Tumor Repressive Functions of Estrogen Receptor β in SW480 Colon Cancer Cells

The majority of sporadic colon cancers are believed to develop through a gradual process of mutations including inactivation of the APC gene. This molecular process corresponds to development of polyps and subsequently adenomas of the colonic epithelium (1). Randomized, placebo-controlled clinical trials, designed to investigate the risk of hormone replacement therapy, indicates that progestin-estrogen hormone replacement therapy reduces the incidence of colorectal carcinoma (2, 3). Interestingly, there are also indications that use of oral contraceptives is associated with a lower incidence of colorectal cancer (4). Several possible explanations have been presented. Estrogens affect the metabolism of bile acids, reduce insulin-like growth factor (IGF-I) levels in the bloodstream, and have direct effects on the colonic mucosa (5–7). In the human colonic epithelium, estrogen receptor β (ERβ) is the predominant ER (8, 9). The ERs mediate their effects by binding to estrogen response elements (ERE) in gene promoters but also indirectly through activator protein (AP-1) and SP-1 sites (10).

A striking reduction of ERβ expression occurs in colon cancer compared with normal colonic epithelium, and furthermore, the level of ERβ seems to be inversely correlated to differentiation grade and Dukes score (11, 12). This raises the question whether ERs in the colonic epithelium could have a function as tumor suppressors in early colonic neoplasms (13). To investigate the function of ERβ in colon cancer, we have generated ERβ-expressing SW480 colon cancer cells by lentiviral transduction of an ERβ expression cassette. Also, we have used an ERβ with mutated DNA-binding domain (DBD-mutation) to investigate if effects of ERβ on gene transcription are dependent on ERE-binding or not. The transformation of normal colonic epithelial cells to colon cancer involves a sequential alteration in gene expression profile, involving induced expression of several cell cycle genes (1, 14). The cell cycle is a central pathway in carcinogenesis and development of a malignant disease. c-Myc is overexpressed in a majority of colon cancers and is believed to be an important factor in the transformation process (15, 16). c-Myc is one of the first genes to be expressed in the cell cycle, in turn regulating the expression of secondary cell cycle genes. However, there are other more or less c-Myc–independent cell cycle factors of importance in colon cancer growth and metastasis. Recently, several studies have shown a correlation of decreased p27^Kip1 protein levels (p27) with poor overall survival compared with patients with high levels of p27 in their tumor (17). The main regulator of p27 protein level is the F-box protein p45^Skp2, which targets the protein for degradation through the proteasome complex. In addition, a correlation of increasing p45^Skp2 levels (p45) with increased number of lymph node metastases has been found (18). Interestingly, both c-Myc and p27 are regulated by ERα in breast cancer cells and we have earlier shown that ERβ induces p27 in breast cancer cells (19). Consequently, we are now investigating the effects of ERβ on the c-Myc and p27 pathways in colon cancer cells.

Cell cultures and generation of stably ERβ-expressing cells. Because colon cancer cell lines do not contain endogenous ERβ, SW480, HCT-116, and HT29 colon cancer cells were lentivirus-transduced with ERβ-construct (SW480-ERβ), (HCT-116-ERβ), and (HT29-ERβ) or with empty vector as control (SW480-Control), (HCT-116-control), and (HT29-control). SW480 cells were also transduced with ERβ-construct with mutated DBD-domain (SW480-ERβ-DBDmut) to investigate the importance of ERE binding in ERβ signaling. Both ERβ constructs were expressed at similar levels in SW480 cells. The plasmid pcDNA3-FLAG ERβ was used as a template for TOPO cloning into pLenti6/V5-D-TOPO according to the manufacturer's instructions (Invitrogen). The pLenti6/V5-D-FLAG ERβ DBDmut was constructed from pLenti6/V5-D-FLAG ERβ using Quick-Change (Stratagene) according to the manufacturer's instructions. Forward oligo ATCACTATGGAG TCTGGTCGTTG CA GCATGTAAGGCCTTTTTTAAAA GA changing amino acid residues E167 and G168, glutamic acid and glycine, respectively, to alanine. Lentivirus was then produced with the ViraPower Lentivirus expression system and infection was performed as described earlier (20). The lentivirus-treated cells were then incubated with 5 μg/mL Blasticidine, resulting in death of nontransduced cells. The corresponding ERα DBD mutant has been characterized earlier. The ERβ DBD mutant was further characterized in an ERE luciferase activation assay where wild-type ERβ, but not the DBD-mutated ERβ, activated ERE-luciferase transcription. The lentivirus transduction was made in independent triplicates for each construct (ERβ, ERβ-DBDmut, and control). The cells were cultured in RPMI 1640 (Life Technologies/Invitrogen) with 5% fetal bovine serum (FBS). Medium was changed to phenol red–free RPMI 1640 with 5% dextran/charcoal-treated FBS (stripped medium) 24 h before start of experiment. The cells were then incubated with 10 nmol/L E₂ or ethanol as control for 24 h in stripped medium.

