Indole-3-carbinol and its N-alkoxy derivatives preferentially target ERα-positive breast cancer cells

Abstract

Indole-3-carbinol (I3C) is a natural anti-carcinogenic compound found at high concentrations in Brassica vegetables. I3C was recently reported to inhibit neutrophil elastase (NE) activity, while consequently limiting the proteolytic processing of full length cyclin E into pro-tumorigenic low molecular weight cyclin E (LMW-E). In this study, we hypothesized that inhibition of NE activity and resultant LMW-E generation is critical to the anti-tumor effects of I3C. LMW-E was predominately expressed by ERα-negative breast cancer cell lines. However, ERα-positive cell lines demonstrated the greatest sensitivity to the anti-tumor effects of I3C and its more potent N-alkoxy derivatives. We found that I3C was incapable of inhibiting NE activity or the generation of LMW-E. Therefore, this pathway did not contribute to the anti-tumor activity of I3C. Gene expression analyzes identified ligand-activated aryl hydrocarbon receptor (AhR), which mediated sensitivity to the anti-tumor effects of I3C in ERα-positive MCF-7 cells. In this model system, the reactive oxygen species (ROS)-induced upregulation of ATF-3 and pro-apoptotic BH3-only proteins (e.g. NOXA) contributed to the sensitivity of ERα-positive breast cancer cells to the anti-tumor effects of I3C. Overexpression of ERα in MDA-MB-231 cells, which normally lack ERα expression, increased sensitivity to the anti-tumor effects of I3C, demonstrating a direct role for ERα in mediating the sensitivity of breast cancer cell lines to I3C. Our results suggest that ERα signaling amplified the pro-apoptotic effect of I3C-induced AhR signaling in luminal breast cancer cell lines, which was mediated in part through oxidative stress induced upregulation of ATF-3 and downstream BH3-only proteins.

Abbreviations

AhR

aryl hydrocarbon receptor

CYP

cytochrome p450 oxidases

DIM

3,3-diindoylmethane

ERα

estrogen receptor α

HMECs

human mammary epithelial cells

I3C

indole-3-carbinol

LMW-E

low molecular weight cyclin E

NE

neutrophil elastase

ROS

reactive oxygen species

RPPA

reverse phase protein array

TNBC

triple-receptor negative breast cancer

Introduction

Epidemiological studies have convincingly demonstrated an association between a diet rich in fruits and vegetables and the reduced incidence of some tumor types, including colon, stomach, esophageal, lung, oral, endometrial, and pancreatic.¹ Total fruit and vegetable consumption was not consistently correlated with reduced breast cancer risk.² However, high dietary consumption of cruciferous vegetables, particularly from the Brassica genus (e.g. broccoli, cauliflower, cabbage, and Brussels sprouts), was specifically associated with lower breast cancer risk.³ Brassica vegetables may contain biologically active phytochemicals with specific chemopreventative properties in the context of breast cancer.

Indole-3-carbinol (I3C), a naturally occurring compound generated from the hydrolysis of glucobrassicin, is found at exceptionally high concentrations in Brassica vegetables. Oral administration of I3C prevented spontaneous⁴ and carcinogen-induced⁵ mammary tumor formation in rodent models. Studies using breast cancer cell lines have demonstrated that I3C possess anti-tumor properties, including the suppression of proliferation and induction of apoptosis.^6,7 I3C readily undergoes acid-catalyzed condensation leading to the generation of numerous oligomeric products, predominately 3,3-diindoylmethane (DIM).⁸ DIM and other I3C oligomers are biologically active and contribute to anti-tumor effects of I3C in experimental models.^8,9 In patients, I3C and its condensation products were readily absorbed by the gut and could be detected in the blood plasma.¹⁰ Collectively, the available evidence indicates that I3C and its oligomeric condensation products have potent chemopreventative and anti-tumor properties, likely contributing to the protective effect of Brassica vegetable consumption against breast carcinogenesis.

Mechanistically, I3C and its oligomeric products have pleiotropic effects on physiology and cell signaling.⁸ In breast cancer cell lines, both I3C and DIM activated the aryl hydrocarbon receptor (AhR),¹¹ which retards estrogen induced cell proliferation through transcriptional downregulation and ubiquitination/proteasome degradation of estrogen receptor α (ERα).^12,13 In human subjects and animal models, I3C-induced activation of AhR increased the expression of cytochrome p450 oxidases (CYP)1A1 and CYP1A2, which altered estrogen metabolism in a manner consistent with reduced breast cancer risk.^4,14-16 AhR-signaling is an important molecular determinate of the chemopreventative effects of I3C in the breast and other reproductive tissues where estrogen signaling plays an important role in tumorigenesis.

A recent publication identified I3C as an inhibitor of neutrophil elastase (NE) activity, which resulted the reduction of NE-mediated tumor-specific processing of cyclin E into low-molecular weight (LMW-E) isoforms.¹⁷ Previously, our laboratory identified 2 NE cleavage sites at the N-terminus of full length cyclin E (50 kDa), which accounted for the generation of LMW-E (45–33 kDa).¹⁸ Compared to full-length cyclin E, LMW-E isoforms bound to CDK2 much more efficiently and conferred resistance to endogenous CDK inhibitors (e.g. p21 and p27), which accounted for their ability to hyperactivate CDK2 and mediated their tumorigenic potential.^18-21 LMW-E isoforms are also strong prognostic indicators of poor breast cancer patient outcome²² and may be an important therapeutic target.^23,24

