INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

CROHN'S DISEASE (CD) and ulcerative colitis (UC) are immunologically mediated inflammatory diseases of the gastrointestinal tract that are characterized by chronic inflammation, mucosal damage, and epithelial cell destruction (26, 31). CD differs from UC, in that it may involve other regions of the gastrointestinal tract in addition to the colon and is characterized by transmural, granulomatous inflammation and fibrosis (7, 26, 31). Fibrosis frequently leads to stricture and bowel obstruction, and these are major causes of surgery and bowel resection in CD (7). Fibrosis in this disorder can be variable in presentation but is typically associated with mesenchymal cell hyperplasia, tissue disorganization, and fibrillar collagen deposition that can occur in lamina propria, muscularis mucosa, submucosa, and muscularis propria (7, 26, 31). Fibrosis in CD is thought to develop as a result of aberrant tissue repair processes, yet the cellular and molecular mechanisms that underlie fibrosis are not well defined.

A number of lines of evidence implicate insulin-like growth factor I (IGF-I) in the fibrogenic process in CD. IGF-I expression is increased in areas of fibrosis in a number of tissues and disease states, including bleomycin-induced pulmonary fibrosis (14), nephrosclerosis (12), and hypertrophic scarring of the skin (5). IGF-I stimulates proliferation of fibroblasts (19, 29), myofibroblasts (24, 29), and smooth muscle cells (13) that each are implicated as cellular mediators of fibrosis in CD (21, 22). IGF-I also stimulates collagen synthesis in intestinal fibroblasts and myofibroblasts (24) and intestinal smooth muscle cells (37). In animal models, local IGF-I expression is increased in the colon during experimental enterocolitis induced by sodium dextran sulfate (28), ethanol trinitrobenzene sulfonic acid (TNBS) (36), or peptidoglycan-polysaccharides (PG-PS) (38, 39). In the PG-PS and TNBS models, the elevated IGF-I expression occurs at sites of increased collagen deposition and fibrosis (36-38). Recent findings suggest that these observations in animal models are relevant to CD. IGF-I immunoreactivity is elevated in lavage fluid obtained from patients with CD but not UC (6). We reported preliminary data that IGF-I mRNA is upregulated in regions of active disease in the intestine of patients with CD (3, 15, 23). One aim of the present study was to compare IGF-I mRNA abundance in involved and uninvolved ileum and colon from patients with CD and involved and uninvolved colon of patients with UC to establish definitively whether local IGF-I overexpression is a particular characteristic of active CD. In addition, we sought to establish whether the cellular sites of IGF-I overexpression colocalize with sites of increased collagen gene expression, collagen deposition, and fibrosis. We therefore examined surgical specimens of ileum or colon obtained from patients with CD or UC for IGF-I and collagen mRNA localization using in situ hybridization histochemistry and histological stains for collagen.

The primary sites of IGF-I expression in the normal or diseased intestine are mesenchymal cells (17-19, 36-38), but the precise mesenchymal cell subtype is not well defined. Mesenchymal cell subtypes can be broadly classified into fibroblasts, smooth muscle cells, or myofibroblasts on the basis of immunostaining properties with antibodies to vimentin (V) and -smooth muscle (SM) actin (A) (21, 22, 27). Typically, fibroblasts are V⁺/A, smooth muscle cells are V/A⁺, and myofibroblasts are V⁺/A⁺ (21, 22, 27). Although SM-actin and vimentin are useful as phenotypic markers, it is increasingly evident that mesenchymal cells in the intestine and other systems may represent a more heterogeneous population than previously suspected (21, 22). Not all myofibroblasts stain positively for SM-actin (21, 22). Desmin, an intermediate-filament protein, is typically found in phenotypically normal smooth muscle but may represent a marker of myofibroblasts in some tissues or disease states (21, 22). Interstitial cells of Cajal (ICC) represent a myofibroblast-related mesenchymal subtype specific to the intestine. ICC are typically located between enteric smooth muscle layers and regulate motility (21, 22). The c-kit receptor, which binds the protooncogene stem cell, or steel factor, is a phenotypic marker of ICC, but immunostaining characteristics with other mesenchymal cell antigens are not well defined, nor is the role of ICC in fibrosis associated with CD (21, 22). In the present study we compared immunostaining patterns for IGF-I precursor, SM-actin, vimentin, desmin, and c-kit in uninvolved and involved ileum and colon of patients with CD and in ileum and colon of patients with no history of inflammatory bowel disease (IBD). Our primary aim was to identify the mesenchymal cell subtypes that show increased IGF-I or collagen expression in involved ileum or colon of patients with CD. We aimed also to assess whether there were qualitative or obvious quantitative differences in mesenchymal cell subtypes in diseased/fibrotic intestine of patients with CD relative to uninvolved or normal intestine. The results demonstrate that expression of IGF-I mRNA and encoded precursor is increased at sites of fibrosis in active CD. Furthermore, the sites of increased IGF-I precursor expression and collagen deposition in ileum and colon are populated by cells that have fibroblast or myofibroblast phenotype and, to a lesser extent, by cells with the phenotype of normal enteric smooth muscle. Compared with normal or uninvolved bowel, active CD and fibrosis are associated with increases in the number of V⁺ cells in muscularis mucosa, submucosa, and muscularis propria, indicating an involvement of fibroblasts and myofibroblasts in the pathophysiology of fibrosis.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Tissue collection. Excess surgical tissue from resected ileum or colon obtained from patients undergoing medically indicated surgery for complications of CD and UC was used in all analyses. This tissue was not required for pathology and would otherwise have been discarded. Use of this tissue is exempt from required approval by the Institutional Review Board for studies on human subjects, but the Institutional Review Board was informed of use of the tissue for these studies. Diagnoses of CD and UC were based on clinical, radiological, endoscopic, and histological criteria (26, 31). Samples from each CD or UC patient were separated into grossly involved, actively diseased tissue and grossly uninvolved tissue at the margins of diseased intestine. Small samples of uninvolved and involved tissue were fixed in 4% paraformaldehyde to allow histological evaluation of disease activity and immunohistochemistry. Additional, adjacent samples of each region were embedded in capsules of optimum cutting temperature compound (Miles, Elkhard, IN) and stored at 80°C for in situ hybridization histochemistry. The remainder of each sample was snap frozen in liquid nitrogen for RNA extraction. Samples were processed further for in situ hybridization histochemistry, immunohistochemistry, or RNA extraction only if histology confirmed the initial categorization as grossly involved or uninvolved. To verify appropriate sample categorization, coded sections of uninvolved and involved ileum and colon, stained with hematoxylin and eosin and Masson's trichrome, were scored for histological abnormalities. Surface epithelial damage, lamina propria inflammation, thickness of muscularis propria, and fibrosis were assigned a score of 0-2, where 0 represents normal, 1 represents mild abnormality, and 2 represents severe abnormality. Total histology score was the sum of these values for each bowel layer. Thickness of muscularis propria was also measured in well-oriented sections as an additional measure of disease. By comparing uninvolved and involved samples from the same patient, we aimed to minimize the impact of interpatient variability on our measured parameters.

