Abstract
We recently demonstrated that endoxifen (4-hydroxy-N-desmethyl-tamoxifen), a pharmacogenetically regulated metabolite of tamoxifen, is equipotent to 4-hydroxy-tamoxifen (4-OH-Tam) with respect to estrogen receptor binding and inhibition of 17β-estradiol (E2)-induced cell proliferation. Endoxifen was also found to be more abundant in human plasma than 4-OH-Tam, and its formation has been shown to be primarily catalyzed by cytochrome P450 2D6 (CYP2D6). Here, we report studies evaluating the effects of endoxifen, 4-OH-Tam, and E2 on gene expression in MCF-7 cells using Affymetrix U133A GeneChip Arrays (Santa Clara, CA). We detected 4062 genes that were E2-regulated (1924 induced; 2138 suppressed), and the ratio of E2-induced versus E2-suppressed genes was consistent regardless of the cutoff value. In the presence of E2, 2444 and 2390 genes were affected by 4-OH-Tam and endoxifen, respectively, when no minimal -fold change cutoff was implemented. The majority of genes regulated by the tamoxifen metabolites were also E2-responsive (74.4 and 73.3%, respectively). Endoxifen and 4-OH-Tam had overlapping effects on 1365 E2-sensitive genes, whose -fold effects between these metabolites were highly correlated (R2 = 0.99). A significant correlation was also found between the -fold effects of 249 E2-insensitive genes coregulated by both metabolites (R2 = 0.99). Hierarchical clustering analysis demonstrated similar gene regulation patterns between these metabolites, which were distinct from E2 or vehicle treatment patterns. Using real time-polymerase chain reaction, we validated the gene expression patterns of five genes that were differentially regulated by endoxifen and 4-OH-Tam. We conclude that endoxifen and 4-OH-Tam have similar effects on global gene expression patterns in MCF-7 cells and that the majority of the affected genes are estrogen-regulated genes.
Tamoxifen, a selective estrogen receptor modulator (SERM), is widely used in the treatment and prevention of hormonal-dependent breast cancer. In addition to its activity in the breast, tamoxifen also affects many other tissues, including beneficial effects on bone (Grey et al., 1995; Powles et al., 1996) and blood lipids (Love et al., 1994; Saarto et al., 1996) as well as undesirable effects on hot flashes, deep venous thrombosis, and rare cases of endometrial carcinoma (Meier and Jick, 1998; Bernstein et al., 1999; Clarke et al., 2001). Although these effects are widely observed, there is a large interindividual variability in the responses (Meier and Jick, 1998). For example, the magnitude of changes in lipids and bone mineral density, as well as the frequency and severity of tamoxifen induced hot flashes, are widely variable among patients. In addition, approximately 70% of the estrogen receptor-positive patients have clinical responses to tamoxifen therapy, but most will eventually develop resistance (Clarke et al., 2001).
The factors responsible for this clinical heterogeneity are poorly understood. Because it is known that tamoxifen is metabolized in vitro and in vivo to several metabolites of varying potency and antiestrogenicity, differences in metabolism may contribute to the clinical interindividual variability in tamoxifen response. Many of the tamoxifen metabolites [e.g., N-desmethyl-tamoxifen, 4-hydroxy-tamoxifen (4-OH-Tam), tamoxifen N-oxide, and α-hydroxy-tamoxifen] have been characterized previously (Lien et al., 1991; Crewe et al., 2002; Lee et al., 2003; Stearns et al., 2003; Desta et al., 2004). Of these metabolites, 4-OH-Tam has been shown to exhibit 30- to 100-fold higher potent antiestrogenic activity than that of tamoxifen (Coezy et al., 1982; Jordan, 1982; Katzenellenbogen et al., 1984). For this reason, tamoxifen has even been referred to sometimes as a prodrug that requires conversion to its hydroxylated metabolite to exert its activity (Furr and Jordan, 1984). A large number of in vitro studies have used this metabolite, rather than the parent compound to characterize and study the effects of the drug in a number of model systems. Our recent studies have implicated another metabolite, 4-hydroxy-N-desmethyl-tamoxifen (endoxifen), as a potentially important factor in mediating the effects of tamoxifen. Although this metabolite was described previously in human breast tumor tissue (Mauvais-Javis et al., 1986) and in human bile (Lien et al., 1988), its pharmacological activity was not investigated in detail. In a series of in vitro experiments, we recently demonstrated that endoxifen has pharmacological activity comparable with that of 4-OH-Tam, with respect to its binding affinity to estrogen receptors (ERs), antiproliferative activity, and inhibitory effects on the expression of typical estrogen-regulated genes in ER-positive breast cancer cells (Stearns et al., 2003; Johnson et al., 2004). We also found a similar potency between endoxifen and 4-OH-Tam in suppressing progesterone receptor expression in MCF-7 cells (Lim et al., 2005). Our recent in vitro and in vivo studies have demonstrated that endoxifen is a secondary metabolite of tamoxifen that is primarily formed from N-desmethyl-tamoxifen by the cytochrome P450 2D6 (CYP2D6), a genetically polymorphic enzyme, and that the plasma concentrations of endoxifen in patients taking a commonly prescribed dose of tamoxifen (20 mg per day) is on the average over 6-fold higher than that of 4-OH-Tam (Stearns et al., 2003; Desta et al., 2004; Jin et al., 2005). These data prompted us to propose a working hypothesis that endoxifen could be a more important contributor to tamoxifen activity than 4-OH-Tam. Therefore, it seems important to characterize fully the pharmacological effects of endoxifen to help better understand the overall activity of tamoxifen and the interindividual variability in its responses.