In this study, we originally hypothesized that inhibition of NE activity and resultant LMW-E generation is critical to the anti-tumor effects of I3C. We observed that, LMW-E expressing breast cancer cell lines were predominately ERα-negative. However, ERα-positive breast cancer cell lines demonstrated greater sensitivity to I3C and its more potent N-alkoxy derivatives.²⁵ Contrary to previously published results,¹⁷ I3C failed to inhibit NE activity or disrupt the generation of LMW-E. To identify pathways that accounted for sensitivity to I3C, we preformed proteomic and gene expression analyzes. We found that AhR, a direct molecular target of I3C,¹¹ mediated sensitivity to the anti-tumor effects of I3C in ERα-positive MCF-7 cells. Furthermore, we identified a role for reactive oxygen species (ROS)-induced upregulation of the stress response transcription factor ATF-3 and pro-apoptotic BH3-only proteins (e.g. NOXA) in the sensitivity of breast cancer cell lines to I3C. Lastly, we show that overexpression of ERα in the TNBC cell line MDA-MB-231 increased sensitivity to the anti-tumor effects of I3C, suggesting that ERα expression plays a direct role in the sensitivity of luminal breast cancer cell lines to the anti-tumor effects of I3C.

Results

ERα-positive breast cancer cell lines demonstrate greater sensitivity to the anti-tumor effects of I3C and its N-alkoxy derivatives

Studies from our laboratory have demonstrated a significant role for LMW-E in tumor growth and progression.^20,21,26 Therefore the report of a natural compound (i.e. I3C) that reduced the tumor-specific processing of cyclin E into LMW-E though inhibition of NE activity¹⁷ was of great interest. We initially hypothesized that inhibition of NE activity and resultant LMW-E generation is critical to the anti-tumor effects of I3C. We also considered the anti-tumor effect of N-alkoxy I3C derivatives, which were previously shown to more potently suppress the proliferation of breast cancer cells when compared to I3C.²⁵ N-methoxy, N-ethoxy, N-propoxy, and N-butoxy derivatives of I3C, characterized by N-alkoxy substitution at the nitrogen position on the indole ring from one to 4 carbons in length, were synthesized according to the previously published methodology.²⁵

We first evaluated the expression of cyclin E (both full-length and LMW-E) in a panel of breast cancer cell lines (including luminal, TNBC, and HER2-positive subtypes) and several immortalized HMECs. Western blot analysis (Fig. S1A) revealed that ERα-negative cell lines expressed LMW-E at significantly higher levels relative to ERα-positive breast cancer cell lines (Fig. S1B). Based on our hypothesis, we expected that the subset of ERα-negative cell lines highly expressing LMW-E would be sensitive to I3C-mediated NE-inhibition.

The sensitivity of breast cancer cell lines to I3C and its N-alkoxy I3C derivatives was determined by MTT assay. We calculated the IC50 of each cell line, as illustrated in Fig. 1A. We observed that ERα-positive breast cancer cell lines demonstrated greater sensitivity to the anti-tumor effects of I3C and its N-alkoxy derivative compared to immortalized HMECs and ERα-negative breast cancer cell lines (Fig. 1B). Significantly, the sensitivity of breast cancer cell lines to the N-alkoxy I3C derivatives increased with the length of the N-alkoxy carbon chain (Fig. 1B). The mean IC50 of all ERα-positive breast cancer cell lines treated with I3C was 204 μM compared to N-methoxy I3C (15 μM), N-ethoxy I3C (5.4 μM), N-propoxy I3C (3.8 μM), and N-butoxy I3C (2.0 μM). For comparison, the mean IC50 of all ERα-negative breast cancer cell lines treated with I3C was 491 μM compared to N-methoxy I3C (93 μM), N-ethoxy I3C (73 μM), N-propoxy I3C (52 μM), and N-butoxy I3C (44 μM).

Figure 1.

ERα-positive breast cancer cell lines demonstrate greater sensitivity to the anti-tumor effects of I3C and its N-alkoxy I3C derivatives. (A) Representative ERα-positive (MCF-7, ZR75.1, T47D) and ERα-negative (MDA-MB-231, MDA-MB-157, MDA-MB-436) breast cancer cell lines were treated with I3C (50–1000 μM) for 48-hours. Cell viability was measured by MTT assay. Absorbance values were normalized to DMSO-treated controls and plotted vs. the logarithm of I3C concentration. The I3C concentration producing a half maximal decrease in cell viability (IC50) was calculated for each cell line from the resultant dose-response curve. (B, C) ERα-positive breast cancer cell lines [HER2-negative: MCF-7, ZR75.1, T47D, MDA-MB-175, and SUM185; HER2-postive: BT474 and UACC-893], ERα-negative breast cancer cell lines [TNBC: MDA-MB-231, MDA-MB-157, MDA-MB-436, HCC-1806, MDA-MB-468, HCC-1937, and BT549; HER2-positive: HCC-1954], and immortalized HMECs [76NF2V and MCF-10A] were treated with I3C (50–1000 μM) or its N-alkoxy derivatives (0.1–1000 μM) for 48-hours. Cell viability was measured by MTT assay. (B) IC50 values were calculated for each cell lines treated with I3C or its N-alkoxy derivatives. X-axis organized by length of the substituted N-alkoxy carbon chain, such that 0 = I3C, 1 = N-methoxy I3C, 2 = N-ethoxy I3C, 3 = N-propoxy I3C, and 4 = N-butoxy I3C (C) Grouped analysis comparing mean IC50 values (I3C and its N-alkoxy derivatives) of immortalized HMECs, ERα-positive, and ERα-negative breast cancer cell lines.

These results demonstrate that ERα-positive breast cancer cell lines are significantly more sensitive to the anti-tumor effects of I3C and it N-alkoxy derivatives compared to ERα-negative breast cancer cell lines or HMEC. LMW-E levels did not predict sensitivity to these compounds, suggesting that this pathway was not a significant molecular target of I3C.

