Comparison of Effects of Diet on Mammary Cancer: Efficacy of Various Preventive Agents and Metabolomic Changes of Different Diets and Agents Free

The rat model of breast cancer induced by the carcinogens, dimethylbenzanthracene or methylnitrosourea (MNU), has been employed for many decades employing a standard diet that had relatively low levels of fat (4%) and employed soy as a protein source (1, 2). The MNU-induced rat model of breast cancer induces ER⁺ cancers that are similar by array analyses to the highly differentiated ER⁺ human breast cancer (3). These tumors respond both in a preventive setting and in a therapeutic setting to treatments that are effective in modulating human ER⁺ tumors including selective estrogen receptor modulators (SERM), aromatase inhibitors, and ovariectomy (4–6). Thus, multiple SERMs (tamoxifen, toremifene, and arzoxifene) and multiple aromatase inhibitors (vorozole, letrozole) have proven to be highly effective in this model. These results are not surprising given that these are ER⁺ cancers and these two classes of agents directly (SERMs) or indirectly (aromatase inhibitors) affect the estrogen receptor. Two other classes of agents that we have previously reported as highly effective in this model are the EGFR inhibitors (gefitinib, erlotinib, and lapatinib) and the RXR agonists (Targretin refs. 7, 8). We have also reported on a variety of agents that are ineffective preventive agents in this model including metformin, statins, antioxidant response element (ARE) agonists, and NSAIDs (9–11).

As stated above, these studies have routinely been performed with rats maintained on a standard (Teklad 7001) diet that has relatively low levels of fat (4% fat by content, 8% fat by calories consumed), and employs a soy mixture to serve as a protein source. The question has frequently been asked whether one might obtain different results if rats are placed on a diet that more closely resembles the human diet routinely employed in the United States; a so-called Western diet. In brief, the Western (high-fat diet, HFD) diet has high levels of fat (21%) and sucrose and low levels of calcium, whereas the standard diet is low in fat, low in sucrose, and has a high soy content. There were multiple reasons for comparing the standard diet and the HFD. First, we have recently reported that the antidiabetic drug metformin, which has gained a great deal of interest both in prevention and therapy, was completely ineffective in the MNU model employing rats on a standard diet (11). It was felt that the HFD might induce a prediabetic state, and, therefore, might indicate whether a HFD would alter the efficacy of metformin. The epidemiologic data supporting the potential efficacy of metformin were primarily based on its use in diabetics (12). Second, we have previously examined the metabolomic changes associated with certain preventive agents (13). Assessing this in animals on a diet more like that consumed by humans might make any results more “relevant.” Third, we were interested in the effect of the HFD on the development of mammary cancers employing different chemopreventive agents. Our hypothesis was that the altered diet might change the chemopreventive efficacy of various agents due to altered physiologic state of the rats, as well as potential alterations in pharmacokinetics in a rat with higher dietary fat intake. For example, a lipophilic drug might partition differently in rats on a HFD.

Metabonomic or metabolomics uses state-of-the-art analytic technology to characterize the metabolome. This is defined as all the detectable biochemicals (or small molecules) present in cells, tissues, and body fluids such as blood or urine. Changes in levels of metabolites may serve as potential biomarkers associated with drug toxicity and various disease states. At the core of metabolomics is the hypothesis that a person's overall metabolic state provides a close representation of that individual's phenotype, which is sometimes referred to as the “metabotype” (14, 15). The “metabotype” is a reflection of genetics, health-to-disease status, biological age, gender, and environmental factors such as drugs, diet, alcohol, and exercise. In this article, we use metabolomics to examine metabolite levels: (i) associated with consumption of HFD or standard diet, (ii) potential metabolic biomarkers of a “prediabetic state” caused by the HFD, (iii) potential biomarkers of cancer, and (iv) identifying metabolite signatures of drug response for Targretin and tamoxifen, which might prove applicable as pharmacodynamic biomarkers in clinical prevention trials. A more complete analysis of the metabolomics changes will be published separately.

