Introduction

Beta-carotene (BC) is a lipid-soluble provitamin, and a major dietary source of vitamin A. BC itself is thought to behave as a health-promoting agent. This has been attributed to its ability to efficiently scavenge radicals and thereby prevent macromolecular damage [1]. Furthermore, epidemiological studies focusing on dietary intakes of BC have indicated that a high BC intake is associated with a reduced risk for the development of cardiovascular diseases and several types of cancer [2, 3]. Moreover, BC is a provitamin A, and therefore its consumption prevents vitamin A deficiency, which is of importance even in Western society. For example, pregnant and lactating women are at risk for vitamin A deficiency in the Western society due to the discouragement of the consumption of food containing high amounts of vitamin A, in combination with their increased requirements [4].

Although BC is mainly considered to be a health-promoting agent, large-scale intervention trials failed to confirm its beneficial properties. Two intervention trials even showed an increased lung cancer risk and increased mortality because of lung cancer in smokers after BC supplementation (CARET study and ATBC study) [5, 6]. A 6-year follow-up of the CARET study showed differences in lung cancer risk in response to BC supplementation between male and female volunteers, while the ATBC study was only executed in male volunteers. In the CARET study, male subjects had an increased lung cancer risk upon BC supplementation, but only for the duration of the BC intervention. On the other hand, BC-supplemented females had an increased lung cancer risk that persisted even after the intervention had stopped, and it took several years before the lung cancer risk in females reached the levels observed in the controls [7]. The BC-induced increase in lung cancer risk is thought to occur only in smokers or asbestos-exposed subjects since one large-scale intervention trial with mainly non-smokers showed no effect of BC on lung cancer [8, 9]. Although the effect of BC on lung carcinogenesis has been studied extensively, the mechanism by which BC increases lung cancer risk is still not precisely known. The main reason for this is that research on molecular pathways influenced by BC is hampered by the lack of an appropriate animal model that also allows for the use of state-of-the-art functional genomic tools, such as commercially available microarrays. Mice and rats are animal models for which these tools are commercially available, but they differ greatly in BC metabolizing activity compared to humans. The main cause for this is the presence of a more active beta-carotene 15, 15′-monooxygenase 1 (Bcmo1) enzyme variant in rodents [10]. Bcmo1 is the key enzyme in BC metabolism, and in rodents, Bcmo1 cleaves almost all absorbed BC. Since rodents metabolize BC to a high degree, BC supplementation results in an increase in BC metabolites rather than the accumulation of BC, and rodents are therefore inadequate animals to study the effects of BC in humans [10].

Not only inter-species differences in Bcmo1 activity have been described, but also large inter-individual differences in BCMO1 activity are present between humans. Around 25–45% of the volunteers participating in BC conversion studies have been classified as poor converters [11–13]. These inter-individual differences in BC metabolism are at least partly related to polymorphisms in the BCMO1 gene, which results in a reduced BCMO1 activity [14, 15]. Of the reported polymorphisms, there are two frequently occurring polymorphisms with variant allele frequencies of 42 and 24%. As a result, individuals with these polymorphisms have relatively low vitamin A and high BC plasma concentrations compared to their wild-type counterparts [14]. Since differences in activity of BCMO1 result in different BC metabolite concentrations, the activity of BCMO1 might therefore have an effect on previously reported BC-induced effects.

The aim of our study was to investigate gene expression changes induced by dietary BC to try to explain the previously found harmful effects of BC in the lung. For this purpose we used a recently described mouse model, which has no functional Bcmo1 enzyme and which therefore displays increased BC plasma concentrations upon BC supplementation [16]. We performed whole-genome microarray gene expression analysis on lung tissue of both female and male wild-type (Bcmo1 +/+) and Bcmol knockout (Bcmo1 −/−) mice with or without 14 weeks of 150 mg/kg BC supplementation. We previously analyzed the effects of BC in female Bcmo1 −/− mice, and the major effect observed was that BC supplementation counteracted inflammation associated gene expression effects, which were most likely caused by an increased vitamin A demand in Bcmo1 −/− mice [17]. This effect was not observed in male Bcmo1 −/− mice, where two genes, Frizzled 6 (Fzd6) and Collagen triple helix receptor containing 1 (Cthrc1) were strongly downregulated by BC supplementation, which was not observed in female mice [18]. Thus, the major BC-affected processes seemed different between both genders. However, both females and males displayed increased cancer risk after BC supplementation during the intervention of the CARET study. Therefore, we aimed to specifically focus on those genes that are commonly regulated by BC supplementation in both genders, i.e., in male as well as female Bcmo1 −/− mice.

