Differential expression of the miR-17-92 cluster and miR-17 family in breast cancer according to tumor type; results from the Norwegian Women and Cancer (NOWAC) study

Breast cancer is a leading cause of cancer-related deaths. Annually, approximately 2.09 million women worldwide are diagnosed with breast cancer whereas an estimated 627.000 women die of the disease [1]. Breast cancer incidence has been increasing over the last decades in the Western world, also in Norway [2]. Mortality rates are falling, leaving an increasing number of women alive with a history of the disease, but also exposed to risk of complications and side-effects due to treatment and with a life-long risk of relapse. There is a need for simple, safe and informative diagnostic tools to better identify the breast cancer tumors with the most aggressive behavior and to diagnose and treat the disease before distant metastases have been established and the disease is beyond curability.

MicroRNAs (miRNAs) are short single-stranded RNAs built up of 18-22 nucleotides after processing of the pri-miRNA by the nuclear RNase III protein Drosha and sequentially cleavage of the hairpin-shaped precursor-miRNA by the RNase III Dicer in the cells’ cytoplasm [3]. MiRNAs are important regulators of gene expression at the post-transcriptional level, usually by either inhibiting translation or inducing mRNA degradation through incomplete or complete binding to a complementary sequence in the 3′ untranslated region (UTR) of their target mRNAs. It is well established that miRNAs are involved in carcinogenesis, invasion and metastasis [4, 5] and display distinct profiles in cancer [6]. Further, miRNAs have properties that make them promising as biomarkers. They can be detected in blood, partly in extracellular vesicles known as exosomes, and in other body fluids such as urine and saliva [7, 8]. MiRNAs are very stable structures and can tolerate freezing and thawing [9]. Further, if miRNAs in blood could give information on the phenotype or aggressiveness of a given tumor, there is a possibility that easily obtained samples could give information on malignant disease, both at the time of primary diagnosis and in the metastatic setting [10].

The Norwegian Women and Cancer study (NOWAC) is a prospective study which started in 1991 and includes 172,000 Norwegian women aged 30–70 years randomly sampled from the Norwegian Central Person Registry. The study is based on questionnaires with information on variables of importance to breast cancer risk such as lifestyle, use of oral contraceptives, hormone replacement therapy, reproductive history and family history of breast cancer. From 2003 the study was expanded to include blood samples for whole-genome expression profiling (the NOWAC postgenome cohort) [11]. 49,633 samples of peripheral blood have been collected. Through linkage to the Cancer Registry of Norway, women in the NOWAC postgenome cohort with a diagnosis of breast cancer have been identified.

The aim of this pilot study was to explore the miRNA expression profile in breast cancer tumors from the NOWAC postgenome cohort and to search for miRNAs that are significantly different in tumor tissue compared to benign breast tissue and could be detected in formalin-fixed paraffin-embedded (FFPE) tissue, collected as part of routine diagnostics. Further, we wanted to identify miRNAs that are differently expressed in tumors with different aggressiveness and prognosis with special focus on high grad tumors and the triple-negative breast cancers.

This study has explored the miRNA expression profile in breast cancer in the NOWAC study. The distribution of tumor characteristics such as receptor status, histologic grade and molecular subtype is as expected based on findings in larger epidemiological studies and the collected national data in the Norwegian Cancer Registry, again underlining that the NOWAC study population is a representative cohort of the Norwegian female population in this age-group [18].

Noteworthy, we found the miRNA expression profile to be significantly different in benign and malignant breast tissue, as illustrated by the principal component analysis. Of the 20 most differentially expressed miRNAs in microarray, presented in Fig. 3, only two miRNAs demonstrated higher expression levels whereas 18 miRNAs demonstrated lower expression levels in tumor compared to benign breast tissue. Of 14 miRNAs quantitated by PCR, 12 miRNAs were significantly differently expressed in breast cancer and benign tissue, of which seven miRNAs had significantly lower expression in breast cancer tissue. Tumors have often been found to have reduced levels of mature miRNA which could be explained by defects in their biogenesis, for instance through loss of key proteins in their synthesis such as DICER, epigenetic silencing through mechanisms such as promoter hypermethylation and/or genetic loss of miRNA loci [5, 19].

