Connecting the dots between different networks: miRNAs associated with bladder cancer risk and progression

Between 2014 and 2016, we collected the tumoral BC tissue along with the adjacent healthy tissue from patients, only after obtaining informed consent from each patient and approval by the institutional ethics committee of the Iuliu Hatieganu University of Medicine and Pharmacy, Cluj-Napoca, Romania (UMPh), with the authorization no. 673A/20.11.2012. We stored transurethral resections of bladder tumor (TURBT) tissues in liquid nitrogen until sample processing and RNA extraction. When surgical and pathological procedures permitted, surgeons collected the healthy tissue, adjacent from the tumor, from each patient. We initially evaluated the expression of Her2 and TP53 using the standard immunohistochemistry staining protocol. The paired (healthy and tumor) tissue samples later used for the microarray and next-generation sequencing analyses are referred to as UMPh patient cohort. The second group of samples collected was for the qRT-PCR validation and named the validation set; we collected this additional patient cohort.

We performed a supplementary analysis at the third level of miRNAs-sequencing from 409 bladder tumors and 19 healthy tissues, adjacent to the tumors, obtained from the TCGA data portal (https://​tcga-data.​nci.​nih.​gov/​tcga/​). The data were analyzed in GeneSpring GX v.13.0, applying the previous defined cut-off value.

We selected to validate three upregulated transcripts (miR-23a, miR-141-3p and miR-205-5p) and two downregulated transcripts (miR-139-5p, and miR-143-5p) from the paired tissue samples. RNA was extracted using TriReagent based method for qRT-PCR validation was performed on 18 healthy bladder tissues and 18 bladder tumor tissues. We performed the cDNA synthesis using a 7.5 μl of reverse transcription mixture containing 0.72 μl of RT primer, 50 ng of total RNA and 0.5 μl of MultiScribe Reverse Transcriptase, 0.75 μl Reverse Transcription Buffer (10×), 0.075 μl dNTPs (100 mM), 0.1 μl of RNase Inhibitor according to Taqman MicroRNA Reverse Transcription Kit (Applied Biosystems) protocol. The cDNA mixture is incubated in PCR tubes for 30 min at 16 °C, 30 min at 42 °C and 5 min at 85 °C. qRT-PCR was performed using the in ViiA7 (Applied Biosystems) PCR machine with a total volume of reaction mix of 12.5 μl; this reaction mix consists of 6.25 μl of cDNA (diluted 1:6 with nuclease-free water), 5.63 μl of SSoAdvanced Universal Probe Supermix (Bio-Rad) and 0.73 μl primers for each miRNA. The reactions were set up as follows: initial denaturation step at 95 °C for 180 s, followed by 39 cycles of 95 °C for 5 s and, lastly, 60 °C for the 30s. The expression level of each miRNA is calculated by the threshold cycle (C_T). The relative expression level was calculated using –ΔΔC_T method and U6 for normalization.

Lab technicians synthesized the cDNA using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). The reaction preparation of qRT-qPCR used the SYBR Select Master Mix (Life Technologies) and executed using ViiA7. The following conditions were used: 95 °C for 2 min, 40 cycles of 95 °C for 10 s and 60 °C for 1 min. The FC of gene expression was calculated with the ΔΔC_T method, using B2M as the housekeeping gene.

A number of 22 bladder cancer samples, the same ones analyzed by microarray (except one samples from microarray patien cohort with low DNA concentration), were sequenced using Ion Ampliseq Cancer Panel and Ion Torrent PGM Next Generation Sequencing (Thermo Fischer Scientific); this panel contains the most relevant hot spot mutation. The amplicon libraries were prepared with 20 ng of DNA and the Ion Ampliseq™ Library Kit 2.0 (Life Technologies) and this was followed by a purification step using AMpure XP Beads (Beckman Coulter). Lastly, Qubit 2.0 was used for the quantification using Qubit HS DNA kit. For sequencing, four bar-coded 100pM-diluted libraries were used for each Ion 316 Chip (Thermo Fischer Scientific). Ion Torrent PGM Machine (Thermo Fischer Scientific) performed the sequencing, using the Ion PGM HI-Q Sequencing 200 kit. The software Torrent Suit 5.6 and Ion Reporter 5.6 executed the bioinformatics analysis, specifically for data trimming alignment and variant calling.

