Introduction

Breast cancer continues to be the most commonly diagnosed cancer and the second most common cause of cancer deaths among US women [1, 2]. Although mutations in several genes, including p53 and BRCA1 [36], have been associated with the development of breast cancer, there remain many unanswered questions regarding the etiology and temporal development of this disease [1]. Evidence supports the role of other genetic and epigenetic influences on breast carcinogenesis, and diet appears to be one of the important factors modulating this effect [2, 7]. Considerable interest has been focused on soy and soy phytochemicals, such as the isoflavones, as candidates for prevention of breast and prostate cancer [8, 9]. Much effort has been devoted to investigating the beneficial effects of soy and soy phytochemicals; however, the precise mechanisms underlying the effects of these compounds on breast and prostate cells remain elusive.

Genistein (5, 7, 4′-trihydroxyisoflavone) is one of the most abundant isoflavones in soy and soy products [10]. Thus, it has been proposed to be a bioactive food component that could contribute to soy’s beneficial effects on cancer [9, 11]. Some have further proposed that genistein could be used as a chemopreventive agent for breast as well as prostate cancer, and several animal and human studies appear to support this idea. For example, Lamartiniere’s group [12, 13] has shown that rats exposed to dietary genistein either before or during puberty have a lower incidence of mammary tumors when challenged by the carcinogen 7,12-dimethylbenz[a]anthracene. This temporal effect appears to occur in humans as well, exemplified in Asian women who consumed higher levels of soy foods during adolescence and were shown to have a lowered incidence of breast cancer compared with those who consumed lower amounts of soy food during the same time period [14, 15]; furthermore, high soy intake during both adolescence and adulthood was associated with a lower incidence of breast cancer risk than high soy intake in adult life only [15].

Using cell culture models, we, as well as others, have reported on the effects of genistein on various pathways, including proliferation, apoptosis, cell cycle and steroid hormone-mediated pathways [1621]. The results of these studies have led some to hypothesize that anti-proliferative, pro-apoptotic effects of genistein may contribute to the cancer preventive properties of a soy diet. However, in most cases, demonstration of the anti-proliferative, pro-apoptotic effects of genistein in cell culture required concentrations of at least 10-fold higher than circulating levels normally seen from oral consumption of genistein [19, 21]. This raises the question of whether the observed anti-proliferative, pro-apoptotic effects of genistein in cell culture are reflective of what occurs in vivo. Of further concern is that, at physiologically relevant concentrations, genistein has been shown to be a weak estrogen, which could contribute to carcinogenesis in an estrogen receptor-positive breast cancer cell [20, 21]. In support of this, genistein has also been shown to promote tumor cell growth in immune-compromised, ovariectomized mice [22].

Recently, in order to investigate more extensively the effects of genistein in various models, we and others have used expression arrays to assess the global effects of this compound on RNA expression [16, 2326]. We have shown that in a prostate cancer cell culture model, physiologically achievable concentrations of genistein can inhibit androgen-dependent pathways and alter cell cycle and stress-related pathways by impacting transcription of multiple genes [25]. These studies, along with other molecular studies, support the idea that genistein can impact multiple cellular pathways. However, its specificity in different cell types and its concentration-dependent effects on cancer-related pathways in mammary cancer cells remain to be delineated. We reasoned that since genistein can potentially act as a weak estrogen, further elucidation of the global pattern of molecular changes it causes in estrogen receptor-positive breast cancer cells would be important. We also reasoned that consideration of concentration-dependent effects would be would be needed. To address these issues, we utilized a microarray gene profiling approach to examine gene expression in estrogen receptor-positive MCF-7 human breast cancer cells exposed to different concentrations of genistein. We further compared our results to published results of gene expression changes found in MCF-7 cells exposed to physiologic concentrations of estradiol [27]. Finally, in order to illustrate the specificity of genistein-induced gene expression changes in breast cancer cells, we discuss our findings in the context of our previous findings in similarly exposed androgen-responsive LNCaP human prostate cancer cells [25].

Materials and methods

Chemicals

Genistein, dimethylsulfoxide (DMSO) and chloroform were purchased from Sigma Chemical Co. (St. Louis, MO, USA). Cell culture media and reagents were from Invitrogen (Carlsbad, CA, USA).

Cell culture and treatments

MCF-7 cells were cultured as described previously [18]. For cell proliferation experiments, cells were plated at 0.5 × 106 cells/well. Twenty-four hours after plating, cells were treated with 1, 10 or 25 μM genistein or vehicle (DMSO) (final concentration, 0.001%) for 72 h. For microarray experiments, 1 × 108 cells/flask were plated into T-175 flasks in media containing 5% charcoal dextran-treated serum (CDS) as previously described [21]. Twenty-four hours after plating, cells (3 flasks each) were treated with either 1, 5, or 25 μM of genistein or vehicle for 48 h prior to isolation of RNA. For polymerase chain reaction (PCR) confirmation of microarray results, 1–2 × 106 cells were plated in 6-well plates. Twenty-four hours after plating, cells were treated in triplicate with vehicle or 1, 5, or 25 μM of genistein. Cells were incubated for 48 and RNA isolated [25].