The studies performed answered a variety of concerns: First, rats on HFD had a shorter cancer latency and greater tumor multiplicity and larger final tumor weights than animals on standard diet. Second, agents that were highly effective (vorozole, tamoxifen, and Targretin) in the standard diet were similarly highly effective in the HFD. Agents ineffective in standard diet (metformin, Lipitor) were similarly ineffective in rats on HFD. Third, we identified metabolomic changes in serum associated with (i) dietary exposure (HFD vs. standard diet); (ii) metabolite changes in rats on HFD that may be associated with a prediabetic state, (iii) multiple metabolite changes associated with exposure to the highly effective chemopreventive agents, tamoxifen and Targretin; and, (iv) metabolite changes that are preferentially observed in mammary cancer–bearing rats.

Vorozole, Targretin and Metformin were supplied by the NCI Cancer Prevention Repository. Tamoxifen citrate and Lipitor were obtained from Sigma Chemical Co. Vorozole and metformin were administered by gavage (0.5 mL/treatment) daily. The vehicle for vorozole was ethanol: polyethylene glycol 400 (10:90; v/v). Metformin was administered in saline. Other agents were incorporated into the diet as described previously (5, 9).

Treatment of female Sprague-Dawley rats was done as described previously (6, 10). In brief, rats were obtained from Envigo, Inc. and were housed in Institutional Animal Care and Use Committee (IACUC)-approved animal facilities at the University of Alabama at Birmingham (Birmingham, AL) beginning at 35 DOA. Rats were placed on a standard Teklad 7001 diet (4% fat, high calcium and high soy protein) or a Western diet (HFD), Envigo TD 88137 (21% fat, low in calcium, and no soy proteins) at 43 DOA. MNU (99.5% pure) was obtained from the NCI and was injected intravenously [75 mg/kg bodyweight (BW)] via the jugular vein when the animals were 50 DOA. At 57 days of age, animal treatments with the various agents were initiated daily and continued throughout the duration of the study. Doses of the various agents were as follows: vorozole (1.25 mg/kg BW/day), tamoxifen (3.3 ppm in diet), Targretin (150 ppm in diet), Lipitor (125 ppm in diet), and metformin (150 mg/kg BW/day). Serum was obtained from rats at both 78 days and at the time of sacrifice (190 DOA).

In all the studies, rats were palpated for mammary tumors twice weekly and weighed once/week. Body weights of the rats did not decrease more than 5% from the controls in any of the prevention studies. Statistical analyses of cancer incidence and latency were determined using log-rank analysis. Differences in cancer multiplicity were determined by the Armitage test, and differences in final tumor weights were determined by Wilcoxon rank test (7,10).

The methods employed were basically as described in our previous publication (13). The nontargeted metabolic profiling platform employed combined three independent platforms: ultra–high-performance LC/MS-MS (UHPLC/MS/MS) optimized for either basic or acidic species, and gas chromatography/mass spectrometry (GC/MS). Samples were processed as described previously (16, 17). For each sample, serum protein was precipitated with methanol. The supernatant was split into equal aliquots for analysis on the three platforms. Aliquots, dried under nitrogen and vacuum-desiccated, were subsequently either reconstituted in 50-μL 0.1% formic acid in water (acidic conditions) or in 50-μL 6.5 mmol/L ammonium bicarbonate in water, pH 8 (basic conditions) for the two UHPLC/MS/MS analyses or derivatized to a final volume of 50 μL for GC/MS analysis using equal parts bistrimethyl-silyl-trifluoroacetamide and solvent mixture acetonitrile: dichloromethane: cyclohexane (5:4:1) with 5% triethylamine at 60°C for 1 hour. A cocktail of standards spiked into every analyzed sample allowed instrument performance monitoring.

Briefly, for UHLC/MS/MS analysis, the serum-extracted aliquots were separated using a Waters Acquity UPLC (Waters) and analyzed using an LTQ mass spectrometer (Thermo Fisher Scientific), which consisted of an electrospray ionization (ESI) source and linear ion-trap (LIT) mass analyzer. The MS instrument scanned 99–1,000 m/z and alternated between MS and MS² scans using dynamic exclusion with approximately 6 scans per second. Derivatized samples for GC/MS were separated on a 5% phenyl-dimethyl-silicone column with helium as the carrier gas and a temperature ramp from 60°C to 340°C and then analyzed on a Thermo-Finnigan Trace DSQ MS (Thermo Fisher Scientific) operated at unit mass resolving power with electron impact ionization and a 50–750 atomic mass unit scan range.

