Background
The luminal subtypes that express the estrogen receptor (ER)α represent approximately 70% of breast cancers, and the majority of these tumors respond to endocrine therapy [
1]. However, resistance to endocrine therapy resulting in relapse is seen in approximately 30% of patients [
1]. ERα
+ breast cancers are heterogeneous with at least two subtypes, luminal A and luminal B [
2]. Luminal A tumors are estradiol (E2)-dependent and responsive to antiestrogens, whereas luminal B tumors display either intrinsic or acquired resistance to antiestrogens with an outcome almost similar to triple negative breast cancers (TNBCs) [
3]. A subgroup of luminal A tumors, particularly those that have metastasized despite expressing luminal A biomarkers (ERα and progesterone receptor (PR)), do not respond to antiestrogen therapies and approximately 55% of these metastases have converted to a different subtype through an unknown mechanism [
4].
Multiple mechanisms of antiestrogen resistance have been documented [
5]. Most of the prior work focused on mechanisms that confer E2-independent activity to ERα, including kinases that phosphorylate ERα, co-activator molecules that enhance ERα activity, pioneer factors that govern chromatin binding of ERα, and growth factor receptor–ERα crosstalk [
6‐
8]. However, to our knowledge, there have been limited attempts to decipher negative regulatory loops that may restrict ERα signaling subsequent to ligand-activated induction and deregulation of these negative regulatory loops leading to prolonged/sustained activation of ERα.
To identify luminal cell-expressed genes that may play a role in restricting E2-dependent proliferation, we scanned gene expression array datasets for E2-inducible genes with ERα binding sites and that have a growth inhibitory activity [
9]. From this search, we focused on the dependence receptor (DR) pathways for their potential role in a negative feedback loop. Under physiological conditions, unliganded DRs elicit cell death and/or growth inhibition but elicit cell survival and proliferation when coupled with their ligands such as Netrin-1 (NTN1) [
10]. DRs are direct transcriptional targets of p53 and integral to p53-dependent apoptotic pathways, particularly in the absence of ligands [
11]. NTN1 belongs to the evolutionary conserved netrin family secreted proteins and is well characterized for its role in the nervous system [
12]. Both netrins and DRs also play crucial roles in other systems, including development of the mammary gland, inner ear, lungs, and pancreas [
12,
13]. Loss of heterozygosity and homozygous deletion of DRs and upregulation of netrins are observed in a variety of cancers including breast cancer [
11,
13]. These aberrations in DR–netrin pathways are believed to confer resistance to p53-dependent apoptosis and enhance proliferation of cancer cells.
In the present study, we show that UNC5A is an E2-inducible gene. Knockdown of UNC5A in ERα+/PR+ cells resulted in defective turnover of phosphorylated ERα, enhanced E2 signaling, cell proliferation, and tumorigenesis independent of E2 supplementation accompanied with multiorgan metastases in xenograft models. Furthermore, UNC5A knockdown cells acquired a hybrid basal/luminal phenotype including elevated expression of epidermal growth factor receptor (EGFR). Thus, UNC5A could serve as a negative feedback molecule in ERα signaling, the deregulation of which could lead to breast cancer progression through enhanced plasticity.
Discussion
UNC5A is a transmembrane receptor that generates cell survival or death signals in a ligand-dependent manner [
10]. UNC5A and NTN1 are described as tumor suppressor and oncogene, respectively, in breast cancer [
51,
52]. However, signaling pathways that control their expression to alter the balance between UNC5A and NTN1 are unknown. Analyses of TCGA dataset showed elevated expression of UNC5A in luminal breast cancers, and NTN1 overexpression in TNBCs and E2 could further enhance luminal expression of UNC5A (Fig.
1). Thus, the UNC5A–NTN1 signaling axis is likely tilted more towards UNC5A-activated signals in ERα
+/PR
+ breast cancers and NTN1-generated signals in TNBC/ER
− tumors. Consistent with this possibility, UNC5A expression was prognostic in ER
+/PR
+/HER2
− breast cancers but not in ER
− tumors, suggesting its critical role in ER
+/PR
+/HER2
− breast cancers. A subgroup of women with ER
+/PR
+ breast cancers develop recurrence, and molecular assays such as the Recurrence Score and Breast Cancer Index are helping to identify ER
+/PR
+ breast cancer patients requiring hormonal and/or chemotherapy [
53]. UNC5A, possibly in combination with EGFR, could be developed as a biomarker to identify such patients [
4].
Molecular events causing variable UNC5A expression in ER
+ tumors are unknown.