Cell proliferation

Cells were treated in quadruplicate as described above and proliferation assessed using the sulfrhodamine-B (SRB) method as described previously [21]. Results are expressed on a relative proliferation index scale (mean ± SD). Significance testing was done using ANOVA followed by Bonferroni posthoc testing, and differences with two-tailed P value of < 0.05 were considered significant.

RNA isolation and PCR analysis

Cells were harvested after treatment (described above) and total RNA isolated for both microarray and PCR analysis using a method described previously [25]. The Taqman real-time PCR method was used to determine mRNA levels of specific genes following previously reported methods [25]. Primers and probes for the glyceraldehyde-3-phosphate dehydrogenase (GAPDH), catchol-O-methyltransferase (COMT), JUN, p21/waf2/cip 1 (CDKN1A) and Bcl-2 interacting killer (BIK) mRNAs were purchased from Applied Biosystems (Foster City, CA, USA). The threshold cycle (Ct) method, as described in the manufacturer’s protocol, was used to generate relative expression values. Results were expressed relative to the vehicle control, and ANOVA followed by Bonferroni post-hoc testing was performed to compare differences between treatments. Two-tailed tests with P < 0.05 were considered significantly different. In all experiments, media containing either test compounds or vehicle were replenished daily, and all treatments were done in triplicate.

Probe synthesis, hybridization and array analysis using Affymetrix GeneChips

Affymetrix GeneChip Human Genome U133A Arrays (Affymetrix, Santa Clara, CA, USA) were used for all gene expression analyses. Probe synthesis, hybridization and initial expression analysis of the Affymetrix GeneChip microarrays were performed according to the Affymetrix recommended protocol and as described previously [25]. Briefly, first-strand and second-strand cDNA syntheses were performed using total RNA (10 μg). The resulting double-stranded cDNAs were then purified using the GeneChip Sample Cleanup Module (Qiagen, Valencia, CA, USA). Following cleanup, biotin-labeled cRNA was synthesized from the cleaned cDNA using the Enzo BioArray HighYield RNA Transcript Labeling Kit (Affymetrix) and purified using the GeneChip Sample Cleanup Module (Qiagen). Final cRNA concentrations were quantified as described in the Affymetrix protocol, fragmented, and hybridized to Affymetrix chips for 16 h. Hybridization and washing of the labeled cRNA to the chips and detection of hybridized signal intensities were performed according to the protocols provided by the manufacturer. Each sample was run on a single chip (i.e., 3 microarray chips were run for each concentration of genistein).

Microarray data analysis

Data preprocessing and normalization were performed using the Bioconductor package for Affymetrix arrays [28]. All statistical analyses were performed with the R program. All spots based on robust multichip analysis (RMA) summary measures were used in class comparison analyses. To identify genes that were different among the different genistein concentrations, an F test was performed separately for each probe set represented on the arrays. Probe sets showing variation between models at the 0.001 level of statistical significance and showing at least 1.5-fold change or greater were identified. We further tested the difference in mean expressions in subgroups through contrast analysis. The contrast tests included control versus each genistein concentration group (all pairwise comparisons). Differentially expressed genes from these comparisons were also identified as those that were significant at the P < 0.001 level and showed at least 1.5-fold change or greater. Additional data processing and analysis was performed using Microarray Suite (MAS) 5.0 software (Affymetrix) as previously described [25]. Heat maps for comparison of gene expression were performed using cluster analysis and visualized with red representing up-regulated and green representing down-regulated genes as compared to vehicle-treated control cells. When data are shown with corresponding fold change numbers listed to the right of the heat maps, significant fold changes (≥ 1.5 fold, P < 0.01) are listed with changes at the other concentrations also included for comparison.

Principal components analysis

Global gene expression similarities/dissimilarities among arrays were examined by principal components analysis (PCA). PCA is a linear transformation that shows individual sample variances along the coordinate axes by minimizing the covariances or correlations [29]. This procedure can be used for dimension reduction when most variance is accounted for by a few principal components. A projection on the largest three principal components accounting for a certain percent of the total variance (computed using software Partekpro 5.0, Partek Inc., St. Charles, MO, USA) is illustrated as a 3-D plot. Each array is shown as a single point on the 3-D plot, and the distance between the points represents a measure of dissimilarity of expression patterns between the arrays. Our PCA analysis was performed on 10,254 probe sets having good replicate correlations. In our analysis, replicate correlation was considered to be good (> 0.7) for each gene at different genistein concentrations by determining the correlation of the individual measurements with their corresponding concentration’s average. This filter allowed us to describe the effect of the genistein concentration on gene expression pattern while simultaneously minimizing the random variation within replicates. The largest three principal components accounted for 66.0% of the total variance.

