Aggressive Prostate Cancer Is Prevented in ERαKO Mice and Stimulated in ERβKO TRAMP Mice

Abstract

Previous evidence suggests soy genistein may be protective against prostate cancer, but whether this protection involves an estrogen receptor (ER)-dependent mechanism is unknown. To test the hypothesis that phytoestrogens may act through ERα or ERβ to play a protective role against prostate cancer, we bred transgenic mice lacking functional ERα or ERβ with transgenic adenocarcinoma of mouse prostate (TRAMP) mice. Dietary genistein reduced the incidence of cancer in ER wild-type (WT)/transgenic adenocarcinoma of mouse prostate mice but not in ERα knockout (KO) or ERβKO mice. Cancer incidence was 70% in ERWT mice fed the control diet compared with 47% in ERWT mice fed low-dose genistein (300 mg/kg) and 32% on the high-dose genistein (750 mg/kg). Surprisingly, genistein only affected the well differentiated carcinoma (WDC) incidence but had no effect on poorly differentiated carcinoma (PDC). No dietary effects have been observed in either of the ERKO animals. We observed a very strong genotypic influence on PDC incidence, a protective effect in ERαKO (only 5% developed PDC), compared with 19% in the ERWT, and an increase in the incidence of PDC in ERβKO mice to 41%. Interestingly, immunohistochemical analysis showed ERα expression changing from nonnuclear in WDC to nuclear in PDC, with little change in ERβ location or expression. In conclusion, genistein is able to inhibit WDC in the presence of both ERs, but the effect of estrogen signaling on PDC is dominant over any dietary treatment, suggesting that improved differential targeting of ERα vs. ERβ would result in prevention of advanced prostate cancer.

Estrogen therapy, specifically diethylstilbesterol treatment, has been used in prostate cancer treatment as early as 1941 (1–5). Initially, this was believed to be an indirect inhibition, involving negative feedback on the hypothalamic-pituitary-gonadal axis to shut down androgen synthesis and thereby the growth of prostate tissue. However, the direct effect of estrogens and estrogen receptors (ER) on the prostate has been recognized both in organ development (6) and disease (7).

The potential for phytoestrogens to act as natural selective estrogen receptor modulators and affect prostate disease has been considered. Dietary consumption of soy products has long been associated with reduced incidence of various diseases. Considerable epidemiological evidence supports the observation that soy foods promote health and reduce chronic ailments, including cardiovascular disease and osteoporosis (8, 9). Soy genistein has been implicated as being cancer preventative through epidemiological studies (10, 11). The cancer incidence for breast (12) and prostate (13) cancer is significantly lower in cultures with high soy consumption (14) and correspondingly high serum genistein concentrations. We and others have previously reported an inhibitory growth effect of genistein on prostate cancer cells, LNCaP and PC3 (15, 16), as well as in the transgenic adenocarcinoma of mouse prostate (TRAMP)-C2 cell line (17). Genistein has a multitude of documented mechanisms of action (8, 17, 18). It has been shown to affect protein tyrosine kinase, hedgehog signaling, and estrogen signaling. Genistein binds to both ERs in the low nanomolar range, with a 7-fold greater specificity to ERβ (19).

The ERs are expressed in the prostate epithelium and stroma (20, 21), and neonatal estrogen exposure has profound effects on prostate development both in the rodent (22–26) and man (27). There is evidence that not only can estrogen signaling affect prostate cancer but that ERα and ERβ may have opposing effects in this regard. Work with receptor knockout (KO) mice showed that diethylstilbesterol treatment led to squamous metaplasia of prostatic epithelium in wild-type (WT) and ERβKO mice but not in ERαKO mice (28).

In this article, we investigate to what degree the cancer protective effects of genistein could be attributed to estrogenic activity. By using either ERαKO/TRAMP or ERβKO/TRAMP mice compared with ERWT/TRAMP mice, we demonstrate that both receptors are required for genistein to have an effect in preventing prostate carcinogenesis and that ERα promotes and ERβ prevents aggressive prostate cancer.