RNA extraction and analyses. Total RNA was extracted from snap-frozen samples by homogenization in guanidine isothiocyanate and centrifugation over 5.7 M cesium chloride, as previously described (37). The abundance of IGF-I mRNA was measured by RNase protection assay (RPA; RPA II kit, Ambion, Austin, TX) as specified by the manufacturer's protocol. Briefly, 40 µg of total RNA were hybridized with ³²P-labeled antisense RNA probes complementary to human IGF-I mRNA (11, 15) and human glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA (Ambion). After hybridization and RNase treatment, reactions were denatured and then electrophoresed on 5% polyacrylamide-urea gels. Gels were dried, and the intensity of protected bands was visualized and quantified by phosphoimager analysis (Image Quant, Molecular Dynamics, Sunnyvale, CA). IGF-I mRNA abundance in each sample was normalized to the abundance of constitutively expressed GAPDH mRNA to control for RNA loading. The normalized values for IGF-I mRNA abundance were used for comparisons and statistical analyses.

mRNA localization. For in situ hybridization, tissue fragments were cryosectioned at 10 µm and processed as previously described (39). Briefly, sections were fixed in 4% paraformaldehyde, treated with proteinase K (0.5 µg/ml) for 10 min, and then acetylated with triethanolamine (0.1 M) and 0.25% (vol/vol) acetic anhydride. Dehydrated and air-dried slides were incubated with 50 µl of hybridization buffer containing 75% formamide and 1 × 10⁶ counts/min of ³⁵S-labeled antisense or sense RNA probes derived from human IGF-I (9, 11) or procollagen 1(I) (20, 30) cDNAs. Sections were hybridized for 18 h at 55°C, treated with RNase, and extensively washed in low-salt buffers (39). Slides were exposed to Ilford K.5F radiographic emulsion (Polysciences, Warrington, PA) at 4°C for 3-14 days. Developed slides were counterstained with hematoxylin and photographed under dark- and bright-field optics. Positive hybridization was defined as clusters of silver grains observed over cells at higher densities than in sections hybridized with sense probes. Adjacent sections were counterstained with Sirius red (1) to localize collagen and with hematoxylin and eosin for histology.

Immunohistochemistry. Immunohistochemistry was performed on fixed, paraffin-embedded samples sectioned at 6 µm. Serial sections were incubated with antibodies specific for pro-IGF-I, SM-actin, desmin, vimentin, or c-kit. A rabbit polyclonal antiserum specific for a carboxy-terminal precursor peptide, or E-domain of pro-IGF-I (anti-human IGF-I Ea), was used at a dilution of 1:500. The pro-IGF-I antibody localizes an intermediate in IGF-I biosynthesis, providing a useful tool to localize sites of increased IGF-I expression (18, 19, 39). Available data indicate that at least a portion of newly synthesized IGF-I is secreted as an E-domain extended form, and so IGF-I precursor may be intracellular or secreted (18, 19, 39). Prior studies have established that the pro-IGF-I antibody yields immunostaining superior to that obtained with antibodies to mature IGF-I, probably because IGF-I is rapidly secreted or associates with tissue IGF binding proteins (IGFBPs) that mask the epitopes recognized by available IGF-I antibodies (18, 19, 39). Mouse monoclonal antibodies to human SM-actin (clone A4), human vimentin (clone V9), and human desmin (clone D33) were purchased from DAKO (Carpinteria, CA) and used at a dilution of 1:200. A rabbit polyclonal antibody to c-kit (SC 168) was purchased from Santa Cruz Biotechnology (Santa Cruz, CA) and used at a dilution of 1:200. Binding of the primary antibodies was detected by the avidin-biotinylated peroxidase method (VectaStain kit, Vector Laboratories, Burlingame, CA). Negative controls, which consisted of omission of primary antiserum, were uniformly negative. Sections adjacent to those used for immunohistochemistry were counterstained with Masson's trichrome to reveal tissue histology and to localize collagen.

