miRGTF-net: Integrative miRNA-gene-TF network analysis reveals key drivers of breast cancer recurrence

Exploring miRNA‑target gene profiles associated with drug resistance in patients with breast cancer receiving neoadjuvant chemotherapy

Exosomal microRNAs (miRNAs) are closely related to drug resistance in patients with breast cancer (BC); however, only a few roles of the exosomal miRNA-target gene networks have been clinically implicated in drug resistance in BC. Therefore, the present study aimed to identify the differential expression of exosomal miRNAs associated with drug resistance and their target mRNAs. In vitro microarray analysis was used to verify differentially expressed miRNAs (DEMs) in drug-resistant BC. Next, tumor-derived exosomes (TDEs) were isolated. Furthermore, it was determined whether the candidate drug-resistant miRNAs were also significant in TDEs, and then putative miRNAs in TDEs were validated in plasma samples from 35 patients with BC (20 patients with BC showing no response and 15 patients with BC showing a complete response). It was confirmed that the combination of five exosomal miRNAs, including miR-125b-5p, miR-146a-5p, miR-484, miR-1246-5p and miR-1260b, was effective for predicting therapeutic response to neoadjuvant chemotherapy, with an area under the curve value of 0.95, sensitivity of 75%, and specificity of 95%. Public datasets were analyzed to identify differentially expressed genes (DEGs) related to drug resistance and it was revealed that BAK1, NOVA1, PTGER4, RTKN2, AGO1, CAP1, and ETS1 were the target genes of exosomal miRNAs. Networks between DEMs and DEGs were highly correlated with mitosis, metabolism, drug transport, and immune responses. Consequently, these targets could be used as predictive markers and therapeutic targets for clinical applications to enhance treatment outcomes for patients with BC.

A novel approach for a joint analysis of isomiR and mRNA expression data reveals features of isomiR targeting in breast cancer

A widely used procedure for selecting significant miRNA-mRNA or isomiR-mRNA pairs out of predicted interactions involves calculating the correlation between expression levels of miRNAs/isomiRs and mRNAs in a series of samples. In this manuscript, we aimed to assess the validity of this procedure by comparing isomiR-mRNA correlation profiles in sets of sequence-based predicted target mRNAs and non-target mRNAs (negative controls). Target prediction was carried out using RNA22 and TargetScan algorithms. Spearman's correlation analysis was conducted using miRNA and mRNA sequencing data of The Cancer Genome Atlas Breast Invasive Carcinoma (TCGA-BRCA) project. Luminal A, luminal B, basal-like breast cancer subtypes, and adjacent normal tissue samples were analyzed separately. Using the sets of putative targets and non-targets, we introduced adjusted isomiR targeting activity (ITA)-the number of negatively correlated potential isomiR targets adjusted by the background (estimated using non-target mRNAs). We found that for most isomiRs a significant negative correlation between isomiR-mRNA expression levels appeared more often in a set of predicted targets compared to the non-targets. This trend was detected for both classical seed region binding types (8mer, 7mer-m8, 7mer-A1, 6mer) predicted by TargetScan and the non-classical ones (G:U wobbles and up to one mismatch or unpaired nucleotide within seed sequence) predicted by RNA22. Adjusted ITA distributions were similar for target sites located in 3'-UTRs and coding mRNA sequences, while 5'-UTRs had much lower scores. Finally, we observed strong cancer subtype-specific patterns of isomiR activity, highlighting the differences between breast cancer molecular subtypes and normal tissues. Surprisingly, our target prediction- and correlation-based estimates of isomiR activities were practically non-correlated with the average isomiR expression levels neither in cancerous nor in normal samples.

