Three 5’isomiRs of miR-183-5p are highly upregulated in breast cancer
Various previous studies have demonstrated the importance of miRNAs as biomarkers or drivers in cancer [43, 44]. However, these reports are mostly limited to investigations of canonical miRNAs, neglecting the functional relevance especially of 5’isomiRs. Therefore, we here aimed to characterize the 5’isomiR expression patterns in TCGA breast cancer patients. We employed the isomiR annotation introduced by Loher et al. [19] where the position and direction of the sequence shift of an isomiR is annotated as exemplified in Fig. 1a. Since 5’isomiRs frequently exhibit functional differences from their canonical counterpart due to a shift in the target-selective seed sequence, we focused on variants with altered 5’ends, neglecting the specific 3’end of an isomiR. For that purpose, all reads with the same 5’end were aggregated to focus on functional entities sharing the same seed sequence (Fig. 1a). To identify miRNAs and their 5’isomiRs with potential tumor-relevance, we analyzed miRNA sequencing data from the TCGA breast cancer dataset. We analyzed the differential expression between tumor and normal tissue samples for 294 5’isomiRs with an average expression of >15 RPM (Fig. 1b, Additional file 2: Supplementary Table 6). Out of these, 11 5’isomiRs were significantly upregulated in tumor tissue (black triangles, log2 fold change > 2, p-adjusted < 0.05) and six 5’isomiRs were significantly downregulated (black dots, log2 fold change < -2, p-adjusted < 0.05). Of note, all three 5’isomiRs of miR-183-5p were among the most significantly and most highly upregulated in tumor. The expression of the three 5’isomiRs of miR-183-5p was upregulated both, in triple-negative breast cancer and other subtypes compared to normal tissue in the TCGA dataset (Fig. 1c). Similarly, the expression of miR-183-5p was also significantly upregulated in both, TNBC and non-TNBC tumors as compared to benign breast tissue in the METABRIC cohort (Fig 1d).
Pre-miR-183 exhibits tumor-suppressive phenotypes in TNBC cell lines
Due to the upregulation of miR-183-5p and its 5’isomiRs, we hypothesized that the overexpression of these three 5’isomiR-183-5p would trigger tumor-promoting phenotypes. Since mature miRNAs and also isomiRs are generated from pre-miRNAs, TNBC cell lines with inducible overexpression of pre-miR-183 were generated to investigate the role of miR-183-5p and its 5’isomiRs. Expression of pre-miR-183 was increased in pre-miR-183 overexpressing MDA-MB-231 and BT-549 compared with two different pre-miRNA negative controls overexpressing TNBC cell lines as confirmed by miScript qRT-PCR (Additional file 2: Supplementary Table 7). Using these model systems, we firstly investigated the effect of pre-miR-183 on cell migration and invasion. In contrast to our initial hypothesis, cell invasion and migration were significantly reduced in both MDA-MB-231 and BT-549 cell lines stably overexpressing pre-miR-183 (Fig. 2a-b).
To dissect which isoforms were responsible for the reduction in cell migration and invasion caused by the overexpression of pre-miR-183, synthetic miRNA mimics were used to overexpress the individual isomiRs. In both TNBC cell lines, the overexpression of all three 5’isomiRs of miR-183-5p reduced cell migration and invasion compared to controls (Fig. 2c-d). To further confirm the reduction in cell migration and invasion in vivo, we injected MDA-MB-231 cells stably overexpressing pre-miR-183 into the mammary gland fat pad of NSG-mice and monitored primary tumor growth along with lymph node and lung metastasis. While three out of six mice in each control group had enlarged lymph nodes, no enlarged lymph nodes were detected in mice harboring tumors overexpressing pre-miR-183, indicating that indeed metastasis to the lymph nodes might be impaired. To further investigate metastasis to distant organs, DNA was isolated from the lungs of the mice and used for Alu qPCR to detect human DNA indicative of micro-metastases in the lungs. There was a trend of increased Ct values indicating decreased metastasis in mice harboring tumors overexpressing pre-miR-183. However, no significant difference in tumor metastasis to lung was observed in our experimental setup (Additional file 1: Supplementary Figure 1).