Statistical analyses. Mean thickness of muscularis propria was compared in involved and uninvolved intestine of patients with CD and UC using Student's t-test. IGF-I mRNA abundance in each sample of involved intestine was expressed as a ratio of the abundance in uninvolved tissue from the same patient. These ratios were compared for a significant difference from 1 using the Mann-Whitney U test. To directly compare IGF-I mRNA abundance in colon of patients with CD and UC, values in uninvolved or involved colon of UC patients were expressed as a ratio of the mean value for uninvolved colon from CD patients assayed on the same gels and compared for a difference from 1 using the mean Mann-Whitney U test. P < 0.05 was considered statistically significant.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Histological verification of tissue categorization. Histological scoring verified that samples of CD and UC intestine that were categorized as involved showed significantly greater evidence of disease than those categorized as uninvolved. In ileum and colon from CD patients, histological scores for the uninvolved samples were 0-2, and scores >0 primarily reflected inflammation within the lamina propria. In involved ileum and colon samples from CD patients, scores were 7-8, indicating significant transmural disease. In addition, the muscularis propria was significantly thicker (P < 0.0001) in involved ileum and colon from patients with CD [4.8 ± 0.39 (SE) mm, range 3-8 mm] than in uninvolved samples (2.4 ± 0.15 mm, range 2-3 mm) from the same patients. For UC patients, histological scores for uninvolved colon were 0-2, and for involved colon the scores were 3-4. In UC samples, mean thickness of muscularis propria did not differ in uninvolved and involved colon (2.8 ± 0.22 and 2.7 ± 0.18 mm, respectively).

IGF-I mRNA expression is upregulated in active CD but not in active UC. RPA detected IGF-I mRNA in all specimens of ileum or colon from CD patients and all colon samples from UC patients. Representative autoradiograms are shown in Fig. 1A. For each patient, IGF-I mRNA abundance in involved intestine was expressed as a ratio of the abundance in adjacent, uninvolved samples of the same region (i.e., ileum or colon). The mean ratio of IGF-I mRNA abundance in involved to that in uninvolved samples from patients with CD was 2.73 ± 0.7 and was significantly (P < 0.05) greater than 1.0. In 10 of 13 CD patients studied, IGF-I mRNA abundance was higher in involved segments of the ileum or colon than in the uninvolved sample from the same patient (Fig. 1B). The 10 involved samples with elevated IGF-I mRNA included 7 of the ileum samples and all 3 of the colon samples. Analyses of CD ileum and colon on the same gels revealed no major differences in IGF-I mRNA abundance in the two segments (data not shown). The mean ratio of IGF-I mRNA abundance in involved to that in uninvolved colon of patients with UC was 1.13 ± 0.11 and did not differ significantly (P = 0.43) from 1. IGF-I mRNA abundance in samples from individual patients with UC is shown in Fig. 1B. Relative to IGF-I mRNA abundance in uninvolved colon from CD patients assayed on the same gels, IGF-I mRNA abundance was 1.05 ± 0.29 for uninvolved UC colon and 1.13 ± 0.35 for involved UC colon. IGF-I mRNA abundance did not differ significantly in involved (P = 0.56) or uninvolved (P = 1.0) colon of UC patients compared with uninvolved colon of patients with CD.

View larger version (63K): [in this window] [in a new window]   Fig. 1.   Insulin-like growth factor I (IGF-I) mRNA abundance in uninvolved and involved intestine of patients with Crohn's disease (CD) and ulcerative colitis (UC). IGF-I and control glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNAs were quantified by RNase protection assay on 40 µg of total RNA extracted from uninvolved and involved samples of intestine from patients with UC and CD. A: autoradiogram of a representative RNase protection assay. Probes, undigested probes; probes + RNase, probes alone incubated with RNase to show complete digestion; other lanes show representative samples of RNA from uninvolved (U) and involved (I) colon from patients with UC or CD. Arrows at right indicate protected fragments corresponding to IGF-I (578 nt) and GAPDH (295 nt) mRNAs. These were evident in each sample of RNA from UC or CD colon shown and all other samples tested. B: IGF-I mRNA abundance in individual uninvolved or involved samples from 7 patients with UC and 13 patients with CD. Values are densitometric units for IGF-I mRNA, normalized to values for the GAPDH mRNA internal control.

Colocalization of IGF-I and procollagen 1(I) mRNA expression at sites of fibrosis in active CD. In situ hybridization histochemistry was used to assess the cellular sites of IGF-I mRNA expression in sections of involved and uninvolved ileum of nine patients with CD and uninvolved and involved colon of three patients with CD. Ileum samples from the 13th CD patient were not analyzed because of poor morphology. In situ hybridization was also performed on samples of uninvolved and involved colon from four patients with UC. Data from the UC samples are not presented, inasmuch as hybridization signals for IGF-I mRNA were low and not convincingly above background in most samples. When detected in UC samples, IGF-I mRNA was observed in scattered cells within the lamina propria (data not shown). In the samples from the 12 CD patients studied, hybridization signals for IGF-I were barely above background in uninvolved ileum or colon. When detected in uninvolved bowel, IGF-I mRNA was localized to submucosa, muscularis propria, or serosa (Table 2). Consistent with RPA data, strong hybridization signals for IGF-I mRNA were observed in sections of involved ileum and colon of the majority of patients with CD. The sites of IGF-I mRNA expression did not differ in involved ileum and colon. Weak but detectable hybridization signals were observed in the lamina propria and muscularis mucosa of the majority of patients (Table 2). The strongest hybridization signals were observed in the thickened, disorganized muscularis propria and submucosa of all patients studied (Table 2). Photomicrographs of regions of uninvolved and involved submucosa (Fig. 2) and muscularis propria (Figs. 3 and 4) of selected CD patients illustrate typical hybridization signals for IGF-I. In uninvolved intestine, hybridization signals for IGF-I mRNA were barely above background levels. In involved submucosa, hybridization signals for IGF-I mRNA were typically detected in mesenchymal cells surrounding aggregates of lymphoid cells or granulomas (Fig. 2). In involved muscularis propria, IGF-I mRNA was localized to areas of tissue disorganization, particularly to cells within or surrounding hypertrophied/hyperplastic smooth muscle cell bundles, rather than to cells with typical smooth muscle morphology (Figs. 3 and 4). Photomicrographs of adjacent sections stained with Sirius red or hybridized with the procollagen 1(I) probe are shown in Figs. 2-4 to permit comparisons between sites of IGF-I mRNA and procollagen 1(I) mRNA expression and collagen deposition. Photomicrographs of procollagen 1(I) mRNA shown in Figs. 2-4 represent short exposure times that reveal only sites of high-level expression to illustrate overlap with sites of IGF-I mRNA expression in involved intestine. In submucosa and muscularis propria of involved CD ileum or colon, sites of procollagen 1(I) mRNA upregulation overlapped with sites of collagen deposition visualized by Sirius red staining and sites of IGF-I expression (Figs. 2-4 ). In submucosal granulomas found in involved ileum of patients with CD, high-level procollagen 1(I) mRNA expression was observed in connective tissue surrounding the granulomas (Fig. 2). In the involved muscularis propria, high-level procollagen 1(I) mRNA was observed in regions of fibrosis (Figs. 3 and 4), and this pattern was typical of all samples of involved CD ileum or colon. In uninvolved CD ileum or colon, procollagen 1(I) mRNA expression was not evident at the short exposure times illustrated in Figs. 2-4, but longer exposure revealed low-level procollagen 1(I) mRNA expression in submucosa and serosa, as would be expected for these connective tissue layers (data not shown). Specificity of hybridization signals detected with antisense IGF-I and procollagen 1(I) probes was verified by the absence of hybridization signal when the same sections were hybridized with sense probes (Fig. 4).