The IPA analysis determined the miRNAs with altered expression levels to identify the most relevant networks, altered pathways and their respective biological significance. For the identification of the target genes most relevant to our miRNAs, the following database and webservers were used: miRTarBase (https://​bio.​tools/​mirtarbase); miRNet (https://​www.​mirnet.​ca/​); and miRtargetLink (https://​ccb-web.​cs.​uni-saarland.​de/​mirtargetlink/​).

We determined the initial miRNA profiling using the WHO-2016 classification and the most frequent histological type, urothelial carcinoma. More specifically for the microarray study, the patients’ samples were graded as low or high and, additionally, through immunohistochemistry (IHC), the expression level of Her2 and TP53 was determined. Figure 1 exemplifies a low grade negative stained case for TP53 and Her2 compared to a high grade positive stained case for TP53 and Her2.

Table 1 presents the demographics and patient characteristic selected for the microarray. Of the 23 patients included in the microarray cohort, the average age is 64.04 ± 11.72, consisting of 5 females with an age average of 56.80 ± 6.30 and 18 males with an age average of 66.05 ± 12.20. The average age for the qRT-PCR cohort of 18 patients (5 females and 13 males), is respectively 66 ± 9,59 (57,2 ± 7,98 for females, respectively 69,38 ± 8.03 for males), which can be seen in Table 2. We classified the patient cohort selected for the profiling and qRT-PCR validations into two main subtypes: low-grade non-invasive tumors and high grade advanced tumors.

The present profiling investigation uses Agilent 60-mer SurePrint microarray technology that contains capture probes to characterize the expression of 2549 mature human miRNAs, obtained from the annotated miRbase 16. This profiling investigation leads to the identification of the most relevant altered miRNAs based on an analysis of paired samples: bladder cancer tumor tissues (TT) versus adjacent healthy tissues (TN). It identified 8 down-regulated and 28 up-regulated miRNAs in the UMPh patient cohort (Additional file 1: Table S1). The Heatmap analysis presenting the altered miRNAs in tumor tissue for UMPh patient cohort is showed in Fig. 2a; the blue color denoting downregulation while the red color signifies overexpression.

To strengthen and improve the accuracy of the miRNAs profiling investigation, we performed an additional TCGA analysis of 409 tumors and 19 tumor-adjacent healthy tissues (Additional file 2: Table S2). The cut-off value was FC ± 2 and p-value≤0,05 for this analysis, from which 18 downregulated and 187 upregulated miRNAs were identified (Additional file 3: Table S3). The Heatmap for this data is shown in Fig. 2b. We overlapped the miRNAs profiling data obtained from the two datasets mentioned above. The Venn diagram represented in Fig. 2c shows 14 upregulated and 4 downregulated miRNAs. For the analysis of the high versus low-grade bladder cancers (UMPh: Additional file 4: Table S4; TCGA: Additional file 5: Table S5), only one commonality, the downregulation of miR-10a, was identified (Fig. 2d).

Given the context from the information displayed in Fig. 3, five miRNAs were selected for further validation based on the statistically significant FC obtained from our paired bladder tissues samples and the following criteria: the two most downregulated miRNAs (miR-143 and miR-139); the highly upregulated miRNAs related to EMT that have been less studied (miR-141 and miR-205); and one intermediary upregulated miRNA (miR-23a).

