Secondary Logo

Journal Logo


Specific Plasma MicroRNA Signatures in Predicting and Confirming Crohn's Disease Recurrence: Role and Pathogenic Implications

Moret-Tatay, Inés PhD1,2; Cerrillo, Elena MD, PhD1,3; Hervás, David PhD4; Iborra, Marisa MD, PhD1,2,3; Sáez-González, Esteban MD1,3; Forment, Javier PhD5; Tortosa, Luis NP1,2; Nos, Pilar MD, PhD1,2,3; Gadea, Jose PhD5; Beltrán, Belén MD, PhD1,2,3

Author Information
Clinical and Translational Gastroenterology: October 2021 - Volume 12 - Issue 10 - p e00416
doi: 10.14309/ctg.0000000000000416



Approximately half of patients with Crohn's disease (CD) undergo surgery within the first 10 years after diagnosis, even in the era of biological therapy; however, surgical resection is not curative. A quarter of these patients will require at least another bowel resection within 5 years (1). Postoperative recurrence (POR) therefore continues to be a major problem for patients with CD.

A number of clinical factors have been described as predictive of an increased risk of early POR and include active smoking, previous intestinal resection, intestinal penetration, perianal disease, and extensive small bowel resection (>50 cm) (2–6). Environmental and genetic factors are known to play key roles in the pathogenesis of CD, although their role in the postoperative progression to recurrence remains unclear. In recent years, a variety of gene regulation mechanisms have been studied, with special focus on epigenetic mechanisms, which are known to modulate gene expression (7). Among these epigenetic mechanisms, microRNAs (miRNAs) are emerging as important regulators in CD, linking genetic and environmental factors (8,9).

MiRNAs are a subgroup of small non-coding RNAs that regulate gene expression at the post-transcriptional level. Gene regulation mediated by miRNAs has been implicated in numerous biological processes, such as the cell cycle, differentiation, proliferation, apoptosis, and immune functions (10,11).

MiRNAs have been found in tissues, serum, plasma, and other body fluids in a stable form that is protected from endogenous RNase activity (12), which is why miRNAs are resistant to harsh conditions and are now being studied as biomarkers for various diseases, including inflammatory bowel disease (IBD). Numerous studies have been published of miRNA expression patterns in patients with IBD, both in peripheral blood and mucosa (13,14). Their use as biomarkers in diagnosis for predicting IBD course and for response to therapy has also been suggested (15–19).

Specifically in CD, differential miRNA expression profiles have been identified in peripheral blood that can be used to distinguish between patients with CD, those with ulcerative colitis (UC), and healthy controls (17,20,21), as well as specific miRNA profiles that can be used to identify the behavior and progression of CD (18,22). Other studies have also identified predictive changes in miRNA expression during the treatment of patients with biological therapies (23). Therefore, the benefits of using peripheral blood to identify specific miRNA profiles in IBD have been previously stated, promoting the use of easy-to-obtain and easy-to-handle samples for biomarker discovery in IBD (24).

A more recent study conducted on ileal biopsies described a role for miRNAs in patients with recurrent CD, highlighting the potential importance of these regulatory RNAs in early disease stages (25). However, their utility as biomarkers for predicting POR in CD has not yet been explored, and no studies have been performed evaluating plasma miRNAs in the context of CD recurrence. The role of the plasma miRNAs implicated in the pathogenesis of CD in the POR scenario has not been previously addressed, and there remains a lack of knowledge about the biological pathways in which miRNAs exert their functions to modulate certain proteins implicated in recurrence.

In the postoperative setting, selecting the most appropriate therapy after resection remains a challenge because not all operated patients are at high risk of early recurrence, and the amount of time to develop a recurrence differs between individuals. There is therefore a real need in clinical practice for predictive biomarkers that help to identify patients at high risk of POR who might require early aggressive therapy after surgery, to establish the best therapeutic strategy, and to optimize patient outcomes. To this end, the aim of the present study was to identify differential miRNA expression profiles in the plasma of operated patients with CD and the potential use of miRNAs for POR prediction. We also reviewed the genes and biological pathways known to be modulated by the miRNAs found in this study. Furthermore, we performed a more in-depth in silico study to detect the most probable genes regulated by these miRNAs and their pathological implications, which could shed light on the future development of more targeted and effective therapies to prevent POR in patients with CD.



We examined a prospective and consecutive cohort of 67 patients with CD who underwent ileocecal or ileocolonic resection (Table 1). Of these, we selected the 44 patients with only ileal resection for this study, considering that this number could provide appropriate comparative groups. Their attending physician followed the patients for at least 18 months after surgery, and medical decisions were taken according to routine clinical practice. Patients were classified according to the presence or absence of POR during follow-up. POR was assessed by ileocolonoscopy (Rutgeerts score ≥ i2b; see definition of score below) or magnetic resonance imaging enterography (MRE; Sailer score ≥ MR2; see definition of score below [see Figure 1, Supplementary Digital Content 4,]). Most patients underwent postoperative therapy to prevent POR (Table 1).

