β-Carotene Oxygenase 2 Genotype Modulates the Impact of Dietary Lycopene on Gene Expression during Early TRAMP Prostate Carcinogenesis

Abstract

Background

Epidemiologic studies suggest lycopene and tomato intake are inversely associated with human prostate cancer incidence. In the genetically driven murine prostate carcinogenesis model transgenic adenocarcinoma of the mouse prostate (TRAMP), prostate cancer is inhibited by feeding of lycopene or tomatoes, and these effects are modulated by the β-carotene oxygenase 2 (Bco2) genotype.

Objective

We sought insight into this interaction through evaluation of prostate gene expression patterns during early TRAMP carcinogenesis.

Methods

Three-week-old TRAMP/+ or TRAMP/– × Bco2^+/+ or Bco2^–/– mice were fed a control, lycopene beadlet, or 10% tomato powder–containing semipurified diet (providing 0, 384 and 462 mg lycopene/kg diet, respectively) for 5 wk. Gene expression patterns were evaluated by prostate cancer- and cholesterol and lipoprotein metabolism-focused arrays at age 8 wk.

Results

The TRAMP genotype profoundly alters gene expression patterns, specifically inducing pathways associated with cell survival [z-score = 2.09, –log(P value) = 29.2, p53 signaling (z-score 1.13, –log(P value) = 13.5], and phosphatidylinositol-3 kinase (PI3K)/protein kinase B (AKT) signaling [z-score = 0.302, –log(P value) = 12.1], while repressing phosphatase and tensin homolog (PTEN) signaling [(z-score = –0.905, –log(P value) = 12.3], cholesterol synthesis [z-score = –1.941, –log(P-value) = 26.2], and LXR/RXR pathway activation [z-score = –1.941, –log(P value) = 23.1]. In comparison, lycopene- and tomato-feeding modestly modulate strong procarcinogenic TRAMP signaling. Lycopene decreased gene expression related to carcinogenesis [ Nkx3-1(NK3 homeobox 1)], tomato feeding increased expression of a gene involved in circadian regulation [Arntl (aryl hydrocarbon receptor nuclear translocator like)], and tomato and/or lycopene increased expression of genes involved in lipid metabolism [Fasn (fatty acid synthase), Acaca(acetyl-CoA carboxylase alpha), Srebf1 (sterol regulatory element binding transcription factor 1), Hmgcr (3-hydroxy-3-methylglutaryl-coA reductase), and Ptgs1 (prostaglandin-endoperoxide synthase 1)] (all P < 0.05). The impact of Bco2 genotype was limited to a subset of lycopene-impacted genes [Apc (adenomatous polyposis coli), Mto1 (mitochondrial TRNA translation optimization 1), Nfkb1 (nuclear factor kappa B subunit 1), andRbm39 (RNA binding motif protein 39)].

Conclusions

The TRAMP genotype strongly impacts procarcinogenic gene expression prior to emergence of histopathologic disease. Dietary tomato and lycopene modestly temper these processes, while Bco2 genotype has a limited impact at this early stage. These observed patterns provide insight into the complex interactions between a dietary variable, here tomatoes and lycopene, genes impacting nutrient metabolism, and their modulating influences on oncogene-driven prostate carcinogenesis. These findings provide further mechanistic support, consistent with cancer outcomes in rodents experiments and human epidemiologic studies.

Introduction

Dietary intake of tomato products and greater blood concentrations of lycopene, a bioactive carotenoid in tomatoes, are inversely associated with the incidence of advanced and lethal prostate cancer (1). An analysis of the Health Professionals Follow-Up Study (HPFS) suggests that intakes of tomato lycopene reported earlier in life are more strongly associated with a reduced risk of total and advanced/lethal prostate cancer than intakes reported immediately before diagnosis (2). Similarly, greater tomato intake during adolescence was recently associated with a reduced risk of non-advanced prostate cancer (3). Preclinical studies of murine and rat prostate carcinogenesis have shown that initiation of tomato and/or lycopene feeding early in life in either genetically-driven models or soon after carcinogen exposure in chemically-induced models reduces cancer incidence and severity (4, 5, 6). Together, these investigations support a hypothesis that tomato components, including lycopene, may be most protective against early molecular events in the prostate carcinogenesis cascade.

The transgenic adenocarcinoma of the mouse prostate (TRAMP) model has been useful for studying dietary impacts on prostate carcinogenesis [reviewed in (7)]. The TRAMP model recapitulates the initiation and progression of aggressive human prostate cancer through androgen-driven prostate-specific expression of the SV40 (Simian virus 40)T-antigen oncogene, primarily through inactivation of Trp53 (transformation related protein 53) and Rb (retinoblastoma). An accumulation of genetic damage in TRAMP mice accelerates following puberty, leading to a predictable progression over several months from normal epithelium to prostatic intraepithelial neoplasia and then to carcinoma (7, 8). Pannellini et al. first reported that feeding 10% tomato powder–containing diets to TRAMP mice from 5 wk of age resulted in significantly improved 45-wk survival (67% survival) compared with control diet–fed mice (11% survival; P = 0.0018), along with lower rates of metastasis (33% compared with 6%, respectively) (5). Tomato powder feeding also led to a shift toward normal and premalignant prostate histology compared with adenocarcinoma that predominated with control diet feeding, an effect that was apparent as early as 12 wk of age in TRAMP mice (5). Importantly, dietary treatments did not affect expression of the SV40 T-antigen which drives TRAMP carcinogenesis.

