Distinct Effects of Calorie Restriction and Exercise on Mammary Gland Gene Expression in C57BL/6 Mice

Energy balance, including diet, weight, adiposity, and physical activity, is associated with carcinogenesis. Epidemiologic studies indicate that obesity and sedentary and/or active behavior are risk factors for breast cancer in postmenopausal women and survival in both premenopausal and postmenopausal breast cancer patients. Thus, understanding the influence of energy balance modulation on changes in gene expression patterns in the normal mammary gland is important for understanding mechanisms linking energy balance and breast cancer. In a 6-week-long study, female C57BL/6 mice (9-week-old) were randomized into four groups: (a) food consumed ad libitum (AL), (b) AL with access to running wheels (AL+EX), (c) 30% calorie restricted (CR), and (d) 30% CR with access to running wheels (CR+EX). CR mice received 70% of calories but 100% of all other nutrients compared with AL mice. Diet and exercise treatments, individually and combined, had significant effects on body composition and physical activity. Affymetrix oligomicroarrays were used to explore changes in gene expression patterns in total RNA samples from excised whole mammary glands. Contrasting AL versus CR resulted in 425 statistically significant expression changes, whereas AL versus AL+EX resulted in 45 changes, with only 3 changes included among the same genes, indicating that CR and EX differentially influence expression patterns in noncancerous mammary tissue. Differential expression was observed in genes related to breast cancer stem cells, the epithelial-mesenchymal transition, and the growth and survival of breast cancer cells. Thus, CR and EX seem to exert their effects on mammary carcinogenesis through distinct pathways.

A growing literature indicates that elements of energy balance, including obesity, caloric intake, and levels of physical activity, influence cancer risk at multiple sites (1, 2). Breast cancer risk seems to be influenced by several of these aspects of energy balance (3–5), particularly energy intake and obesity, but an emerging literature indicates that exercise (EX) also delays postmenopausal breast cancer, as well as colon and prostate cancer (6–8). A major and unsolved question concerning these diverse results is to what extent epidemiologic associations among diet, body weight, physical activity, and carcinogenesis arise through the same mechanistic pathways.

Animal models have proved useful in characterizing the mechanisms underlying the epidemiologic associations between energy balance and carcinogenesis. Calorie restriction (CR), the most commonly recommended dietary strategy in humans to prevent or reverse obesity, dramatically reduces or inhibits the incidence of spontaneous, chemically, and virally induced mammary tumorigenesis in diverse animal models (1, 9–12). Experimental evidence indicates that changes in hormone/growth factor signaling and immune function mediate (at least in part) the anticancer effects of CR (1, 12). EX is the most commonly recommended way to increase energy expenditure. However, the effects of EX on mammary carcinogenesis in rodent models have not been as consistent as results of CR interventions (2, 13–22). Differences between chemically induced and spontaneous tumors, voluntary versus involuntary EX treatments, age of initiation of physical activity, and rodent model used for the study do not seem to account for these mixed results (2, 18, 21–23). In mouse models of carcinogenesis, beneficial effects of CR are mediated in part by reductions in serum insulin-like growth factor (IGF)-I levels (1). However, EX does not seem to reduce serum IGF-I levels in humans (24), and in some mouse studies, IGF-I is increased by EX (2). One recent study of rats indicates that the preventive effect of EX for lung and liver cancer may be explained by increased activity of antioxidant and phase 2 enzymes (25).

A review of 22 reports of microarray analyses of the effects of CR in various organisms and tissues, most with a focus on processes related to aging, concluded that no particular genes were altered in common across all the studies (26). Nevertheless, multiple genes involved in energy metabolism, stress responses (such as heat shock and oxidative stress), and inflammation pathways consistently showed changes in these microarray analyses in response to CR. We are unaware of any past microarray studies directly comparing the effects of CR or EX, alone or in combination, using mammary tissue. One study of gene expression in mouse skin tissue reports that CR and EX have unique effects; however, EX with ad libitum (AL) consumption had smaller and fewer effects on gene expression than EX with controlled feeding (27). Importantly, there are also distinct gene expression patterns in AL and CR-fed animals in different tissues such as adipose, muscle, and brain (28, 29).