The data collected so far suggest that endoxifen might be a potent selective estrogen receptor modulator. As described above, endoxifen has been shown in vitro to suppress two E2-sensitive genes (progesterone receptor and pS2) in breast cancer cells (Johnson et al., 2004; Lim et al., 2005); however, the whole genome-wide effect of endoxifen has not yet been reported. This is particularly important to understand because it is known that estrogens regulate a variety of physiological events by affecting various genomic and nongenomic signaling pathways. In the present study, we used Affymetrix U133A GeneChip Arrays (Santa Clara, CA) to test the effects of endoxifen on global gene expression in breast cancer cells and to compare its activity with that of the other known potent metabolite, 4-OH-Tam.
Materials and Methods
Chemicals. E2, pure Z-(trans)-isomers (>98% Z) of 4-OH-Tam, and all other drugs were purchased from Sigma (St. Louis, MO). Endoxifen (4-hydroxy-N-desmethyl-tamoxifen; mixture of 75% Z-isomers and 25% E-isomers) was synthesized by Dr. Ross Weatherman in the Department of Medicinal Chemistry and Molecular Pharmacology, Purdue University (West Lafayette, IN) and its synthetic method has been described in our previous report (Johnson et al., 2004). All compounds were dissolved in 0.1% ethanol, which does not produce any significant effect on gene expression.
Cell Culture. The ERα-positive human breast carcinoma cells, MCF-7, were maintained at 37°C in a 5% CO2 atmosphere in improved minimum essential medium (IMEM; Biosource International, Inc., Camarillo, CA) supplemented with 10% fetal bovine serum (Mediatech, Inc., Herndon, VA). Before drug treatments, cells were grown in steroid depleted medium containing phenol red-free IMEM supplemented with 10% charcoal-stripped calf serum (Valley Biomedical, Winchester, VA) for 3 days, as described previously (Skaar et al., 1998).
Drug Treatment. Cells were treated with various combinations (vehicle, E2, 4-OH-Tam, endoxifen, E2 + 4-OH-Tam, E2 + endoxifen, or E2 + 4-OH-Tam + endoxifen) for 24 h, the time that showed E2-induced maximal expression of several genes, such as progesterone receptor genes (May et al., 1989; Lim et al., 2005) pNR-17, pNR-25, pNR-100 genes (May and Westley, 1988), pNR-1, and pS2 (May and Westley, 1986). To obtain relevant information on differential regulation of gene expression profiles by the test drugs, optimal dose selection for each drug is required. The E2 dose used was 10-10 M (100 pM), and it was selected based upon our previous results in which this dose induced mid-to-submaximal effects on the dose-response curves of cell proliferation and gene expression of progesterone receptor (Lim et al., 2005) and pS2 in MCF-7 cells (Johnson et al., 2004). The dose range in dose-response experiments of E2-induced progesterone receptor expression (Lim et al., 2005) was very similar with those of E2-induced proliferation curves in MCF-7 cells (Olsen et al., 2002; Johnson et al., 2004). Based upon our previous dose-response data on the inhibitory effect of 4-OH-Tam or endoxifen on E2-induced cell proliferation and gene expression of progesterone receptor (Lim et al., 2005) and pS2 genes in MCF-7 cells (Johnson et al., 2004), the dose of both tamoxifen metabolites (endoxifen and 4-OH-Tam) was chosen as 10-7 M (100 nM) because this dose was shown to effectively inhibit E2-induced responses.