We previously found that diets containing tomato powder or purified lycopene significantly reduced prostate cancer incidence in TRAMP mice (6). Furthermore, we found that this effect was modified by the expression of the eccentric carotenoid cleavage enzyme gene, β-carotene-9′,10′-oxygenase (aka β-carotene oxygenase 2, Bco2), such that lycopene exerted a stronger protective effect in wild-type Bco2^+/+ mice, compared with Bco2^–/– mice (10% compared with 50% cancer incidence, respectively, genotype effect P = 0.0004; control diet–fed mice showed 70–80% incidence) (6). In contrast, the protection of tomato feeding was less dependent on Bco2 genotype (15% incidence in Bco2^+/+and 30% in Bco2^–/–, genotype effect P = 0.0383) (6). BCO2 is 1 of 2 known mammalian carotenoid cleavage enzymes that can metabolize lycopene (9, 10). Ablation of BCO2 in mice results in greater circulating and tissue concentrations of lycopene (6, 11, 12). Thus, the aforementioned results may suggest that BCO2-generated lycopene metabolites are potential mediators of lycopene's prostate cancer preventive bioactivity. The effect of tomato feeding on prostate carcinogenesis may be less dependent on Bco2 genotype, because tomatoes deliver a variety of potentially bioactive compounds (13), possibly offering anticancer activity redundant to that of Bco2-generated lycopene metabolites.

Specifics regarding how tomato components inhibit carcinogenesis remain elusive, although several hypotheses focus on antioxidant and anti-inflammatory properties [reviewed in (14, 15)] and decreasing intracellular cholesterol pools available for steroidogenesis and proliferation (16, 17). Lycopene metabolites may also act as weak nuclear receptor agonists or antagonists to directly alter gene expression (15, 18). In 10-wk-old Bco2^+/+ TRAMP mice, we found that both tomato and lycopene feeding during early carcinogenesis led to similar changes in expression of genes associated with androgen metabolism, MAPK (mitogen-activated protein kinase 1) signaling, p53 signaling, cell adhesion, TR/RXR (thyroid hormone receptor/retinoid X receptors) and VDR/RXR (vitamin D receptor/retinoid X receptors) activation, phase II detoxification, and endocytosis (19). However, whether Bco2 genotype modulates the prostatic gene expression response to dietary lycopene in the early phases of carcinogenesis, when tomato components may be most impactful, is unknown.

The objective of this study was to model, in a well-controlled experiment, a scenario that is likely very relevant to the early prostate cancer cascade in humans. We examined how 3 factors relevant to human cancer may interact in a controlled study examining 2 targeted gene arrays as the outcome. First, we examined the impact of the strongly procarcinogenic effects of p53 and Rb inhibition, by comparison of wild type and TRAMP. Second, we imposed a nutritionally relevant dietary intervention of tomato components and explored the capacity to inhibit or attenuate procarcinogenic gene expression networks. Finally, we examined a mechanistically relevant gene–diet interaction, examining the impact of inhibiting lycopene metabolism through genetic loss of Bco2, with the hypothesis that the effects of lycopene or tomato intake may be dependent upon Bco2 genotype. We examined this 3-factor interaction on a targeted prostate cancer gene expression array as the primary outcome. As a secondary outcome, we examined the impact of TRAMP genotype, Bco2 status, and tomato components on a panel of genes related to cholesterol and lipoprotein metabolism. This effort is based on studies suggesting that tomato lycopene impacts host lipid metabolism (12, 15, 16, 17, 19) and that carotenoid uptake in a target tissue such as the prostate is linked to lipoprotein uptake and metabolism (20).

Methods

Animal breeding strategy, diets, and experimental design

All animal protocols and procedures were reviewed and approved by the Ohio State University Institutional Animal Care and Use Committee and conducted according to ethical standards. A 2 × 2 × 3 factorial study design was completed with 4 genotypes of mice [TRAMP/–:Bco2 ^+/+, TRAMP/+:Bco2 ^+/+, TRAMP/–: Bco2 ^–/–, TRAMP/+:Bco2 ^–/–]. Bco2^–/–: (B6; 128S6-Bco2tm1Dnp) mice were generated as previously described (6, 21). Breeding pairs were generously provided by Johannes von Lintig (Case Western Reserve). TRAMP [C57BL/6-Tg(TRAMP)824Ng/J; The Jackson Laboratory] and Bco2^–/– breeding colonies were established and maintained at The Ohio State University. Mice were crossed as previously described to obtain the 4 desired genotypes (6). Mice were provided 1 of 3 dietary interventions: a control AIN-93G diet, the control diet containing 10% (wt/wt) tomato powder, and the control diet supplemented with 0.25% (wt/wt ) water-soluble lycopene beadlets (10% lycopene beadlets) (6), upon weaning at age 3 wk (n = 8/TRAMP genotype × Bco2 genotype × diet group). Diets were formulated to provide similar lycopene concentrations (250 mg lycopene⋅kg diet^–1), and concentrations quantified by HPLC (6) were found to deliver 384 and 462 mg lycopene/kg diet from the tomato and lycopene diets, respectively. Diet storage and provision of fresh diet were as described previously (6). After 5 wk of feeding, nonfasted mice were terminated by CO₂ asphyxiation, blood was collected by cardiac puncture for serum carotenoid analysis, and prostate lobes were microdissected, snap frozen, and processed for RNA and protein analysis.

Serum carotenoid analysis

Serum carotenoids were analyzed as previously described (12). Lycopene concentrations were measured for each TRAMP genotype × Bco2 genotype diet group (n = 3). In tomato-fed mice, an additional sample for each TRAMP genotype × Bco2 genotype group (n = 4) was analyzed for quantitation of the less abundant tomato carotenoids phytoene, phytofluene, ζ-carotene, and β-carotene.

mRNA and Protein isolation

mRNA was isolated using spin columns from microdissected dorsolateral prostate lobes (RNA/Protein Purification Plus Kit, Norgen Biotek) per the manufacturer's protocols to include on-column DNase I treatment and RNA cleanup and concentration. Isolated mRNA was used to derive cDNA using the RT2 Easy First Strand Kit (Qiagen Science Inc).