Table 1. - Demographic and clinical characteristics of the study patients
Presurgery All patients (n = 32) No recurrence (n = 16) Recurrence (n = 16) P values
Age at onset, yr
 Mean (SD)
 Median (1st–3rd Q)
34.2 (13.0)
29.5 (24.0–44.5)
32.1 (11.5)
26.5 (24.0–38.0)
36.3 (14.3)
33.0 (23.5–50.3)
Sex, n (%)
7 (21.9)
25 (78.1)
2 (12.5)
14 (87.5)
5 (31.3)
11 (68.8)
Disease duration until surgery, yr
 Mean (SD)
 Median (1st–3rd Q)
6.5 (8.4)
3.0 (0.0–10.5)
5.8 (7.0)
2.5 (0.0–11.0)
7.2 (9.8)
3.5 (0.0–10.5)
Smoking, n (%)
16 (50.0)
8 (25.0)
8 (25.0)
8 (50.0)
4 (25.0)
4 (25.0)
8 (50.0)
4 (25.0)
4 (25.0)
Previous intestinal surgeries, n (%)
28 (87.5)
3 (9.4)
1 (3.1)
14 (87.5)
1 (6.3)
1 (6.3)
14 (87.5)
2 (12.5)
0 (0.0)
Indication for surgery, n (%)
 Failure of medical treatment
 Penetrating pattern
20 (62.5)
4 (12.5)
8 (25.0)
6 (37.5)
2 (12.5)
8 (50.0)
14 (87.5)
2 (12.5)
0 (0.0)
Disease behavior, n (%)
 B1. Nonstricturing/nonpenetrating
 B2. Stricturing
 B3. Penetrating
 P. Perianal
4 (12.5)
19 (59.4)
9 (28.1)
7 (21.9)
2 (12.5)
6 (37.5)
8 (50.0)
3 (18.8)
2 (12.5)
13 (81.3)
1 (6.3)
4 (25.0)
Presurgery therapy, n (%)
 Anti–TNF-α (IFX)/(ADA)
21 (65.6)
2 (6.3)
14 (43.8)
6 (18.8)/8 (25.0)
9 (56.3)
1 (6.3)
8 (50.0)
3 (18.8)/6 (37.5)
12 (75.0)
1 (6.3)
6 (37.5)
3 (18.8)/2 (12.5)
Fecal calprotectin, µg/g
 Mean (SD)
 Median (1st–3rd Q)
434.9 (371.4)
353 (99.3–742.5)
363.5 (356.3)
207.9 (41.5–698.0)
477.8 (385.8)
409.0 (115.0–870.0)
C-reactive protein, mg/L
 Mean (SD)
 Median (1st–3rd Q)
29.7 (45.1)
9.5 (3.0–48.0)
34.5 (60.4)
14.8 (3.6–27.3)
26.1 (31.6)
9.0 (2.0–48.0)
Fibrinogen, mg/L
 Mean (SD)
 Median (1st–3rd Q)
471.4 (83.8)
488.0 (400.0–526.0)
487.5 (87.1)
510.0 (413.5–535.0)
456.5 (80.8)
450.5 (386.3–524.5)
Postsurgery (1 yr) All patients (n = 32) No recurrence (n = 16) Recurrence (n = 16)
Smoking, n (%)
26 (81.3)
6 (18.8)
15 (93.8)
1 (6.3)
11 (68.8)
5 (31.3)
Postoperative therapy, n (%)
 Anti–TNF-α monotherapy
16 (50.0)
7 (21.9)
1 (3.1)
3 (9.4)
5 (15.6)
10 (62.5)
4 (25.0)
1 (6.3)
1 (6.3)
0 (0.0)
6 (37.5)
3 (9.4)
0 (0.0)
2 (12.5)
5 (31.3)
Fecal calprotectin, µg/g
 Mean (SD)
 Median (1st–3rd Q)
292.8 (321.4)
158.5 (74.8–512.5)
81.9 (54.2)
73.0 (42.0–105.5)
475.7 (345.8)
496.0 (215.0–592.0)
C-reactive protein, mg/L
 Mean (SD)
 Median (1st–3rd Q)
7.5 (18.3)
2.7 (1.0–5.0)
1.8 (1.3)
1.2 (1.0–2.8)
13.5 (25.3)
5.0 (2.7–11)
Fibrinogen, mg/L
 Mean (SD)
 Median (1st–3rd Q)
389.2 (96.6)
382.0 (308.0–444.0)
328.3 (49.0)
315.0 (290.0–377.8)
464.2 (87.9)
450.0 (398.0–495.5)
ADA, adalimumab; IFX, infliximab; TNF-α, tumor necrosis factor alpha; Q, quartile.

This study was conducted at the La Fe University and Polytechnic Hospital (Valencia, Spain) from 2013 to 2018. Ethical approval for the study was obtained from the hospital's Clinical Research Ethics Committee (ref: 2010/0342), in compliance with the Declaration of Helsinki. All patients provided written informed consent before their enrollment.

Endoscopic and radiological assessment

Morphological recurrence was assessed by ileocolonoscopy or MRE within 6–12 months after surgery. Endoscopic activity in the neoterminal ileum was graded according to the Rutgeerts score: i0, no lesions; i1, ≤5 aphthous lesions; i2a, lesions confined to the ileocolonic anastomosis; i2b, >5 aphthous lesions with normal mucosa between lesions; i3, diffuse aphthous ileitis with diffusely inflamed mucosa; and i4, diffuse inflammation with larger ulcers, nodules, and/or narrowing, as previously indicated (26). Recurrence was defined as a Rutgeerts score ≥ i2b (27). MRE was used to assess those asymptomatic patients who declined to undergo endoscopy during the first year or when the neoterminal ileum could not be properly assessed by ileocolonoscopy. The appearance of de novo CD-related lesions on the MRE was graded according to the Sailer score. Recurrence was defined as a Sailer score ≥ MR2 as previously described (27,28).

Blood samples

Peripheral blood samples were prospectively collected in ethylenediaminetetraacetic acid tubes before surgery (within 1 week) and during the postoperative period (scheduled at 1, 3, 6, 9, 12, and 18 months after surgery) or until the onset of morphological recurrence. Blood was acquired after 12 hours of fasting, simultaneously with the routine blood test (C-reactive protein, fibrinogen, white blood cell count, and platelet count) indicated by the clinician in charge. Plasma was obtained within 1 hour of sampling by gradient centrifugation with Histopaque 1077 solution (Sigma) at 213 g for 30 minutes at room temperature. The plasma layer was subsequently transferred to new tubes, aliquoted, and spun down at 2,375g for 10 minutes. The supernatants were placed in new coded tubes and stored at −80 °C until use. Samples with no visible signs of hemolysis at all stages of plasma preparation were considered for the study. In addition, hemolysis assessment was performed in all samples by quantitative reverse transcription–polymerase chain reaction (PCR).

The study was conducted in 2 parts. First, pooled samples from patients who developed recurrence (R) and patients who remained in remission (NR) were used for an exploratory analysis (exploratory cohort), which was undertaken to identify (after data normalization and fold-change comparison) the best plasma miRNAs for discriminating the 2 groups (R vs NR). We established 3 time points: presurgery (PS), 3 months after surgery, and the time of morphological POR detection (R) or, in NR, 1 year after surgery. The blood samples were therefore retrospectively classified as R or NR at PS and at 3 months after surgery, accordingly to the presence (or not) of recurrence at 1 year.

In the second analysis, the inferential cohort, individual plasma samples from the 32 patients included were used to estimate the recurrence risk of the selected miRNA with their associated confidence intervals (CIs) at the time of PS and 1 year after surgery or when the morphological POR was detected.

Profiling of miRNAs

RNA isolation and genetic analysis were performed with Exiqon (Vedbaek, Denmark) reagents and services. Briefly, after RNA extraction from frozen/thawed plasma samples, miRNAs were polyadenylated and reverse transcribed into complementary DNA in a single reaction step. ExiLENT SYBR Green master mix and complementary DNA were transferred to primer-preloaded quantitative PCR panels.

In the initial exploratory cohort stage, the miRNA profiling platform, Exiqon miRCURY LNA Universal RT miRNA PCR Human Panel I + II (Exiqon miRNA qPCR panel, Denmark; 752 miRs), was applied for the selection of candidate miRNAs. We prepared 3 NR pools (PS, 3 months, and 1 year) from 16 gathered plasma samples and 3 R pools (PS, 3 months, and at the time of morphologic recurrence) from 16 gathered plasma samples to identify differently expressed miRNAs between NR and R pools at those fixed time points using a LightCycler 480 (Roche Diagnostics, Mannheim, Germany). The specificity of the PCR products was evaluated by the melting curve analysis. Normalization was based on calculating the mean Cq (the maximum second derivative of the amplification curve) of all the assays in a sample minus each miRNA Cq in that sample.

