Introduction
Breast cancer is one of the most frequently diagnosed cancers and the leading cause of cancer related deaths among females of the Western world [
1]. Patients do not die from the primary tumor, but from metastases, which already are resistant or acquire resistance to systemic therapy. Metastasis is a complex, multi-step process in which malignant cells undergo sequential molecular changes helping them to disengage from primary sites, intravasate into blood vessels, extravasate to distant organs and finally colonize secondary sites. Each of these metastatic steps is affected by aberrant expression of a variety of transcription factors and among them, TWIST homologue 1
(TWIST1) is considered an important regulator of disease progression [
2].
The TWIST1 protein, encoded by the
TWIST1 gene, is a member of a large protein family called basic helix-loop-helix (bHLH) transcription factors [
3]. Most family members contain a bHLH domain, which enables it to target specific DNA sequences and thereby allowing them to regulate developmental processes in many organs and tissues.
TWIST1 plays a key role in the regulation of embryogenesis, gastrulation and mesoderm formation during early embryonic development of Drosophila and many other species [
4,
5]. An autosomal mutation pattern in the
TWIST1 gene leads to Saethre-Chotzen syndrome, a genetic condition characterized by premature fusion of skull bones affecting symmetrical growth of the head and face [
6]. In children, TWIST1 protein is involved in adequate maturation of the skull and spine bones and normal development of arms and hind legs.
More recently, TWIST1 protein has been implicated in various carcinomas, including breast cancer, where it plays a role in metastasis through activation of a biologically latent developmental process called epithelial to mesenchymal transition (EMT) [
7,
8]. In the EMT process, malignant epithelial cells undergo cytoskeletal changes, including the down-regulation of epithelial markers, such as E-cadherin and co-expressed catenins and up-regulation of mesenchymal markers, such as vimentin, N-cadherin and fibronectin. EMT transformed malignant cells are more motile and can be more efficient in invading the surrounding tissues and as a result metastasize to distant organs [
9].
In this large retrospective study of 1,427 primary breast cancer patients, we determined whether the
TWIST1 gene expression level is a prognostic marker. To avoid possible confounding effects of therapy and to study the natural course of the disease, we particularly focused on the subgroup of 778 lymph node-negative (LNN) patients who did not receive any adjuvant systemic therapy. Additionally, to understand the biological context of TWIST1, we have identified, using available Affymetrix U133A gene expression data [
10,
11], genes and biological pathways co-expressed with
TWIST1. By doing this, we identified a clear link between the tumor microenvironment and
TWIST1 expression in clinical breast cancer .
Discussion
The bHLH transcription factor TWIST1, which is essential in developmental processes, such as gastrulation, has been shown to be oncogenic in various cancers [
26‐
33]. Functional studies pin-pointed a pivotal role for this protein in EMT, a fundamental biological process considered to be crucial for metastatic spread in various carcinomas, including breast cancer. This prompted us to analyze mRNA expression of this gene in primary breast tumors from a large retrospective cohort of patients with complete clinical follow-up and address two basic questions. First, can
TWIST1 mRNA expression in breast cancer tissue predict disease progression? Second, as we had Affymetrix U133A gene-chips data available for a subset of samples [
10,
11,
22], can we reveal which biological pathways are co-expressed with
TWIST1 in breast cancer?
Our current study shows that high
TWIST1 mRNA expression is an independent marker of poor outcome in breast cancer patients. As
TWIST1 mRNA expression was also a predictor of poor outcome in patients with LNN disease who did not receive any adjuvant systemic treatment, we may conclude that
TWIST1 mRNA expression is a pure prognostic marker and is associated with the natural course of disease progression. This suggests that tumors with high expression of
TWIST1 are more aggressive, a finding that fits with experimental data where overexpression of
TWIST1 resulted in increased extravasation during the dissemination of tumor cells [
7]. Also in line with this is the increased lymph node involvement of breast tumors exhibiting higher
TWIST1 expression as observed by us and others [
34]. The first seminal paper on
TWIST1 in breast cancer also showed higher
TWIST1 mRNA expression in infiltrative lobular carcinoma [
7], which we also confirmed, though the implications of this remain unclear.
What we further unexpectedly observed was that the association of
TWIST1 expression with poor prognosis was confined to ER-positive cases and not seen in ER-negative breast cancer. Similar observations were recently made by Van Nes and colleagues [
21]. This group, by studying
TWIST1 expression in a smaller-sized cohort using IHC on tissue microarrays, also found significantly more cumulative relapses in the ER-positive subgroup of tumors. This observation was not significant in their entire cohort, which also included ER-negative cases. Using the same antibody, we observed, in a subset of formalin-fixed specimen available from our cohort, clear nuclear TWIST1 expression. This protein expression was not only confined to the nuclei of invasive tumor cells but also present in nuclei of other cells, such as fat cells, fibroblasts, endothelial cells and inflammatory cells. Furthermore, although based on small numbers, TWIST1 protein staining was high in those specimens in which we had measured high
TWIST1 mRNA expression in a matched fresh-frozen sample of the same tumor, while in specimens with low
TWIST1 mRNA expression the protein staining was overall weaker or absent.