isomiRTar: a comprehensive portal of pan-cancer 5'-isomiR targeting

Inaccurate cleavage of pri- and pre-miRNA hairpins by Drosha and Dicer results in the generation of miRNA isoforms known as isomiRs. isomiRs with 5'-end variations (5'-isomiRs) create a new dimension in miRNA research since they have different seed regions and distinct targetomes. We developed isomiRTar (https://isomirtar.hse.ru)-a comprehensive portal that allows one to analyze expression profiles and targeting activity of 5'-isomiRs in cancer. Using the Cancer Genome Atlas sequencing data, we compiled the list of 1022 5'-isomiRs expressed in 9282 tumor samples across 31 cancer types. Sequences of these isomiRs were used to predict target genes with miRDB and TargetScan. The putative interactions were then subjected to the co-expression analysis in each cancer type to identify isomiR-target pairs supported by significant negative correlations. Downstream analysis of the data deposited in isomiRTar revealed both cancer-specific and cancer-conserved 5'-isomiR expression landscapes. Pairs of isomiRs differing in one nucleotide shift from 5'-end had poorly overlapping targetomes with the median Jaccard index of 0.06. The analysis of colorectal cancer 5'-isomiR-mediated regulatory networks revealed promising candidate tumor suppressor isomiRs: hsa-miR-203a-3p-+1, hsa-miR-192-5p-+1 and hsa-miR-148a-3p-0. In summary, we believe that isomiRTar will help researchers find novel mechanisms of isomiR-mediated gene silencing in different types of cancer.

Evolution of miRNA binding sites and regulatory networks in cichlids

The divergence of regulatory regions and gene regulatory network (GRN) rewiring is a key driver of cichlid phenotypic diversity. However, the contribution of miRNA binding site turnover has yet to be linked to GRN evolution across cichlids. Here, we extend our previous studies by analysing the selective constraints driving evolution of miRNA and transcription factor (TF) binding sites of target genes, to infer instances of cichlid GRN rewiring associated with regulatory binding site turnover. Comparative analyses identified increased species-specific networks that are functionally associated to traits of cichlid phenotypic diversity. The evolutionary rewiring is associated with differential models of miRNA and TF binding site turnover, driven by a high proportion of fast-evolving polymorphic sites in adaptive trait genes compared to subsets of random genes. Positive selection acting upon discrete mutations in these regulatory regions is likely to be an important mechanism in rewiring GRNs in rapidly radiating cichlids. Regulatory variants of functionally associated miRNA and TF binding sites of visual opsin genes differentially segregate according to phylogeny and ecology of Lake Malawi species, identifying both rewired e.g. clade-specific and conserved network motifs of adaptive trait associated GRNs. Our approach revealed several novel candidate regulators, regulatory regions and three-node motifs across cichlid genomes with previously reported associations to known adaptive evolutionary traits.

MIO: MicroRNA target analysis system for Immuno-Oncology

Summary

MicroRNAs have been shown to be able to modulate the tumor microenvironment and the immune response and hence could be interesting biomarkers and therapeutic targets in immuno-oncology; however, dedicated analysis tools are missing. Here, we present a user-friendly web platform MIO and a Python toolkit miopy integrating various methods for visualization and analysis of provided or custom bulk microRNA and gene expression data. We include regularized regression and survival analysis and provide information of 40 microRNA target prediction tools as well as a collection of curated immune related gene and microRNA signatures and processed TCGA data including estimations of infiltrated immune cells and the immunophenoscore. The integration of several machine learning methods enables the selection of prognostic and predictive microRNAs and gene interaction network biomarkers.

Availability and implementation

https://mio.icbi.at, https://github.com/icbi-lab/mio and https://github.com/icbi-lab/miopy.

Supplementary information

Supplementary data are available at Bioinformatics online.

Isoforms of miR-148a and miR-203a are putative suppressors of colorectal cancer

MicroRNAs are short non-coding molecules which regulate translation in a gene-specific manner. MicroRNA isoforms that differ by few extra or missing nucleotides at the 5'-terminus (5'-isomiR) show strikingly different target specificity. This study aimed to identify functional roles of 5′-isomiR in colorectal cancers. Transcriptomic targets of microRNA isoforms were predicted using bioinformatics tools miRDB and TargetScan. The sets of putative targets identified for 5′-isomiR were integrated with mRNA and microRNA sequencing data for primary colorectal tumors retrieved from The Cancer Genome Atlas Colon Adenocarcinoma (TCGA-COAD) database. The network of interactions among miRNA, their targets and transcription factors was built using the miRGTF-net algorithm. The results indicate that microRNA isoforms highly expressed in colorectal cancer and differing by a single nucleotide position at the 5'-terminus have ≤ 30% common targets. The regulatory network of interactions enables identification of the most engaged microRNA isoforms. Anti-correlated expression levels of canonical microRNA hsa-miR-148a-3p and its putative targets including CSF1, ETS1, FLT1, ITGA5, MEIS1, MITF and RUNX2 proliferation regulators suggest an anti-tumor role for this molecule. The canonical microRNA hsa-miR-203a-3p|0 and its 5′-isoform bind different sets of anti-correlated putative targets, although both of them interact with genes involved in the epithelial-mesenchymal transition: SNAI2 and TNC.