To understand more fully how energy balance influences mammary carcinogenesis, it is critical to understand the effects of energy balance on normal mammary gland growth and development. The present study was designed to determine the individual and combined effects of CR and EX on normal gene expression patterns in RNA samples from whole mammary glands of C57BL/6 mice. A combination of detailed energy balance phenotyping and measures of gene expression in normal mammary tissue were designed to explore the distinctive effects of CR and EX on gene expression in normal, non-neoplastic mammary glands. Such expression patterns represent the biological background in which carcinogenesis might occur as well as one of several potential mechanistic pathways through which tissue level preventive effects could be occurring.

Six-week-old female C57BL/6 mice (n = 96) were purchased from Charles River Laboratories. Upon arrival in the main animal facility, the mice were kept on a reverse light/dark cycle (22:00/10:00). The mice were active during the dark cycle, under red lights. During the experiment, all mice were housed individually in a specific pathogen-free animal facility at the Frederick laboratories of the National Cancer Institute. The study was done under an approved animal study protocol, and animal care was in accord with National Cancer Institute Animal Care and Use Committee guidelines.

Mice (at 9 wk of age) were randomized into four experimental groups (n = 24 mice/group) as follows: (a) food consumed AL; (b) AL with free access to running wheel (AL+EX); (c) 30% CR; and (d) 30% CR with free access to running wheels (CR+EX). Twelve mice per group received implanted transponders that detected body temperature and spontaneous activity levels (see below). The study was done in four blocks. Blocks a and b included 12 mice each with implanted transponders, 3 in each treatment. Blocks c and d each comprised 12 mice with transponders (n = 3 in each treatment group) and 24 mice without transponders (n = 6 in each treatment group). Each block started ∼8 wk after the end of the previous one.

Mice on the AL diet regimen received AIN-76A diet (Bio-Serv) or a CR version of the diet (Bio-Serv) delivered in 1-g dustless precision pellets and formulated such that the reduction in calories was entirely from carbohydrates. All other nutrients were increased to match AL consumption (details in ref. 30). CR mice were fed with AL diet until the beginning of the study and then switched to daily aliquots of the CR diet and restricted by 30% (compared with average daily intake of the AL or AL+EX groups separately).

Mice in the EX groups were housed in cages equipped with a running wheel (MiniMitter, Inc.). Running wheel revolutions were recorded and separate measures of spontaneous locomotor activity and body temperature were made using transponders surgically implanted (2 wk before study onset) in the abdominal cavity.

Body weights and food consumption (in the AL fed animals) were recorded weekly. After 6 wk on study, mice were sacrificed by continuous CO₂/O₂ inhalation, and serum and various tissues, including a pair of thoracic mammary glands, were collected and immediately frozen and stored at −80°C. Body composition and bone characteristics were determined (details in refs. 30, 31) using dual-energy X-ray absorptiometry (GE Lunar Piximus II).

Serum IGF-I concentration was measured with a rat/mouse RIA IGF-I kit (Diagnostic Systems Laboratories, Inc.). Serum IGF-I concentrations were determined for only those mice with implanted transponders. We report average values for two replicate determinations from a single sample per animal.

Total mammary gland RNA was isolated from pooled left and right thoracic mammary glands (∼60 mg each) for only those mice with implanted transponders, using TRIzol extraction reagent (Invitrogen) and the RNeasy Midi Cleanup kit (Qiagen). Expression data were thus obtained for 9 to 10 animals per group. The samples represent the entire cellular milieu of the mammary glands, including a mix of epithelial cells, adipocytes, mesenchymal cells, and immune cells. Total RNA concentration was determined by spectrophotometric evaluation of absorbance at 260 nm, and RNA integrity was confirmed by 1% agarose gel electrophoresis. RNA quality of random samples was also tested on an Agilent 2100 Bioanalyzer (Agilent Technologies).

