed by Western blotting (not shown)]. The number of viable cells 24 h after the last treatment was reduced slightly after treatment with KRSCR compared with lipid alone (88% ± 4%; n = 3). The viability of cells with KR4 was also slightly reduced (80% ± 4%; n = 3) but was not significantly different from cells treated with KRSCR (P = 0.07). Neither treatment resulted in increased cell death because the small number of floating apoptotic cells was unchanged. We also examined cell cycle distribution 24 and 48 h posttreatment. Fig. 3A shows that the cell cycle distributions from the controls and from the KR4 antisense-treated cells were similar, and there was no evidence for a cell cycle arrest in any specific phase of the cell cycle.

A number of distinct signaling pathways operate downstream of Ras and the different Ras isoforms have been reported to interact with different effectors that regulate distinct pathways (11-14). We examined the activity of the Raf→Erk and PI3K→c-Akt pathways after the inhibition of Ki-ras expression by measuring c-Akt and Erk1/2 phosphorylation using antibodies specific to the phosphorylated forms of these proteins (Fig. 3B). There was no difference in the low levels of phosphorylation of c-Akt on serine-473 after treatment with either KR4 or the control oligonucleotide KRSCR. Phosphorylation of Erk1/2 was slightly decreased by the control KRSCR oligonucleotide, but was completely inhibited by KR4. The Western blots were stripped and reprobed with antibodies that recognize both phosphorylated and unphosphorylated Erk1/2 or c-Akt. The levels of total c-Akt and Erk1/2 were not decreased by treatment with KRSCR or KR4 (Fig. 3B). Cyclin D1 expression is regulated at the transcriptional level by signaling through Ras→Erk and would be predicted to be decreased after the inhibition of Erk1/2 phosphorylation (47). However, expression of cyclin D1 protein was not significantly decreased after treatment with KRSCR or KR4. In view of this observation, we also treated SW480 cells with 10 μm U0126, a Mek inhibitor whose action results in the inhibition of Erk phosphorylation (Fig. 3C). Similar to Ki-ras antisense, inhibition of Erk phosphorylation did not result in decreased cyclin D1 expression (Fig. 3B) and did not significantly inhibit cell proliferation (data not shown).

The acquisition of mutated Ki-ras has been reported to increase VEGF production (48). Therefore, we also measured VEGF-A₁₆₅ secretion by ELISA after treatment with Ki-ras antisense. Treatment with lipid alone or with the control KRSCR oligonucleotide did not effect secretion of VEGF-A₁₆₅ into the tissue culture media (Fig. 3D). Cells were also treated with KR1, KR4, and KR5 antisense; under the conditions used, these oligonucleotides reduced Ki-ras gene expression to 47, 7, and 25% of the control respectively. KR1 had no effect on VEGF-A₁₆₅ secretion, KR5 slightly reduced VEGF-A₁₆₅ secretion to 73% of the control KRSCR oligonucleotide (P = 0.088; n = 3), but only KR4, the most active oligonucleotide against Ki-ras mRNA, significantly reduced VEGF-A₁₆₅ secretion to 56% of the control oligonucleotide (P < 0.001; n = 3).

Gene Expression Profiling after Inhibition of Ki-ras Expression. mRNA was extracted from the cells treated with the control oligonucleotide KRSCR and KR4 and was used for expression profiling by gene array carrying 4132 I.M.A.G.E/LLNL cDNA clones. This ensured that the cells were exposed to as similar conditions as possible, the only difference being the sequence of the oligonucleotides that the cells were exposed to. Direct comparison of cells treated with KR4 to cells treated with KRSCR demonstrated that the expression of the majority of genes was unchanged following the inhibition of Ki-ras expression by KR4 (Fig. 4). Table 1 details genes the expression of which was changed by 2.0-fold or greater. Using this commonly used criterion, KR4 decreased the expression of 18 genes and increased the expression of 15 genes. If the gene array data were valid one would expect Ki-ras to be among the genes the expression of which was decreased by KR4. KR4 decreased Ki-ras expression by 2.4-fold to near background levels; in contrast, N-ras expression was unchanged (1.1-fold) after KR4 treatment (Table 1; Fig. 4). We also examined the expression of Ki-ras protein in an aliquot of cells taken from the same experiment, and we clearly confirmed that, as before, Ki-ras protein expression was also inhibited after KR4 treatment when compared with KRSCR.

