Background
Breast cancer (BC) is the most common malignant pathology affecting women worldwide [
1‐
3]. As BC accounts for an increasing number of deaths each year, efforts are being made to develop more efficient methods for early diagnosis, stratification and prediction of therapy response. The complexity of this disease comes from the diversity of environmental factors along with various inhered or acquired genomic, transcriptomic or proteomic alterations [
4]. In general, BC is classified based on the expression levels of estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 protein (HER-2). Triple negative breast cancer (TNBC) represents about 15–20% of BC cases [
5,
6], and is characterized by the absence of ER, PR and Her-2 proteins [
2,
4,
5]. This BC subtype poses major clinical challenges due to the lack of specific diagnostic/prognostic biomarkers and the failure of standard therapy to provide a targeted effect [
2,
6‐
8].
MicroRNAs (miRNAs) are short noncoding RNAs of about 19–25 nucleotides in length [
9‐
11]. MiRNA profiling studies have identified specific miRNA signatures in a wide range of cancer types [
12‐
14]. These transcripts can either be overexpressed (oncomiRs) or underexpressed (tumor suppressor miRs) [
4,
11,
15,
16]. These alterations are specific for each malignancy, including various BC subtypes [
4,
5,
7,
9,
15‐
18]. Thereby, circulating miRNAs are potential biomarkers in the case of numerous diseases [
19], such as BC [
15,
20‐
22]. The studies undertaken to prove the causative effect of miRNA first perform a general profiling of clinical samples, then are followed by controlled experiments [
22‐
26]. Still many questions remain regarding the exact mechanisms, biological functions, and clinical implication of miRNAs in the BC subtypes [
11,
17,
21].
The Cancer Genome Atlas (TCGA) is a large database of sequencing results generated from studies involving genome analysis in a rigorous and consistent manner [
27]. This allowed us to perform a direct comparison between the TCGA data and the results from our PCR-array plasma profiling study of TNBC and DPBC. We evaluated a panel of miRNAs related to BC and we identified the most specific miRNAs for TNBC and DPBC. The validation was done in a new independent patient cohort with the help of qRT-PCR technology. Furthermore, by overlapping the miRNA patterns, we identified either common or specific miRNA signatures for the two selected subtypes of Her-2 negative BC. Based on the expression level of the transcripts, miRNAs survival curves were generated. The results revealed the prognostic potential of some miRNAs, as well as their interdependence with some metastasis related genes.
Methods
TCGA miRNA expression pattern evaluation
We downloaded level 3 TCGA data from the University of California Santa Cruz cancer genomics data portal in the form of data matrices documenting patterns of miRNA expression for 112 TNBC tissue samples, 358 DPBC tissue samples, and 44 normal tissues (Table
1). Differential expression analysis was performed using the GeneSpring GX software from Agilent Technologies. The volcano plot module was applied, using a fold change > 1.5 and a
p-value of < 0.05. An additional validation step was performed for miR-200b in normal (
n = 19), DPBC (
n = 47) and TNBC (
n = 21) tissues (Table
2), in order to sustain the plasma expression profiling and the TCGA data, displayed as Pirate Plot generated in R programme.
Table 1
TGGA patient cohort characteristics
Sex |
Males | 0 | 3 |
Females | 112 | 355 |
Age |
Median, Range | 54, 29–90 | 58, 28–90 |
Median, Range ♂ | – | 68, 44–84 |