Treatment with I3C increased the apoptotic cell death and decreased the proliferation of ERα-positive breast cancer cell lines

We next examined if the anti-tumor activity of I3C in breast cancer cells is due to inhibition of cell proliferation, induction of apoptotic cell death or both. Examination of BrdU incorporation into DNA revealed that I3C treatment consistently decreased the proliferation of ERα-positive cell lines compared to vehicle treated controls (Fig. 2A). However, proliferation of ERα-negative cell lines was minimally affected by I3C at the concentration (200 μM) examined (Fig. 2A). TUNEL analysis revealed that I3C treatment consistently increased apoptotic cell death in ERα-positive breast cancer cell lines compared to vehicle treated cells (Fig. 2B). Apoptotic cell death in the ERα-negative breast cancer cell lines remained virtually unaffected at the I3C concentration (200 μM) examined.

Figure 2.

Treatment with I3C increased the apoptotic cell death and decreased the proliferation of ERα-positive breast cancer cell lines. (A, B) Representative ERα-positive (ZR75.1, MCF-7, and T47D) and ERα-negative (MDA-MB-157, MDA-MB-231, and MDA-MB-436) breast cancer cell lines were treated with I3C at a concentration of 200 μM or an equal volume of vehicle (DMSO) for 24 hours. (A) Cells were pulsed with 10 μM BrdU for 30 minutes and BrdU incorporation into DNA (marking proliferating cells) was measured by flow cytometry. BrdU positive cells are indicated as a percentage of total cells. (B) Cells were subjected to the TUNEL staining (marking apoptotic cells and quantified by flow cytometry; TUNEL positive cells are indicated as a percentage of total cells. (C) Cells were treated with PBS, vehicle (DMSO), or increasing concentrations of I3C (100–500 μM) for 24-hours. Cell lysates were subjected to protein gel blot analysis using antibodies against PARP and phosphorylated Rb (S807/811). MDA-MB-436 cells do not express Rb due to mutational inactivation of the RB1 gene⁵⁸. Actin, loading control.

To further evaluate the anti-tumor effect of I3C, we examined the levels of PARP cleavage and phosphorylated Rb, biomarkers of apoptosis and cell proliferation respectively. Western blot analysis revealed a dose-dependent increase in cleaved PARP, in ZR75.1 and T47D cell lines (Fig. 2C). MCF-7 cells lack caspase 3²⁷, which may explain the lack of PARP cleavage in these cells (Fig. 2C). A dose dependent decrease in Rb phosphorylation was also observed in the ERα-positive breast cancer cell lines (Fig. 2C). However, in the ERα-negative breast cancer cell lines, I3C had a minimal effect on PARP cleavage and Rb phosphorylation at the concentrations (100–500 μM) of I3C examined (Fig. 2C).

Collectively, these results show that the anti-tumor effects of I3C on ERα-positive breast cancer cell lines were associated with increased apoptotic cell death and suppression of cell proliferation. In ERα-negative breast cancer cells, the changes in cell proliferation and apoptosis mediated by I3C were negligible at the concentrations examined.

I3C is not an inhibitor of NE and does not disrupt cyclin E processing into LMW-E

To address if NE activity and LMW-E generation are important molecular targets mediating the anti-tumor effects of I3C in breast cancer cell lines, we first evaluated the ability of I3C to inhibit NE and reduce the tumor-specific processing of cyclin E into LMW-E.¹⁷ Western blot analysis revealed that treatment with I3C consistently increased full length cyclin E expression in all breast cancer cell lines examined and LMW-E in the ZR75-1, MDA-MB-157, and MDA-MB-468 cell lines (Fig. 3A). I3C was also unable in inhibit the activity of purified NE, measured using the chromogenic substrate Methoxysuccinyl-Ala-Ala-Pro-Val-p-nitroanilide, which is specifically hydrolyzed by NE. The concentration range of I3C (0.2 nanomolar to 25 micromolar) used in this experiment was consistent with the concentrations used in the disputed publication.¹⁷ Attesting to the validity of this assay, well-characterized inhibitors of NE (sivelestat, GW311616, and SSR69071) effectively inhibited NE activity (Fig. 3B).

Figure 3.

I3C is not an inhibitor of NE and does not disrupt cyclin E processing into LMW-E. (A) Cells were treated with PBS, vehicle (DMSO), or increasing concentrations of I3C (100–500 μM) for 24-hours. Protein gel blot analysis was performed using an antibody against cyclin E. Actin, loading control (for all cell lines except MDA-MB-468, compare to actin from Fig. 2C, these blots were stripped and re-probed for cyclin E). (B) Dose-response curve demonstrating NE activity in the presence of I3C, sivelestat, GW311616, and SSR69071 (concentration range: 100–0.2 μM). NE -activity was measured by hydrolysis of chromogenic substrate Methoxysuccinyl-Ala-Ala-Pro-Val-p-nitroanilide. (C, D) MCF-7 cells stably overexpressing PCDNA 4.0, full length cyclin E (EL), LMW-E truncation 1 (T1), or LMW-E truncation 2 (T2) were treated with PBS, vehicle (DMSO), or I3C (200 μM). (C) Cell lysates were subjected to protein gel blot analysis using antibodies against phosphorylated Rb (S807/811) and cyclin E. Actin, loading control. (D) BrdU incorporation into DNA, as described Fig. 2A.

Treatment of MCF-7 cells overexpressing full length cyclin E (EL), LMW-E truncation 1 (T1) or LMW-E truncation 2 (T2) with I3C (200 μM) reduced Rb phosphorylation (Fig. 3C) and BrdU incorporation into DNA (Fig. 3D) similar to the vector (PCDNA 4.0) expressing control cells. This result suggests that I3C-mediated inhibition of cell proliferation was independent of the ability of I3C affect LMW-E levels.

