All procedures were performed in accordance with the standards of the European Community guidelines for the Care and Use of Laboratory Animals and were approved by the Institutional Ethics Committee at the University of Szeged (XX./4799/2015). To investigate the molecular background of IBD we used the 2,4,6-trinitrobenzene sulfonic acid (TNBS) induced rat model of experimental colitis as described previously (26). Briefly, male Wistar rats (180–220 g) were randomly divided into two groups: the first group served as control and the second group was treated with TNBS (colitis-induced) based on the method described by Morris et al. (27). Following overnight fasting, animals were challenged with a single dose of 0.25 mL TNBS (10mg in 0.25ml of 50% ethanol, w/v) intracolonically with an 8 cm long soft polyethylene tube through the anus under mild anesthesia (thiopental, i.p. 40 mg/kg). In the case of the control group (n=2), samples were taken from random colon sections; samples from colitis-induced animals were taken from inflamed colon region (n=4) as well as from non-adjacent uninflamed region (n=4). The damage score and lesion measurement (extents of macroscopically apparent inflammation, ulceration, and tissue necrosis) were determined using computerized planimetry as described previously (28). All samples were kept in TRIzol reagent (Thermo Fisher) at -80°C.

Colonic biopsies were obtained from 21 consenting patients with IBD [13 females and 8 males; median age 41 years, 25-74 years (for more details see (24)] undergoing colonoscopy for diagnostic purposes approved by Hungarian Medical Research Council (ETT) Medical Research Council’s Committee of Scientific and Research Ethics (TUKEB) (license number 59-32-PTE2015/IBD) as described previously (29). Human colonic biopsies were obtained in accordance with the guidelines set out by the Medical Research Council of Hungary. Sample collection and classification was performed according to disease status of patients, active/relapsing or inactive/remission phase. Furthermore, samples from relapsing patients were subdivided as uninflamed or inflamed according to the status of colon tissue. To classify uninflamed from inflamed samples, macroscopic diagnostic features were used in accordance with the recommendations of the European Crohn’s and Colitis Organisation (30). Multiple samples were obtained both from visible lesions and from mucosa which is normal on gross inspection. Classically, the mucosa in UC has a friable granular appearance and shows superficial ulcers. The earliest visible mucosal lesions of CD are small aphthous ulcers after which they coalesce to large deep serpiginous or linear ulcers (24).

The human THP-1 monocytic cells and HT-29 colonic epithelial cells were maintained in RPMI-1640 (Gibco) or DMEM (Gibco) medium, respectively, supplemented with 10% fetal bovine serum (FBS; Gibco) and 1% antibiotic/antimycotic solution containing 10,000 units/ml of penicillin, 10,000 µg/ml of streptomycin and 25 µg/ml of Amphotericin B (Gibco). Cells were cultured at 37°C in an atmosphere of 5% (v/v) CO₂ in air. For TNFα or LPS treatment, cells were seeded into six-well cell culture plates (Sarstedt) and stimulated with 1 μg/ml LPS (Sigma) or 10 ng/ml TNFα (R&D Systems) for the indicated time, followed by RNA extraction.

Samples from rat colons were homogenized in TRIzol reagent by ULTRA-TURRAX T-18 (IKA) instrument as described previously (26). 0.1 ml of chloroform (Sigma) was added to 0.3 ml homogenized sample with vigorous vortexing. Samples were centrifuged at 13000 rpm for 10 minutes. Total RNA was than extracted from the upper aqueous phase. RNeasy Plus Mini Kit (Qiagen) was used to purify total RNA from rat colon samples, as well as human THP-1 and HT-29 cells according to the manufacturer’s protocol. Total RNA from human biopsies was isolated using NucleoSpin RNA Kit (Macherey-Nagel) according to the manufacturer’s protocol. The quality and the quantity of the extracted RNAs were determined by TapeStation (Agilent) and Qubit Fluorometer (Thermo Fisher).

Reverse transcription was performed by SuperScript VILO Master Mix (Thermo Fisher) according to the manufacturer’s instructions. Gene and microRNA expression was measured by quantitative real-time PCR using the StepOne PCR Systems (Thermo Fisher). SybrGreen technology based QPCR reactions were performed by SYBR Select Master Mix (Thermo Fisher) with specific exon spanning primer sets ( Table 1 ); TaqMan technology based reactions were performed by TaqMan Fast Advanced Master Mix and TaqMan Gene Expression Assays (Thermo Fisher) ( Table 2 ). The ratio of each gene relative to the 18S rRNA was calculated using the 2^-ΔΔCT method.