Microarray analysis was done using Affymetrix GeneChip Mouse Genome 430A expression arrays (details in ref. 32). These arrays contain ∼23,000 Affymetrix probe sets corresponding to ∼13,000 mouse genes. Signal values and detection calls of the probe sets were determined using Affymetrix GCOS (ver. 1) software. Relative intensity variation across arrays was normalized by scaling to an average target signal level of 500 counts, excluding lowest 2% and highest 2% signals. Signal values were log₂ transformed for statistical analysis.

Class comparisons were first explored using two-way ANOVA considering EX and CR as independent variables (R Foundation for Statistical Computing) after adjusting for block variation. P values were adjusted by the Benjamini & Hochberg method for all features on the array (22,691), and 1,978 features were found with a false discovery rate (FDR) of <0.05. For the interaction between EX*CR, none were found to be significant at FDR < 0.05. This finding indicated a limited interaction between EX and CR. Thus, we examined CR and EX effects separately using Affymetrix Microarray Suite Version 5.0 (MAS5) Software. First, block-dependent scale factors and shifts of data were adjusted by transforming to Z-scores. Class comparisons were done using two-sample t tests. T he FDRs and random probability of number of significant genes were estimated by multivariate permutation testing using BRB-ArrayTools.⁵

Average gene expression changes between classes were calculated using untransformed data. Genes altered by at least 1.5-fold at P ≤ 0.005, and with the geometric mean signal value of one of the classes >100, were selected for further examination. We note that by this pair-wise comparison method, a different list of features of potential interest resulted than the one from the two-way ANOVA approach. For example, by two-way ANOVA, CR effects indicated that 469 features were altered by ≥1.5-fold at FDR < 0.05 and with at least one signal value above 100. Sixty-seven percent of these features were shared between this analysis and the two sample comparison. The two-way ANOVA list can be calculated from our data, accessible through GEO Series accession number GSE14202,⁶ or is available from the authors upon request.

Global expression changes in each intervention group compared with AL mice were examined by principal component analysis (PCA) using Partek Pro discovery software (Partek, Inc.). PCA analysis included all the genes detected in at least 50% of the arrays (12,814 probe sets). Further filtering was done by eliminating expression data that were associated with high SDs. Specifically, we eliminated genes having a SD of >1 in any sample—AL, AL+EX, CR, or CR+EX. PCA of the remaining 1,649 probe sets found not to alter the patterns. Ellipsoids of 2 SDs (95% probability) enclosed most of the arrays (Fig. 1).

Hierarchical clustering was also used to explore changes in groups of these genes. Genes were clustered by the similarity of their expression profiles based on logarithmic values of expressions using (1-correlation) as the distance metric (33, 34). To investigate functional categories of genes altered by CR and CR+EX interventions, we used (a) comprehensive literature searches and (b) gene ontology terms.⁷

Genes were grouped into functional categories using the Expression Analysis Systematic Explorer software application for rapid biological interpretation of gene lists that result from microarray analysis (35).

Real-time reverse transcriptase-PCR (RT-PCR) following manufacturers' recommended protocols was used to confirm array results of gene expression changes (Applied Biosystems) using two microgram of total RNA for the first-strand reaction and VIC-labeled β-actin as the reference. Messenger RNA expression of the following genes was quantified: leptin, Wee1, Elov-6, Igfbp4, and Uble1a. Reactions used the TaqMan universal PCR master mix (Applied Biosystems) in a total volume of 30 μL on an iQ5 Real-time PCR Detection System (Bio-Rad Laboratories). Amplifications were done in triplicate for each sample; PCR optimal conditions were 50°C for 2 min and 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. The threshold cycle (Ct) method was used to generate expression values, and results for individual genes were quantified relative to the β-actin control. Three replicate measurements on five to eight mice were completed.