Materials and methods

Animals and treatment

The experimental setup has been described previously (females, [17] and males [18]). Twelve female (fe) and 12 male (ma) B6129SF1 (Bcmo1 +/+) and 12 female and 12 male B6;129S-Bcmo1 tm1dnp (Bcmo1 −/−) mice [16] were used for the dietary intervention, which was conducted in accordance with the German animal protection laws by the guidelines of the local veterinary authorities. During the breeding and weaning periods of the mice, mothers were maintained on KLIBA 3430 chow containing 14,000 IU vitamin A/kg diet (Provima Kliba AG, Kaiseraugst, Switzerland). Five-week-old female and male Bcmo1 +/+ and Bcmo1 −/− mice were caged in groups containing two to four siblings per group and were maintained under environmentally controlled conditions (temperature 24°C, 12 h/12 h light/dark cycle). Mice had ad libitum access to feed and water. Basic feed consisted of the pelletized diet D12450B (Research Diets Inc, USA) with a fat content of 10%. The diet was modified to contain 1,500 IU vitamin A/kg of diet, which is a vitamin A-sufficient diet, and the control diet (control) was supplemented with water-soluble vehicle beadlets (DSM Nutritional Products Ltd., Basel, Switzerland) containing dl-alpha-tocopherol and ascorbyl palmitate as stabilizers, as well as carriers such as gelatin, corn oil sucrose and starch. The BC diet (BC) was supplemented with identical water-soluble beadlets containing BC (DSM Nutritional Products Ltd., Basel, Switzerland) to generate 150 mg BC/kg diet. Beadlets were added by the manufacturer before low temperature pelletting. Feed pellets were color marked and stored at 4°C in the dark.

After 14 weeks of dietary intervention, six female and six male Bcmo1 +/+ mice on the control diet (Bcmo1 +/+ Co), six female and six male Bcmo1 +/+ mice on the BC diet (Bcmo1 +/+ BC), three female and six male Bcmo1 −/− mice on the control diet (Bcmo1 −/− Co), and three female and six male Bcmo1 −/− mice on the BC diet (Bcmo1 −/− BC) were randomly killed on three subsequent mornings. Due to an insufficient number of female Bcmo1 −/− mice in the original breeding pool, six additional female Bcmo1 −/− mice were used that were born 2 weeks later from an identical experiment, three were fed the control diet and three the BC diet, to generate n = 6 for each group. These mice were killed 2 weeks after the first group of mice. Blood was collected from the vena cava after isoflurane and ketamin anesthesia. Blood was coagulated for at least 20 min at room temperature, cooled to 4°C and centrifuged. Lung tissue was removed, rinsed in phosphate-buffered saline (PBS) and snap frozen in liquid nitrogen. The lung tissues were stored at −80°C.

RNA isolation

Left lung lobes were homogenized in liquid nitrogen using a cooled mortar and pestle. Total RNA was isolated using TRIzol reagent (Invitrogen, Breda, The Netherlands) followed by purification using RNeasy columns (Qiagen, Venlo, The Netherlands) using the instructions of the manufacturer. RNA concentration and purity were measured using the Nanodrop system (IsoGen Life Science, Maarsen, The Netherlands). Approximately 30 μg of total RNA was isolated with A260/A280 ratios above 2 and A260/A230 ratios above 1.9 for all samples, indicating good RNA purity. RNA degradation was checked on the Experion (Bio-Rad, Veenendaal, The Netherlands) using Experion StdSense chips (Bio-Rad). Two samples did not meet RNA quality (female Bcmo1 +/+ mice, one on BC diet and one on control diet) and were omitted from the experiment.