For miRNA detection TaqMan MicroRNA Reverse Transcription Kit (Thermo Fisher) and TaqMan Universal PCR Master Mix (Thermo Fisher) were used according to the manufacturer’s instructions. The specific miRNA assays were purchased from Thermo Fisher; assay numbers are shown in Table 2 . The ratio of each miRNA relative to the endogenous U6 or RNU48 snRNA for rat or human, respectively, was calculated using the 2^-ΔΔCT method.

miRNA-Seq was performed by TruSeq Small RNA Library Preparation Kit (Illumina), according to the manufacturer’s instructions and sequenced on Illumina MiSeq instrument. Briefly, during 3’ adapter ligation reactions achieved by the T4 RNA Ligase 2 (Deletion Mutant) enzyme, 1 μg total RNA was used from each rat colon sample. After 5’ adapter ligation by T4 RNA Ligase, reverse transcription was performed by SuperScript II Reverse Transcriptase (Thermo Fisher), and the constructs were amplified by PCR. Amplification products were verified on TapeStation High Sense DNA tape (Agilent). cDNA contracts were purified and size selected by gel electrophoresis on 6% Novex TBE gels (Thermo Fisher). To elute the DNA, gel slices were broken and incubated in elution buffer overnight. Gel slurry was removed by Ambion Spin Column, and the DNA was precipitated with isopropanol. Resuspended DNA libraries were quality checked on TapeStation High Sense DNA tape. Final libraries were quantified by Qubit (Thermo Fisher) and sequenced on Illumina MiSeq instrument using the 50-base sequencing chemistry.

The quality checking and filtering of raw Illumina reads was carried out by the QC for Sequencing Reads, Trim Reads modules of CLCBio Genomic Workbench v11.0.1 (Qiagen). The bioinformatics and statistical analysis of the miRNA expression was done by the Small RNA Analysis, Annotate and Merge Counts, Principal Component Analysis and the Empirical Analysis of DGE modules of the same software. Sequenced reads were annotated and counted according the mature Rattus norvegicus miRNA sequence records of the miRBase Database (Release 22.1). The differential expression of miRNAs was considered as significant when the abs(fold change) > 2x and the FDR < 0.05 selection criteria were fulfilled. The downstream gene network studies were performed by Ingenuity Pathway Analysis (IPA, Qiagen) software. Hierarchical clustering and the generation of the heatmap representation of miRNA expression were done by the heatmap.2 function of gplot R package.

Statistical evaluations were performed using the IBM SPSS Statistics program for Windows. Graphs were plotted with GraphPad Prism 6 software. Quantitative data are presented as the mean ± SEM and the significance of difference between sets of data was determined by one-way analysis of variance (ANOVA) following LSD post-hoct test; a p value of less than 0.05 was considered significant.

We started by exploring the extent of global pri-miRNA changes in a rat in vivo model of experimental colitis. To this end, we analyzed RNA-Seq data of the well-characterized 2,4,6-trinitrobenzene sulphonic acid (TNBS) induced rat model of colitis, in which the inflammatory response is due to the generation of transmural oxidative stress and release of proinflammatory mediators. Specifically, in our previous study, we reported that the whole transcriptome analysis of the TNBS induced rat model of IBD revealed significant differences between the inflamed and uninflamed colon samples of TNBS treated rats that associated with the activation of inflammation and carcinogenesis related canonical signaling pathways (24). Here, we again used this RNA-Seq dataset (GEO accession GSE149517) and focused on the expression of pri-miRNAs, the precursors of miRNAs.

The number of known microRNAs is continuously increasing, the number of annotated human and rat precursor/mature miRNA at the time this study was conducted was 1917/2564 and 496/764, respectively, according to the reference repository miRBase release 22 (31). In the TNBS treated rat colons, we identified 28 miRNA precursors with significantly altered expression in the inflamed tissues ( Figure 1 ). Notably, 27 out of 28 pri-miRNAs (eg miR-192, -200b, -375) showed a downregulated expression profile in the inflamed rat colon samples, with only miR-223 exhibiting an upregulated expression ( Figure 1 ).