Materials and Methods

TRAMP mouse studies

The TRAMP model was developed by placing the simian virus 40 large and small T-antigen genes under the control of the androgen-regulated rat probasin promoter (PBTag transgene), which has been shown to be highly and specifically expressed in the mouse prostate epithelium (29, 30). The expression of simian virus 40 T-antigens disrupts function of p53 and Rb proteins, tumor suppressors important in prostate cancer pathology (31, 32), leading to spontaneous development of prostate tumors by 3 months of age. The biology of the TRAMP model makes it a useful tool to study the progression of the disease, as well as prevention and treatment options.

Male TRAMP mice on a C57BL6/J background were raised in-house as described earlier (17). All University of Missouri institutional guidelines for animal care and use were followed.

To generate the ERα/βKO-TRAMP mice, female C57Bl/6J mice, heterozygous for the ERα or ERβ gene and positive for the PBTag transgene (TRAMP), were crossed with male C57Bl/6J mice that were heterozygous for either the ERα or ERβ gene and negative for the PBTag transgene. All breeder pairs were maintained on a casein-based diet as previously described (33). ERWT, ERαKO, and ERβKO offspring of this breeding scheme that were positive for the PBTag transgene were used in this study.

The mice were fed the casein diet until weaning and then randomly assigned to three groups fed the casein diet or the casein diet, to which 300 or 750 mg of genistein/kg (LC Laboratories, Woburn, MA) had been added. The mice were maintained on these diets from 5 to 6 wk until 5 months of age (18–22 mice per treatment group). The concentration of genistein was selected after analyzing serum from mice consuming a range of diets from 0 to 500 mg genistein/kg diet in 100 mg increments (data not shown). The goal was to provide a concentration of dietary genistein that resulted in serum concentrations lower than 1 μm range. These concentrations are below the threshold for inhibition of tyrosine kinase (34) but are sufficient to saturate both ERs. We also used another estrogenic phytoestrogen in soy, daidzein, which lacks tyrosine kinase inhibitory activity.

At 5 months of age, mice were euthanized and tissues collected. The reproductive tract (testes, vas deferens, urinary bladder, seminal vesicles, and prostate lobes) was removed and weighed. A portion of the dorsal lobe and the ventral prostate were fixed in neutral buffered formalin and paraffin embedded for histological analysis and the remainder snap frozen in liquid nitrogen and stored at −80 C for future studies. Tissues sections were stained with hematoxylin and eosin and examined by light microscopy for assessment of cancer stages (35). All lobes of the prostates were scanned by trained veterinary pathologists, who were unaware of the treatment groups, and the dorsal prostate was staged as either 1) normal, 2) hyperplasia (HYP), 3) prostatic intraepithelial neoplasia (PIN), 4) well differentiated carcinoma (WDC), 5) moderately differentiated carcinoma (MDC), or 6) poorly differentiated carcinoma (PDC), “neuro-endocrine-like carcinoma,” although identification of PDC in any lobe of the prostate prompted a stage of PDC to be assigned to the animal regardless of the status of the dorsal prostate. This protocol follows a procedure described previously (36, 37).

Serum measurements

Serum concentrations of total genistein were analyzed by HPLC-multiple ion-monitoring mass spectrometry (34). Serum testosterone was separated by HPLC before measurement by RIA. Serum estradiol was similarly measured in samples from ERWT and ERαKO female mice fed control or 1 g genistein per kg casein diet for 34 wk.

Immunohistochemistry (IHC)

Slides were prepared using heat-mediated antigen retrieval by immersing and heating in 0.1 m citrate buffer (pH 6.0). IHC reactions were performed using the DakoCytomation Autostainer (Dako, Carpinteria, CA). Primary antibodies were used in the following dilutions: androgen receptor (AR) RG-21 1:50 (Upstate, Billerica, MA), ERα MC-20 1:300 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), and ERβ PPG5/10 1:50 (Serotec, Raleigh, NC). Sixty-six slides (which included 39 ERWT mice, 11 ERαKO mice, and 16 ERβKO mice) were evaluated for AR staining, 49 slides (which included 32 ERWT mice and 17 ERβKO mice) were evaluated for ERα staining, and 45 slides (which included 34 ERWT mice and 11 ERβKO mice) were evaluated for ERα staining. We also looked at 12 B6FVBF1 non-TRAMP mice as a control for ER expression in normal prostate epithelium.