Microarray hybridization procedure

The 4 × 44 k Agilent whole mouse genome microarrays (G4122F, Agilent Technologies, Inc. Santa Clara, CA) were used. Preparation of the sample and the microarray hybridization were carried out according to the manufacturer’s protocol with a few exceptions as described previously [19, 20]. In brief, cDNA was synthesized from 1 μg lung RNA using the Agilent Low RNA Input Fluorescent Linear Amplification Kit for each animal without addition of spikes. Thereafter, samples were split in two equal amounts to synthesize Cyanine 3-CTP (Cy3) and Cyanine 5-CTP (Cy5) labeled cRNA using half the amounts per dye as indicated by the manufacturer (Agilent Technologies). Labeled cRNA was purified using RNeasy columns (Qiagen). Yield, A260/A280 ratio, and Cy3 or Cy5 incorporation were examined for every sample using the nanodrop. All samples met the criteria of a cRNA yield higher than 825 ng and a specific activity of at least 8.0 pmol Cy3 or Cy5. Then 1,200 ng of every Cy3-labeled cRNA sample was pooled and used as a common reference pool. Individual 825 ng Cy5-labeled cRNA and 825 ng pooled Cy3-labeled cRNA were fragmented in 1× fragmentation and 1× blocking agent (Agilent Technologies) at 60°C for 30 min and thereafter mixed with GEx Hybridization Buffer HI-RPM (Agilent Technologies) and hybridized in a 1:1 ratio at 65°C for 17 h in an Agilent Microarray hybridization Chamber rotating at 4 rpm. After hybridization, slides were washed according to the wash protocol with Stabilization and Drying solution (Agilent Technologies). Arrays were scanned with an Agilent scanner with 10 and 100% laser power intensities (Agilent Technologies).

Data analyses of microarray results

Signal intensities for each spot were quantified using Feature Extraction 9.1 (Agilent Technologies). Median density values and background values of each spot were extracted for both the experimental samples (Cy5) and the reference samples (Cy3). Quality control for every microarray was performed visually, by using quality control graphs from Feature Extraction, M-A plots and boxplots that were made using limmaGUI in R (Bioconductor) [21]. Data were imported into GeneMaths XT 2.0 (Applied Maths, Sint-Martens-Latem, Belgium).

The median Cy5 signal intensity per spot per array was used for the individual gene expression, while the median Cy3 signal intensity was used for normalization. Since values close to the background may give aberrant fold changes between groups, spots with an average Cy5 or Cy3 median density value below two times the average of the background density were discarded. Spots with an average of the median Cy5 and Cy3 densities twice above the average of the background density were selected and log transformed. Thereafter, the log transformed Cy5 median density value was normalized against the fold change in the log-transformed Cy3 median density value compared to the average log-transformed Cy3 median density value for each spot as described before [22]. Pathway analysis was performed using Gene-Ontology (GO) overrepresentation analysis (ErmineJ) [23], and classification of genes was based on OMIM (www.ncbi.nlm.nih.gov), genecards (www.genecards.org) and literature mining.