View larger version (63K): [in this window] [in a new window]   Fig. 2.   Localization of IGF-I and procollagen 1(I) mRNAs in submucosa of uninvolved and involved ileum from a patient with CD. Bright-field (BF and SR) and dark-field (IGF-I and Col I) photomicrographs illustrate in situ hybridization data. IGF-I and Col I, dark-field image of sections of uninvolved or involved ileal submucosa from a CD patient hydbridized with IGF-I or procollagen 1(I) antisense RNA probes, respectively. Exposure times were 14 days for sections hybridized with the IGF-I probe and 3 days for sections hybridized with the procollagen 1(I) probe. BF, hematoxylin-stained section after hybridization with the IGF-I antisense probe to show histology and submucosal immune cell aggregates in involved ileum; SR, Sirius red-stained section adjacent to that hybridized with procollagen 1(I) probe to indicate localization of collagen protein (appears darker) surrounding the submucosal lymphoid aggregates in involved submucosa. Note higher-level expression of IGF-I mRNA in the involved ileum adjacent to the lymphoid aggregate and high-level expression of procollagen 1(I) mRNA and collagen deposition in the same region.

View larger version (58K): [in this window] [in a new window]   Fig. 3.   Localization of IGF-I and procollagen 1(I) mRNAs in muscularis propria of uninvolved and involved ileum of a patient with CD. Bright-field (BF and SR) and dark-field (IGF-I and Col I) photomicrographs illustrate in situ hybridization data for uninvolved and involved ileal muscularis propria from a CD patient. IGF-I and Col I, dark-field images of sections hybridized with IGF-I or procollagen 1(I) antisense RNA probes, respectively. Exposure times were 14 days for sections hybridized with the IGF-I probe and 3 days for sections hybridized with the procollagen 1(I) probe. BF, hematoxylin-stained section after hybridization with the IGF-I antisense probe to show histology; SR, Sirius red-stained section adjacent to that hybridized with the procollagen (I) probe to indicate localization of collagen protein (appears darker). Note increased IGF-I and procollagen 1(I) mRNA expression in the intermuscular connective tissue of involved muscularis propria, where increased collagen deposition is also evident.

View larger version (81K): [in this window] [in a new window]   Fig. 4.   Localization of IGF-I and procollagen 1(I) mRNAs in muscularis propria of uninvolved and involved colon of a patient with CD and sense controls. Bright-field (BF and SR) and dark-field (IGF-I and Col I) photomicrographs illustrate in situ hybridization data. IGF-I and Col I at left and middle, dark-field images of sections of muscularis propria of uninvolved and involved colonic muscularis propria from a CD patient hybridized with IGF-I or procollagen 1(I) antisense RNA probes, respectively; IGF-I and Col I at right (Sense), dark-field images of sections of involved colonic muscularis propria from the same CD patient hybridized with IGF-I or procollagen 1(I) sense RNA probes, respectively. Exposure times were 14 days for sections hybridized with the IGF-I and 3 days for sections hybridized with the procollagen 1(I) antisense or sense probes. BF, hematoxylin-stained section after hybridization with the IGF-I probe to show histology; SR, Sirius red-stained section adjacent to that hybridized with the procollagen 1(I) probe to indicate localization of collagen protein (appears darker). Note increased IGF-I and procollagen 1(I) mRNA expression in the regions between 2 muscle layers and throughout 1 muscle layer in involved muscularis propria. SR shows increased collagen deposition in these same regions. Lack of detectable hybridization signal with sense control IGF-I and procollagen 1(I) probes indicates specificity of signals obtained with the antisense probes.

Colocalization of IGF-I precursor immunoreactivity and collagen in intestine of patients with CD. Immunohistochemistry was used to establish whether pro-IGF-I, like the mRNA encoding the IGF-I precursor, was colocalized to areas of fibrosis in involved intestine from patients with CD. It was not possible to directly compare localization of IGF-I mRNA and encoded protein on the same tissue sections. This is because IGF-I mRNA localization has proven successful only on frozen, postfixed sections using isotopically labeled probes, whereas immunohistochemistry to localize IGF-I precursor has proven successful only on fixed, paraffin-embedded tissue. As illustrated in Fig. 5, pro-IGF-I was detected within regions of collagen deposition in the submucosa and muscularis propria of involved ileum or colon of patients with CD. Within the submucosa, IGF-I precursor immunoreactivity was observed in mesenchyme-like cells surrounding lymphoid aggregates similar to the distribution of IGF-I mRNA (cf. Figs. 2 and 5). Within the muscularis propria, distribution of pro-IGF-I was particularly striking. As shown in examples of involved ileum or colon in Fig. 5, strong immunostaining for IGF-I precursor was observed in regions of disorganized/fibrotic muscularis propria visualized by Masson's trichrome stain on adjacent sections. Immunostaining for IGF-I precursor was less extensive or intense in regions of muscularis propria that appeared relatively normal.

