Lack of Efficacy of the Statins Atorvastatin and Lovastatin in Rodent Mammary Carcinogenesis

The statins reduce serum cholesterol by inhibiting the upstream enzyme 3-hydroxy-3-methylglutaryl CoA (HMG-CoA) reductase. Recently, there have been epidemiologic reports that this class of agents may reduce the incidence of cancer (1–3). Although perhaps the most convincing epidemiologic data have been associated with colon cancer (2) and prostate cancer (3), there are data implying that statins may also alter the incidence of breast cancer (4–6). One early hypothesis was that by inhibiting cholesterol synthesis, the statins would decrease the production of farnesyl PPi and geranyl PPi (7). These metabolites are used in the prenylation and activation of a wide variety of proteins. Included among the prenylated proteins are the known oncogenes Ha Ras and Ki Ras, and proteins such as RHO A,B,C and CEBPE/F. Many of these are either known to be or are hypothesized to contribute to the oncogenic process (8). In addition, the statins have been shown, at least in vitro, to increase cellular levels of the two cyclin-dependent kinase inhibitors p21 and p27, which diminish cell proliferation (7).

The methylnitrosourea (MNU)–treated Sprague-Dawley rats develop multiple estrogen receptor–positive (ER+) cancers that have a histopathology (9) and gene expression profile (10) similar to highly differentiated ER⁺ breast cancer in women. Tumors in this model have proven to be highly susceptible to a wide variety of antihormonal agents (e.g., tamoxifen, aromatase inhibitors) that are effective against human ER⁺ breast cancer. In addition, the cancers are susceptible to other agents such as retinoid X receptor (RXR) agonists (11–13) and epidermal growth factor receptor inhibitors (14) that may not work directly by inhibiting the hormonal axis. Unlike human breast cancer where Ras mutations are rarely observed, approximately half of the MNU-xinduced cancers have Ha Ras mutations (15). Tumors with Ha Ras mutations are particularly sensitive to the preventive and therapeutic effects of the farnesyl transferase inhibitor tipifarnib (16). Because statins should decrease prenylation, the presence of Ha Ras in 50% of MNU-induced mammary tumors (15) should have made this model particularly sensitive to prevention by statins.

As indicated earlier, chemically induced cancers in rats are primarily ER⁺ (17), whereas many of the mammary cancer models in transgenic mice yield ER⁻ tumors. The mouse mammary tumor virus (MMTV)-Neu transgenic model uses overexpression of NEU under the control of a MMTV promoter to induce the development of multiple ER⁻ mammary carcinomas. However, the majority of the resulting tumors have a mutation in the transmembrane domain in Neu unlike most human Neu-expressing tumors (18). By making a bitransgenic animal that expresses the MMTV-Neu transgene and that also has an alteration in the p53 tumor suppressor gene, the resulting tumors overexpress Neu and have an altered p53 (19). However, the resulting tumors do not have a mutation in the transmembrane domain. This lack of mutation in Neu and an alteration in p53 are characteristic of human ER⁻ tumors that overexpress Neu.

In the present experiments, the ability of two structurally distinct statins (lovastatin and atorvastatin) to prevent the development of ER⁺ cancers in rats was evaluated. The statins were tested either alone or in combination with suboptimal doses of tamoxifen or the RXR agonist bexarotene. Both of these agents have proven to be highly effective in inhibiting tumor development (multiplicity and incidence) in this model when used at higher doses. In addition, we examined the ability of the statins to modulate serum triglyceride levels in normal rats as well as in rats treated with bexarotene. Finally, the efficacies of atorvastatin and bexarotene to prevent ER⁻ mammary cancers in heterozygous MMTV-Neu^+/−/p53KO^+/− bitransgenic mice were also determined.