Prostatic Expression of Genes Associated with Prostate Cancer

We measured the expression of 84 prostate cancer–related genes (Supplemental Table 1) using a commercially available qRT-PCR array (RT² Profiler Mouse prostate cancer PCR array) per the manufacturer's protocols. The tool is a targeted array of preselected genes that cover prostate cancer–related processes and gene expression features, including differentially methylated promoters, genes for which expression is upregulated or downregulated in prostate cancer, metastatic potential, androgen receptor signaling, AKT and PI3 kinase signaling, PTEN signaling, apoptosis, cell cycle, transcription factors, fatty acid metabolism, and other prostate cancer–related genes.

Expression of Genes Related to Cholesterol and Lipoprotein Metabolism

We measured the expression of 90 genes related to cholesterol and lipoprotein metabolism (Supplemental Table 2) digitally by fluorescent molecular barcode detection technology (Cholesterol and Lipoprotein Metabolism Plex Set, nCounter Elements, Nanostring).

Pathway Analysis of Prostate Cancer and Lipid Metabolism-Related Gene Expression

Ingenuity Pathway Analysis (IPA, Qiagen) software was used (12) for separate pathway analyses of each set of array data. In brief, IPA was used to predict the effects of TRAMP genotype across all Bco2 genotype and diet groups on canonical pathway signaling pathways and biological functions (predicted directional biological outcomes based on a potentially multiple pathways) that can be inferred from prostatic gene expression. Relative expression ratios of TRAMP/+:TRAMP/− mice and TRAMP genotype main effect P values acquired from 3-way ANOVA tests were used for analysis.

Statistical Analysis

The effects of dietary treatment, Bco2 genotype, TRAMP genotype, and corresponding interactions on serum lycopene concentrations and end-of-study body masses were evaluated by 3-way ANOVA (α = 0.05). The effects of Bco2 genotype and TRAMP genotype on serum concentrations of the tomato carotenoids phytoene, phytofluene, and ζ-carotene were analyzed using 2-way ANOVA in only the tomato powder–fed mice.

Expression of 84 prostate cancer–related genes (n = 5/group) was normalized to the mean expression of 4 genes [Actb (beta-actin), B2m (beta-2 microglobulin), Gusb(glucuronidase, beta), Hsp90ab1(heat shock protein 90 alpha (cytosolic) class B member 1)]. These 4 normalizer genes had low variability in cycle number at threshold (Ct) across treatment groups, and the mean value of the 4 was not subject to significant Bco2 genotype, TRAMP genotype, or diet main effects or interactions. The effects of TRAMP genotype, Bco2 genotype, and diet on natural log–transformed, normalized gene expression (2^-dCt) was determined by 3-way ANOVA (α = 0.05). Expression of 90 cholesterol and lipoprotein metabolism–related genes was analyzed (n = 5/group) within the nSolver software package (NanoString). Normalizer probes {Hprt1 [hypoxanthine guanine phosphoribosyl transferase], Rplp0 [ribosomal protein, large, P0], Polr1b [polymerase (RNA) I polypeptide B], Abcf1 [ATP-binding cassette, sub-family F (GCN20), member 1], Gusb, and Ldha (lactate dehydrogenase A]} were selected according to the geNorm algorithm (2,2). For this semi-targeted analysis of cholesterol and lipoprotein metabolism–related gene expression, the overall main and interaction effects of Bco2 genotype, TRAMP genotype, and diet treatment were evaluated by 3-way ANOVA, with an uncorrected significance cutoff of 0.05 with Holm-Sidak post hoc testing using Sigmaplot. Least squared means for treatment effects were estimated and used to estimate the effect size of experimental conditions by calculating the ratio of TRAMP/+ compared with TRAMP/–, Bco2^–/–: Bco2^+/+, and the average ratio of tomato-fed:control-fed and lycopene-fed:control fed mice. Previously, we found that a main effect of tomato and lycopene feeding led to a 15% decrease in expression of Igf1 (insulin-like growth factor 1), among other genes, during early TRAMP carcinogenesis (19). Assuming a 66% SD, a combined n = 10 (from both TRAMP/+ and TRAMP/−) provides 89% power to detect a significant post hoc difference between Bco2^+/+ and Bco2^–/–: mice fed the same diet at α = 0.05; thus, group sizes were set to n = 5/diet × Bco2 × TRAMP group. Post hoc group and interaction comparisons for the cholesterol and lipoprotein metabolic gene panel were considered significant if P < 0.05. False discovery rate–corrected P values were calculated by Benjamini-Hochberg (23) and were inspected. All ANOVA and post hoc tests were conducted using SigmaPlot 13.0 (Systat Software Inc.).

Results

Effect of Diets and Genotypes on Growth

There were no significant impacts of diet or genotype on body weights at the termination of the study (25.1 ± 3.4 g; mean ± SD).

Effects of Diets and Genotypes on Serum Carotenoid Concentrations

Serum carotenoid concentrations are presented in Table 1. Carotenoids were not detected in the serum of mice fed the control diet. Serum lycopene concentrations in lycopene- and tomato-fed mice increased similarly. Phytoene, phytofluene, and ζ-carotene were detectable only in tomato-fed mice, while β-carotene was not detected in any mice. Lycopene concentrations in Bco2^–/– mice were greater than those in Bco2^+/+ mice (P = 0.004) for both tomato- and lycopene-fed mice and were not impacted by TRAMP genotype.

TABLE 1.