Next, we applied miRNA-Seq approach for two reasons. First, non-coding RNAs, including miRNAs and their precursors, form a very small portion of the total RNA content of a mammalian cell (32), hence their detection and quantification by RNA-Seq highly depends on the sequencing depth and abundance of the pri-miRNA. Second, pri-miRNAs are processed, first into pre-miRNAs and subsequently to mature miRNAs, which are known regulators of protein expression. In light of these, and in order to monitor the global, unbiased expression profile of functional miRNAs, miRNA-Seq was conducted from the same input samples we used previously for transcriptome (RNA-Seq) profiling. Next, we performed in silico principal component analysis (PCA) of individual samples to visually summarize normalized miRNA expression profiles revealing similarities and differences between samples. We determined that TNBS-treated inflamed rat colon samples (n=4) form a close group that clearly clusters apart from TNBS-treated uninflamed (n=4) as well as control (n=2) samples ( Figure 2A ). In addition, heat map representation of significantly altered miRNAs confirmed differences between inflamed and uninflamed samples of TNBS treated animals, whereas control and uninflamed samples proved to be very similar at miRNA expression level ( Figure 2B ). These data are in perfect agreement with those previously reported at the global transcriptome level (24).

We report that 9% (68) of the 764 rat miRNAs are significantly up- or downregulated in inflamed samples (i.e. show at least a two-fold change with an FDR-adjusted p-value < 0.05; Figure 2B and Supplementary Table 1 ). Interestingly, nearly 65% (44/68) of the miRNAs with altered expression show decrease in the inflamed colon regions. General down-regulation of mature miRNAs as determined by miRNA-Seq ( Figure 2 ) is in perfect agreement with down-regulation of pri-miRNAs observed in RNA-Seq experimental setup ( Figure 1 ).

We previously reported decreased expression of miR-375, miR-192, miR-200b, miR-30a and miR-199a in the inflamed colon regions of TNBS-treated rats and in IBD patients (22, 26, 29). Here we sought to validate the newly established miRNA-Seq dataset by monitoring the expression of these five miRNAs. We found a decreased expression pattern ( Supplementary Table 1 ) of all five miRNAs (miRNA-Seq) that perfectly resembled the previously reported decreased expression pattern these miRNAs (QPCR), indicating that the miRNA-Seq data are of high quality, thus allowing further analysis.

We next sought to assess the extent of the association of RNA-Seq and miRNA-Seq data with respect to pri-miRNA and mature miRNA expression. To this end, we examined the correlation between the previously performed transcriptome analysis (RNA-Seq; Figure 1 ) and miRNA-Seq ( Supplementary Table 1 ) data, resembling pri-miRNA and mature miRNA expression, respectively. We determined that the expression change of numerous miRNAs was detected at both precursor and mature miRNA level (eg. miR-375, miR-194, miR-192, miR-215, miR-26a, miR-27b, miR-200b, miR-196c, miR-205 and miR-223). These data show that the expression pattern of mature miRNAs perfectly follows that of the precursors. It is important to notice, however, that miRNA-Seq detected more significantly changed mature microRNAs. This may be due to 1) the fact that several mature miRNAs are originating from the very same precursor (eg. the miRBase contains 496 precursor and 764 mature rat miRNAs) and 2) the limitations of the RNA-Seq technology which seemingly detected only the most abundant microRNA precursors.

For the genome-wide (RNA-Seq and miRNA-Seq) examination of inflammatory processes in the bowel, we used the 2,4,6-trinitrobenzene sulphonic acid (TNBS) induced rat in vivo model of experimental colitis. Here we aimed to validate the findings made using the murine model on colon biopsies derived from IBD patients. Both the chemically triggered rat tissues and the colons of IBD patients are characterized by the alteration of inflamed and intact regions (24, 26, 29). In order to get a comprehensive view of processes in these tissues, we simultaneously examined the expression of miRNAs both in macroscopically normal (uninflamed) and in inflamed regions.

The expression of several miRNAs, including miR-200b known to play a role in the epithelial to mesenchymal transition (EMT), a fundamental process in wound healing as well as in formation of metastasis (33, 34), is downregulated in the model of in experimental colitis (26). MiR-200b, along with miR-200a, -200c, -141 and -429, is a member of the miR-200 family, which forms clusters in the genome. In human, cluster 1 consists of miR-200a/b and -429 and cluster 2 includes miR-200c and -141. To further investigate the possible dysregulation of the EMT in intestinal inflammation, we next examined the expression of the miR-200 family. We determined that the expression of the rat precursor and mature miRNAs belonging to the miR-200 family (miR-200a/b/c, -141 and -429) strongly decreased in the inflamed tissues ( Figure 1 and Supplementary Table 1 ). To further validate these data, we next measured the expression of mature miRNAs by QPCR and observed perfect overall correlation with our current miRNA-Seq results both in the inflamed colons of TNBS treated rats ( Figures 3A–E ) and, more importantly, in IBD patients ( Figures 3F–J ).

