Introduction
Metastasis of primary tumors to distant sites is a complex process that involves a sequence of interdependent events including intravasation, survival within the circulation, extravasation, and colonization. Epithelial-mesenchymal transition (EMT) has been strongly implicated in metastasis [
1,
2]. During EMT, epithelial cells dissociate from each other, in part due to loss of E-cadherin expression, upregulate mesenchymal markers, acquire a fibroblast-like morphology, reorganize their cytoskeleton, and become more motile and invasive [
1,
2]. Several transcription factors, including members of the SNAI family have been shown to promote EMT and thus tumor dissemination [
3‐
5]. Bone metastasis is a frequent complication of breast cancer (BCa), with distinct gene signatures defining bone-seeking tumors [
6‐
9].
Prolonged exposure to estradiol (E2) is associated with an increased risk of BCa [
10‐
14]. The mechanisms through which estrogens contribute to BCa initiation and progression are complex and implicate estrogen receptor (ER)-mediated genomic and nongenomic signaling as well as the action of genotoxic estrogen metabolites [
12]. In contrast to E2-mediated carcinogenesis, the presence of ERα is a favorable prognostic marker associated with less invasive tumors, and those negative for ERα are more aggressive [
15]. A randomized clinical trial of postmenopausal women receiving equine estrogen treatment revealed a decrease in BCa incidence [
16] and low-dose estradiol treatment has been proposed as a treatment modality for advanced ERα-positive BCa that does not respond to aromatase inhibition [
17]. Indeed, introduction of ERα into ERα-negative BCa cells attenuated their pro-cancerous properties
in vitro [
18‐
21]. Thus, anti-estrogen therapy for BCa patients, while antagonizing the oncogenic properties of estrogens, may inadvertently result in loss of their anti-metastatic property. Better understanding of mechanisms underlying the beneficial role of estrogen signaling in advanced disease may inform the development of novel therapeutic approaches and improved treatment plans for BCa patients.
Runx2 is a lineage-specific transcription factor with crucial roles in both bone biology and carcinogenesis [
22‐
24]. During development Runx2 is involved in the process of osteogenesis. Targeted disruption of Runx2 in mice leads to failure of osteoblast differentiation and bone formation [
25,
26], and Runx2 haploinsufficiency in humans results in the skeletal disorder cleidocranial dysplasia, with a similar phenotype observed in Runx2 haploinsufficient mice [
25]. Although Runx proteins have tumor suppressor properties [
24], recent studies assigned a role for Runx2 in promoting breast and prostate cancer metastasis [
27‐
32]. Thus, Runx2 and E2 signaling play dual roles in BCa, with each functioning to either promote or suppress tumor progression. The mechanisms underlying these contrasting manifestations in cancer are poorly understood.
We previously showed that in the presence of ligand, ERα physically binds Runx2 and inhibits expression of several Runx2 target genes [
33]; our recent study revealed that in MCF7 BCa cells E2 independently regulated about half of the Runx2-responsive genes [
34]. We therefore hypothesized that gene(s) stimulated by Runx2 and inhibited by E2 may contribute to manifestations of the pro-metastatic or tumor suppressor functions of Runx2 and E2, respectively. Using a combination of tissue culture modeling and bioinformatics analysis of gene expression in BCa biopsies, we present evidence suggesting that SNAI2/SLUG plays a role in mediating the pro- and anti-metastatic effects of Runx2 and E2, respectively.
Discussion
Apart from promoting BCa progression, estrogen signaling has paradoxical anti-metastatic properties [
15,
17,
19‐
21]. Potentially contributing to this, E2 antagonized the transcription factor Runx2, whose role in metastasis is being increasingly recognized [
22,
24,
30‐
32]. Specifically, Runx2 promoted EMT and invasion of BCa cells
in vitro, and this was antagonized by E2 (Figure
2). At the center stage of the opposing effects of Runx2 and E2 signaling on EMT and invasion was the transcription factor SNAI2. Consistent with previous reports [
30,
48,
49], SNAI2 was stimulated by Runx2 and inhibited by E2 (Figure
1A and
2A); Runx2 no longer enhanced EMT and invasion after SNAI2 knockdown; SNAI2 expression in BCa biopsies positively correlated with a metagene that reports on Runx2 activity and negatively correlated with ERα mRNA (Figure
1B), and expression of SNAI2 was associated with BCa metastasis in general and bone metastasis in particular (Figure
5A-D).
SNAI2 likely plays different roles during various stages of BCa progression, and its ability to transcriptionally repress E-cadherin and induce EMT is well documented [
3‐
5,
49‐
51]. The role of SNAI2 downstream of Runx2 in BCa is reminiscent of the role of its homologue, SNAI1, in EMT and metastasis of breast, ovarian, colon, lung and squamous cell carcinomas [
4]. However, SNAI1 is not stimulated by Runx2 in either MCF7 BCa [
34] or C4-2B PCa cells [
30]. Instead, it is SNAI2 that is strongly stimulated by Runx2 in BCa and PCa cells (Figure
2). Furthermore, SNAI2 and not SNAI1 was repressed by E2 in our study (data not shown), and SNAI2 and not SNAI1 exhibited an inverse correlation with ERα and E-cadherin in human BCa tumors [
49]. Like Runx2, Twist1 has recently been shown to induce SNAI2 expression, and SNAI2 was essential for Twist1-mediated EMT of human mammary epithelial cells [
50]. Thus, depending on context, either SNAI1 or SNAI2 regulate BCa metastasis, with the latter mediating the pro-metastatic activity of Runx2 and Twist1 as well as the anti-metastatic activity of E2.
