Tumor Genotype-specific Growth Inhibition in Vivo by Antisense Oligonucleotides against a Polymorphic Site of the Large Subunit of Human RNA Polymerase II¹ Free

The central problem in cancer therapy is to kill cancer cells without damaging normal cells. To successfully solve this problem, it is crucial to identify a consistent and absolute difference between cancer cells and normal cells and to have a therapeutic approach that exploits this difference to selectively target and kill cancer cells. In many cases, cancer cells differ from normal cells because they have lost large segments of DNA. The loss of large chromosomal regions, or even whole chromosomes, is an early event in the clonal evolution of cancers. LOH³ can involve >20% of the total genome in certain cancers (1). Hence, the genetic difference between normal cells and cancer cells extends beyond the loss of a tumor suppressor gene. Many genes will be reduced to hemizygosity in cancer cells because of LOH, and some of these genes may be essential for cell survival. This irreversible difference between normal and tumor cells forms the basis of a new approach for anticancer drug development, called ASI.

We have identified a large number of SNPs in genes that are essential for cell survival and are localized in chromosomal regions often involved in LOH in various cancer types (2). Over 80% of these SNPs have a high degree of heterozygous expression in a panel of normal individuals. In principle, these polymorphisms can be targeted selectively by antisense ODNs that will discriminate the two alleles. ASI of the allele of an essential gene left as the only allele in the cancer cells as a result of the LOH will lead to cell death. The heterozygous normal cells, however, will only lose 50% of the expression of the gene by this ODN and, provided that 50% gene activity is enough for cell survival, will not be damaged by this antisense ODN. This novel approach to cancer treatment has been described in greater detail by Basilion et al. (3).

The large subunit of RNA polymerase II (POLR2A) is a potential target gene for ASI because it fulfills three important criteria (4): (a) POLR2A encodes a M_r 220,000 protein, which is the heart of the cellular transcription machinery; the POLR2A gene product is essential for cell survival as has been demonstrated by the use of the mushroom toxin α-amanitin, which binds specifically to the POLR2A M_r 220,000 protein (5, 6); (b) POLR2A contains many SNPs with a high frequency of heterozygous expression in normal individuals (2); and (c) POLR2A is localized on chromosome 17p13.1 in close proximity to p53 (7), a region involved frequently in LOH in many tumor types (8).

We have described the development of a set of antisense phosphorothioate ODNs that can discriminate between the two alleles of a polymorphic site in POLR2A (4). These ODNs were designed to selectively target the matched allele, whereas the mRNA levels of the mismatched allele, which differs in one nucleotide, is only affected slightly. However, a major question left unanswered in our previous in vitro study was whether these phosphorothioate ODNs can arrest tumor cell growth in vivo. Therefore, we set up an in vivo model in which a continuous dose of antisense ODN can be administered systemically to tumor xenografts using Osmotic mini-pumps. We show here that phosphorothioate antisense ODNs directed against POLR2A can inhibit tumor growth in vivo as efficiently as a well-described antitumor antisense ODN against Ha-ras. In addition, we show that single bp mismatches can be sufficient to obtain ASI of tumor growth, giving proof of principle for allele-specific targeting as cancer therapy and demonstrating that the observed effects are true antisense effects.

All phosphorothioate ODNs were purchased from Isogen (Maarssen, the Netherlands) except the ISIS 2503 and 12790 antisense ODNs, which were a kind gift of Variagenics, Inc. (Cambridge, MA). L5Tas17: 5′ TCGCCATAGCGCAGCTG 3′, L5Cas16:5′ TCGCCGTAGCGCAGCT 3′, ISIS 2503 is a 20 mer targeted to the initiation codon (AUG) of c-Ha-ras mRNA: 5′ TCCGTCATCGCTCCTCAGGG 3′ (9). ISIS 12790 is a 20 mer targeted to the M_r 70,000 subunit of replication protein A (3): 5′ TGGTCTGCAGTTAGGGTCAG 3′. Tritium labeling of ODNs was performed using the heat exchange method described by Graham et al. (10).

The pancreatic carcinoma cell line MiaPaca II and the prostate cancer cell line 15PC3 were maintained by serial passage in DMEM. Cells were grown at 37°C and 5% CO₂. Media were supplemented with 10% FCS, 2 mm L-glutamine, 100 units/ml penicillin, and 100 μg/ml streptomycin.