RNA Isolation. Total RNA was isolated from 2.5 × 105 cells using the RNeasy Mini Kit (Qiagen Inc., Valencia, CA) according to the manufacturer's instructions. Yield and quality were determined using spectrophotometry (Beckman DU640; Beckman Coulter, Inc., Fullerton, CA) and a Bioanalyzer RNA 6000 Nano Chip (Agilent Technologies, Palo Alto, CA), and samples were stored at -80°C. RNA integrity data will be provided on-line with the microarray data.
Affymetrix Microarray and Data Analysis. Gene expression patterns were determined using Affymetrix U133A GeneChip Arrays according to the manufacturer's instructions by the Indiana University Center for Molecular Genetics Core Facility. The average intensity on each array was normalized by global target intensity of 1000. The gene expression levels of samples were analyzed using Microarray Suite (MAS) 5.0 (Affymetrix) and entered into a custom-designed analysis platform (Microarray Data Portal) at the Center for Medical Genomics in Indiana University. To reduce false-positive results, four replicates were arrayed from four biologically independent cultures of treated MCF-7 cells performed on separate days per each drug treatment group. The genes, whose absence/presence call is less than 0.75 in all treatment groups, were filtered out before further analysis. This absence/presence filtering requires that a gene be detected on at least three of the four individual arrays per experimental condition, and this could avoid data that primarily represent “noise” at or near background intensities (McClintick et al., 2003). Only genes, whose p values of log-transformed data are less than 0.05 on a t test comparison (unequal variance) between two treatment groups, were considered as specific drug-regulated genes significantly. Correlation of -fold changes between endoxifen and 4-OH-Tam-induced E2-regulated genes are analyzed by linear regression, and R2 is reported. Hierarchical clustering analysis was also performed between expression levels induced by various drugs. One minus the Pearson correlation was used as the similarity measurement. In particular, the data for the cluster analysis contained all of the genes passing absence/presence filters. The microarray data of the present study will be available on the PharmGKB (The Pharmacogenetics and Pharmacogenomics Knowledge Base) website (http://www.pharmgkb.org/).
Gene Ontology Analysis. The microarray data were analyzed to determine whether the genes in each list are statistically enriched with genes of related functions. For every annotation gene ontology (GO) term, the hypergeometric distribution was applied to calculate a p value. This p value describes the relative enrichment of the genes (i.e., over-representation) in the list that are annotated with this GO term compared with the total genes on the microarray with the same GO annotation. We used the GeneNotes (http://combio.cs.brandeis.edu/GeneNotes/index.htm) (Hong and Wing, 2005) to implement this GO analysis.
Quantitative Real-Time Polymerase Chain Reaction. Primers were designed using the Primer3 software (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi) and purchased from Integrated DNA Technologies, Inc. (Coralville, IA). One microgram of DNase-treated (DNA-free; Ambion Inc., Austin, TX) RNA was reverse-transcribed using the Promega Reverse Transcription System (Madison, WI). Each real time-PCR reaction consisted of 2× Platinum Quantitative PCR Supermix-UDG (Invitrogen, Carlsbad, CA), 5 μl of cDNA (equivalent to 50 ng of total RNA), SYBR Green I in 1:1000 dilution (Invitrogen), fluorescein (10 nM), 0.5 μl of both forward and reverse primers (0.2 μM), and 6 μl of water. Reactions were carried out using an iCycler Thermal Cycler (Bio-Rad Laboratories, Hercules, CA) with thermal cycling conditions: initial step of UDG carryover decontamination (one cycle of 50°C for 2 min), an activation step at 95°C for 2 min and then 42 cycles at 95°C for 15 s and 60°C for 1 min.