Dorsal prostate was examined, and the percent of immunopositive cells was estimated for 1) normal, 2) HYP, 3) PIN, 4) WDC, 5) MDC, and 6) PDC, using the following scale: 1 = 0% staining, 2 = 1–25% staining, 3 = 25–50% staining, 4 = 50–75% staining, 5 = 75–99% staining, and 6 = 100% staining.

Histological images were captured using DP2-BSW software with a DP72 camera (Olympus Corp., Center Valley, PA) attached to an Eclipse E6000 (Nikon, Tokyo, Japan) microscope.

Statistical analysis

The stages of tumor incidence were classified into noncancer stages (normal, HYP, and PIN) and cancer stages (WDC, MDC, and PDC or neuro-endocrine-like carcinoma). Tumor incidence data were analyzed as a 2 × 2 factorial with genotype (ERαWT or ERαKO) and diet (casein or genistein) as main effects using Fisher's exact test. Values that achieved two-tailed P < 0.05 were considered to be statistically significantly different. Comparison within the same tumor stage among different dietary groups and genotypes has also been done. Body weight, reproductive tract weight, testicular weight, prostate weight, and IHC scores were analyzed using a two-sample t test assuming unequal variance or a one-way ANOVA with Tukey post hoc test. GraphPad Prism4 software was used to perform the analyses (GraphPad, La Jolla, CA).

Results

Genistein's effect on cancer incidence in TRAMP

Genistein significantly reduced overall cancer incidence in ERWT animals compared with control diet in a dose-dependent manner (Table 1 and Supplemental Table 1, published on The Endocrine Society's Journals Online web site at http://endo.endojournals.org). Specifically, PIN incidence increased with genistein treatment, and WDC incidence was dramatically decreased from 50% in the control to 19% with low-dose (300 mg/kg) and 8% with high-dose (750 mg/kg) genistein (Table 1). Although there was a slight increase in PDC incidence in the genistein treatment groups, the change was not statistically significant, nor was an effect of genistein seen on tumor stage in the ERαKO or ERβKO animals.

Table 1.

Effect of diet and genotype on the incidence of prostate cancer in 5-month-old TRAMP mice

Genotype	Diet	n	Tumor stage
			Noncancer			Cancer
			Normal	HYP	PIN	WDC	MDC	PDC
ERWT	Casein	175	2 (1%)	10 (6%)a	41 (23%)c	88 (50%)g	0	34 (20%)k
ERWT	Genistein 300 mg/kg	81	1 (1%)	13 (16%)b	29 (36%)d	15 (19%)h	1 (1%)	22 (27%)k,l
ERWT	Genistein 750 mg/kg	25	0	0a	17 (68%)e	2 (8%)i	0	6 (24%)k,l
ERαKO	Casein	80	0	3 (4%)a	4 (5%)f	68 (85%)j	1 (1%)	4 (5%)m
ERαKO	Genistein 300 mg/kg	25	0	0a	1 (4%)f	23 (92%)j	1 (4%)	0m
ERβKO	Casein	51	0	0a	13 (26%)c,d	18 (35%)g	0	20 (39%)l
ERβKO	Genistein 300 mg/kg	23	0	0a	5 (22%)c,d	8 (35%)g,h	0	10 (43%)l

Incidence of tumor stage were analyzed using Fisher's exact test two-tailed P value. Both doses of genistein had an extremely significant (P < 0.0001) effect on WDC incidence within the ERWT mice but not in ERKO mice. There was also an extremely significant genotypic effect on WDC between ERWT and ERαKO but not quite significant (P = 0.0572) compared with ERβKO mice. There was a very significant (P < 0.005) genotype effect in the observed PDC incidence, between both, ERαKO and ERβKO, compared with ERWT. However, there was no significant dietary effect on PDC incidence. Tumor incidences within columns are being compared with each other; different superscript letters indicate a statistically significant (P < 0.05) difference between groups.

