Identification of Epithelial to Mesenchymal Transition as a Novel Source of Fibroblasts in Intestinal Fibrosis*

Intestinal fibrosis is a major complication of Crohn disease (CD), but the precise mechanism by which it occurs is incompletely understood. As a result, specific therapies to halt or even reverse fibrosis have not been explored. Here, we evaluated the contribution of epithelial to mesenchymal transition (EMT) to intestinal fibrosis associated with a mouse model of CD and also human inflammatory bowel disease. Mice administered intrarectal 2,4,6-trinitrobenzene sulfonic acid (TNBS) develop inflammation and fibrosis that resembles CD both histologically and by immunologic profile. We utilized this model to molecularly probe the contribution of EMT to intestinal fibrosis. Additionally, we utilized double-transgenic VillinCre;R26Rosa-lox-STOP-lox-LacZ mice, in which removal of the STOP cassette by Cre recombinase in villin+ intestinal epithelial cells activates permanent LacZ expression, to lineage trace epithelial cells that might undergo EMT upon TNBS administration. TNBS-induced fibrosis is associated with the presence of a significant number of cells that express both epithelial and mesenchymal markers. In the lineage tagged transgenic mice, the appearance of LacZ+ cells that also express the fibroblast marker FSP1 unequivocally demonstrates EMT. Transforming growth factor (TGF)-β1, a known inducer of EMT in epithelial cells, induces EMT in rat intestinal epithelial cells in vitro, and bone morphogenic protein-7, an antagonist of TGF-β1, inhibits EMT and fibrosis both in vitro and in the TNBS-treated mice. Our study demonstrates that EMT contributes to intestinal fibrosis associated with the TNBS-induced model of Crohn colitis and that inhibition of TGF-β1 with recombinant human bone morphogenic protein-7 prevents this process and prevents fibrosis.

Crohn disease (CD) 5 is a disorder of chronic, transmural inflammation that can affect any part of the gastrointestinal tract from the mouth to the anus (1,2). Chronic inflammation combined with dysregulated wound healing is thought to result in a number of complications including fistula and fibrostenotic stricture formation (3,4). Over one-third of patients with CD undergo surgery to relieve obstruction related to fibrotic strictures, and many will require repeat interventions because of recurrent disease (3).
As with fibrosis in other organ systems, intestinal fibrosis is considered to follow a common pathway of fibrogenesis (4,5).
In what is presumably a protective adaptation, activated fibroblasts are recruited to sites of severe inflammation for the purpose of wound healing. Although fibroblasts are considered a heterogeneous population (6,7), those that express fibroblastspecific protein 1 (FSP1) or ␣-smooth muscle actin (␣-SMA) are the main types mediating fibrosis (8 -10). With persistent inflammation, fibrosis may result from excessive deposition of the extracellular matrix (4,5,11,12). A better understanding of why and how this happens will likely have a dramatic impact on the treatment of CD and its myriad complications.
Epithelial to mesenchymal transition (EMT) is a key contributor to the pool of activated fibroblasts in multiple organ systems including the heart, lung, liver, and kidney (13)(14)(15)(16)(17). EMT is a process in which epithelial cells lose their phenotypic and functional characteristics while acquiring mesenchymal features (13, 18 -20). The role of EMT in intestinal fibrosis has yet to be investigated.
Although organ fibrosis was once considered irreversible, recent data from animal models of renal, hepatic, and cardiac fibrosis demonstrate the anti-fibrotic effect of recombinant human bone morphogenic protein-7 (rhBMP-7) (14,17,21,22). BMP-7 is a member of the transforming growth factor-␤ (TGF-␤) superfamily and is one of 15 proteins with structural and functional homology (21,23,24). BMP-7 was first recognized for its regulatory effect on bone and cartilage formation and is currently employed clinically to accelerate fracture healing (25,26). With respect to organ fibrosis, the anti-fibrotic effect of rhBMP-7 stems from its ability to counteract the profibrotic action of TGF-␤1 (17,22), which has been implicated in the fibrosis of various organs including the gut (27). Receptors for BMP-7 are present on colonocytes (28), and systemically administered rhBMP-7 binds with high affinity to the inflamed colon and prevents inflammation in a rat model of CD induced by rectally administered 2,4,6-trinitrobenzene sulfonic acid (TNBS) (29). The aim of our study was to assess the contribution of EMT to intestinal fibrosis associated with both a mouse model of Crohn colitis and human inflammatory bowel disease as well as to explore the mechanisms underlying the anti-fibrotic properties of rhBMP-7 in the inflamed colon.