MiR-183-5p|+2 is functionally different from the other two isoforms
While we could not conclusively confirm a role of pre-miR-183 in metastasis in our descriptive set-up of the mouse experiment, we observed a trend towards reduced tumor growth upon pre-miR-183 overexpression as indicated by reduced tumor volumes 33 days after injection (Additional file 1: Supplementary Figure 1b and c). This was also underlined by a trend of prolonged survival, i.e. the time to pre-defined humane endpoints, in the pre-miR-183 overexpression group, compared to the controls (Additional file 1: Supplementary Figure 1d). This difference was statistically significant when compared to control 1, but not compared to control 2. Yet, we decided to validate a potential role of pre-miR-183 in tumor cell proliferation. For that purpose, we next investigated the impact of synthetic miRNA mimics on cell viability and cell cycle in vitro. A microscopy-based cell counting assay was utilized to unravel the role of the three 5’isomiRs of miR-183-5p on cell proliferation in three TNBC cell lines, namely MDA-MB-231, HCC-1806 and BT-549. Of note, the overexpression of miR-183-5p|+2 strongly reduced the number of viable cells compared to siAllstar as miRNA mimics negative control for all tested cell lines (Fig. 3a-c). The other two 5’isomiRs, miR-183-5p|0 and miR-183-5p|+1, had slightly milder effects on cell viability. Mechanistically, this was partly caused by increased rates of cell death triggered especially by miR-183-5p|+2. Specifically, we investigated the percentage of cells taking up propidium iodide which is considered as a sign of disintegration of the cytoplasmic membrane during cell death. Here, we observed that cells overexpressing miR-183-5p|+2 displayed a 1.5-fold (MDA-MB-231) to 3-fold (HCC-1806) increase in the percentage of propidium iodide positive cells compared to control transfection (Additional file 1: Supplementary Figure 2 a-c). This effect was less prominent upon overexpression of the other two isomiRs of miR.183-5p. Next, to investigate the effect of all isoforms on the ability of cells to re-enter cell cycle after starvation-induced arrest in G1-phase, BrdU/7AAD staining and subsequent FACS analysis were performed 24 h after release from starvation. In all three TNBC cell lines, the overexpression of miR-183-5p|+2 arrested cells in G0/G1-phase compared to siAllstar and to the other two isoforms. At the same time, the cells in S-phase (BrdU-positive) were significantly decreased by miR-183-5p|+2 overexpression in all cell lines in comparison to the other two 5’isomiR-183-5p and negative control (Fig. 3d-f), indicating an impaired capability to overcome starvation-induced cell cycle arrest. In addition, S-phase was significantly reduced by miR-183-5p|0 in HCC-1806 cells. The results of the statistical comparison of all isomiRs to the control by Dunnett's multiple comparisons test and comparison of all conditions against each other by Tukey's multiple comparisons test are summarized in Additional file 2: Supplementary Table 8. Briefly, S-phase is significantly reduced by miR-183-5p|+2 compared to the other isomiRs in MDA-MB-231 cells, whereas these differences do not reach significance in BT-549 cells. In summary, among the investigated 5’isomiRs, miR-183-5p|+2 displayed the strongest inhibitory effect on cell cycle progression and cell viability in three genetically diverse TNBC models.
Proteins encoded by E2F target genes are depleted in cells overexpressing miR-183-5p|+2
Having observed a strong reduction in cell proliferation and cell cycle progression specifically upon overexpression of miR-183-5p|+2, we next aimed to unveil the molecular mechanism and the associated specific direct targets underlying this phenotype. Hence, we hypothesized that miR-183-5p|+2 would regulate specific proteins or signaling pathways resulting in the differential regulation of growth-associated phenotypes. To address this hypothesis, label-free mass spectrometry was performed to examine alterations in the proteome upon overexpression of each isoform in an unbiased way. We reasoned that a protein-level profiling would be most informative as miRNAs exert their function at the post-transcriptional level and do not always affect the abundance of target mRNAs.
In total, 5864 proteins were detected, and 3690 of them were detected in all samples. For a general understanding, we first focused on proteins detected in all samples and significantly deregulated by one of the isomiRs as compared to the controls (355 proteins, Fig. 4a, Additional file 2: Supplementary Table 9). Of note, protein expression levels upon overexpression of miR-183-5p|+2 were clearly distinct in comparison to the other two isoforms and miRNA negative controls.