Overall, out results demonstrate that I3C does not inhibit NE, contrary to previously published results.¹⁷ Furthermore, the growth inhibitory effects of I3C on breast cancer cells were not mediated by interference with NE-activity or disruption of cyclin E processing.

Microarray and RPPA analysis of I3C treated breast cancer cells reveals pathways preferentially altered in the sensitive cells

To identify the molecular mechanisms contributing to the anti-tumor effects of I3C on sensitive ERα-positive breast cancer cell lines, we preformed proteomic (RPPA) and gene expression (microarray) analyzes.

RPPA is an antibody-based proteomic technology for the analysis of proteins and phosphorylation events commonly associated with cancer cell signaling pathways (see supplemental materials for full data set). We identified alterations in ERα-positive cell lines indicative of decreased hormone receptor signaling (i.e. ERα, PR, GATA3, AR, and AIB1),^28-30 increased apoptosis (i.e. cleaved caspase 3, SMAC, and IAP),⁸ proliferative arrest (i.e. phosphorylated Rb at S807/S811, p21, and c-myc),^31,32 increased p53-signaling (including p53, p21, and 53BP1),³² and increased protein translation (phosphorylated 4E-BP1 at S65 and ribosomal protein S6 at S235/S236) (Table S1).

Microarray analysis identified a large number of alterations in gene expression (see supplemental materials for full data set), consistent with the pleotropic effects of I3C on cell signaling.⁸ Ingenuity pathway analysis identified significant overlap between specific alterations in gene expression mediated by I3C in sensitive ERα-positive breast cancer cell lines and several canonical pathways of interest. Several of these pathways were consistent with previous findings, including p53 signaling,³² apoptosis signaling,⁸ aryl hydrocarbon receptor signaling,^12,13 estrogen receptor signaling,^28-30 cell cycle,^31,32 role of BRCA1 in DNA damage response,³³ and NRF2-mediated oxidative stress response ³⁴ (Table S2).

Gene expression analysis found that CYP1A1 was highly upregulated by I3C in the ERα-positive cells (70-fold) compared to ERα-negative cell lines (11-fold). CYP1A1 is the canonical target gene of the AhR, which is known to be a direct target of I3C.¹³ The stress response transcription factor ATF3 is also highly upregulated in ERα-positive cells (28-fold) compared to ERα-negative cell lines (3.6-fold). Finally, pro-apoptotic factor NOXA (PMAIP1) is highly upregulated in ERα-positive cells (26-fold), but not significantly altered in ERα-negative cell lines (Fig. 4A). Based on these alterations in gene expression, we proposed a model accounting for the sensitivity of breast cancer cells to I3C (Fig. 4B). In this model, I3C activates AhR, which synergizes with the ERα signaling pathway, resulting in the transactivation of AhR target genes, including CYP1A1. Extensive crosstalk between the AHR and ERα has been reported and was required for transactivation of AhR target genes following exposure to the potent ligand 2,3,7,8-tetracholorodibenzo-p-dioxin (TCDD)^35-38. Several AhR-target genes, including CYP1A1, increase intracellular ROS ^39-41. ATF-3 is upregulated by oxidative stress,^42,43 resulting in the transactivation of p53 target genes including NOXA (PMAIP1).⁴⁴ We next interrogated if key nodes in this proposed model (Fig. 4B) account for the increased anti-tumor activity of I3C in ERα-positive compared to ERα negative breast cancer cells.\raster(94%)="rgFigKCCY_A_942210_F0004_B"

Figure 4.

Microarray analysis of I3C treated breast cancer cells reveals pathways preferentially altered in the sensitive cells. (A) I3C sensitive (ERα-positive: MCF-7, ZR75.1, and T47D) and I3C insensitive (ERα-negative: MDA-MB-231, MDA-MB-157 and MDA-MB-436) breast cancer cell lines were treated with I3C at a concentration of 200 μM or vehicle (DMSO) and subjected to microarray analysis. Normalized microarray data sets were fit with linear mixed effects models on a gene-by-gene basis. We used a FDR of 0.001 to identify genes that were altered differentially between sensitive and insensitive cell types treated with I3C. (B) Hypothetical molecular model accounting for sensitivity of ERα-positive breast cancer cells based on gene expression analysis.

AhR signaling mediates sensitivity to the anti-tumor effects of I3C

Microarray analysis revealed that several AhR target genes, CYP1A1, CYP1A2, ALDH1A3, and ALDH3A1, were preferentially upregulated in ERα-positive cells compared to ERα-negative cells (Fig. 5A). We validated these results by treating MCF-7 cells (an ERα-positive cell line) with I3C and observed a dose-dependent increase in CYP1A1 protein expression concomitant with downregulation of ERα (Fig. 5B). We hypothesized that AhR mediates the anti-tumor effects of I3C in ERα-positive breast cancer cell lines.

Figure 5.

AhR and ERα signaling mediated sensitivity to the anti-tumor effects of I3C. (A) Fold change (FC) in AhR and its target genes in I3C sensitive and insensitive breast cancer cell lines, from microarray analysis. (B) MCF-7 cells were treated with vehicle (DMSO) or I3C (100, 250, 500 μM). Cell lysates were subjected to protein gel blot analysis using antibodies against AhR, ERα, and CYP1A1. Actin, loading control. (C–E) MCF-7 cells stably expressing non-targeting shRNA (control), shAhR.1, and shAhR.2 were generated. (C) The cells were treated with vehicle (DMSO) or I3C (100 and 250 μM) and lysates were subjected to protein gel blot analysis using an antibody against AhR. Actin, loading control. (D) Dose response curve showing cell viability (measured by MTT assay) following treatment with I3C (50–600 μM) for 48-hours. (E) IC50 values were calculated for each cell lines treated with I3C (50–600 μM) or its N-alkoxy derivatives (0.1–200 μM). X-axis organized by length of the substituted N-alkoxy carbon chain, see Fig. 1A (F–G) MDA-MB-231 parental cells and cells stably overexpressing either GFP or ERα were generated. (F) The cells were treated with vehicle (DMSO) or I3C (100 and 250 μM) and lysates were subjected to protein gel blot analysis using an antibody against ERα. Actin, loading control. (G) Dose response curve showing cell viability (measured by MTT assay) following treatment with I3C (50–1000 μM) for 48-hours.