Female Sprague-Dawley rats were obtained from Harlan Sprague-Dawley, Inc., at 4 wk of age. The bitransgenic (MMTV-Neu^+/−/p53 KO^+/−) female mice were generated in the Chemoprevention Center at the University of Alabama at Birmingham. The p53-deficient line p53N5-T was purchased in a C57BL background (Taconic) and backcrossed at least five times onto a FBV/N background (The Jackson Laboratory). MMTV-Neu transgenic mice [strain FVB/N-Tg (MMTVneu) 202Mul/J] were purchased from The Jackson Laboratory. For generation of MMTV-Neu^+/−/p53KO^+/− females, p53^−/− males were crossed to MMTV-Neu^+/+ females (20, 21). All animals were housed in groups of five per cage in a room maintained at 22 ± 2°C and artificially lighted 12 h/d. Atorvastatin, lovastatin, and bexarotene were obtained from the National Cancer Institute Prevention Repository. Tamoxifen was purchased from Sigma Chemical Company, and MNU was from the National Cancer Institute Chemical Carcinogen Repository. All agents evaluated for prevention activity were incorporated into the diet by mixing with mash feed using a liquid-solid blender with intensifier bar (Patterson-Kelly).

When rats were 50 d of age, they were injected i.v. (via the jugular vein) with MNU (75 mg/kg body weight) as previously described (13). Rats (15 per group) were given the statins, bexarotene, and/or tamoxifen (either as a single agent or in combination) beginning 5 d after MNU. Rats were weighed weekly, palpated for mammary tumors twice per week, and checked daily for signs of toxicity. The studies were terminated 126 d after MNU. Mammary tumors were removed at necropsy of the rats and examined histopathologically as previously described (9).

During the study that evaluated atorvastatin, blood (0.5 mL) was collected from the jugular vein of anesthetized rats (5 per group per time period) at 63, 94, and 126 d after the administration of the carcinogen to determine serum triglyceride levels. For the study that evaluated lovastatin, blood was collected only at termination of the study. In both studies, blood was collected from rats not receiving the carcinogen. After centrifugation of the blood, the serum was frozen at −85°C until it was analyzed for triglycerides (22). The Infinity triglyceride assay kit was purchased from Thermo DMA.

Female MMTV-Neu^+/−/p53KO^+/− mice were placed on a diet supplemented with atorvastatin (200 mg/kg diet) or bexarotene (250 mg/kg diet) beginning at 60 d of age. The numbers of mice of group were as follows: controls, 31; atorvastatin, 27; and bexarotene, 28. The animals remained on the diet until the termination of the study (at 310 d of age). The mice were weighed weekly, palpated for mammary tumors twice per week, and checked daily for signs of toxicity. At necropsy, all mammary lesions were removed and histologically evaluated (20).

Chemopreventive effects of the statins on cancer incidence and latency were determined using the log-rank analysis, and differences in cancer multiplicity were determined using the Armitage test (23, 24). The Student's t test was used to compare differences in animal body weights and serum triglyceride levels.

In an initial study (data not shown), the efficacy of lovastatin and atorvastatin against MNU-induced mammary cancers were evaluated at relatively low doses: 100 and 125 mg/kg diet, respectively. Because these dose levels did not alter tumor multiplicity or incidence, additional studies using higher doses of the agents were done. Statin, tamoxifen, or the combination of these agents did not alter the body weight gain of the animals in any of the studies >7% from the respective controls. Atorvastatin was given at 500 mg/kg diet beginning 5 days after MNU and continuing for the duration of the studies (Table 1; Fig. 1). The number of mammary cancers at the end of the study in the MNU-treated-only group was 3.8 per rat. Atorvastatin caused a 30% increase in tumor number, whereas bexarotene (80 mg/kg diet) and tamoxifen (0.4 mg/kg diet) caused 76% and 24% decreases, respectively. The combinations of atorvastatin with bexarotene or tamoxifen did not alter the preventive efficacy from that observed when bexarotene or tamoxifen were given alone. The effects of the agents on serum triglyceride levels were determined at three time intervals during the study (at 63, 94, and 126 days after initiating treatment in rats not receiving the carcinogen). As indicated in Table 2, bexarotene greatly increased triglycerides when given alone. Atorvastatin treatment did not reduce triglyceride levels in rats when given alone, but diminished the increase in triglycerides by bexarotene when the two agents were given in combination. Tamoxifen, as expected, did not modify triglycerides either when given alone or in combination with atorvastatin.

In the second experiment, lovastatin (400 mg/kg diet) and bexarotene (80 mg/kg diet) were given alone or in combination to determine their effects on mammary cancer prevention. In this study, the control group developed 3.6 cancers per rat with a mammary cancer incidence of 93% (Table 3; Fig. 2). Lovastatin treatment caused a slight increase (19%) in cancer multiplicity. As seen with atorvastatin, the combination of lovastatin with bexarotene was no more or less effective than the administration of the RXR agonist alone. Lovastatin also prevented the large increase in serum triglycerides caused by bexarotene but did not alter these levels from controls when given alone (Table 3).