Median, Range ♀ | 54, 29–90 | 58, 28–90 |
Menopausal status |
Pre-menopausal | 30 | 89 |
Peri-menopausal | 5 | 16 |
Post-menopausal | 68 | 225 |
Unknown, N/A | 9 | 28 |
TNM |
T1 | 27 | 110 |
T2 | 70 | 189 |
T3 | 11 | 48 |
T4 | 4 | 10 |
Tx | – | 1 |
N0 | 72 | 168 |
N1 | 25 | 121 |
N2 | 11 | 39 |
N3 | 4 | 25 |
Nx | – | 5 |
M0 | 95 | 308 |
Mx | 17 | 50 |
Turmor grade |
I | 20 | 72 |
II | 70 | 195 |
III | 18 | 82 |
IV | 1 | 3 |
X / unknown | 3 | 6 |
Table 2
Clinical characteristic of patients with TNBC and DNBC patient cohort for PCR-array screening profile and plasma qRT-PCR validation lot
TNBC |
1 | T4bN1 M0 | 56 |
2 | T2N0M0 | 59 |
3 | T4bN2Mx | 40 |
4 | T2N0M0 | 52 |
5 | T2 N1 M0 | 46 |
6 | T2N0M0 | 53 |
7 | T2 N1 M0 | 56 |
8 | T3 N1 M0 | 46 |
9 | T4bN1 M0 | 57 |
10 | T3 N1 M0 | 50 |
11 | T4bN2Mx | 57 |
12 | T4bN2M0 | 55 |
13 | T2 N1 M0 | 35 |
14 | T4cN2Mx | 59 |
15 | T2 N1 M0 | 48 |
16 | T4bN1 M0 | 50 |
17 | T2 N1 M0 | 51 |
18 | T3 N1 M0 | 59 |
19 | T3 N1 M0 | 45 |
20 | T4bN1 M0 | 56 |
21 | T3 N1 M0 | 53 |
DPBC |
1 | T2N1aMx | 59 |
2 | T2 N1 M0 | 69 |
3 | T3N1Mx | 60 |
4 | T2N0Mx | 39 |
5 | T4bN3aMx | 73 |
6 | T2N0M0 | 49 |
7 | T2N0Mx | 42 |
8 | T3N1Mx | 58 |
9 | T2 N1 M0 | 41 |
10 | T1N0Mx | 67 |
11 | T4bN1 M0 | 66 |
12 | T3N1Mx | 52 |
13 | T2N2aMx | 57 |
14 | T4bN1 M0 | 52 |
15 | T1N0Mx | 42 |
16 | T4bN1 M0 | 38 |
17 | T2N1Mx | 62 |
18 | T2N0M0 | 46 |
19 | T3N0M0 | 57 |
20 | T2N2aMx | 48 |
21 | T2 N1 M0 | 64 |
22 | T3N1Mx | 63 |
23 | T2N0M0 | 62 |
24 | T4N3bMx. | 70 |
25 | T2N0M0 | 62 |
26 | T3N1aMx | 66 |
27 | T1N0M0 | 69 |
28 | T3N1Mx | 45 |
29 | T2 N1 M0 | 44 |
30 | T3N1Mx | 36 |
31 | T2N0M0 | 42 |
32 | T3N0Mx | 47 |
33 | T2N1Mx | 47 |
34 | T2 N1 M0 | 41 |
35 | T4N2Mx | 51 |
36 | T2 N1 M0 | 44 |
37 | T4N2Mx | 45 |
38 | T3N0Mx | 37 |
39 | T4N2Mx | 73 |
40 | T3N1Mx | 40 |
41 | T4N2Mx | 49 |
42 | T4N1Mx | 56 |
43 | T3N3Mx. | 80 |
44 | T3N3Mx. | 49 |
45 | T4N2Mx | 59 |
46 | T3N0Mx | 49 |
47 | T2 N1 M0 | 59 |
Survival analysis for the TCGA patients
We extracted the patient survival data from the TCGA clinical information file. In the case of miR-200b, miR-200c, miR-210, and miR-29, the survival was estimated in days from the date of diagnosis until date of last contact. Survival analysis was performed by using Kaplan Meier curves, in the GraphPad Prism program. In addition, we assessed the correlation of miR-200b to the most relevant metastatic markers, as described in literature [
28,
29].
Sampling procedures
The sampling for all biological specimens was done after we received the approval from the Oncology Institute “Prof. Dr. Ion Chiricuta” Ethics Committee and the informed consent form signed by the patient. The patients were diagnosed at the Oncology Institute “Prof. Dr. Ion Chiricuta” in Cluj-Napoca, Romania. The clinical characteristics of patients are presented in Table
3. The blood samples were collected from patients with TNBC or DPBC prior to treatment, between November 2010 and August 2013. In addition, blood samples from eight healthy female controls, free of any chronic diseases, were obtained in the second half of 2013. Sampling for all biological specimens was performed according to Romania’s laws and accompanied by an informed consent signed by every donor. The peripheral blood samples were collected in 3 ml tubes with EDTA for plasma isolation, and prepared by centrifuging the blood at 3000× rpm for five minutes. The plasma supernatant was carefully removed, placed in 2 ml Eppendorf tubes, and stored at − 80 °C. The qRT-PCR for miRNA-39 was used as quality control for extraction efficiency and as an indicator of miRNA recovery rate from plasma.