Metabolites were identified by automated comparison of the ion features in the experimental samples to an in-house reference library of chemical standard entries that included retention time, molecular weight (m/z), preferred adducts, and in-source fragments as well as associated MS spectra, and curated for quality control using software (18). For statistical analyses, any missing values were assumed to be below the limit of detection and these values were imputed with the compound minimum (minimum value imputation). Statistical analysis of the log-transformed data was performed using “R” (http://cran.r-project.org/), a freely available, open-source software package. A Welch two-sample t test was used to identify biochemicals that differed significantly between groups. P ≤ 0.05 was considered statistically significant and P < 0.10 was reported as trend. Multiple comparisons were accounted for by estimating the FDR using q-values (19).

As mentioned above, rats on either diet were fed ad libitum beginning at 43 DOA, administered MNU at 50 DOA, and followed for tumor development until sacrifice (190 DOA). Not surprisingly, animals on the HFD diet gained weight more rapidly than rats on the standard diet, although the final difference in weights was only 6% at the time of sacrifice (269 g for rats on HFD; 253 g for rats on standard diet). However, it should be noted that this reflects almost an 11% difference in weight gains because both groups at 43 DOA weighed approximately 120 g (Supplementary Fig. S1). Interestingly, the HFD strikingly altered tumor development. Although rats on the standard diet developed mammary cancer with an average latency of 98 days after MNU treatment, rats on the HFD exhibited an average latency of 64 days (P < 0.01). At the time of sacrifice (140 days post MNU), these diet-related differences were quite striking (cancer multiplicity: standard, 2.7; HFD, 5.2. Final cancer weights: standard diet, 5.7 g; HFD, 11.4 g). All three parameters (latency, multiplicity, and weights) were statistically different (P < 0.05) when comparing groups on the two different diets.

We evaluated three effective agents (tamoxifen, vorozole, and Targretin) and two ineffective agents (Lipitor, metformin) that we had tested previously (and published results) in the standard diet. The primary objective of including the current studies with standard diet was to serve as concurrent controls for the studies using HFD. Nevertheless, it became an indirect test of the reproducibility when compared with our previously published results. It was found that the three positive preventive agents (Targretin, vorozole, and tamoxifen) yielded strong positive results as expected. These positive agents reduced cancer multiplicity and final cancer weights by 85% to 100%, and all increased tumor latency (Table 1; Fig. 1A). Results with the two negative agents (Lipitor, metformin) also reproduced our prior results. Specifically, neither of the agents was effective in reducing cancer formation, and metformin actually increased final cancer multiplicity and final cancer weights. Lipitor, in contrast, had minimal effects on cancer multiplicity and final weights and was clearly ineffective. Data for three agents (Lipitor, Targretin, and vehicle control) are presented in Fig. 1A to illustrate typical data. The current data with metformin confirms our prior data with this agent, which also showed that metformin significantly increased tumorigenesis in rats on a standard diet (ref. 11; Table 1; Fig. 1A).

The same five agents that were tested in the standard diet were also evaluated for mammary cancer–preventive activity in rats on the HFD. It was found that the three positive agents also yielded strongly positive results in the rats fed HFD. These positive agents reduced cancer multiplicity and final cancer weights by 85% to 95%, and all increased tumor latency (Table 1), in agreement with our results using a standard diet. Results with the negative agents (Lipitor, metformin) showed that both significantly increased final tumor weights. Interestingly, Lipitor significantly increased tumor multiplicity and decreased tumor latency. Data for three agents (Lipitor, Targretin, and vehicle control) are presented in Fig. 1B to illustrate typical data (Table 1; Fig. 1B).