Pathway analysis

The genes significantly altered by genistein treatment were categorized by gene ontology using expression analysis systematic explorer (EASE) software [30]. Significantly altered genes are first compared to all the genes present on the array and over-represented ontologies are identified, along with the genes representing each ontology. The probability of finding the number of significant genes of an ontology from the list of significant genes is then calculated as the Fisher exact probability, given the frequency of population hits in the total population on array. The EASE score represents this probability of over-representation, which is the upper boundary of the distribution of Jackknife Fisher exact probabilities as determined by EASE.

Results and discussion

DNA microarray analysis of global gene expression in MCF-7 cells treated with genistein

As we have previously demonstrated [21] and as illustrated in Fig. 1, after 72 h of treatment, genistein exerts a concentration-dependent effect on the growth of MCF-7 human breast cancer cells. At lower, physiologically achievable concentrations (1–5 μM) [21], genistein increases cell proliferation, whereas at a higher pharmacological concentrations (>10 μM), it inhibits cell proliferation. To further describe the earlier molecular events that contribute to this effect, we analyzed global gene expression changes in MCF-7 cells exposed for 48 h to low (1 μM), moderate (5 μM) and high (25 μM) concentrations of genistein using the Affymetrix GeneChip platform as outlined in Materials and methods. Cells were cultured in 5% CDS to minimize the effects of estradiol in serum as described in Material and methods.

Fig. 1
figure 1

Effects of genistein on MCF-7 cell growth. MCF-7 cells were treated with 0, 1, 10, or 25 μM genistein for 72 h and cell growth determined using the sulfrhodamine-B method as described previously [21]. Results are expressed as relative proliferative index (mean ± SD, n = 4). Bars with different letters indicate values significantly different from each other at P < 0.05 using ANOVA follow by posthoc testing

RMA was used to create a summary list of Affymetrix probe sets, and these were analyzed for significant changes (P < 0.001) of 1.5 fold or greater in the different comparisons. Increasing numbers of genes were changed with increasing concentrations of genistein as compared to control: 1 μM genistein versus control yielded 207 significantly altered genes; 5 μM genistein versus control yielded 232 significantly altered genes; and 25 μM genistein versus control yielded 248 significantly altered genes. In contrast, no significant differences were found between 1 versus 5 μM genistein, and only 39 and 17 genes were found to be significantly altered when comparing 25 μM genistein to 1 and 5 μM genistein, respectively.

We further compared the different treatment groups by PCA of the RMA probe sets (Fig. 2). Each point on the PCA plot corresponds to one array, with arrays having similar gene expression patterns clustering closer together. We found that genistein treatment at all concentrations led to expression profiles that deviated significantly from that of untreated cells. There were smaller differences in overall gene expression profiles seen when comparing the two lowest concentrations (1 and 5 μM) of genistein to one another versus the large differences seen between 25 μM genistein-treated cells and all other treatments.

Fig. 2
figure 2

Principal components analysis of MCF-7 cell gene expression after genistein exposure. MCF-7 cells were treated with 0, 1, 5, or 25 μM genistein and global gene expression similarities/dissimilarities among arrays were examined by principal components analysis as described in Materials and methods. An array is shown as a single point in the 3-D plot, with the distance between the points representing a measure of dissimilarity of expression patterns between the arrays. Letters indicate independent replicates and numbers represent the concentration of genistein for that replicate in μM. Results from the same concentration are denoted by the same color. The three largest principal components accounted for 66.0% of the total variance

As the RMA procedure is only one statistical procedure for selecting significant gene changes [31], we complemented our initial list of significantly altered genes by performing another analysis using Affymetrix MAS 5.0 software and thresholds of P < 0.01 and altered by ≥ 1.5 fold, as previously described [25]. This analysis yielded 345, 372, and 434 additional genes of interest for 0 versus 1, 0 versus 5 and 0 versus 25 μM genistein, respectively. Since we were able to confirm the results from this latter analysis for multiple genes (see below), we include them in the following results and discussion points and, for consistency, present fold change numbers only from the MAS 5.0 analysis in all subsequent heat maps. Of note was the fact that genes found significantly altered by RMA were consistently represented in the list of genes shown to be significantly altered by MAS 5.0 analysis (one exception is noted below). We indicate if a gene was significant only in the latter procedure.

We next performed an EASE analysis on those genes found to be significantly altered by MAS 5.0 analysis in order to assess the gene categories with the greatest representation of genes significantly altered by genistein treatment based on Gene Ontology (http://www.geneontology.org) and then canonical KEGG (http://www.genome.jp/kegg/pathway.html) and BioCarta (http://cgap.nci.nih.gov/Pathways/BioCarta_Pathways) pathway annotations. We observed that, among others, apoptosis-related, steroid biosynthesis and xenobiotic metabolism gene categories had multiple members significantly altered (data not shown). These genes, along with genes involved in cell proliferation are discussed below.