The genes that were assayed and their PCR primer sequences that were used are noted as follows: 1) KRT13 gene: forward primer, 5′-CAG CAG ATC CAG GGA CTC AT-3′, and reverse primer, 5′-TCT GGC ACT CCA TCT CAC TG-3′; 2) JAG1 gene: forward primer, 5′-ACC TGC CAG TGC CTG AAT G-3′, and reverse primer, 5′-AGG CAA GGT CGA GGG CC-3′; 3) SPINK4 gene: forward primer, 5′-GTC TCC AAC CTG TTC CCA GA-3′, and reverse primer, 5′-AAG CAG AGC TGG CAT TCA TT-3′; 4) SLC7A11 gene: forward primer, 5′-TCC TAA CCA TTC AGT CAG GAA-3′, and reverse primer, 5′-GGC TAT CAA GGT GAC CCA TT-3′; 5) OGG1 gene: forward primer, 5′-AAA TTC CAA GGT GTG CGA CTG-3′, and reverse primer, 5′-GCG ATG TTG TTG TTG GAG GA-3′; and 6) XRCC4 gene: forward primer, 5′-ATG CAG CTC AAG AAC G-3′, and reverse primer, 5′-CTT GGT GCA ATA TCA GTG AC-3′. Quantitative values were calculated based on the threshold cycle (Ct) obtained during the PCR reaction. The ratio of -fold induction of each target gene was calculated using relative quantification method (Pfaffl, 2001) in comparison with housekeeping gene glyceraldehyde-3-phosphate dehydrogenase.
Results
Comparison of Expression Levels between Active Tamoxifen Metabolites. We used gene expression microarrays as a broad test to determine whether there are differences in the activities of two highly active tamoxifen metabolites, 4-OH-Tam and endoxifen. MCF-7 breast cancer cells were cultured in defined E2 conditions (10-10 M E2) and treated with 10-7 M 4-OH-Tam or endoxifen for 24 h before RNA isolation. These conditions were chosen based on our prior work with endoxifen regulation of the progesterone receptor expression (Lim et al., 2005). Using the Affymetrix U133A GeneChip, we detected 11,378 genes as present (absence/presence call is ≥0.75, i.e., detectable in three of four replicate arrays) in cells grown in the presence of E2. Using these detectable genes, we compared the gene expression levels of the two anti-estrogen-treated groups (4-OH-Tam versus endoxifen) in the presence of E2. Genes were considered significantly different if the p value of log-transformed data for the comparison was <0.05. Using this p value, only 127 genes showed a significantly different expression level between endoxifen- and 4-OH-Tam-treated cells. Among the 127 genes, the expression of 65 and 62 was higher in the 4-OH-Tam- and endoxifen-treated groups, respectively. This initial analysis was carried out without using a -fold change cutoff, because we do not know what -fold change in a gene is important to a cell with respect to its antiestrogen effects. In a second analysis, we did use -fold change cutoffs of 1.5- and 2-fold to further characterize the effects of these metabolites on the gene expression patterns. Furthermore, because the probability of finding these many differentially expressed genes due to chance alone with this type of analysis is high, we added -fold change cutoffs and verified the differential expression additional experiments. None of the gene expression differences was greater than 2-fold. Furthermore, there were only five genes that were >1.5-fold higher in the endoxifen compared with 4-OH-Tam and only five genes that were >1.5-fold higher in 4-OH-Tam compared with endoxifen. Because there are many different ways other investigators may wish to analyze and visualize these results, the data from all of the experiments reported here will be freely accessible at the NIH-PharmGKB database (http://www.pharmgkb.org).
Verification of Differentially Expressed Genes. To verify the differential regulation of the genes with >1.5-fold difference between endoxifen and 4-OH-Tam, we chose six of the genes (the three with the greatest -fold difference that were elevated in the endoxifen-treated cells and the three with the greatest -fold difference that were elevated in the 4-OH-Tam cells) and compared their mRNA expression levels by quantitative real time-PCR in samples from three additional replicate cell culture experiments. These cell culture experiments were performed on 3 separate days and were distinct from those used for the microarrays. Because these experiments were designed to verify the changes observed in the microarray experiments, we chose the same doses and time points as those for the microarray studies. To understand the mechanism of the differential regulation of these genes, additional studies, including multiple time points, dose responses, and inclusion of RNA translation inhibitors, will be conducted for each of these genes, but they are beyond the scope of this report. Figure 1A illustrates the data for three of the genes whose expression level was determined by microarray analysis to be greater in 4-OH-Tam- than endoxifen-treated cells. The quantitative PCR data confirmed the differential regulation of the KRT13 gene. The PCR data for the SPINK4 gene expression showed a trend in the same direction as the microarray data, but the difference was not as clear, which reflects the small differences observed by microarray. However, the JAG1 gene expression level appeared to be similar between the two treatments by real-time PCR, and thus the differential expression was not confirmed by the quantitative PCR data. Figure 1B illustrates the data for three of the genes whose expression level was determined by microarray analysis to be greater in endoxifen- than 4-OH-Tam-treated cells. The quantitative PCR data confirmed the differential expression for all three genes (SLC7A11, OGG1, and XRCC4). The expression of SLC7A11 appeared to be up-regulated by endoxifen (relative to E2 alone) and down-regulated by 4-OH-Tam, indicating that there may even be a different direction to their regulation; however, the -fold induction is near to one, which indicates that the differentials expression seemed to be mainly due to the greater suppression by 4-OH-Tam.
