Opposing effects of Runx2 and estradiol on breast cancer cell proliferation: in vitro identification of reciprocally regulated gene signature related to clinical letrozole responsiveness

Abstract

Purpose: To assess the clinical significance of the interaction between estrogen and Runx2 signaling, previously shown in vitro.

Experimental design: MCF7/Rx2(dox) breast cancer cells were treated with estradiol and/or doxycycline to induce Runx2, and global gene expression was profiled to define genes regulated by estradiol, Runx2, or both. Anchorage-independent growth was assessed by soft-agar colony formation assays. Expression of gene sets defined using the MCF7/Rx2(dox) system was analyzed in pre- and on-treatment biopsies from hormone receptor-positive patients undergoing neoadjuvant letrozole treatment in two independent studies, and short-term changes in gene expression were correlated with tumor size reduction or Ki67 index at surgery.

Results: Reflecting its oncogenic property, estradiol strongly promoted soft-agar colony formation, whereas Runx2 blocked this process suggesting tumor suppressor property. Transcriptome analysis of MCF7/Rx2(dox) cells treated with estradiol and/or doxycycline showed reciprocal attenuation of Runx2 and estrogen signaling. Correspondingly in breast cancer tumors, expression of estradiol- and Runx2-regulated genes was inversely correlated, and letrozole increased expression of Runx2-stimulated genes, as defined in the MCF7/Rx2(dox) model. Of particular interest was a gene set upregulated by estradiol and downregulated by Runx2 in vitro; its short-term response to letrozole treatment associated with tumor size reduction and Ki67 index at surgery better than other estradiol-regulated gene sets.

Conclusion: This work provides clinical evidence for the importance of antagonism between Runx2 and E2 signaling in breast cancer. Likely sensing the tension between them, letrozole responsiveness of a genomic node, positively regulated by estradiol and negatively regulated by Runx2 in vitro, best correlated with the clinical efficacy of letrozole treatment.

Figure 1

Conditional expression of Runx2 and colocalization with ERα in MCF7 BCa cells. A, Western blot analysis of Runx2 in MCF7/Rx2^dox cells treated for 48h with 0.5 μg/ml Dox or vehicle. Runx2-positive T47D BCa cells and ROS17/2.8 osteosarcoma cells were used as control. Tubulin was analyzed to confirm equal loading. B, Confocal immunofluorescence analysis of Runx2 and ERα in MCF7/Rx2^dox cells treated for 24h with vehicle (C; control), 0.5 μg/ml Dox (D) to induce Runx2, 10 nM E2 (E), or both together (DE) as indicated in the legend.

Figure 2

Runx2 diminished E2-stimulated anchorage independent growth of MCF7/Rx2^dox cells in soft agar. A, MCF7/Rx2^dox cells were incubated for three weeks in soft agar containing vehicle control (C), dox (D), E2 (E), or both (DE), and the colonies formed were enumerated (Mean±SEM; n=3). B, Representative phase contrast images form one of the three experiments (4X magnification).

Figure 3

Genome-wide reciprocal modulation of Runx2 and E2 signaling in MCF7/Rx2^dox BCa cells. MCF7/Rx2^dox cells were maintained for two days in medium supplemented with CSS prior to treatment with either vehicle control (C) dox (D), E2 (E), or both stimulants (DE). After 48h of treatment, RNA was isolated and subjected to microarray analysis using Illumina HumanRef-8 v3.0 expression Beadchips. A, Scatter plot describing the log₂ response of each gene to Runx2 in the presence of E2 (Log₂DE/E) versus absence of E2 (Log₂D/C). Data points representing multiple genes are proportionally darker. The bars illustrate the fraction of genes whose response to Runx2 was attenuated in the presence versus absence of E2 for each of the indicated levels of stimulation or repression. B, Log₂ fold changes by E2 in the presence (y-axis,) versus absence of Runx2 (x-axis) are plotted for each gene. The bars illustrate the fraction of genes whose response to E2 was attenuated in the presence versus absence of Runx2 for each of the indicated levels of stimulation or repression.

Figure 4

Negative interaction of E2- and Runx2 signaling in BCa tissues. A, Venn diagram of genes responsive to dox (Runx2) or estradiol by ≥1.3-fold (FDR p<0.05) in MCF7/Rx2^dox cells. Genes responsive to dox but not E2, or vice versa, are indicated as “D-only” or “E-only”, respectively, and those responsive to each are indicated as “common”. B, Linear regression analysis of the correlation between the D-only^↑ and the E-only^↑ metagenes in BCa biopsies from the three indicated cohorts. The metagenes were defined as average of strandardized expression of all genes within the gene list. The D-only^↑ metagene was represented by 34 (out of 52) genes which were upregulated by ≥2 fold in response to dox treatment in the MCF7/Rx2^dox cell culture model and were also present in the HG-U133A Affymetrix array used in Gene Expression Omnibus datasets GSE11121 (39), GSE2034 (3) and GSE7390 (38). The E-only^↑metagene was represented by 37 (out of 53) genes upregulated by ≥2.5 fold in response to E2 treatment of MCF7/Rx2^dox cells and present in the above public datasets. C, Negative correlation between the E-only^↑ and D-only^↑ metagenes at baseline in patients from the WU-POL study. D, Changes in expression of the E-only^↑ (left) and D-only^↑ (right) metagenes in inidividual patients in the POL study between base line (BL) at the time of initial biopsy and four weeks after initiation of letrozole treatment (4WK). Two-sided paired sample t-test was used to examine the significance of the mean paired difference.

Figure 5

Various relationships between effects of E2 and Runx2 on “common” genes. A, Unsupervised hierarchical clustering of the “common” genes (Figure 4A) based on their normalized expression in MCF7/Rx2^dox cells treated with vehicle control (C), Dox (D), E2 (E), or both stimulants (DE). Pathway enrichment of genes in each cluster was assessed by DAVID. B, RT-qPCR analysis of representative genes from each of the four clusters from panel A.

Figure 6

Letrozole responsiveness of the Cluster IV metagene predicts future Ki67 index. A. Changes in expression of the Custer I-IV metagenes in the WU-POL study between baseline (BL) at the time of initial biopsy, and 4 weeks (4WK) after initiation of letrozole treatment. Two-sided paired sample t-test was used to examine the significance of the mean paired difference. B. Distribution of Ki67 (%) at surgery in the lower and higher expressing group of patients defined based on the Cluster I-IV metagenes. Patients were dichotomized into low and high expressing groups based on their response to letrozole (4WK-BL) in each cluster and further examined if the two groups differed in their Ki67 status determined at surgery. Significance of the difference between the two groups was determined by Wilcoxon rank sum test. C, Distribution of Wilcoxon rank sum test p-values of correlation of high and low 4WK-BL difference of 1000 random metagenes (50 genes each) sampled from the list of the E2-stimulated genes in MCF7/Rx2^dox cells (N=1061) with the Ki67 index at surgery. the double arrowed vertical line indicates the p value of 0.0043 returned for the Cluster IV metagene. D, Analysis of an independent letrozole study from the GEO dataset GSE5462 (41). Distribution of the letrozole-induced changes in Cluster IV metagene exprtession (2WK-BL) in patients that later showed a decrease in tumor volume (responders, R) versus non-responders (Non-R).