Genistein-induced alterations in genes involved in maintaining cell number homeostasis in MCF-7 cells

Given our initial observations that cell growth varied significantly by concentration of genistein (Fig. 1), we first examined genistein-induced gene expression changes in cell cycle/proliferation- and apoptosis-related pathways, as these pathways contribute to overall cell number homeostasis.

Effects of genistein on expression of cell cycle-related genes

Shown in Fig. 3a are the cell cycle-related genes found to be significantly affected by at least one concentration of genistein. Results are shown as a heat map, with red representing up-regulated and green representing down-regulated genes as compared to vehicle-treated control cells. Corresponding fold-change numbers are listed to the right of the heat map for each concentration of genistein. If a gene was significantly changed by at least one concentration, the changes at the other concentrations are also shown.

Fig. 3
figure 3

Genistein-responsive genes in MCF-7 cells. MCF-7 cells were treated with 0, 1, 5 or 25 μM genistein for 48 h, and cluster analysis of microarray results was performed as described in Materials and methods. Findings for selected gene categories are depicted as heat maps with red indicating up-regulated and green indicating down-regulated genes (as compared to vehicle-treated control). Columns represent different genistein concentrations and rows represent different genes, with mean fold-change ratios shown to the right of each gene. If a gene was significantly changed by at least one concentration of genistein (≥ 1.5-fold vs. control, P < 0.01), the changes at the other concentrations are also shown. a Cell cycle-related genes, b apoptosis-related genes; and c genes differentially altered at high versus low genistein concentrations

Initial inspection revealed cell cycle-related gene expression changes were primarily in the positive direction. For example, the CDC28 protein kinase regulatory subunit 1B (CKS1B) and CDC28 protein kinase regulatory subunit 2 (CKS2) were both significantly up-regulated, the former even at low genistein concentrations. CKS1 has been reported to promote p27 cyclin inhibitor degradation [32]. CKS2 is an orthologue of CKS1 and appears to be involved in meiosis [33]. Both of these proteins regulate cyclin-dependent kinase (cdk) activity and promote proliferation [34]. The DNA replication-related protein (CDT1) gene and CAAT enhancer binding protein gamma (C/EBPG) gene were also both significantly up-regulated, and, interestingly, fold changes were greatest at low (1 and 5 μM) genistein concentrations. These proteins are known to be associated with proliferative activity of cells [35, 36]. Further, the gene encoding the dual-specificity phosphatase CDC14 cell division cycle 14 homolog A (CDC14A), required for centrosome separation and productive cytokinesis during cell division [37], was also up-regulated, though this was seen at all genistein concentrations and was found significant by MAS 5.0 analysis only.

In addition to the genes described above, there appeared to be a global effect of genistein on several genes involved in the p53-related pathway, primarily at the highest genistein concentration, that would be expected to contribute to a decrease in cell proliferation (Fig. 3a). For example, MAS 5.0 analysis revealed that CDKN1A or p21/waf1/cip1, a cyclin inhibitor and a classic p53 responsive gene [38], was significantly up-regulated by 25 μM genistein. Figure 4a illustrates real-time PCR confirmation of array results for CDKN1A with significantly increased expression appearing only at 25 μM genistein. Two additional genes, cyclin G1 (CCNG1), a cyclin known to respond to p53 [39], and DNA-damage inducible protein GADD45 alpha (GADD45A), were both up-regulated at 25 μM genistein. The former was significant only by MAS 5.0 and the latter only by RMA; thus, GADD45A does not appear on the heatmap (1.7-fold increase). GADD45 beta (GADD45B) was also up-regulated at high genistein concentrations, and both GADD45 alpha and beta are known p53-responsive genes [40]. Finally, the gene encoding B-cell lymphoma 3-encoded protein (Bcl-3), a p53 down-regulated protein involved in the NF-kappa B pathway [41], was down-regulated by both 5 and 25 μM genistein treatment.

Fig. 4
figure 4

Real-time PCR analysis of selected genes altered in microarray analyses by exposure to genistein. MCF-7 cells were treated with 0, 1, 5 or 25 μM genistein for 48 h as described in Materials and methods. Total RNA was isolated and real-time PCR analysis performed, also as described in Materials and methods. Results are expressed as mean ± SD relative to vehicle control (n = 3). *Represents significantly different from control at P < 0.05

The results described above are consistent with the up-regulation of the p53 protein that we have seen in MCF-7 cells after 48 h of 25 μM, but not 1 or 5 μM, genistein treatment (data not shown), and support the idea that activation of p53-mediated pathways can occur at higher, pharmacological levels of genistein. This activation would then alter gene transcription and ultimately result in cell cycle arrest and therefore decreased cell proliferation at high genistein concentrations. Thus, though several cell proliferation genes are significantly increased at each concentration of genistein, induction of growth inhibitory pathways (p53-mediated) at higher genistein concentrations would result in an overall decreased cell number at these concentrations.