Figure 3a shows the Pirate plots that were generated using the TCGA data for the five selected miRNAs (overexpressed transcripts: miR-23a, miR-141 and miR-205; downregulated transcripts: miR-139 and miR-143). The related survival curves for each miRNA transcript can be seen Fig. 3b while Fig. 3c displays the survival curves for combinations of two miRNAs. We observed an overall superior survival rate for the following three cases: high expression of miR-141 (p = 0.0045); low expression of miR-143 (p = 0.0184); and low expression of miR-143 combined with high expression of miR-141 (p = 0.0163). These altered miRNAs are directly interconectd via key genes regulated apoptosis or invasion related genes as can be emphases by network generated using miRtargetlink (Fig. 3d).

qRT-PCR is one of the most relevant and standardized approaches to validate microarray data. The new patient cohort used for the qRT-PCR consisted of 18 healthy tissues and 18 tumor tissues. From the common list of miRNAs between UMPh patients’ cohort and TCGA data, we selected two downregulated (miR-139-5p and miR-143-5p) and three up-regulated (miR-23a-3p, miR-141-3p, and miR-205-5p) to validate as independent prognostic markers. For miRNAs normalization, the ΔΔCt method with U6 as reference/normalizer; qRT-PCR validation data of the new additional patient cohort is presented in Fig. 4a. A ROC curve was constructed to estimate the sensitivity and specificity of these transcripts as biomarkers for bladder cancer, the highest AUC value being for miR-141 (0.8599) and miR-205 (0.8865). Circos diagram can be seen in Fig. 4b and the following observations were made: miR-205-5p and miR-141-3p are very heterogeneous providing expression level value, that should be considered as possible biomarker. miR-143-5p and miR-139-5p are very homogenous offering a potential advantage as biomarkers; and as expected with our relatively intermediary-expressed upregulated microRNA, miR-23a lies between the previously mentioned miRNAs.

We analyzed the mutation patterns for tumor tissues from the UMPh cohort of 22 patients diagnosed with bladder cancer (Additional file 9: Table S6). An increased mutation rate is shown for TP53, FGFR3, KDR, PIK3CA and ATM. Figure 5a presents the number of mutations identified in each gene; meanwhile, Fig. 5b shows the mutations frequency for each gene in the analyzed samples. Using NCBI ClinVar or FATHMM, the mutations exhibited were classified based on pathogenicity, meaning as pathogenic, benign/non-pathogenic or not applicable/unknown (NA) (Fig. 5c). A significant percentage of the mutations were classified with unknown (NA) role (55%) and 7% are pathogenic based on the ClinVar classification. In comparison, using the FATHMM scoring 52% were classified as unknown (NA) while 37% of the mutations were classified as pathogenic (Fig. 5d). The mutated genes in bladder cancer have both intronic and exonic localizations. The mutation localizations for the most relevant genes (ATM, FGFR3, KDR, TP53 and PIK3CA) are presented in Fig. 5e. For the case of TP53, an increased survival rate for patients expressing TP53 wild-type versus those expressing mutant was observed; for the rest of the frequently mutated genes there was no statistically significant data. The frequency of mutations in our pattern exhibited is consistent with the frequent mutations in 200 bladder cancer cases from the TCGA; OncoGrid summarizes the 200 most mutated bladder cancer cases and top 50 mutated genes. From the TGCA data and the frequent mutated genes obtained, the overall survival rate for the case of patients expressing mutated and wild-type genes can be seen Additional file 6: Figure S1. No statistically significant correlation between the survival rate and mutational status was observed for AKT, ATM, CTNNB1, FGFR3, TP53, KIT, MET, PIK3CA and SMO.

The calculation of gene expression FC used the ΔΔC_T method and B2M as the housekeeping gene (Fig. 5f). Figure 5f presents relative expression level of TP53 when comparing bladder tumor and normal tissues. There was no correlation among the expression level of TP53and their status (Fig. 5g). An additional graph of TP53 expression level using a median cut-off (50% above or low) to separate the expression levels into high or low and TP53 mutational status can be seen in Fig. 5e or based on the expression level (Fig. 5i). We observed an increase in survival rate for the case of TP53 wild type patients versus mutant TP53 (p = 0.0091).