Proliferation assay. To investigate the ERβ-mediated effects on cellular proliferation, 2 × 10⁵ SW480, HCT-116, and HT29 cells, respectively, were virally transduced with either ERβ-construct or empty control vector and were plated onto 6-cm cell culture dishes in RPMI 1640. The plating efficiency was similar in the different plates. After 72 h, the cells were counted manually with a Bürkner chamber. Experiments were performed with three different clones, each in three separate dishes.

Flow cytometry. Flow cytometry was performed to observe different cell cycle distributions of the SW480 clones. Cells were treated with mevinolin at 40 μmol/L (Sigma) for 36 h before the start of the experiment to ensure cell synchronization. Cells were then released into media containing 5% serum, harvested at 0, 16, 24, and 36 h by trypsinization and fixed in 70% cold ethanol for 30 min on ice. The fixed cells were washed in PBS and stained with 50 μg/mL propidium iodide (Sigma) supplemented with RNase A (50 μg/mL; Sigma) for 30 min at 37°C. Fluorescence-activated cell sorting (FACS) was performed using a FACS Calibur flow cytometer (Becton Dickinson). Cell cycle analysis was performed using CELLQuest program (Becton Dickinson).

Western immunoblotting and antibodies. SDS-PAGE was performed as described earlier (21), with the primary intention to investigate regulation of G₁-phase cell-cycle factors. The following primary antibodies were used for incubations: against ERβ(Abcam), β-actin (Sigma), c-Myc sc-764, Cyclin A sc-751, Cyclin E sc-247, p27^Kip1 sc-1641, p21^(Waf1/Cip1) sc-19, p45^Skp2 sc-7164, p15 sc-612, p16 sc-468, and p14^ARF sc-8613 (all from Santa Cruz Biotechnology); pRb-antibody sampler kit (Cell Signaling); and Cdc25A Ab-3 (from NeoMarkers).

RNA-extraction and real-time PCR. RNA extraction with TRIzol (Invitrogen) was done according to standard protocol. RNA was purified with Qiagen RNeasy spin columns (Qiagen) with on-column DNase I digestion (for subsequent Microarrray-experiment). cDNA was synthesized with NBS First Strand System (Nordic Bioservice).

Real-time PCR was performed with NBS Probe or NBS SYBR Mastermixes (Nordic Bioservice). Primer/probe sequences will be provided on request. The real-time PCR reactions were performed in an ABI PRISM 7500 (Applied Biosystems) under the following conditions: 10 min hot start at 95°C, followed by 15 s at 95°C, and 60°C for 50 s in totally 40 cycles. The optimum concentration of primers was determined in preliminary experiments and all SYBR-Green primer pairs were checked with melting curve analysis.