Analysis of mRNA expression by real-time Q-PCR

Differential gene expression of lecithin-retinol acyltransferase (phosphatidylcholine-retinol-O-acyltransferase) (Lrat), glutamate receptor, ionotropic, AMPA1 (alpha 1) (Gria1) and endothelin 1 (Edn1) was analyzed using quantitative PCR (Q-PCR) to validate microarray results. Then 1 μg RNA isolated from lung of individual animals was converted into cDNA using the iScript cDNA Synthesis Kit (Bio-Rad). One sample was taken along without reverse transcriptase to examine the presence of DNA (without RT reaction). An equal amount of cDNA of all individual samples was pooled and diluted (10×, 31.6× 100×, 316× 1,000×, 3,160×, 10,000×) for the generation of a calibration curve to test the efficiency of the Q-PCR reaction. Each PCR reaction (25 μl) contained 1× iQ SYBR green supermix (Bio-Rad), 10 μM sense primer, 10 μM antisense primer and 2 μl 100 times diluted cDNA. Primers for Lrat were sense 5′-GCTCTCGGATCAGTCCACAGGC-3′ and antisense 5′-TCCCAAGACAGCCGAAGCAAGA-3′, for Gria1 sense 5′-TGGTGGTGGTGGACTGTGAATC-3′ and antisense 5′-CAGGTTGGCGAGGATGTAGTGG-3′ and for Edn1 sense 5′-AGACCAGGCAGTTAGATGTCAGTG and antisense 5′-TCGTAGTTTCCTTCCTTCCACCAG. For the amplification of the reference genes we used syntaxin 5a (Stx5a) with sense 5′-TTAAAGAACAGGAGGAAACGATTCAGAG-3′ and antisense 5′-CAGGCAAGGAAGACCACAAAGATG-3′, and ring finger protein 130 (Rnf130) with sense 5′-ACAGGAACCAGCGTCGTCTTG-3′ and antisense 5′-ACCCGAACAACATCATTCTGCTTATAG-3′, which showed equal gene expression levels for all individual animals on the microarray. These intron-spanning primers were designed using Beacon designer 7.00 (Premier Biosoft International, Palo Alto, CA) or using primer-BLAST (http://www.ncbi.nlm.nih.gov/). Amplification was performed in duplicate using the MyIQ single-color real-time PCR detection system (Bio-Rad) using the following temperature cycles: 1× 3 min at 95°C, 40× 2-step amplification (15 s 95°C, 45 s 60°C), 1× 1 min 95°C, and 1× 1 min 65°C followed by melting curve analysis (60× 10 s 65°C with an increase of 0.5°C/10 s. A negative control without cDNA template and the minus reverse transcriptase sample were taken along with every assay. Using the standard curves for every gene, the relative level of expression of all genes was calculated. Normalized Gene Expression (ΔΔCt) analysis was calculated using the IQ5 software version 2.0 (Bio-Rad).

Testosterone measurement

Testosterone concentrations were measured using radio immunoassays (DSL 4100, Diagnostic Systems Laboratories, Beckman Coulter, Assendelft, The Netherlands) and were performed according to the manufacturer’s protocol using quarters of all volumes as has been described previously [24]. Gamma-radiation was measured on a Wallac 1470 Wizard Gamma Counter (Perkin-Elmer). The recovery of testosterone spikes and linearity of the dilution were tested before use.

Statistical analysis

General effects of diet, gender and genotype on concentrations of BC, retinol and retinyl esters in lung and serum were analyzed using 2 × 2 × 2 factorial univariate ANOVA (SPSS version 15.0) and considered statistically significant when p < 0.05. Effects of BC compared to control diet were tested using a Student’s t-test and considered statistically significant at p < 0.05). Fold changes for both microarray gene expression and Q-PCR gene expression were calculated using mean log signal intensities. P-values for differential expressions were calculated between two groups using two-tailed Student’s t-test statistics on log intensity values. Changes were considered statistically different at p < 0.05. Regression analysis of gene expression in males versus females was performed in GraphPad Prism software (version 5.0; GraphPad Software, Inc, San Diego, CA) using nonlinear regression analysis.

Results

BC supplementation and knockout of Bcmo1 results in changes in BC and BC metabolites

BC concentrations in lung and serum were increased after 14 weeks of BC treatment compared to control diet-fed mice, and as expected, to a higher level in Bcmo1 −/− mice as in Bcmo1 +/+ mice (as has been described separately for female [17] and male [18] mice). There were no significant differences in retinol serum concentrations between the different groups. In lung tissue, retinol and retinyl ester concentrations were increased upon BC supplementation compared to the control diet fed mice, but there was no interaction between BC and the Bcmo1 genotype concerning the concentration of these metabolites (Fig. 1).