We examined by gene array the expression of a number of genes, such as c-fos, cyclin D1, and c-myc, which might be expected to be decreased after the inhibition of Erk1/2 phosphorylation by KR4, and these were unchanged (not shown). Ki-ras activation has been associated with increased VEGF expression (48); although we observed a 44% reduction in the secretion of VEGF₁₆₅, no reduction in VEGF mRNA was detected on the array. However, the expression of a number of other genes of interest did change (Table 1). Genes whose products regulate proliferation or induce cell death (INK4C, MKK7, LTA4H, and TIEG) exhibited increased expression after treatment with Ki-ras antisense KR4. In addition, the expression of three genes the products of which influence protein trafficking and export (KDELR2, CAP2, and RNP24) were increased after the inhibition of Ki-ras expression. The expression of a number of genes the products of which are involved in cell cycle progression (RRM1, PSMD8, CPR2, and CCNC) was decreased after the inhibition of Ki-ras expression. Other genes decreased by KR4 treatment included those the products of which are involved in detoxification (ALDH1 and GSTP1), protein trafficking (ARF3 and ARF5), and transcription (ATF4, ETV5, and CCNC).

Discussion

A number of therapeutic strategies are evolving that take advantage of knowledge of the deregulated signal transduction pathways responsible for driving the tumor phenotype (2). Ki-ras is frequently mutated in colorectal cancers and may be an important therapeutic target, especially because the acquisition of an activating mutation is associated with a poorer outcome (8, 18). A major challenge in the development of therapeutics with specific molecular targets, such as Ki-ras, is that one understands the consequences of specifically inhibiting the target so that ‘on-’ and ‘off-’ target effects can be distinguished. Therefore, understanding the contribution of Ki-ras to the biology of colorectal cancer and the consequences of inhibiting its expression or functional activity is potentially important to the development of anticancer strategies.

Our initial goal was to identify an antisense oligonucleotide that specifically inhibited Ki-ras expression. The specificity of antisense oligonucleotides is influenced by a number of factors, including the delivery of oligonucleotide to the correct intracellular compartment and the accessibility of the oligonucleotide to its target structure. The problem of uptake can be by-passed in cell culture by using cationic lipids to ensure that the oligonucleotides reach their destination. However, the effects of RNA structure are not easily predicted. Only a limited number of studies are available comparing oligonucleotide screening strategies based on computer modeling, cell-free assays, and cell line testing; however, experimental screening approaches rather than predictive strategies seem to show better correlation with cell line data (49). Therefore, we used an oligonucleotide library screening approach to identify regions of Ki-ras mRNA accessible to oligonucleotide and RNase H. The RNA used for this approach encompassed exons 1–3 and, therefore, had 5′ and 3′ termini that corresponded to sequence positions internal to a full-length transcript. The local structure in the vicinity of unnaturally positioned ends may be different from the corresponding sequences in full-length transcripts. These differences may be detected by enzymatic mapping or hybridization affinity and are referred to as end effects (50). The results of our oligonucleotide screening approach showed that the sites accessible to oligonucleotide and RNase H were located in exon 2 in the middle of the 462-nucleotide transcript. Therefore, it was unlikely that our findings were attributable to end effects. In addition, local structural features of RNA tend to be maintained irrespective of length until the size of RNA reaches 100 nucleotides or below (51). Our predicted structure of Ki-ras RNA was compact, and the regions identified as potential targets for antisense were not localized to exclusively single-stranded structures but were located at the junction of two stems and the region around a small stem-loop that branched off from the main stem. These observations again point to the limitations of predicting sites to target with oligonucleotides in the absence of experimental data.