Serum carotenoid concentrations of TRAMP/+ or TRAMP/– × Bco2^+/+ or Bco2^–/–: mice fed control or lycopene- or tomato powder-containing diets from weaning to 8 wk¹

	Control Diet				Lycopene				10% Tomato Powder
	Bco2+/+		Bco2–/–		Bco2+/+		Bco2–/–		Bco2+/+		Bco2–/–		P values for main effects2
	TRAMP/−	TRAMP/+	TRAMP/−	TRAMP/+	TRAMP/−	TRAMP/+	TRAMP/−	TRAMP/+	TRAMP/−	TRAMP/+	TRAMP/−	TRAMP/+	Diet	Bco2	TRAMP
Lycopene	ND3	ND	ND	ND	0.16 ± 0.02 a	0.19 ± 0.03a	0.37 ± 0.11b	0.34 ± 0.08b	0.20 ± 0.04a	0.21 ± 0.03a	0.32 ± 0.05b	0.26 ± 0.07b	0.656	0.004	0.772
Phytoene	ND	ND	ND	ND	ND	ND	ND	ND	0.03 ± 0.02	0.03 ± 0.01	0.04 ± 0.01	0.04 ± 0.01	—	0.652	0.783
Phytofluene	ND	ND	ND	ND	ND	ND	ND	ND	0.06 ± 0.02	0.06 ± 0.00	0.06 ± 0.01	0.06 ± 0.01	—	0.747	0.826
Zeta-carotene	ND	ND	ND	ND	ND	ND	ND	ND	0.02 ± 0.01	0.02 ± 0.00	0.02 ± 0.00	0.03 ± 0.00	—	0.634	0.89

¹Values are mean ± SD μmol/L unless otherwise indicated. Lycopene model included lycopene- and tomato powder–fed groups since no carotenoids were detected in control-fed mice. Models for phytoene, phytofluene, and ζ-carotene included only the tomato powder–fed group, as the other carotenoids were not detected in control- or lycopene-fed mice. Bco2, β-carotene-9′,10′-oxygenase; TRAMP, transgenic adenocarcinoma of the mouse prostate; ND, not detected.

²No significant 2- or 3-way interactions were detected. P < 0.05 was considered significant. n = 3–4/group. Means with differing superscript letters are significantly different. β-Carotene was not detected in any sample.

³ND, not detected.

Main Effects of TRAMP Genotype on Prostate Carcinogenesis Gene Expression

We detected expression of 79 of the 84 genes included in the prostate cancer-focused PCR array (Figure 1). Five genes were deemed undetectable with Ct values >35 {Ccna1 [cyclin A1], Il6[interleukin 6], Klkb1[kallikrein B, plasma 1], Shbg[sex hormone binding globulin], and Slc5aa8 [solute carrier family 5 (iodide transporter), member 8]}. Twenty-six of the 79 detected genes (33%) in the carcinogenesis array were not significantly changed by the genetic or dietary factors. Forty-nine of the 79 expressed genes (62%) were significantly altered by TRAMP genotype at this early time point in the carcinogenesis cascade (Figure 1). Canonical pathways and biological or disease functions impacted by TRAMP genotype are shown in Figure 1 and Supplemental Table 3. The TRAMP genotype showed its strongest impact on prostate cancer signaling pathways, as expected, with significant effects noted on p53, estrogen receptor, phosphatase and tensin homolog (PTEN), and phosphatidylinositol-3 kinase (PI3K)/protein kinase B (AKT) signaling, among others (Figure 1, Supplemental Table 3). Biological functions predicted to be upregulated in TRAMP/+ prostates included cell survival [z-score = 2.091, –log(P-value) = 29.2] and cell viability [(z-score 2.168, –log(P-value) = 28.0] among others (Figure 1, Supplemental Table 3). Biological functions predicted to be downregulated in TRAMP/+ included mortality [z-score = –3.708, –log(P-value) = 25.1] and organismal death [z-score = 2.091, –log(P-value) = 29.2] and were consistent with a survival advantage observed in cancer cells and suppression of gene expression patterns associated with a benign cancer phenotype. Genes included in each biological function category are shown in Supplemental Table 3.

FIGURE 1.

Effect of genetic and dietary variables on prostate cancer-related gene expression in the dorsolateral prostate in TRAMP/+ or TRAMP/- × Bco2^+/+ or Bco2^–/–: mice. (A) Venn diagram of the 84 prostate cancer–related genes evaluated demonstrating main effects of TRAMP genotype, Bco2 genotype, or diet, P < 0.05, n = 5, TRAMP genotype × Bco2 genotype × diet group. (B) Genes impacted by a main effect of the TRAMP genotype, n = 30, TRAMP genotype group. (C) Top 10 (of 363 calculated) canonical pathways predicted to differ between TRAMP/+ compared with TRAMP/− mice, across diets, based on prostate cancer–related gene expression strength, –log(P value) and direction (z-score), n = 30, TRAMP genotype group. (D) Biological functions predicted to differ by TRAMP genotype, |z-score| >2, n = 30, TRAMP genotype group. (E) Prostate cancer–related gene expression impacted by a main effect of lycopene feeding, and (F) tomato feeding, n = 10/diet group. Symbols indicate significant differences relative to control. Bco2, β-carotene-9′,10′-oxygenase; TRAMP, transgenic adenocarcinoma of the mouse prostate.

Main Effects of Diet on Prostate Carcinogenesis Gene Expression

As hypothesized for a modulatory effect of diet on the carcinogenesis cascade, we observed that 11 of 79 expressed genes (14%) demonstrated a change in expression due to diet (Figure 1). Eight of the 11 diet-sensitive genes showed significant post hoc differences between lycopene or tomato compared with control diets and are shown in Figure 1, while 3 genes [Akt1(Thymoma viral proto-oncogene 1), Socs3(Suppressor of cytokine signaling 3), Tgfb1i1(transforming growth factor beta 1 induced transcript 1), not shown] did not differ significantly by post hoc comparison.