From the 752 total miRNAs in Panel I + II, 47 miRNAs were detected in all samples, with an average of 119 miRNAs per sample. Assays with several melting points or with melting points deviating from assay specifications were flagged and excluded from the analysis. For the negative control evaluation, sample Cq values were clearly discerned from the background when they were at least 5 Cq below the negative control (see Table 1, Supplementary Digital Content 1,

Unsupervised and supervised data analyses were performed to identify and select candidate miRNAs. We selected 34 miRNAs (after normalization with the mean of Cq values of each sample) for the inferential cohort where the large sample size (32 samples; 16 in NR and 16 in R) was of importance, given it is the most critical step in determining the final predictive models. The miRNA sequences are shown in Supplementary Table 2 (see Supplementary Digital Content 2,

Statistical analysis

The mean value of the technical replicates (samples with at least 2 similar Cq values) in the inferential cohort was analyzed to identify differentially expressed miRNAs across the samples to establish an association between miRNA levels and the presence of clinical and endoscopic CD recurrence.

We first performed an exploratory analysis using unsupervised techniques, such as clustering methods, to classify miRNAs results into groups with a high degree of similarity. We used hierarchical clustering with dendrograms (tree structures in which miRNAs are shown as leaves) and heatmaps (graphical representations of the data in which values are color coded) to visualize and interpret the results.

To assess which miRNAs were able to discriminate between the 2 patient groups, we used an elastic net penalized logistic regression model, which was adjusted to identify the most influential variables in the differentiation between recurrence and nonrecurrence in patients with CD. The shape parameter of the elastic net was set at 0.8, and the penalization factor was selected using 500 repetitions of 10-fold cross validation. From each repetition, the highest lambda at 1 standard error from the minimum error was selected, and the median of the 500 lambda values was used as the final penalization factor. We also used the relaxed elastic net approach to mitigate the overshrinkage bias associated with penalization methods. The corresponding area under the receiver operating characteristic curve (AUC-ROC) for each comparison (PS and 1 year) was estimated. We also performed an internal validation of the selected miRNAs by estimating an optimism-corrected AUC for the various elastic net models using bootstrap with 200 replicates (29). To ease the interpretation of results and facilitate predictions, we developed nomograms for each model. We used R software (version 3.6.2) ( for the statistical analysis and glmnet package (v4.0-2) (30), NMF package (v0.23) (31), and clickR package (v0.5.27).

Functional and pathway enrichment analysis for intersection genes

Predicted targets for the differentially expressed miRNAs were identified using the mirWalk database v3.0 (32), applying mirDB ( and TargetScan ( algorithms for filter options. The threshold cutoff was a 5% false discovery rate and a P < 0.01 using Gene Ontology (GO) biological process and Kyoto Encyclopedia of Genes and Genomes categories enriched in blood-expressed targets (KEGG: Release 95.1, August 1, 2020; GO: Release 95.1, September 1, 2020).

Predicted targets were further filtered using the blood expression data set by Mo et al. (33), downloaded from Pubmed_Geo (a public repository of genetic platform results, which helps users download experiments and curated gene expression profiles). A predicted target was considered expressed in blood if at least 75% of the 13 samples from patients with CD in the GEO GSE112057 series yielded 0.5 or more reads per kilobase million in the RNA-seq data set. Functional annotation was performed using the Database for Annotation, Visualization, and Integrated Discovery (34). We used the Deseq2 algorithm to estimate anticorrelation patterns between target and miRNA expression in the CD and control samples of the Mo et al. data set (33). Heatmaps were generated using the Heatmapper tool (35).


During the 4-year study period, 67 patients with CD underwent surgery, 44 of whom presented exclusive ileal disease localization (L1) and were therefore eligible for the study. Patients with other disease localization and/or other comorbidities were excluded. Three patients dropped out before completing the follow-up for various reasons (1 patient was diagnosed with lung neoplasia 16 months after surgery, and 2 were lost to follow-up because they stopped attending the clinic during the first year). Endoscopic and radiological evaluation of recurrence were performed after a mean interval of 12.13 (SD, 5.1) months after surgery (median, 12 months; 1st–3rd Q, 8.3–13.8). Of the 41 remaining patients, 19 developed recurrence, 16 of whom completed all the visits and had samples available for the epigenetic study. Of the 22 patients who did not present disease recurrence after 1.5 years of follow-up, we selected 16 with the closest clinical and demographic characteristics to the recurrence group to have a balanced cohort, also eliminating those samples whose plasmas presented hemolysis signals or other technical problems that prevented their use in the present project (see Figure 1, Supplementary Digital Content 4,

Table 1 shows the demographic and clinical characteristics of all patients included in the study. The mean age (SD) of the included patients was 34.2 (13.0) years at the time they underwent surgery. The main indication for surgery was stenosis (20, 62.5%). Almost half of the patients were undergoing immunosuppressive therapy (14, 43.8%), and the other half were undergoing biological therapy (14, 43.8%). Glucocorticoids were taken by 61.5% of the patients. We also indicate the clinical characteristics according to the presence (or not) of recurrence during the follow-up (data shown in Table 1). All clinical variables were similar for the 2 groups except for the more penetrating pattern in the nonrecurrence group.

After intestinal resection, the patients underwent postoperative maintenance therapy (Table 1), which was established between the second and third week after surgery. Thiopurines were the most frequently used therapy (16; 50%), followed by anti–tumor necrosis factor (TNF) alpha (7; 22%). As expected, fecal calprotectin (FC), C-reactive protein, and fibrinogen values were higher in the recurrence group at the time of morphological recurrence confirmation. We also observed that FC values were progressively increasing from the third month onward after surgery in those patients who later developed recurrence (26).

We first studied miRNA expression in 3 different scenarios: PS, 3 months (when FC values started to increase in the patients who later developed recurrence), and 1 year for patients who were in remission (NR) or at the time morphological recurrence was detected (R). In this exploratory analysis, we used pooled samples from each time set (PS, 3 months, and 1 year) and a 752 miR detection panel (Exiqon). Reliable detection (fluorescence signals) was achieved in 47 miRNAs (of the 752), and their results were compared between the pooled samples in a 2-way hierarchical clustering (see Figure 2, Supplementary Digital Content 4, The greatest miR differences were observed at PS between the patients with CD who later developed recurrence and the comparative group of NR during the follow-up period. The hierarchical miR profile for NR was shown to be similar to that of the 1-year.

We selected the 34 miRNAs that showed the most consistent differential expression between the pooled samples at PS and 1 year for further experiments, mainly due to economic and experimental limitations. Next, we proceeded with the inferential cohort to obtain a reliable estimation of recurrence risk and the associated CIs. Samples were analyzed individually. Some miRNAs (such as miR-451a and miR-15b-5p) showed differences at this stage (before normalization) that were confirmed after normalization (see Figure 3, Supplementary Digital Content 4, However, for other miRNAs, such as miR.424.5p, differences were lost during data normalization. Those miRNAs with Cq values close to background were excluded from the analysis (8 in PS [25%]; 5 in 1 year [16%]).