To test this hypothesis, we generated AhR knockdown MCF-7 cells (Fig. 5C). AhR knockdown reduced the proliferation rate in MCF-7 cells, increasing doubling time from 24 hours in controls to an average of 34 hours in AhR knockdown cells measured over a 96 hour period (data not shown). A ligand independent role for AhR in cell proliferation was consistent with previous publications.^45,46 AhR knockdown significantly decreased sensitivity to the anti-tumor effects of I3C in these cells (Fig. 5D). Compare the IC50 of shRNA control cells treated with I3C (263 μM) to that of shAhR.1 (390 μM) and shAhR.2 (448 μM) cells. The decreased sensitivity of shAhR.1 and shAhR.2 MCF-7 compared to control MCF-7 was also evident when these cells were treated with the more potent I3C derivatives, N-methoxy I3C (IC50 control: 20 μM, shAhR.1: 48 μM, and shAhR.2: 40 μM) N-ethoxy I3C (IC50 control: 9 μM, shAhR.1: 42 μM, and shAhR.2: 38 μM), N-propoxy I3C(IC50 control: 7 μM, shAhR.1: 26 μM, and shAhR.2: 31 μM), and N-butoxy I3C (IC50 control: 4 μM, shAhR.1: 18 μM, and shAhR.2: 22 μM) (Fig. 5E).

Notably, knockdown of AhR in the ERα-negative breast cancer cell line MDA-MB-157 (Fig. S2A) only modestly reduced the sensitivity of this cell line to the anti-tumor effects of I3C (Fig. S2B). Compare the IC50 of shRNA control cells treated with I3C, 491 μM, to that of shAhR.1 (585 μM) and shAhR.2 (570 μM). These results provide evidence that AhR mediates the sensitivity of MCF-7 cells to the anti-tumor effects of I3C and its N-alkoxy derivatives.

ERα sensitizes cells to the anti-tumor effects of I3C

Extensive crosstalk between the AhR and ERα signaling pathways has been described.³⁷ Therefore, we hypothesized that ERα controls the sensitivity of breast cancer cell lines to the anti-tumor effects of I3C via AhR. To test this hypothesis, we generated MDA-MB-231 cells overexpressing ERα (Fig. 5F). I3C addition resulted in the downregulation of exogenously expressed ERα, however the level of ERα in this system remained sufficient to increase sensitivity to the anti-tumor effects of I3C. Compare the IC50 of parental MDA-MB-231 cells treated with I3C (437 μM), pLenti-control cells (425 μM), and pLenti-ERα cells (269 μM) (Fig. 5G). Comparatively, the average IC50 of all ERα-positive cell lines treated with I3C was 203 μM (Fig. 1B).

This result suggests that ERα-status accounted for the differential sensitivity of breast cancer cell lines to I3C. Given a role for ERα in the sensitivity of breast cancer cell lines to I3C, we next examined the ability of I3C to target cells resistant to anti-estrogens. We generated tamoxifen resistant MCF-7 cells by culturing MCF-7 cells for 6 months in α-MEM supplemented with tamoxifen at a concentration of 1 μM. MCF-7 TAM-R cells are 6 fold more resistant to tamoxifen (IC50 5.2 μM) as compared to parental MCF-7 cells (IC50 0.9 μM) (Fig. S3A). By comparison, MCF-7 parental (IC50 277.9 μM) and TAM-R (IC50 253.1 μM) cells demonstrated similar sensitivity to the anti-tumor effect of I3C (Fig. S3B). This result suggests that resistance to anti-estrogen therapy does not decrease sensitivity of luminal ERα-positive breast cancer cells to the anti-tumor effects of I3C.

ROS induced ATF-3 enhances the sensitivity to the anti-tumor effects of I3C

Microarray analysis also identified ATF-3 as an I3C-induced gene that was preferentially upregulated in ERα-positive breast cancer cell lines (Fig. 4A). Western blot analysis revealed that I3C caused a dose-dependent (100–500 μM) increase in ATF-3 levels in MCF-7 cells (Fig. 6A). We hypothesized that ATF-3 is a mediator of the anti-tumor effects of I3C in ERα-positive breast cancer cell lines. Supporting this hypothesis, knockdown of ATF-3 in MCF-7 cells (Fig. 6B) decreased their sensitivity to the anti-tumor effects of I3C. Compare the IC50 of shRNA control cells treated with I3C (209 μM) to that observed in shATF3.1 (310 μM), shATF3.2 (298 μM) and shATF3.3 (304 μM) (Fig. 6C).

Figure 6.

