Epithelial down‐regulation of the miR‐200 family in fibrostenosing Crohn's disease is associated with features of epithelial to mesenchymal transition

Abstract Intestinal mesenchymal cells deposit extracellular matrix in fibrotic Crohn's disease (CD). The contribution of epithelial to mesenchymal transition (EMT) to the mesenchymal cell pool in CD fibrosis remains obscure. The miR‐200 family regulates fibrosis‐related EMT in organs other than the gut. E‐cadherin, cytokeratin‐18 and vimentin expression was assessed using immunohistochemistry on paired strictured (SCD) and non‐strictured (NSCD) ileal CD resections and correlated with fibrosis grade. MiR‐200 expression was measured in paired SCD and NSCD tissue compartments using laser capture microdissection and RT‐qPCR. Serum miR‐200 expression was also measured in healthy controls and CD patients with stricturing and non‐stricturing phenotypes. Extra‐epithelial cytokeratin‐18 staining and vimentin‐positive epithelial staining were significantly greater in SCD samples (P = 0.04 and P = 0.03, respectively). Cytokeratin‐18 staining correlated positively with subserosal fibrosis (P < 0.001). Four miR‐200 family members were down‐regulated in fresh SCD samples (miR‐141, P = 0.002; miR‐200a, P = 0.002; miR‐200c, P = 0.001; miR‐429; P = 0.004); miR‐200 down‐regulation in SCD tissue was localised to the epithelium (P = 0.001‐0.015). The miR‐200 target ZEB1 was up‐regulated in SCD samples (P = 0.035). No difference in serum expression between patient groups was observed. Together, these observations suggest the presence of EMT in CD strictures and implicate the miR‐200 family as regulators. Functional studies to prove this relationship are now warranted.

of fibrotic strictures in fibrostenosing CD are not well defined. The estimated lifetime risk of surgery for patients with CD is approximately 70% 1,2 and fibrostenosing disease is the most common surgical indication. Surgery is not only costly for healthcare systems 3 but is associated with unemployment and poorer quality of life scores for patients. 4 Current medical therapies are unable to reverse established fibrosis and the prediction of stricture development remains a significant clinical challenge. Moreover, the rates of progression to stricturing disease have not altered despite the introduction and subsequent widespread use of biological therapy. 5 Intestinal mesenchymal cells play a pivotal role in deposition of extracellular matrix that characterises fibrosis in CD. 6,7 Manipulating the accumulation and function of intestinal mesenchymal cells is an attractive therapeutic option for fibrostenosing CD. Potential mechanisms of mesenchymal cell accumulation include expansion of the resident cell population, recruitment of bone marrow-derived circulating fibrocytes, and epithelial to mesenchymal transition (EMT). 8 EMT is a process in which epithelial cells lose cellular adhesion and acquire mesenchymal morphology, together with an increased migratory phenotype. Epithelial-specific cellular markers are lost whilst mesenchymal markers are gained. The loss of e-cadherin, a cell adhesion molecule responsible for adherens junction integrity, 9,10 appears to be a necessary step for EMT to occur. 11 EMT has been heavily implicated in both cancer progression (type 3 EMT) [12][13][14] and the development of tissue fibrosis (type 2 EMT). The latter has been demonstrated in renal, [15][16][17][18][19] hepatic, [20][21][22] pulmonary 23,24 and cardiac 25,26 models. Two recent studies of type 2 EMT have provided evidence for the concept of "partial EMT" 27,28 whereby cells lose epithelial characteristics but promote fibrosis by means other than the active deposition of extracellular matrix within the deeper tissue layers. These include the secretion of exosomes and cytokines to promote the differentiation of fibroblasts to myofibroblasts and the recruitment of bone-marrow derived mesenchymal cells. In one study these transitioning cells were found close to the epithelial layer from which they had originated. 27 This concept of "partial EMT" is thought occur predominantly in type 2, rather than type 3, EMT. 27,29 It is possible that this is the case in intestinal fibrosis, although comparatively few studies have been performed examining the role of EMT in this setting. Cell fate mapping in a murine colitis model demonstrated double-labelled intestinal epithelial cells for both epithelial-specific and mesenchymal-specific cellular markers. 30 In addition, high levels of nuclear β-catenin localisation and expression of the transcription factor SLUG, which can both be associated with EMT, were reported in human CD intestinal samples. 31 MiRNAs regulate EMT by negative post-transcriptional control of target gene expression. The miR-200 family impacts on fibrosis-related EMT in organs other than the intestine. [32][33][34][35] MiR-200b overexpression leading to suppression of TGFβ-induced EMT alterations has been demonstrated in an intestinal epithelial cell line. 36 In addition, downregulation of the miR-200 family and altered expression of the EMT effectors SNAIL and SLUG has been reported in colonic tissue from patients with CD. 37 In this study, we examined well-characterised intestinal CD specimens for the presence of EMT by quantifying cellular marker expression and related this to the histological degree of fibrosis. We also confirmed dysregulation of the miR-200 family and downstream effectors in fibrotic samples and demonstrate that the former is specific to the intestinal epithelium.  38 CD patients were either classified as SCD (with a stricturing phenotype) or NSCD (with either an inflammatory or penetrating phenotype).