Immunohistochemistry. To examine the number of proliferating cells in ± ERβ-expressing tumor xenografts, immunohistochemistry with a Ki-67 antibody was performed. Antigen retrieval was done by microwave boiling in 0.01 mol/L citric acid (pH 6.0) for 120 s and then slides were left undisturbed for 20 min. Endogenous peroxidase activity was blocked by incubating sections shaking in 3% hydrogen peroxide in methanol for 10 min and then blocked in 2% Bovine Serum Albumin (BSA) and 0.1% NP40 for 1 h at room temperature. Primary antibody Ki-67 (DakoCytomation) was diluted 1:100 in 2% BSA and 0.1% NP40 and incubated overnight at 4°C. Sections were then washed consecutively in PBS, 0.1% NP40 for 3 × 5 min, followed by incubation with a biotinylated antibody BA-9200 1:200 in 0.1% NP40 for 1 h (Vector Laboratories). After this, sections were incubated in the streptavidin-horseradish peroxidase ABC complex (Vectastain Elite; Vector Laboratories) for 1 h, stained in 3,3′-diaminobenzidine and counterstained with Mayer hematoxylin (Sigma) before dehydration through ethanol, and mounted in Pertex (Histolab). The number of Ki-67–positive cells was counted on two independent fields per slide with ×20 objective. The average number of positive cells and SD were calculated for each group.

Experimental animals and xenograft model. SW480 colon cancer cells ± ERβ expression were implanted into immunodeficient mice to investigate tumor growth. SW480-ERβ/control cells (5 × 10⁶) were diluted in 200 μL normal medium + 200 μL Matrigel (BD Biosciences) and injected into the abdominal fat close to the mammary tissue of 9- to 12-wk-old pathogen-free severe combined immunodeficient (SCID)/beige mice (Taconic) on day 0. E₂ pellets, 0.72 mg/pellet (IRA), were inserted s.c. into the neck with a pellet trochar (IRA). After 14 d, the mice were sacrificed and the tumor weight was measured. All tumors were fixed in 4% paraformaldehyde and stored in 75% ethanol. After this, tissues were paraffin embedded and subsequently sliced into 4.5-μm sections according to standard protocol.

Statistics. Values are expressed as means with 95% confidence intervals. For real-time PCR analysis, unpaired two-tailed t test was used to compare differences between two parallel treatment groups of identical origin. Mann-Whitney U test was used to test differences between two groups of nonparametric data. Significance is presented as P value of <0.05 (*), <0.01 (**), and <0.001 (***); nonsignificant differences are presented as NS.

Colon cancer epithelial cells do not contain ERα but varying amounts of ERβ. The SW480 as well as most other colon cancer cell lines grown in vitro contain very low levels of ERβ. To confirm this, we performed real-time PCR with primers/probe specific for ERα and ERβ mRNA, respectively. As expected, we found no ERα and very low levels of ERβ expression in SW480 cells (data not shown). To be able to study the effect of ERβ in these cells, we used lentiviral transduction of an ERβ expression cassette. In Fig. 1, we show the expression of ERβ mRNA and protein in these cells (A). The level of ERβ was not affected by E₂ treatment. Expression of exogenous ERβ decreased the proliferation compared with SW480 cells transduced with empty expression cassette (SW480-control) as shown in Fig. 1B. As a comparison, two other colon cancer cell lines, HCT-116 and HT29, were transduced with an ERβ expression cassette (HCT-116-ERβ and HT29-ERβ). The proliferation of HCT-116 cells was reduced in presence of ERβ expression. However, the proliferation of HT29 cells was not reduced to a significant level (Fig. 1C). We then performed flow cytometry to investigate the cell cycle distributions of the SW480-clones. Thirty-six hours after start of experiment, 75% of SW480-ERβ cells were distributed in the G₁ phase, whereas 52% of SW480-control cells were in the G₁ phase (Fig. 1D). Mutation of the DNA-binding domain of ERβ (SW480-ERβ-DBDmut) resulted in a nearly complete removal of the ERβ-mediated G₁ arrest, resulting in 58% of the cells distributed in the G₁ phase. In agreement with this, expression of ERβ-DBDmut did not reduce the proliferation of SW480 cells as measured with proliferation assay (data not shown).

We went on to investigate the regulation of the most important genes governing proliferation in the G₁-S phase. Western blot was performed and the results are shown in Fig. 2A. The immediate early gene c-Myc was down-regulated by ERβ expression and, as expected, the c-Myc target gene Cdc25A was also repressed by ERβ. p14^ARF, a tumor suppressor involved in both cell-cycle arrest and apoptosis, was induced by ERβ. p14^ARF functions by down-regulating MDM2 and, in this way, stabilizing p53-protein. As a consequence, we found p53 to be stronglyinduced by ERβ.