MiRNA-Seq identified two other, highly conserved microRNAs, miR-223 and miR-27b, with significantly altered expression level ( Supplementary Table 1 ). These miRNAs act as tumor promoting oncogenes or suppressors, the function being context- or tissue-dependent (35, 36). In colorectal cancer (CRC), the expression of miR-27b is downregulated while the level of miR-223 directly correlates with disease severity with higher expression associated with more severe symptoms (36, 37). Here we sought to assess the role of miRNA dysregulation in CRC formation, both at microRNA biogenesis and species (rat, human) level. To this end, we first measured the expression changes of pri-, pre- and mature miR-27b and identified a significant decrease at all levels in rat samples ( Figure 4A ) and a moderate decrease of human miR-27b ( Figures 5A–C ). In contrast, the expression of pri- and mature miR-223 markedly increased in both inflamed rat colons ( Figure 4B ) as well as active inflamed human samples ( Figures 5D, E ). This analysis confirmed that the alteration of miRNA expression occurs already at the precursor level and that this phenomenon takes place upon both rat and human colon inflammation.

In addition to miRNAs that significantly changed at precursor level, we also determined the expression pattern of miR-31. Similarly to miR-223 and -27b, miR-31 has a dual function in cancer progression: while in e.g. ovarian, prostate or gastric cancer it acts as tumor repressor, in CRC tissues its elevated expression promotes carcinogenesis (38). As revealed by miRNA-Seq, miR-31 was upregulated in the inflamed rat colon samples ( Supplementary Table 1 ) that is in concordance with the elevated expression of miR-31 in the colon biopsies derived from IBD patients ( Figure 5F ).

Overall, these findings suggest that an important step in the development of IBD-related colorectal cancer is the dysregulated microRNA expression, which is present at all levels (including pri-, pre- and mature miRNA expression) of microRNA biogenesis.

Transcription factors (TFs) known to play a role in EMT (e.g. ZEB2, SNAI1) exhibit increased expression in the inflamed rat and human colon regions which, in turn, is associated with the downregulation of their inhibitory miRNAs (e.g. miR-200b, -192, -375, -143, -30a, -107, -199a and -34a) (22, 26, 29). To further investigate the parallel expression patterns of transcription factors and miRNAs in inflamed colon, we compared the transcriptome and miRNA expression profiles in silico. For this, we searched known transcriptional regulators of the significantly altered miRNAs using the TransmiR database (39, 40). We primarily focused on those miRNAs for which both the precursor and the mature miRNA exhibited an expression change (miR-200 family, miR-27b and miR-223) as we hypothesized that their altered expression may be directly associated with a given transcription factor. We narrowed the list to those transcription factors that have experimentally validated regulatory roles in miRNA expression; furthermore, listed only those that have altered expression in inflamed colon as determined by RNA-Seq ( Table 3 ). Consistent with the hypothesis, we observed a clear correlation between expression change of TFs and their regulated miRNAs. Namely, decreased expression of enhancer (e.g. E2F1, HNF1B) and increased expression of repressor (e.g. NFKB1, ZEB2) TFs associated with the downregulation of miR-200 family members and miR-27b. In contrast, elevated miR-223 and miR-31 expression appears to be the consequence of the upregulation of activator (e.g. CEBPB) and decrease of repressor (e.g. E2F1) TFs in the inflamed colon.

The results above show an intimate relation of TF and miRNA expression in inflamed colon, in which numerous molecules with altered expression can be grouped into a single regulatory circuit ( Figure 6A ). Notably, we identified the miR-223-E2F1-miR-27b slice of this network as a possible central node that includes a feed-back loop, inhibitory and enhancing units. E2F transcription factors have a role in cell cycle regulation, differentiation and development. Canonical E2F protein, E2F1 is able to bind miR-27b promoter and induces miR-27b transcription (49). In contrast, E2F1 and miR-223 form a feed-back loop, in which E2F1 acts as a transcriptional repressor of miR-223 which, in turn, inhibits E2F1 translation (50). In the inflamed colon, upregulation of miR-223 and downregulation of E2F1 is consistent with the reduced expression of miR-27b ( Figure 6A ). Overall, this expression pattern indicates the potential role of miR-27b target genes in the regulation of colon inflammation.