Two other cell cycle-related kinases, cell cycle division protein 20 and 2 homolog (CDC20 and CDC2) [42, 43], were also found to be up-regulated at 25 μM genistein. However, the implication of these genes and their corresponding proteins in genistein’s biological effects is less clear.

Effects of genistein on apoptosis-related genes

Apoptosis, or programmed cell death, was determined by EASE analysis to be significantly altered by genistein and is another important process that determines overall cell number homeostasis; therefore, we also examined the effects of genistein on expression of genes belonging to two major families of proteins that are involved in apoptosis pathways: the BCL-2 family [44] and the tumor necrosis factor receptor (TNFR) family [45, 46]. As shown in Fig. 3b, genistein appeared to affect the expression of multiple Bcl-2 family genes, and both pro- and anti-apoptotic genes were altered. For example, RNA expression of BIK, an apoptosis inducing protein [47], was significantly decreased by 2.3-fold, 4.3-fold, and 2.3-fold at 1, 5, and 25 μM genistein, respectively. Further examination of this effect using real-time PCR on RNA isolated from different cultures of MCF-7 cells treated at the same concentrations of genistein for 48 h (Fig. 4b) confirmed this finding to some extent, in that low concentrations of genistein induced large decreases in BIK expression, but 25 μM genistein had no effect. This difference could be related to slightly different culture conditions since the experiments were done at different times. Large decreases in BIK expression at low concentrations may be explained by the fact that regulation of BIK can occur through estrogen-mediated pathways [48]; exposure of MCF-7 cells to estradiol inhibits BIK expression (T. Wang, unpublished observation). Genistein at low concentrations may be acting on this particular gene in an estrogen-like fashion (see also “estrogen-responsive gene” section below), inhibiting BIK expression and contributing to decreased apoptosis.

The Bcl-2L1 (or Bcl-XL) gene, which encodes an anti-apoptotic protein [44], was significantly inhibited by all genistein concentrations. In contrast, several other Bcl-2 family genes were significantly altered at only 25 μM genistein. These included genes encoding the pro-apoptotic proteins Bcl-2 antagonist of cell death (BAD) and Bcl-2/adenovirus E1B 19 kDa interacting protein 3 (BNIP3) [49], which were significantly inhibited and induced, respectively (BAD was significant by MAS 5.0 only). BNIP is reported to respond to hypoxia, a situation that induces p53 [50]. The gene for the anti-apoptotic protein Bcl-2-associated athanogene (BAG-1) [51] was, by MAS 5.0 analysis, significantly inhibited by 25 μM genistein. Thus, there is some indication from these gene changes that at 25 μM genistein there are some unique pro-apoptotic signals and inhibition of some anti-apoptotic signals.

The effects on TNFR family genes [46] were also examined, and the results are illustrated in Fig. 3b. As with the Bcl-2 family, treatment of cells with genistein resulted in RNA expression changes in multiple TNFR family genes. Tumor necrosis factor receptor superfamily member 1A (TNFRSF1A or TNFR1) and 11B were both significantly down-regulated by genistein, the latter only at concentrations < 5 μM. The TNFRSF1A protein binds to TNF-α and triggers apoptosis [52], while the TNFRSF11B protein, whose transcript was decreased 1.6- and 2.8-fold by 5 and 25 μM genistein, respectively, appears to serve as decoy receptor and is anti-apoptotic [53]. By contrast, death receptors that activate apoptosis, such as TNFRSF6 (or FAS) and TNFRSF10b, were induced by genistein, the latter only at 25 μM, and both results were significant by MAS 5.0 analysis only. TNFRSF6 is a receptor for TNFSF6/FASLG that activates apoptosis through recruitment of caspase-8 [54]. TNFRSF10b is one of the receptors that can bind the cytotoxic ligand TNFSF10/TRAIL [55] and also activates apoptosis through recruitment of caspase-8. Both the receptors are known to be p53 inducible [53]. However, TNFSF10, the ligand for TNFRSF10 [56], was down-regulated at 25 μM genistein, as identified by MAS 5. TNF receptor-associated factor (TRAF)s 4 and 5, both adaptor proteins that mediate apoptosis signal transduction from TNF receptors to intracellular components [57, 58], were also significantly down-regulated at the transcript level by genistein, though by RMA the results were significant only for TRAF5 and for that gene only at ≥5 μM. Thus, there are suggestions of differential effects of genistein on apoptosis-related genes at different concentrations, and this, combined with differential cell proliferation transcript and p53 protein effects, may contribute to overall differences in cell number homeostasis after genistein exposure at different concentrations.