The cdk4/6-inhibitors p15 and p16 were not affected by ERβ at the protein level.Cyclin E and A, important in the late stage of the G₁ phase, were not affected significantly, whereas Cyclin D1 was up-regulated at the protein level. The cdk2-inhibitors p21^(Waf1/Cip1) (p21) and p27 showed strong induction following ERβ expression and the expression of p45, which is the regulator of p27-degradation, was down-regulated by ERβ. Forced expression of Retinoblastoma protein (pRb) efficiently blocks cell cycle transition and pRb is often mutated or inactivated in tumorigenesis resulting in free E2F1 (22). Interestingly, we found that total amount of pRb was increased in the ERβ-expressing cells compared with the control. Phosphorylation status of pRb was changed as well, resulting in increased phospho-Ser780 and decreased phospho-Ser795 following ERβ expression.

However, SW480 cells expressing ERβ-DBDmut showed a completely different protein expression pattern (Fig. 2B). The DBD-domain is responsible for binding of ERβ to EREs. Therefore, the DBD-mutant is crucial to investigate potential ERE-dependent activation/inactivation by ERβ. Cdc25A was induced by ERβ-DBDmut, whereas p21 and p27 were not induced to the same extent as by wild-type ERβ. Furthermore, c-Myc was not repressed to as large extent as in the SW480-ERβ cells.

In HT29 cells (Fig. 2C), c-Myc, p21, p27, and p53 were all regulated by ERβ expression in a similar fashion as in SW480 cells. However, ERβ expression in HCT-116 cells caused a reduction of p53 and p21 levels but also of c-Myc, whereas p27 was not significantly changed.

To investigate the regulation of these proteins at the mRNA level, we performed real-time PCR (Fig. 3) on SW480-ERβ, SW480-control, or SW480-ERβ-DBDmut. We found that c-Myc mRNA was down-regulated >70% by ERβ expression independently of E₂, and this effect was dependent on a functional DBD (Fig. 3A). Cdc25A mRNA expression was also reduced by ERβ, but not by the ERβ-DBDmut (Fig. 3C). There were no significant differences in mRNA expression of Cyclin E or Cyclin A by ERβ (Fig. 3A), whereas Cyclin A expression was reduced by ERβ-DBDmut. Cyclin D1 mRNA expression was up-regulated in SW480-ERβ in a DBD-dependent fashion. p21-mRNA was up-regulated in SW480-ERβ independent of the DBD, indicating that the regulation of p21 by ERβ occurs independently of ERE-binding. p27 mRNA was not regulated by ERβ (Fig. 3B) but was induced by ERβ–DBDmut; however, this induction was nonsignificant (P = 0.32, bars 2 versus 6). There was a strong reduction of p45-mRNA in SW480-ERβ cells dependent on a functional DBD. As expected, pRb was not regulated at the mRNA-level by ERβ but was strongly repressed by the ERβ-DBDmut, an effect which we cannot explain at present (Fig. 3C). p14^ARF and p53 were not affected significantly at the mRNA-level (data not shown).

To obtain a more complete view of the antitumorigenic role of ERβ, we investigated how ERβ expression affected growth of SW480-xenografts (Fig. 4). SW480-control or SW480-ERβ cells were transplanted into the right flank of SCID/beige mice (Fig. 4A). At end point, the average weight of SW480-control xenografts was 704 mg, whereas average weight of SW480-ERβ xenografts was 245 mg, a reduction of tumor weight by 65% following ERβ expression (Fig. 4A and B). Immunohistochemistry with a Ki67-specific antibody was performed. The xenografts in the mice injected with ERβ-expressing cells showed a significant down-regulation of the number of Ki67-positive cells (Fig. 4C and D). A TUNEL assay was performed on the xenografts to detect apoptosis, but there was no significant difference in apoptosis between SW480-ERβ and SW480-control xenografts (data not shown).

Moreover, microarray studies were performed to investigate genome-wide transcriptional effects as a result of long-term expression (3 weeks) of ERβ in SW480 cells. ERβ expression resulted in dramatic changes of the transcriptome; 623 differentially regulated transcripts were detected and the effect was mainly repressive (Supplementary Data).