Main Effects of Bco2 Genotype on Prostate Carcinogenesis Gene Expression

Only 2 genes in the prostate cancer array were impacted by a main effect of Bco2 genotype: Ppp2r1b [protein phosphatase 2 (formerly 2A), regulatory subunit A (PR 65), beta isoform] and Ptgs1(prostaglandin-endoperoxide synthase 1) were both modestly upregulated in Bco2^–/–: compared with Bco2^+/+ mice (0.2- to 0.4-fold; P = 0.021 and P = 0.042, respectively).

Interactions of Genotypes × Diet on Prostate Carcinogenesis Gene Expression

The potential for intricate interactions is high when 3 variables are evaluated (diet × Bco2 genotype × TRAMP genotype) and when employing a gene expression array. Seven of the 79 genes (9%) demonstrated significant 3-way interactions (Supplemental Figure 1). These include Cdkn1a [cyclin-dependent kinase inhibitor 1A (P21), P = 0.046], Nkx3-1 (P = 0.036), Ppp2r1b (P = 0.047), Sox4 (SRY-box containing gene 4, P = 0.027), Tfpi2 (tissue factor pathway inhibitor 2, P = 0.012), Tgfb1i1 (P = 0.038), and Timp3 (tissue inhibitor of metalloproteinase 3,P = 0.025).

Four genes were significantly impacted by a 2-way Bco2 genotype × diet interaction [Apc(adenomatous polyposis coli),Mto1(mitochondrial TRNA translation optimization 1),Nfkb1(nuclear factor kappa B subunit 1), andRbm39(RNA binding motif protein 39)] (to slightly increase the score (Figure 2). All 4 were subject to a main effect of TRAMP genotype (Figure 1), in which they were upregulated in TRAMP/+ mice. Post hoc testing revealed that only the effect of lycopene was significantly modulated by Bco2 genotype, such that expression of all 4 genes was lower in the Bco2^–/–: mice than the Bco2^+/+ mice.

FIGURE 2.

(A–D) Prostatic gene expression significantly impacted by an interactive effect of diet with Bco2 genotype, in TRAMP/+ or TRAMP/− × Bco2^+/+ or Bco2^–/–: mice. Bars represent the average gene expression (2^-dCt) in all Bco2^+/+ and Bco2^–/–: mice, across TRAMP genotypes, n = 10/Bco2 genotype × diet group. Asterisks indicate a significant Holm-Sidak post hoc difference between groups. Apc, adenomatous polyposis coli;Bco2, β-carotene-9′,10′-oxygenase; Mto1, mitochondrial TRNA translation optimization 1;Nfkb1, nuclear factor kappa B subunit 1; Rbm39,RNA binding motif protein 39; TRAMP, transgenic adenocarcinoma of the mouse prostate.

The procarcinogenic effect of the TRAMP genotype was not impacted by Bco2 genotype. As seen in Figure 3, we detected interactions between diet and TRAMP genotype for 3 genes. Abcb1b [ATP-binding cassette, sub-family B (MDR/TAP), member 1B] and Cav2 (caveolin 2) exhibited lower expression in lycopene-fed TRAMP/+ mice than in lycopene-fed TRAMP/– mice (P = 0.007 and P = 0.007, respectively) (Figure 3). In TRAMP/+ mice, tomato increased expression of the Arntl and Cav2 genes relative to mice fed lycopene (P = 0.001 and P = 0.011, respectively). Finally, in TRAMP/+ mice, relative to mice fed the control diet, tomato feeding increased prostatic Arntl (aryl hydrocarbon receptor nuclear translocator-like) expression nearly 2-fold (P < 0.05).

FIGURE 3.

(A-C) Prostatic gene expression significantly impacted by an interaction effect of diet type with TRAMP genotype in TRAMP/+ or TRAMP/– × Bco2^+/+ or Bco2^–/–: mice. Bars represent the average gene expression (2^-dCt) in all Bco2^+/+ and Bco2^–/–: mice, across all TRAMP/– and TRAMP/+ mice, across Bco2 genotypes, n = 10, TRAMP genotype × diet group. Asterisks indicate a significant post hoc difference between groups. Bco2, β-carotene-9′,10′-oxygenase; TRAMP, transgenic adenocarcinoma of the mouse prostate.

Main and Interactive Effects of TRAMP Genotype, Bco2 Genotype, and Diet on Expression of Cholesterol and Lipoprotein Metabolism–Related Genes

We observed main effects of the TRAMP genotype on expression of 39 of 90 (43%) lipid metabolism genes (Figure 4). Clearly, the main impact of TRAMP/+ genotype was a reduction in the expression of lipid metabolism genes, with 35 genes showing significantly reduced expression. This was in marked contrast to the predominantly upregulatory impact of the TRAMP/+ genotype carcinogenesis-related genes (Figure 4, Supplemental Table 4). Canonical pathway analysis revealed that TRAMP/+ mice demonstrate a decrease in multiple lipid synthetic and signaling pathways (activation z-score = 2) at this early stage of carcinogenesis (Figure 4, Supplemental Table 5). In addition, we saw a decrease in 2 bioactive lipid–nuclear receptor signaling pathways: LXR/RXR activation and PPAR-alpha (peroxisome proliferator activated receptor alpha)/RXR-α activation (Figure 4, Supplemental Table 5). Some metabolism-related biological functions were predicted to be increased (activation z-score >2), including insulin resistance and disordered glucose metabolism, while other functions predominantly related to lipid and cholesterol transport were predicted to be decreased (activation z-score = 2) (Figure 4, Supplemental Table 5).

FIGURE 4.