Among the most differentially expressed miRNAs, we found downregulation in tumor tissue of miRNAs such as miR-143-3p and miR-145-5p in the miR-143/145 cluster, miR-10b-5p, miR-99a and let-7a-2-3p which is in line with other studies comparing malignant and benign breast tissue [20‐25]. Among the 20 most deregulated miRNAs, we observed significantly higher levels of miR-4419b and miR-2964a-5p. MiR-4419b has been found to be upregulated in small cell tumors of the esophagus with rapid relapse after surgery [26] and miR-2964a in pancreatobiliary adenocarcinoma [27], but they are not functionally described in breast cancer. However, as established by miRNA profiling across cancer types, many of the deregulated miRNAs are common in different malignancies.

Analyses of the breast cancer cases revealed significant differences between cancer subtypes. ER is of special interest in breast cancer as a prognostic and predictive marker, and we found 23 miRNAs to be significantly different according to ER status. Among these, miR-155 has previously been shown to be upregulated in breast cancer compared to normal tissue and upregulated in ER− compared to ER+ tumors [22, 24, 28], as verified in our study. Noteworthy, miR-342-3p, which we found to be most significantly different with higher expression in ER+ compared to ER− tumors, has also been found by others to be strongly associated with ER+ status and to predict ER+ receptor status [29, 30].

Interestingly, when using an agnostic approach and exploring the miRNA expression profile in breast cancer tissue in an epidemiological study using microarray, four of the six miRNAs found to be most significantly different between tumors of different grade, were miR-17-5p, miR-20a-5p, miR-106b-5p and miR-93-5p, belonging to the miR-17-92 cluster and its paralogue miR-106b-25, and all members of the miR-17 family. Further analyses demonstrated that these miRNAs were significantly higher in tumors with high histologic grade and triple-negative status. Other studies have also found higher levels of miRNAs in these clusters in the most aggressive breast cancers using fresh-frozen tumor tissue and a bead-based flow cytometric miRNA expression method, analyzing on a smaller number of miRNAs than in our study [28]. Calvano Filho et al. found higher levels of miRNAs in the miR-17-92 cluster and miR-17 family in selected triple-negative tumors compared to luminal A breast cancers, using RT-PCR [31]. MiR-18a and -18b have been shown to have higher expression levels in ER− compared to ER+ tumors, and to directly target ERα [32].

Of note, when comparing the clustered miRNAs’ expression in breast cancer to benign tissue, we found significantly higher expression of miR-106b-5p and miR-93-5p in breast cancer tissue compared to benign tissue using PCR; by using microarray we found no significant differences. Higher levels of miR-17 family members such as miR-93, miR-25 and miR-106b in cancer have also been demonstrated in other studies using deep sequencing, microarray and/or PCR [33‐35]. Several of these studies were smaller than ours and used tissue adjacent to the breast tumor as benign tissue controls. As in our study, using microarray, others have also found members of the miR-17-92 cluster and miR-17 family such as miR-17-5p, miR-20a-5p and miR-92a-3p to be downregulated in solid cancers compared to benign tissue [19, 24]. We could, however, only validate this result for miR-92a-3p using PCR; for miR-17-5p and miR-20a-5p no significant difference was observed. Noteworthy, in the PCR-assays, a relative larger proportion of the cancers was triple-negative compared to the entire study cohort included in the microarray, underlining that differences between breast cancer subgroups could influence comparisons between cancer and benign tissue. Although our study includes more tumor and benign tissue samples compared to many other studies, our study material is still small, and the results must be interpreted with caution. Further, when performing expression analyses on tissue, one must be aware that the tissue cores contain other cellular elements contributing to the RNA pool such as immune cells, endothelial cells and fibroblasts, where malignant tissue would be expected to be more heterogenic compared to benign tissue. Hence, differences in miRNA expression between malignant and benign tissue could also, in part, be attributed to differences in tissue composition where tumor heterogeneity could influence the results [36].