As mentioned, preclinical studies have shown reduced expression of ERβ in colon cancer tissue compared with normal colonic epithelium. ERβ is expressed even in metastatic colon cancer but to a lower extent compared with normal colonic epithelium. This points toward a potential role of ERβ as a protein with tumor suppressive function. Earlier studies by other researchers have shown effects of ERβ on proliferation in another colon cancer cell line (23), but detailed analyses in combination with in vivo xenograft studies are lacking. Flow cytometry with propidium iodide staining revealed a clear difference in cell cycle distribution of SW480-ERβ compared with SW480-control as well as SW480-ERβ–DBDmut cells. There was a nearly 50% increase of cell proportion in the G₁ phase in presence of ERβ expression compared with the control. The cell cycle distribution of SW480-ERβ-DBDmut cells was very similar to that of SW480-control cells, pointing toward an ERE-dependent, antiproliferative effect of ERβ. Furthermore, expression of ERβ also reduced the proliferation of HCT-116 colon cancer cells but not that of HT-29 cells. The reason behind the lack of antiproliferative effect in HT-29 cells in not clear at present. However, all these cell lines express very low levels of endogenous ERβ mRNA. Preliminary experiments and published data indicate that the relative ERβ-mRNA level is higher in HCT-116 and SW480 compared with HT-29 cells (24). This finding might reflect alternative signaling pathways in proliferation of HT29 cells. Interestingly, this type of mechanism is well known in breast cancer cells, where ERα- cancers are frequently dependent on tyrosine-kinase signaling to proliferate (25). Moreover, the total ERβ level in the transduced cells was lower in HT-29 cells compared with HCT-116 and SW480.

We performed both real-time PCR analysis of rate-limiting cell cycle factors in the G₁ phase as well as microarray analysis for the overall effects by ERβ expression (Supplementary Data). We found several proliferation-regulating genes to be transcriptionally inhibited by ERβ. c-Myc is an important immediate early gene, and an activator of numerous G₁-S-phase genes. Expression of c-Myc has been shown to counteract a dnTCF-4 expressed in the colon cancer cell line HT29 with respect to effects on proliferation. This points toward c-Myc as a central factor in colon cancer growth, neutralizing the effect of the APC tumor suppressor (16). We found that ERβ was a strong inhibitor of c-Myc expression at both mRNA and protein levels, which could be an important step in the repression of proliferation by ERβ. c-Myc expression affects the transcription of many proliferation-associated genes and Cdc25A is one of the most well-known targets (26). We found Cdc25A to be repressed at mRNA and protein-levels by ERβ, most likely dependent on c-Myc repression. In SW480-ERβ-DBDmut, c-Myc was induced at mRNA level but not as strongly regulated at protein level; therefore, it is likely that ERβ-mediated repression occurs through ERE binding.

In response to c-Myc overexpression, normal cells undergo p53-dependent apoptosis and one pathway is through induction of p14^ARF (27). Interestingly, p14^ARF protein but not mRNA was induced in response to ERβ, despite its strong inhibition of c-Myc. This indicates that ERβ may induce p14^ARF through an alternative pathway. It is known that p14^ARF is regulated through proteasomal degradation (28). Accordingly, it is possible that ERβ regulates p53 by affecting p14^ARF degradation. Induction of p53 by ERβ through p14^ARF could be a novel pathway important for ERβ-mediated apoptosis in premalignant lesions with wild-type p53.

p21 is an important downstream effector of p53-signaling in parallel to its proapoptotic effects and p21 was found to be up-regulated in the ERβ-expressing cells. SW480 cells contain a mutated form of p53, presumably without proapoptotic potential but with some transcriptional activity such as the ability to induce p21 (29). In line with this, we could not find increased apoptosis in the ERβ-expressing tumor xenografts, but a strong induction of p21-protein in vitro.

We also analyzed the ERβ-mediated regulation of cell cycle genes in two other colon cancer cell lines: HCT-116 and HT29. Cell cycle regulation at protein level by ERβ was very similar in HT29 cells compared with SW480. Interestingly, in HCT-116 cells, which contain a nonmutated, wild-type p53, this protein was strongly reduced by ERβ, for unknown reasons. Furthermore, the p53 target gene p21 was reduced. Most interestingly, the expression of c-Myc was reduced in both cell lines, further strengthening the role of c-Myc as a central ERβ-regulated cell cycle gene.