View larger version (121K): [in this window] [in a new window]   Fig. 5.   Localization of IGF-I precursor immunoreactivity and collagen within involved submucosa and muscularis propria of patients with CD. Bright-field photomicrographs of involved ileal submucosa (left), involved ileal muscularis propria (middle), and involved colonic muscularis propria (right) of patients with CD: H + E, hematoxylin-and-eosin-stained sections to show histology; MT, Masson's trichrome-stained adjacent sections to show histology and collagen as blue stain; IGF-I, adjacent section incubated with an antibody to IGF-I precursor followed by localization of sites of antibody binding (brown stain) using the avidin-biotin-peroxidase method. Note that IGF-I precursor localizes predominantly to the same sites as collagen.

Distribution of phenotypic markers of mesenchymal cell subtypes and IGF-I precursor in normal intestine and uninvolved and involved intestine of patients with CD. We wished to identify the mesenchymal cells associated with increased IGF-I expression and fibrosis in involved intestine of patients with CD. To achieve this we performed immunohistochemistry on involved and uninvolved CD ileum and colon using antibodies to pro-IGF-I in conjunction with a panel of antibodies to antigens associated with different mesenchymal cell subtypes. Inasmuch as information was limited about the localization of these different mesenchymal cell antigens in normal compared with CD intestine, we included in the immunohistochemical analyses samples of ileum and colon from patients with no history of IBD as a "normal" comparison group. Representative data in Figs. 6-8 reflect findings in samples of normal ileum from two patients, normal colon from two patients, uninvolved and involved ileum from three CD patients, and uninvolved and involved colon from two CD patients. Figures 6 and 7 show data obtained in normal, uninvolved and involved mucosa/submucosa of the ileum (Fig. 6) or colon (Fig. 7) using all antibodies except c-kit, which did not immunostain cells in the mucosa/submucosa. Muscularis propria of uninvolved ileum and colon is also represented in Figs. 6 and 7. Figure 8 compares immunostaining of muscularis propria in ileum from a noninflammatory control and involved ileum from a CD patient. Immunostaining patterns in normal and involved colonic muscularis propria were essentially the same as those observed in ileum, and so data are shown only for ileum. Figure 9 shows high-power views of selected regions of involved ileum from different CD patients to illustrate colocalization of particular antigens and IGF-I immunostaining in samples with particularly severe fibrosis of the muscularis propria.

View larger version (84K): [in this window] [in a new window]   Fig. 6.   Mesenchymal cell subtypes and IGF-I precursor localization in ileum. Bright-field photomicrographs show sections of ileum of a patient with no history of imflammatory bowel disease and sections of uninvolved and involved ileum from a patient with CD. Serial sections were treated as follows: MT, Masson's trichrome-stained sections to show histology and collagen as blue stain; other sections show immunostaining with antibodies to -smooth muscle (SM) actin, desmin (des), vimentin (vim), and IGF-I precursor (IGF-I). mm, Muscularis mucosa; sm, submucosa; mp, muscularis propria (represented only in uninvolved ileum). White arrows in normal and uninvolved ileum point to the typical location of subepithelial myofibroblasts, where positive immunostaining for SM-actin (A), vimentin (V), and IGF-I precursor is evident, but only a few cells are desmin (D) positive. Muscularis mucosa of normal and uninvolved ileum and muscularis propria of uninvolved ileum are A+ and D+ but V. Note increased numbers of V+ cells in muscularis mucosa and submucosa of involved ileum. In involved ileum, black arrows indicate a region populated by V+/A cells (see Fig. 9A), and white arrows point to regions with V+/A+ cells. Note increase in IGF-I precursor immunostaining in muscularis mucosa and submucosa of involved ileum.

View larger version (90K): [in this window] [in a new window]   Fig. 7.   Mesenchymal cell subtypes and IGF-I precursor localization in colon. Bright-field photomicrographs show sections of colon from a patient with no history of inflammatory bowel disease (normal) and sections of uninvolved and involved colon from a patient with CD. Some disease activity is evident in uninvolved colon. Serial sections were treated as follows: Masson's trichrome-stained sections show histology and collagen as blue stain; other sections show immunostaining with antibodies to SM-actin, desmin, vimentin, and IGF-I precursor. mm, Muscularis mucosa; sm, submucosa; mp, muscularis propria (represented only in uninvolved and involved colon). In normal and uninvolved colon, white arrows point to the typical location of subepithelial myofibroblasts that are A+ and V+ but D. Normal muscularis mucosa is A+ but D and V; uninvolved muscularis mucosa is A+ and contains some D+ and V+ cells. In involved colon the mucosa is largely destroyed, and it is not possible to clearly delineate the epithelial layer, muscularis mucosa, or submucosa. Note the greater number of V+ cells in involved mucosa/submucosa than in normal tissue. In the involved sample, black arrows indicate an area populated largely by V+/A cells that is strongly positive for IGF precursor, and white arrows point to regions at the surface of the mucosa where SM-actin and vimentin immunostaining overlap.

View larger version (150K): [in this window] [in a new window]   Fig. 8.   Mesenchymal cell phenotypes and IGF-I localization in muscularis propria of normal and involved CD ileum. Bright-field photomicrographs show sections of muscularis propria from a sample of normal ileum and from involved ileum from a patient with CD. Serial sections were treated as follows: Masson's trichrome-stained sections show histology and collagen as blue stain; other sections show immunostaining with antibodies to SM-actin, desmin, c-kit, vimentin, and IGF-I precursor. Normal muscularis propria is A+ and D+ but largely V. Note c-kit-positive/V+ cells at the junction between circular and longitudinal muscle layer in normal muscularis propria, typical of the location of interstitial cells of Cajal. IGF-I precursor may weakly immunostain these c-kit-positive cells, but this was not definitive in any sample of normal ileum or colon studied. In involved muscularis propria, note the larger number of V+ cells and no c-kit-positive cells. IGF-I precursor immunostaining is clearly higher in involved muscularis propria than in normal muscularis propria, especially in the more disorganized/fibrotic layer, throughout the V+/A/D region between circular and longitudinal layers and in collagenous septa between muscle bundles.