Table 3
Clinical characteristic of patients with TNBC and DNBC patient cohort for PCR-array screening profile and plasma qRT-PCR validation lot
PCR-array plasma |
TNBC |
1 | T4bN3M0 | 58 |
2 | T2 N1 M0 | 47 |
3 | T3N2M0 | 59 |
4 | T4bN1 M0 | 51 |
5 | T3 N1 M0 | 45 |
6 | T4bN2M0 | 51 |
7 | T2 N1 M0 | 51 |
8 | T2 N1 M0 | 56 |
9 | T4bN2M0 | 43 |
10 | T2 N1 M0 | 35 |
11 | T2N2Mx | 53 |
12 | T2 N1 M0 | 40 |
13 | T4cN2Mx | 59 |
14 | T4bN2M0 | 55 |
15 | T1N0M0 | 48 |
16 | T1 N1 M0 | 56 |
17 | T2N2Mo | 54 |
18 | T4bN2Mx | 40 |
19 | T2N0M0 | 52 |
20 | cT2 N1 M0 | 59 |
DPBC |
1 | T2N2M0 | 54 |
2 | T2N2M0 | 52 |
3 | T4bN2Mx | 72 |
4 | T4bN2M0 | 62 |
5 | T3 N1 M0 | 62 |
6 | T2N1Mx | 52 |
7 | T2N1Mo | 51 |
8 | T2 N1 M0 | 45 |
9 | T3N0Mx | 43 |
10 | T3 N1 M0 | 57 |
11 | T2N0M0 | 48 |
12 | T1N0M0 | 56 |
13 | T4aN0M0 | 53 |
14 | T2N0M0 | 62 |
qRT-PCR plasma |
TNBC |
1 | T4bN1 M0 | 56 |
2 | T2N0M0 | 59 |
3 | T2N3cM0 | 58 |
4 | T2 N1 M0 | 57 |
5 | T2 N1 M0 | 46 |
6 | T2N0M0 | 53 |
7 | cT2N2M0 | 59 |
8 | T4bN2M0 | 73 |
9 | cT1N0M0 | 70 |
10 | T2N2Mx | 49 |
11 | cT4bN2M0 | 61 |
12 | cT4bN2Mx | 57 |
13 | cT2N1Mx | 74 |
14 | T2N0M0 | 53 |
15 | T2N0M0 | 34 |
16 | T2N1cM0 | 62 |
17 | T4bN2M0 | 46 |
18 | T1 N1 M0 | 38 |
19 | T3 N1 M0 | 40 |
20 | T2 N1 M0 | 35 |
21 | T2N0M0 | 36 |
22 | T2N0M0 | 37 |
23 | T2N0M0 | 34 |
24 | T2N0M0 | 36 |
DPBC |
1 | T2 N1 M0 | 54 |
2 | T2NoMo | 59 |
3 | T2 N1 M0 | 52 |
4 | T2N0M0 | 46 |
5 | T4bN2Mx | 60 |
6 | T3N1Mx | 63 |
7 | T2N0M0 | 67 |
8 | T4bN2M0 | 53 |
9 | T3N1Mx | 43 |
10 | T2N0M0 | 51 |
11 | T2 N1 M0 | 64 |
12 | T2N1Mo | 57 |
13 | T4bN2M0 | 45 |
14 | T3N0Mx | 69 |
15 | T2N1Mx | 52 |
16 | T2 N1 M0 | 44 |
17 | T2 N1 M0 | 55 |
18 | T2 N1 M0 | 62 |
19 | T1 N1 M0 | 49 |
20 | T3N1Mx | 40 |
21 | T3N0Mx | 45 |
22 | T3N1Mx | 60 |
23 | T4N2M0 | 63 |
24 | T4 N1 M0 | 50 |
25 | T2N1Mx | 65 |
26 | T2 N1 M0 | 60 |
27 | T2NoMo | 44 |
28 | T4bN2Mo | 47 |
miRNA isolation from plasma samples
Before use, plasma samples were thawed for five minutes on ice. Total circulating miRNAs were isolated from a 200 μl plasma aliquot using a commercially available column-based assay, according to the manufacturer’s instructions (Qiagen miRNeasy Serum/Plasma Kit). Spike-in control, containing lyophilized C. elegans miR-39 miRNA mimic was added to each sample, used as a PCR normalization control. In the final elution stage, 14 μl of RNase-free water were added to the membrane of the MinElute spin column. This was incubated for 1 min at room temperature and centrifuged at 1200 g for another minute. The isolated miRNA samples were stored at − 20 °C before processing.