Effects of diet and EX on body composition, running wheel performance, and IGF-I serum levels were examined using one- or two-way ANOVA to account for different experimental designs. We included body weight as a covariate in the analysis of bone mineral density (BMD) and bone mineral content. Tukey's Honestly Significant Difference test was used to compare means of all treatment groups. All models included a block effect. Block had small but statistically significant effects on body weight and other phenotypic variables (P = 0.025); interaction effects between block and treatment were not statistically significant (P > 0.1). These analyses were done using SAS JMP (SAS Institute, Inc.).

Table 1 summarizes effects of CR and EX on body composition. CR resulted in significantly decreased body mass (P < 0.0001), with reductions in both lean (P < 0.0001) and fat mass (P = 0.0087). Access to running wheels resulted in a small decrease in body weight in both AL+EX and CR+EX mice; this difference was not statistically significant. CR significantly reduced overall BMD (P < 0.0001), tibial BMD (P < 0.0001), and vertebral BMD (P < 0.0001) and also decreased bone mineral content (P = 0.0202). In contrast, EX increased BMD in both AL+EX and CR+EX mice (P = 0.0009). The increase in BMD was greater in AL+EX mice than in CR+EX mice, even after adjustment for the effects of body size (P_interaction = 0.04). Qualitatively similar results were obtained for mice with and without implanted transponders.

Weekly food consumption was recorded for AL and AL+EX mice. AL+EX mice consumed significantly more (27.2 ± 0.4 g) food than AL mice (24.7 ± 0.4 g; P < 0.0001). Mice with implanted transmitters consumed less food (25.2 ± 0.3 g) than mice without transmitters (26.6 ± 0.4 g; P = 0.0154). There was no interaction between EX and transmitter status (P = 0.8280). CR and CR+EX mice received 70% of the AL and AL+EX food consumption level.

AL+EX mice with free access to running wheels ran 4.3 ± 0.5 km per day, whereas CR+EX mice ran 1.4 ± 0.5 miles per day(Table 2). CR significantly reduced body temperature (P = 0.0001), and there was a significant interaction between calorie intake and EX (P = 0.0004). Running wheel access increased body temperature in AL+EX mice but reduced it in CR+EX mice (P = 0.008).

IGF-I serum levels were 302.7 ± 30.4 ng/mL in AL mice and declined significantly in response to the diet intervention with or without EX (CR, 181.2 ± 30.4 ng/mL; CR+EX, 176.3 ± 33.8 ng/mL; P = 0.007). EX alone did not significantly affect IGF-I serum levels (AL+EX, 325.3 ± 30.4 ng/mL), and there was no statistically significant interaction between CR and EX (P = 0.66).

Global gene expression patterns in a subset of AL-, AL+EX-, CR-, and CR+EX-treated mice were examined by PCA. All genes detected in at least half of the arrays (12,814 probe sets) revealed that the CR interventions are more dissimilar from AL than EX from AL. Elimination of data with high SDs (leaving 1,649 genes) did not alter these patterns. The first three principal components accounted for 51.5% of the total variance in these 1649 genes and illustrate a strong diet effect (Fig. 1). The ellipsoids of 2 SDs (95% probability) enclosed most of the arrays. CR-dependent changes were observed in the variance along principal component #1 (35.8%), and EX-dependent changes were seen in the variance along principal component #2 (9.5%).

To further explore these data, we first performed two-way ANOVA. This analysis revealed that 418 genes were altered by ≥1.5-fold and had at least one signal value above 100. There were 267 features at P < 0.005 for the interaction of EX*CR, all with an FDR of >0.05. This finding indicated a limited interaction between EX and CR. As described in the Materials and Methods section, we therefore chose to examine CR and EX effects separately using signals from MAS5 software to identify a more complete list of candidate genes influenced by CR and/or EX that could then be later followed for validation.

