Background
Despite significant advances in diagnosis and treatment of early breast cancer, almost 30% of patients will eventually have local or distant recurrence [
1‐
3]. Recurrence is considered to result from cancer cells that persist after surgery and systemic therapy and remain in a dormant state for many years before they start proliferating and form local or distant metastases [
4,
5]. Strategies to improve the management of patients with early disease should include the development of novel biomarkers for the early recognition of patients at high risk of relapse.
Clinicopathological parameters are commonly used for the prediction of patients’ prognosis; however, they often lack individualized validity for the identification of patients at high risk, due to significant inter-patient heterogeneity [
6]. In addition, molecular profiling tests have been developed for prognostication but their routine clinical implementation is problematic [
7]. Furthermore, the genetic profiling of solid tumors is currently performed on biopsies that might fail to reflect intra-tumoral heterogeneity and limit the opportunity to track genetic alterations occurring during cancer evolution [
8]. Therefore, there is an unmet need to identify novel non-invasive biomarkers for the better prediction of the risk of recurrence in breast cancer.
MicroRNAs (miRNAs), a large family of small (20–22 nucleotides) non-coding RNAs, regulate approximately 30% of the genes in the human genome at the post-transcriptional level, by binding to the complementary sequences of the 3′- untranslated region (3’-UTR) of their target messenger RNAs (mRNAs), leading to either mRNA degradation or inhibition of protein translation [
9]. miRNAs are deregulated in cancer, acting as both oncogenes and tumor suppressor genes [
10]. The altered expression of miRNAs has been associated with poor clinical outcome in patients diagnosed with a variety of tumors [
11]. In the past decade miRNAs have emerged as promising biomarkers in breast cancer and have been increasingly identified in biological fluids such as serum or plasma as circulating miRNAs [
12]. Circulating miRNAs are significantly stable in biological fluids [
13,
14] and could potentially serve as a “liquid biopsy” for the real-time evaluation of tumor status.
The assessment of dormancy and metastasis-related miRNAs could be of importance for the identification of patients at high risk of relapse. The mechanisms that lead to dormancy or enable the formation of metastases remain poorly understood. Data from in vitro models or expression analysis in patients with breast cancer suggest that miR-21, miR-23b, miR-190 and the miR-200 family members, such as miR-200b and miR-200c, are important in cancer dormancy and metastasis. An epithelial to mesenchymal transition (EMT)-related gene signature in the primary tumor has been associated with both stromal activation and escape from dormancy in breast cancer [
15], suggesting that intrinsic EMT features may regulate the transition of disseminated tumor cells into a dormant phenotype with the ability to outgrow as recurrent disease. In another report, the activation of the EMT program, as orchestrated by the key regulator of EMT, Zeb1, was sufficient to promote escape from latency and stimulate the development of metastases [
16]. The miR-200 family regulates EMT by targeting the ZEB1/2-E-cadherin axis [
17], whereas in other studies, elevated levels of miR-200 family have induced EMT and promoted metastasis in breast cancer [
18]. Several lines of evidence suggest that miR-21 is oncogenic in various types of cancer by suppressing several apoptotic and tumor suppressor genes [
19] and by inducing cell proliferation, migration, invasion and metastasis. miR-23b has been shown to promote tumor dormancy in the metastatic niche [
20], whereas miR-190 upregulation has been associated with prolonged tumor dormancy in fast-growing tumors such as osteosarcomas and glioblastomas [
21].
Based on the above, the aim of the present study was to investigate the expression of miR-21, miR-23b, miR-190, miR-200b and miR-200c in the plasma of patients with early breast cancer and evaluate their role in the prediction of patients’ outcomes.
Discussion
An important area in current breast cancer research is the identification of novel biomarkers for the prediction of outcome in patients with early disease. In the present study we investigated the predictive capacity of the dormancy and metastasis-related miR-21, miR-23b, miR-190, miR-200b and miR-200c when determined in the plasma of patients with early breast cancer. We found that miR-21, miR-23b, miR-190 and miR-200c, evaluated before the initiation of adjuvant therapy, were differentially expressed among patients who subsequently experienced disease recurrence, compared to patients who did not relapse. High expression of miR-21 and miR-200c was associated with shorter DFS compared to patients with low expression, whereas high miR-21 was also associated with shorter OS. Interestingly, miR-21, miR-23b, miR-190 and miR-200c discriminated patients who relapsed from non-relapsed patients. The combination of miR-21, miR-23b and miR-190 in ROC curve analyses had higher sensitivity and specificity compared to each miRNA alone; accuracy was further improved by adding lymph node infiltration and tumor grade to the panel of three miRNAs. Furthermore, the combination of miR-200c, lymph node infiltration, tumor grade and ER status predicted late relapse.
In breast cancer, clinically detectable metastases emerge after a period of dormancy and can last for varying and frequently prolonged periods of time. As miRNAs regulate tumor progression and metastasis we hypothesized that dormant tumors could be distinguished from faster-growing tumors by the differential expression of miRNAs [
26]. We show for the first time that miR-190 expression was lower in patients with early relapse, suggesting a potential role for this miRNA in sustaining tumor dormancy in breast cancer. Indeed, miR-190 was among the most upregulated miRNAs in a dormancy-related miRNA signature [
21]. miR-190 is involved in the regulation of the transforming growth factor (TGF)β pathway and in breast cancer TGFβ has been shown to promote bone and lung metastases [
27,
28]. Thus, miR-190 could induce tumor dormancy through the modulation of TGFβ signaling [
29].