In order to identify the molecular mechanisms potentially underlying the effect of each isomiR on cancer cell phenotypes, gene set enrichment analysis was performed using proteomic data (Additional file 2: Supplementary Table 10) and the Hallmark Gene Set Collection (h.all.v7.3) [39]. Here, we observed highly significant associations with the gene sets ‘G2/M Checkpoint’, ‘Hypoxia’ and ‘E2F Targets’ (Fig. 4b, Additional file 2: Supplementary Table 11). The latter had the highest NES and was most significantly downregulated specifically by overexpression of miR-183-5p|+2. No such enrichment was observed with the other two isoforms (Additional file 1: Supplementary Figure 3). The specific downregulation of ‘E2F Targets’ by miR-183-5p|+2 was apparent also at the level of individual proteins compared to the other two isoforms or the negative controls (Fig. 4c, Additional file 2: Supplementary Table 12). This was further strengthend by a Fisher’s exact test using GO-term based genesets, namely GOBP_NEGATIVE_REGULATION_OF_CELL_CYCLE_G1_S_PHASE_TRANSITION and GOBP_POSITIVE_REGULATION_OF_CELL_CYCLE_G1_S_PHASE_TRANSITION to specifically test the hypothesis that the G1-arrest observed in our cell cycle analysis can also be confirmed on protein level specifically for miR-183-5p|+2. While we did not observe any significant association of miRNA overexpression with the positive regulation set, we confirmed a significant enrichment of negative regulators of G1-S transition within the set of proteins upregulated upon overexpression of miR-183-5p|+2. (Additional file 2: Supplementary Table 13). Together, the impaired ability of cells overexpressing miR-183-5p|+2 to re-enter cell cycle after starvation and the marked downregulation of a set of E2F target proteins in these cells let us hypothesize that there might be a mechanistic link.
E2Fs, a group of genes that encode a family of transcription factors (TFs), regulate G1/S transition in the cell cycle. All E2Fs are involved in cell cycle regulation, but only three of them, E2F transcription factor 1 (E2F1), E2F2 and E2F3a, are transcriptional activators. Hence, we hypothesized that miR-183-5p|+2 would directly target at least one of these activating TFs to achieve coordinated downregulation of downstream transcriptional targets.
Thus, for all three isomiRs of miR-183-5p, we predicted direct target sites within 3’UTR regions supported by breast cancer cell line RNA sequencing data to reduce the occurrence of false-positive predictions. To be more stringent, only consensus predictions from the TargetScan and miRanda3.3a algorithms were considered (Additional file 2: Supplementary Table 14). While there was a strong overlap in the predicted targets of each of the 5’isomiR-183-5p, also isomiR-specific targets were identified (Fig. 4d). Of note, the spectrum of uniquely predicted targets was the largest for miR-183-5p|+2. E2F1 was among those genes that were specifically predicted as target of miR-183-5p|+2 with a seven base pairs binding site (7mer-m8; Additional file 1: Supplementary Figure 4). Therefore, we hypothesized that E2F1 might be directly regulated by miR-183-5p|+2 and went on to test this experimentally.
E2F1 is a direct target of miR-183-5p|+2
In order to validate direct targeting of E2F1, the full-length 3’UTR of E2F1 was cloned into a dual luciferase reporter assay (Fig. 5a). To prove that the effect of miR-183-5p|+2 was indeed due to direct binding to the seed-matching sequences in the 3’UTR of E2F1, we mutated the seed region in the E2F1 3’UTR and tested whether the mutation in the miRNA-binding site would abrogate the downregulation by miR-183-5p|+2. As shown in Fig. 5b, the luciferase activity was specifically repressed by miR-183-5p|+2, while no decrease was observed upon overexpression of the other two 5’isomiR-183-5p. In addition, the significant downregulation of relative luciferase activity by miR-183-5p|+2 was completely rescued by the mutations of the seed-matching bases. Hence, the 3’UTR of E2F1 was directly and specifically targeted by miR-183-5p|+2.
To validate the downregulation of E2F1 by miR-183-5p|+2 at the mRNA and protein levels, MDA-MB-231 and BT-549 cells were transfected with miR-183-5p|0, miR-183-5p|+1, miR-183-5p|+2 mimics or a negative control. The E2F1 mRNA level was significantly decreased upon miR-183-5p|+2 overexpression in both cell lines (Fig. 5c). Accordingly, E2F1 protein expression level was slightly but significantly reduced in both cell lines (Fig. 5d).