To test for potential interactions between 4-OH-Tam and endoxifen, MCF-7 cells cultured in the presence of 10-10 M E2 were treated with a combination of 4-OH-Tam (10-7 M) and endoxifen (10-7 M). The gene expression pattern elicited by the combination was virtually identical to those observed with each metabolite alone, suggesting that there is no interaction (data not shown). In all, these results indicate that endoxifen and 4-OH-Tam have very similar effects on MCF-7 gene expression, although there were a few genes that were confirmed to be differentially regulated by the two tamoxifen metabolites.
Effects of Tamoxifen Metabolites on E2-Regulated Genes. Next, we compared the genes that were altered by 4-OH-Tam and endoxifen relative to E2-treated controls rather than to each other and determined how many of these genes were E2-sensitive. MCF-7 cells were treated with 10-10 M E2 or 0.1% ethanol (vehicle) for 24 h. A total of 11,793 genes were scored as present (absence/presence filtering ≥0.75). We considered only the genes that had a p value of log-transformed data <0.05 (E2 versus vehicle) to be E2-regulated genes.
Because the biologically significant -fold change may vary between genes, and for most of them, we do not know the -fold change that is biologically important, we conducted our analyses with several cutoff values. When we analyzed the data without a minimal -fold change cutoff (i.e., difference between vehicle and E2), the total number of genes that were significantly regulated by E2 was 4062 (Fig. 2A). Of these “E2-sensitive” genes, 1924 were up-regulated and 2138 were down-regulated (the ratio of E2 up- to down-regulated genes, 0.90). When a 2.0-fold change cutoff was used, which includes genes whose -fold changes are ≥2.0 or ≤-2.0, the numbers of E2 up- and down-regulated genes were 110 and 126, respectively (the ratio of up- to down-regulated genes was 0.87). Using several other -fold change cutoffs (1.25-3.0-fold), the ratio of up- to down-regulated genes by E2 was generally comparable (i.e., range was from ∼40 to 60% E2-regulated genes that were up-regulated). A similar pattern of regulation was also observed when other filtering criteria were changed by reducing the absence/presence call to 0.5 or by using more conservative p values (p < 0.01 or p < 0.001; data not shown), suggesting that the observed ratio of E2-inducible and E2-suppressible genes is not dependent on the filtering cutoffs or p values.
We then determined what proportion of the endoxifen- and 4-OH-Tam-regulated genes are within this group of estrogen-regulated genes. The data illustrating this comparison are shown in a Venn diagram in Fig. 2A. We chose not to include a -fold change cutoff restriction in this analysis to allow a more comprehensive comparison and accepted the possibility of having some false-positive genes. The total numbers of genes regulated by 4-OH-Tam or endoxifen were 2444 and 2390, respectively. Of these, 74.4% (n = 1819) of the 4-OH-Tam- and 73.3% (n = 1752) of the endoxifen-altered genes were also determined to be E2-sensitive in the analysis above.
When we compared the 4-OH-Tam-regulated/E2-sensitive genes to the endoxifen-regulated/E2-sensitive genes, there were 454 genes that were affected by 4-OH-Tam but not by endoxifen. Likewise, 387 genes were affected by endoxifen and not by 4-OH-Tam. Of the E2-sensitive genes regulated by the tamoxifen metabolites, there were an additional 1365 genes that were coregulated by both endoxifen and 4-OH-Tam.
We made a similar comparison between the E2-insensitive genes. There were 376 E2-insensitive genes that were regulated by 4-OH-Tam and not by endoxifen. Likewise, there were 389 that were regulated by endoxifen and not by 4-OH-Tam. There were an additional 249 E2-insensitive genes that were coregulated by both 4-OH-Tam and endoxifen.