Daidzein is often used as a control for genistein due both to lack of tyrosine kinase inhibition and lower estrogenic activity (38). In separate animals, we also tested daidzein, in ERWT and ERαKO mice, and found a nonsignificant decrease in overall cancer incidence (Supplemental Table 2).

ER status and cancer incidence

Approximately 70% of TRAMP mice WT for both ER developed cancer by 5 months of age (Supplemental Table 1), with WDC incidence of 50% and PDC incidence of 19% (Table 1). ERαKO TRAMP mice had higher WDC incidence at 85% but reduced PDC at 5%. ERβKO TRAMP mice had moderate levels of PDC incidence at 39%, indicating a possible protective role for ERβ in tumorigenesis, or a tumorigenic role of ERα, or a combination of both (Table 1).

ERα, ERβ and AR profile changes with cancer progression

To have a better understanding of the involvement of ER in cancer progression, we performed IHC analyses of hormone receptor expression in the prostate from TRAMP mice. IHC analysis for ERα, ERβ, and AR was performed on a total of 66 representative mice from the above mentioned studies.

The ERα IHC score in PDC epithelium was significantly reduced compared with other stages of prostate adenocarcinoma (Fig. 1). Just as striking was the change in distribution of ERα, from a nonnuclear location in the more benign cells to an essentially exclusive nuclear presence in PDC lesions. No significant difference in the percent of ERβ positive epithelial cells was observed, although there was a predilection for nuclear staining in normal/HYP/PIN/WDC/MDC lesions and predominately cytoplasmic staining in PDC (Fig. 2). In addition, although there was no change in low stromal ERα expression across different stages of prostate cancer, the expression of stromal ERβ was significantly lower in normal prostate (see figure 4 below). Neither diet nor genotype affected stromal expression of ERα or ERβ (Supplemental Fig. 1).

Fig. 1.

ERα immunostaining in TRAMP mice. Mice were euthanized at 5 months, and prostates were analyzed for ERα immunostaining. Normal prostates from non-TRAMP mice were stained as control. A, PIN lesion. B, WDC lesion. C, PDC lesion. D, Plot of IHC scores. n = 45, from studies 1 and 5–7 (see Supplemental Table 2). ERα expression decreases with cancer progression. In PDC tumors, even though quantitatively lower, the expression shifts from mainly nonnuclear to nuclear. Scale bar, 20 μm. Inset in A shows negative control with specific blocking peptide. *, P < 0.05.

Fig. 2.

ERβ immunostaining in TRAMP mice. Mice were euthanized at 5 months, and prostates were analyzed for ERβ immunostaining. Normal prostates from non-TRAMP mice were stained as control. A, PIN lesion. B, WDC lesion. C, PDC lesion. D, Plot of IHC scores. n = 49, from studies 1, 2, 6, and 7 (see Supplemental Table 2). ERβ expression did not significantly change with tumor progression. Scale bar, 20 μm. Inset in A shows negative control with specific blocking peptide.

AR expression in the epithelium was stable across HYP, PIN, WDC, and MDC and comparable with normal prostates from non-TRAMP mice. A significant decrease in AR was observed in PDC samples (Fig. 3), which correlates with the loss of responsiveness to androgen. In addition, AR decreased in the stroma of PDC compared with other stages (Fig. 4C).

Fig. 3.

AR immunostaining in TRAMP mice. Mice were euthanized at 5 months, and prostates were analyzed for AR immunostaining. Normal prostates from non-TRAMP mice were stained as control. A, PIN lesion. B, WDC lesion. C, PDC lesion. D, Plot of IHC scores. n = 49, from studies 1 and 5–7 (see Supplemental Table 2). There is a very significant reduction of AR expression in PDC tumors. Scale bar, 20 μm. Inset in A shows negative control without antibody. *, P < 0.05.

Fig. 4.