The supervised analysis was performed using a logistic regression analysis penalized with the elastic net algorithm. This method provided a list of miRNAs that are predictive of recurrence based on their expression's changes. In the PS comparison among all the analyzed miRNAs, 5 were selected to discriminate between the groups (R and NR): miR-191-5p, miR-15b-5p, miR-106b-5p, miR-451a, and miR-93-5p (Figure 1a and Table 2). All 5 miRNAs were upregulated (Table 2) in the group of patients who later developed CD recurrence. Similarly, at the 1-year after surgery comparison for identification of recurrence in operated patients with CD, 5 miRNAs (miR-15b-5p, miR-451a, miR-93-5p, miR-423-5p, and miR-125b-5p) were selected for discriminating the 2 patient groups (Figure 1b, Table 2). In this comparison, 3 miRNAs were upregulated in the group of patients who presented CD recurrence (miR-15b-5p, miR-451a, and miR-125b-5p), whereas the other 2 (miR-93-5p and miR-423-5p) were downregulated in the R group. Furthermore, 3 miRNAs appeared in both comparisons: miR-15b-5p, miR-451a, and miR-93-5p. Interestingly, the increased levels of miR-15b-5p and miR-451a were shown to be a risk factor at both PS and 1 year. However, the prediction of recurrence risk associated with miR-93-5p differs between PS and 1 year (Table 2), being upregulated before surgery and downregulated in the patients who later presented recurrence.

Figure 1.
Figure 1.:
Heatmap of differential miRNA expression in patients with CD who underwent surgery. Yellow indicates higher than mean intensity (black); blue represents lower than mean intensity. Each row represents 1 miRNA, and each column represents 1 sample (blue for patients in remission; pink for patients with disease recurrence; n = 16 each). (a) Five miRNAs were identified in the miRNA microarray profiling at the time of presurgery. (b) Five miRNAs were identified in the miRNA microarray profiling at the time of morphological POR or, in those who remained in remission, 1 year after surgery. CD, Crohn's disease; POR, postoperative recurrence; miRNA, microRNA.
Table 2. - Prediction of recurrence
Presurgery prediction of recurrence
Variables Estimate Odds ratio
Intercept 0.163
hsa-miR-191-5p 0.642 1.90
hsa-miR-15b-5p 0.517 1.678
hsa-miR-106b-5p 0.369 1.447
hsa-miR-451a 0.519 1.681
hsa-miR-93-5p 0.219 1.244
lambda 0.15
1-year prediction of recurrence
Variables Estimate Odds ratio
Intercept −0.088
hsa-miR-15b-5p 1.231 3.426
hsa-miR-451a 0.74 2.097
hsa-miR-93-5p −0.344 0.709
hsa-miR-423-5p −0.537 0.584
hsa-miR-125b-5p 0.583 1.791
lambda 0.14

For the PS comparison (Figure 2a), the model (relaxed elastic net) achieved an apparent AUC-ROC of 0.88 (95% CI [0.79, 0.98]). For the 1-year comparison (Figure 2b), the same model achieved an apparent AUC-ROC of 0.96 (95% CI [0.91, 1]). To quantify the possible generated optimism due to overfitting, we performed an internal validation of the model using the bootstrap method with 200 replicates. This analysis resulted in a bootstrap-validated AUC-ROC for PS of 0.88, with optimism close to 0, and a bootstrap-validated AUC-ROC of 0.95, with optimism close to 0 for the 1-year comparison (see Figures 4b, Supplementary Digital Content 4,

Figure 2.
Figure 2.:
ROC curve for predicting CD recurrence. (a) A prediction model for early CD recurrence risk at the time of presurgery, with an area under the ROC curve of 0.88 (95% CI [0.79, 0.98]) and a bootstrap-validated AUC of 0.88. (b) A prediction model of CD recurrence risk at the time of morphological POR or, in those who remained in remission, 1 year after surgery, with an area under the ROC curve of 0.96 (95% CI [0.91, 1]) and a bootstrap-validated AUC of 0.95. AUC, area under the ROC; CD, Crohn's disease; CI, confidence interval; POR, postoperative recurrence, ROC, receiver operating characteristic.

To further facilitate the clinical use of the logistic regression results, nomograms for calculating the risk of recurrence score were developed. We used the coefficients of the selected miRNAs from the multivariate analysis as weights to develop the nomograms. This tool facilitates the model's practical application for making predictions of the expected risk of recurrence for a given patient before the patient undergoes a surgical procedure (Figure 3a). We also created a confirmatory nomogram for recurrence (1 year) at the time of morphological assessment by ileocolonoscopy (Figure 3b).

Figure 3.
Figure 3.:
Nomograms for risk prediction of recurrence at the time of presurgery (a), and for confirmation of recurrence (1 year) at the time of morphological assessment by ileocolonoscopy/MRE (b), during the first year after surgery. MRE, magnetic resonance imaging enterography.

We performed an in silico analysis for predicting potential candidate target genes of the selected miRNAs (Table 2). We identified 22 pathways for the upregulated miRNAs on the PS data set, 11 pathways for the upregulated miRNAs on the 1-year data set, and 4 for the downregulated miRNAs on the 1-year data set (Table 3). At PS, miRNAs had enriched GO terms related to tumoral processes (e.g., pancreatic cancer and chronic myeloid leukemia), intracellular elements (e.g., endocytosis and mitogen-activated protein kinase signaling pathway), and the TNF cytokine (e.g., TNF signaling pathway). For 1 year, the upregulated miRNAs were highly enriched in the corresponding GO terms for intracellular processes (e.g., cell cycle, ubiquitin-mediated proteolysis, and activation of JUN kinase kinase activity), whereas downregulated miRNAs were enriched for transcription and endocytosis.

Table 3. - Significant GO (Biological process) and KEGG categories enriched in blood-expressed targets of the miRNAs
Category Term P value FDR (%)
Presurgery data set (upregulated miRNAs)
 Pancreatic cancer hsa05212 1.79E-05 0.02
 Chronic myeloid leukemia hsa05220 3.82E-05 0.05
 Renal cell carcinoma hsa05211 1.60E-04 0.20
 Endocytosis hsa04144 3.45E-04 0.42
 Non–small-cell lung cancer hsa05223 4.52E-04 0.56
 TNF signaling pathway hsa04668 6.18E-04 0.76
 Negative regulation of translation GO:0017148 5.44E-04 0.90
 Bladder cancer hsa05219 7.54E-04 0.92
 Glioma hsa05214 0.0010 1.23
 Positive regulation of nuclear-transcribed mRNA poly(A) tail shortening GO:0060213 8.13E-04 1.35
 Negative regulation of transforming growth factor beta receptor signaling pathway GO:0030512 9.23E-04 1.53
 Melanoma hsa05218 0.0016 1.96
 MAPK signaling pathway hsa04010 0.0018 2.20
 Retrograde transport; endosome to Golgi GO:0042147 0.0014 2.26
 Estrogen signaling pathway hsa04915 0.0019 2.29
 cGMP-PKG signaling pathway hsa04022 0.0020 2.43
 Positive regulation of nuclear-transcribed mRNA catabolic process; deadenylation-dependent decay GO:1900153 0.0017 2.84
 Cell cycle GO:0007049 0.0017 2.84
 FoxO signaling pathway hsa04068 0.0026 3.21
 Endosome organization GO:0007032 0.0023 3.83
 Positive regulation of GTPase activity GO:0043547 0.0026 4.28
 Focal adhesion hsa04510 0.0037 4.43
1-yr data set (upregulated miRNAs)
 Cell cycle GO:0007049 5.21E-04 0.83
 Ubiquitin-mediated proteolysis hsa04120 0.0011 1.32
 Thymus development GO:0048538 0.0011 1.73
 Chronic myeloid leukemia hsa05220 0.0015 1.83
 Activation of JNKK activity GO:0007256 0.0016 2.58
 Sphingolipid signaling pathway hsa04071 0.0028 3.36
 Neurotrophin signaling pathway hsa04722 0.0028 3.36
 Focal adhesion hsa04510 0.0028 3.36
 ErbB signaling pathway hsa04012 0.0035 4.18
 Prostate cancer hsa05215 0.0037 4.38
 Non–small-cell lung cancer hsa05223 0.0041 4.93
The 1-yr data set (downregulated miRNAs)
 Positive regulation of transcription; DNA-templated GO:0045893 9.81E-06 0.02
 Endocytosis hsa04144 4.00E-04 0.48
 Pathways in cancer hsa05200 0.0024 2.91
 Bladder cancer hsa05219 0.0035 4.22
cGMP-PKG, cGMP-dependent protein kinase; ErbB, erythroblastic leukemia viral oncogene homolog; FDR, false discovery rate; FoxO, forkhead box O; GO, gene ontology; JNKK, JUN kinase kinase; KEGG, Kyoto Encyclopedia of Genes and Genomes; MAPK, mitogen-activated protein kinase; miRNA, microRNA; TNF, tumor necrosis factor.
clickR package (v0.5.27).