EXPERIMENTAL PROCEDURES
Mice-CD1 male mice were purchased from Charles River Laboratories (Cambridge, MA). VillinCre mice were a gift from Dr. Raju Kucherlapati (Harvard Medical School, Boston, MA). R26Rosa-lox-Stop-lox-LacZ (RStopLacZ) transgenic mice have been described previously (30). VillinCre;R26Rosalox-Stop-lox-LacZ (Villin-LacZ) double transgenic mice were generated by mating VillinCre ϩ with homozygous RStopLacZ mice. In these double transgenic mice, LacZ becomes irreversibly expressed in all cells that have at one time or currently express villin. In the adult mouse intestine, this expression is limited to the epithelium (31). All of the mice were on a C57Bl/6 genetic background. The mice were maintained in the animal research facility at Beth Israel Deaconess Medical Center under standard conditions and with ad libitum access to pelleted chow and tap water. The Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee reviewed and approved all of the animal studies.
Cell Culture-The rat small intestinal cell line IEC-6 was obtained from the ATCC (Manassas, VA). IEC-6 cells were grown in Dulbecco's modified Eagle's medium (DMEM; ATCC) supplemented with 10% heat-inactivated fetal bovine serum (FBS). Stock cultures of the cells were subcultured at appropriate intervals to maintain the cells at subconfluent densities.
Induction of EMT in Vitro-Rat enterocytes were cultured for 48 h until they reached 50% confluency in DMEM supplemented with 10% FBS. To induce EMT, the medium was FIGURE 1. Inflammation and fibrosis in a model of TNBS-induced Crohn colitis. Chronic inflammation and fibrosis were induced by weekly rectal administration of TNBS for 7 weeks. Control mice received a weekly enema of 45% ethanol alone (vehicle) for the duration of the study. Treatment with rhBMP-7 was initiated in a subset of mice 1 day prior to the first TNBS enema and continued every other day for the duration of the study. Paraffin-embedded colonic tissue sections were labeled with H&E or MTS. Immunofluorescent labeling with antibodies to FSP1 (red) and ␣-SMA (green) was done on frozen sections. The nuclei were stained with DAPI and appear blue. Representative photomicrographs of each treatment group are displayed. Ten high power fields were assessed in four mice/treatment group for quantification. A, colonic tissue from mice treated with ethanol alone served as a control. Labeling with H&E and MTS demonstrates normal architecture and no fibrosis, respectively. Immunofluorescent labeling with antibodies to FSP1 (red) and ␣-SMA (green) show that ϳ10 cells/each high power field are fibroblasts. Of these 10 cells, the majority are FSP-1-positive. B, colonic tissue from mice treated with TNBS demonstrate architectural distortion with submucosal thickening, epithelial destruction, and inflammatory infiltrate on H&E. The sections labeled with MTS display diffuse mucosal collagen deposition and fibrosis. The frozen sections labeled with antibodies to FSP1 (red) and ␣-SMA (green) show an approximately 2.5-fold increase in fibroblasts compared with control tissue. Fibroblasts expressing FSP-1 exhibited a more pronounced increase in response to TNBS than did the cells expressing ␣-SMA. C, administration of rhBMP-7 in conjunction with TNBS treatments substantially protects against epithelial injury and fibrosis as assessed by H&E and MTS, respectively. Furthermore, there is a statistically significant decrease in both subtypes of fibroblasts labeled on immunofluorescence. D, quantification of immunofluorescent labeling. ***, p Ͻ 0.001.
replaced with DMEM supplemented with 0.5% FBS containing TGF-␤1 (10 ng/ml) for 7 days (14). To evaluate the effect of BMP-7 on EMT in vitro, a subset of IEC-6 cells in culture were treated with rhBMP-7 (100 ng/ml) with or without medium containing TGF-␤1. The medium was changed every 2 days.
Immunocytochemistry-EMT was visualized in vitro by immunofluorescent double labeling of cultured cells as described in our previous publications (32). IEC-6 cells were cultured on fibronectin-coated chamber slides and fixed in icecold 100% acetone. The slides were incubated with primary antibodies to E-cadherin (BD Biosciences, Rockville, MD) and FSP1 (research gift from Dr, Eric G. Neilson, Vanderbilt University, Nashville, TN) at 4°C overnight. Following incubation with primary antibody, the cells were washed with Tris-buffered saline and incubated for 45 min at room temperature with Alexa Fluor 594-and Alexa Fluor 488-conjugated secondary antibodies (Invitrogen). The nuclei were counterstained using DAPI, and sections were covered with a glass coverslip using Vectashield mounting medium (Vector Laboratories, Burlingame, CA). Images were captured for analysis using a Zeiss Axioskop 2 fluorescent microscope attached to a Zeiss AxioCam HRC camera.