Finally, atorvastatin (and a positive control bexarotene), when given as single agents to MMTV-Neu^+/−/p53KO^+/− bitransgenic mice, was evaluated for its preventive efficacy against ER⁻ mammary cancers. Other lesions (e.g., skin cancers, lymphomas) were less than three per group and were not related to treatment. As seen in Fig. 3, atorvastatin had minimal effects on the development of mammary cancers when compared with control bitransgenic mice, whereas bexarotene was effective in preventing the appearance of tumors (P < 0.05).

A carcinogen-induced model of mammary cancer in rats was developed five decades ago by Huggins and coworkers (17). The resulting mammary cancers appear histologically similar to invasive ductal adenocarcinoma in women. Recent gene array analyses of these tumors have shown them to be similar to highly differentiated human ER⁺ tumors (10). As expected, these tumors are responsive to a wide variety of agents that are effective in prevention and/or therapy of human cancers, including selective ER modulators (SERM), aromatase inhibitors, and pregnancy (25, 26). Because of its relative ease of performance, this model has been used to screen for potential chemopreventive efficacy using the widest range of agents. Various nonhormonal agents (e.g., RXR agonists and, more recently, farnesyltransferase inhibitors and epidermal growth factor receptor inhibitors) have all been shown to be highly active (11–14).

Lovastatin and atorvastatin are two small-molecule inhibitors of HMG-CoA reductase that alter cholesterol metabolism (1). Both have KIs in the low nanomolar range when used against purified HMG-CoA. The impetus for testing the statins is the relatively recent and somewhat conflicting epidemiology reports regarding the efficacy of this class of agents in preventing breast cancer (4–7). In the present experiments, the statins were first examined as potential preventive agents in a chemically induced model of mammary carcinogenesis in rats. Each of the statins was initially tested at lower doses (atorvastatin, 125 mg/kg diet, and lovastatin, 100 mg/kg diet) and found to have minimal effects on tumor incidence and multiplicity. They were subsequently tested at higher, albeit nontoxic, doses (atorvastatin, 500 mg/kg diet, and lovastatin, 400 mg/kg diet), which similarly failed to significantly decrease cancer multiplicity (Tables 1 and 3). In fact, increases in tumor multiplicity of 20% to 30% were observed; however, these increases were not statistically significant. The results were somewhat unexpected because of one unusual characteristic of the model: ∼50% of the MNU-induced cancers have mutations in the Ha Ras oncogene, specifically at codon 12 (15). We have previously shown that the Ha Ras–mutated tumors are highly sensitive to the preventive and therapeutic activity of the farnesyltransferase inhibitor tipifarnib (R115777; ref. 16). This is presumably due to blocking of farnesylation of Ha Ras, which is necessary for activation of Ras proteins. Because statins would be expected to block the prenylation process upstream of an FTI inhibitor, one might expect these cancers to be sensitive to prevention by statins. Finally, atorvastatin and bexarotene were tested as single agents in an ER⁻ model of mammary cancers in bitransgenic mice. The MMTV/Neu model was developed almost 15 years ago by Mueller and colleagues (21). The specific bitransgenic mice used in this study were heterozygous both for the MMTV-Neu transgene and KO of p53. The resulting mice developed mammary carcinomas that overexpress Neu and have an alteration in p53 (19), which is similar to human ER⁻ cancers expressing Neu. As observed in the rat, no effect of atorvastatin on cancer formation was found but bexarotene (at the dose used) reduced tumor multiplicity by ∼60%. The efficacy of bexarotene is even more striking given that only 23% of the bexarotene-treated bitransgenic mice had developed a tumor at 310 days, whereas 50% of control mice had developed tumors at 274 days of age.

We chose the two specific statins (lovastatin and atorvastatin) because they both have been commonly used in humans and because atorvastatin is considered relatively lipophilic and lovastatin less lipophilic. Some investigators had proposed that relatively lipophilic statins (e.g., atorvastatin) might exhibit significantly greater chemopreventive activity compared with less lipophilic statins (7).