The miRNAs with an altered expression level were input into the specialized software IPA (Ingenuity Pathways Analysis). This analysis highlights the main altered networks indicating disease or biological function based on the miRNA signatures from the bladder cancer patients and the interconnection with target genes. Therefore, 9 relevant molecular networks were identified for bladder cancer as can be seen in Table 3. Furthermore, Table 4 shows the statistically significant biological, cellular and molecular functions as well as diseases that these altered miRNAs are associated to. This statistical correlation is based on the number of molecules altered within a given network. The data from Table 4 was used to generate graphical representation of molecular networks using IPA, which displays the nodal genes for the altered miRNAs in bladder cancer. This improves identification of miRNA-gene interactions and determines their nature as direct or indirect. Figure 6a presents an overlap of the most important networks in terms of the connection between the altered miRNAs and target genes. Altered miRNAs target genes critical to regulating tumour progression and cell fate are particularly important for patient prognosis. Once again, this highlights the connection of the uncovered genes TP53, AKT, RAS, CDKN2A or CDK2 to cell fate, as important regulators. These genes are interconnected with the validated miRNAs which are highlighted with red circles. Another important mechanism in carcinogenesis is the transition from epithelial to mesenchymal (EMT) phenotype (Fig. 6b). It was determined that AKT is directly related to ZEB2 that interacts with miR-205-5p, miR-141-3p and miR-200b-3p. These miRNAs have highest expression levels, extracted from the top altered transcripts of both patient cohorts (found in the TCGA as well as our experimental data). Thus, genes were identified that are involved in carcinogenesis or responsible to apoptotic resistance. These genes are recognized to have prognostic role in colorectal cancer (RAS) or related to the progression of bladder cancer (AKT). RAS and AKT have been shown to be related to miR-21. Figure 6c highlights the relevant transcripts related to inflammation, based on the literature information including the IPA.

In order to identify the biological significance of the altered miRNA expression for bladder cancer, miRTarBase was used to identify validated target genes. The target genes relating to the most upregulated miRNAs were grouped into specific families. These miRNA families have an important role in the regulation of translational processes and, through their identification, relevant hallmarks of cancer can be correlated. Using the miRnet database, we determined the initial miRNA-target interactions and represented the network in Additional file 7: Figure S2. Following this initial inquiry, we looked at every individual interaction between miRNA-target and these can be seen in Additional file 7: Figure S2.

In Additional file 8: Figure S3A, the target genes for miR-23a-3p involved in PI3K-AKT signaling pathway and tight junctions are presented, and respectively emphasized in red and blue. Additional file 7: Figure S2B shows the target genes for miR-139-5p involved in Chemokine signaling pathway colored as blue, while the targeted genes involved focal adhesion colored as red. Additional file 8: Figure S3C presents the target genes for miR-141-3p that regulate the cell cycle in blue and the most frequently mutated miRNA-regulated genes in cancer in red. Additional file 8: Figure S3D presents the target genes for miR-143-5p involved in focal adhesion in blue and the most frequently mutated miRNA-regulated genes in cancer in red. Additional file 8: Figure S3E presents the target genes for miR-205-5p involved in adherent junctions in yellow, focal adhesion in green, the most frequently modified cancer pathways in blue and the most frequently mutated miRNA-regulated genes in cancer in red. In its entirety, Additional file 8: Figure S3 emphasizes the important relationship between EMT and miRNAs in terms of the frequently mutated genes or dysregulated pathways; with the chosen miRNAs, governing some aspect of cell adhesion. Multiple core regulatory processes or signaling pathways have demonstrated to be of functionally relevance to bladder cancer tumor progression, particularly in terms of invasiveness and reoccurrence. This has essential impact for bladder cancer in terms of diagnosis and/or prognosis.

The interconnection between mutated genes and altered miRNAs evaluated can been seen in Fig. 7. As was expected, miRNAs are not isolated transcripts. This is exemplified by the large number of miRNAs, with altered expression levels in bladder cancer, connected to the most frequently mutated genes (TP53, FGFR3, KDR, PIK3CA and ATM). Figure 7 also emphasizes the direct or indirect mode of affecting the transcriptomic pattern, which includes other miRNAs. Alterations in some genes, including TP53, FGFR3, ATM, KDR and PIK3CA, have been shown to trigger carcinogenesis and become intertwined with a high number of altered miRNA transcripts in bladder cancer (Additional file 8: Figure S3).