Whereas this type of analysis is a valuable first approach to the data, it is susceptible to a high rate of false-positive results. Therefore, we determined the correlation between the -fold changes induced by the two metabolites. We correlated the -fold change induced by 4-OH-Tam (in the presence of estradiol) relative to estradiol alone with the -fold change induced by endoxifen (in the presence of estradiol) relative to estradiol alone. Thus, genes that are regulated to the same magnitude by both metabolites would be highly correlated. Furthermore, even genes that are found to be significantly different between endoxifen and 4-OH-Tam could be correlated if they are regulated in the same direction but to different degrees. As shown in Fig. 2B, the -fold changes induced by 4-OH-Tam and endoxifen in the levels of expression of E2-sensitive genes were highly correlated (R2 = 0.99). Furthermore, even in the genes that were classified as specifically regulated by 4-OH-Tam or endoxifen, the -fold changes were highly correlated (R2 = 0.98 for the 4-OH-Tam-specific genes and 0.97 for endoxifen-specific genes). Although there were five genes that appeared to be changed in the opposite direction by the two metabolites (i.e., up-regulated by one metabolite and down-regulated by the other metabolite), the -fold change in the levels of expression of these genes is small. Therefore, many of the genes designated as specifically regulated by one of the tamoxifen metabolites in Fig. 2A may actually have similar regulation patterns. Thus, it seems that the regulation of gene expression by these two potent tamoxifen metabolites is highly correlated.
We also assessed the correlation between effects of the two metabolites on the expression of E2-insensitive genes (Fig. 2C). The -fold changes induced by 4-OH-Tam and endoxifen were highly correlated when the E2-insensitive genes were examined (R2 = 0.99). The majority of genes that were initially determined to be differentially affected by the tamoxifen metabolites (376 genes for 4-OH-Tam and 389 genes for endoxifen; Fig. 2A) were also highly correlated between the two metabolites (R2 = 0.89 for 4-OH-Tam and 0.84 for endoxifen-regulated genes). However, there were a few genes that showed small differences (<2-fold).
To further examine the possibility that these two metabolites had different effects on gene expression, we conducted an additional analysis that was focused only on the genes that had at least 2-fold changes in levels of expression (i.e., -fold changes were ≥2.0 or ≤-2.0 (Fig. 3). Both tamoxifen metabolites induced patterns of gene expression similar to those observed in the unfiltered data, as indicated by 1) the similar ratio of the E2-sensitive to E2-insensitive genes, 2) the proportion of coregulated genes, and 3) the proportion of genes affected specifically by one tamoxifen metabolite and not the other.
In Fig. 4, we present the list of genes induced or suppressed by at least 2-fold by E2 and where these effects were reversed by the treatment with 4-OH-Tam, endoxifen, or the combination of both metabolites. The magnitude of antagonizing effects of 4-OH-Tam was similar to that of endoxifen for the majority of these genes. Within the E2-sensitive genes, there was approximately the same number of genes that were increased as those decreased by the metabolites. However, when we examined the E2-insensitive genes (using a slightly less stringent -fold change cutoff of ≥1.5 or ≤-1.5 due to the lower number of E2-insensitive genes that are regulated by the metabolites), most of the genes (44 of 47 genes) were up-regulated by the metabolites.
In breast cancer, the regulation of the nuclear receptor coactivators and corepressors by E2 and the antiestrogens is of particular interest because they affect the cellular estrogen responses and may contribute to antiestrogen resistance. E2 induced the expression of two corepressors (BRCA1 and BRCA2 genes) and suppressed the expression of three coactivator genes (NCoA1, NCoA3, and NCoA4 genes) and one corepressor gene (NcoR1 gene). These E2-induced changes were counteracted by both tamoxifen metabolites (data not shown). Estrogen regulation of the nuclear receptor coactivators and corepressors has also been shown by others (Thenot et al., 1999; Nephew et al., 2000; Lauritsen et al., 2002; Frasor et al., 2003).
Hierarchical Clustering Analysis. We used unsupervised hierarchical clustering as an additional test to evaluate the variability in gene expression between the 4-OH-Tam- and endoxifen-treated cells (Fig. 5). If the expression patterns were substantially different between the treatments, it would be expected that they would cluster into separate groups. We included the genes that were detectable in at least three of the four replicates within a treatment (absence/presence filter of ≥0.75). Because we knew that the E2-treated group should segregate as a separate group, we first ran the analysis to verify that this group did in fact separate from the others to verify the ability of our algorithm to separate groups within our data set. The E2-treated replicate arrays clearly segregated as a separate group from other drug-treated groups or vehicle group. Three of the four vehicle replicates also segregated together. It is not clear why one of these replicates did not segregate with the others, and we considered rerunning it; however, because the goal of this was to see whether they segregated together, it would not have been appropriate to rerun samples until they did all segregate together.