The other major cell cycle pathway regulated by ERβ was the p45-p27 pathway. p27-protein but not p27-mRNA was induced by ERβ expression. Therefore, we investigated whether p45, governing p27 degradation, was affected by ERβ. We found that p45 was strongly reduced at protein and mRNA-levels. We believe that the combined c-Myc reduction and p27/p21 induction efficiently inhibits cell cycle progression. We have earlier studied the effects of ERβ in breast cancer cells (19, 30) where c-Myc was reduced and p21 and p27 were induced. Remarkably, Cyclin D1 was induced by ERβ in colon cancer cells and this was seen in breast cancer cells as well, but even under these conditions, proliferation was inhibited.

pRb is an important tumor suppressor, often mutated and inactivated in human malignancies. We found that total pRb level was up-regulated by ERβ expression. Phosphorylation status of pRb was changed as well with increased phospho-Ser780, which requires Cyclin D-cdk4/6 activity (31) and decreased phospho-Ser795 following ERβ expression. This indicates that the increase in Cyclin D1-level causes an activation of the Cyclin D1-cdk4/6-complexes, and a partial phosphorylation of pRB. Still, the antiproliferative and antitumorigenic effects of ERβ expression are strong, which means that ERβ probably has its major effect on a parallel or downstream target. It is believed that complete inactivation of pRb and dissociation of E2F1 requires sequential phosphorylation of pRb by the Cyclin E/A-cdk2 complexes as well (32). Therefore, we suggest that the combined action of increased cdk2-inhibitor levels (p21 and p27), decreased Cdc25A-level and, not least, increased pRB-levels results in strong inhibition of proliferation by inactivation of the Cyclin E/A-cdk2 complexes.

We have used an ERβ construct with mutated DNA-binding domain (ERβ-DBDmut) to investigate ERβ effects in absence of functional, classic ERE binding. Several genes are regulated by ERα and ERβ through nonclassic binding sites such as AP-1 and SP-1, independently of a functional DBD region. Therefore, the ERβ-DBDmut does not work as a negative control for ERβ signaling. It is likely that in genes where the ERβ effects are completely dependent on ERE binding, the effects can be reversed with the ERβ-DBDmutant. On the other hand, when the ERβ effects are mediated through other promoter elements, the expression of the ERβ-DBDmut can result in both induction and repression of transcription. Based on the findings in this article, we conclude that the antiproliferative effect of ERβ in colon cancer cells is dependent on ERE binding and, hence, a functional DBD region. Most interestingly, to our knowledge, this is the first study to indicate a ligand-independent, ERE-mediated regulation by ERβ. However, we cannot rule out the possibility that the E₂ independence is a result of ERβ overexpression. But today, no colon cancer cell lines with endogenous ERβ are available.

Our in vivo xenograft experiments indicate that ERβ efficiently blocks tumor growth. However, because all xenografts grew in presence of E₂, we do not know whether estrogens further strengthen the antiproliferative function of ERβ. This should be investigated in future studies. Loss of ERβ is seen in the progression of several types of malignancies, e.g., in breast, ovarian (33), and colon cancer. Konstantinopoulos and colleagues (11), among many authors, showed that reduction in ERβ expression was correlated with the degree of tumor differentiation, poorly differentiated tumors exhibiting a larger loss of ERβ compared with moderately or well-differentiated tumors. It is interesting to note that reduced p27 and increased p45 are correlated to higher aggressiveness in many types of cancer (34, 35) in line with a regulation of these factors by ERβ.

In conclusion, ERβ inhibits proliferation and tumor growth of colon cancer cells by regulating G₁-phase cell cycle genes. Moreover, this regulation is dependent on a functional DBD-domain of ERβ.

J-A. Gustafsson: commercial research grant, consultant, and ownership interest, KaroBio AB. The other authors disclosed no potential conflicts of interest.

Grant support: Swedish Cancer Fund, KaroBio AB, Lars Hierta Memorial Foundation, and the Magnus Bergvall foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.