Differential gene expression compared with AL or CR treatments [expression changed by ≥1.5-fold (P < 0.005)] is illustrated using Venn diagrams (Fig. 2A-C). Diet and EX altered distinctly different sets of genes. More than 400 genes changed in the CR versus AL comparison with only ∼1/10th as many changes seen in the AL versus AL+EX and CR versus CR + EX comparisons (Fig. 2A-C). The CR+EX intervention resulted in significant changes in the expression of 295 genes compared with the AL group (134 upregulated and 161 downregulated), and of these, 139 were significantly altered in the same direction by CR (Fig. 2B). Of the 30 genes changed in the CR versus CR + EX comparison, 14 were found to be common to the list of genes significantly different between AL and CR without EX; however, the changes in expression of all 14 genes (either increased or decreased by CR) were in the opposite direction in CR+EX.

Hierarchical clustering of the 425 genes significantly altered by the CR intervention was done to visualize the patterns of changes among the experimental groups. The gene expression data for the AL+EX, CR, and CR+EX groups are compared with AL and presented as a heat map (Fig. 3). Only genes with statistically significant changes in the AL versus CR comparison are shown. CR, with or without EX, had a stronger effect on expression of this group of genes than EX alone and changes were in a similar direction for both CR and CR+EX, whereas the AL+EX intervention affected a different and distinct group of genes.

Genes were categorized according to functional categories (using Expression Analysis Systematic Explorer analysis and literature review) and further sorted by their transcription fold change (Table 3). Gene expression changes occurred in diverse functional categories, including cell cycle regulation and proliferation, differentiation and morphogenesis, apoptosis, DNA repair, and lipid metabolism. Additionally, there were statistically significant changes in genes in cancer-related pathways [e.g., in the WNT, mitogen-activated protein kinase (MAPK), and IGF-I pathways].

Differences in expression of certain genes for the CR and CR+EX interventions compared with AL were not always both statistically significant, but the direction of change was always consistent (upregulated or downregulated versus AL), as illustrated in the heat map presentation (Fig. 3). Notable examples include the cell cycle regulation and cell proliferation genes Wee1 kinase, Cyclin-G1 (Ccng1), Cyclin D1 (Ccnd1), and Mycn (Table 3). Furthermore, death-associated protein kinase 1 (Dapk1), caspase 7 (Casp7), and Rad23a were all significantly upregulated by CR versus AL, whereas GADD45 was significantly downregulated in mammary glands of both CR and CR+EX mice compared with AL mice.

CR and/or CR+EX compared with AL influenced the expression of genes involved in fatty acid synthesis and elongation (Table 3). These included elongation of long-chain fatty acids (ELOVL5 and ELOVL6); fatty acid binding protein 5, epidermal (fabp5); endothelial lipase (lipg); and fatty acid desaturase 2 (fads2). Several genes involved in mammary gland development showed changes in expression. Growth hormone receptor (GHr) was significantly overexpressed compared with AL, and CD24a antigen (CD24a), which plays an important role in cell differentiation, was significantly downregulated. Stk 16 serine/threonine kinase 16 (stk16/Krct) and heat shock protein 9A (hspa9a) were upregulated in mammary glands of both CR and CR+EX mice, and claudins 3 and 4 (cldn3 and cldn4) and prolactin receptor (prlr) were all downregulated (although not all comparisons were significant for both treatments).

Finally, several genes involved in pathways (e.g., WNT, MAPK, and IGF-I) shown to play an important role in carcinogenesis were also significantly altered by CR compared with AL. As mentioned above, these genes were altered in the same direction by CR+EX versus AL, and several also reached statistical significance. In addition, the leptin (lep) gene was significantly downregulated in mammary tissue of CR mice compared with the AL group.