Stromal immunostaining in TRAMP mice. Mice were euthanized at 5 months, and prostate stroma was analyzed for AR (A), ERα (B), and ERβ (C) immunostaining. Relative immunostaining was noted for prostates with a weighted lesion stage. Normal prostates from non-TRAMP mice were stained as control. Error bars show sd. Statistical significance (P < 0.05) between different stages is noted with differing letters.

No significant difference in AR expression due to genotype or ER expression was found among the diet groups (Supplemental Figs. 1 and 2).

Weight differences

WT and ERαKO mice fed the high genistein diet had increased total body weight compared with those fed casein (Fig. 5).

Fig. 5.

Mouse body and prostate weights from all studies combined. A, ERWT mice fed the high genistein diet and ERαKO mice fed the low genistein diet had significantly increased total body weight compared with those fed the casein diet. When combined as in B, there are no significant differences in prostate weights. When we applied a cut-off of 1.5 g for very large prostate, which were exclusively PDC, we obtained a more differentiated pattern for prostate weight above 1.5 g (C) and under 1.5 g (D).

ERαKO mice had significantly lower testes weights, which corresponds with previous reports (39), but no difference between ERWT and ERβKO mice (Supplemental Fig. 3).

Total prostate weights did not vary between groups due to significant variance in tumor size, specifically within the PDC tumors (Fig. 5B and Supplemental Table 3). To clarify the data, we used a cut-off weight of 1.5 g to define two groups. The separation by weight was used to determine whether there were differences between groups within the same cancer stage. Prostates over 1.5 g, which were exclusively PDC, were most highly represented in the ERWT casein group (8/175) and in the ERWT low genistein group (7/81). The largest prostates were found in the ERWT low genistein group, followed by ERWT casein, and ERWT high genistein (1/25) and ERβKO genistein (3/23). It is interesting to note that even though ERβKO mice had a significantly higher PDC incidence relative to WT mice, their prostate weights were not higher. ERαKO mice had only one prostate above 1.5 g corresponding to the near absence of PDC in ERαKO (n = 105) (Fig. 5C). Within prostates under 1.5 g, ERαKO mice had significant higher weights compared with the other genotypes (Fig. 5D), corresponding to the increased WDC incidence in ERαKO mice (Table 1).

Serum hormone concentrations

Serum testosterone concentrations were measured, and no significant differences in serum testosterone were found due to genotype (in αKO casein vs. αKO genistein, 13,540 ± 11,090 vs. 9570 ± 3120 pg/ml) or diet (WT casein vs. WT genistein, 7670 ± 3440 vs. 16,440 ± 8350 pg/ml). Serum was also analyzed for total genistein concentration in a subset of mice (n = 11 for each treatment group). Serum concentrations for ERαWT and ERαKO mice were not significantly different (0.44 ± 0.11 and 0.64 ± 0.15 μm, respectively). Genistein in the diet was associated with a decrease in serum estradiol, although the effect was not statistically significant due to small number of animals tested (Supplemental Fig. 4).

Concentrations of genistein in the serum of mice not fed genistein were not detected.

Discussion

We present here the cumulative results from seven studies with a total of 460 ERWT, ERα, and ERβKO TRAMP mice (Supplemental Table 2). This work found opposing roles for the two ERs in carcinogenesis (Table 1). ERαKO animals had almost no PDC incidence (5%), whereas ERβKO mice had approximately double the PDC incidence (39%) compared with WT mice (19%). This is in agreement with previous suggestions that ERα increases and ERβ decreases prostate cancer risk (28). Recently, a polymorphism in ERα has been correlated with increased prostate cancer incidence in a north Indian population (40), supporting the idea of ERα being the “bad” ER.

It is likely that the genotypic effect of ERKO that we observed is dominant over any dietary treatment. In our many dietary studies in TRAMP mice, we found no dietary effect on PDC in ERαKO animals compared with animals WT for both receptors. This work supports the conclusion that genistein is protective against prostate tumorigenesis (Table 1), although contrary to a previous report, it did not affect PDC incidence (41). Although the percent of mice with PDC was slightly greater in the genistein group, the change is not statistically significant. Furthermore, if we look at the individual studies (Supplemental Table 2), we can see that there are several instances where the incidence of PDC is lower in the genistein treatment groups. The higher incidence of PIN in mice fed genistein may indicate that genistein may delay progression of early prostate cancer. This may imply that genistein would be more appropriate as a preventative or perhaps as a therapy in early prostate cancer.