To explore the biological functions of the selected miRs, we also constructed a hierarchical clustering analysis of connections between intersection genes (downloaded from Pubmed_Geo public repository, Mo, et al. (33)) and miRNAs (Figure 4). CCND2 and BCL9L genes appeared related to PS miR profiles, whereas SENP5 and AKT3 were shared between PS and 1-year upregulated miR results. In the 1-year upregulated group, the genes SUV39H1 and MAPK3K10 also seemed to be implicated, whereas for the 1-year downregulated group, only the CREB5 gene seemed to be involved.

Figure 4.
Figure 4.:
In silico analysis with a hierarchical clustering analysis of intersection genes. Each row represents one of the intersection genes, and each column represents a blood sample from patients with Crohn's disease or healthy controls. Blue represents lower expression and yellow higher expression in each comparative group. CD, Crohn's disease; miRNA, microRNA.

For further clarification of the interactions between the selected miRNAs and the genes they regulate, we constructed miRNA-intersection gene networks using the mirWalk database for each comparison set (see Figures 5a, b, Supplementary Digital Content 4,


We found specific plasma miRNA signatures that can predict which operated patients with CD will develop recurrence within 1 year after surgery. We found a specific miRNA signature that can be used at the moment of surgery to identify patients at high risk of POR (miR-191-5p, miR-15b-5p, miR-106b-5p, hsa.miR-451a, and miR-93-5p), with excellent discriminative capacity. This novel finding could be useful in clinical practice, enabling clinicians to adjust the treatment to prevent POR and optimize patient outcomes. Furthermore, we focused the study on plasma miRNAs to simplify the obtention of samples for the analysis and not involve an invasive procedure such as endoscopy. Similarly, another signature (miR-15b-5p, miR-451a, miR-93-5p, miR-423-5p, and miR-125b-5p) has shown a strong capacity to confirm the presence of recurrence within 1 year of surgery, which could also help spare patients from colonoscopies. Given the high discriminative capacity of the AUC-ROC for the miRNA groups described, we developed nomograms to facilitate the clinical use of these results. Recently, the ability to accurately risk stratify patients has garnered considerable interest, especially given that clinicians have turned toward a top-down form of management to promote mucosal healing and minimize exposure to corticosteroids. This approach requires accurate identification of high-risk patients to minimize overtreatment of lower-risk patients, with the associated risk of treatment-related adverse events and costs. The POR scenario is a drug-demanding and cost-demanding situation in which orientation for personalized management is needed. The inclusion of nomograms in the algorithm for managing operated patients with CD could therefore be considered.

Researchers have been interested in miRs, not only because of their potential use as biomarkers but also as an aid in unveiling the molecular mechanisms that participate in disease progression. Studies have dealt with the potential use of plasma miRNAs in the diagnosis of IBD (10,22,36) and in differentiating active from nonactive ileal CD (37). In the context of CD recurrence, however, little is known. Several studies have been performed on mucosal miRNA expression in the scenario of recurrence. To our knowledge, however, this is the first study to explore plasma miRNA in the recurrent CD scenario. Verstockt et al. (25) compared intestinal miRNAs from patients with postoperative recurrent CD with non-IBD controls and with newly diagnosed or late-stage patients with CD to study the early events in the course of CD, observing a few differentially expressed miRNAs between those groups, suggesting that miRNA dysregulation plays an important role at the mucosal level after the disease is reset by surgery. We showed that 2 miRNAs (miR-15b-5p and miR-451a) are upregulated both before surgery and after recurrence and could therefore indicate their importance in the disease's pathogenesis and chronic nature. Two miRNAs (miR-125b-5p and miR-423-5p) then appear with the establishment of recurrence and could therefore be participating in the development of active lesions. Last, 1 miRNA (miR-93-5p) is upregulated before surgery and seems to be downregulated in patients with recurrent CD. The reason for this postoperative change remains unclear but could also be involved in the pathogenic events. None of the miRNAs described in our study coincide with those reported by Verstockt et al. (25); however, the fact that the authors studied ileal mucosa samples while we used plasma samples could be sufficient reason to justify such a different profile.

The miRNA miR-15b-5p, previously associated negatively with apoptosis and drug resistance, is a key inflammatory mediator of the nuclear factor kappa-light-chain-enhancer of activated B cells family in colon cancer cells (38). In other cell types, miR-15b-5p can restrain triggered apoptosis by downregulating the Bax proapoptotic protein, cleaving caspase-3, and upregulating the antiapoptotic protein Bcl-2 (39). This miRNA might also repress induced oxidative stress by reducing malondialdehyde content and NOX4 expression and enhancing activities of superoxide dismutase and catalase (39). Our group had previously shown the involvement of oxidative stress and its regulatory enzymes in the pathogenesis of CD (40) together with resistant apoptotic capacities in peripheral lymphocytes (41), which could be influenced by the effect of miR-15b-5p on apoptosis. Catalase was permanently inhibited in the peripheral blood cells of patients with CD. Although this result would be contrary to the effect described for miR.15b.5p on catalase, it could be argued that catalase inhibition is not reverted in the lymphocytes of patients with CD, even if miR.15b.5p is overexpressed. This fact points to another mechanism of action specifically responsible for maintaining catalase inhibition despite the oxidative stress present in patients with CD.

Although miR-451 has not previously been described in the context of IBD, this study revealed it to be permanently increased. The role of miR-451 in cancer has been reported to function by repressing colorectal cancer by inhibiting tumor growth and IL6R expression (42) and the PI3K/AKT/mTOR signaling pathway (43), being dysregulated in primary colorectal tumors and in patients' sera (44). It therefore seems that this miRNA could have opposite functions to those of miR-15b-5p, and a compensatory balance between the 2 miRNAs is probably a biological target of the cells to prevent recurrence, although this hypothesis needs further investigation.