TNBS-induced Intestinal Colitis and Fibrosis-7-8-weekold mice were lightly anesthetized by an intraperitoneal injection of 100 mg/kg ketamine and 10 mg/kg xylazine. Approximately 3 cm of a polyethylene cannula (Intramedic PE-20 polyethylene tubing; BD Biosciences) attached to a 1-ml tuberculin syringe (BD Biosciences) was inserted into the colon, and 100 l of TNBS was infused. The mice received increasing doses of TNBS throughout the study. During the first 2 weeks, a dose of 1 mg/0.1 ml TNBS in 45% ethanol was used. The dose was increased to 1.5 mg/0.1 ml TNBS in 45% ethanol for the third and fourth weeks and again to 2 mg/0.1 ml TNBS in 45% ethanol for weeks 5-7. The control mice received either a 100-l enema of 45% ethanol alone (vehicle) or were untreated (33). Treatment was carried out every 7 days for a total of 7 weeks.
To evaluate the effect of BMP-7 on intestinal fibrosis in this model of CD, a subset of mice in the TNBS treatment group were administered concomitant rhBMP-7 (research gift from Curis). Based on our previous experience, a dose of 300 g/kg rhBMP-7 was delivered by intraperitoneal injection 24 h before the initial enema and then every other day until week 7 (22). At the conclusion of the treatment period, the mice were euthanized using CO 2 , and the colon was collected.
Tissue Preparation-The colon of each mouse was removed en bloc, cleared of feces, and washed in phosphate-buffered saline before being cut into smaller sections. The samples were either fixed in 10% formalin for future embedding in paraffin or directly embedded in Tissue Tek TM and snap frozen in liquid nitrogen for immunohistochemistry.
Histologic Assessment of Intestinal Fibrosis and Inflammation-Tissue samples fixed in formalin were delivered to the Histology Core at Beth Israel Deaconess Medical Center where they were embedded in paraffin, cut into 5-m sections, and stained with both Masson trichrome stain (MTS) and hematoxylin and eosin (H&E) under standard conditions.
Immunohistochemistry for TNF-␣-Immunohistochemistry of 10% formalin-fixed, paraffin-embedded intestinal sections was performed as described previously with minor modifications (34). Briefly, the sections were first deparaffinized and then microwaved in citrate buffer for antigen retrieval. After blocking, the sections were incubated with a primary antibody to TNF-␣ (rabbit polyclonal; Abcam) at 4°C overnight. After incubation, the sections were washed in Tris-buffered saline, blocked with 3% hydrogen peroxide, and then incubated with a biotinylated secondary antibody from the Vectastain Elite ABC kit (Vector Laboratories) at room temperature for 30 min. Positive staining was visualized with the 3,3Ј-diaminobenzidine peroxidase substrate kit also from Vector Laboratories, and the sections were counterstained with methylene blue after rehydration.
Immunofluorescent Labeling for EMT in Vivo-EMT was visualized by immunofluorescent double-labeling as described previously (35). Frozen tissue was cut into 5-m-thick cross-sections that were fixed in ice-cold 100% acetone for 5-10 min. The sections were incubated with the following primary antibodies at 4°C overnight depending on the experiment: mouse anti-E-cadherin (BD Bioscience), rabbit anti-FSP1 (polyclonal antibody; research gift from Eric Neilson, Vanderbilt University, Nashville, TN), fluorescein isothiocyanate-conjugated mouse anti-␣-SMA (Sigma), or rabbit anti-␤galactosidase (␤-gal; polyclonal antibody; Cappel, MP Biomedicals). After incubation, the sections were washed with Tris-buffered saline before exposure to secondary antibody. Primary antibodies raised in mice were detected using the M.O.M. kit (Vector Laboratories). For ␤-gal FSP1 double-labeling experiments, the primary antibodies were directly conjugated with Alexa Fluor 488 and 594 according to the manufacturer's recommendations (Invitrogen). The nuclei were counterstained with DAPI (Vectashield). The sections were covered with a glass coverslip using Vectashield mounting medium (Vector Laboratories) and analyzed using either the Zeiss Axioskop 2 fluorescent microscope attached to an AxioCam HRC camera or the Zeiss LSM 510 Meta scanning confocal microscope.