A similar analysis with PubMed searches and Expression Analysis Systematic Explorer was also done on the genes altered in the comparisons AL versus AL+EX (45 genes) and CR versus CR+EX (30 genes). The complete list of the 75 genes altered, the direction of change, and gene functions are reported in Table 4; some of the genes identified were related to motile and catabolic processes, but other categories of gene function were not obvious in this small sample of genes.

Real-time RT-PCR analysis was used to validate array results. The genes were selected to include targets related to energy balance that increased, decreased, or did not change in the microarray analysis (Elov6, Leptin, Igfbp4, Uble1A, and Wee1). Gene expression levels of the selected genes were qualitatively similar in RT-PCR and microarray analyses. For example, contrasts between CR and AL fed animals were 3.52 (0.72), 0.17 (0.07), 1.11 (0.19), 0.92, and 1.41 (0.21) based on replicate RT-PCR measurements from multiple animals [SEMs in parentheses for the five genes listed above and 2.18, 0.23, 2.77, 0.97 (N.S.), and 4.84 from microarray analysis. N.S., not significantly different from unity]. CR + EX versus AL contrasts also indicated that RT-PCR and microarray estimated expression changes were in the same direction.

This study shows, for the first time, that CR and EX have distinct effects on normal mammary gland gene expression. In addition, the study clearly showed that CR in combination with or without EX has profound effects on animal phenotype, with the response to EX contingent on calorie intake. The results of this study highlight the complex effects of calorie intake compared with EX on gene expression in mammary gland and contribute to efforts aimed at understanding molecular and epidemiologic associations between energy balance and breast cancer.

EX alone (AL+EX) altered an almost completely distinct set of genes in our mouse mammary glands compared with CR alone (Fig. 2A). Furthermore, the effects on gene expression in CR+EX mice differed substantially from the effects of CR (Fig. 2B); no genes were common to both the CR+EX or AL+EX versus AL comparisons. These results and our observation that there were statistically significant interactive effects of diet and running wheel access on major aspects of metabolism, including body temperature, spontaneous locomotor activity, and bone characteristics, all support the argument that metabolic effects of physical activity depend on energy balance. Lu et al. (27) used a very similar experimental design to examine the interactive effects of CR and EX on gene expression in skin tissues from SENCAR mice. Although Lu et al. (27) used treadmill EX instead of running wheels, a different target tissue, 20% CR, and 10 weeks of EX, these authors also found that CR significantly altered many more genes than EX and that there was little overlap in the specific genes altered in the different treatment groups.

The relatively large number of genes unique to AL versus CR (286) compared with AL versus CR+EX (156), and the limited effects of EX in CR mice, are most parsimoniously explained as the consequences of a moderate EX intervention. However, CR+EX did reverse some of the gene expression changes caused by CR alone or AL+EX alone. Further work, possibly in a model with more intense EX treatment, is needed to determine if these represent interactive as well as distinct effects of CR versus EX. A more intense EX regimen might reveal more complete changes of gene expression pathways in a few major areas. The results of this study provide a strong case for such work. An alternative EX regimen such as swimming, or the use of mice bred to be more active, may be required to elicit such EX patterns more consistently than the present voluntary regimen.

The moderate-intensity EX done by mice in this study had significant effects on normal mammary gland gene expression. Most of the transcripts identified as responsive to EX alone are involved in metabolism (catalytic activity, ion transporter activity, transferase activity, etc.) and in particular in metabolism of lipids (Table 3). The effect of EX on lipid metabolism may be more general; a recent study in male C57BL/6J mice reports that EX reverses many of the effects of a high-fat diet on hepatic gene expression, particularly for a number of genes related to fatty acid metabolism (36).