We found that miR-106b-5p was upregulated only in the PS sample and has also been found to have a role in controlling apoptosis by inhibiting the block of TNF-α–induced activation of caspase-3 (45). Treg differentiation has been observed to be favored when miR-106b-5p was silenced, whereas miR-106b-5p overexpression promoted an immune imbalance toward Th17 response (46). It therefore seems that, in the PS scenario, patients with CD harboring higher miR-106b-5p levels would be predisposed to later develop recurrence by favoring a more proinflammatory immune phenotype.

In experiments to identify blood miRNAs able to distinguish CD from UC, miR-191 emerged as specific (hyperexpressed) in active CD with respect to active UC and healthy individuals (47). This miRNA presented a similar expression in the intestinal biopsies of patients with IBD (47). Circulating miRNAs could therefore be a reflection of their expression at the intestinal site. In addition, the modulation exerted over the Wnt Signaling, a pathway related to CD, also indicates its importance in this pathology (48).

Increased miR-125b-5p levels have been observed when CD4+ T-cell differentiation is inhibited, mainly by targeting interferon-gamma, interleukin-2, and interleukin-10 (44). This miRNA has also been associated with the inhibition of B-cell activation, the promotion of macrophage activation, and the induction of apoptosis or proliferation through the modulation of p53 expression (49). Therefore, miR-125b-5p emerges as an important miRNA to consider as a target for future personalized therapies to prevent recurrence in operated patients with CD.

The miRNA miR-423-5p has previously been associated with IBD and with colorectal carcinoma. Its levels were increased in patients with IBD but were lower when they developed colorectal carcinoma (50). In our patients, miR-423-5p was downregulated, although it was active in the neoileum. The fact that our study did not involve the colon could at least partly explain this discrepancy with previous studies.

On the other hand, miR-93-5p can block the transforming growth factor beta signaling pathway that, in turn, has a critical role in regulating cell growth and differentiation (51). Whereas the increase in this miRNA has a role in PS risk, the decrease in this miRNA confers a risk of recurrence in the postsurgery scenario. No explanation for this can be concluded from our study and therefore warrants further investigation.

The in silico analysis led to the identification of 22 biological pathways related to the selected miRNAs on the PS data set. The tumorigenic pathways were the most representative, consistent with the body of evidence derived from the initial studies for miRNA, which focus predominantly on cancer research. However, other pathways critically important in IBD emerged, such as the TNF-α and transforming growth factor beta pathways, and are possibly involved. We observed 7 genes that can be regulated by the selected miRNAs. Four genes were involved in the PS data (CCND2, BCL9L, SENP5, and AKT3), and 5 were involved at 1 year (SUV39H1, MAP3K10, CREB5, SENP5, and AKT3), although 2 of them were present at both times (SENP5 and AKT3).

CCND2, a gene of the cyclin D family, is silenced by epigenetic mechanisms when the histone demethylase, Jumonji domain-containing protein 3, is inhibited (52), promoting Th2 and Th17 cells while inhibiting Treg and Th1 differentiation in the small intestine (53). These observations are in agreement with our in silico analysis, in which CCND2 was downregulated in patients with CD who have a high risk of developing recurrence, possibly influenced by the higher proinflammatory phenotype of Th cells. BCL9L, the other gene that appeared in the in silico PS analysis, presents similarities to CCND2, promoting intestinal tumor progression, as a cofactor of Wnt signaling, and its inhibition increases cytotoxic lymphocytes while inhibiting Tregs (54).

SENP5 and AKT3 were both downregulated at PS and at 1 year. SENP5 belongs to the SUMO isopeptidases and is necessary for maintaining normal mitochondrial morphology and regulating intracellular levels of reactive oxygen species (ROS). The reduction of SENP5 gives rise to an increase in the production of free radicals (55). The presence of H2O2, a ROS present during active CD (40), inhibits SUMO activity (55). Thus, our results are in agreement with the ROS status described in CD. Akt signaling also regulates cellular ROS metabolism (56) and is implicated in epithelial regeneration after inflammation through the Wnt/β-catenin pathway (57).

The only upregulated gene was CREB5 (cyclic adenosine monophosphate response element regulating a variety of cellular responses), a gene previously associated with CD but not with UC (58). CREB5 expression was significantly higher in active CD than in inactive CD (59), which agrees with our in silico analysis in patients with recurrent CD. All these referred pathways have been summarized up in a supplementary Table 3 (see Supplementary Digital Content 3, (35,36,39–42,44–46,48,60–65).

Our study had a number of limitations. Our sample involved only ileal CD resections, which is an advantage for performing comparisons between homogenous phenotypic patients with CD; however, it could also be a handicap, given that the study does not include patients with colonic CD. Whether the same miRNA signatures are valid for the patients with colon involvement remains to be clarified. One-quarter of the patients had perianal disease, and although it has not been a factor risk for recurrence (26), a non-miRNA subanalysis specifically for these patients has been performed. None of the patients with fistulizing disease developed POR; however, these patients are those undergoing more intense therapy for POR prevention (26). Although the miRNA signatures have been validated in individual patients, the use of these signatures and the developed nomograms in clinical practice require a multicentric study that could validate their use in clinical management together with a cost-efficacy analysis. Similarly, future work should be performed to develop the technology for an easy-to-use kit that could facilitate the measurement of the specific plasma miRs, offering quick and reliable results. Whether this strategy could increase the safety of colonoscopies in operated patients deserves further investigation.

In conclusion, the miRNAs that can predict or confirm POR are involved in the regulation of various cellular processes, although they seem to be particularly relevant for the regulation of apoptosis, autophagy, the establishment of proinflammatory immunological T-cell clusters, ROS production, and metabolism. We have shown that a plasma miR signature can identify patients at high risk of POR at the time of surgery, which could help clinicians decide on postoperative therapeutic management and optimize patient outcomes. Another miRNA signature can be used to confirm POR after its establishment, which could help spare patients from invasive explorations. We developed nomograms to facilitate the clinical use of both signatures. Our results showed that an miRNA signature can strongly predict POR at the time of surgery and thus could be a powerful tool for use in clinical practice, which we intend to explore in the near future.


Guarantor of the article: Belén Beltrán, MD, PhD, and Inés Moret-Tatay, PhD.

Specific author contributions: I.M-T. and B.B.: study design, data collection, experimental analysis and interpretation, manuscript writing, and critical review. J.G., E.C., and M.I.: study design, data collection, and manuscript review. E.S.-G.: data interpretation and critical review. D.H. and J.F.: statistical analysis and interpretation. L.T.: data collection and critical review. P.N.: study design and critical review. All the authors reviewed and approved the final manuscript.

Financial support: Supported by grants from the Spanish Healthcare Institute Carlos III [PI18/01552] and the Valencian Society of Digestive Pathology [SVPD, Valencia, Spain].

Potential competing interest: None to report.