Changes in the Expression of miRNA Isoforms and Their Targets in HT-29 Cells after Hypoxic Exposure

Tumor hypoxia is one of the main causes of progression and metastasis of colorectal cancer. Changes in the expression of miRNA responsible for post-translation regulation of gene expression is an important molecular mechanism of cell response to hypoxia. We performed sequencing of miRNA and mRNA of human colorectal adenocarcinoma HT-29 cells treated with two chemical agents mimicking hypoxia: cobalt (II) chloride and oxyquinoline. Bioinformatics analysis revealed differentially expressed miRNA isoforms (hsa-miR-210-3p|0, hsa-miR- 22-3p|0, hsa-let-7a-3p|0, hsa-miR-615-3p|0, and hsa-miR-4521|0) and their targets that changed their expression in both models of hypoxia. Thus, we identified new regulatory mechanisms of cell response to hypoxia.

Expression of CD44 Isoforms in Tumor Samples and Cell Lines of Human Colorectal Cancer

Detection of colorectal cancer biomarkers (CRC) remains an urgent task for the diagnosis and prediction of the disease course. A promising approach is the study of cancer stem cell markers. The cell surface glycoprotein CD44 is very important for CRC and its stem cells. Alternative splicing of 9 variable exons of CD44 mRNA leads to the formation of various isoforms of the protein with different roles in the progression of cancer. Studies of the functions of CD44 isoforms require adequate models considering the distribution of CD44 isoforms in real tumor samples. In the present study, the expression profile of CD44 isoforms in CRC was assessed based on the publicly available mRNA sequencing data of patient tumors from the TCGA-COAD database. It was shown that normal tissues predominantly expressed isoforms 3 and 4 at nearly equal levels, whereas tumors mainly expressed isoforms 2, 3, and 4; isoform 3 was expressed at the highest level. Further, the most relevant cell lines for studying the role of CD44 in CRC were identified based on the analysis of mRNA sequencing data of 55 CRC cell lines form CCLE database.

miRModuleNet: Detecting miRNA-mRNA Regulatory Modules

Increasing evidence that microRNAs (miRNAs) play a key role in carcinogenesis has revealed the need for elucidating the mechanisms of miRNA regulation and the roles of miRNAs in gene-regulatory networks. A better understanding of the interactions between miRNAs and their mRNA targets will provide a better understanding of the complex biological processes that occur during carcinogenesis. Increased efforts to reveal these interactions have led to the development of a variety of tools to detect and understand these interactions. We have recently described a machine learning approach miRcorrNet, based on grouping and scoring (ranking) groups of genes, where each group is associated with a miRNA and the group members are genes with expression patterns that are correlated with this specific miRNA. The miRcorrNet tool requires two types of -omics data, miRNA and mRNA expression profiles, as an input file. In this study we describe miRModuleNet, which groups mRNA (genes) that are correlated with each miRNA to form a star shape, which we identify as a miRNA-mRNA regulatory module. A scoring procedure is then applied to each module to further assess their contribution in terms of classification. An important output of miRModuleNet is that it provides a hierarchical list of significant miRNA-mRNA regulatory modules. miRModuleNet was further validated on external datasets for their disease associations, and functional enrichment analysis was also performed. The application of miRModuleNet aids the identification of functional relationships between significant biomarkers and reveals essential pathways involved in cancer pathogenesis. The miRModuleNet tool and all other supplementary files are available at https://github.com/malikyousef/miRModuleNet/

ExhauFS: exhaustive search-based feature selection for classification and survival regression