Treatment effects and pathway analysis of cholesterol and lipoprotein metabolism–related prostatic gene expression in TRAMP/+ or TRAMP/– × Bco2^+/+ or Bco2^–/–: mice. (A) Venn diagram demonstrating the number of genes for which expression is affected by either TRAMP genotype, Bco2 genotype, or diet treatment main effects, P < 0.05, n = 5, TRAMP genotype × Bco2 genotype × diet group. (B) Genes impacted by a main effect of the TRAMP genotype, n = 30, TRAMP genotype group. (C) Canonical pathways predicted to differ between TRAMP and wild type mice based upon expression of cholesterol and lipoprotein metabolism–related genes, strength [log(P-value)] and direction (no direction determined, gray bars; positive z-score, black bars; negative z-score, white bars) n = 30/TRAMP genotype group. (D) Biological functions predicted to differ between TRAMP/+ and TRAMP/– mice based upon expression of cholesterol and lipoprotein metabolism–related genes, n = 30, TRAMP genotype group. Symbols indicate significant differences. Bco2, β-carotene-9′,10′-oxygenase; TRAMP, transgenic adenocarcinoma of the mouse prostate.

We observed that dietary tomato and lycopene had a limited impact on genes affecting lipid metabolism, with 4 of 90 genes (4%) showing a main effect of tomato/lycopene feeding [Apoa5 (apolipoprotein A-V), Srebf1, Lcat(lecithin cholesterol acyltransferase), and Ppard(peroxisome proliferator activator receptor delta)] (Table 2). The Bco2 genotype demonstrated a main effect on the expression of 2 genes [Stard3(START domain containing 3) and Cnbp(cellular nucleic acid binding protein)].

TABLE 2.

Prostatic gene expression of cholesterol and lipoprotein metabolism–related genes significantly impacted by tomato and/or lycopene compared with control feeding in TRAMP/+ or TRAMP/– × Bco2^+/+ or Bco2^–/–: mice¹

	Diet main effect unadjusted P value2	Lycopene: control		Tomato: control expression ratio
Gene		Expression ratio3	P value4	Expression ratio3	P value4
Apoa55	0.036	1.107*	0.031	1.056	0.332
Ppard	0.031	1.089	0.283	0.918	0.199
Lcat	0.038	1.028	0.731	1.204	0.052
Srebf1	0.022	1.237*	0.019	1.103	0.228

¹Apoa5, apolipoprotein A-V;Bco2, β-carotene-9′,10′-oxygenase; Srebf1, sterol regulatory element binding transcription factor 1; Lcat, lecithin cholesterol acyltransferase;Ppard, peroxisome proliferator activator receptor delta; TRAMP, transgenic adenocarcinoma of the mouse prostate.

²Main effect P-values were determined by 3-way ANOVA models.

³Expression ratio calculated from the least squares mean across all treatment groups.

⁴Post hoc pairwise difference between lycopene compared with control or tomato compared with control determined by Holm-Sidak test.

⁵For a complete list of all genes assayed on the cholesterol and lipoprotein metabolism focused expression panel, refer to Supplemental Table 2.

^*Indicates significant differences between treatment and control mean expression.

There were several genes influenced by significant 2-way or 3-way interactions, which also demonstrated significant post hoc group differences (Figure 5). The interactions are complex and suggest that there are a limited number of diet effects that differed by both TRAMP and Bco2 genotype {Cela3b[chymotrypsin-like elastase family, member 3B], Lpa [lipoprotein(A)], and Sorl1 [sortilin-related receptor, LDLR class A repeats-containing)]}, TRAMP genotype [Hmgcr (3-hydroxy-3-methylglutaryl-Coenzyme A reductase)], or Bco2 genotype [Npc1(Niemann-Pick type C1) and Lipa (lysosomal acid lipase A)].

FIGURE 5.

Effects of diet type and genotype on prostatic gene expression related to cholesterol or lipoprotein metabolism in TRAMP/+ or TRAMP/– × Bco2^+/+ or Bco2^–/–: mice. (A–C) Genes subject to a significant 3-way interaction of diet, Bco2 genotype, and TRAMP genotype with significant post-hoc group differences; n = 5, TRAMP genotype × Bco2 genotype × diet group. (D) Significant 2-way interactions of TRAMP genotype and diet, n = 10/TRAMP genotype × diet group. (E) Bco2 genotype and diet, n = 10, Bco2 genotype × diet group, and (F) Bco2 genotype and TRAMP genotype with a significant post hoc group difference, n = 15/Bco2 genotype × TRAMP genotype group. *Significant post hoc group differences are indicated by blunt-ended lines. Blunt-ended lines spanning multiple bars indicate that the variable corresponding to that level of data was significantly different. Bco2, β-carotene-9′,10′-oxygenase; Cela3b, chymotrypsin-like elastase family, member 3B; Hmgcr, 3-hydroxy-3-methylglutaryl-Coenzyme A reductase; Lpa, lipoprotein(A); Lipa, lysosomal acid lipase A; Npc1, Niemann-Pick type C1;Sorl1, sortilin-related receptor, LDLR class A repeats-containing; TRAMP, transgenic adenocarcinoma of the mouse prostate.

Discussion

Several studies have illustrated that dietary lycopene or tomato components inhibit prostate carcinogeneisis in rodent models [e.g., (4, 5, 24, 25, 26)]. Interestingly, we previously observed that the protective impact of lycopene feeding in TRAMP carcinogenesis is lessened when the Bco2 gene is ablated (6). Thus, this finding supports a hypothesis that BCO2 cleavage of lycopene produces metabolites that may interact with specific nuclear receptors, such as RAR and RXR (27, 28), either as partial agonists or antagonists to impact gene expression relevant to the prostate carcinogenesis cascade. We previously found that tomato and lycopene feeding can alter prostatic gene expression at as early as 10 wk of age in TRAMP/+ C57Bl/6 mice (19), coinciding with the very early morphological changes of hyperplasia and carcinoma in situ (29). Thus, in this study we add the Bco2 variable and focus our attention on 2 distinct, targeted gene expression arrays, prostate carcinogenesis and lipid metabolism, providing an informative context to study the impact of TRAMP genotype, diet, and Bco2 status.