LacZ Staining-␤-gal activity was visualized by staining as described in our previous publications (14). Colon tissue (ϳ1 mm 3 ) was fixed in 4% paraformaldehyde at 4°C for 60 min and then washed three times in phosphate-buffered saline before being incubated at 37°C in 1 mg/ml ␤-gal (Sigma), 5 mM potassium ferrocyanide, 5 mM potassium ferricyanide, 2 mM MgCl 2 , 0.2% Nonidet P-40, and 0.1% sodium deoxycholate in phosphate-buffered saline for 72 h as previously described (34). The tissues were mounted in paraffin and counterstained with eosin.
Human Tissue-Samples of human colon and terminal ileum demonstrating active inflammatory bowel disease (either CD or ulcerative colitis) were obtained from the Ardais/Beth Israel Deaconess Medical Center Biomaterials and Information for Genomic Research Tissue Library (Boston, MA). All of the diseased samples were derived from surgical resection specimens and were cryopreserved. Histologically normal intestinal tissue FIGURE 3. Evidence for EMT associated with intestinal fibrosis in CD1 mice. Frozen colonic tissue sections from control CD1 mice that received treatment with either ethanol alone, TNBS alone, or TNBS plus rhBMP-7 were labeled with antibodies to E-cadherin (green) and FSP1 (red). The nuclei were stained with DAPI and appear blue. In the merged images, double positive cells appear yellow. Representative immunofluorescent images from each group are displayed. Five high power fields were assessed in three mice/treatment group for quantification. A, frozen colonic tissue samples from mice treated with ethanol alone exhibit few FSP1 ϩ fibroblasts with no detectable colocalization of E-cadherin and FSP1. B, fibrotic colonic tissue samples from mice treated with TNBS demonstrate increased FSP1 ϩ fibroblasts compared with control tissue as well as colocalization of E-cadherin and FSP1. This colocalization indicates EMT and appears yellow (arrows). C, treatment with TNBS plus rhBMP-7 results in a statistically significant decrease in both the total number of FSP1 ϩ fibroblasts and the degree of colocalization with E-cadherin. D, quantification of immunofluorescent labeling. **, p Ͻ 0.01.
at the margin of a surgical resection specimen for cancer or from an unaffected area in a patient with inflammatory bowel disease was obtained from the same tissue library and used as a normal control. Immunofluorescent labeling was performed as described above, and the images were analyzed using either the Zeiss Axioskop 2 fluorescent microscope or the Zeiss LSM 510 Meta scanning confocal microscope.
Western Blot Analysis-Western blots were performed using standard procedures. Total cell culture and tissue lysates were made using radioimmune precipitation assay buffer plus a pro-tease inhibitor mixture (Roche Applied Science). Antibodies to ␤-actin and E-cadherin were purchased from Sigma and BD Biosciences, respectively. A polyclonal antibody to TNF-␣ was purchased from Abcam (Cambridge, MA).
Statistical Analysis-The results are expressed as the means Ϯ S.E. Multiple comparisons were performed by one-way analysis of variance with a Bonferroni correction. A p value Յ 0.05 was considered statistically significant.

TNBS-induced Chronic Colitis Is Associated with Intestinal Fibrosis-
Paraffin-embedded colonic tissue samples from TNBS-treated and ethanol control mice were assessed for inflammation and fibrosis after labeling with H&E and MTS (Fig. 1). Treatment with TNBS but not the vehicle control alone (45% ethanol) was associated with colitis marked by considerable thickening of the submucosa, infiltration by immune cells, and destruction of the epithelial layers on H&E (Fig. 1, A and B). Treatment with TNBS but not the vehicle was similarly associated with diffuse collagen deposition in the mucosa and submucosa on labeling with MTS (Fig. 1, A and B).
Given the heterogeneous nature of activated fibroblasts (6, 7), we next sought to identify which populations were involved in this model of intestinal fibrosis. Fresh frozen sections from control and TNBStreated mice were labeled with antibodies to FSP1 and ␣-SMA. Frozen sections from mice with colitis and fibrosis on H&E and MTS, respectively, showed considerably more FSP1 ϩ and ␣-SMA ϩ fibroblasts as compared with similar sections from control mice (Fig. 1, A and B). Because FSP1 ϩ fibroblasts were more prominent at base line and increased in number more dramatically following treatment with TNBS ( Fig. 1), we chose to use this mesenchymal marker in our subsequent murine experiments.