Metabolomics was first used to look for metabolite changes between the control groups on standard diet and HFD at 78 days (T1) and 190 days (T2). Fig. 2A shows box plots for myristate (14:0) and Fig. 2B shows 2-aminooctanoate levels for control diet (CD, standard diet), control HFD, Lipitor CD, and Lipitor-HFD studies at both T1 and T2. Myristate is increased in the control HFD at both T1 and T2 compared with control CD and the similar changes are observed for Lipitor HFD compared with Lipitor CD. The inclusion of the Lipitor samples is purely illustrative to show that the metabolite changes (myristate and 2-aminodecanoate) are independent of agent treatment. Myristate along with several other medium-chain saturated fatty acids [caprate (10:0), laurate (12:0), and pentadecanoate (15:0), several monounsaturated fatty acids (MUFA; 10-undecnoate (11:1n1), myristoleate (14:1n5), and 10-heptadecenoate (17:1n7)] and several methylated fatty acids (13-methylmyristic acid, 15-methylpalmitate) had similar trends to those as observed in myristate. 2-aminooctanoate showed the opposite trend to myristate and was decreased in control HFD and Lipitor-HFD compared with the normal CD at both time points. 2-hydroxyocyaboate and 2-hydroxydeconate had similar trends as 2-aminooctanoate. The metabolomic data also show expected high levels of saturated fatty acids (caprate laurate, myristate, etc.), which were substantially higher in the HFD at both time points. Interestingly, levels of palmitate, stearate, and oleate, which similarly were higher in serum from rats on the HFD at the first-time point (78 DOA) were not increased in levels at the second-time point (Supplementary Table S1) Likewise, a variety of metabolites associated with dietary soy consumption (genistein, daidzein, equal, and their derivatives) were higher in sera from animals on the standard diet at either time point (Supplementary Table S1).

In Fig. 2C, it was observed that 2-aminobutyrate was increased in animals on a HFD, whereas in Fig. 2D, glycine was decreased in levels on the HFD at both time points. 2-Aminobutyrate has been reported as a biomarker of diabetes risk and prevalence in humans (20). Glycine is reduced in control HFD compared with control standard diet at both time points. Reduced levels of glycine have similarly been reported as a biomarker of prediabetes (20) and may be related to dimethylglycine (DMG) deficiency observed in diabetes (21).

Figure 2E shows box plots of 5-methylcytidine and Fig. 2F shows N-acetylcitrulline in rats on standard diet, or HFD at T2 from control, metformin, Lipitor, tamoxifen, or Targretin-treated groups. The rats from the control, metformin, and Lipitor groups all had multiple tumors at T2, whereas rats in the tamoxifen or Targretin treatment groups had few, if any. 5-methylcytidine and other methylated DNA bases (7-metylguanine, N2-N2-dimethylguanosine) were increased at T2 in serum from groups of rats with tumors compared with groups of rats with minimal or no tumors. Increased methylated DNA bases have been observed previously in the clinic (22, 23). The urea cycle metabolite, N-acetylcitrulline, and another N-acetylated urea cycle metabolite (N-acetyl arginine) were decreased in serum from rats with minimal or no tumors at T2. The reduction in urea cycle metabolites is compatible with the recent finding that breast tumors in mice use ammonia as a source of nitrogen during tumor growth (24).

The second set of Fig. 3A–D identify metabolites that are potential pharmacodynamic biomarkers related to the two highly positive chemopreventive agents, Targretin and tamoxifen. Fig. 3A shows box plots of citrulline and Fig. 3B taurocholate in control CD, control HFD, Targretin-CD, and Targretin-HFD studies at both time points. Citrulline was increased in Targretin-treated samples in both CD and HFD and at both timepoints when compared with their controls. Citrulline is an amino acid associated with the urea cycle and other members of the urea cycle were modulated by Targretin treatment. In a prior study examining Targretin and another RXR agonist UAB-30 (13), we observed that Targretin not only increased citrulline in serum but increased levels of this metabolite in liver and urine as well. Levels of the bile acid taurocholate were strikingly decreased in the Targretin samples in both CD and HFD at both timepoints when compared with their controls. Besides taurocholate, striking alterations in the levels of other bile acids were observed including decreases of the bile acids (glycocholate, chenodeoxycholate, glycochenodeoxycholate, and taurochenodeoxycholate) and increases in α and β muricholate were observed with Targetin dosing. We had previously observed alterations in various bile acids in rats on standard Teklad diet treated with Targretin (13).