Eleven oligonucleotides encompassing the identified sites were tested. Although based on a single concentration, there appeared to be a range of activities in the in vitro RNase H assay, with oligonucleotides KR2, KR3, KR4, and KR8 being particularly active. The results are consistent with previous observations that shifting the site of oligonucleotide hybridization by as little as one nucleotide can significantly alter activity (34, 39). Oligonucleotides targeting the region of major cleavage in the randomer screen (KR3–KR11) generally exhibited similar activities in the cell-free RNase H assay and the cell-based assay. In contrast, the two oligonucleotides that targeted the sites of minor cleavage exhibited different activities in the two assays. KR2 efficiently cleaved Ki-ras mRNA in the RNase H assay but was inactive in cells, whereas KR1 appeared less active in the RNase assay but showed activity in cells. This suggested that the region encompassing the sites of minor cleavage may exhibit different RNA structure and oligonucleotide accessibility in the cell-free and cell-based assays, whereas the site of major cleavage had a more stable structure that is conserved between the two model systems. Six of the 11 oligonucleotides decreased Ki-ras expression in SW480 cells by greater than 50%. Comparison to the RNase protection assay showed that three of the four oligonucleotides (KR3, KR4, and KR8) particularly active in the RNase H assay also reduced Ki-ras expression in SW480 cells by greater than 50%. Five of the most active oligonucleotides in cells, KR3, KR4, KR5, KR6, and KR7, targeted a major stem that encompassed a four-nucleotide single-strand bulge and two stem junctions. The most effective oligonucleotide, KR4, targeted the four-nucleotide bulge at its 5′ terminus and a stem junction and a single-strand bulge at its 3′ terminus. This oligonucleotide was selected for cellular studies and was shown to completely inhibit Ki-ras protein expression in the treated tumor cells. The control oligonucleotide, KRSCR, was ineffective, which indicated a true antisense effect.

Unexpectedly, given the role of Ras in signal transduction and the frequency and prognostic significance of Ki-ras mutation in human colorectal cancer, inhibition of Ki-ras expression by KR4 did not significantly inhibit proliferation or have any major effects on the cell cycle distribution of cells. Previous studies of cells that were transfected with a Ki-ras antisense gene construct or with oligonucleotides that inhibited Ki-ras expression demonstrated modest and different effects on growth rates. Cell lines with point-mutated Ki-ras were generally more affected, with doubling times being extended by between 1.2- and 1.9-fold, depending on the cell line, whereas cell lines without a point mutation were generally unaffected (27, 28). Another study has shown that SW480 cells transfected with a Ki-ras antisense gene construct has a 2- to 3-fold reduced rate of proliferation (52). ISIS 6957, a 20-mer phosphorothioate oligonucleotide targeted to the 5′-untranslated region of Ki-ras, has been demonstrated to reduce proliferation of MRC-5 diploid human lung fibroblasts ∼1.5- to 2-fold but not the growth of a bladder carcinoma cell line (Ha-ras point mutated, Ki-ras wild type), although in both cases Ki-ras mRNA was decreased (16). Consequently, the minimal change in growth rate that we observed with KR4 is perhaps not as surprising as first considered.

In transfected COS cells, Ki-ras has been reported to be a more effective activator of the c-Raf-1→Erk pathway, possibly by efficiently recruiting c-Raf-1 to the plasma membrane; in contrast, Ha-ras is a more potent activator of PI3K (12). Consistent with this, we show here that treatment with KR4 effectively inhibits Erk1/2 phosphorylation, but does not influence c-Akt phosphorylation. This is the first time that such a difference has been shown in a human tumor cell line. Inhibition of Erk1/2 could be expected to inhibit expression of genes such as cyclin D1 (47); however, cyclin D1 protein was not decreased after the inhibition of Ki-ras expression and Erk phosphorylation by KR4. This may be a reflection of experimental design, because these experiments were measuring constitutive Erk phosphorylation and cyclin D1 under normal growth conditions rather than serum stimulation of starved cells that is frequently used to examine signal transduction pathways. In addition, this may also be related to the observation that cyclin D1 expression can also be regulated posttranscriptionally by signaling through PI3K→c-Akt (53) and may also explain the lack of effects on proliferation and cell cycle after the inhibition of Erk1/2 phosphorylation by KR4. In agreement with this, inhibition of Erk phosphorylation after treatment with the Mek inhibitor U0126 gave similar results to the inhibition of Ki-ras expression by antisense: cyclin D1 levels were unchanged, and there was no significant reduction in cell proliferation. The potential dependence of high constitutive phopshorylation of Erk1/2 on Ki-ras may also explain a previous observation of elevated Erk1/2 phosphorylation in a high proportion of tumor cell lines and primary cell cultures derived from colon cancers (19).