Genetic Evidence Demonstrates EMT Is a Source for Intestinal Fibrosis-To provide evidence for EMT involving epithelial cells in intestinal fibrosis, we performed lineage tracing of intestinal epithelium. We generated double transgenic mice that express Cre recombinase under control of the villin promoter . Evidence for EMT associated with intestinal fibrosis in Villin-LacZ mice. Frozen colonic tissue sections from ethanol control and TNBS-treated Villin-LacZ mice were either labeled with antibodies to ␤-gal (red) and FSP1 (green) or with antibodies to E-cadherin (red) and FSP1 (green). The nuclei are stained with DAPI and appear blue. Representative immunofluorescent images of each group are displayed. In the merged panels ␤-gal ϩ FSP1 ϩ and E-cadherin ϩ FSP1 ϩ cells indicate EMT and appear yellow (arrows). A, colonic tissue samples from Villin-LacZ mice treated with ethanol alone exhibit few FSP1 ϩ fibroblasts with no colocalization of FSP1 and ␤-gal. B, fibrotic colonic tissue samples from mice treated with TNBS demonstrate increased FSP1 ϩ fibroblasts and colocalization of FSP1 and ␤-gal, indicating EMT. C, colonic tissue samples from Villin-LacZ mice treated with vehicle alone exhibit few FSP1 ϩ fibroblasts and no colocalization of FSP1 and E-cadherin. D, fibrotic colon tissue from TNBS-treated Villin-LacZ mice shows colocalization of E-cadherin and FSP1, suggesting EMT.
(VillinCre) and that also harbor an allele in which a STOP cassette flanked by two LoxP sites is inserted between the ubiquitous Rosa promoter and the LacZ gene (RStopLacZ) (Fig. 2A). Within the gut, the villin gene is expressed in the epithelium. In VillinCre mice, the Cre recombinase is expressed under the villin promoter and thus is present in intestinal epithelial cells (31). In the RStopLacZ reporter strain, the LacZ reporter gene is activated only after Cre-mediated excision of a LoxP stop cassette. Thus, in the Villin-LacZ double transgenic mice, the LacZ gene remains activated in intestinal epithelial cells and in all cells of intestinal epithelial origin regardless of any further phenotypic changes.
To demonstrate the lineage-specific expression of LacZ in our double transgenic mice, colonic tissue from control RStopLacZ and Villin-LacZ mice was labeled for ␤-gal. Whereas ϳ80% of the colonic epithelial cells from the double transgenic mice were ␤-gal ϩ (Fig. 2E), there was no detectable labeling in RStopLacZ control mice (Fig. 2B).
Colonic sections from TNBS-treated Villin-LacZ mice demonstrating fibrosis on MTS (Fig. 2D) were assessed for LacZ activity (Fig. 2F) and compared with the control samples labeled for the same enzyme (Fig. 2E). Within the area of max-imal fibroblast accumulation, there were several ␤-gal ϩ cells outside the epithelium in the TNBS-treated Villin-LacZ mice (Fig. 2F).
To provide evidence for EMT, we double-labeled fibrotic colon sections from CD1 mice treated with ethanol alone or TNBS with antibodies to the epithelial cell adhesion molecule E-cadherin and the fibroblast marker FSP1. In CD1 mice, treatment with TNBS but not vehicle alone was associated with the appearance of E-cadherin ϩ FSP1 ϩ cells (Fig. 3, A, B, and D).
To provide further evidence of EMT, we next performed double-labeling experiments to assess for colocalization of FSP1 with either ␤-gal (Fig. 4, A and B) or E-cadherin (Fig. 4, C and D) in the Villin-LacZ double transgenic mice. Although double-positive cells were absent in control colon tissues (Fig. 4, A and C), FSP1 ϩ cells of gastrointestinal epithelial origin (␤-gal ϩ or E-cadherin ϩ ) were detected within the colon of mice treated with TNBS (Fig. 4, B and D).

BMP-7 Inhibits TGF-␤1-induced EMT in Intestinal Epithelial Cells in
Vitro-EMT was induced in rat IEC-6 cells by exposure to TGF-␤1 for 7 days once they became subconfluent. IEC-6 cells in control medium displayed cuboidal morphology (Fig. 5A) and had robust E-cadherin expression by both immunofluorescent labeling and Western blot (Fig. 5, B and C). In contrast, the same cells became spindle-shaped (Fig. 5A), lost E-cadherin expression, and increased FSP1 expression upon incubation with TGF-␤1 (Fig. 5, B-D). The addition of rhBMP-7 to the medium containing TGF-␤1 completely prevented EMT (Fig. 6A).