ROS-induced ATF-3 enhances, while overexpression of BCL2 decreases sensitivity to the anti-tumor effects of I3C. (A) MCF-7 cells were treated with vehicle (DMSO) or I3C (100–500 μM) for 24-hours. Cell lysates were subjected to protein gel blot analysis using an antibody against ATF-3. Actin, loading control. (B, C) MCF-7 cells stably expressing non-targeting shRNA (control), shATF3.1, shATF3.2, and shATF3.3 were generated. (B) The cells were treated with vehicle (DMSO) or I3C (250 μM) and lysates were subjected to protein gel blot analysis using an antibody against ATF-3. Actin, loading control. (C) Dose response curve showing cell viability (measured by MTT assay) following treatment with I3C (50–500 μM) for 48-hours. (D, E) MCF-7 cells were pretreated with vehicle (PBS) or NAC at a concentration of 5 mM. (D) The cells were treated with vehicle (DMSO) or I3C (100 and 250 μM) and lysates were subjected to protein gel blot analysis using an antibody against ATF-3. Actin, loading control. (E) Dose response curve showing cell viability (measured by MTT assay) following treatment with I3C (50–500 μM) for 48-hours (F) Fold change (FC) in BH3-domain containing pro- and anti-apoptotic genes in I3C sensitive breast cancer cell lines, microarray analysis. (G, H) MCF-7 parental and cells stably overexpressing either GFP or BCL-2 were generated. (G) The cells were treated with vehicle (DMSO) or I3C (100 and 250 μM) and lysates were subjected to protein gel blot analysis using an antibody against BCL-2. Actin, loading control. (H) Dose response curve showing cell viability (measured by MTT assay) following treatment with I3C (50–500 μM) for 48-hours.

ATF-3 is a key modulator of the response to oxidative stress.^42,47 We found that the potent antioxidant NAC prevented the upregulation of ATF-3 in cells treated with I3C (Fig. 6D), suggesting that ATF-3 upregulation in this system was dependent on oxidative stress. Pre-treatment of MCF-7 cells with NAC decreased their sensitivity to the anti-tumor effects of I3C. Compare the IC50 of MCF-7 cells treated with I3C (248.1 μM) and the IC50 of MCF-7 cells pretreated with NAC (398.5 μM) (Fig. 6E). These results demonstrate that I3C-induced oxidative stress resulted in the upregulation of ATF-3, which contributing to the anti-tumor effect of I3C in breast cancer cell lines.

Overexpression of BCL2 blunts the anti-tumor effects of I3C

In our microarray analysis we found that another feature of ERα-positive breast cancer cell lines treated with I3C, was alteration of BH3-only domain containing proteins, including pro-apoptotic NOXA, BAK1, and BAX, and anti-apoptotic BCL-XL (Fig. 6F). NOXA upregulation was of particular interest given previous studies that identified a role for ATF-3 in NOXA upregulation and apoptotic cell death.⁴⁴ NOXA is a canonical p53 target gene. Although ERα-positive cells upregulated p53 in response to I3C (Fig. S4A), knockdown of p53 in MCF-7 cells (Fig. S4B) failed to alter sensitivity to the anti-tumor effects of I3C (Fig. S4C), suggesting that the anti-tumor properties of I3C were independent of p53.

We hypothesized, that upregulation of pro-apoptotic BH3-only proteins (Fig. 6F) is required for I3C induced apoptosis. To test this hypothesis, we generated MCF-7 cells overexpressing anti-apoptotic, BCL-2 (Fig. 6G). Overexpression of BCL-2 decreased the sensitivity of cells to the anti-tumor effects of I3C. Compare the IC50 of parental MCF-7 cells treated with I3C (254 μM), GFP overexpressing control cells (266 μM), and BCL-2 overexpressing cells (412 μM) (Fig. 6H).

These results further validated our model (Fig. 4B), which proposed that I3C treatment caused apoptotic cell death by increased expression of pro-apoptotic BH3-only proteins, including NOXA, following oxidative stress induced upregulation of ATF-3.⁴⁴

Discussion

The NE-mediated cleavage of full length cyclin E at 2 N-terminal sites accounts for the tumor specific generation of LMW-E isoforms.¹⁸ Compared to full-length cyclin E, LMW-E isoforms have a significant role in tumorigenesis.^18-21,26 Inhibition of LMW-E generation, using an inhibitor of NE-activity could positively impact the relatively poor outcome of breast cancer patients whose tumors express LMW-E.^22,23 I3C is a naturally occurring anti-carcinogenic compound found at high concentrations in brassica vegetables.⁸ A recent report identified I3C as an inhibitor of NE-activity, such that treatment of breast cancer cell lines with I3C diminished LMW-E levels in favor of full length cyclin E¹⁷. We originally hypothesized that inhibition of NE-activity and resultant LMW-E generation in breast cancer cell lines is critical to the anti-tumor effects of I3C. However, we could not reproduce the evidence that I3C inhibited NE-activity (Fig. 3B) or LMW-E generation (Fig. 3A),¹⁷ which served as the basis for our original hypothesis. In fact, we found that I3C increased the levels of both full length cyclin E and LMW-E in breast cancer cell lines (Fig. 3A). Control of LMW-E generation was not a major determinant of I3C-induced growth arrest (Fig. 3C).

By evaluating the sensitivity of a wide range of breast tumor cell lines to I3C, we found that ERα-positive (luminal-subtype) breast cancer cell lines demonstrated greater sensitivity to the anti-tumor effects of I3C compared to ERα-negative breast cancer cell lines and immortalized HMECs (Fig. 1B and C). We determined that the anti-tumor properties of I3C in ERα-positive breast cancer cell lines included both increased apoptotic cell death (Fig. 2A) and suppression of cell proliferation (Fig. 2A). The anti-tumor effects of I3C were evident only at to relatively high micromolar concentration (Fig. 2B), likely restricting its use as a therapeutic modality. For this reason, several research groups have modified the chemical structure of I3C to enhance its anti-tumor properties. In cell culture models, N-alkoxy derivatives of I3C (N-methoxy I3C, N-ethoxy I3C, N-propoxy I3C, and N-butoxy I3C)²⁵ have more potent anti-tumor properties compared to unmodified I3C (Fig. 1A) and also demonstrated a significant bias for ERα-positive breast cancer cell lines (Fig. 1B).