Figure 3C shows box plots of palmitoyl-linoleoyl-glycerophosphocholine (2) and Fig. 3D shows plots of stearoyl-arachidonoyl-glycerophosphocholine (1) in CD, Control HFD, tamoxifen-CD, and tamoxifen-HFD studies at both time points. Palmitoyl-linoleoyl-glycerophosphocholine (2) and palmitoyl-linoleoyl-glycophosphoinositol (1) were increased in tamoxifen samples in both CD and HFD and at both timepoints when compared with their controls. Stearoyl-arachidonoyl-glycerophosphocholine (1) and stearoyl-arachidonoyl-glycerophosphocholine (2) shown were decreased in tamoxifen samples in both CD and HFD diets and at both timepoints when compared with their controls. The glycolipid changes listed above are consistent with the results reported previously in tamoxifen-treated rats (Fig. 2A–F; Supplementary Table S1; ref. 25).

These studies were driven by four questions: (i) can a HFD that more closely mimics the human diet alter tumor development in a chemically induced cancer model and be more “relevant” than a standard chow diet that is low in fat and has high levels of soy?; (ii) because we recently found that the antidiabetic drug metformin showed no efficacy in this model in animals on standard diet (11), would metformin show the same lack of efficacy in rats on a HFD, which more closely approximates the standard human Western diet?; (iii) would a variety of preventive agents (effective and ineffective) show similar activity in rats on HFD, given the potentially altered physiology of the rats and potential alterations in pharmacokinetics in these rats?; and (iv) could metabolomics be used to define potential pharmacodynamic (but not efficacy) biomarkers in serum that might be used in clinical prevention trials of chemopreventive agents? Thus, certain of the metabolomic changes observed in Fig. 3A–D were significantly altered at highly effective doses of Targretin and tamoxifen, respectively. We feel that the levels of these metabolites might be useful pharmacodynamic biomarkers in clinical trials of these agents and, furthermore, the basic metabolomics approach should be relevant to other potential effective agents. Metabolite changes were determined both in standard diet as well as in the HFD.

If one compares MNU-treated rats given vehicle, there is a higher tumor multiplicity, tumor weight, and shorter latency period in animals on HFD versus standard diet (Fig. 1A and B; Table 1). These data are not surprising, based on extensive data showing that giving HFD ad libitum yields a higher tumor incidence and multiplicity in this model (26, 27). This may be due in some part to the increased body weights achieved in rats on a HFD, although this weight gain was not profound.

If one examines the results of the various agents in the rats on a standard diet (Table 1), it is observed that two of the agents (vorozole, Targretin) caused profound (80%–90%) decreases in tumor multiplicity and final tumor weights. Interestingly, at a dose of 3.3-ppm tamoxifen in the diet (which is far below the human equivalent dose), it was observed that there were virtually no cancers in any of the rats. These two parameters (multiplicity and final cancer weights) are closely associated because the tumors that do grow out following treatment with highly effective agents tend to be smaller. These agents also cause an increase in the average cancer latency (time to appearance). Parenthetically, our results in this study are in close agreement with our prior published studies with these agents (5, 6, 8). In contrast, the ineffective agents (9, 11) remained ineffective (metformin and Lipitor). In fact, metformin (see above) increased both tumor multiplicity and final tumor weights in rats on the standard diet. Comparison of these parameters employed a Wilcoxon rank test because the tumor numbers and average tumor weights per rat did not follow a normal distribution. Lipitor caused a 16% decrease in tumor multiplicity that was not significant, and an increase in final tumor weights of approximatively 40%, although this difference was not significant (P > 0.05). Thus, the results achieved in the standard diet are like those that we have published with these agents (9, 11). Similarly, the chemoprevention results in rats on a HFD were like those on the standard diet (Table 1). Specifically, increases in tumor latency and decreases in tumor multiplicity and final tumor weights were observed for tamoxifen, vorozole, and Targretin groups. The ineffective agents were not active in either diet. However, the absolute number of cancers (as well as the average tumor weights) were higher in rats on a HFD regardless of agent treatment. As we stated earlier, we thought that there were at least two major considerations that might alter the efficacy of certain of the preventive agents when administered to rats on a HFD versus a standard diet. First, it was hypothesized that one might achieve either a prediabetic state in animals on a HFD and that the altered physiology might affect efficacy. In fact, it appears that we did achieve a prediabetic state in the animals on a HFD (see below; ref. 28). Second, we hypothesized that altered physiology and fat content of rats on the HFD might alter the pharmacokinetics and, subsequently, the efficacy of various agents, although we did not perform pharmacokinetics in these studies.