Colocalization of FSP1 with an epithelial marker was next assessed in our Villin-LacZ double transgenic mice. In mice treated with either vehicle alone, TNBS, or TNBS plus rhBMP-7, colocalization of FSP1 with either ␤-gal or E-cadherin was measured (Fig. 6, B and D). In mice treated with TNBS, 31% of the total FSP1 ϩ population of cells was also ␤-gal ϩ (Fig. 6C). The percentage of FSP1 ϩ ␤-gal ϩ cells was significantly decreased (p Ͻ 0.01) upon rhBMP-7 treatment

. EMT involving intestinal epithelial cells in vitro.
Rat intestinal epithelial cells (IEC-6) in culture were exposed either to control medium (DMEM plus 10% FBS) or to DMEM plus 0.5% FBS with TGF-␤1 for 7 days. A, on phase contrast, IEC-6 cells display cuboidal morphology in control medium but acquire a spindle-shaped morphology on exposure to TGF-␤1. B, total protein lysates were analyzed by immunoblot using an antibody to E-cadherin. The blots were stripped and reprobed with a ␤-actin antibody as a control. A representative experiment is displayed. E-cadherin expression is considerably decreased upon exposure to TGF-␤1. C and D, the cells in culture were labeled with antibodies to E-cadherin (green) and FSP1 (red). The nuclei were labeled with DAPI and appear blue. The photomicrographs display representative immunofluorescence pictures of each group at an original magnification of 60ϫ. In control medium, IEC-6 cells are E-cadherin ϩ FSP1 Ϫ . Upon exposure to TGF-␤1, the IEC-6 cells in culture become E-cadherin Ϫ FSP1 ϩ , indicative of EMT. (Fig. 6C). Similar statistically significant results were obtained when FSP1 was colocalized with E-cadherin (Fig. 6E). In fibrotic colon tissues from Villin-LacZ mice, 24% of the FSP1 ϩ fibroblasts were also E-cadherin ϩ , and this percentage was significantly decreased (p Ͻ 0.001) if the mice received concomitant rhBMP-7 (Fig. 6E). Together, these data suggest that BMP-7 inhibits EMT in vivo, similar to its effect in the in vitro culture experiments.
Treatment with rhBMP-7 Reduces Colonic TNF-␣ Expression-To investigate the effect of rhBMP-7 on inflammation, paraffin-embedded colonic tissue sections were stained for TNF-␣ (Fig.  7A). Mice treated with ethanol alone (vehicle control) demonstrated low levels of TNF-␣ staining. The presence of TNF-␣ was significantly increased in the colonic tissue of mice treated with TNBS ( Fig. 7B; p Ͻ 0.01). Concomitant treatment with rhBMP-7 resulted in substantially less TNF-␣ staining in the colonic epithelium than those treated with TNBS alone (Fig. 7B; p Ͻ 0.01). The immunohistochemical results were confirmed by colonic tissue Western blot for TNF-␣ (Fig. 7C).

FIGURE 6. BMP-7 inhibits EMT of intestinal epithelial cells in vitro and in vivo.