However, the evaluation of how lipophilicity might influence statin chemoprevention is complex. Lipophilicity ranking of compounds is routinely expressed by the use of either experimentally determined or computed log P values (log of the n-octanol/water partition coefficient). Although an approximate log P value is readily generated, extrapolation of the measured log P in a highly complex organism is fraught with questions. The log P value is premised on the basis of a neutral molecule partitioning between two separate but chemically homogenous static solvent systems. When ionizable groups are present in a molecule, the protonation state will change with pH. Therefore, the log P value must be adjusted to reflect the pH of the medium and the ionization constant of the acid. Thus, a standard log P value does not capture true lipophilicity of ionizable compounds such as the statins. The log D value (log D = log P− [1 + 10 ^(pH-pKa)] for an acid) is a more accurate measure of true lipophilicity. However, even with the addition of pH, factors such as multiple phases, temperature fluxes, specific drug receptors, transport mechanisms, multiple ionized species, and drug metabolism are not modeled by the log P or log D equation.

Another factor that complicates interpreting the true influence of lipophilicity on the activity of the statins in a biological matrix is the ability of statins to equilibrate between a cyclic lactone (non-ionized) and free acid (ionized) structure. The lactone and open acid can be interconverted spontaneously both through pH-dependent chemical hydrolysis and enzymatic actions. Each individual form undergoes separate binding and partitioning with specific lipophilicity influences operating on each form. These many issues are illustrated in the Scheme and highlight the complexity involved in attempting to correlate lipophilicity with tissue concentrations and relative efficacy.

Although atorvastatin and lovastatin given alone were ineffective, we evaluated the combination of statins with other agents with known preventive efficacy. This was to test for possible enhanced or decreased efficacy of the combinations of agents, as well as to determine whether the expected physiologic effect of statins (decreased serum triglycerides) was achieved. In these studies, the statins were combined with suboptimal doses of two agents, the SERM tamoxifen and the RXR agonist bexarotone. Tamoxifen is a SERM that functions as an ERα antagonist in the breast (27). Tamoxifen is highly effective in prevention of ER⁺ breast cancer, both in this animal model and in human cancer. As seen (Table 1), there was no improvement in efficacy by combining atorvastatin with a low suboptimal dose of tamoxifen. The second agent used was the RXR agonist bexarotene. Bexarotene and related RXR agonists have proven to be highly effective in preventing ER⁺ tumors as well as ER⁻ mammary tumors in rodent models (11–13, 28). As shown in Tables 1 and 3, no increase in efficacy was observed when combining a suboptimal dose of bexarotene with either of the statins. Bexarotene and various RXR agonists strongly increase triglyceride levels (29). We, therefore, determined whether the combination of bexarotene and either lovastatin or atorvastatin would diminish the increase in triglyceride levels induced by bexarotene. As shown in Tables 2 and 3, both lovastatin and atorvastatin significantly decreased the bexarotene-induced increase in serum triglycerides. These studies also showed that the statins do not interfere with the efficacy of this RXR agonist. This is important because the present clinical use of bexarotene in cutaneous T-cell lymphoma has used this agent in combination with a statin.

In summary, the statins were not effective by themselves in this model of breast cancer despite the fact that the doses were relatively high. In contrast, these doses were effective in decreasing azoxymethane-induced colon cancer in rats (30). Our data demonstrating lack of efficacy of the statins in two mammary models is in contrast to a prior study examining statins in transgenic mice (31), which showed significant, albeit limited, activity of lovastatin. The dose used was lower than our dose on a mg/kg body weight basis. However, the methods of administration (diet in the present study versus i.p.) make it difficult to directly compare because i.p. administration circumvents typical concerns of absorption and hepatic metabolism. Similarly, two articles examining the effects of statins on growth of grafted tumor cells (32, 33) also used i.p. administration of the statins and showed limited efficacy. This may contribute to the different results obtained in the various studies. The present experiments, however, clearly show the lack of efficacy of these statins following dietary dosing in two in vivo models of mammary cancer used routinely in screening for preventive agents.

No potential conflicts of interest were disclosed.

We thank Jeanne Hale and Mary Jo Cagle for editorial services; Tom Morgan, Bonnie Mould, and Caroline Kirkner for performing all aspects of the animal studies; and Julie Gray for analytic determinations on diets and serum.