Regulation of other genes that are differentially affected at high versus low genistein concentrations

Numerous other genes were significantly altered by genistein, many at all concentrations. However, we next chose to consider only those transcripts that were significantly altered at either high or low concentration genistein (Fig. 3c), since these genes may also contribute to the differential growth patterns seen in Fig. 1 and likely represent pharmacological (25 μM) versus physiological (1 or 5 μM) effects. For example, consistent with our previous observation [21], estrogen receptor alpha (ESR1) expression was significantly decreased by 25 μM genistein, and this decrease could limit genistein’s ability to signal through this receptor, resulting in decreased proliferation at higher genistein concentrations. Furthermore, nuclear receptor coactivator 2 (NCOA2) and NCOA3, ubiquitous nuclear receptor co-activators that facilitate hormone-dependent transcription of steroid hormones [59, 60], were also significantly down-regulated at high- and low-concentration genistein, respectively. These findings suggest that genistein could differentially alter hormone signaling, depending on its concentration.

Changes in additional genes indicate that several other metabolic enzymes are differentially affected by high and low concentrations of genistein; many of these genes were significantly altered (up-regulated) only by 25 μM genistein (Fig. 3c). For example, the aldo-keto reductases (AKRs), AKR1C1 and AKR1C3, were up-regulated by 25 μM genistein. AKR1C1 catalyzes the inactivation of progesterone [61], and AKR1C3 has been shown to catalyze the reduction of prostaglandins [62]. Cytochrome P450 1B1 (CYP1B1), another enzyme that can contribute to steroid metabolism in addition to xenobiotic metabolism [63], was also up-regulated at 25 μM. The 7-dehydrocholestrol reductase (DHCR7) gene, which encodes an enzyme that catalyzes production of the hormone precursor cholesterol by reduction of the C7–C8 double bond of 7-dehydrocholesterol [64], and the gene for stearol-CoA desaturase (SCD), which can contribute to increased cholesterol synthesis [65], were increased 1.6- and 2.0-fold by 25 μM genistein, respectively. In contrast to the above, the 24-dehydrocholestrol reductase (DHCR24), also known as seladin-1, which contributes to cholesterol synthesis by catalyzing the reduction of the delta-24 double bond of sterol intermediates [66], was down-regulated by low-concentration genistein only. Thus, the main effect on these steroid and other metabolizing enzymes appears to be an increase in gene transcription at high genistein concentrations that could contribute to increased cholesterol synthesis, increased hormone metabolism and decreased hormone signaling that potentially contribute to decreased growth in cells exposed to ≥25 μM genistein.

Interestingly, JUN and JUN D, two JUN transcription factor family members [67], appeared to be up-regulated at 25 μM genistein. Real-time PCR confirmation of JUN up-regulation is illustrated in Fig. 4c. Thus, in addition to the activation of p53-dependent events as discussed earlier, activation of JUN family genes would support an overall stress-responsive effect of genistein at high concentrations.

Finally, another interesting change was the down-regulation of fibroblast growth factor 12 (FGF12) by 25 μM genistein. Though the specific function of FGF12 has not yet been determined, FGF family members possess broad mitogenic and cell survival activities, and signaling through FGFRs is associated with many processes, including embryonic development, cell proliferation, morphogenesis, cell migration, cell survival, tumor growth, and invasion [68]. This change could contribute to the decreased cell proliferation observed with 25 μM genistein exposure.

Effects of genistein on estrogen-responsive genes

Genistein has been shown to be a weak estrogen in some model systems, including MCF-7 cells [21]. To understand whether genistein elicited an estrogen-like response on gene transcription at the concentrations used in this study, we compared a list of transcripts that have been shown by Frasor et al. [27], to be altered ≥2.5-fold in MCF-7 cells exposed to 10 nM 17β-estradiol 48 h to our results for genistein (Fig. 5). We also included their results with 17β-estradiol + the antiestrogen ICI 182,780 on this group of genes, as combining estrogen with ICI 182,780 allows for the differentiation between estrogen receptor (ER)-α and non-ER-α-mediated transcription changes, and we were interested in whether genistein was likely impacting gene expression via the estrogen receptor. Genes were included in the figure based on the following criteria: those that were significantly up or down regulated (Fig. 5, clusters a and b, respectively) by 17β-estradiol after 48 h (as defined by Frasor et al.) and significantly altered by genistein treatment at one or more concentrations (as defined by our criteria). For the latter group, all genes were altered by 1.5-fold at P < 0.01 in one or more comparisons of 1, 5, 25 μM genistein versus control. For two genes there is no number in the E2 + ICI column, indicating that ICI had an independent effect on the gene and was not an agonist or antagonist of the E2 response [27].