A, IEC-6 cells were exposed to control medium alone (DMEM plus 10% FBS), DMEM plus 0.5% FBS with TGF-␤1, or medium containing rhBMP-7 with or without TGF-␤1 for 7 days. The cells were labeled with antibodies to E-cadherin (green) and FSP1 (red). The nuclei are labeled with DAPI and appear blue. The photomicrographs display representative immunofluorescence pictures of each group (original magnification, 60ϫ). Control cells are E-cadherin ϩ but FSP1 Ϫ . Upon incubation with TGF-␤1, cells become E-cadherin Ϫ FSP1 ϩ , indicating EMT. Incubation with rhBMP-7 alone does not affect E-cadherin or FSP1 expression. Coincubation with rhBMP-7 and TGF-␤1 protects the IEC-6 cells from undergoing EMT, as can be seen in the far-right panel. B, colon tissue sections from ethanol-treated Villin-LacZ mice, fibrotic TNBS-treated Villin-LacZ mice, and Villin-LacZ mice that received rhBMP-7 in addition to TNBS were labeled with antibodies to ␤-gal (red) and FSP1 (green). The nuclei stained with DAPI appear blue. ␤-gal ϩ FSP1 ϩ cells, which are indicative of EMT, appear yellow (arrows in the middle panel). In tissues from TNBS-treated mice that received rhBMP-7, ␤-gal ϩ FSP1 ϩ cells were only sporadically present. C, the bar graph summarizes both the total number of FSP1 ϩ fibroblasts and the number of FSP1 ϩ ␤-gal ϩ cells in each treatment group. Treatment with rhBMP-7 results in a statistically significant decrease in both the total number of FSP1 ϩ fibroblasts and the amount of colocalization with ␤-gal. **, p Ͻ 0.01. Ten high power fields were assessed in three mice/group for quantification. D, colon tissue sections from ethanol-treated Villin-LacZ mice, fibrotic TNBS-treated Villin-LacZ mice, and Villin-LacZ mice that received rhBMP-7 in addition to TNBS were labeled with antibodies to E-cadherin (red) and FSP1 (green). The nuclei stained with DAPI appear blue. E-cadherin ϩ FSP1 ϩ cells, which are indicative of EMT, appear yellow (arrows in the middle panel). In tissues from TNBS-treated mice that received rhBMP-7, E-cadherin ϩ FSP1 ϩ cells were only sporadically present. E, the bar graph summarizes both the number of FSP1 ϩ fibroblasts and the amount of EMT as evidenced by colocalization of FSP1 and E-cadherin in each group. Treatment with rhBMP-7 results in a statistically significant decrease in both the total number of FSP1 ϩ fibroblasts (**, p Ͻ 0.01) and the amount of colocalization with E-cadherin (***, p Ͻ 0.001). Ten high power fields were assessed in three mice/group for quantification.

Demonstration of Epithelial to Mesenchymal Transition in
Human Inflammatory Bowel Disease-Colocalization of ␣-SMA and E-cadherin was demonstrated in human samples of active inflammatory bowel disease but not in normal-appearing intestinal tissue (Fig. 8, A and B). There was a statistically significant increase in the percentage of crypts harboring doublepositive cells in the samples from patients with inflammatory bowel disease compared with controls (Fig. 7C). These results favor the notion that EMT may also be a source for activated fibroblasts associated with fibrosis in inflammatory bowel disease in general. Further studies are required to assess whether this finding is unique to CD, but the data likely validate the earlier observation that EMT may play a role in human fistulizing CD (36).

DISCUSSION
Intestinal fibrosis in the form of fibrostenotic strictures is a well described complication of longstanding CD and is thought to occur as a result of chronic inflammation and dysregulated wound healing (2,3). The mechanism by which inflammation leads to fibrosis is only beginning to be understood but likely involves EMT to some degree (13). Although various anti-in-flammatory and immunosuppressive agents have demonstrated efficacy for active CD, they have little impact on fibrosis once it occurs (3).
EMT is a complex process in which epithelial cells lose their phenotypic and functional characteristics and develop features of the mesenchyme. Specifically, the epithelial cell loses polarity, intercellular adhesion molecules, and membrane-bound transporters to develop a specialized cytoskeleton and invasive capacity (19,37).
To provide evidence for EMT involving epithelial cells in intestinal fibrosis, we generated double transgenic mice in which intestinal epithelial cells can be lineage-traced by expression of LacZ. By inducing colitis in the Villin-LacZ mice, we were able to demonstrate for the first time that EMT contributes to the pool of activated fibroblasts that accumulate in a murine model of Crohn colitis generated by treatment with TNBS.
Activated fibroblasts are the main mediators of fibrogenesis in the colon as in other organs (38,39). Fibroblasts are a heterogeneous cell population (6) and the major populations that are known to mediate fibrosis express FSP1 or ␣-SMA (8). FSP1 ϩ fibroblasts have been well characterized during the fibrogenesis of kidney, liver, and heart (14,16,17) and often derive from resident epithelia via EMT (18). In our model of colitis, the numbers of both FSP1 ϩ and ␣-SMA ϩ fibroblasts increase upon disease induction. The effect of treatment with TNBS is more dramatic on the population of fibroblasts expressing FSP1. Thus, we used this marker for our subsequent experiments.