Downregulation of E2F1 is partially responsible for the effect of overexpression of miR-183-5p|+2
Having shown that miR-183-5p|+2 directly targets E2F1, we next hypothesized that the effect on cell proliferation and cell cycle upon overexpression of miR-183-5p|+2 was in part caused by downregulation of E2F1. To this end, a set of four different siRNAs targeting E2F1 was used. Based on the knockdown efficiency of each individual E2F1 siRNAs, we chose siE2F1_9 and siE2F1_11 for further study. The knockdown efficiency of siE2F1_9 and siE2F1_11 was confirmed (Additional file 1: Supplementary Figure 5) and, as anticipated, the knockdown of E2F1 resulted in a significant inhibition of cell proliferation in the MDA-MB-231 and BT-549 cell lines (Fig. 6a). The cells were arrested in G0/G1-phase upon transfection of siE2F1_9 or siE2F1_11 (Fig. 6b) and, in MDA-MB-231, S-phase was significantly decreased. In summary, downregulation of E2F1 phenotypically resembles overexpression of miR-183-5p|+2 regarding cell proliferation and cell cycle in two TNBC cell lines indicating that this interaction might be partially responsible for the phenotype.
To further validate that miR-183-5p|+2 arrest cells in G0/G1-phase due to targeting within the 3’UTR of E2F1, we tested whether ectopic overexpression of the E2F1 ORF (Additional file 1: Supplementary Figure 6) would rescue the effect of miR-183-5p|+2 overexpression on cell cycle. Indeed, the effect of miR-183-5p|+2 on the cell cycle was rescued in E2F1 ORF overexpressing cell lines (Fig. 6c). In summary, these results indicate that miR-183-5p|+2 directly targets – among others – the 3’UTR of E2F1, leading to the downregulation of E2F1, thereby causing reduced cell proliferation and repression of cell cycle progression. Of note, despite minor downregulation of E2F1 by miR-183-5p|+2, the impact on viability and cell cycle progression was comparable to the effects of E2F1 knockdown indicating coordinate regulation of additional, yet unknown, targets cooperatively causing the phenotype of the isomiR.
Pri-miR-183 transcription is regulated by E2F1
Having observed direct targeting of E2F1 by miR-183-5p|+2 in an in vitro setup, we hypothesized that the expression and activity of E2F1 should be inversely correlated with the expression of miR-183-5p|+2 in breast cancer patients. Therefore, we investigated the correlation of expression between miR-183-5p|+2 or miR-183-5p and E2F1 mRNA levels in the TCGA and METABRIC datasets, respectively (Additional file 1: Supplementary Figure 7). However, the expression of E2F1 itself did not exhibit a strong correlation with the expression of miR-183-5p|+2 (Spearman r = 0.19 in TCGA TNBC patients and r = 0.01 in METABRIC TNBC patients; Additional file 1: Supplementary Figure 7a). Since the post-transcriptional regulation via miRNAs may not affect the mRNA but only protein levels and thereby activity, we next assessed whether miR-183-5p|+2 expression would inversely correlate with the activity of E2F1 in the patient data. To this end, gene set enrichment analysis was applied to batch-corrected TCGA miRNA and mRNA sequencing data [33] for TNBC patients (Additional file 2: Supplementary Table 15 and Supplementary Table 16). In contrast to our expectations, E2F and MYC proto-oncogene, bHLH transcription factor (MYC) activities were positively correlated with the expression of all three isomiRs in TNBC patients in TCGA as well as with the overall expression of miR-183-5p in the METABRIC cohort (Fig. 7a, Additional file 1: Supplementary Fig. 7b). Of note, associated pathways differed substantially between PAM50 subtypes highlighting the importance of differentiated analysis of individual subtypes (Additional file 1, Supplementary Figure 8). Active E2F signaling was increased in tumor as compared to normal tissue and the expression of miR-183-5p|+2 was also higher in TNBC than in non-TNBC and normal tissue (Fig. 7b, Spearman correlation coefficient within TNBC patients r = 0.19 in TCGA and r = 0.36 in METABRIC; Additional file 2: Supplementary Table 14). The other two 5’isomiR-183-5p as well as miR-183-5p expression in the METABRIC cohort showed a similar trend (Fig. 7b, Additional file 1: Supplementary Figure 9a). Of note, the expression levels of the 5’isomiRs are highly correlated in patients suggesting that their expression is mainly regulated by transcriptional activity. Therefore, GSEA results, which are based on the correlations of genes with the respective isomiRs, are expected to be very similar for the three isomiRs and not indicative of similarities or dissimilarities between them. Similarly, activities of E2F and MYC are highly positively correlated in both TCGA and METABRIC breast cancer patient samples irrespective of their subtype (Spearman correlation coefficient r=0.82, both in TCGA and METABRIC datasets, Additional file 1: Supplementary Figure 9b). Hence, it cannot be inferred from this data if E2F, MYC or both are involved in the transcriptional regulation of pri-miR-183 – and thereby consistently of all three 5’isomiRs – in breast cancer patients.