To study the potential biological mechanisms in disease progression connected to
TWIST1 expression, we identified genes co-expressed with
TWIST1 in LNN/ER-positive tumors from which genome wide gene expression data were available. Three related pathways were revealed. The first pathway was "Focal adhesion", which involves interaction of ECM with integrin-A and -B ligands. The second pathway, "ECM-receptor interaction", is related to ECM and involves collagen, laminin, fibronetic, tenascin and integrin-A and -B proteins. The third pathway is the "TGF-beta signaling". How these pathways, involving well-known cancer-related genes, are connected to
TWIST1 expression in relation to disease progression in LNN/ER-positive cancer warrants functional validation. A common denominator among these three pathways, however, is their involvement in ECM remodeling. There is a growing body of evidence showing that changes in the tumor microenvironment can promote proliferation of the tumor cells and, therefore, can influence outcome of patients [
35], and our data connect this to
TWIST1 expression. Furthermore, our work also shows that the stroma is a significant source of
TWIST transcript expression. Indeed, the current data reinforce recent observations made by Roman-Perez and colleagues after evaluating gene expression patterns in 72 breast tissue samples [
36]. They identified two distinct subtypes, Active and Inactive, in the cancer-adjacent extra-tumoral microenvironment of breast tissues and demonstrated that
TWIST1 was highly expressed along with other stromal associated genes, such as
VIM, ADAMTS2, COL4A2, COL4A1, ITGA7 and
ITGB1, in the Active subtype. This could suggest a potential involvement of TWIST1 in the activation of stroma in malignant breast tissue. Furthermore, the Active subtype was associated with expression of TGF-beta induced fibroblast activation signatures, and showed a strong association with OS among specifically ER-positive patients. Intriguingly, also in our data set, the association of
TWIST1 with the TGF-beta signaling pathway appears to be more prominent in ER-positive tumors.
Another recent study also demonstrated that TWIST1 protein can be found in both compartments of malignant breast tissue, with a predominant expression in the stromal compartment [
37]. More importantly, they found that, although rare in their hands, nuclear expression of TWIST1 protein in the epithelial tumor cell compartment to be associated with poor prognosis (OS). TWIST1 protein expression in the stromal compartment was not associated with OS, but positively associated with a positive ER or PGR status of the tumors. This mutual dependence of TWIST1 expression on the presence of a notable stromal component and an ER-positive status of the tumor tissue in our and other studies is in line with the observed prognostic value of
TWIST1 in ER-positive disease only. Although our study is a correlative study and cannot reveal causality, it is intriguing and points to a novel role of
TWIST1 in breast cancer.
What is surprising about our results, which associate high
TWIST1 mRNA expression with poor prognosis in ER-positive breast cancer and with ECM and stromal content, is the fact that up until now the process of EMT has been more frequently assigned to ER-negative cancer and was rarely observed in ER-positive breast cancer. Also, previously studied breast cancer cell lines with clear EMT-like features were ER-negative [
38]. This left us with an apparent contradiction. We, therefore, specifically explored whether markers of EMT present on gene expression arrays [
11] were associated with
TWIST1 expression. Of the eight markers analyzed, we found Zinc finger E-box-binding homeobox 1 (
ZEB1), Zinc finger E-box-binding homeobox 2
(ZEB2), Snail2 (
SNAI2), Smooth muscle actin, also known as Alpha-actin-2 (
ACTA2) and
VIM clearly, positively correlated with
TWIST1 expression (Rs = 0.34, 0.34, 0.35, 0.42 and 0.28, respectively; all
P-values <0.001; Additional file
3: Table S15). This does suggest, in line with the recent study of Roman-Perez and colleagues [
28] that markers of EMT are co-expressed with
TWIST1. However, the lack of a clear negative association of
E-cadherin mRNA (
CDH1) with
TWIST1 mRNA expression in our data set could indicate that EMT may not be fully established. Alternatively, and more in line with our findings, is that some markers of EMT, such as Vimentin and Smooth muscle actin, are also expressed in re-activated stroma [
39]. Therefore, our findings are a starting point for more functional studies to be certain about TWIST1 possible role in EMT as well as in re-activated stroma in ER-positive breast cancer.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
AMS, JAF and JWMM designed the study. MR, JAF and JWMM wrote the manuscript. AMS performed the mRNA expression studies. MAT performed the immunohistochemistry. MR, AMS, JAF and JWMM analyzed the data. MR, MPL and MS did the statistical data analyses. All authors approved the final version of the manuscript.