Expression and secretion of VEGF is induced by activation of Ras and phosphorylation of Erk1/2 (54). VEGF-A₁₆₅ secretion was not affected by the control oligonucleotide but was inhibited by treatment with Ki-ras antisense oligonucleotides. The level of inhibition generally correlated quite well with the ability of the antisense oligonucleotide to inhibit Ki-ras expression. KR4, the most active oligonucleotide, was the most effective and significantly reduced secretion of VEGF-A₁₆₅ into the growth media. Two groups have disrupted oncogenic Ki-ras function by homologous recombination to evaluate the likely benefits of therapeutic strategies in vivo. In one study, these cells failed to form palpable tumors after 2 months; in the other study, tumors formed but grew at a significantly slower rate than the controls (23, 24). Inhibition of Ki-ras expression by antisense gene constructs also slowed or prevented tumor formation (26-28). Especially given the generally modest effects of Ki-ras inhibition on cell proliferation seen here and elsewhere, one contributing factor could be reduced vascularization, inasmuch as the targeted disruption of Ki-ras reduced VEGF mRNA and protein by 4- to 5-fold (54), whereas another study has noted a 2- to 3-fold reduction in the secretion of VEGF associated with reduced tumor formation (48). Gene transfer of VEGF into these cells partially rescued the ability to form tumors in vivo (48).

To obtain a more global understanding of the consequences of inhibiting Ki-ras with the specific antisense oligonucleotide, we also used microarrays to profile gene expression after treatment with KR4 and the control KRSCR oligonucleotide. Ki-ras gene expression served as an internal control for the array data and was decreased by KR4 treatment compared with the KRSCR control oligonucleotide. The expression of the majority of genes remained unchanged. However, a few changes of potential biological significance were apparent.

Of particular interest were two groups of genes, one group encoding products involved in drug resistance/activation and another group encoding products involved in protein processing and trafficking. ALDH1 and GSTP1 were reduced by Ki-ras antisense. The products of these genes, aldehyde dehydrogenase 1 and glutathione-S-transferase π are involved in detoxification and have also been reported to be involved in drug resistance (55-59). Ki-ras antisense also induced the transcript of another gene, UP; the product of this gene, uridine phosphorylase, is involved in the activation of 5FU, an agent commonly used in the treatment of colon cancer (60). Therefore, Ki-ras activation may contribute to decreased sensitivity or resistance to commonly used chemotherapeutic agents. Although little is known of the role, if any, of mutant Ki-ras in resistance to 5FU or chemotherapeutic agents per se.

The group of genes that encoded products involved in protein export and processing included decreased expression of ARF-3 and ARF-5, the products of which are GTP-binding proteins involved in trafficking. KDEL2 and RNP24 expression were increased by KR4; these genes encode a protein required for retrieval of endoplasmic reticulum proteins from the Golgi and a protein involved in the export of proteins from the endoplasmic reticulum to the Golgi, respectively. In addition, the transcript for CAP2 was also increased; this gene encodes an inhibitor of the endoproteinase furin, a protein found in the Golgi that has a role in processing growth factors and membrane proteins. The absence of functional furin results in the loss of proteolytic processing of integrins α6, α3, and αV in LoVo colon cancer cells (61). Coincidentally, inactivation of Ki-ras by transfection of antisense gene constructs into SW480 has been reported to increase the presence of integrins α1 and α5 and to decrease α3 and αV at the cell surface (52). Similarly, integrin β1 fails to undergo complete maturation in the Golgi of colon cells transfected with activated Ki-ras, but not activated Ha-ras (62). Thus, activation of Ki-ras may change the adherence properties of the cell by altering the processing of key adherence molecules that results in reduced adherence, polarization, and differentiation (62). Also of note, we detected decreased expression of a gene that encoded a non-integrin, high-affinity, laminin receptor after antisense treatment. The product of this gene is frequently overexpressed in colon tumors and may contribute to tumorigenesis and metastasis (63).