To investigate this further, we quantified the levels of mature miR-183-5p upon knockdown of E2F1 and MYC. Indeed, knockdown of E2F1 led to reduced expression of both mature miR-183-5p and pri-miR-183 (Fig. 7c-d). In contrast, knockdown of MYC did not significantly affect the expression of mature miR-183-5p indicating that MYC is not the relevant transcription factor in our model systems. Accordingly, the expression of miR-183-5p and of pri-miR-183 was increased upon overexpression of E2F1 (Fig. 7e-f). Together, these results point towards a regulatory circuit comprising of transcriptional upregulation of miR-183 which in turn can lead to post-transcriptional downregulation of E2F1.
Discussion
MicroRNAs, as post-transcriptional regulators, influence tumor behavior and progression [16]. High throughput sequencing has demonstrated the existence of isomiRs, which are length and/or sequence variants of miRNAs. Of note, 5’isomiRs exhibit a shifted seed sequence compared to their canonical counterparts leading to an altered target spectrum, which has been rarely taken into consideration in previous research [14]. The functional relevance of 5’isomiRs might thus be underestimated, but should be considered towards a more holistic understanding of the mechanisms underlying tumor progression. Sequencing data from the TCGA breast cancer cohort indicated that three variants of miR-183-5p are highly upregulated in breast cancer compared to normal breast tissue, namely miR-183-5p|0, miR-183-5p|+1 and miR-183-5p|+2. Due to the substantial intrinsic differences between breast cancer subtypes, we focus here specifically on TNBC for mechanistic studies. Patient data analysis indicated that phenotypic traits of tumor cells associated with expression levels of miR-183 strongly differ between the PAM50 subtypes of breast cancer (Additional file 1: Supplementary Figure 8). Yet, this does not imply that other subtypes are not similarly affected by expression of miR-183-5p.
In vitro, overexpression of pre-miR-183 in both MDA-MB-231 and BT-549 cell lines reduced cell migration and invasion. These results are in line with a previous study on the tumor-suppressive role of miR-183-5p in lung cancer, where Wang et al. found that overexpression of miR-183 inhibited migration and invasion by targeting EZR [45]. Another study in HeLa cells indicated that overexpression of miR-183 caused a significant decrease in cell migration and invasion [46]. In contrast, several other studies have claimed that miR-183, functioning as an oncogene, contributes to the tumorigenesis of various cancers. For instance, Sarver et al. showed that downregulation of miR-183 suppressed EGR1 and PTEN in synovial sarcoma, rhabdomyosarcoma, and colon cancer cell lines, and was accompanied by decreased cell migration and invasion in both systems [47]. Besides, in gastric cancer cells, miR-183 increased cell proliferation and decreased cell autophagy and apoptosis by regulating UVRAG [48]. Taken together, this controversy emphasizes the context-dependent function of miRNAs as their functions depend on both expression levels and the importance of their targets in the cell. Importantly, previous studies on miR-183 in breast cancer and other tumor entities did not consider 5’isomiRs [49, 50], which adds another layer of complexity and regulatory potential. In our study, overexpression of each individual 5’isomiRs of miR-183-5p inhibited cell migration and invasion to similar extents. However, overexpression of pre-miR-183 in an in vivo xenograft model did not alter the potential of MDA-MB-231 cells to metastasize to the lung in our experimental setup. Yet, we observed a trend of reduced tumor growth that failed to reach statistical significance with the limited number of mice used.