It remains to be investigated whether Runx2 directly regulates SNAI2 transcription. On the one hand, there are several Runx-binding motifs upstream of the SNAI2 transcription start site (Additional file
3: Supplemental Figure 1A). However, Runx2 ChIP-seq analysis in C4-2B cells, where SNAI2 expression is also stimulated by Runx2 (Figure
1B), did not suggest occupancy of Runx2 at these SNAI2 upstream Runx motifs [
52]. Instead, we believe that Runx2 indirectly stimulates SNAI2 expression via modulation of ETS signaling. This hypothesis is based on the observation that Runx2 down-regulated expression of SPDEF (Additional file
3: Supplemental Figure 1B and [
30]), a transcriptional inhibitor belonging to the ETS family, which suppresses tumor progression and metastasis [
53] in part through direct inhibition of SNAI2 expression [
54]. Indeed, our Runx2 ChIP-seq analysis in C4-2B/Rx2
dox cells [
52] demonstrated strong occupancy of the SPDEF transcription start site by Runx2 (Additional file
3: Supplemental Figure 1C). We are currently investigating how such Runx2 occupancy represses SPDEF transcription, and which of the many ETS sites upstream of the SNAI2 transcription start site (Additional file
3: Supplemental Figure 1A) mediate its repression by SPDEF and/or its activation by other members of the ETS family. Regardless of the precise mechanism by which Runx2 controls SNAI2, expression of the two is tightly correlated in BCa tumors (Figure
1D). In fact, much of the reported association between Runx2 expression and BCa metastasis [
22] is attributable to SNAI2. Indeed, the results of our independent analysis validating the association of Runx2 with metastasis (Additional file
4) are remarkably similar to those demonstrating the association of SNAI2 with metastasis (Figure
5).
The anti-Runx2/SNAI2, anti-metastatic properties of E2 signaling in BCa cells may be relevant for early
in vivo metastatic events (Figure
5E), and possibly to targeting BCa metastasis to non-osseous tissues (Figure
5F-G). However, they appear less relevant for bone metastasis. Whereas SNAI2 expression is associated with bone-seeking primary BCa tumors and BCa bone metastasis even more strongly than with lung or brain metastasis, ERα is negatively associated with only non-osseous metastasis (Figure
5). In fact, consistent with previous reports [
9,
55], the correlation between ERα expression and bone metastasis was positive, not negative in our cohort (Figure
5H); BCa tumors that metastasized to bone were mostly ER-positive (Figure
5Q) and BCa bone metastases co-expressed high levels of Runx2/SNAI2 and ERα (Figure
5V). The bone-seeking property that ERα bestows on BCa cells may be counteracted during the first two years after surgery by the general anti-metastatic property of estrogens, resulting in a total neutral effect, but at the end it is the former that prevails (Figure
5H). Possibly, activation of bone-seeking pathways in primary BCa cells, such as Src1 [
8,
9], results in the stimulation of SNAI2 and other Runx2-regulated genes despite high levels of ERα.
Conclusions
Runx2 stimulates EMT and invasiveness in MCF7 BCa cell cultures, adding to the mounting evidence for its role in metastasis. Our data suggest that E2 attenuates BCa metastasis by antagonizing Runx2. Indeed, EMT and invasiveness of MCF7 cells were severely compromised by E2 specifically after induction of Runx2. The linkage between the pro- and anti-metastatic properties of Runx2 and E2, respectively, is attributable in part to SNAI2, which is upregulated by Runx2 and downregulated by E2. Furthermore, we observe a strong positive association between Runx2 and SNAI2 expression, a negative association between ESR1 and SNAI2, and a positive correlation between SNAI2 expression and metastasis in BCa tumors. These observations are consistent with the hypothesis that unopposed stimulation of Runx2 target genes such as SNAI2 in ER-negative tumors contributes to their aggressive metastatic phenotype. However, the relationship between ER expression and metastasis is more complicated in bone-seeking tumors. Here, the negative correlation between SNAI2 and ER is weak, and the anti-metastatic property of ER is likely masked by an opposite property, which ultimately prevails. Mimicking ER signaling to specifically antagonize Runx2 and SNAI2 offers a research avenue towards the development of novel therapeutic approaches for the management of BCa patients who fail first line therapy.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
NOC, SKB and GHL performed the experiments. NOC and YBC performed the bioinformatics analyses. NOC, MK, ZB, DT and BF designed the study and wrote the manuscript. All authors read and approved the final manuscript.