Fig. 1
figure 1

Beta-carotene (BC) (a, c), retinol (b, d) concentrations in serum and lung tissue, respectively, and retinyl ester concentrations in lung tissue (e) of female and male Bcmo1 +/+ mice and Bcmo1 −/− mice fed a control diet (white bars) or a BC diet (black bars). Data are expressed as mean ± SEM; significance was tested for diet (BC), gender (S) and genotype (G) using ANOVA and considered significant at p < 0.05. Significant effects for the different factors or any interaction between those factors are displayed on top of each figure. Student’s t-test was used between BC and control groups when there was a significant diet effect by ANOVA and considered significant when p < 0.05. Using Student’s t-test statistics: *p < 0.05, **p < 0.01 and ***p < 0.001

Full size image

Gene expression of all groups was analyzed, and of the 43,379 spots (number of spots minus control spots) on the whole genome microarrays, 31,128 spots had an average signal twice above background and were regarded as positive spots and included in the analysis. There were two genes involved in downstream BC metabolism that were significantly regulated by BC in Bcmo1 +/+ mice (Table 1). The biggest fold change was found for lecithin-retinol acyltransferase (phosphatidylcholine-retinol-O-acyltransferase (Lrat; 1.6 fold increase after BC supplementation). Five genes involved in downstream BC metabolism were regulated because of the knock-out of Bcmo1. Most striking was the high fold change up regulation of Lrat (1.7–2.1 fold increase in Bcmo1 −/− mice) and the high fold change down regulation of alcohol dehydrogenase 7 (class IV), mu or sigma polypeptide (Adh7; 1.6–2.7 fold decrease in Bcmo1 −/− mice).

Table 1 Genes involved in BC metabolism or retinoic acid catabolism that were significantly regulated after BC supplementation or Bcmo1 knockout
Full size table

Gene expression is changed in the opposite direction by BC in male and female Bcmo1 −/− mice

BC supplementation resulted in a considerably higher number of significantly regulated spots in the Bcmo1 −/− mice than in the Bcmo1 +/+ mice (Fig. 2a, b). Only 20 regulated genes overlapped in male and female Bcmo1 +/+ mice and 89 genes in Bcmo1 −/− mice. These two groups of regulated genes commonly affected in both genders showed no overlap between Bcmo1 −/− and Bcmo1 +/+ mice. To further determine whether these genes were similarly regulated by BC in both genders, fold changes of gender-independent BC-regulated genes were plotted for female mice against the fold changes in male mice. The 20 BC-regulated genes in male and female Bcmo1 +/+ mice were distributed in all four quadrants of the plot (Fig. 2c). Regression analysis resulted in a significant correlation [R = 0.66; slope = 0.787; 95% CI slope = (0.2988–1.276)]. This correlation was however based on only two data points (−1.93; −1.87) and (1.34; 1.56). Without these two points, there was no significant correlation [R = 0.14; slope = −0.193; 95% CI slope = (−0.8832; 0.4960)]. To our surprise, genes regulated by BC in female Bcmo1 −/− were regulated in opposite directions compared to BC-regulated genes in male Bcmo1 −/− mice (upwards regulated genes in males were down in females and vice versa), with only 4 out of the 89 probes being regulated in the same direction (in italics in Table 3). We found >95% of the probes being regulated in the opposite direction by BC between males and females, suggesting that this is a real effect of BC. Randomly affected gene expression would theoretically result in 50% of the probes being regulated in a similar direction and 50% in opposite directions. Regression analysis resulted in a significant negative regression [R = 0.76; slope = −0.436; 95% CI slope = (−0.5144; −0.3582)]. Genes regulated by BC in both male and female Bcmo1 +/+ mice (Table 2) or Bcmo1 −/− mice (Table 3) were categorized into biological processes, with endothelial functioning, immune response and nervous response being most prominently regulated by BC in Bcmo1 −/− mice (Fig. 3).