| Immunohistochemistry
Archived small intestinal resection specimens from patients with stricturing CD were identified from the Barts Health NHS Trust histopathology database. Pathology reports were interrogated and blocks containing SCD and NSCD tissue for each patient chosen.

| Immunohistochemistry image analysis
Stained slides were scanned with the NanoZoomer 2.0 slidescanner (Hamamatsu, Japan). Tissue areas of interest for each cellular marker were outlined in ImageJ software in order to block out the rest of the image. Predetermined thresholds for pigment colour on the redgreen-blue spectrum, size and circularity were then applied to these areas of interest in order to capture both background cellular counterstaining and positive cellular antibody staining separately (Table S1). These thresholds were used consistently across all samples. Images generated by the ImageJ program were also checked visually to ensure staining included in the analysis was truly cellrelated and not simply as a result of cellular debris or tissue artefact.
Extra-epithelial staining of e-cadherin and cytokeratin-18 was measured using three randomly selected fields of lamina propria and submucosa at 20× magnification of each sample and measuring the proportion of positive cellular staining from total counterstaining.
Vimentin staining was measured in 2 randomly selected regions of 100 consecutive surface epithelial cells as well as 3 randomly selected crypts per slide, again at 20× magnification. Results were expressed as area of positive staining as a percentage of total background staining. Positively stained percentage area (rather than percentage of cells) was used owing to merging of cell nuclei upon cellular parameter inputting within the ImageJ program. Other easily identifiable cell types known to also express vimentin, such as pericryptal fibroblasts and neuronal cells were visually excluded from ImageJ analysis wherever possible. according to their severity and depth. If more than one ulcer was present, the score of the widest and deepest ulcer was taken forward. Other features that were recorded included the severity of the transmural chronic inflammatory infiltrate (if present) and the presence or absence of serositis.

| Laser capture micro-dissection
Pathology reports for archived intestinal resection specimens from patients with CD were interrogated and blocks containing SCD and NSCD tissue for each patient chosen. Two serial 5 μm sections of FFPE tissue were cut from each block in an RNAse-free manner onto nuclease-and human nucleic acid-free slides covered with polyethylene naphthalate membrane (Zeiss, Germany). Sections were very briefly counterstained before specific tissue compartments (epithelium, submucosa and muscle) were dissected and collected into collection tubes with adhesive caps (Zeiss, Germany) using the PALM Microbeam Laser Dissector (Zeiss, Germany). Samples were standardised by micro-dissecting and collecting the same surface areas for each compartment for each slide, to within a 1.5% error: 4 000 000 μm 2 epithelium, 6 000 000 μm 2 submucosa, 10 000 000 μm 2 muscle. Tissue from two serial sections from each block was combined and incubated overnight with proteinase K (10 μL) and Buffer KD (150 μL) prior to total RNA extraction using the miRNeasy FFPE Kit according to the manufacturer's protocol (Qiagen, Hilden, Germany).

| RNA extraction, reverse transcription & quantitative real-time PCR of gene expression
Total RNA, including and excluding small RNAs, was extracted using the miRNeasy and RNeasy kits (Qiagen), respectively. Reverse transcription (RT) and quantitative real time PCR (qPCR) were performed on RNA extracted from intestinal tissue using the High-Capacity-RNA to cDNA kit and Taqman assays (Applied Biosystems, Foster City, CA, USA), respectively. MiRNAs extracted from FFPE samples using laser capture micro-dissection (LCM) underwent RT and qPCR using the miScript system (Qiagen) and a pre-amplification step owing to low RNA concentrations. Fold changes were calculated using the 2 −δδCT method normalised to miR-26b as an endogenous control for miRNA assays and GAPDH for mRNA assays. The miScript system was used for miRNA extracted from serum without a pre-amplification step. The synthetic RNA spike-in, Ce_miR-39_1, was added to aid data normalisation. Analysis of melt curves was undertaken to ensure a single PCR product was obtained at each reaction. The normalised data were then log10

| Statistics
Graphpad Prism analysis software was used to perform Student's t tests and Pearson's correlation coefficients as denoted. A P-value <0.05 was considered statistically significant. Experiments were performed in triplicates except for the serum quantifications which were performed in duplicates.