Next, the most important comparison was to determine whether the 4-OH-Tam and endoxifen groups segregated together. The 4-OH-Tam and endoxifen did not cluster as separate groups, indicating that the two metabolites have broadly similar gene regulation patterns. The cluster containing the two tamoxifen metabolites was segregated from the E2 and vehicle alone-treated groups, suggesting that they have different regulation patterns.
Gene Ontology Analysis. To identify pathways that may be specifically regulated by 4-OH-Tam and endoxifen, we conducted a GO analysis of the array data. We set the p value at p < 0.01 and the required number of differentially regulated genes in each pathway to be at least 5. The results from this analysis are in Table 1. Using these criteria, we identified several GO groups in the genes that were regulated by estradiol and 4-OH-Tam and endoxifen. As expected, many of these were related to cell proliferation and signal transduction. In contrast, there were no groups revealed that were regulated by estradiol and one of the metabolites, but not the other metabolite. This probably indicates their similar mechanism of action. However, there were groups that were regulated by one metabolite but not the other and not estradiol. It was of interest to note that four of the six GO groups that were differentially regulated were transporters.
Discussion
The results of our microarray analysis strongly suggest that the active tamoxifen metabolites, endoxifen and 4-OH-Tam, have almost identical effects on the global pattern of gene expression in human breast cancer cells as: 1) both regulate nearly the same genes in the ERα-positive MCF-7 breast cancer cell line; 2) there is a high correlation in -fold change magnitudes between the two metabolites; 3) hierarchical clustering analysis did not separate endoxifen from 4-OH-Tam; 4) there was a similar ratio of E2-sensitive to E2-insensitive genes between them; and 5) our analysis was able to distinguish compounds that exert different effects on gene expression. These findings are consistent with our previous data that showed that endoxifen and 4-OH-Tam exhibited similar ER binding affinity and suppression of E2-mediated cell proliferation as well as expression of individual E2-sensitive genes in breast cancer cell models (Stearns et al., 2003; Johnson et al., 2004; Lim et al., 2005). Taken together, our data strongly suggest that 4-OH-Tam is not the only active tamoxifen metabolite and that endoxifen is also a potent selective estrogen receptor modulator that induces a similar expression profile.
Our previous in vitro and in vivo studies have unequivocally shown that endoxifen formation is dependent on CYP2D6 activity (Stearns et al., 2003; Desta et al., 2004). The gene that encodes the CYP2D6 protein is highly polymorphic in the general population (over 50 reported allelic variants, many of which result in a loss of enzyme function) with ∼5 to 10% of the population exhibiting the poor metabolizer phenotype (Gaedigk et al., 2002). Therefore, we observed in clinical trials that the plasma concentration of endoxifen is highly variable among breast cancer patients on chronic tamoxifen therapy (20 mg/day orally), but on the average, it was over 6-fold greater than 4-OH-Tam (10-14-fold higher in extensive CYP2D6 metabolizers) (Stearns et al., 2003; Jin et al., 2005). Given its comparable efficacy to 4-OH-Tam and its abundance in plasma, it is reasonable to speculate that endoxifen might contribute more than 4-OH-Tam in mediating tamoxifen activity in vivo. It follows that inherited polymorphisms in the CYP2D6 gene and drugs that influence CYP2D6 enzyme activity may alter individual responses to tamoxifen.
In our current study, ∼74% of the genes regulated by both active tamoxifen metabolites were estrogen-regulated genes, suggesting that the active tamoxifen metabolites do work mainly through the estrogen receptor. The remaining genes that were regulated by endoxifen and 4-OH-Tam were E2-insensitive. These data support the possibility that some effects of tamoxifen metabolites may involve estrogen-independent pathways. Currently, little information is available on the biological role of E2-insensitive genes with regard to the therapeutic efficacy and adverse effects of tamoxifen, thus requiring further studies on the role of these genes. Further studies will be required to determine whether these are direct effects on the expression of these genes or indirect effects mediated by the regulation of other genes or proteins.