Still, our results from breast cancer subgroup analyses, using both microarray and PCR, point to the miR-17-92 cluster and miR-17-family as overexpressed in aggressive breast tumors. The miR-17-92-cluster is located on chromosome 13 in the locus of the non-protein coding gene MIR17HG (miR-17-92 cluster host gene) and was first identified as the gene “chromosome 13 open reading frame 25” (C13orf25) found to be amplified in human B-cell lymphoma [37]. The cluster is transcribed as a polycistronic primary transcript that give rise to six mature miRNAs: miR-17, miR-18a, miR-19a, miR-19b, miR-20a, and miR-92a-1 [38, 39]. Transcriptional regulation of C13orf25 can be part of the molecular basis for the coordinated expression of cluster members as observed in our study. A correlation in expression of the individual miRNAs in the miR-17-92 cluster [29] and also correlation of expression of the miR-106b-25 cluster members and their host gene MCM7 on chromosome 7 has been shown [28]. The miR-17-92 cluster has been shown to be regulated by the transcription factor and proto-oncogene MYC which binds to the promoter region directly upstream of the miR-17 locus [40]. The cluster is highly expressed in a range of hematopoietic malignancies including MYC-rearranged Burkitt’s lymphomas [41]. High MYC-activation is also found in triple-negative breast tumors [42, 43] which could partly explain our findings of high miR-17-92-expression in triple-negative cancers. Similarly, the miR-17-92 promoter has binding sites for HES1 [39], a transcriptional repressor in the Notch signaling pathway which is overexpressed in triple-negative breast cancer [44]. N-myc has also been found to induce miR-17-92 expression in medulloblastomas [45]. NDRG2, N-myc downstream-regulated gene 2, has been found to be significantly higher in triple-negative breast cancers compared to other subtypes [46], again indicating that differences in expression and activity of transcription factors targeting the miR-17-92 promoter vary between breast cancer subtypes and can explain differences in miR-17-92 expression.

However, we also observed significant changes in the expression of miRNAs that are members of the paralogue cluster miR-106b-25 on chromosome 7 comprising miR-106b, miR-93 and miR-25. The miR-106b-25 cluster and its host gene, MCM7, as well as miR-20a in the miR-17-92 cluster, are induced by the transcription factors E2F1 and E2F3 which are regulated by MYC [47‐49]. The cluster has also been shown to be regulated by bromodomain protein 4 (BRD4) which is increased in MYC-driven cancers [50]. Similar to the miR-17-92-cluster, miR-106b-25 is transcriptionally regulated by N-myc [51]. Of note, the expression of the miRNAs in the clusters is also regulated by mechanisms such as epigenetic modifications induced by hypoxia [52], independent transcription of pri-miRNAs from an alternative promoter, alternative splicing [53], and post-transcriptional modifications of the long primary transcripts based on their tertiary structure [54, 55]. These mechanisms allow for differential expression of the individual miRNAs within the clusters and miRNA families. Further, differences in expression of DICER1, AGO and DROSHA, all crucial to miRNA biosynthesis, between breast cancer subtypes have been shown [28]. Summarized, the paralogous clusters seem to have important transcription factors and regulatory pathways in common. Indeed, it has been shown that the clusters are evolutionary conserved and it is suggested that they derive from a single gene that underwent duplication, mutations and losses of individual miRNAs [56, 57].