View larger version (133K): [in this window] [in a new window]   Fig. 9.   Colocalization of IGF-I precursor and antigens for particular mesenchymal cell subtypes in involved ileum of patients with CD. Bright-field photomicrographs of sections of muscularis mucosa (A) and muscularis propria (B-D) of involved CD ileum are shown. Masson's trichrome-stained sections show histology and collagen as blue stain. Other sections show immunostaining with antibodies to SM-actin, vimentin, and IGF-I precursor. Black arrows indicate regions that are V+/A; white arrows indicate regions that are V+/A+. A: comparison of SM-actin and vimentin immunostaining in involved muscularis mucosa indicates a majority of V+/A+ cells as well as some V+/A cells. Note strong IGF-I precursor immunoreactivity in the region of V+/A cells but also IGF-I precursor located in regions containing V+/A+ cells. B: involved muscularis propria sample showing intense IGF-I immunostaining in a highly fibrotic region containing more V+ than A+ cells and in V+/A cells located between muscle layers. C and D: low- and higher-power views, respectively, of a highly fibrotic involved muscularis propria. Note intense IGF-I immunostaining in smooth muscle-like bundles populated by V+/A+ cells (1 indicated by white arrow) as well as the surrounding fibrotic area that contains V+/A cells. Desmin immunostaining of these sections revealed patterns identical to SM-actin immunostaining (data not shown).

Immunostaining patterns of mesenchymal cell subtypes in normal and uninvolved ileum. In normal and uninvolved ileum (Figs. 6 and 8), the patterns of immunostaining indicate that vimentin and SM-actin antibodies can distinguish subepithelial myofibroblasts (SEMF), which are V⁺/A⁺, enteric smooth muscle cells in muscularis mucosa and muscularis propria, which are V/A⁺, and scattered fibroblasts in the submucosa or lying between smooth muscle bundles in muscularis propria, which are V⁺/A. SM-actin immunostaining was sometimes not observed or was not very strong in the circular or the longitudinal layer of the muscularis propria, as shown for the longitudinal muscle layer of the normal ileum in Fig. 8. Across different samples analyzed, the relative intensity of SM-actin immunostaining of the two layers of muscularis propria varied depending on the orientation of smooth muscle fibers in the section, and this seemed to reflect the orientation/accessibility of antigen in the plane of the section rather than a true difference in relative expression of SM-actin in circular and longitudinal muscle layers. Desmin antibodies immunostained cells in the lamina propria within villi and in pericryptal regions of normal and uninvolved ileum, but these were fewer in number than observed with SM-actin or vimentin antibodies (Fig. 6). This indicates that either a subset of ileal SEMF are V⁺/A⁺/D⁺ or desmin antibodies immunostain only smooth muscle cells in villus lamina propria or pericryptal regions. Desmin antibodies immunostained the muscularis mucosa and muscularis propria in normal and uninvolved ileum (Figs. 6 and 8), consistent with a consensus in the literature that normal enteric smooth muscle has V/A⁺/D⁺ phenotype (21, 22). Cells positive for c-kit were detected between the muscle layers of normal ileal muscularis propria, typical of the location of ICC, and these were also V⁺/A/D (Fig. 8).

Immunostaining patterns of mesenchymal cell subtypes in normal and uninvolved colon. In normal colonic mucosa, patterns of immunostaining with vimentin and SM-actin were similar to those observed in normal and uninvolved ileal mucosa, in that SEMF were V⁺/A⁺ and the muscularis mucosa was V/A⁺ (Fig. 7). Normal colonic mucosa did, however, differ from normal ileal mucosa in patterns of desmin immunostaining. In normal colon, the mucosa was essentially D, indicating that colonic SEMF and muscularis mucosa do not express desmin (Fig. 7). The uninvolved CD colon samples had more inflammatory cells in lamina propria and somewhat thickened muscularis mucosa relative to normal colon, indicating some mild disease activity (Fig. 7). In contrast to normal colon, the samples of uninvolved colon showed D⁺ cells in the muscularis mucosa (Fig. 7). As well as desmin staining, the thickened muscularis mucosa in uninvolved colon showed increased numbers of V⁺ cells relative to normal muscularis mucosa. These findings in uninvolved colonic mucosa, albeit on a small sample number, support a possibility that in CD a subset of smooth muscle cells of the colonic muscularis mucosa may change phenotype to express desmin or transform toward a V⁺/A⁺/D⁺ myofibroblast phenotype. Vimentin, SM-actin, desmin, and c-kit immunostaining patterns in normal and uninvolved colonic muscularis propria were indistinguishable from those in ileal muscularis propria (data not shown).

Immunostaining patterns of mesenchymal cell subtypes in involved CD ileum and colon. Masson's trichrome staining of involved CD ileal mucosa revealed major disease activity (Figs. 6 and 9A). The muscularis mucosa was thickened and disorganized, with islands of what appeared to be smooth muscle bundles surrounded by collagen. The most striking feature was the large number of V⁺ cells in the subepithelial region of ileal mucosa and the disorganized muscularis mucosa. V⁺ cells were evident even in A⁺/D⁺ smooth muscle bundles in the muscularis mucosa of involved CD ileum (Figs. 6 and Fig. 9A). Figure 9A clearly reveals coexpression of SM-actin and vimentin in the bundles of smooth muscle-like cells evident in sections stained with Masson's trichrome. In addition, there were clearly regions of fibrotic ileal muscularis mucosa surrounding the smooth muscle-like cells that were V⁺ but A and D (Figs. 6 and 9A). The patterns shown for involved ileal mucosa in Figs. 6 and 9A were similar to those in ileum of all other CD patients analyzed. Although they are not strictly quantitative, these data provide qualitative/semiquantitative data to indicate an increase in relative numbers of V⁺/A⁺/D⁺ myofibroblasts and V⁺/A/D fibroblasts in diseased ileal muscularis mucosa.