Our experiments suggest that approximately one-third of FSP1 ϩ fibroblasts are derived from intestinal epithelial cells. This result is consistent with our studies of EMT in mouse models of renal (ϳ35%) (40) and liver (ϳ45%) (14) fibrosis but may actually be an underestimate because the LacZ transgene is only active in 80% of the cells. Although E-cadherin FSP1 double labeling is insufficient to detect epithelial cells that have fully acquired a fibroblast phenotype, and lost expression of epithelial cell markers, these data further support the idea that EMT occurs in the model of TNBS-induced intestinal fibrosis.
To explore the anti-fibrotic effects of rhBMP-7 in this model of CD, a subset of mice receiving TNBS enemas were concomitantly treated with rhBMP-7. As we and others have previously shown, BMP-7 exerts its anti-fibrotic action in the setting of chronic disease of the kidney, liver, and heart at least partially via inhibition of EMT (14,17,22). Therefore we speculated that enterocytes could contribute to the pool of FSP1 ϩ fibroblasts in the setting of intestinal fibrosis and that inhibition of EMT by rhBMP-7 would inhibit fibrosis. Administration of rhBMP-7 in conjunction with TNBS prevented the epithelial injury and fibrosis seen in the mice treated with TNBS.
Previous studies have demonstrated that the TNBS-induced model of Crohn colitis is associated with substantially increased BMP receptor expression (29) This finding suggests that administration of rhBMP-7 exerts its anti-fibrotic action via utilization of a pre-existing repair pathway in which colonocytes play a pivotal role (29). Our studies do not exclude the possibility that other BMPs are equally effective to prevent FIGURE 8. Evidence for EMT in human inflammatory bowel disease. E-cadherin (red) and ␣-SMA (green) colocalization (yellow) was assessed in three intestinal sections from patients with active inflammatory bowel disease (IBD, either CD of the terminal ileum or ulcerative colitis) (A) and in four frozen sections of normal-appearing human intestine (B). The number of crypts harboring a double positive cell was assessed for each section and expressed as a percentage of the total number of crypts/section. The representative images were acquired using a Zeiss Axioskop 2 fluorescent microscope (original magnification, 60ϫ). The arrow indicates colocalization, which is present in the inflammatory bowel disease sample but not in normal tissue. The bar graph (C) summarizes the percentage of crypts harboring a double-positive cell in diseased versus control tissue. **, p Ͻ 0.01.
Anti-TNF-␣ therapy is associated with a loss of inhibitory effect of TNF-␣ on TGF-␤1 signaling (43). In this regard, our studies demonstrate that TGF-␤1 is a potent inducer of EMT and that BMP-7 significantly inhibits TGF-␤1-induced EMT even while inhibiting TNF-␣. Additional studies are needed to explore possible synergistic effects of the combination of anti-TNF-␣ therapy and rhBMP-7 in the effective suppression of inflammation and fibrosis in CD. Although controversial, there is some evidence that treatment with infliximab, an anti-TNF-␣ monoclonal antibody, may enhance stricture formation in some patients (44,45). In these patients, dual treatment may be safer and more efficacious.
TGF-␤1 is regarded as a principal mediator of intestinal fibrosis (4), and it is the most potent inducer of EMT in various organ systems (13). Therefore, we tested the capacity of TGF-␤1 to induce EMT in adult rat intestinal epithelial cells. The findings in vivo were confirmed in the IEC-6 cell line.
Although repeated treatment with TNBS leads to a T H 1 T cell inflammatory process that resembles CD both in histopathology and immunology (46 -48), the role of EMT in human inflammatory bowel disease remains virtually unexplored. A recent paper demonstrated a potential role for EMT in fistulas from a small cohort of patients with CD (36). Our finding of colocalized E-cadherin and ␣-SMA in intestinal resection specimens from patients with active inflammatory bowel disease suggests that EMT likely occurs in human intestinal fibrosis as well. The pooling of samples from patients with CD and ulcerative colitis clearly limits the scope of our interpretation of these human data. Nevertheless, our data support a role for EMT in active human inflammatory bowel disease. Whether the findings are unique to CD or merely a result of active inflammation associated with any number of colitides remains to be established. Analysis of additional samples collected in a prospective and nonpooled manner will be necessary to further characterize this process as it applies to human intestinal fibrosis in general and CD in particular.
Our study demonstrates a clear role for EMT in the development of TNBS-induced intestinal fibrosis. As has been demonstrated in the kidney and heart, treatment with rhBMP-7 both in vivo and in vitro prevents EMT associated with interstitial fibrosis. This finding suggests that rhBMP-7 may be an important future therapy for intestinal fibrosis in general and CD in particular.