Fig. 5
figure 5

Effects of genistein on estrogen-responsive genes. Cluster analysis was performed on transcripts that had been shown by Frasor et al. [27], to be altered ≥2.5-fold in MCF-7 cells exposed to 10 nM 17β-estradiol (E2) for 48 h [27] and ≥1.5-fold in MCF-7 cells exposed to 1, 5, or 25 μM genistein for 48 h. Also included are the Frasor et al. results for the genes after exposure to 17β-estradiol + the antiestrogen ICI 182,780 (E2 + ICI) for 48 h. Genes were included in the figure that were significantly up regulated (cluster a) or down regulated (cluster b) by 17β-estradiol after 48 h (as defined by Frasor et al.) and significantly altered by genistein treatment at one or more concentration as compared to control. No number in the E2 + ICI column indicates that ICI had an independent effect on the gene and was not an agonist or antagonist of the E2 response [27]. All genes were altered at P < 0.01 in one or more comparisons of 1, 5, 25 μM genistein versus control. Findings are depicted as heat maps with red indicating up-regulated and green indicating down-regulated genes (as compared to vehicle-treated control), columns representing the different treatments, and rows representing different genes, with mean fold-change ratios shown to the right of each gene. If a gene was significantly changed by at least one concentration of genistein, the changes at the other concentrations are also shown

Not surprisingly, classic estrogen-responsive genes such as TFF1 (also known as pS2, cluster a), were induced by genistein at each concentration as was previously demonstrated by us and others [21, 69]. The gene for the GREB1 protein (GREB1, cluster a), an estrogen-inducible gene that may be involved in regulation of hormone-dependent breast cancer growth [70], was up-regulated by all studied concentrations of genistein as well. The EGR3 gene (cluster a), which encodes a transcriptional factor reported to be involved in estrogen-mediated pathways [71], was also up-regulated by 1, 5 and 25 μM genistein. Additionally, the estrogen-inducible chemokine CXCL12 gene (cluster a) was up-regulated at each of these concentrations. CXCL12 has been shown to affect breast cancer cell migration through its receptor, CXCR4 [72]. Transcription of AREG (amphiregulin, cluster a), an estrogen-responsive growth factor [73], was also up-regulated in response to genistein exposure, and PCR confirmation of AREG expression is illustrated in Fig. 4d.

Many genes, including BIK, an apoptosis-related gene described above, and the interleukin 1 receptor (IL-1R1) gene (both in cluster b) were inhibited by both estrogen and genistein. PCR confirmation of IL-1R1 expression is illustrated in Fig. 4e. The IL-1 system appears to be vital in the local control of tumor growth [74], is important in regulating “protumorigenic” activities within the tumor microenvironment [75], and contributes to angiogenesis, tumor proliferation, and tumor invasion [76].

The majority of significant changes in the genes included in Fig. 5 occurred at all, including low (1 μM), genistein concentrations. The effects are likely occurring via estrogen receptor signaling, since, in the presence of 17β-estradiol, almost all changes in these genes were at least partially inhibited by the antiestrogen ICI 182,780. Despite the estrogen-like effects illustrated in Fig. 5, genistein actually exerted a differential effect compared to 10 nM 17β-estradiol [27], as only approximately 35% of the estrogen-responsive genes shown to be altered ≥2.5-fold by 17β-estradiol in the study by Frasor et al., were also altered by genistein in our system. Furthermore, the magnitude of changes was most often lower for genistein than 17β-estradiol. For example, for TFF1, genistein treatment resulted in a 2-fold induction, while 17β-estradiol treatment induced close to a 14-fold induction. GREB1 was an exception in that genistein’s effect appeared to be greater than that of 17β-estradiol. Thus, genistein did not completely mimic 17β-estradiol’s effects.

We observed additional changes in other documented estrogen-responsive genes that do not appear in our comparative analysis in Fig. 5. For example, the argininosuccinate synthetase (ASS) and TPD52L1 genes were induced by genistein treatments (ASS: 4.00-fold, 3.73-fold, and 2.83-fold at 1, 5, and 25 μM genistein, respectively; TPF52L1: 1.87-fold, 1.87-fold, and 2.00-fold at 1, 5, and 25 μM genistein, respectively). ASS is an arginine-synthesizing enzyme that has been explored as a therapeutic target [77], and TPD52L1 is involved in G2/M progression [78]. The COMT gene, for which we performed real-time PCR confirmation (Fig. 4f), was down-regulated by genistein treatment (0.65-fold, 0.57-fold, and 0.53-fold at 1, 5, and 25 μM genistein, respectively) and is another estrogen down-regulated gene [79]. Hence, our results not only confirmed previously demonstrated estrogenic effects of genistein at low concentrations (1 μM, Fig. 5), but also contribute the novel findings that the estrogen-responsive genes GREB, COMT, IL-1R1, ASS and TPD52L1 respond similarly to genistein exposure. Furthermore, we report that 1, 5, or 25 μM genistein treatments only partially mimic the gene expression patterns observed with 10-nM estradiol treatment of MCF-7 cells.