NMRI nu/nu mice( 8–10 weeks old; Charles River, Maastricht, the Netherlands) were injected s.c. in the flank with 10⁶ Miapaca II cells or 10⁶ 15PC3 cells in 300 μl of Matrigel (Collaborative Biomedical Products, Bedford, MA). The cells were injected in ≤1 h after harvesting by trypsin treatment. Before injection, the cells were washed with cold PBS, counted with a hemocytometer, and mixed subsequently with the Matrigel on ice. One week after tumor cell injection, when tumor take was positive, an Osmotic mini-pump (Alzet model 1002; Alzet Corp., Palo Alto, CA) was implanted dorsally according to the instructions of the manufacturer. The Osmotic mini-pumps were incubated in PBS 20 h at 37°C before implantation to start up the pump and quickly reach a steady delivery rate after implantation. In vitro testing showed that the Alzet 1002 mini-pumps reached a steady pumping rate in ≤24 h. The Osmotic mini-pumps were filled with phosphorothioate ODNs (5 mg/kg bodyweight) or 0.9% saline. Tumor growth was monitored daily after the implantation of the Osmotic mini-pump. Tumor volume was measured and calculated as described previously (11). Tissue distribution studies of tritiated ODNs were performed according to Bijsterbosch et al. (12). The radioactivity in the different organs was corrected for serum present at the time of sampling as determined by the distribution of ¹²⁵I-BSA.⁴ Liver samples were fixed in 4% formaldehyde in PBS and embedded subsequently in paraffin according to standard procedures. H&E stains were used to visualize the effects of ODN treatment on the liver. Bilirubin and alkaline phosphatase levels in the serum were determined using standard diagnostic procedures with the H747 (Hitachi/Roche Diagnostics) with the appropriate kits (Roche Diagnostics).

15PC3 tumors expressing GFP were created as described (13) with the following modifications. Enhanced GFP was cloned into the retroviral LNCX2 vector (Clontech), and 70% confluent Phoenix packaging cells were used to generate the viral supernatant (14). G418 (Life Technologies, Inc.) was used to select GFP-expressing 15PC3 cells. The stability of GFP expression was verified by using FACScan analysis. The GFP-expressing 15PC3 cells were injected into NMRI nude mice as described above and were imaged using diffused laser light excitation (488 nm, argon laser, model 2030; Spectra-Physics) and CCD camera (Telecam SL; Karl Storz Gmbh, Tuttlingen, Germany) equipped with a 515-nm longpass filter.

We have described a complete set of ODNs around the SNP at position 2673 of POLR2A (4). The ODNs L5Tas17 and L5Cas16 could discriminate between the two alleles of the targeted SNP in vitro. Fig. 1 shows the sequence around the polymorphic site and the two antisense ODNs tested. A tritiated phosphorothioate ODN was used to monitor the delivery of the ODNs in vivo to the tissues of nude mice bearing 15PC3 or MiaPaca II tumor xenografts 48 h after implantation of the Osmotic mini-pumps. Both liver and kidneys were found to take up most of the recovered amount of radioactivity (15.7 ± 5.7% and 8 ± 0.1%, respectively, for the complete organs). Less than 1% of the radioactivity could be recovered in either the 15PC3 or MiaPaca II xenografts. The specific uptake per gram tissue is shown in Table 1. The liver, kidneys, and, to a lesser extent, the lungs were found to have a relatively high specific uptake per gram tissue. The specific uptake of radioactive ODN by tumor xenografts is similar to the uptake by the skin, muscles, and intestinal tissues.

To test whether the amount of ODN found within the tumor xenografts was sufficient to deliver an effective dose, a well-described and potent antisense ODN against Ha-ras (9) was administered to the tumor xenograft-bearing nude mice. One week after injection of 10⁶ MiaPaca II or 15PC3 cells, the pumps were implanted dorsally delivering a dose of 5 mg/kg/day the anti Ha-ras phosphorothioate ODN (ISIS 2503). Both Miapaca II and 15PC3 tumor growth was inhibited by the treatment with ISIS 2503 as compared with saline and sense (5 mg/kg/day) ODN controls (Fig. 2).