Fig. 2
figure 2

Venn diagrams representing the number of genes significantly (p < 0.05) regulated by BC in female and male mice and the number of genes regulated by BC in both females as well as males (overlap) in Bcmo1 +/+ (a) and Bcmo1 −/− (b) mice. The scatter plots represent the gene expression fold changes of the genes regulated by BC independent of gender in females (x-axis) and males (y-axis) in Bcmo1 +/+ (c) and in Bcmo1 −/− (d) mice

Full size image
Table 2 Classification of genes that were significantly regulated by BC in both female and male Bcmo1 +/+ mice
Full size table
Table 3 Classification of genes that were significantly regulated by BC in both female and male Bcmo1 −/− mice
Full size table
Fig. 3
figure 3

Classification of genes that are regulated by BC (p < 0.05) in male as well as in female Bcmo1 −/− mice into different biological process categories

Full size image

We used Q-PCR to validate the microarray experiment. For validation of the effects in Bcmo1 +/+ mice, lecitin retinol acyltransferase (phosphatidylcholine-retinol O-acyltransferase) (Lrat), an enzyme involved in downstream BC metabolism, was selected. For validation of effects induced by BC supplementation in Bcmo1 −/− mice, the genes glutamate receptor, ionotropic, AMPA1 (alpha 1) (Gria1) and endothelin 1 (Edn1) were chosen because of their relatively high expression level and relatively high fold changes. All three selected genes showed an identical expression using the microarray technique and Q-PCR (Fig. 4).

Fig. 4
figure 4

The fold change gene expression of Lrat in Bcmo1 +/+ mice (a), Gria in Bcmo1 −/− mice (b) and Edn1 in Bcmo1 −/− mice (c) in BC supplementated mice relative to the control diet fed mice in the microarray, validated by Q-PCR using the reference genes Stx5a and Rnf130. Data represent the average gene expression ± SEM compared to control diet fed mice. Using Student’s t-test statistics on log transformed data: *p < 0.05 and **p < 0.01

Full size image

BC supplementation results in the overrepresentation of the steroid biosynthetic process in male Bcmo1 −/− mice

Gene-ontology processes affected by BC supplementation in male and female Bcmo1 −/− mice were examined to further explain the opposite direction of gene expression induced by BC. This overrepresentation analysis was performed using ErmineJ overrepresentation analysis [23] on all genes and their corresponding p-value. We found a significant overrepresentation of the steroid biosynthetic process (GO:0006694) in male Bcmo1 −/− mice by BC supplementation (p = 4.6 × 10−6), but not in female Bcmo1 −/− mice (p = 0.17). Genes regulated by BC resulting in the overrepresentation of the steroid biosynthetic process in male Bcmo1 −/− mice are listed in Table 4 (genes presented until p = 0.1).

Table 4 Genes regulated by BC in male Bcmo1 −/− mice resulting in a significant overrepresentation of the steroid biosynthetic process (GO: 0006694)a
Full size table

Effect of BC on serum testosterone levels

We determined serum concentrations of the sex hormone testosterone. Since only limited amounts of serum were available, other sex hormones requiring bigger serum amounts could not be determined. As expected, testosterone levels were significantly higher in Bcmo1 +/+ males than in Bcmo1 +/+ females (Fig. 5a). In Bcmo1 −/− mice, however, testosterone levels showed a high inter-individual variability (Fig. 5b). Due to this large variation, there was no statistically significant difference in testosterone levels between male and female Bcmo1 −/− mice. However, BC increased testosterone levels in both Bcmo1 −/− female and male mice with a fold change of 4.7 and 4.0, respectively (Fig. 5c), although, because of the high variation, differences did not reach statistical significance (ANOVA; effect of diet p = 0.12 and interaction effect of diet with Bcmo1 genotype p = 0.14).