In contrast to the modest effects seen for EX alone in terms of overall number of genes changed relative to AL, CR alone significantly altered the expression of a large number of genes (425 total). This finding is consistent with the notion that CR affects cancer risk in both humans and animal models by changing the expression of genes in many pathways and in many tissues, including the mammary gland. Analysis of specific gene changes revealed that there were 139 gene expression changes common to both CR and CR+EX versus AL mice. These genes represent 32.7% of the overall gene changes in the first comparison and 47.1% in the second. It is also noteworthy that all of them were altered in the same direction in both comparisons and generally fall into categories/pathways relevant to mammary gland development and carcinogenesis (Fig. 3 and Table 3). The categories/pathways into which they grouped included differentiation and morphogenesis; regulation of cell cycle and proliferation; apoptosis; DNA repair/response to DNA damage stimulus; lipid metabolism and biosynthesis; the WNT, MAPK, and IGF-I pathways; and the leptin pathway. We discuss some aspects of these results below.

Genes that are differentially expressed in the mammary glands of CR and CR+EX mice are involved in mammary gland growth, differentiation, and morphogenesis. The CD24a antigen gene was downregulated in both CR versus AL and CR+EX versus AL. This protein plays a pivotal role in cell differentiation and was originally identified as downregulated in primitive ectoderm, mesoderm, and ventral endoderm when organogenesis is completed (37). In addition, the Krct and heat shock protein 9 genes were upregulated in mammary glands of both CR and CR+EX mice. Krct (or Stk16/Krct) is a member of the family of serine/threonine protein kinases that is involved in end bud morphogenesis in murine mammary gland development and plays an important role in regulating stromal-epithelial interactions during ductal morphogenesis (38). In a transgenic mouse model, a modest overexpression of Krct in the mammary gland during puberty results in duplication of the terminal end bud axis (38). Future studies examining the effects of CR and/or EX on the growth and morphologic development of the mammary gland and whether there are downstream effects on cancer risk would contribute to our mechanistic understanding of the influence of these interventions.

The claudin 3 and 4 genes were both downregulated by CR and CR+EX, although the results reached statistical significance only in the CR+EX mice. The claudins are members of a family of transmembrane proteins that are critical in the maintenance of epithelial and endothelial tight junctions and may play a role in the cytoskeleton in cell signaling (39). Claudins 3 and 4 are differentially expressed during normal mammary gland development in the mouse and mammary gland differentiation processes of pregnancy, lactation, and involution, suggesting a role for these proteins at different stages of mammary gland function (40). Specific claudins are overexpressed in a wide variety of cancer types, although their functional role in cancer progression remains unclear (39).

Prolactin receptor gene expression was also decreased by CR and CR+EX compared with AL mice, although the effect was not statistically significant in the former. This gene is an important member of the set of prolactin-regulated genes that mediate prolactin-driven mammary development (41, 42). Decreased prolactin levels are associated with decreased risk of breast cancer in women (43), but this association is not found consistently in epidemiologic studies (44). Ongoing work may shed additional light on this association.

Several genes involved in cell cycle regulation and cell proliferation, including Wee1 kinase, Cyclin G1, Cyclin D1, and Mycn, were also all significantly overexpressed in mammary glands from CR and/or CR+EX mice (Table 3). These genes are negative regulators of different checkpoints in the cell cycle, during the G₁-S transition as well as the G₂ checkpoint (45–48). Overexpression of Dapk1 and Casp7 (significant only for CR) and the inhibition of GADD45 (significant both in CR and CR+EX mice) also show a possible role for CR and CR+EX in promoting the apoptotic process (49). Dunn et al. (50) reported that CR increases the rate of apoptosis in mouse bladder cancers, supporting a role for CR in increasing apoptosis. As for EX, the GADD45 transcript was also found to be downregulated by EX training in hepatic transcriptional profiles of mice submitted to a high-fat diet (36).