Previous presentation: These data were presented at the 15th Congress of European Crohn's and Colitis Organisation (ECCO); February 12-15, 2020; Vienna, Austria.

Study Highlights


  • ✓ MicroRNAs in plasma have shown their potential use in the diagnosis of inflammatory bowel disease.
  • ✓ Postoperative recurrence is a major problem for patients with Crohn's disease (CD).
  • ✓ There is a need for biomarker discovery in the CD recurrence scenario to help clinicians assess recurrence and manage the disease.


  • ✓ Our specific microRNA signatures can preoperatively predict which patients will develop CD recurrence and confirm recurrence within 1 year postsurgery.
  • ✓ These signatures are involved in the regulation of various cellular processes.


1. Frolkis AD, Lipton DS, Fiest KM, et al. Cumulative incidence of second intestinal resection in Crohn's disease: A systematic review and meta-analysis of population-based studies. Am J Gastroenterol 2014;109:1739–48.
2. Ng SC, Lied GA, Arebi N, et al. Clinical and surgical recurrence of Crohn's disease after ileocolonic resection in a specialist unit. Europ J Gastroenterol Hepatol 2009;21:551–7.
3. Simillis C, Yamamoto T, Reese GE, et al. A meta-analysis comparing incidence of recurrence and indication for reoperation after surgery for perforating versus nonperforating Crohn's disease. Am J Gastroenterol 2008;103:196–205.
4. Hofer B, Bottger T, Hernandez-Richter T, et al. The impact of clinical types of disease manifestation on the risk of early postoperative recurrence in Crohn's disease. Eur J Gastroenterol Hepatol 2001;48:152–5.
5. Parente F, Sampietro GM, Molteni M, et al. Behaviour of the bowel wall during the first year after surgery is a strong predictor of symptomatic recurrence of Crohn's disease: A prospective study. Aliment Pharmacol Ther 2004;20:959–68.
6. Bernell O, Lapidus A, Hellers G. Risk factors for surgery and postoperative recurrence in Crohn's disease. Ann Surg 2000;231:38–45.
7. Moret-Tatay I, Cerrillo E, Sáez-González E, et al. Identification of epigenetic methylation signatures with clinical value in Crohn's disease. Clin Transl Gastroenterol 2019;10:e00083.
8. Yung RL, Julius A. Epigenetics, aging, and autoimmunity. Autoimmun 2008;41:329–35.
9. Feil R, Fraga MF. Epigenetics and the environment: Emerging patterns and implications. Nat Rev Genet 2012;13:97–109.
10. Iborra M, Bernuzzi F, Invernizzi P, et al. MicroRNAs in autoimmunity and inflammatory bowel disease: Crucial regulators in immune response. Autoimmun Rev 2012;11:305–14.
11. Ruan K, Fang X, Ouyang G. MicroRNAs: Novel regulators in the hallmarks of human cancer. Cancer Lett 2009;285:116–26.
12. Lodes MJ, Caraballo M, Suciu D, et al. Detection of cancer with serum miRNAs on an oligonucleotide microarray. PLoS One 2009;14(4):e6229.
13. Wu F, Zhang S, Dassopoulos T, et al. Identification of microRNAs associated with ileal and colonic Crohn's disease. Inflamm Bowel Dis 2010;16:1729–38.
14. Fasseu M, Treton X, Guichard C, et al. Identification of restricted subsets of mature microRNA abnormally expressed in inactive colonic mucosa of patients with inflammatory bowel disease. PLoS One 2010;5:e13160.
15. Archanioti P, Gazouli M, Theodoropoulos G, et al. Micro-RNAs as regulators and possible diagnostic biomarkers in inflammatory bowel disease. J Crohns Colitis 2011;5:520–4.
16. Coskun M, Bjerrum JT, Seidelin JB, et al. MicroRNAs in inflammatory bowel disease - pathogenesis, diagnostics and therapeutics. World J Gastroenterol 2012;18:4629–34.
17. Jensen MD, Andersen RF, Christensen H, et al. Circulating microRNAs as biomarkers of adult Crohn's disease. Eur J Gastroenterol Hepatol 2015;27:1038–44.
18. Peck BC, Weiser M, Lee SE, et al. MicroRNAs classify different disease behavior phenotypes of Crohn's disease and may have prognostic utility. Inflamm Bowel Dis 2015;21:2178–87.
19. Chen WX, Ren LH, Shi RH. Implication of miRNAs for inflammatory bowel disease treatment: Systematic review. World J Gastrointest Pathophysiol 2014;5:63–70.
20. Wu F, Guo NJ, Tian H, et al. Peripheral blood microRNAs distinguish active ulcerative colitis and Crohn's disease. Inflamm Bowel Dis 2011;17:241–50.
21. Mohammadi A, Kelly OB, Smith MI, et al. Differential miRNA expression in ileal and colonic tissues reveals an altered immunoregulatory molecular profile in individuals with Crohn's disease versus healthy subjects. J Crohns Colitis 2019;13:1459–69.
22. Koliani-Pace JL, Siegel CA. Prognosticating the course of inflammatory bowel disease. Gastrointest Endosc Clin N Am 2019;29:395–404.
23. Fujioka S, Nakamichi I, Esaki M, et al. Serum microRNA levels in patients with Crohn's disease during induction therapy by infliximab. J Gastroenterol Hepatol 2014;29:1207–14.
24. Iborra M, Bernuzzi F, Correale C, et al. Identification of serum and tissue micro-RNA expression profiles in different stages of inflammatory bowel disease. Clin Exp Immunol 2013;173:250–8.
25. Verstockt S, De Hertogh G, Van der Goten J, et al. Gene and miRNA regulatory networks during different stages of Crohn's disease. J Crohns Colitis 2019;13:916–30.
26. Cerrillo E, Moret I, Iborra M, et al. Nomogram combining fecal calprotectin levels and plasma cytokine profiles for individual prediction of postoperative Crohn's disease recurrence. Inflamm Bowel Dis 2019;25:1681–91.
27. Domènech E, Mañosa M, Bernal I, et al. Impact of azathioprine on the prevention of postoperative Crohn's disease recurrence: Results of a prospective, observational, long-term follow-up study. Inflamm Bowel Dis 2008;14:508–13.
28. Cerrillo E, Beltrán B, Pous S, et al. Fecal calprotectin in ileal Crohn's disease: Relationship with magnetic resonance enterography and a pathology score. Inflamm Bowel Dis 2015;21:1572–9.
29. Steyerberg EW, Harrell FE, Borsboom GJ, et al. Internal validation of predictive models: Efficiency of some procedures for logistic regression analysis. J Clin Epidemiol 2001;54:774–81.
30. Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models via coordinate descent. J Stat Softw 2010;33:1–22.