In vitro, we observed that overexpression specifically of miR-183-5p|+2 led to a decrease in cell viability and impaired ability of cells to re-enter cell cycle employing three genetically diverse models of TNBC, emphasizing the generality of this finding. Especially in the BT-549 cell line, not only miR-183-5p|+2 but also the other two 5’isomiRs of miR-183-5p inhibited cell proliferation. This might be explained by the genetic and phenotypic heterogeneity of the model cell lines. Potentially, BT-549 expresses a different set of target genes or is addicted to a specific pathway regulated by miR-183-5p|0 and |+1. Taken together, among the miR-183-5p 5’isomiRs, miR-183-5p|+2 displays the strongest tumor-suppressive effect in cell viability and cell cycle. These results are in line with our hypothesis that 5’isomiRs with a shifted seed sequence have different target spectra potentially resulting in a functional difference compared to their canonical counterparts as well as to other 5’isomiRs. This is in line with a study of Telonis et al. who surveyed transcriptome-differences of three different 5’isomiR-183-5p in MDA-MB-231 [51]. Overexpression of distinct miR-183-5p isomiRs in MDA-MB-231 cells followed by microarray analysis revealed that each isomiRs of miR-183-5p had a different targetome and the effect of the isomiRs on the transcriptome differed drastically. However, these authors restricted their investigation to computational analyses and did not functionally validate their findings [51]. In our present study, we have extended this analysis to the proteome of MDA-MB-231 cells. In addition, we have proven that 5’isomiR-183-5p exhibit functional differences in cell viability and cell cycle of TNBC. This is well in line with a growing number of reports pointing out that 5’isomiRs exhibit differential functions. A previous study in our group showed that 5’isomiR-140-3p specifically inhibits proliferation and migration by directly regulating targets distinct from the canonical miRNA [15]. Another study conducted by Bhardwaj et al. showed that miR-140-3p|+1 directly regulates HMGCR and HMGCS1 leading to inhibition of cell growth [52] thus confirming that 5’isomiRs frequently have different functions than canonical miRNAs.
Since it is widely accepted that the functions of miRNAs rely on their target genes, we hypothesized that the phenotypic differences were due to distinct target genes. Commonly, potential targets of miRNA are predicted based on sequence complementarity using computational tools, such as miRanda, PicTar and TargetScan. However, these prediction tools frequently identify a vast amount of false-positive candidates [53] and predictions for miR-183-5p|+2 were not available from these databases. To overcome shortcomings of computational methods, we performed custom target predictions combining consensus predictions from miRanda and TargetScan restricted to bona-fide expressed 3’UTRs. Since the regulation of target genes by a miRNA ultimately leads to changes in protein expression, we chose mass spectrometry analysis to identify direct and indirect targets of miR-183-5p in this study. Mass spectrometry analysis showed that overexpression of miR-183-5p|+2 resulted in protein expression patterns distinct from the other two 5’isomiR-183-5p and also from the controls. While this analysis cannot discriminate between direct and indirect miRNA-effects, it clearly indicates a global functional difference between the isoforms. Here, we demonstrate the post-transcriptional regulatory effects of the isomiRs on protein level. Yet, we cannot make claims about their impact on global target mRNA stability due to lack of transcriptomics data. Gene set enrichment analysis based on mass spectrometry data suggested that the E2F target signature was specifically decreased by the overexpression of miR-183-5p|+2, indicating a potential molecular mechanism underlining the functional difference among the three investigated 5’isomiR-183-5p. Of note, we also observed a highly significant downregulation of proteins belonging to the hallmark gene set ‘G2/M Checkpoint’ by miR-183-5p|+2 suggesting a second, yet unexplored, impact of miR-183-5p|+2 on cell cycle progression. Similarly, the ‘Hypoxia’ gene set appeared to be positively associated with overexpression of this 5’isomiR based on our proteomic data. However, these observations were out of the scope of the current study and require further validation to confirm their relevance. It is important to note that we here employed gene sets derived from transcriptomic signatures and transcription factor binding site predictions together with proteomic data. Therefore, associations observed in this analysis require additional validation as provided in this study in case of the regulation of E2F by miR-183-5p.