To exclude the developmental effect of mice lacking ERα or ERβ, studies are planned to repeat this work with conditional KO animals whose ER are turned off at puberty. An alternative approach would be to perform the study with selective ERα and ERβ ligands to recreate the KO results. Examples of such compounds include the such compounds the ERβ-specific agonist 2,3-bis (4-hydroxyphenyl) propionitrile (DPN), to mimic an ERαKO, and the ERα-specific agonist 4,4′,4″-(propyl-[(1)H]-pyrazole-1,3,5-triyl) trisphenol (PPT), to mimic an ERβKO, or the methoxychlor metabolite 2,2-bis(p-hydroxyphenyl)-1,1,1-trichloroethane (HPTE), which is an ERα agonist, ERβ antagonist, and AR antagonist (42), which would silence any confounding effects through AR. The difficulty with using these compounds is their current high cost and that their selectivity for the specific receptor is between 70- and 300-fold for 2,3-bis (4-hydroxyphenyl) propionitrile (DPN) (43, 44), 400-fold for 4,4′,4″-(propyl-[(1)H]-pyrazole-1,3,5-triyl) trisphenol (PPT) (45), and 20-fold for 2,2-bis(p-hydroxyphenyl)-1,1,1-trichloroethane (HPTE) (46), which will require careful titrating and monitoring of the uptake of the ligands in the prostate.

A surprising finding of the current study was that genistein does not act solely through ERβ. Because genistein had no effect on total cancer incidence in either ERαKO or ERβKO animals, it may be that genistein requires both ERs to exert its protective action. Alternatively, genistein could be acting via a non-ER pathway. This effect is likely not via genistein's action on tyrosine kinases, because the genistein serum concentrations achieved in the mice were 10-fold lower than the reported Ki (inhibitory binding affinity) of genistein for tyrosine kinases (34). The phytoestrogen, daidzein, had no significant effect on cancer incidence. Daidzein has a 10-fold lower binding affinity to mouse uterine cytosol than does genistein (47) and, in our study, was able to decrease WDC incidence, although to a smaller degree than did genistein. This implies that the conformation of ER when bound to daidzein is different than when bound to genistein, resulting perhaps in altered interactions with necessary coregulatory proteins. Regardless, genistein's effects are nullified or overshadowed by the stronger effects of both ERs.

We observed that a diet containing a high dose of genistein significantly increased total body weights in the WT and ERαKO mice (Fig. 5 and Supplemental Table 3). This is consistent with previous reports that genistein contributes to increased lean body mass in neutered cats (48) and food intake and total body mass in quail (49).

Comparing prostate weights by tumor stage among the groups, we observed an interesting phenomenon. Besides having an increased WDC incidence (Table 1), which accounts for the overall increase in cancer in the ERαKO group, when looking at prostates under 1.5 g, the ERαKO mice had significantly higher prostate weights compared with both ERWT and ERβKO mice (Fig. 5D). Interestingly, although ERαKO mice had a more rapidly growing cancer, they did not acquire the PDC stage within the time frame of the study. Additionally, aged ERαKO mice did not develop PDC even after 8 months (data not shown).

ERβKO/TRAMP mice had fewer tumors over 1.5 g than ERWT, even though they had twice the PDC incidence (Fig. 5). It would be interesting to determine whether the PDC tumors differed in other aspects, such as expression of neuroendocrine or proliferation markers, or had a different metastatic potential. We plan to examine this question using allograft of PDC tumors from different ER genotypes implanted in young ERWT TRAMP mice. Differing biological conditions in the WDC and PDC tumors might be possible if they arose from different types of cells within the prostate epithelial cell population. This idea has been proposed by others (50) and is supported by our observations. The difference in cell origin for WDC and PDC may be part of the reason why there was an increased incidence of WDC in the ERαKO groups, opposite of what was seen with PDC.