We performed proteomic (RPPA) and gene expression (microarray) analyzes to identify molecular pathways altered by I3C that may account for the sensitivity of ERα-positive breast cancer cell lines to I3C. Microarray analysis identified several transcriptional targets of the ligand-activated transcription factor AhR, which were preferentially upregulated in ERα-positive breast cancer cell lines (Fig. 4A). Previous studies in breast cancer cell lines found that I3C and its metabolites activated the AhR signaling pathway.^12,13 Knockdown of AhR decreased the sensitivity of MCF-7 cells to the anti-tumor properties of I3C (Fig. 5D), providing evidence that AhR mediated the anti-tumor effects of I3C in luminal breast cancer cell lines. AhR knockdown also decreased the sensitivity of MCF-7 cells to the anti-tumor properties of the N-alkoxy derivatives of I3C (Fig. 5E), suggesting that the increased potency of the N-alkoxy derivatives was due to stronger transactivation of AhR target genes. Consistent with this result, a previous report found that N-methoxy I3C more strongly induced the expression of the canonical AhR target gene, CYP1A1, when compared to I3C.⁴⁸ Possible explanations of the increased potency of N-alkoxy derivatives of I3C include greater cellular absorption, enhanced binding to AhR, and increased stability in cell culture.

Extensive crosstalk exists between the AhR and ERα signaling pathways, leading us to hypothesize that ERα expression increases enhances the vulnerability of luminal breast cancer cell lines to I3C. Luminal breast cancer cell lines are dependent on ERα for growth and survival. Activation of AhR by I3C interferes with ERα-induced gene expression.⁴⁹ AhR ligands, including I3C and its metabolites, induced the AhR-dependent ubiquitination and degradation of ERα in breast cancer cells.^12,13,50,51 Consistent with these reports, we observed the downregulation of endogenously expressed ERα in MCF-7 cells (Fig. 5B) and exogenously expressed ERα in MDA-MB-231 cells (Fig. 5F) following addition of I3C.^12,13,50 ERα downregulation at the mRNA level was also previously described in response to I3C⁵² and was apparent in our microarray analysis of luminal breast cancer cell lines treated with I3C (Fig. 4A). I3C also alters estrogen metabolism through the AhR-dependent upregulation of CYP1A1 and CYP1A2, which catalyzes the 2-hydroxylation of estradiol and generates metabolites antagonistic toward ERα signaling with anti-carcinogenic properties.^4,14,16

Although ligand-activated AhR induced the downregulation of ERα levels in breast cancer cell lines,^12,13,50,51 several studies suggest that ERα also has an important role in AhR signaling. ERα and AhR were found to physically interact, such that ERα was recruited to AhR target genes in the presence of AhR ligands.^35,36 ERα expression in breast cancer cell lines was correlated with AhR responsiveness to the potent ligand TCDD.³⁷ Overexpression of ERα in MDA-MB-231 cells, which normally lack ERα expression, was shown to restore the responsiveness of the CYP1A1 promoter, a canonical AhR target gene, to TCDD.³⁸ Using MDA-MB-231 cells, which are relatively insensitive to anti-tumor effects of I3C, we demonstrated that exogenous expression of ERα enhanced sensitivity to I3C (Fig. 5G). This result suggests that ERα directly mediates the sensitivity of breast cancer cell lines to the AhR-dependent anti-tumor effects of I3C.

In our microarray analysis, the stress response transcription factor ATF-3 and the BH3-only protein NOXA (PMAIP1) were prominently upregulated in I3C-sensitive luminal breast cancer cell lines (Fig. 6F). We incorporated these factors into our model (Fig. 4B), based on their previously identified role in the anti-tumor response.⁴⁴ Knockdown of ATF-3 in MCF-7 cells (Fig. 6B) decreased their sensitivity to the anti-tumor effects of I3C (Fig. 6C). Oxidative stress is an important inducer of ATF-3.^42,43 Several AhR-target genes, including CYP1A1, are known to increase intracellular ROS.^39-41 In MCF-7 cells, the antioxidant NAC prevented the upregulation of ATF-3 in response to I3C (Fig. 6D) and decreased the sensitivity of MCF-7 cells to the anti-tumor effects of I3C (Fig. 6E). ATF-3 activates the expression of a large repertoire of genes involved in cell cycle arrest and apoptosis, including the BH3-only protein NOXA (PMAIP1).⁴⁴ We found that the overexpression of the anti-apoptotic factor BCL-2 (Fig. 6G) decreased the sensitivity of luminal breast cancer cells to the anti-tumor effects of I3C (Fig. 6H). This result suggests that the upregulation of BH3-only genes, including NOXA, accounted for the sensitivity of ERα-positive breast cancer cell lines to I3C-induced apoptotic cell death.

In conclusions, NE-activity was not a molecular target of I3C as previously reported.¹⁷ Instead, we found that the increased sensitivity of ERα-positive breast cancer cell lines to the anti-tumor properties of I3C were mediated through AhR. Further analysis revealed a role for the oxidative stress-induced upregulation of the transcription factor ATF-3 and pro-apoptotic BH3-only proteins, including NOXA. Patients with ERα-positive breast cancer may benefit from a therapeutic moiety derived from I3C.

Material and Methods

Cell lines and culture conditions

All tumor cell lines and MCF-10A HMECs were obtained from ATCC and were routinely authenticated. 76NF2V HMECs were obtained from Dr. V. Band.⁵³ Tumor cell lines were cultured in DMEM (MDA-MB-468) or α-MEM (HyClone) containing 10% fetal calf serum (Atlanta Biological).⁵⁴ HMECs were cultured in DFCI-1, as previously described.⁵⁵ All cell lines were maintained in a humidified tissue culture incubator at 37ºC and 6.5% CO_2.