PCR array analysis
To generate the cDNA, we used the miScript HiSpec Buffer and 2 μl of total RNA. The 20 μl amplification mixture was incubated at 37 °C for 60 min, then at 95 °C for 5 min. The cDNA was then diluted and mixed with the miScript miRNA PCR array kit, containing specific miRNA primers and QuantiTect SYBR Green PCR Master Mix. For the PCR array analysis, we worked with the 96-well Human Breast Cancer miScript miRNA PCR Array (SABiosciences), containing replicates for miRNA reverse transcription control assay (miRTC) and a positive PCR control (PPC). The plate contains probes for 84 miRNAs whose expression is known or expected to be altered in breast cancer. The miScript SYBR Green PCR Kit was used following the manufacturer protocol, with one exception: only half of the cDNA volume was used and therefore 50 μl of RNase free water was added at the total volume of the reaction mixture. For the PCR-array determination, the Roche LightCycler480 instrument was used, following the cycling conditions indicated by the producer.
The miRNA PCR-array data analysis is displayed as fold-change mean for TNBC group, compared with the healthy female controls. For the interpretation of data, we used a web analysis tool provided by Qiagen, USA (
https://www.qiagen.com/us/shop/genes-and-pathways/data-analysis-center-overview-page/), based on the ΔΔc
t method for the calculation of relative miRNA expression. The normalization was done with the help of the average Ct value and the reference expression of cel-miR-39, SNORD68, SNORD95, SNORD96A, RUN6–2.
qRT-PCR data validation
To perform data validation, samples from 28 healthy controls, 24 TNBC and 24 DPBC were analyzed. For the cDNA protocol, we took a total of 50 ng of isolated RNA and mixed it with the Taqman microRNA Reverse Transcription Kit (Cat. No. 4366596, Life Technologies) in a reaction volume of 7.5 μl. Then the following cycling parameters were utilized: 16 °C for 30 min, 42 °C for 30 min, 85 °C for 5 min. The qRT-PCR reaction was performed on the ViiA7 instrument (Applied Bio systems) by using 5 μl of SsoFast Supermix (Biorad cat no. 172–5230), 4.5 μl of 5X diluted cDNA and 0.5 μl of TaqMan Primer. The evaluated miRNAs were: miR-10a, miR-125, miR-193b, miR-200b and miR-489. For data normalization of miRNA expression levels, U6 was used. The same protocol was used for the miR-200b tissue validation. When normalizing this data set, we used U6, RNU48 and miR-16. The qRT-PCR cycle was set at: 98 °C for 3 min, 40 cycles of 95 °C for 15 s, 60 °C for 30 s. The data were analysed by applying the ΔΔCt method and presented as Pirate Plot using R.
Discussion
Despite the late transition from pan-genomics to the post-genomics era, BC still remains one of the main causes of cancer related deaths [
30]. TNBC is the most aggressive subtype of BC and it presents the worse clinical outcome among BC cases [
2]. As follows, there is undeniable need for the development of novel diagnostic/prognostic markers that may also constitute therapeutic targets. Over the last few years, different research teams have explored the variation of miRNA profiles in relation to its diagnostic or prognostic potential [
11,
21,
24,
31‐
33].
Certain miRNAs have a distinct expression profile specific for each BC subtype, which could prove to be a valuable diagnostic/prognostic tool. The bioinformatic analysis of the TCGA dataset is a powerful approach for characterizing miRNA expression patterns in large patients cohorts [
27]. This allowed us to perform a comparison between tissue and circulating miRNAs. A partial correlation with the literature data was observed, especially in the case of miR-200 family members. This correlation was confirmed in both tissue and plasma samples. Specific patterns of plasma miRNAs appear to have distinct roles in metastasis. Furthermore, they can be related to the EMT, to invasion, or to late metastatic events, such as the establishment of metastatic tumors. However, different miRNA profiling studies failed to reach a consensus regarding the local versus systemic levels.
The miR-200 family members are regarded as the main regulators of EMT, invasion and metastasis. Moreover, it was recently discovered that miR-200 s contribute to the angiogenic process by targeting VEGF-A and its receptors [
34,
35]. The inhibition of TGFβ receptor restores the normal ZEB/miR-200 balance and it leads to the overexpression of E-cadherin, resulting in reduced tumor dissemination [
36]. As follows, miR-200 family is considered an early biomarker of metastasis [
37,
38]. Our data supports this role of miR-200 as a general prognostic tool and a specific biomarker of early metastasis. This miRNA can be considered as a single evaluation tool or it can be correlated with the expression level of other coding or non-coding transcripts. Additionally, these other transcripts may function as direct or indirect targets, which can be seen in Fig.
6e.