In addition, the PIN and WDC are likely more androgen responsive than is PDC, given the difference in AR expression (Fig. 3D). The ERαKO mice had slightly higher serum testosterone levels than the WT mice, and an increased androgen signaling in this group may have favored WDC growth, although the differences in testosterone concentrations were not statistically significant. Mentor-Marcel et al. (41) reported in WT/TRAMP mice a dose-dependent effect of genistein treatment at stage 6 (PDC) but not at any lower stages. An explanation for why this differed from our study may have been the different genetic background of the TRAMP mice and duration (7-month vs. our 5-month end date) of genistein administration.

The IHC analysis of the tumors is in agreement with the hypothesis of the importance of ERα and ERβ in cancer progression. The noncancerous, hyperplastic prostate expresses both ERs, with ERα being predominately nonnuclear. It also expresses AR, corresponding with the initial responsiveness to androgens in the prostate. In PDC tumors, however, the expression pattern changes to low but mostly nuclear expression of ERα. The role of ERα in the developing prostate and its association with metaplasia of prostate epithelium may fit with its shift in localization from nonnuclear in the benign adult prostate epithelium to a nuclear location (51). This may indicate a change in function of ERα in PDC, which could be consistent with a dedifferentiated cell line having ER characteristics similar to embryonic prostate tissue (7).

Nonnuclear signaling of ER has been reported and is recognized as a means of rapid estrogenic response in tissues (52–54). The ER has been identified tethered to lipid rafts at the cell membrane, where it can interact with components of other signaling pathways and affect, and potentially be affected by, these other pathways (55). In addition to being localized in the nucleus or cell membrane, ERs have also been found in the mitochondria (56), where they may affect energy metabolism or apoptosis. One group saw ERα in the cytoplasm and the nucleus of normal human prostate stem/progenitor cells (57). This group also noticed that these stem/progenitor cells have a nuclear receptor expression profile more similar to androgen-independent cell lines, such as PC3, than the androgen-responsive cell line, LNCaP.

The hypothesis of protective ERβ and tumorigenic ERα (58, 59) is supported by our data. The associations of PDC incidence in the ERβKO and ERαKO mice may be due to an alteration of estrogen signaling from membrane or cytoplasmic ERα (and conversely, nuclear ERβ) in PIN and WDC to nuclear (and cytoplasmic) in PDC, which may result not only in changes in the expression of expression of estrogen response element-containing genes, but also in G protein and other membrane and cytoplasmic cellular signals that may affect cell growth, hormone response, and cell adhesion (60, 61). We thus speculate that the nonnuclear localized receptors in TRAMP are not necessarily inactive; they may in fact be interacting with other pathways.

We did not observe any changes in ERα, ERβ, or AR expression associated with genistein diets, nor with genotype (Supplemental Fig. 2). We did, however, see a significant decrease in AR at the PDC (Fig. 3D). This decrease in AR expression, along with a loss of androgen responsiveness, has been seen by others in the TRAMP mice (62, 63). Considering that both AR and ERα change so dramatically in PDC, we attribute the effects of genistein on PDC to ERα rather than AR. The affinity constant of genistein for AR is 400 μm (64), and the average serum genistein concentration that we measured was four orders of magnitude below this, which is still well above genistein's binding affinity for ERα (64). Therefore, it is unlikely that genistein is acting directly through AR in these studies. In a study of Japanese men, testosterone concentrations were inversely correlated with soy intake (65), possibly indicating indirect action via negative feedback on the hypothylamic-pituitary axis, or genistein inhibition of aromatase activity (66), but we did not see genistein affecting serum testosterone in our mice. Serum estradiol tended to be higher in the ERαKO compared with the ERWT mice and was slightly decreased in both groups by feeding genistein (Supplemental Fig. 4). However, due to small amounts of serum available, estradiol concentrations were measured in only a few animals. As reported previously, elevated estrogen concentrations in the ERαKO mice are presumed to occur due to the absence of negative feedback from the ERs (67). It is possible that a genistein-bound ERα or ERβ may affect androgen signaling. Interestingly, genistein decreased AR protein concentration and PSA expression in LNCaP cells (which naturally lack ERα) (68). This effect was relieved by cotreatment with ICI, an ER antagonist. And because LNCaP cells lack ERα, it is possible that genistein's effect on AR in the aforementioned example was via ERβ. There have been other accounts of genistein modulating AR expression. Wang et al. (69) reported genistein causing a decrease in ER and increase in AR expression. Another lab had earlier shown genistein treatment of LNCaP cells resulted in decreased AR mRNA with a concurrent increase in AR protein (70). And yet others have seen genistein have no affect on AR concentration (71).