Feature selection is one of the main techniques used to prevent overfitting in machine learning applications. The most straightforward approach for feature selection is an exhaustive search: one can go over all possible feature combinations and pick up the model with the highest accuracy. This method together with its optimizations were actively used in biomedical research, however, publicly available implementation is missing. We present ExhauFS-the user-friendly command-line implementation of the exhaustive search approach for classification and survival regression. Aside from tool description, we included three application examples in the manuscript to comprehensively review the implemented functionality. First, we executed ExhauFS on a toy cervical cancer dataset to illustrate basic concepts. Then, multi-cohort microarray breast cancer datasets were used to construct gene signatures for 5-year recurrence classification. The vast majority of signatures constructed by ExhauFS passed 0.65 threshold of sensitivity and specificity on all datasets, including the validation one. Moreover, a number of gene signatures demonstrated reliable performance on independent RNA-seq dataset without any coefficient re-tuning, i.e., turned out to be cross-platform. Finally, Cox survival regression models were used to fit isomiR signatures for overall survival prediction for patients with colorectal cancer. Similarly to the previous example, the major part of models passed the pre-defined concordance index threshold 0.65 on all datasets. In both real-world scenarios (breast and colorectal cancer datasets), ExhauFS was benchmarked against state-of-the-art feature selection models, including L₁-regularized sparse models. In case of breast cancer, we were unable to construct reliable cross-platform classifiers using alternative feature selection approaches. In case of colorectal cancer not a single model passed the same 0.65 threshold. Source codes and documentation of ExhauFS are available on GitHub: https://github.com/s-a-nersisyan/ExhauFS.

Identification of key miRNAs in prostate cancer progression based on miRNA-mRNA network construction

Prostate cancer (PC) is one of the major male cancers. Differential diagnosis of PC is indispensable for the individual therapy, i.e., Gleason score (GS) that describes the grade of cancer can be used to choose the appropriate therapy. However, the current techniques for PC diagnosis and prognosis are not always effective. To identify potential markers that could be used for differential diagnosis of PC, we analyzed miRNA-mRNA interactions and we build specific networks for PC onset and progression. Key differentially expressed miRNAs for each GS were selected by calculating three parameters of network topology measures: the number of their single regulated mRNAs (NSR), the number of target genes (NTG) and NSR/NTG. miRNAs that obtained a high statistically significant value of these three parameters were chosen as potential biomarkers for computational validation and pathway analysis. 20 miRNAs were identified as key candidates for PC. 8 out of 20 miRNAs (miR-25-3p, miR-93-3p, miR-122-5p, miR-183-5p, miR-615-3p, miR-7-5p, miR-375, and miR-92a-3p) were differentially expressed in all GS and proposed as biomarkers for PC onset. In addition, "Extracellular-receptor interaction", "Focal adhesion", and "microRNAs in cancer" were significantly enriched by the differentially expressed target genes of the identified miRNAs. miR-10a-5p was found to be differentially expressed in GS 6, 7, and 8 in PC samples. 3 miRNAs were identified as PC GS-specific differentially expressed miRNAs: miR-155-5p was identified in PC samples with GS 6, and miR-142-3p and miR-296-3p in PC samples with GS 9. The efficacy of 20 miRNAs as potential biomarkers was revealed with a Random Forest classification using an independent dataset. The results demonstrated our 20 miRNAs achieved a better performance (AUC: 0.73) than miRNAs selected with Boruta algorithm (AUC: 0.55), a method for the automated feature extraction. Studying miRNA-mRNA associations, key miRNAs were identified with a computational approach for PC onset and progression. Further experimental validations are needed for future translational development.