The EMT process is considered as an efficient strategy adopted by epithelial cancer cells to promote local invasion and dissemination to distant organs [
29]. This is supported by our evaluation of the miR-200 as an important metastatic marker, with a particular correlation in lung metastasis. The TFF1 gene was negatively correlated with the expression level for miR-200b in both breast cancer subtypes, meanwhile RARA gene was negatively correlated only in TNBC. We integrated these metastasis associated genes in a complex regulatory network. This could prove to be a useful tool for further experiments studying the mechanism of their action or the way they affect the clinical therapeutic outcomes in these Her-2- BC subtypes (Fig.
6e).
MiR-130 overexpression in breast cancer is related to EMT, invasion and metastasis. In addition
, this microRNA is also connected with the downregulation of miR-200 [
39,
40]. MiR-130 is known to have an active role in angiogenesis by modulating the expression of VEGF [
41]. Another stand-out was miR-22, associated with poor clinical outcomes and the silencing of the TET-miR-200 axis in human breast cancer patients [
42]. This microRNA was found to be specific for TNBC, when compared with DPBC.
The miR-29 family members were downregulated in various types of cancers and have been recognized mainly due to their tumor suppressive roles [
43]. Lately, these molecules are presented as possible new biomarkers or therapeutic targets in BC, but with no direct implications in the TNBC pathogenesis [
44,
45]. What’s more, the altered plasma levels of miR-29c and miR-200 were suggested to promote brain metastasis [
46]. However, our results showed no correlation between the miR-200 expression level and the evaluated brain metastasis markers (BRCA1 and PARP1).
The miR-210 is another microRNA considered to have an effect over the clinical outcome of cancer patients [
47]. The overexpression of this microRNA is correlated with a higher proliferation rate of the cancer cells. For BC patients, it was associated with an unfavorable prognostic [
48], especially for Tamoxifen-treated patients [
49]. The miR-210 up-regulation was observed specifically in patients with unresected tumours, lymph node involvement and metastases [
50]. Some studies have established a correlation between miR-210 and the therapeutic response to Trastuzumab [
50]. The miR-210 expression in TNBC was significantly higher than in DPBC [
51]. A meta-analysis revealed that the increased level of miR-210 was related with a reduced overall survival [
52]. In our study, the overlap analysis based on the TCGA data confirmed the results from previous studies. The miR-210 expression levels are similar in the plasma as well as the tumor tissue in both TNBC and DPBC.
In order to provide a more comprehensive overview of the interaction established between miRNA and mRNA, we constructed an IPA network. This is a helpful step towards a better understanding of the carcinogenic mechanisms as well as affected cellular pathways in TNBC and DPBC. As it was previously mentioned, EMT is an essential step in the metastatic cascade, because it leads to the activation of invasion and migration (Fig.
4d). Our study revealed a panel of miRNAs related to EMT that could become non-invasive biomarkers.
In this study, further details were revealed regarding the molecular basis of miR-200b involvement in BC metastasis, which can become a future clinical tool for establishing a more accurate prognostic. Our results demonstrated the increased sensibility of combined miRNA signature or miRNA-gene interaction.
The process of implementing a miRNA-based biomarker remains a challenge, the main problem being represented by the small patient cohort and the lack of a standardized method for evaluation. In addition, we need to take into account some of the patient characteristics such as dietary habits, environmental exposure, immune status and age. In this context, one miRNA with an altered expression level does not automatically have an oncogenic or a tumor suppressive role.
Conclusion
We identified an aberrant miRNA expression pattern in the plasma of TNBC and DPBC patients. Our investigation found several miRNAs deregulated in the plasma of these patients, most of them being common for the HER2- subtypes of breast cancer. The miRNA specific signature for TNBC versus DPBC includes the downregulation of four miRNAs belonging to the miR-17-92 cluster (miR-17-5p, miR-20a, miR-20b, and miR-93), along with other miRNAs, such as miR-130, miR-22 and miR-29a/c. The overlap of circulating plasma and tissue miRNAs emphasizes the important role of miR-200b/c, miR-210 and miR-29c in TNBC tumorigenesis.
The regulatory mechanisms in cancer are more complex than one simple biomarker; miR-200b is a key element for the future answers given to the breast cancer mystery, especially considering that this microRNA is integrated in a regulatory network which acts in conjunction. As follows, not a single node, but the whole network affects the patient prognosis and response to therapy.
Nevertheless, the fluctuating levels of miR-200b provide a deep understanding over some of the mechanisms which drive the metastatic spread from the primary tumour. Controlling these EMT transcripts may increase the survival rate of the TNBC patients, due to their link with metastatic markers that promote cell adhesion, migration, and motility.
Further studies on a larger cohort of patients are needed to validate our findings. Also, much remains to be learned about the application of miRNA-based evaluation of treatment response and the early detection of recurrences.