Previous studies showed that miR-23b induced dormant phenotypes in a bone marrow, metastatic, human breast cancer cell line, induced cell cycle arrest in glioma cancer stem cells and suppressed glioma cell migration and invasion [
20,
30,
31] . On the contrary, the miR-23b/27b/24 cluster correlated with increased metastatic potential in human breast cancer cell lines and was upregulated in lung metastases from breast cancer [
32]. Moreover, high miR-23b/27b/24 expression was associated with poor outcome in breast cancer [
33]. Our results demonstrate higher plasma miR-23b expression in patients who relapsed, indicating that it is more likely associated with the development of metastases in breast cancer. Interestingly, the mature sequence of miR-23a differs by just one nucleotide in comparison to its paralog miR-23b, therefore they could share the same putative target genes and similar biological functions. However, there are reports showing distinct function between miR-23a and miR-23b and in contrast to miR-23b, we detected no variations in miR-23a expression levels among the different patient cohorts [
34,
35].
Various preclinical studies have established that miR-21 is involved in tumor growth, invasion and migration, extracellular matrix modification and survival [
36]. In primary breast cancer, miR-21 expression is associated with tumor progression, advanced clinical stage, lymph node metastasis and poor patient outcome [
37,
38]. In support of the tumor-promoting role of miR-21, serum miR-21 distinguishes patients with breast cancer from healthy controls and patients with distant metastasis from those with locoregional disease, and it is associated with poor prognosis in breast cancer [
36,
39,
40]. Accordingly, we show that high circulating miR-21 discriminated between patients with early breast cancer who relapsed and those who remained disease-free and specifically, high expression was associated with late relapse. Importantly, patients with high plasma miR-21 expression levels had worse DFS and OS compared to patients with low expression, whereas high miR-21 also emerged as an independent predictive factor for shorter DFS (
p = 0.003). Iorio et al., demonstrated that the TGFβ gene was a target for miR-21 and Yan et al. showed that TGFβ1 and the receptor TGFβR2 were identified among the putative target genes of miR-21 [
37,
38]. These data suggest that the tumor promoting effects of miR-21 in breast cancer, could be exerted through the regulation of TGFβ signaling.
The miR-200 family (miR-200a, miR-200b, miR-200c, miR-141 and miR-429) has opposing roles in the regulation of EMT and metastasis [
41]. On one hand, they negatively regulate the E-cadherin transcriptional repressors ZEB1/2 preventing EMT and on the other, they have been associated with global shifts in gene expression which promote metastatic colonization in breast cancer mouse models [
17,
42]. Conflicting results have been also reported on the clinical relevance of miR-200 family members in breast cancer [
43,
44]. By adopting a global profiling approach, Madhavan et al. showed that miR-200b and miR-200c were among the panel of six miRNAs with significantly increased expression in patients with early breast cancer who developed metastases [
45]. Our results also support the association between the plasma miR-200 family and metastatic progression in breast cancer. Importantly, high miR-200c was associated with late relapse and emerged as an independent prognostic factor for worse DFS (
p = 0.037).
ROC curve analysis confirmed the value of the plasma miRNAs in the prediction of disease recurrence in breast cancer. The combination of miR-21, miR-23b and miR-190 had higher accuracy compared to each miRNA alone. Moreover, the addition of common clinicopathological prognostic factors further improved the discriminatory capability of the three miRNAs. These results provide novel opportunities for breast cancer therapeutics employing the aforementioned miRNAs in a combinatorial miRNA approach [
46]. From a network analysis perspective, further insights might be achieved through the incorporation of information on the expression of the protein-coding mRNA associated to the involved miRNA. The formulation of a model of intervention efficiency based on a combination of miRNA, their gene targets and associated pathways would thus provide complementary information orthogonal to the one obtained from pathological characteristics.
In breast cancer, late relapses are common and impose considerable concern among disease-free patients, and there are no accurate tools to identify patients at risk. Importantly, in our study miR-200c expression combined with the clinical information on axillary lymph node status, tumor grade and ER status yielded an AUC of 0.89 with sensitivity of 75% and specificity of 89% for the prediction of late relapse (p < 0.001).
Our study is among the first to demonstrate the potential of metastasis-promoting miRNAs to serve as circulating predictive markers in early breast cancer. Importantly, (a) this patient cohort had long-term follow up, (b) plasma samples and clinical information were obtained prospectively, (c) the prediction of relapse was possible years before metastasis emerged and (d) circulating miRNAs added independent predictive value to common clinicopathological parameters. Furthermore, we considered pre-analytical and analytical parameters very carefully, taking into account the variables that could lead to bias in miRNA quantification [
22,
47].
Limitations of our study include that results are derived from the analysis of a relatively small group of patients and lack validation in an independent cohort. However, by performing cross-validation analysis of our data [
48], the predictive performance of the aforementioned miRNAs was confirmed, therefore it could probably be verified in an independent dataset. Nevertheless, our results should be viewed as preliminary and warrant prospective validation in a larger cohort of patients with early disease.