The results with the ISIS 2503 antisense ODN showed that the Osmotic mini-pumps delivered enough antisense ODN to arrest the growth of both the 15PC3 and MiaPaca II tumor xenografts. Next, the anti-POLR2A ODNs (Fig. 1) were tested in this model to determine whether POLR2A is a suitable target for ASI in vivo. Nude mice with either MiaPaca II xenografts, which are T at position 2673, or 15PC3 tumor xenografts, which are C at position 2673, received via the Osmotic mini-pumps either the matched L5Tas17 ODN (5 mg/kg/day) or the one base mismatched L5Cas16 (5 mg/kg/day). Fig. 3 shows the effects of the anti-POLR2A ODNs on tumor growth. The matched L5Tas17 ODN significantly reduced the MiaPaca II tumor growth as compared with the mismatched L5Cas16 and sense (5 mg/kg/day) and saline controls (P < 0.01, two-way ANOVA). In contrast, the ODN L5Tas17 did not inhibit the mismatched 15PC3 xenograft growth at all, whereas L5Cas16 inhibited 15PC3 growth significantly (P < 0.01, two-way ANOVA). These results indicate that the ODNs against POLR2A are also allele specific in vivo and that POLR2A is a suitable target for tumor growth inhibition. Furthermore, these results indicate that the observed effects are true antisense effects, because a single base mismatch determines the efficacy of the antisense ODN.

To exclude the possibility that the external measurements during the experiment of the tumor xenografts were biased or affected by the coinjection of Matrigel or to s.c. fluid depositions, fluorescent GFP-expressing 15PC3 cells were created. When injected s.c. into nude mice, the resulting xenograft is fluorescent and can be imaged externally (15). Only the tumor cells are fluorescent and can be distinguished from other cells and tissues of the mouse. The fluorescence remained localized in the tumor xenograft, and the growth characteristics were similar to the nonfluorescent 15PC3 xenografts (data not shown). Nude mice with fluorescent 15PC3 (C) tumors were treated with the anti-POLR2A ODNs L5Tas17 and L5Cas16 using the Osmotic mini-pumps. The xenografts were imaged after 10 days of treatment. In Fig. 4, two representative fluorescent 15PC3 tumors are shown. The fluorescent 15PC3 tumors treated with the matched L5Cas16 ODN were markedly smaller (90 ± 10 mm³, n = 5) and less fluorescent as compared with tumors treated with the mismatched L5Tas17 ODN (200 ± 20 mm³, n = 5).

However, increasing the dose of ODNs to 10 mg/kg/day resulted in a loss of allele specificity. L5Tas17 and L5Cas16 reduced tumor growth independent of the genotype (data not shown), indicating that allele specificity of these ODNs is dose dependent. Similar observations were done in vitro (4). The POLR2A antisense treatment did not result in any overt visible problems for the mice. During the antisense treatments, the mice were weighted daily. No weight loss occurred during the ODN treatment. Although the POLR2A sequence is highly conserved, the polr2a sequence of the mice differs at one nucleotide position from the human sequence within the targeted mRNA stretch (Fig. 1). Thus, for L5Tas17, there is one mismatch, and for L5Cas16, two mismatches occur with the mouse sequence. Because the biodistribution studies (Table 1) indicated clearly that most ODN is taken up by both the liver and the kidneys and because we have seen that an increase in ODN dosage resulted in loss of the allele specificity, it is feasible that at high concentrations, the ODNs also affect the mouse polr2a mRNA. To investigate this possibility, we examined the effects of ODN treatment on the mouse liver, one of the principle sites of uptake of phosphorothioates in the mouse. No effects could be found on the levels of polr2a mRNA in the mouse liver attributable to treatment with ODNs L5Tas17 and L5Cas16 (Fig. 5). We also examined the histology of the liver after ODN treatment. Only minor and diffuse mixed mononuclear cell infiltrates were discovered within the liver in all mice treated with phosphorothioate ODNs, including the sense controls. These effects are nonsequence specific and probably caused by the phosphorothioate chemistry of the ODNs as they are also described previously (16). In addition, we also examined the effects of antisense treatment on general liver function by measuring the levels of billirubin and alkaline phosphatase in the serum of treated mice. Additionally in this case, no sequence-specific effect could be found after ODN treatment as compared with mice treated with saline and sense controls (data not shown). Therefore, we conclude that the treatment with antisense against the human POLR2A sequence is not overtly toxic for the mice and can effectively inhibit xenograft tumor growth in a genotype selective manner.

Previous studies by Basilion et al. (3) and ten Asbroek et al. (4) have shown that ASI of gene expression is possible in vitro. POLR2A is one of the target genes for which allele-specific ODNs were designed, which discriminate between the two alleles differing only in a SNP (4). We now demonstrate that these ODNs can inhibit the growth of human tumor xenografts in nude mice.

The ODNs were administered continuously through Osmotic mini-pumps. Tissue distribution studies indicate that only a limited amount of the administered ODN reached the tumors. This is in accordance with previous data, which show rapid clearance of phosphorothioate ODNs mainly through high uptake in liver and kidney (12, 17).