Fig. 5
figure 5

Testosterone levels in serum of Bcmo1 +/+ mice (a), Bcmo1 −/− mice (b) and the corresponding fold change in testosterone (c) induced by BC supplementation (black bars) as compared to control mice (white bars). Data are expressed as mean ± SEM; significance was tested for diet (BC), gender (S) and genotype (G) using ANOVA and considered significant at p < 0.05. Significant effects using ANOVA are displayed on top of the figure

Full size image

Discussion

We showed for the first time that BC supplementation to Bcmo1 −/− mice resulted in the regulation of gene expression in opposite directions in male and female mice. We hypothesize that BC alters systemic hormone production, which was supported by the overrepresentation of the steroid biosynthetic process (GO:0006694) in male Bcmo1 −/− mice upon consumption of a BC diet (p = 4.6 × 10−6). Moreover, testosterone concentrations were increased, but highly variable after BC supplementation in Bcmo1 −/− mice. The presence of an active Bcmo1 enzyme was of importance in the observed opposite direction of gene expression changes induced by BC supplementation in male and female mice, since these effects were not observed in Bcmo1 +/+ mice.

BC supplementation to Bcmo1 −/− mice resulted in an opposite direction of gene regulation in males and females. A hypothesis to explain this is that BC supplementation induces alterations in sex hormone synthesis or metabolism in male and female mice. Interestingly, the biosynthetic pathway for BC production in plants partly overlaps with the steroid biosynthetic pathway in animals [25]. To support this hypothesis, we measured testosterone levels in serum, which were highly variable in BC-supplemented Bcmo1 −/− mice, but not in the other groups. Another important sex hormone that might have been changed after BC supplementation to Bcmo1 −/− is 17β-estradiol, but the amount of serum left did not allow us to measure this. Although we were not able to technically prove a change in sex hormones, a change in sex hormones would result in a gender-specific change since the lung contains all necessary elements for sex hormone action, such as enzymes involved in androgen and estrogen metabolism as well as estrogen and androgen receptors [26–28]. In line with this, the GO process for steroid biosynthesis was significantly regulated in male lung tissue, as can be seen in Table 4. Several enzymes involved in steroid conversion were shown to be significantly regulated in male Bcmo1 −/− mice. For example, testosterone can be converted in the lung into estrone and 17β-estradiol. 17β-Estradiol can bind to the estrogen receptor to induce estrogen-related responses [29]. Interestingly, the gene expression of 17-beta-hydroxysteroid dehydrogenase 12 (Hsd17b12), which converts estrone into estradiol, was significantly upregulated [30], and steroid 5 alpha-reductase 1 (Srd5a1), an important enzyme in the inactivation of testosterone [27], was downregulated (p = 0.055) in lungs of male Bcmo1 −/− mice upon BC supplementation. Moreover, 17-beta-hydroxysteroid dehydrogenase 7 (Hsd17b7), which oxidizes or reduces estradiol, and thus decreases the estrogenic potency, was significantly downregulated by BC in female Bcmo1 −/− mice. These data indicate that BC might have had an effect on sex hormone synthesis and had an effect on hormone conversion in male Bcmo1 −/− mice. Further research is needed to elucidate the exact mechanism underlying the observed opposite change in gene regulation induced by BC supplementation in male and female Bcmo1 −/− mice.

Although the number of studies that have assessed effects of BC on hormone production is limited, these studies actually support a relation between BC and steroid hormone alterations. BC supplementation to female dogs resulted in increased plasma progesterone concentrations [31], and cats supplemented with BC had increased estradiol concentrations [32]. In addition, it has been reported that downstream BC metabolites, such as vitamin A and the bioactive metabolite retinoic acid, also have an effect on steroidogenesis and that these effects are, at least in part, due to an increased expression of enzymes involved in steroidogenesis [33, 34]. It is not known whether previously described effects of downstream BC metabolites on steroidogenesis are regulated by binding of retinoic acid to the transcription factors retinoic acid receptor (RAR) or rexinoid X receptor (RXR).