| Patient demographics
The clinical characteristics of patients providing tissue from intestinal resections are shown in Table 1. All patients presented with either ileal (55.6%) or ileocolonic (44.4%) disease, and all had a stricturing phenotype. Table 2 summarises the clinical characteristics of CD patients from whom serum samples were used for miRNA expression profiling. Both the median age at sample collection and disease duration were higher in the SCD group compared to the NSCD group, although this difference was not statistically significant. This is perhaps intuitive, given that the development of fibrosis usually occurs later in the disease course in CD. CDAI scores were almost identical between the two groups (NSCD and SCD).
Similarly, there was no significant difference in the proportion of patients prescribed systemic 5-ASA, antibiotic, thiopurine or anti-TNFα treatment.

| Quantified expression patterns of cell-specific markers are consistent with EMT in fibrotic CD intestinal tissue
Immunohistochemistry (IHC) staining for the epithelial-specific cellular marker e-cadherin revealed a highly specific epithelial pattern in all sections of nine CD surgical sample pairs ( Figure S1). However, not a single cell positive for e-cadherin staining was localised outside of the epithelial architecture (0% for both NSCD and SCD sample groups). Table 3 and Figure 1A Table S3. Epithelial expression of the mesenchymal marker vimentin was significantly higher in the SCD group (17.39% vs 6.65%, P = 0.003). Representative IHC images are shown in Figure 2C-H.

| Histological tissue scoring allows objective assessment of fibrosis and inflammation and correlation with EMT marker expression
As expected, all blocks containing stricture had significantly higher total fibrosis scores than blocks containing no stricture (9.55 vs 5.11, P = 0.003; Table 4, Table S4). The mild to moderate scores for NSCD blocks confirmed the field change effect of histological fibrosis progression (Table S4). No dysplasia was found in any section. Total ulcer scores, used as correlates for inflammation, were Median age at resection (IQR) (y) 33.8 (20) Median disease duration at resection (IQR) (y) 10 not significantly different between the two groups (P = 0.7; Table 4).  For cytokeratin-18 the ImageJ program was used as described to count the total number of cells and number of positively stained cells in 3 randomly selected fields of ×20 magnification. The sum of the numbers of cells for each field was totalled and a percentage of positive staining cells calculated for each sample as shown below. For vimentin, the ImageJ program was used to calculate the total area (in pixels) assessed and the area positively stained for vimentin (in pixels) in three randomly selected crypts and two randomly selected regions of one hundred consecutive surface epithelial cells at ×20 magnification. The latter was expressed as a percentage of the former. Each data point is an average of percentage values across the three crypts and two surface epithelial areas for each slide. The mean percentages are shown for each group (NSCD vs SCD) and the standard error of the mean (SEM) with the P value as calculated using a Student's t test showing statistical significance.
compartment from each tissue section ( Figure S3). As predicted RNA yield was low, so standardisation of RNA input involved dissecting a consistent area per tissue compartment (Table S5 shows the areas harvested for each sample; Table S6 shows total area grouped by NSCD and SCD samples). The expression data were also normalised to miR-16 to facilitate comparison of relative expression levels.

| Epithelial down-regulation of miR-141 is associated with the histological degree of fibrosis in intestinal CD samples
The  (Table S7).

| Known EMT effectors and downstream targets of the miR-200 family are dysregulated in fibrotic CD intestinal tissue
The

| The miR-200 family is lowly expressed in sera of CD patients and profiles do not differ between disease phenotypes
The observation that the miR-200 family is down-regulated within the intestinal epithelium in SCD is potentially clinically relevant for the development of biomarkers of intestinal fibrosis in CD, in that the epithelium is readily sampled by endoscopic mucosal biopsy.
However, non-invasive testing is clearly more desirable given the potential risks with endoscopic procedures and the potential inaccessibility of small intestinal lesions. Therefore, given that serum miRNA profiles can be stable over time and assayed successfully, 31 were observed between the clinical groupings (one-way analysis of variance; Figure 4).