IGF-I precursor immunoreactivity in normal, uninvolved and involved ileum and colon. In normal and uninvolved ileum and colon, IGF-I precursor immunostaining was generally weak but was present in SEMF (Figs. 6 and 7). Little, if any, specific immunoreactivity was observed in mucosal epithelial cells, muscularis mucosa, or muscularis propria (Figs. 6-8). Samples of uninvolved colon that had mild disease activity showed stronger immunostaining for IGF-I precursor in submucosa than in normal colon or normal and uninvolved ileum (Figs. 6 and 7).

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Accumulating evidence from animal models (19, 28, 36, 39) indicates that upregulation of IGF-I mRNA occurs locally within the intestine during inflammation, particularly inflammation associated with fibrosis. The present findings extend on preliminary data (3, 15, 23) that elevated IGF-I mRNA expression occurs in involved intestine of patients with active CD. Our study, in an independent patient population, provides definitive evidence that the majority of patients with CD leading to resection exhibit elevated IGF-I mRNA expression in diseased intestinal segments and that this characteristic is not shared by the other major IBD, UC. Proinflammatory cytokines may induce IGF-I mRNA in involved intestine of patients with CD. Interleukin-1 and tumor necrosis factor- have, for example, been shown to increase IGF-I expression in some cell types (19, 25). However, increased expression of these proinflammatory cytokines would be expected in CD and UC and cannot readily account for the specific increase in IGF-I expression in CD. IGF-I expression may increase in CD, because transmural inflammation exposes mesenchymal cells in submucosa or muscularis propria to proinflammatory cytokines. Support for this possibility stems from our observations that submucosa and muscularis propria are the sites of strongest and most frequently observed IGF-I mRNA upregulation (Table 2).

Our data demonstrate that the increase in IGF-I mRNA in active CD is accompanied by increased expression of IGF-I precursor. This occurs in mesenchymal cells in regions of fibrosis within the muscularis mucosa, submucosa, and muscularis propria and overlaps with sites of increased procollagen 1(I) mRNA and collagen deposition. Increased IGF-I synthesis may therefore contribute to the development of fibrosis or represent an epiphenomenon of fibrosis. Support for the former possibility stems from in vitro studies in which IGF-I was shown to induce collagen expression in intestinal myofibroblasts (24) and intestinal smooth muscle cells (37), and IGF-I was shown to be a potent mitogen for these cells (13, 29). In CD, locally expressed IGF-I could contribute to fibrosis by expanding a population of phenotypically modified, collagen-producing intestinal mesenchymal cells and/or by directly stimulating collagen synthesis. Our data suggest that it will be of interest to develop strategies to experimentally modulate local IGF-I expression or action in intestinal mesenchymal cells in vivo to determine the functional relevance of local upregulation of IGF-I during intestinal inflammation. Transgenic mice have been developed recently in which the SM-actin promoter was used to target overexpression of IGF-I to enteric smooth muscle and myofibroblasts in vivo (17, 33). Such models could prove useful to better define whether IGF-I overexpression in these two mesenchymal cell subtypes alters mesenchymal cell hyperplasia and hypertrophy or collagen synthesis and fibrosis in response to experimental enterocolitis. If so, this would point to avenues for future therapeutic intervention in CD. A number of high-affinity IGFBPs can inhibit IGF-I action (18, 19) and could possibly be useful as experimental or therapeutic agents to inhibit IGF-I action during inflammation-induced fibrosis in the intestine. Defining the potential of IGFBPs as therapeutic agents requires a better understanding of the complex interplay between IGF-I and endogenous IGFBPs during intestinal inflammation and fibrosis. Most information about intestinal IGFBPs has been obtained in animal models. IGFBP3, IGFBP4, and IGFBP5 are the primary IGFBPs expressed in rodent intestine postnatally (18, 19). Expression of IGFBPs is altered in some situations of intestinal inflammation. The mRNA encoding IGFBP4, an IGFBP that inhibits IGF-I action in most systems tested (18, 19), is elevated in colon during TNBS-induced enterocolitis (36). Expression of colonic IGFBP5 mRNA also is increased at similar sites as IGF-I in TNBS and PG-PS models of entercolitis and in active CD (15, 36, 37). IGF-I is known to induce IGFBP5 in colonic smooth muscle cells in culture (37). IGFBP5 generally potentiates IGF-I action in mesenchymal cells (2, 18, 19, 37) and so could amplify any effects of IGF-I during inflammation-induced fibrosis. More information about the in vivo actions and interactions of IGF-I and IGFBPs in intestinal mesenchymal cells is clearly required before their role in inflammation-induced fibrosis can be defined. A transgenic mouse line with SM-actin promoter-mediated overexpression of IGFBP4 in intestinal myofibroblasts and enteric smooth muscle represents a promising model to address the in vivo consequences of IGFBP4 upregulation in experimental entercolitis (32). Indeed, in light of our findings about the mesenchymal cell subtypes associated with fibrosis and IGF-I overexpression in active CD, use of promoters to target overexpression of particular peptides to specific mesenchymal cell subtypes in transgenic models represents a generally attractive strategy to analyze the molecular mediators of inflammation-induced fibrosis in the intestine (17).