One inconsistency that we noted was that, although VEGF-A₁₆₅ secretion was reduced by Ki-ras antisense, there was no corresponding decrease in VEGF-A mRNA expression in the microarray experiment. A somewhat similar observation was made by Feldkamp et al. (64), who noted that treatment of hypoxic astrocytoma cells with L-744,832, a farnesyltransferase inhibitor, resulted in the inhibition of the secretion of VEGF-A but did not inhibit expression of VEGF-A at the mRNA level. Previous experiments have shown by Northern blotting that VEGF transcription is regulated not only by Ki-ras (48) but also by other Ras isoforms, such as Ha-ras (48, 54, 65), as well as a number of ras-independent pathways (66). Therefore, our failure to detect decreased VEGF-A mRNA may be because the change in mRNA levels that result in a 2-fold decrease in VEGF-A secretion may be below the detection threshold of the microarray. Alternatively, because we detected an altered expression of genes the products of which regulate protein trafficking and processing, it is possible that altered protein processing may also explain our observation of decreased levels of VEGF-A₁₆₅ that were detected in the culture supernatant in the absence of decreased VEGF-A mRNA as measured by array. Consistent with the latter hypothesis is the observation that the intracellular pool of VEGF has been localized to secretory vesicles, Golgi, and rough endoplasmic reticulum, and, specifically, VEGF-A₁₆₅ is localized to the Golgi apparatus (67, 68). In addition, a recent study has shown that the inhibition of the expression of orp150, a novel endoplasmic reticulum chaperone, resulted in the inhibition of VEGF secretion (69).

In addition to failing to detect the expected decrease in VEGF-A expression at the RNA level, we did not detect altered expression of genes such as cyclin D1, c-fos, c-myc, and others that might be expected to decrease after the inhibition of Erk1/2 phosphorylation. Interestingly, two other studies (15, 70) have investigated the relationship between gene expression profile and Ras activation and have also made similar observations. One used overexpression of different Ras isoforms in fibroblasts and identified Ras-regulated gene expression by subtractive suppression hybridization (15). This approach identified ∼400 genes that were regulated by Ha-ras, of which the majority were also regulated by N-ras and Ki-ras. Another used differential display of a pancreatic line transfected with a Ki-ras antisense gene construct; in this study only a few genes were altered by inhibition of Ki-ras expression (70). However, neither of these two screening studies detected the expected changes in genes such as cyclin D1, fos, myc, or VEGF, despite demonstrating altered Erk1/2 phosphorylation (15, 70). Other studies have identified specific individual genes regulated by Ki-ras (71-73). Comparison of all of these studies, including our own, indicates little if any overlap between the genes identified as Ki-ras regulated. These differences may be a reflection of the different methodologies used to assess gene expression; or, alternatively, inasmuch as Ki-ras has been shown to have different biological effects in different cell lines, the differences in expression may result from the use of different cell types (14). Of note, however, two genes for which we detected altered expression, uridine phosphorylase and glutathione-S-transferase π, have previously been demonstrated to be differentially expressed between normal colon epithelium and colon tumors or cell lines using SAGE methodology.⁴

In conclusion, we have used a simple oligonucleotide screening strategy to identify an effective antisense reagent for inhibiting Ki-ras expression and have investigated some of the biological consequences of Ki-ras inhibition in a human colon cancer cell line that harbors a Ki-ras mutation. Complete inhibition of Ki-ras expression did not significantly inhibit proliferation, and there were no discernible cell cycle effects. However, for the first time, we show that the inhibition of Ki-ras expression in a human tumor cell line resulted in an inhibition of Erk1/2, but not c-Akt, phosphorylation. We also demonstrated reduced secretion of VEGF-A₁₆₅; this effect correlated with the ability of different antisense oligonucleotides to inhibit Ki-ras expression. We have used the novel combination of antisense inhibition coupled with gene expression profiling. These experiments detected changes in a number of genes that would favor the malignant phenotype. Of particular interest for future study were changes in genes that encode products involved in the trafficking of proteins, including growth factors and proteins involved in adherence. The data presented here were limited to adherent cell growth on a plastic tissue culture dish. Future studies will investigate the role of Ki-ras in spheroid growth and basement membrane adherence. The ability to specifically inhibit the expression of different isoforms of the ras family will also allow comparative studies of Ki-, N- and Ha-ras inhibition. In addition, colon tumors have been reported to express more of the Ki-ras4A splice variant relative to the more ubiquitously expressed 4B splice variant (11, 14, 45). This may be important, because the different splice variants may also have different functions; for example, Ki-ras4B has been reported to be a very effective raf-1 activator when compared with Ki-ras4A (11, 12, 14). In addition, Ki-ras4A, but not Ki-ras4, efficiently induces transformed foci and enables anchorage-independent growth, whereas Ki-ras4B, but not Ki-ras4A, induces cell migration (14). This suggests that Ki-ras4A and Ki-ras4B may interact with different effector proteins. An antisense strategy would be the ideal way to specifically inhibit the expression of these different splice variants and investigate their function in colon cancer.