ECM-Receptor Regulatory Network and Its Prognostic Role in Colorectal Cancer

Interactions of the extracellular matrix (ECM) and cellular receptors constitute one of the crucial pathways involved in colorectal cancer progression and metastasis. With the use of bioinformatics analysis, we comprehensively evaluated the prognostic information concentrated in the genes from this pathway. First, we constructed a ECM-receptor regulatory network by integrating the transcription factor (TF) and 5'-isomiR interaction databases with mRNA/miRNA-seq data from The Cancer Genome Atlas Colon Adenocarcinoma (TCGA-COAD). Notably, one-third of interactions mediated by 5'-isomiRs was represented by noncanonical isomiRs (isomiRs, whose 5'-end sequence did not match with the canonical miRBase version). Then, exhaustive search-based feature selection was used to fit prognostic signatures composed of nodes from the network for overall survival prediction. Two reliable prognostic signatures were identified and validated on the independent The Cancer Genome Atlas Rectum Adenocarcinoma (TCGA-READ) cohort. The first signature was made up by six genes, directly involved in ECM-receptor interaction: AGRN, DAG1, FN1, ITGA5, THBS3, and TNC (concordance index 0.61, logrank test p = 0.0164, 3-years ROC AUC = 0.68). The second hybrid signature was composed of three regulators: hsa-miR-32-5p, NR1H2, and SNAI1 (concordance index 0.64, logrank test p = 0.0229, 3-years ROC AUC = 0.71). While hsa-miR-32-5p exclusively regulated ECM-related genes (COL1A2 and ITGA5), NR1H2 and SNAI1 also targeted other pathways (adhesion, cell cycle, and cell division). Concordant distributions of the respective risk scores across four stages of colorectal cancer and adjacent normal mucosa additionally confirmed reliability of the models.

HIF-Dependent NFATC1 Activation Upregulates ITGA5 and PLAUR in Intestinal Epithelium in Inflammatory Bowel Disease

Intestinal epithelial cells exist in physiological hypoxia, leading to hypoxia-inducible factor (HIF) activation and supporting barrier function and cell metabolism of the intestinal epithelium. In contrast, pathological hypoxia is a common feature of some chronic disorders, including inflammatory bowel disease (IBD). This work was aimed at studying HIF-associated changes in the intestinal epithelium in IBD. In the first step, a list of genes responding to chemical activation of hypoxia was obtained in an in vitro intestinal cell model with RNA sequencing. Cobalt (II) chloride and oxyquinoline treatment of both undifferentiated and differentiated Caco-2 cells activate the HIF-signaling pathway according to gene set enrichment analysis. The core gene set responding to chemical hypoxia stimulation in the intestinal model included 115 upregulated and 69 downregulated genes. Of this set, protein product was detected for 32 genes, and fold changes in proteome and RNA sequencing significantly correlate. Analysis of publicly available RNA sequencing set of the intestinal epithelial cells of patients with IBD confirmed HIF-1 signaling pathway activation in sigmoid colon of patients with ulcerative colitis and terminal ileum of patients with Crohn's disease. Of the core gene set from the gut hypoxia model, expression activation of ITGA5 and PLAUR genes encoding integrin α5 and urokinase-type plasminogen activator receptor (uPAR) was detected in IBD specimens. The interaction of these molecules can activate cell migration and regenerative processes in the epithelium. Transcription factor analysis with the previously developed miRGTF tool revealed the possible role of HIF1A and NFATC1 in the regulation of ITGA5 and PLAUR gene expression. Detected genes can serve as markers of IBD progression and intestinal hypoxia.

Hairpin sequence and structure is associated with features of isomiR biogenesis

MiRNA isoforms (isomiRs) are single stranded small RNAs originating from the same pri-miRNA hairpin as a result of cleavage by Drosha and Dicer enzymes. Variations at the 5'-end of a miRNA alter the seed region of the molecule, thus affecting the targetome of the miRNA. In this manuscript, we analysed the distribution of miRNA cleavage positions across 31 different cancers using miRNA sequencing data of TCGA project. As a result, we found that the processing positions are not tissue specific and that all miRNAs could be correctly classified as ones exhibiting homogeneous or heterogeneous cleavage at one of the four cleavage sites. In 42% of cases (42 out of 100 miRNAs), we observed imprecise 5'-end Dicer cleavage, while this fraction was only 14% for Drosha (14 out of 99). To the contrary, almost all cleavage sites of 3'-ends (either Drosha or Dicer) were heterogeneous. With the use of only four nucleotides surrounding a 5'-end Dicer cleavage position we built a model which allowed us to distinguish between homogeneous and heterogeneous cleavage with the reliable quality (ROC AUC = 0.68). Finally, we showed the possible applications of the study by the analysis of two 5'-end isoforms originating from the same exogeneous shRNA hairpin. It turned out that the less expressed shRNA variant was functionally active, which led to the increased off-targeting. Thus, the obtained results could be applied to the design of shRNAs whose processing will result in a single 5'-variant.