| DISCUSSION
Although published data suggest a contribution of EMT to the development of intestinal fibrosis in mouse models, 30  The absolute percentages of extra-epithelial cells positive for cytokeratin-18 were very low, perhaps because cytokeratin-18 is also lost relatively early in EMT. Given that tissue used in this study was taken from clinically relevant strictures, it is likely that fibrosis was at a late stage of development. It is also possible therefore that fully transitioned cells will have been missed, or that the contribution of EMT at this stage is less than at the initiatory stages of fibrosis.
In contrast, the epithelial vimentin staining percentage values were relatively high, potentially representing a population of cells primed for transitioning. It is possible that some of these positively stained cells represented intra-epithelial lymphocytes, although both NSCD and SCD groups would have been affected by this to a similar degree given that the histological inflammation scores were not significantly different between the two groups. The acquisition of mesenchymal markers, therefore, may occur earlier in EMT than purported, and certainly earlier than in animal models. 30   Nine sample sets of archived small intestinal specimens from patients with CD were graded by an independent gastrointestinal histopathologist who was blinded to clinicopathological details and given total fibrosis and total ulcer scores. The mean total fibrosis scores were significantly higher in the stricture sets than the non-stricture sets, with no significant difference in the total ulcer scores between the two groups.
thought to occur more frequently in fibrosis-related EMT than cancer-related EMT. 27,29 The significant alterations in EMT cellular marker expression between NSCD and SCD tissue occurred in sample groups between which objective inflammatory scores were not significantly different.
This contrasts with previous data showing reduced e-cadherin expression and increased vimentin expression in inflamed colonic mucosal biopsy samples from patients with IBD. 36 It is unclear whether the specimens in that study also had evidence of fibrosis, although this would seem less likely given that patients with UC were included and all samples were colonic, rather than ileal, in origin. In addition, the samples in the present study were chosen carefully to avoid including dysplasia or established cancer, given the well-characterised association of EMT with gastrointestinal cancer progression. Therefore, it can be assumed that the EMT marker alterations demonstrated here were not being driven by malignancy or inflammation.
The histological fibrosis scoring system used here was not validated on an external cohort, nor by a second pathologist. Despite these limitations, a significant correlation was observed between cytokeratin-18 expression and the degree of subserosal fibrosis. This would seem counterintuitive given that, at a tissue level, the epithelial origin of transitioning cells and the serosa are some distance apart. It is possible, therefore, that the contribution of EMT to intestinal fibrosis is minimal or that it occurs in ways other than the delivery of active matrix-producing cells. In models of renal fibrosis, injured epithelial cells have been found to undergo partial EMT but remain close to the tubular epithelium and secrete cytokines and exosomes that aid fibrocyte recruitment and macrophage stimulation. 27,28,42 It is possible that a similar paradigm exists in the gut, whereby epithelial cells undergoing partial EMT modulate fibrosis in the interstitium by paracrine, rather than migratory and matrix-producing, functions. The extra-epithelial cells identified by cytokeratin-18 staining could therefore represent "stranded" epithelial cells following end-stage fibrosis, "partial" EMT cells effecting fibrosis through paracrine channels or "migrant" cells invading deeper tissue layers for matrix deposition. Further work is required to clarify this.
Dysregulation of miRNA-200 family members was shown to occur not only between NSCD and SCD tissue, but within the epithelial layer specifically. This is a novel finding, and one which makes biological sense as the altered miRNA expression is juxtaposed to the origin of transitioning cells. Given the relatively small sample size, this finding should be explored in larger throughput studies. In addition, significant negative correlation was found between miR-141 expression and the subserosal fibrosis score.
Hence, alterations in molecular expression within the superficial mucosa may reflect cellular events occurring within deeper intestinal layers in CD, evidence for which has been presented before. 43   One way ANOVA testing revealed no significant differences between the clinical groups for any member of the miR-200 family nor the control U6. Mir-429 was excluded from this analysis as correlation between duplicate readings was poor. AU, arbitrary units; HC, healthy controls; NSCD, non-stricturing Crohn's disease; SCD, stricturing Crohn's disease Loss of e-cadherin and subsequent cellular adhesion is a hallmark of EMT 16,29,44 and this may occur under the control of the miR-200 family via the transcription factors ZEB1 and perhaps also ZEB2, SNAIL and TWIST; this loss may be an "EMT trigger" in this setting.
Further work is clearly required to demonstrate a functional link between the miR-200 family and control of type 2 EMT in the intestine.
If this mechanism was supported by further studies, potential therapeutic strategies may involve miRNA manipulation. For example, recent work has demonstrated that microvesicle-delivered miR-200b reduces colonic fibrosis in TNBS-induced experimental colitis. 45 Furthermore, preclinical trials that assess the therapeutic impact of targeting miRNAs have already begun. 46,47 Limitations to therapeutic use of miRNA modulation, include off-target effects [48][49][50] and drug delivery, which is a particular problem when dealing with the gastrointestinal tract.
Serum expression of the miR-200 family did not differ between healthy controls, NSCD or SCD, possibly due to an overlap of fibrotic burden or generally low miR-200 expression levels. The miR-200 family is likely to be functionally relevant in CD fibrosis, but does not appear to be amenable to non-invasive serum sampling for assessment of fibrotic burden or monitoring therapeutic intervention.
The focus of future research therefore should be on improving the understanding of the processes underpinning intestinal fibrosis and identifying potential therapeutic targets.
In conclusion, our results indicate the occurrence of EMT in fibrotic intestinal CD tissue, and that this occurs independently of the effect of inflammation. Additional work is now required to confirm these findings using a wide range of epithelial and mesenchymal

CONF LICTS OF INTEREST
The authors confirm that there are no conflicts of interest.