Fibrosis in CD is associated with hyperplasia and disorganization of enteric smooth muscle layers and excessive collagen deposition around and within the smooth muscle. The present study suggests that in active CD the muscularis mucosa and muscularis propria show an increase in the relative numbers of cells with fibroblast (V⁺/A/D) or myofibroblast (V⁺/A⁺/D⁺) phenotype. Although our findings indicate that V⁺/A/D and/or V⁺/A⁺/D⁺ cells are associated with regions of fibrosis in diseased intestine of patients with CD, they do not define the lineage that gives rise to this fibrogenic cell population. In the mucosa, expansion of the SEMF and/or transformation of smooth muscle cells within the muscularis mucosa toward collagen-expressing fibroblasts/myofibroblasts may contribute to fibrosis. In muscularis propria, our findings support a concept that excessive collagen deposition in CD may result from a phenotypic switch of resident enteric smooth muscle cells toward fibroblast/myofibroblast phenotype or infiltration/proliferation of collagen-producing fibroblasts/myofibroblasts within enteric smooth muscle layers. Our inability to detect normal c-kit/vimentin-positive ICC cells in involved muscularis propria is intriguing. This finding is consistent with a study in isolated canine circular smooth muscle, in which damage and structural alterations in ICC were reported in response to inflammatory stimuli (16). Our detection of V⁺ cells in regions of collagen deposition between circular and longitudinal smooth muscle layers raises the possibility that transformation of ICC toward fibroblast phenotype may accompany fibrosis of muscularis propria in CD. Further analyses of phenotypic or functional changes in ICC during IBD and experimental enterocolitis are warranted, inasmuch as this could contribute to motility disorders in IBD as well as fibrosis (16, 21, 22).

Even the most comprehensive immunohistochemical analyses of resected bowel from patients with CD can provide only a snapshot at one particular point in time and cannot define the phenotypic or functional changes in mesenchymal cells during initiation or progression of fibrosis. To gain a better understanding of the cellular basis of inflammation-induced fibrosis of the intestine, it will be necessary to first study animal models in which disease is more homogeneous and phenotype of mesenchymal cells may be better correlated with the onset and progression of fibrosis. In this regard, it is noteworthy that mesenchymal cell phenotype has not been examined in detail in any animal model of experimental enterocolitis and fibrosis, such as the rat PG-PS and TNBS models. Furthermore, it is clearly desirable to develop mouse models of inflammation-induced intestinal fibrosis so that the power of mouse genetics may be used to define mechanisms of fibrosis. There is no mouse model of intestinal inflammation in which fibrosis has been well documented. Our present findings in clinical samples provide important information that there are indeed changes in the phenotype of intestinal mesenchymal cells during active CD and suggest that more detailed studies of the cellular basis of fibrosis are warranted in appropriate animal models.

Mature 70-residue IGF-I is the predominant form of IGF-I in the circulation (18, 19), probably derived primarily from liver (35). However, available evidence indicates that a higher-molecular-weight form of IGF-I extended at the carboxy terminus by an E-domain peptide present in pro-IGF-I is the predominant form of IGF-I secreted from nonhepatic cells (10, 18, 19), including mesenchymal cells in the intestine (38, 39). Most reports documenting the sites of IGF-I precursor expression using immunohistochemistry have localized IGF-I precursor at sites of injury or inflammation in a number of nonhepatic tissues (18, 19, 39). Little or no IGF-I precursor is detected in the circulation. Neither IGF-I nor IGF-I precursor is readily detected or detected at high levels by immunohistochemistry in normal tissues, even though the IGF-I mRNA is expressed (18, 19, 39). We have speculated that the IGF-I precursor may accumulate at sites of tissue damage due to association with extracellular matrix (ECM) as well as increased synthesis by mesenchymal cells (18, 19, 39). Indirect support for this possibility stems from the present findings of high levels of IGF-I precursor localized within regions of increased ECM deposition in involved intestine of patients with CD. It also is known that some IGFBPs associate with ECM (2, 18, 19). Upregulation of IGFBPs during enterocolitis could serve to sequester IGF-I precursor onto ECM and/or modulate IGF-I action on mesenchymal cells. Ultrastructural or in vitro studies will be required to establish whether IGF-I precursor associates with ECM. In vitro studies could establish whether this is direct or mediated by IGFBPs and its biological relevance. Our present data demonstrate increased IGF-I precursor expression or accumulation at regions of fibrosis that were populated by cells with fibroblast and myofibroblast phenotype and, to a lesser extent, regions containing phenotypically normal enteric smooth muscle cells. High levels of IGF-I precursor could also be expressed in modified ICC in involved CD intestine, inasmuch as particularly high levels of IGF-I precursor were observed in V⁺ cells between circular and longitudinal layers of fibrotic muscularis propria. These findings suggest that future analyses aimed at defining the role of IGF-I precursor or IGF-I in conversion of mesenchymal cells to fibrogenic phenotype or in regulating proliferation of different intestinal mesenchymal cell subtypes are warranted. In this regard, it is of interest that preliminary studies indicate that IGF-I does stimulate proliferation of intestinal fibroblasts that were previously converted to myofibroblast phenotype by pretreatment with transforming growth factor- (24). Additional in vitro studies and analyses of mesenchymal cell responses to inflammation in the intestine of transgenic mice with targeted overexpression of IGF-I precursor in enteric smooth muscle and myofibroblasts (17, 33) should provide further insights into the functional significance of IGF-I precursor overexpression in intestinal mesenchymal cells during active CD.

In conclusion, increased IGF-I expression in mesenchymal cells at sites of fibrosis is a feature of active CD. In CD, severe fibrosis is associated with increased numbers of cells that exhibit fibroblast or myofibroblast phenotype in regions of intestine usually populated by smooth muscle cells. IGF-I precursor is localized primarily to regions of diseased and fibrotic bowel populated by fibroblasts and myofibroblasts and, to a lesser extent, phenotypically normal smooth muscle. IGF-I may regulate fibrosis in CD by actions on mesenchymal cell phenotype, proliferation, or collagen expression.