Anti-estrogen Resistance in Human Breast Tumors Is Driven by JAG1-NOTCH4-Dependent Cancer Stem Cell Activity

Abstract

Breast cancers (BCs) typically express estrogen receptors (ERs) but frequently exhibit de novo or acquired resistance to hormonal therapies. Here, we show that short-term treatment with the anti-estrogens tamoxifen or fulvestrant decrease cell proliferation but increase BC stem cell (BCSC) activity through JAG1-NOTCH4 receptor activation both in patient-derived samples and xenograft (PDX) tumors. In support of this mechanism, we demonstrate that high ALDH1 predicts resistance in women treated with tamoxifen and that a NOTCH4/HES/HEY gene signature predicts for a poor response/prognosis in 2 ER+ patient cohorts. Targeting of NOTCH4 reverses the increase in Notch and BCSC activity induced by anti-estrogens. Importantly, in PDX tumors with acquired tamoxifen resistance, NOTCH4 inhibition reduced BCSC activity. Thus, we establish that BCSC and NOTCH4 activities predict both de novo and acquired tamoxifen resistance and that combining endocrine therapy with targeting JAG1-NOTCH4 overcomes resistance in human breast cancers.

Graphical abstract

Figure 1

Tamoxifen or Fulvestrant Treatment of ER+ Patient-Derived Samples and PDXs Selectively Enriches for Cells with CSC Properties High BCSC frequency is associated with worse outcomes for tamoxifen-treated BC patients. (A) Mammosphere self-renewal of freshly isolated ER+ early and metastatic patient-derived samples. Primary mammospheres cultured in the presence of ethanol (Control) or 10⁻⁶ M 4-hydroxy-tamoxifen (Tamoxifen) were dissociated and re-plated in secondary mammosphere suspension culture for a further 7–9 days to measure self-renewal of mammosphere-initiating cells treated in the first generation. p value was calculated with Wilcoxon signed-rank test. (B) Representative micrographs of metastatic BC cells before fluorescence-activated cell sorting (FACS) analysis of ALDH1 enzymatic activity (ALDEFLUOR assay). ALDH-positive cells were discriminated from ALDH-negative cells using the ALDH inhibitor DEAB. (C) Percentage of ALDH-positive cells in nine ER+ metastatic BC patient-derived samples. Cells were grown in adherence with ethanol (Control), tamoxifen (10⁻⁶ M), or fulvestrant (10⁻⁷ M) for 7–9 days. Arrows indicate fold change greater than 20% compared to control. (D–G) Early (HBCx34) and metastatic (BB3RC31) BC estrogen-dependent PDX tumors treated in vivo for 14 days with tamoxifen (10 mg/kg/day, oral gavage; red bars) or fulvestrant (200 mg/kg/week, subcutaneous injection; blue bars). Gray bars correspond to vehicle control. FFPE, formalin-fixed paraffin-embedded. (E) Representative micrographs and quantification of Ki67 expression determined by immunohistochemistry (IHC). (F) Percentage of MFE. (G) ALDH-positive cells (%) determined using the ALDEFLUOR assay. (H) ALDH1 expression was assessed by immunohistochemistry in breast tumor epithelial cells, and the percentage of positive cells was scored. Representative micrographs of ALDH-high (ALDH^hi) and -low (ALDH^lo) epithelial expression are shown. Kaplan-Meier curves represent cumulative survival for the ALDH^lo population and ALDH^hi population of a cohort of 322 pre-menopausal ER+ BC patients who participated in a randomized trial of 2 years of adjuvant tamoxifen treatment versus no systemic treatment (control). Vertical bars on survival curves indicate censored cases. p values are based on a log-rank (Mantel-Cox) test of equality of survival distributions. Scale bars, 100 μm. Data are represented as mean ± SEM. ^∗p < 0.05; ^∗∗p < 0.01. See also Figure S1.