E2Fs, a group of genes that encode a family of transcription factors, are vital for cell cycle regulation, especially for the transition from G0/G1- to S-phase [54]. Based on the stringent target prediction, we found that E2F1 is a predicted specific target of miR-183-5p|+2 and confirmed this by a 3’UTR luciferase reporter assay. Besides, we showed that siRNA-mediated knockdown of E2F1 partially phenocopied the effect of miR-183-5p|+2 on cell proliferation and cell cycle re-entry. This is congruent with a study demonstrating that upregulation of E2F1 promoted G1/S transition and induced tumorigenesis of TNBC [55]. Taken together, these results indicate a tumor-suppressive function of miR-183-5p|+2 by directly targeting E2F1. On the other hand, these results also highlight that miRNA effects cannot be explained by individual direct targets but rather by coordinate regulation of multiple proteins contributing to the same phenotype. This concept has been demonstrated for several example miRNAs [56, 57]. Therefore, it is plausible that also in case of miR-183-5p, various downregulated targets contribute to the observed phenotypes. Yet, it is out of scope of the present study to identify and validate these additional targets.
To investigate the potential clinical relevance of miR-183-5p|+2 targeting E2F1, gene set enrichment analysis was used to determine the correlation of E2F1 activity and miR-183-5p in TCGA and METABRIC TNBC patients’ data. These analyses showed that E2F and MYC activities positively correlated with the expression of all three 5’isomiR-183-5p, which contrasts our expectation to observe an inverse correlation of E2F1 and miR-183-5p|+2. Therefore, we hypothesized that expression of miR-183 might, in turn, be regulated by either E2F or MYC resulting in the observed positive correlation in patients. Along this line, we validated that the expression of mature miR-183 and its precursor pri-miR-183 was indeed regulated by E2F1. This is in line with recent miRNA sequencing data demonstrating a 60% decrease in expression of miR-183-5p upon knockdown of E2F1 in HeLa cells [58]. Of note, the oncogenic miRNAs miR-182 and miR-96 originating from the same primary transcript as miR-183 are coordinately downregulated by knockdown of E2F1 in this dataset. This further supports transcriptional regulation of these miRNAs by E2F1. Together, we demonstrated both direct targeting of E2F1 by miR-183-5p|+2, and transcriptional regulation of miR-183 by E2F1 in vitro in TNBC cell lines, and patient data analysis supports the upregulation of miR-183 in tumors with high E2F activity. This strongly indicates the presence of a negative feedback loop between these two components. Yet, the analysis of steady state expression data of breast cancer patients does not enable a more precise characterization of this feedback loop in patient tumors. In addition, the exact characteristics of the regulation likely differ between individual patients based on their mutational backgrounds (e.g. in the E2F1 regulating RB1 gene). This and if additional targets of miR-183 are involved in the feedback regulation cannot be concluded here and would require additional research.
In our system, MYC seemed not to significantly influence the expression of miR-183. This does not exclude the possibility that MYC might be a regulator of miR-183 expression in other cellular contexts as well as in patients. Given its oncogenic roles in tumor progression, it is not surprising that E2F1 activity is higher in breast cancer tissue as compared to normal tissue with the highest activity in aggressive TNBC tumors. Consequently, this results in the increased expression of miR-183 which might not be expected given its rather tumor-suppressive role in different systems [45, 46]. Interestingly, we could demonstrate that miR-183-5p|+2 directly targets E2F1 and leads to downregulation of E2F1 expression within a negative feedback loop, thereby potentially restricting cell proliferation by repressing cell cycle progression. This might be especially relevant in cells deficient in pRB protein. Under physiological conditions, this protein is crucial in the control of cell cycle entry by sequestering E2F transcription factors. In cancer, especially in TNBC, the expression and functionality of pRB are frequently disrupted by various mechanisms [59]. Hence, these cells can enter cell cycle independent from activating signals. Here, it could be speculated that the feedback regulation by miR-183-5p|+2 might be part of an evolutionary fail-safe mechanism to initially prevent overactive E2F, potentially also during extensive mitogenic signalling.
Our study highlights that due to the heterogeneity and complexity of tumorigenesis, the high expression of a miRNA in tumors is not necessarily connected to an oncogenic function of this miRNA. Such elevated expression might rather be an indication of a feedback loop that is supposed to counteract the growth-promoting activity of the miRNA-targets under physiological conditions. Given the success of tumor cells in overcoming negative regulation of growth, the observed elevated expression of miR-183-5p could be seen as a reminiscence of a negative control mechanism that has obviously failed in the progressed tumor system. Concretely, the suppression of E2F1 by miR-183-5p|+2 could potentially be compensated by upregulation of other E2F transcription factors or loss of pRB, which might in turn further upregulate expression of miR-183.