In conclusion, our results suggest that at least in SW480 human colon cancer cells, constitutive phosphorylation of Erk1/2 2 requires Ki-ras. However, the inhibition of Ki-ras expression does not appear to affect cell proliferation. Nonetheless, strategies targeted to Ki-ras, including antisense, may still be effective via inhibition of biological processes such as angiogenesis, invasion, and metastasis.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

A, 8% polyacrylamide/8-m urea gel analysis ofki-ras RNA, incubated for 5 and 15 min with the optimal concentration of randomer oligonucleotide library and E. coli RNase H. The controls were incubated with randomer oligonucleotides in the absence of RNase H. Cut sites were identified by parallel analysis with sequence and structure specific RNases (not shown). B, long 100-cm 5% polyacrylamide/8-m urea gel analysis magnifying the accessible regions identified in section A. RNaseT₁Lane, incubation with RNase T₁ at 68°C for 5 min. C, mapping of RNase H cut sites on the computer predicted secondary structure of Ki-ras mRNA; only the relevant region of Ki-ras mRNA is shown. D, location of target sites for oligonucleotides KR1–KR11 on the predicted secondary structure of Ki-ras mRNA. E, 8% polyacrylamide/8-m urea gel analysis of the cell-free assay examining cleavage of 3′-end-labeled Ki-ras mRNA by RNase H and the control oligonucleotide KRSCR and Ki-ras antisense oligonucleotides.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

A, SW480 cells were treated for 24 h with KRSCR or KR1–11. Expression of Ki-ras and gapdh were measured by RNase protection at 24 and 48 h. B, summary of three individual experiments of Ki-ras expression after treatment of SW480 cells for 24 h. ∗, P < 0.05, ∗∗, P < 0.001. C, Western blot analysis of Ki-ras and N-ras expression at 72 h after single (1X) and multiple (2X) treatment of SW480 cells with KR4 or KRSCR.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

A, cell cycle analysis of SW480 cells 24 and 48 h after treatment with 1 μm control KRSCR or antisense KR4 oligonucleotides on 2 consecutive days. B, SW480 cells treated with control KRSCR or antisense KR4. Western blot analysis of Erk1/2 phosphorylation and c-Akt phosphorylation at serine-473. Blots were stripped and reprobed for total Erk1/2, c-Akt, and cyclin D1. Increased total Erk1/2 in KR4-treated cells was attributable to increased total protein loading as cyclin D1 levels were also increased; this effect was not observed after treatment with the control oligonucleotide or U0126 Mek inhibitor. C, SW480 cells were treated with 10 μm U0126 for 24 h. Expression of cyclin D1, phosphorylated and total Erk1/2 were analyzed by Western blotting. D, analysis of VEGF-A₁₆₅ secretion by ELISA 24 h after treatment of SW480 cells with 1 μm KRSCR, KR1, KR4, or KR5 oligonucleotides on 3 consecutive days.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Gene expression profiling of SW480 cells treated with 1 μm control KRSCR or antisense KR4 on 2 consecutive days. A, plot (arbitrary phosphorimaging units) of KRSCR (X axis) versus KR4 (Y axis). B, gene array data from cells treated with control KRSCR was overlaid on data from cells treated with KR4 and pseudocoloured. Red, genes decreased by KR4; green, genes increased by KR4, yellow, genes not changed by KR4; brown, expression not detected. Spots corresponding to Ki-ras and N-ras genes are indicated. Control genomic DNA spots (tgDNA) are also indicated. C, Western blot analysis of Ki-ras after antisense treatment of SW480 cells.