To test this, we next applied Ingenuity Pathway Analysis (IPA) to identify mRNAs that may be affected by this regulatory circuit. We focused on miR-27b targets which exhibited altered expression, again limiting hits to the experimentally validated connections ( Figure 6B ). We identified seven miR-27b targets with altered expression, yet excluded Pmaip1 (Phorbol-12-myristate-13-acetate-induced protein 1, also known as NOXA) from further analysis because its interaction with miR-27b is indirect according to Ingenuity Knowledge Base.

IFI16 (interferon gamma inducible protein 16) is an intracellular DNA sensor that regulates interferon expression, thereby playing a role in innate immune responses (63). Its pathological expression has been reported from colons of IBD patients where, as a target of autoantibodies, it triggers the malfunction of immune homeostasis (64). Similarly, GCA (grancalcin) induces interferon production by the interaction with TLR9 (Toll-like Receptor 9); in addition, it plays a role in autophagy regulation (65, 66). CYP1B1 (cytochrome P450 family 1 subfamily B member 1) enzyme is a member of the cytochrome P450 (CYP450) family that plays a role in the conversion of procarcinogens to carcinogens, in addition, it is involved in the process of CRC progression (67). RUNX1 (Runt-related transcription factor 1) also acts as an oncogene in epithelial tumors; it induces CRC metastasis through the Wnt/β-catenin and EMT pathway activation (68). Transcription factor MEF2C (Myocyte Enhancer Factor-2 C) regulates B- and T-cell development, has an oncogenic function in acute myeloid leukemia and its upregulation is directly associated with CRC progression (69). Finally, MMP13 (matrix metallopeptidase 13) is a collagenase that takes part in the activation of inactive TNFα (tumor necrosis factor alpha), a crucial cytokine in IBD pathogenesis and therapy (70).

Since all six genes were upregulated in the inflamed colon of TNBS induced rats ( Figure 6B ), we next examined the expression of CYP1B1, RUNX1, MEF2C and MMP13 (71–74) in the colon samples of IBD patients. We observed a marked upregulation in the inflamed colon tissues of all four genes ( Figure 7 ), which may be a direct consequence of the lack of repression due to the downregulated miRNA-27b expression ( Figure 5C ).

Finally, we asked whether the global expression changes we detected in colon tissues indeed originate from different cell types. Importantly, miR-223 is the predominant cargo of microvesicles and exosomes in peripheral blood and neutrophil-derived extracellular vesicles; furthermore, polymorphonuclear leukocytes are able to transfer miR-223 containing cargo to alveolar-epithelial cells (35). Therefore, we hypothesized that some of the genes and miRNAs with altered expression in inflamed colon tissues are cell type specific alterations of molecules which exhibit their effects on a different cell type.

To test this hypothesis, we examined the miR-223-E2F1-miR-27b regulatory node - for which direct relationship between the molecules has been previously demonstrated (49, 50, 74) - by monitoring the expression pattern of the members using in vitro cell cultures. Since Caco-2, HT29 and THP-1 cell are well known human-derived in vitro models of IBD (75) we used these cell lines in our further experiments. miRNAs often have cell type specific manifestation, such as the expression pattern of miR-223 which is primarily expressed by myeloid cells (76, 77). Importantly, the initial source of miR-223 in inflamed colon tissue of DSS-induced colitis is from infiltrating neutrophils and monocytes (78). In line with this, we could not detect miR-223 expression either in untreated or in TNFα activated HT-29 colonic epithelial cells. Hence, we used THP-1 leukemia monocytic cells and stimulated them with lipopolysaccharide (LPS) to monitor the inducibility of miR-223. While 6 h of LPS induction had no effect on miR-223 expression, 24 h of LPS stimulation significantly increased its expression ( Figure 8A ). Since we could not detect miR-27b expression neither in unstimulated nor in LPS stimulated Caco-2 cells (data not shown), we next activated HT-29 colonic epithelial cells with TNFα to mimic an inflammatory environment. We detected significantly decreased expression of both miR-27b and its transcriptional activator E2F1 ( Figures 8 B, C ). As a consequence of the possible deregulation, the mRNA level of the miR-27b target MMP13 was markedly upregulated ( Figure 8D ).

Collectively, these results suggest that while cell-type specific differences in miRNA pattern exist between various gut components both in human patients and chemically induced colitis in rat, cellular or soluble cytokine-mediated interactions during inflammation may further modulate the miRNA expression profiles in IBD.