31. Gaujoux R, Seoighe C. A flexible R package for nonnegative matrix factorization. BMC Bioinformatics 2010;11:367.
32. Sticht C, De La Torre C, Parveen A, et al. MiRWalk: An online resource for prediction of microRNA binding sites. PLoS One 2018;13:e0206239.
33. Mo A, Marigorta UM, Arafat D, et al. Disease-specific regulation of gene expression in a comparative analysis of juvenile idiopathic arthritis and inflammatory bowel disease. Genome Med 2018;10:48.
34. Huang DW, Sherman BT, Lempicki RA. Systematic and integrative analysis of large gene lists using DAVID Bioinformatics Resources. Nature Protocols 2009;4:44–57.
35. Babicki S, Arndt D, Marcu A, et al. Heatmapper: Web-enabled heat mapping for all. Nucleic Acids Res 2016;44:W147–53.
36. Netz U, Carter J, Eichenberger MR, et al. Plasma microRNA profile differentiates Crohn's colitis from ulcerative colitis. Inflamm Bowel Dis 2017;24:159–65.
37. Guo Z, Wu R, Gong J, et al. Altered microRNA expression in inflamed and non‐inflamed terminal ileal mucosa of adult patients with active Crohn's disease. J Gastroenterol Hepatol 2015;30:109–16.
38. Zhao C, Zhao Q, Zhang C, et al. MiR-15b-5p resensitizes colon cancer cells to 5-fluorouracil by promoting apoptosis via the NF-κB/XIAP axis. Sci Rep 2017;7:4194.
39. Fu Y, Wang C, Zhang D, et al. MiR-15b-5p ameliorated high glucose-induced podocyte injury through repressing apoptosis, oxidative stress, and inflammatory responses by targeting Sema3A. J Cell Physiol 2019;234:20869–78.
40. Beltrán B, Nos P, Dasí F, et al. Mitochondrial dysfunction, persistent oxidative damage, and catalase inhibition in immune cells of naïve and treated Crohn's disease. Inflamm Bowel Dis 2010;16:76–86.
41. Moret I, Rausell F, Iborra M, et al. Apoptosis resistance of Crohn's Disease blood T-cells depends on catalase activity inhibition. Gastroenterology 2012;142;5, S1, S-885.
42. Bai H, Wu S. MiR-451: A novel biomarker and potential therapeutic target for cancer. Onco Targets Ther 2019;12:11069–82.
43. Streleckiene G, Inciuraite R, Juzenas S, et al. MiR-20b and miR-451a are involved in gastric carcinogenesis through the PI3K/AKT/mTOR signaling pathway: Data from gastric cancer patients, cell lines and ins-gas mouse model. Int J Mol Sci 2020;21:3.
44. Zhang Z, Zhang D, Cui Y, et al. Identification of microRNA-451a as a novel Circulating biomarker for colorectal cancer diagnosis. Biomed Res Int 2020:5236236.
45. Zhang J, Li SF, Chen H, et al. MiR-106b-5p inhibits Tumor Necrosis Factor-α-induced apoptosis by targeting phosphatase and tensin homolog deleted on chromosome 10 in vascular endothelial cells. Chin Med J (Engl) 2016;129:1406–12.
46. Li JQ, Tian JM, Fan XR, et al. MiR-106b-5p induces immune imbalance of Treg/Th17 in immune thrombocytopenic purpura through NR4A3/Foxp3 pathway. Cell Cycle 2020;19:1265–74.
47. Paraskevi A, Theodoropoulos G, Papaconstantinou I, et al. Circulating microRNA in inflammatory bowel disease. J Crohns Colitis 2012;6:900–4.
48. Pehlivan M, Soyoz M, Cerci B, et al. sFRP1 expression induces miRNAs that modulate Wnt signaling in chronic myeloid leukemia cells [in Russian]. Mol Biol (Mosk) 2020;54:626–33.
49. Zhu Y, Zhang S, Li Z, et al. MiR-125b-5p and miR-99a-5p downregulate human γδ T-cell activation and cytotoxicity. Cell Mol Immunol 2019;16:112–25.
50. Fang Z, Tang J, Bai Y, et al. Plasma levels of microRNA-24, microRNA-320a, and microRNA-423-5p are potential biomarkers for colorectal carcinoma. J Exp Clin Cancer Res 2015;34:86.
51. Hu B, Mao Z, Du Q, et al. miR-93-5p targets Smad7 to regulate the transforming growth factor-β1/Smad3 pathway and mediate fibrosis in drug-resistant prolactinoma. Brain Res Bull 2019;149:21–31.
52. Ray G, Longworth MS. Epigenetics, DNA organization, and inflammatory bowel disease. Inflamm Bowel Dis 2019;25:235–47.
53. Li Q, Zou J, Wang M, et al. Critical role of histone demethylase Jmjd3 in the regulation of CD4+ T-cell differentiation. Nat Commun 2014;5:5780.
54. Feng M, Jin JQ, Xia L, et al. Pharmacological inhibition of β-catenin/BCL9 interaction overcomes resistance to immune checkpoint blockades by modulating Treg cells. Sci Adv 2019;5:eaau5240.
55. Zunino R, Schauss A, Rippstein P, et al. The SUMO protease SENP5 is required to maintain mitochondrial morphology and function. J Cell Sci 2007;120:1178–88.
56. Tokuhira N, Kitagishi Y, Suzuki M, et al. PI3K/AKT/PTEN pathway as a target for Crohn's disease therapy. Int J Mol Med 2015;35:10–6.
57. Moparthi L, Koch S. Wnt signaling in intestinal inflammation. Differentiation 2019;108:24–32.
58. Drobin K, Assadi G, Hong MG, et al. Targeted analysis of serum proteins encoded at known inflammatory bowel disease risk loci. Inflamm Bowel Dis 2019;25:306–16.
59. Qiao YQ, Huang ML, Xu AT, et al. LncRNA DQ786243 affects Treg related CREB and Foxp3 expression in Crohn's disease. J Biomed Sci 2013;20:87.
60. Liu X, Dong Y, Song D. Inhibition of microRNA-15b-5p attenuates the progression of oral squamous cell carcinoma via modulating the PTPN4/STAT3 Axis. Cancer Manag Res 2020;12:10559–72.
61. Gu H, Gu S, Zhang X, et al. miR-106b-5p promotes aggressive progression of hepatocellular carcinoma via targeting RUNX3. Cancer Med 2019;8:6756–67..
62. Zhou LY, Zhang FW, Tong J, et al. MiR-191-5p inhibits lung adenocarcinoma by repressing SATB1 to inhibit Wnt pathway. Mol Genet Genomic Med 2020;8:e1043.
63. Chen B, Zheng ZY, Yang JZ, et al. MicroRNA-191-5p promotes the development of osteosarcoma via targeting EGR1 and activating the PI3K/AKT signaling pathway. Eur Rev Med Pharmacol Sci 2019;23:3611–20.
64. Wang M, Guo J, Zhao YQ, et al. IL-21 mediates microRNA-423-5p/claudin-5 signal pathway and intestinal barrier function in inflammatory bowel disease. Aging (Albany NY) 2020;12:16099–110.
65. Dawidowicz M, Kula A, Mielcarska S, et al. miREIA - An immunoassay method in assessment of microRNA levels in tumor tissue-pilot study. The impact of miR-93-5p, miR-142-5p and IFNγ on PD-L1 level in colorectal cancer. Acta Biochim Pol 2021;68:247–254.

Supplemental Digital Content

© 2021 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of The American College of Gastroenterology