Our results suggest that the BC-induced opposite direction of gene expression in the different genders of Bcmo1 −/− mice was due to an increase in BC concentration and not due to a change in BC metabolites such as retinoic acid, since this opposite direction of gene expression regulation was only observed in Bcmo1 −/− mice. Compared to Bcmo1 −/− mice, male and female Bcmo1 +/+ mice had higher concentrations of the BC metabolites; retinol and retinyl esters accumulated in lung tissue after BC supplementation. Furthermore, although still not fully understood, it is generally believed that retinoic acid concentrations are well regulated [35]. Nevertheless, it may be that the concentration of the bioactive metabolite retinoic acid, able to bind to RAR and RXR, is altered only in Bcmo1 −/− mice and not in Bcmo1 +/+ mice. We previously reported that the Bcmo1 −/− mice were more prone to a phenotypic vitamin A deficiency than Bcmo1 +/+ mice due to alterations in the expression of downstream BC-metabolizing enzymes important in the production of retinoic acid [17]. Supplemental BC was able to restore this phenotype. We similarly show in this study that downstream BC metabolism was altered in Bcmo1 −/− mice. Most striking was the upregulation of Lrat combined with the downregulation of Adh7 in Bcmo1 −/− mice. Lrat is involved in the storage of retinol by the formation of retinyl esters [36], while Adh7 is involved in the conversion of retinol into retinal, which is the precursor for retinoic acid [37]. This therefore implicates a more vitamin A-storing phenotype of Bcmo1 −/− mice compared to Bcmo1 +/+ mice. Moreover, more enzymes involved in downstream BC metabolism were altered, and altogether these data indicate that also a change in retinoic acid, possibly because of a less tight regulation of retinoic acid stores in Bcmo1 −/− mice compared to Bcmo1 +/+ mice, might explain the observed alterations in hormone production.

Another question is whether the processes that were regulated by BC supplementation in lung tissue of Bcmo1 −/− mice are known to respond to hormonal changes. The BC-regulated genes in lungs of Bcmo1 −/− mice in this study were mainly categorized into three biological categories: endothelial functioning, immune response and nervous response. A literature search revealed that these three categories are all affected by one gaseous mediator: nitric oxide (NO). NO is a radical and can be produced endogenously by the iron-containing enzyme nitric oxide synthase (Nos). Three distinct NOS isoforms have been characterized, neuronal NOS (nNOS or NOSI), inducible NOS (iNOS or NOSII) and endothelial NOS (eNOS or NOSIII), and all three NOS isoforms have been detected in the airway epithelium [38, 39]. The functions of NO in the airway include neurotransmission, vascular tone homeostasis and regulation of immune functioning [40]. In our microarray, only NOSIII had an expression twice above background, and there were no significant differences in gene expression after BC supplementation. It is known that estrogen and testosterone can have opposite effects on NO production resulting from NOS activity [41]. An effect of BC on sex hormone production may therefore result in a change in expression of genes that are mediated by NO in opposite directions in male and female lungs. Thus, when BC-induced changes in hormone production, this possibly results in differences in NO production, which might explain that particularly genes involved in immune response, nervous response and endothelial functioning were changed after BC supplementation to male and female Bcmo1 −/− mice.

A last point that we would like to address is whether changes in hormone production or metabolism induced by BC could possibly explain an increase in lung cancer risk as observed in the CARET and ATBC studies. If this is the case, than hormones are a determining factor in lung carcinogenesis, resulting in gender-related differences in lung cancer risk. Indeed, females are thought to have a higher risk for the development of smoke-induced lung cancer than males [42, 43], and lung cancer is more frequent in never-smoking females than in never-smoking males [44]. It has been hypothesized that especially estrogens might be an important contributor to the differences in lung cancer risk between the different genders. For example, estrogen replacement therapy increases lung cancer risk, and early menopause, which results in decreased estrogen levels, which is associated with a reduced risk for lung cancer in women [45]. Therefore, BC-induced changes in hormone production could be an explanation for part of the increases in lung cancer risk in the CARET and ATBC studies.

In summary, we used Bcmo1 −/− mice to investigate BC-induced alterations in gene expression in lung. We unexpectedly found that BC supplementation resulted in an opposite direction of gene expression in male and female Bcmo1 −/− mice. We hypothesize that BC supplementation to Bcmo1 −/− mice resulted in changes in hormone production or metabolism, which would explain our opposite effects on lung gene expression. To fully understand and explain these effects, further research is needed specifically focusing on reproductive organs as potential target organs for BC and on quantitative measurements of sex hormones in response to BC supplementation in males and in females.