We recently reported that the antidiabetic agent metformin was ineffective in rats on a standard diet (11). This result was somewhat unexpected given the great enthusiasm for metformin as a potential preventive/therapeutic agent, although the limited data in humans (in a placebo controlled trial) do not look strikingly positive (28). However, the current data in rats fed a standard diet are a repeat of our prior published studies showing a significant increase both in tumor multiplicity and final tumor weights in rats given metformin while on a standard diet. We had the expectation that positive results with metformin in rats on a HFD would be observed. Our rationale was that a prediabetic state would be achieved on a HFD and that this might make the rats more susceptible to the physiologic effects of metformin (e.g., altered IGF1 levels, which might contribute to its potential efficacy). To routinely achieve definitive insulin resistance in rats by diets has routinely required diets in which roughly 60% of the consumed calories are from fat employing either lard or vegetable oil shortening and exposure periods of 4 to 6 months (29). As can be seen in Table 1, efficacy with metformin in rats on the HFD was not achieved. There was no increase in tumor multiplicity; however, an increase in final tumor weights was observed. Lipitor, although clearly not effective in rats on either diet, significantly increased tumor formation (decreased latency, increased multiplicity, and final tumor weight) in animals on the HFD. The rationale for these results warrants further investigation.

As stated in the background and previously outlined in our prior data examining metabolomic changes in rats given RXR agonists (13), we examined metabolomics in serum to determine potential pharmacodynamic biomarkers that might be used in clinical studies. We were not attempting to identify biomarkers that were necessarily associated with the mechanism of action of a given agent or class of agents. The objective of the metabolomic analysis was to identify metabolic changes and specific metabolites that were indicative of the physiologic or pharmacologic effects of a given agent at an effective preventive dose in the animal model. In the human model, similar changes would imply that a pharmacologically significant dose had been achieved that would hopefully be effective in prevention of the target cancer. There are two comments that need to be made about metabolomics. First, if one includes markers for the agent itself or a major metabolite of the agents, one can quantitate these levels. We identified metformin, Lipitor, and Targretin in their respective treatment groups. However, because this study was performed at a single time point, one could not determine true pharmacokinetic results that require multiple time points to be examined. Implicit in the metabolomic studies is that the changes are relatively constant if the preventive agent is administered. Because metabolomics were performed at 28 and 140 days after the beginning of treatment with the agent, and the changes were observed at both time points, these changes were not transient.