Materials

I3C was purchased from LKT Laboratories. N-Methoxy-I3C, N-Ethoxy-I3C, N-Propoxy-I3C, and N-Butoxy-I3C were synthesized according to previously published methodology ²⁵ by the Translational Chemistry Core Facility at the University of Texas MD Anderson Cancer Center (UT MDACC). N-Acetyl Cysteine (NAC) and tamoxifen were purchased from Sigma.

MTT assay

Cells were cultured in a 96-well plate (1500 or 5000 cells per well depending on proliferation rate) and treated with I3C or its N-alkoxy derivatives. MTT (50 μl/well at a concentration of mg/mL) was added for the final 4-hours of the 48-hour incubation. The media was then aspirated and 100 μl of solubilization solution (20mL 1N HCL, 50 mL 10% SDS, 430 mL isopropyl alcohol) was added to each well. Absorbance was quantified using a spectrophotometer (Victor3, Perkin-Elmer) at a wavelength of 590 nm, as previously described.⁵⁴

Brdu incorporation

Cells were pulsed for 30 minutes in media containing 10 μM BrdU (Invitrogen) then fixed in ice-cold 70% ethanol. The cells were labeled with a FITC-conjugated anti-BrdU monoclonal antibody (BD Bioscience). The percentage of cells incorporating BrdU was measured using a FACScalibur flow cytometer (BD Bioscience) and analyzed using FloJo software (Version 8), as previously described.⁵⁴

TUNEL assay

Cells were fixed in 2.5% paraformaldhyde (Sigma) and then post-fixed in ice-cold 70% ethanol. Apoptosis was measured using the APO-BrdU TUNEL Assay Kit (Invitrogen), according to manufactures instructions.⁵⁴

Western blot analysis

As previously described,⁵⁴ cells were lysed by sonication in the presence of protease/phosphatase inhibitors and cleared by ultra-centrifugation (125 000 g). The protein fraction was separated by gel electrophoresis, and transferred to Immobilon P membrane (Millipore). We probed with rabbit polyclonal antibodies to phosphorylated Rb site Ser807/811 (Cell Signaling Technology), PARP (Cell Signaling Technology), AhR (H-211,Santa Cruz), CYP1A1 (Santa Cruz), ATF-3 (C-19, Santa Cruz) and mouse monoclonal antibodies to actin (Chemicon), cyclin E (HE12, Santa Cruz), ERα (Novacastra), and BCL-2 (100; Santa Cruz).

NE activity assay

GW 311616 (Sigma), SSR69071 (Torcis), Sivelestat (Sigma), or I3C (LKT Laboratories) were incubated with 0.0001 U/μL NE (Calbiochem) and the NE specific substrate Methoxysuccinyl-Ala-Ala-Pro-Val-p-nitroanilide (Calbiochem) at a final concentration of 1 mM. Assay buffer (100 mM Tris HCL pH 7.5, 250 mM NaCl) was added for a total reaction volume of 100 μL. The reaction proceeded for 10 minutes at 37°C and absorbance values were measured at 410 nM using a spectrophotometer (Victor3, Perkin-Elmer).

Lentiviral shRNA

Target specific shRNA in the GIPZ lentiviral vector system (Thermo, OpenBiosystems) were co-transfected into 293T cells with pCMV deltaR8.2 and pMD2.G (produced by the Didier Trono Lab and made available through the Addgene Repository). Target cells were infected with the virus containing media in the presence of 8 μg/mL polybrene and selected in 1ug/ml puromycin. See supplemental material for targeting sequences.

Lentiviral cDNA

ERα and GFP cDNA were cloned into plenti CMV Blast DEST vector (obtained from the Addgene repository) ⁵⁶ using the Gateway cloning system (Invitrogen). See the supplemental materials for primer sequences We obtained precision LentiORF Human BCL-2 (Clone ID:PLOHS_100004100; Thermo, OpenBiosystems) . Packaging and infection were preformed as described for the pGIPZ lentiviral vectors. Cells were selected in 20 μg/mL blasticidin.

Microarray

Total RNA was extracted from 2 × 10⁶cells using the RNAeasy kit (Qiagen) and subjected to on column DNase I (NEB) digestion. The Genomics Core Facility at UT MDACC performed the cDNA labeling, hybridization to the Illumina HT-12 v4 BeadChip, and image acquisition. Raw signal intensities were obtained using the Beadstudio analysis software from Illumina and imported into the lumi Bioconductor package (R version 2.15 and Bioconductor version 2.11). The lumi package was used to perform quality control analysis and apply the robust spline normalization (RSN) algorithm to normalize between arrays.

Reverse phase protein array

RPPA analysis performed by the Functional Proteomics core facility at UT MDACC has been previously described.⁵⁷ Protein samples were serially diluted and arrayed on nitrocellulose coated slides. The slides were probed with 174 primary antibodies against critical nodes in cancer cell signaling pathways and scanned with ImageQuant (Molecular Dynamics). Spot intensity was determined by MicroVigene software (VigeneTech Inc.), relative protein abundance was estimated by supercurve fitting, and normalized for protein loading.

Microarray and RPPA analysis

To identify alteration in gene expression that could account for the differential sensitivity of breast cancer cell lines to I3C, we fitted the normalized RPPA and microarray data sets with linear mixed effects models on a gene-by-gene basis. Additional analysis details can be found in the supplemental materials.

Statistical considerations

All experiments were performed at least in triplicate. Error-bars represent the standard deviation from the mean. All pair-wise comparisons were analyzed using the unpaired, 2-sided, t-test using GraphPad Prisim Software (Version 6.0b); p < 0.05 was considered significant.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Supplementary Material

Supplemental data for this article can be accessed on the publisher's website.