The importance of stroma in hormonal responsiveness of prostate epithelium has been shown by several groups (71–73). In one striking example, proliferation of prostate epithelium in response to varying dihydrotestosterone to estradiol ratios was seen when the cells were cocultured with prostate stromal cells, but this growth response was lost when the epithelial cells were grown without stromal cells (71). We saw low expression of ERα in the stroma, which did not change between tumor stages. However, we did see lower ERβ in normal prostate stroma and a decrease of AR expression in PDC stroma, when comparing different stages with each other. Others have noted a decrease in stromal AR in association with advanced prostate cancer, although it is not known whether a low AR stroma is a more favorable environment for PDC tumors or if the PDC is silencing the expression of AR in the surrounding stroma. Microarray analysis of normal and reactive prostate stroma showed differential expression of many genes (73), including Gli2, a key mediator of hedgehog signaling, which has been implicated in advanced prostate cancer (74–77). Potentially, these changes in ERβ and AR expression in the stroma of the TRAMP prostate are important in the development of prostate cancer.

As mentioned above, the fact that genistein treatment was associated with a decrease in cancer incidence in the WT mice, but not in either of the KO groups, suggests the possibility that genistein requires both ER present to enact its protective action. This is supported by observations that knocking down ERα and/or ERβ abrogated the effects of genistein in vitro (78–80). The IHC results are consistent with the idea that a heterodimer may be important in the role that ER plays in prostate cancer. If a heterodimer is protective, then the localization of ERα to the nucleus would leave ERβ to form a homodimer in other areas of the cell. Levin and co-workers (81) previously found the ER heterodimer to be present and active in the membrane of breast cancer and endothelial cells.

Another player in the prostate in addition to estradiol is the dihydrotestosterone metabolite, 5-androstene-3β, 17β-diol, which is able to bind to both ERs in vitro (19), with affinities of 6 nm for ERα and 2 nm for ERβ. However, this metabolite was reported to preferentially activate ERβ (82). We speculate that the protective effects seen in ERαKO mice against PDC are mediated through 5-androstene-3β, 17β-diol via ERβ. In support of this idea, Dondi et al. (83) recently demonstrated that treatment with 5-androstene-3β, 17β-diol decreased proliferation of DU145 and PC3 cells in vitro and PC3 xenografts in vivo. Previously, the Poletti lab had shown that 5-androstene-3β, 17β-diol inhibited the migratory effects of DU145 (negative for ERα and positive for ERβ) and that this inhibitory effect could be counteracted with ICI or an ERβ-selective antagonist (84). In a mouse model, 5-androstene-3β, 17β-diol treatment resulted in decreased proliferation of prostate epithelium (82). However, the effect of the compound was not evident in ERβKO mice. This inhibitory effect on prostate growth by a natural ERβ ligand is one way in which removal of the “good” receptor in the ERβKO mice leads to a doubling of PDC incidence and is an additional reason to consider ERs as potential targets for human prostate cancer therapy.

In summary, we propose that ERβ has a protective role in prostate tumorigenesis, and/or ERα has a tumorigenic influence. This report offers a unique insight into the effects of ER signaling on prostate carcinogenesis. Many questions remain to be answered. However, the comprehensive data demonstrate clearly the dualistic involvement of ERα and ERβ in the aggressive PDC phenotype and suggests that targeting both ERα and ERβ would be useful in preventing and treating prostate cancer.