Specificity of genistein in cells from different tissue origins

From the analysis described above, it is clear that genistein is active in altering gene transcription in MCF-7 cells and that the effects are dependent on the concentration of genistein. However, it is not known if some or all of these changes are cell-type specific. To begin to address this question, we compared our present results in MCF-7 cells to those from our previous microarray experiment in LNCaP human prostate cancer cells exposed to the same concentration range of genistein for the 48 h [25] (Fig. 6). Genes were selected for the comparison if they were significantly differentially expressed (P < 0.01, ≤1.5-fold) in both cell lines when comparing the 25 μM genistein treatment group to the equivalent cell line control or 1 or 5 μM genistein. Significantly altered genes were then clustered together by similarity of expression pattern, resulting in five clusters of different patterns, a–e (Fig. 6).

Fig. 6
figure 6

Comparison of genistein’s effects in MCF-7 human breast cancer cells and LNCaP prostate cancer cells. MCF-7 cells were treated with genistein (0, 1, 5, 25 μM) and microarray analysis performed as described in Materials and methods. Microarray anlysis of LNCaP cells treated with the same concentrations of genistein was described previously [25]. Cluster analysis was performed on genes that were significantly differentially expressed (P < 0.01, ≥1.5-fold) in both cell lines when comparing the 25 μM genistein treatment group to the equivalent cell line control or 1 or 5 μM genistein and clusters were separated by expression pattern (clusters ae). Results are depicted as heat maps with red indicating up-regulated and green indicating down-regulated genes (as compared to vehicle-treated control), columns representing the different treatments in the two cell lines, and rows representing different genes. If a gene was significantly changed by at least one concentration of genistein, the changes at the other concentrations are also shown

Possibly because LNCaP cells were cultured in 10% FBS and MCF-7 cells were in 5% CDS, the two cell lines displayed quite different patterns at the lower genistein concentrations at which serum factors may be contributing to gene expression (evident in all clusters); however, at the highest genistein concentration, similar gene expression patterns emerged (see clusters b, d and e). For example, the induction of CDKN1A (cluster b) [38] as well as JUN (cluster d) [67] mRNA expression, discussed above for MCF-7 cells, occurred in both cell lines with 25 μM genistein exposure. For a number of metabolic genes (all in cluster d), including G6PD [80], ME1 [81], TXNRD1 [82], SCD [65], DHCR7 [64], and AKR1C1 [61], transcription was also up-regulated at 25 μM genistein in both cell lines, though AKR1C1 was also up-regulated by 1 and 5 μM genistein in MCF-7 cells. The dual specificity phosphatase 4 (DUSP4) gene [83] was also affected by 25 μM genistein in both cell lines, but in the case of this gene, expression was inhibited (cluster e). In LNCaP cells it was down-regulated in a concentration-dependent fashion, while in MCF-7 cells DUSP4 was down-regulated at 1 μM and remained at the same level throughout the concentration range. Another category of genes, including GREB1 [70] and IGF-1R [84] (both in cluster c), was regulated differently in the two cell lines (Fig. 6). GREB1 was up-regulated in MCF-7 but was down-regulated in LNCaP cells. On the other hand, IGF-1R was down-regulated in LNCaP cells but up-regulated in MCF-7 cells.

Thus, despite some gene expression differences at 25 μM genistein exposure (∼1/3 of genes at 25 μM were differentially regulated), we suggest that when present at pharmacological levels, genistein is non-specific and alters the expression of genes involved in multiple pathways similarly in both cell lines, whereas at lower concentrations, genistein acts in a more cell-type specific manner. Consistent with this hypothesis, we have observed growth inhibition for both cell lines at high (25 μM) but not low genistein concentrations [85]. Thus, at the level of cell growth, the effect supports what we have observed at the genetic level: differences at low and similarities at high concentrations of genistein.

Conclusions

The current study capitalizes upon oligo micoarray technology to examine genistein’s effects on global gene expression patterns across physiological and pharmacological concentrations in a human breast cancer cell line. Overall, we found that distinct sets of genes were regulated by exposure to different concentrations of genistein: at physiologically relevant lower concentrations, genistein elicited an expression pattern that may contribute to an increase in cell proliferation, while at a pharmacologically relevant concentration, genistein effected a pattern that may contribute to decreased total cell number. The latter was true in both MCF-7 breast cancer and LNCaP prostate cancer cell lines. These results are suggestive of the complex nature of genistein’s biological activity as well as the unique effects of genistein in different biological systems. Correlation of specific biological activity to beneficial or detrimental effects of genistein exposure in humans remains to be elucidated.