Our analyses also identified changes in many genes common to the CR and CR+EX groups that are involved in lipid metabolism, in particular upregulation of genes related to the elongation of fatty acids chains (Tables 3 and 4). A higher level of fatty acid serum concentration, usually due to an increased release from adipose tissue in obesity, is a possible early effect of obesity related to the development of insulin resistance and compensatory hyperinsulinaemia (3), an effect opposite of what might be happening in our model.

The gene-encoding leptin was downregulated in the CR (P < 0.05) and CR+EX (N.S.) mice compared with the AL mice. Circulating leptin concentrations are proportional to body fat content; thus, this finding supports earlier findings of decreased leptin in the serum of CR mice (1). Although most studies have failed to relate leptin serum levels to breast cancer risk (51, 52), in vitro studies show that leptin stimulates the proliferation of benign and malignant epithelial breast cells and tumor cell invasion in breast cancer cell lines (53–55). Leptin is produced by breast cancer cells, and thus local leptin production by malignant cells as well as adipocytes rather than serum leptin may play a critical role for breast cancer development (56).

Finally, we investigated the effects of EX in the context of CR compared with CR alone (CR+EX versus CR) and identified 30 genes that were significantly altered (4 up and 26 down). The genes differentially expressed are primarily involved in motile and catabolic processes (Table 4). No genes were found in common between CR versus CR+EX and AL versus AL+EX comparisons. Again, the EX effect on gene expression seems to be dependent on caloric intake level. However, we cannot determine in this experimental design if this dependence is mediated through the amount of EX (CR animals ran less than AL animals) or through some other moderating effect of nutritional status on EX effects.

There are some limitations to the present study. Notably, we analyzed gene expression changes in whole, undissected samples of mammary tissue that included epithelial cells, adipocytes, and other cell types. Adipocytes alone, as well as other cell types, are known to exhibit distinct changes in gene expression in response to diet (28, 29). The present study does not allow us to determine which specific cell types are most responsive to the effects of CR versus EX. However, the tissue analyzed in this study does represent the combined cellular milieu in which mammary tumors arise. The hypothesis being tested by these analyses is that CR and EX differentially affect pathways in the mammary gland related to breast cancer susceptibility. Recent evidence suggests that multiple cellular compartments within the mammary gland, in addition to the epithelium, contribute to the development and progression of breast cancer. We therefore chose to characterize global gene expression patterns in the entire mammary gland to begin to understand the common and unique effects of CR and EX on breast cancer susceptibility.

Second, past studies of EX in rodents have had mixed results (57), and it is currently not clear what factors underlie the heterogeneity of EX effects in mouse models. The effects of CR in rodents have been much more consistent and robust, consistent with the notion that CR and EX exert there effects through different mechanisms. Thus, it is not clear to what extent rodents and mice in particular represent models for diet and physical activity effects on carcinogenesis. Nonetheless, mice have been used as models for the influence of CR on mammary carcinogenesis (1), and the mice in this study showed significant phenotypic responses to both diet and EX treatments. Thus, they represent at least a model for influences of preventive interventions on the biological environment found in the mammary gland and a source of ideas and hypotheses concerning targets for further understanding mechanisms for breast cancer prevention related to energy balance.

In summary, this study shows that CR has specific and significant effects on the expression of genes in the normal mouse mammary gland, including many genes implicated in carcinogenesis. Furthermore, we observed distinctive effects of calorie intake and EX on gene expression in mouse mammary glands. Studies such as this one complement epidemiologic evidence indicating that physical activity reduces risk of breast cancer in postmenopausal women independent of obesity and provide clues for future work aimed at identifying the mechanist bases of such associations and designing physical activity–related interventions for breast cancer prevention.

No potential conflicts of interest were disclosed.

We thank Dr. Stuart H. Yuspa and the members of the Laboratory of Cancer Biology and Genetics (LCBG) for housing and mentoring M. Padovani during her final year at the National Cancer Institute and for generously offering their valuable advice. Lisa Riffle and Dan Logsdon ably assisted with animal care and surgery and Heather L. Hill performed the IGF-I assays.