Further, clustered miRNAs seem to cooperate by regulating similar sets of genes belonging to specific signaling pathways [58] which fits with the sequence homology and conserved seed sequences of the miRNA within the clusters [59]. The miR-17 family of miRNAs share the same seed sequence of special importance for binding and targeting mRNAs, and include miR-17-5p, miR-20a-5p, miR-20b-5p, miR-106a-5p, miR-106b-5p and miR-93-5p from clusters miR-17-92, 106b-25 and the third paralogue cluster miR-106a-363 [38]. Of the three paralogues, involving four miRNA families, the miR-17-92 cluster is best described. MiR-17 and miR-19a have been shown to target mitogen activated kinases (MAPKs) such as extracellular signal-regulated kinase (ERK) 1/2, and key signaling molecules in the MAPK signaling pathway such as KRAS and RAF1 [60]. The MAPK signaling pathways regulate cellular proliferation, migration, differentiation and cell death and are dysregulated in many cancers, including breast cancer [61]. Note that both miR-17 and miR-20a have been shown to target E2F1, thereby taking part in a negative feed-back loop where the E2F transcription factors induce transcription of miRNAs that have the same transcription factors as their target [40, 49]. In addition, miR-17 and miR-20a targets the type II transforming growth factor β (TGF-β) receptor II whereas miR-18a targets Smad 4 downstream in the TGF-β signaling pathway, thereby opposing the tumor-suppressive effects of TGF-β and promoting angiogenesis [62]. Interestingly, miR-17-92 has also been found to target the cyclin-dependent kinase inhibitor p21 and the apoptosis facilitator BCL2L11 which are mediators of TGF-β effects on proliferation and apoptosis [57]. Noteworthy, BCL2L11 is also targeted by miR-106b-25, again underlining how the miRNA clusters share targets [63]. MiR-17-5p has been shown in cellular assays to play an important role in cancer cell invasion and migration by suppressing HBP1 and consequently Wnt/β-catenin [64]. The miR-17-92 cluster members can also suppress the specificity protein (Sp) repressor ZBTB4, which in turn facilitates upregulation of Sp transcription factors and their target genes, thereby displaying tumor promoting functions [65]. MiR-19 has been shown to excert oncogenic activity through binding to and repression of the tumor suppressor PTEN and activation of the Akt-mTOR (mammalian target of rapamycin) pathway to promote cell survival [66]. Further, the tumor suppressor p53 is a direct target of miR-25 [67]. MiR-25 has also been shown to promote proliferation in triple-negative breast cancer cells by repression of the BTG anti-proliferation factor 2 [68] whereas miR-106b has been shown to induce proliferation by targeting RB proteins in various cancers [69, 70]. MiR-93 has been shown to increase proliferation, migration and invasion potential of MCF7 breast cancer cells and to have many potential targets involved in tumor growth, including the large tumor suppressor homologue 2 (LATS2) [33]. In summary, the miR-17-92 cluster and the miR-17 family of miRNAs have been demonstrated to regulate functions at the very core of malignancy: invasion, metastasis, cellular proliferation and resistance to apoptosis.

Microarray miRNA profiling of human breast cancer has been demonstrated to be an informative tool which can be used to classify human breast cancers [19]. However, interpretation of research data, and implementation of miRNA-profiling into clinical practice are complicated by the apparent lack of consistency between studies. Variation in study size, design and experimental factors such as sample type, RNA quality, methods and technology platform used is challenging when trying to summarize clinically relevant data on miRNA profiling. Further, investigated miRNAs are not identical between studies. However, this study using FFPE-tissue, immunohistochemical analyses used in everyday diagnostics and a microarray with 94% coverage of human miRNAs, indicates that the miR-17-92 cluster and miR-17-family of miRNAs are of special interest in the high grade, triple-negative tumors with the worst prognosis. Further studies, including validation in an independent cohort, in situ hybridization in tissues and analyses of prediagnostic blood samples could be of interest to further evaluate the miRNA expression within tumor cells, the tumors’ microenvironment and in the circulation, and the potential of these miRNAs in diagnostic pathology and clinical oncology.

In the NOWAC postgenome cohort, archived FFPE tissue was found suitable for miRNA analyses using microarray and PCR technology. We found the miRNA expression to be significantly different between benign and malignant breast tissue. When exploring the miRNA expression profiles according to receptor status, histologic grade and molecular subtypes, we found a coordinated clear upregulation of miR-17-5p, miR-20a-5p, miR-92a-3p, miR-106b-5p and miR-93-5p of the miR-17-92 cluster and the miR-17-family in the most aggressive tumors with higher tumor grade and triple-negative receptor status. A validation of this pilot study in an independent cohort is a focus for future research. Further, the diagnostic and functional role of the miR-17-family of miRNAs in breast cancer should be further explored, where miRNAs’ potential as markers of response to specific treatments or as biomarkers of breast cancer relapse or metastatic disease is highly interesting.