Other investigators have reported microarray data on the effects of E2 and 4-OH-Tam (Inoue et al., 2002; Coser et al., 2003; Frasor et al., 2003, 2004; Wang et al., 2004). In general, there were many similarities between our study and others. For example, as observed by Frasor et al. (2004), estrogen regulated gene ontology groups, including those related to the cell cycle, apoptosis, DNA repair, signal transduction, transporters, and nucleotide processing. They also reported a similar number of up-regulated genes (132 compared with our 110) using a 2-fold cutoff. Likewise, Creighton et al. (2006) also observed regulation of gene ontology group related to the cell cycle, DNA replication, metabolism, and development. They also observed a similar number of total genes up-regulated by estradiol (1989 compared with our 1856) when no -fold difference cutoff was applied.
There were also differences between our study and others, which may be due to differences in the details of the studies. For example, when the ratio of E2-up-regulated to E2-down-regulated genes was compared among the related microarray studies that used the same cell lines as those in our studies (time points of expression were not identical between them), there were differences both in magnitude of changes and in the number of genes involved. For example, some microarray studies, which used E2 in doses of 10 nM, reported a higher ratio of down-regulated genes than up-regulated genes (Frasor et al., 2003, 2004), whereas others that used the same E2 doses (10 nM) showed a higher ratio of E2 up-regulated genes (Inoue et al., 2002; Wang et al., 2004). We used a relatively lower dose of E2 (100 pM) and observed approximately equal numbers of E2-induced and E2-suppressed genes regardless of the cutoff values implemented. It seems that the ratio of E2-inducible to E2-suppressible genes varies depending on the dose of E2 (Coser et al., 2003). Likewise, the -fold changes seem to also depend on the doses used. The E2 dose (100 pM) that we used in the present study is within the physiologically relevant range in the serum of premenopausal women and is very close to the range in the serum of postmenopausal women. The dose ratio of 4-OH-Tam to E2 was 10 times lower in the study conducted by Frasor et al. (2004) (4-OH-Tam 10-6 M:E2 10-8 M = 100-fold) than that of ours (4-OH-Tam 10-7 M:E2 10-10 M = 1000-fold), raising the possibility that the higher 4-OH-Tam (in the study of Frasor et al., 2004) could more effectively block the estrogen receptor or could have greater activity on the nongenomic estrogen effects and/or estrogen receptor-independent effects than in our study.
The present microarray study was primarily designed to investigate the differences in gene regulation by two similar tamoxifen metabolites. Using DNA microarray analysis, we could clearly demonstrate that there were similar (e.g., 4-OH-Tam versus endoxifen in the presence of E2) and different (e.g., E2 versus vehicle or E2 versus either Tam metabolite) gene regulation patterns. This provides additional evidence that microarray analysis can be a useful tool to examine drug-specific effects using global gene expression patterns. This application of microarrays has been used by several investigators (Gunther et al., 2003; Frasor et al., 2004). For instance, Frasor et al. (2004) also recently demonstrated the distinct agonistic or antagonistic activities among several SERMs using microarray analysis in breast cancer cells.
In conclusion, the present data, together with our earlier work, demonstrate that the spectrum of pharmacological activity and potency of endoxifen is quite similar to that of 4-OH-Tam. Based on these data and abundance in plasma, we suggest that endoxifen is a major component of the total tamoxifen activity in women taking tamoxifen.
Footnotes
-
This study was supported in part by Pharmacogenetics Research Network Grants 2U-01 GM61373 (to D.A.F.) and K24RR020815 (to D.A.F.), National Institutes of Health Clinical Pharmacology Training Grant 5T32-GM-08425 (to D.A.F.), the Indiana University Center for Medical Genomics Core Facility, a grant of the Korea Health 21 R&D Project, Ministry of Health and Welfare, Republic of Korea (A050174; to Y.C.L.) and a grant from the Basic Research Program of the Korean Science and Engineering Foundation (R01-2006-000-11087-0; to Y.C.L.).
-
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
-
doi:10.1124/jpet.105.100511.
-
ABBREVIATIONS: SERM, selective estrogen receptor modulator; endoxifen, 4-hydroxy-N-desmethyl-tamoxifen, IMEM, improved minimum essential medium; PharmGKB, the Pharmacogenetics and Pharmacogenomics Knowledge Base; E2, estradiol; PCR, polymerase chain reaction; GO, gene ontology; ER, estrogen receptor; 4-OH-Tam, 4-hydroxy-tamoxifen.
- Received December 22, 2005.
- Accepted May 9, 2006.
- The American Society for Pharmacology and Experimental Therapeutics