To contextualize the impacts of diet, we first characterized gene expression changes observed during early TRAMP carcinogenesis. The TRAMP genotype drives prostate carcinogenesis and thus it is not surprising to see a major impact of this genotype on gene expression, even at 8 wk of age (Figure 1), ∼4 wk past sexual maturity, when rising testosterone drives androgen receptor signaling and activation of the transgene (29). The dominant histopathological phenotype at this time point of 8 wk of age is hyperplasia and foci of prostatic intraepithelial neoplasia (29), and thus, we found that 62% of the assayed genes in the highly targeted prostate cancer–focused array show clear changes. Pathway analysis showed that TRAMP/+ expression patterns are consistent with downregulation of PTEN signaling, upregulation of p53 and estrogen receptor signaling and changes in prostate cancer signaling canonical pathways, and increases in biological functions of survival and viability with decreases in mortality, death, and connective tissue proliferation (Figure 1).

The impact of the TRAMP genotype that we observed on cholesterol and lipoprotein metabolism (30) stood in remarkable contrast to our results with the carcinogenesis-focused array. The TRAMP genotype did impact the expression of a significant proportion of assayed cholesterol and lipoprotein metabolism–related genes [39 of 90 genes (43%)], and in contrast to the carcinogenesis panel, the gene expression and pathway analysis revealed a strong downregulation of genes involved in cholesterol and lipoprotein metabolism. For example, the cholesterol biosynthesis superpathway, LXR/RXR activation, and PPARα/RXRα activation were all downregulated (Figure 4). LXR/RXR is involved in regulating lipid metabolism, cholesterol export and conversion to bile acids, and inflammatory pathways, while PPARα/RXRα activation plays a role in fatty acid uptake and oxidation, as well as oxidative stress and inflammation. When analyzed in a biological function framework, the TRAMP/+ genotype is associated with a predicted reduction in prostatic cholesterol transport and flux and reduced lipid uptake functions (Figure 4). This finding, coupled with diminished LXR/RXR activation, may indicate increased intracellular cholesterol levels in TRAMP/+ prostate tissue, as would be conducive to growth, inducing feedback downregulation of the cholesterol synthetic pathway (31). Regarding individual gene expression changes, the most significantly modulated (P = 9.6 × 10^–7) expression in TRAMP/+ compared with TRAMP/– mice was Stard3, which was modestly, but significantly upregulated in TRAMP (expression ratio of TRAMP/+:TRAMP/–, 1.04:1.0) (Supplemental Table 4). Stard3 is involved in cholesterol transport from the endoplasmic reticulum to the endosome (32), and interestingly STARD3 binds the carotenoid lutein in the macula of the eye (33). While it is not known whether STARD3 binds other carotenoids, this finding raises the hypothesis that increased Stard3 expression could contribute to the phenomenon of carotenoid accumulation in cancerous prostate tissue (34), which warrants further investigation. Although we chose this array based upon a hypothesis that lycopene or tomato products may impact cholesterol and lipoprotein metabolism (12, 15, 16, 17, 19), the multitude of changes due to the TRAMP genotype in this array are enlightening and align with some previous findings in prostate cancer (30, 35).

To evaluate the effect of the experimental diets on the internal exposure to carotenoids, we measured serum carotenoid concentrations (Table 1). Serum lycopene concentrations were similar between lycopene- and tomato-fed mice, with the Bco2^–/– genotype being associated with greater serum lycopene concentrations, as previously reported (6, 11). The serum lycopene concentrations are slightly lower than those previously observed in 18-wk-old mice fed the same diets containing lycopene and tomato powder (6), likely due to the shorter feeding duration of 5 compared with 15 wk. Similar to our previous findings (6), phytoene, phytofluene, and ζ-carotene, biosynthetic precursors of lycopene within tomatoes, were only detected in tomato-fed mice and were less abundant than lycopene.

As expected, the demonstrated effects of tomato or lycopene on prostate cancer–related gene expression were more modest than those of the TRAMP genotype, impacting a subgroup (16%) of the genes driven by the TRAMP genotype (Figure 1). The prostate cancer–related genes for which expression is increased by tomato and lycopene feeding are associated with cellular lipid metabolism (Fasn, Acaca, Srebf1, & Hmgcr, Ptgs1), cell cycle regulation (Cdkn1a), and prostate cancer (Arntl), and a gene for which expression is typically associated with prostate cancer (Nkx3–1) showed decreased expression with lycopene feeding. Of the 8 genes impacted by diet, Fasn expression was most robustly increased with both tomato and lycopene feeding. Fasn was also impacted by a main effect of the TRAMP genotype, and its expression was associated with canonical pathways for PParα/RXRα activation, fatty acid biosynthesis initiation II, LXR/RXR activation, and AMPK signaling. Some of the gene expression changes associated with diet were in the same direction as those induced by the TRAMP/+ genotype. At this early stage of carcinogenesis, some of the changes due to TRAMP genotype may be a response of the hyperplastic host cells to compensate for detrimental impacts of the mutant cascades, yet such changes may be lost in later stages of carcinogenesis. A longer-term goal is to better understand gene expression in the early stages of the cancer cascade and how diet impacts such patterns. Empirically, however, we must keep in mind that tomato and lycopene feeding are associated with reduced cancer burden in TRAMP/+ mice (4, 5, 6) and therefore the precise biological implications of the gene expression changes warrant further investigation.