To demonstrate the validity of both our tumor model and our antisense ODN delivery system, we targeted tumors of both test cell lines with a well-known antitumor Ha-ras antisense ODN (9). The antisense ODNs directed against POLR2A mediated a similar reduction in growth rate when compared with the Ha-ras antisense ODN in these tumors. Furthermore, the growth of the tumors was only affected by the respective matched ODNs. MiaPaca II tumors, which are homozygous for a T in the target region, were only inhibited in their growth rate by the matched ODN L5Tas17 and not by the single base mismatched L5Cas16. Similarly, 15PC3 tumors, homozygous for a C in the target region, are only affected by the matched L5Cas16 and not by the single base mismatched L5Tas17. Therefore, the antisense ODNs can discriminate between the two alleles of a SNP in POLR2A and cause genotype-specific reduction of tumor growth in vivo. These results indicate that ASI therapy is a potentially viable approach to develop a tumor-specific anticancer therapy and that POLR2A is a good target gene for antisense ODN-mediated inhibition of tumor growth. However, the allele specificity of the ODNs was reduced when the daily dose of ODNs was increased to 10 mg/kg. At this higher dosage, both L5Tas17 and L5Cas16 could inhibit tumor growth similarly, irrespective of the tumor genotype. This is in accordance to our previously published in vitro experiments where allele specificity was shown to be dependent on the ODN concentration (4). The differences in affinity for the targeted sequence as compared with the mismatched sequence are dependent on only a single nucleotide difference.

In view of the variation in tissue uptake, we analyzed the effect of ODNs on the mouse tissue. The mouse polr2a sequence differs in one base from the human sequence in the targeted region of the mRNA. Thus, depending of the human genotype, one or two mismatches are present. We observed no effect of the antisense ODNs on the mice. The liver, which takes up a much larger amount of the phosphorothioate ODN, was not affected in a sequence-specific manner. This may bode well for when ASI is to be used in an authentic therapeutic setting, when there is only a single base mismatch between the tumor and the organism in which it resides. In our model, we also have shown that two different homozygous tumor types are protected by a single base mismatch at the effective concentration. In addition, it must be noted that at least a large fraction of the ODNs taken up by the liver and kidneys might be sequestered from the body in such a way that they are not effective anymore (i.e., degraded or rapidly sequestered to the urine and bile for excretion). Future studies will have to incorporate heterozygous tumor xenograft controls, as well as syngeneic mouse tumor models to obtain full proof of principle for ASI as therapy.

Essential genes, like POLR2A, which are highly conserved through evolution, have been shown to posses SNPs (2) that occur frequently in the human population. Hence, there are enough SNPs within essential genes that are affected by LOH in cancers that may serve as target for ASI. Most of these polymorphisms are silent and do not cause changes in the peptide sequence of the protein. This makes it difficult to find allele-specific inhibitors that act on protein polymorphisms that occur frequently enough within the human population to be of any therapeutic use. Therefore, we chose to use antisense ODNs to target SNPs. Antisense ODNs have been proven to inhibit gene expression specifically both in vitro and in vivo and have entered clinical trials or are already registered as a drug (18). The most important problem for ASI remains the effective discrimination of the two alleles of a SNP. For each target SNP, this will be a process of trial and error, because both the antisense ODN and the target mRNA will influence this process. Shen et al. (19) showed that SNPs can cause marked changes in the secondary structure of an mRNA. This may lead to an altered accessibility of the mRNA for antisense ODNs and, therefore, affect the allele specificity of an ODN. For the POLR2A target SNP presented in this study, we found that shortening the ODN, as well as shifting the SNP position within the ODN, were beneficial for the allele specificity (4). The number of antisense ODNs that can be used for ASI is limited, because the target is limited to a specific region containing the SNP. Therefore, we envisage that ASI might not be feasible for every potential target SNP. Our studies were performed using simple phosphorothioate antisense ODNs. Novel antisense chemistries have become available recently (20, 21). These advanced chemistries might further increase the allele specificity and efficacy of ODNs. In the future, ASI therapy might be an attractive addition to current chemotherapies, because an absolute genetic difference between cancer cells and normal cells is targeted. The fact that essential genes are targeted in this approach may even further increase the efficacy of the treatment. Cancer cells cannot retrieve the allele that is lost by LOH during tumorigenesis; therefore, the targeted allele can not be replaced.

We thank Dr. M. Kool for technical assistance in creating the GFP-expressing cells. We thank Dr. S. D. Troost and J. Aten for evaluation of the tissue sections. We also thank Drs. P. Borst and D. Housman for their comments on the manuscript.