UNC5A is a TP53-inducible gene, and TP53 is infrequently mutated in ER
+/PR
+ breast cancer [
2,
54]. Deregulated p53 activity instead of mutations may lead to loss of UNC5A expression in ER
+/PR
+ tumors, although this remains to be investigated. In addition, there is potential for p53 to control UNC5A activity since we noted a differential effect of
UNC5A knockdown on proliferation in wild-type p53 containing MCF7 cells compared with mutant p53 containing T-47D cells, although knockdown had a similar effect on BCL2 and TP63 expression in both cell lines. cBioPortal analyses revealed frequent missense and truncating mutations in
UNC5A [
55]. Additionally,
UNC5A expression is regulated through allele-specific DNA methylation [
56]. Thus, mutations and DNA methylation could be other mechanisms leading to inactivation/silencing of
UNC5A during breast cancer progression.
One of the consequences of reduced UNC5A expression is significant changes in basal gene expression and altered ERα signaling. GO analyses revealed a specific effect of
UNC5A knockdown on a network of transcription factors including the stem cell-associated transcription factor
SOX2, which may be a reason for the altered expression of 20% of genes in
sh-UNC5A MCF7 cells and 7% in
sh-UNC5A T-47D cells compared with
sh-Control cells (Fig.
4). It is interesting that, in both cell lines,
UNC5A knockdown affected the expression of genes linked to the plasma membrane and extracellular region composition (Fig.
4), which can explain the aggressive growth characteristics of
sh-UNC5A compared with
sh-Control MCF7 cells in vivo. We also observed a distinct effect of UNC5A on E2-regulated gene expression, with several genes gaining E2-regulated gene expression (Fig.
4). These results suggest a role for UNC5A in restricting the activity of unliganded ERα in a gene-specific manner, which could involve the following mechanisms. One possibility is the direct effect of UNC5A-activated signals on chromatin organization since we observed an effect of
UNC5A knockdown on the histone demethylation network in E2-treated cells (Fig.
4).
UNC5A knockdown increased the E2-inducible expression of KDM4B, which is a master regulator of ERα activity [
57]. Elevated KDM4B in
sh-UNC5A cells could further amplify the E2 signaling axis as evident from more than 400 genes gaining E2-regulated expression in
sh-UNC5A cells. Robustness at which
UNC5A knockdown altered BCL2 expression further suggests a direct link between UNC5A and chromatin organization. This drastic effect of
UNC5A knockdown on BCL2 expression is reminiscent of permanent gene expression changes observed upon transient knockdown of DNMT1 [
38]. However, we did not observe an effect of
UNC5A knockdown on the expression of any
DNMTs, although there was a modest but statistically significant effect on
TET1 and
TET3 which antagonize DNMTs (Additional file
6).
UNC5A knockdown may have an effect on histone acetylation/deacetylation since
sh-UNC5A MCF7 cells expressed significantly higher levels of the epigenetic regulator
HDAC9 compared with
sh-Control cells (Additional files
6). The second possibility is the involvement of
ELF5. ELF5 suppresses E2 sensitivity by reducing the expression of
ESR1 and the pioneer factors
FOXA1 and
GATA3 [
36]. ERα, FOXA1, and GATA3 constitute a positive lineage-restricted hormone responsive regulatory loop in luminal cells [
58]. We observed the effect of
UNC5A knockdown on
ELF5 expression, and reduced
ELF5 expression in
sh-UNC5A MCF7 cells correlated with elevated
GATA3 expression (Additional file
6). The third possibility is the involvement of AKT.
UNC5A knockdown caused upregulation of activated AKT, which confers ligand-independent activity to ERα [
59]. The fourth possibility involves ERα–EGFR crosstalk since
UNC5A knockdown cells contained higher levels of EGFR protein (Fig.
7), and EGF through EGFR has been shown to alter ERα cistrome and transcriptome [
60].
UNC5A knockdown in MCF7 cells resulted in a hybrid phenotype with cells expressing luminal (ER,
PGR), myoepithelial (TP63), and stem cell markers (
SOX2), which in part is due to altered ERα signaling. Recent studies have identified similar hybrid cells in primary breast cancers, potentially generated through Notch-Jagged signaling [
42,
61]. Based on cell surface marker profiles and KRT14/KRT19 expression patterns in
sh-UNC5A cells, we propose that the gradual loss of
UNC5A results in cancer cells acquiring hybrid phenotype without expressing classic markers of EMT. Since there is still a controversy related to in-vivo detection of cancer cells with EMT features, it is possible that primary tumors contain cells with hybrid phenotype which functionally behave like cancer cells with EMT features. Characterizing primary tumors for ER, PR, UNC5A, EGFR, and additional basal cell markers would allow the detection of such hybrid cells. Collectively, results presented in this study provide novel insights into pathways that restrict ERα signaling and metastatic progression of ERα
+ breast cancer, which potentially involves luminal to luminal/basal hybrid conversion due to an aberrant DR pathway.