We initially screened for metabolites that were differentially expressed in the serum of rats on HFD versus standard diet and found metabolites from numerous pathways altered, including decreases in glycine, urea, and phenylalanine pathways as well as increases in certain medium-, long-chain, and polyunsaturated fatty acids. The latter may be directly related to the fat content of the HFD. The two metabolites that we show results are myristate and 2-aminooctanoate (Fig. 2A and B). The second question addressed was whether a prediabetic state was achieved (Fig. 2C and D). Increased levels of 2-aminobutyrate and decreased levels of glycine were observed in rats on the HFD at both time points, which agrees with published metabolite changes associated with prediabetes and insulin resistance in humans (30), implying that the rats on the HFD may be progressing to prediabetes. However, as mentioned above, clear insulin resistance in a rat normally requires a diet in which roughly 60% of calories are supplied by animal fat (lard) or vegetable shortening (29). One of the most intriguing aspects of these studies were the results shown in Fig. 2E and F with 5-methylcytidine and N-acetyl citrulline whose changes in serum appeared to be associated with the presence or absence of tumors in the various groups at the time of final sacrifice. Thus, rats eating control/Teklad diet or HFD (and in the control metformin or Lipitor groups), all developed a high tumor burden and had higher relative levels of 5-methyl-cytidine than rats exposed to Targretin or tamoxifen, which developed few, if any, cancers. Similarly, these same groups had lower levels of acetycitrulline than Targretin or tamoxifen. We will examine this in greater detail in a follow-up article. Defined potential pharmacodynamic biomarkers were also associated with the effective chemopreventive agents. In Targretin-treated rats (Fig. 3A and B), citrulline (an amino acid derived from the urea cycle) increased, whereas the levels of taurocholate were profoundly decreased irrespective of diet. These may be related to downstream effects of Targretin RXR agonist activity. For example, alterations in bile acids (e.g., Taurocholate) have been attributed to activation of the FXR receptor (13). For tamoxifen (Fig. 3C and D), levels of the glycolipid palmitoyl-linoleoyl-glycerophosphocholine (2) were increased, whereas the levels of stearoyl-arachidonoyl-glycerophosphocholine (1) were decreased in the Tamoxifen-treated rats on either diet. The glycolipids changes in tamoxifen have been observed previously (25) and may be related to known alterations in glycolipid levels associated with tamoxifen in humans (31).

The metabolomic approach also reflected specific components due to the two different diets. Thus, we observed increased levels of various fatty acids in rats on the HFD as well as high levels of various soy components in the serum from rats on standard diet (Supplementary Table S1). These differences support the fact that one can see dietary changes in the metabolite profile. Although one should be aware that the two diets are strikingly different and were examined in a relatively homogeneous system.

In summary, we have investigated: (i) tumor development following MNU administration in Sprague-Dawley rats on HFD or standard diets; (ii) examined the efficacy of the known effective or ineffective chemopreventive agents in both diets, and (iii) determined the serum metabolomics in rats on either diet to determine the pharmacodynamic biomarkers that might be useful for clinical trials. The results show that animals on the HFD were more susceptible to tumor development, yielding a shorter tumor latency and increased cancer multiplicity. This is of importance because the altered diet may be applicable to a wide variety of cancer models and may be more relevant in determining metabolic changes (see below). Furthermore, our studies comparing various effective (tamoxifen, vorozole, and Targretin) or ineffective (metformin, Lipitor) agents showed that results were strikingly similar in animals on standard diet or HFD. One interesting aspect of this study is that the results in standard diet yielded the same results as our prior studies with these agents that had been collected over a 20-year period using this diet. This shows that the basic model is highly reproducible when testing strongly positive or negative agents. The similar results in rats on either standard diet or HFD indirectly imply the reproducibility of the MNU model despite some changes in the protocol. In metabolomics, we had many observations. First, and not unexpectedly, we observed a variety of differences in serum directly related to the content of the HFD, including alterations in saturated fatty acids. Changes in certain metabolites, including glycine and 2-aminobutyrate, implied that a prediabetic state with the HFD was achieved, but other metabolites associated with a prediabetic state like branched-chain amino acids were not changed. Finally, we have defined potential pharmacodynamic biomarkers associated with the highly effective chemopreventive agents. Further studies may determine whether some algorithm employing multiple biomarkers or pathway-related algorithms may have the greatest use.

No potential conflicts of interest were disclosed.

The views presented here do not necessarily reflect those of the U.S. Food and Drug Administration.

Conception and design: C.J. Grubbs, R.A. Lubet, M.S. Miller

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.E. Seifried, C.J. Grubbs

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.A. Lubet, R.D. Beger, H.E. Seifried

Writing, review, and/or revision of the manuscript: R.A. Lubet, R.D. Beger, M.S. Miller

Study supervision: M.S. Miller, J. Luster

This work was supported by NCI grant HHSN261201200021I (contract support).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.