Compared with the TRAMP genotype, diet had less of an effect on cholesterol and lipoprotein metabolism-related gene expression (Table 2). Lycopene-feeding was associated with increased expression of Apoa5 and Srebf1 (Table 2), which are involved in cholesterol and lipid transport and catabolism (Supplemental Table 5). While marginally significant, the effect of lycopene on Apoa5 is opposite to the effect of the TRAMP/+ genotype (Supplemental Table 4), which may warrant future investigation. It is important to add that carotenoids are lipid-soluble compounds, sharing transport and metabolic pathways with cholesterol and other lipids; thus, the metabolic intersections of carotenoids, lipids, and cholesterol are intriguing. The hypothesis stands that lycopene and tomato may impact prostate carcinogenesis, in part by affecting intracellular lipid metabolism.

We observed that the Bco2 genotype very minimally impacted the prostate independently as well as in the context of tomato and lycopene feeding. Serum and tissue lycopene concentrations (Table 1) were similar between tomato- and lycopene-fed mice, but were greater in Bco2–/– mice, consistent with our previous findings (6, 11, 12, 19). Lycopene, but not tomato, led to lower expression of Apc, Nfkb, Mto1, and Rbm39 in Bco2^–/– mice than in the Bco2^+/+ mice. While the biological significance of a subtle change in any one gene is unclear, these Bco2-dependent alterations by lycopene do parallel the pattern we observe for Bco2-dependent effects of lycopene on TRAMP tumor incidence (6). In other words, the TRAMP genotype is the strongest effect, with diet shifting cancer risk, and the BCO2 genotype having small modulatory impact on the diet effect. Indeed, it may be that such subtle gene expression changes mediate the relationship between long-term dietary lycopene and prostate cancer prevention observed in this model and in human epidemiology. We hypothesize that the effect of tomato feeding on gene expression is less sensitive to Bco2 genotype because the array of bioactive components in tomato offers overlapping activity, such that the lycopene × Bco2 interaction is overcome or that lycopene cleavage products act by genes not evaluated in our targeted arrays.

The lycopene × Bco2 effect on Nfkb1 is intriguing, as lycopene's potential inhibitor effect on oxidative stress and inflammatory signaling, partially mediated by NFkB, may be one process impacting carcinogenesis (14). Nfkb1 is found in a number of the canonical pathways that differed between TRAMP genotypes, including prostate cancer signaling, PTEN signaling, and PI3K/AKT signaling (Supplemental Table 3). The increase in Nfkb1 expression in Bco2^+/+ mice was unexpected, but could suggest that 1) lycopene promoted a protective inflammatory response in Bco2^+/+ mice, enhancing repair processes or 2) accumulation of lycopene in Bco2^–/– mice (Table 1) may have led to more nonenzymatically generated lycopene metabolites that may serve as agonists or antagonists to nuclear receptors, such as LXR/RXR to regulate Nfkb1 (36). While little is known regarding lycopene metabolite activation of retinoid X receptors, other lycopene and nonlycopene carotenoid metabolites have been shown to antagonize retinoic acid receptors and retinoid X receptors, respectively (27, 28, 37).

Consistent with the prostate cancer–focused array, we observed an effect of diet (tomato and lycopene compared with control) on Hmgcr with the cholesterol and lipoprotein metabolism–focused array; however, on the latter platform, we detected that the diet effect differed by TRAMP genotype (Figure 4). Nonetheless, the results indicate an effect of tomato and lycopene feeding on Hmgcr expression. Hmgcr is critical to the cholesterol biosynthesis and the LXR/RXR activation canonical pathways, which are predicted to differ between TRAMP genotypes. Thus, Hmgcr may be a mediator by which lycopene and tomato bioactives affect prostatic cholesterol balance. Compared with tomato feeding in Bco2^–/– mice, lycopene feeding was associated with greater expression of Npc1 (Figure 5), which is involved in intracellular lipoprotein trafficking; however, this effect was not present in Bco2^+/+ mice.

There are several strengths and limitations to be considered. Strengths include a factorial study design that allowed for simultaneous observation of gene expression changes in early carcinogenesis and the evaluation of a diet × Bco2 genotype interaction effect on those genes. Such a design, although adding complexity to analysis, is reflective of interactions that are relevant to the human condition, which is even more complex. Furthermore, comparing whole tomato powder and lycopene alone allows the interpretation of the relative importance of these dietary components in impacting early carcinogenesis. While there were some 3-way interactions between the variables, due to the inherent complexity of these interactions and small group sizes, effect sizes, and significance levels, we viewed these findings as suggestive and warranting further investigation. Another limitation is that we examined gene expression in the whole tissue and not in isolated cell types, as gene expression may differ in matrix and epithelial cells, and within various subpopulations of epithelial cells destined to progress more rapidly into prostatic intraepithelial neoplasia or adenocarcinoma. Also, by narrowing the genes assayed to those contained in the prostate and lipid panels, the interpretation of the results is biased toward data obtained from these gene array panels. Future studies may benefit from unbiased and cell-specific approaches.

In conclusion, we report new findings providing insight into the dietary prevention of prostate cancer in the TRAMP model. First, we observed at 8 wk of age, early in the cancer cascade, a very strong impact of the TRAMP genotype on prostatic gene expression, and that a subset of these genes associated with pathways related to cholesterol, lipid metabolism, and nuclear receptors associated with lipids were sensitive to tomato and lycopene feeding. Nearly all diet–Bco2 interactions revealed that the effects of lycopene differed by Bco2 genotype, while the effects of tomato were less sensitive to Bco2. This suggests a mechanistic basis by which Bco2-generated lycopene metabolites may exert some early cancer protective effects via changes in gene expression, to modulate later cancer incidence as we previously reported (6). Future studies should investigate the effects of tomato and lycopene feeding on cholesterol and lipid transport and catabolism, as potential mechanisms by which carcinogenesis is impacted.