Figure 2

Tamoxifen or Fulvestrant Treatment Upregulates Notch Target Genes in Patient-Derived Samples and PDXs JAG1-NOTCH4 receptor signaling in ALDH-positive cells drives Notch activity in endocrine-resistant BC. (A and B) Expression of Notch target genes HEY1 and HES1 was assessed by real-time qPCR analysis and compared to control to determine fold change. (A) Metastatic BC patient-derived cells were treated for 7–9 days with ethanol (control), tamoxifen (10⁻⁶ M), or fulvestrant (10⁻⁷ M) and a correlation between fold change of expression of HEY1 and HES1 and fold change of percentage of ALDH-positive cells is shown. (B) Early (HBCx34) and metastatic (BB3RC31) BC PDXs: the effect of in vivo treatment for 14 days with tamoxifen (10 mg/kg/day, oral gavage) or fulvestrant (200 mg/kg/week, subcutaneous injection) on HEY1 and HES1. (C) NOTCH4, HES1, and JAG1 protein expression levels determined by western blot in metastatic (Met) (BB3RC31) BC PDX. β-actin was used as a reference for the loading control. (D) NOTCH4, HES1, and JAG1 protein expression levels were determined by western blot in MCF-7 ALDH-negative and ALDH-positive sorted cells. MCF-7 cells were treated with tamoxifen or fulvestrant for 6 days before ALDH sorting. (E) Wild-type MCF-7 cells (filled bars) and a CRISPR clone containing a disruption of NOTCH4 exon 2 (N4EX2 cells, hatched bars) treated in adherence with ethanol (Control, gray bars), 10⁻⁶ M tamoxifen (red bars), and 10⁻⁷ M fulvestrant (blue bars) for 6 days. N4EX2 cells’ fold change of MFE and ALDH-positive cells after treatments was compared to that of the wild-type cells. (F and G) NICD4 and JAG1 rescue tamoxifen- or fulvestrant-inhibited growth: cell number (using sulforhodamine B [SRB] assay, y axis) of MCF-7 overexpressing (F) NICD4-GFP, (G) JAG1-GFP, or GFP control incubated with tamoxifen or fulvestrant for 1, 3, and 5 days (x axis) compared to the respective cell line treated with control ethanol. p values are for the 5-day treatment. Data are represented as mean ± SEM. ^∗p < 0.05; ^∗∗p < 0.01. See also Figures S2 and S4.

Figure 3

NOTCH4 Inhibition using RO4929097 Abrogates Tamoxifen and Fulvestrant Enrichment of CSC Activities (A–C) Early (HBCx34) and metastatic (Met) (BB3RC31) PDX tumors treated in vivo for 14 days with tamoxifen (10 mg/kg/day, oral gavage) or fulvestrant (200 mg/kg/week, subcutaneous injection) in the presence or absence of the NOTCH4 inhibitor RO4929097 (3 mg/kg/day, oral gavage). (A) MFE (%). (B) Percentage of ALDH-positive cells. (C) Secondary tumor formation. 100,000 cells of metastatic (BB3RC31) PDX were re-implanted subcutaneously in NSG mice with 90-day slow-release estrogen pellets. Tumor growth (>100 mm³) was determined at day 90 after cell injection. (D and E) MCF-7, T47D, and ZR-75-1 cells were pre-treated in adherence with 10⁻⁶ M tamoxifen (red bars) and 10⁻⁷ M fulvestrant (blue bars) with RO4929097 (10 μM; hatched bars) or DMSO (filled bars) for 72 hr. (D) MFE and (E) percentage of ALDH-positive cells were assessed after pre-treatments. (F–H) MCF-7 cells were pre-treated in adherence for 6 days in the presence of RO4929097 (10 μM; hatched bars) or DMSO (filled bars). (F) In vivo experiments were carried out in NSG mice with 90-day slow-release estrogen pellets. Tumor growth (>100 mm³) was assessed at day 60 and is represented as mice positive for growth/mice tested for each cell number tested. ELDA of tumor-initiating cell frequency is shown. (G) Expression of HEY1 and HES1 by real-time PCR was compared to the control. (H) HES1 protein expression levels determined by western blot. Data are represented as mean ± SEM. p values refer to hatched bars compared to filled control bars. ^∗p < 0.05; ^∗∗p < 0.01. See also Figures S3 and S4.

Figure 4

HBCx22 and HBCx34 TAMR PDXs Express High Levels of HES1 NOTCH4 inhibitor RO4929097 targets CSCs in tamoxifen-resistant (TAMR) PDXs. (A) Representative micrographs and quantification of HES1 expression determined by immunohistochemistry. Scale bars, 100 μm. (B–D) HBCx22 and HBCx34 TAMR PDXs treated in vivo for 14 days in the presence or absence of the GSI RO4929097 (10 mg/kg/day, oral gavage). (B) MFE (%). (C) Percentage of ALDH-positive cells. (D) Representative FACS plots of ALDEFLUOR assay. ALDH-positive cells were discriminated from ALDH-negative cells using the ALDH inhibitor DEAB. Data are represented as mean ± SEM. ^∗p < 0.05; ^∗∗p < 0.01. See also Figure S4.

Figure 5

NOTCH4 Receptor Activity Predicts for Resistance to Tamoxifen Treatment and Prognosis in ER+ Tumors (A and B) NOTCH4, HES1, HEY1, and HEY2 genes in ER+ primary tumors from (A) tamoxifen-treated or (B) untreated patients are co-expressed in the heatmap ranked from left to right using the four-gene signature. Colors are log₂ mean-centered values; red indicates high, and green indicates low. All significant cut-points (p < 0.05) are shown in gray. Kaplan-Meier analysis using the optimum cut-point (dashed white line) demonstrates that elevated expression of the Notch genes is significantly associated with an increased rate of (A) distant metastasis and (B) decreased overall survival. Vertical bars on survival curves indicate censored cases. p values are based on a log-rank (Mantel-Cox) test. (C) Diagram suggesting that endocrine therapies do not target BCSCs and emphasizing the need of targeting residual drug-resistant cells to eliminate all cancer cells and prevent long-term recurrences of ER+ BC.