Abstract
Using a CD4+ T-cell-transplanted SCID mouse model of colitis, we have analyzed TGF-β transcription and translation in advanced disease. By in situ hybridization, the epithelium of both control and inflamed tissues transcribed TGF-β1 and TGF-β3 mRNAs, but both were expressed significantly farther along the crypt axis in disease. Control lamina propria cells transcribed little TGF-β1 or TGF-β3 mRNA, but in inflamed tissues many cells expressed mRNA for both isoforms. No TGF-β2 message was detected in either control or inflamed tissues. Immunohistochemistry for latent and active TGF-β1 showed that all cells produced perinuclear latent TGF-β1. The epithelial cell basal latent protein resulted in only low levels of subepithelial active protein, which co-localized with collagen IV and laminin in diseased and control tissue. Infiltrating cells expressed very low levels of active TGF-β. By ELISA, very low levels (0–69 pg/mg) of soluble total or active TGF-β were detected in hypotonic tissue lysates. TGF-β1 and TGF-β3 are produced by SCID mouse colon and transcription is increased in the colitis caused by transplantation of CD4+ T-cells, but this does not result in high levels of soluble active protein. Low levels of active TGF-β may be a factor contributing to unresolved inflammation.
Transforming growth factor-β (TGF-β) is a multifunctional growth factor that influences growth and differentiation in many cell types. There are three isoforms in mammals (reviewed in McCartney–Francis and Wahl 1994), the mRNA for which is differentially expressed both temporally and spatially (Miller et al. 1989; Pelton et al. 1989, 1990) during development and in adults, indicating that they have distinct functions. Regulatory roles for the different isoforms have been proposed in wound healing (O'Kane and Ferguson 1997) and epithelium-matrix interactions (Streuli et al. 1993), but little is known regarding their function in vivo. The biological activity of TGF-β is controlled through its interaction with the extracellular matrix (ECM), both by specific binding proteins (Miyazono et al. 1991) and by nonspecific interactions with, e.g., decorin (Ständer et al. 1999) or collagen IV (Paralkar et al. 1991). Release of the active dimer is mediated by serine proteases (reviewed in Flaumenhaft et al. 1993; Keski–Oja et al. 1995) or by thrombospondin (Crawford et al. 1998). TGF-β is involved in many aspects of intestinal physiology (Van Vlasselaer et al. 1992; Dignass and Podolsky 1993; Planchon et al. 1994; Halttunen et al. 1996). It can have both pro- and anti-inflammatory effects (Brandes et al. 1991; Reibman et al. 1991; Kitamura et al. 1996; Bright et al. 1997; Takeuchi et al. 1997). Because of its wide-ranging influence on gut physiology, transcription and translation of TGF-β in the intestine have been investigated in several studies. In mice, TGF-β1 and TGF-β3 messages have been demonstrated in small and large intestine, but there are conflicting reports for TGF-β2 transcription (Barnard et al. 1993; Wang et al. 1998).
Contradictory results have been obtained for small intestinal epithelial cell TGF-β mRNA distribution at different stages of enterocyte differentiation. It has been demonstrated to be restricted to crypt epithelial cells in some studies (Koyama and Podolsky 1989; Wang et al. 1998). In other studies (Barnard et al. 1989), transcription has been shown to be restricted to upper villous epithelial cells. TGF-β protein for all three isoforms has been demonstrated in small intestinal upper villous and colonic surface crypt epithelium (Barnard et al. 1993).
Although the inflammatory bowel diseases (IBD), Crohn's disease and ulcerative colitis, are of unknown etiology, an inappropriate and uncontrolled immune response involving infiltrating T-cells and macrophages, probably responding to intestinal luminal antigens, is believed to be involved. A feature of chronic disease is increased matrix deposition, probably modulated via myofibroblast TGF-β (Van Tol et al. 1999). It has been proposed by Strober et al. (1997) that the occurrence of mucosal inflammation depends on a balance between the pro-inflammatory effects of IFNγ and modulation by TGF-β. On the other hand, levels of TGF-β message and protein are increased in active human IBD (McCabe et al. 1993; Babyatsky et al. 1996). Several murine models of intestinal inflammation that probe the imbalance in mucosal T-cell reactivity have been generated. The mice develop a severe colitis after about 3 months due to infiltration by both T-cells and macrophages. Using one such model of CD4+ T cell-transplanted SCID mice (Reimann et al. 1995), this study examines the distribution of TGF-β mRNA and protein in advanced disease and investigates the hypothesis that TGF-β levels are elevated in disease in an attempt to resolve the inflammation, but that resolution fails for some reason. Elevated TGF-β may therefore contribute to both the inflammatory and the connective tissue disease pathology.
Materials and Methods
Animals and Development of Disease
C.B-17+/+ mice and congenic C.B-17 scid/scid (SCID) mice were bred and housed under specific pathogen-free conditions in the animal colonies of the University of Ulm, Germany and the Panum Institute, University of Copenhagen, Denmark. Experiments conformed to local and national guidelines. Transplanted and non-transplanted mice were monitored under identical conditions. Mice were injected IP with 5 × 105 non-fractionated CD4+ splenic T-cells and monitored for signs of the disease as previously described (Rudolphi et al. 1994; Bonhagen et al. 1996). Samples were taken at autopsy approximately 3 months after transfer. Tissue was scored (1, mild to 3, severe) in a blinded fashion for six parameters of disease development: tissue hypertrophy; mononuclear cell infiltration; crypt hyperplasia; crypt branching; crypt abscesses; and ulceration.
Immunohistochemistry, Antibodies, and Reagents
Colon tissue samples were taken from 11 CD4+ T-cell-transplanted, diseased SCID mice and from five non-transplanted age- and sex-matched control SCID mice. Samples were placed on labeled cork discs (RA Lamb; London, UK), covered with OCT (RA Lamb), and snap-frozen in isopentane cooled over liquid nitrogen. Samples were stored at –70C. Sections for immunohistochemistry (5 μm) and in situ hybridization (10 μm) were cut at −20C and air-dried. Spleen tissue from transplanted mice was used as a positive control for TGF-β staining.
Sections for β1-LAP immunohistology were fixed in 100% methanol at –20C for 10 min. Sections to be stained with other antibodies were fixed in acetone at 4C for 10 min and then rehydrated in PBS for 10 min. Nonspecific staining was blocked with 50% (v/v) methanol in PBS plus 0.6% (v/v) hydrogen peroxide, followed by 10% (v/v) normal goat serum plus 5% (w/v) non-fat milk powder (for TGF-β) or 10% normal horse serum (for β1-LAP, collagen, and laminin) in PBS for 1 hr at 20C, followed by an avidin-biotin block (Vector Laboratories; Peterborough, UK). Chicken anti-TGF-β1 (10 μg/ml) (R&D Systems Europe; Abingdon, UK), goat anti-β1-LAP (10 μg/ml) (R&D), or rabbit anti-mouse collagen IV, or rabbit anti-mouse laminin (both 1 μg/ml) (TCS Biologicals; Buckingham, UK) were applied in PBS. The anti-β1-LAP antibody was centrifuged at 10,000 × g for 5 min to remove debris, possibly caused by interaction between the antibody and latent TGF-β bound to immunoglobulin. Biotinylated secondary antibodies were goat anti-chicken 1:500 (Vector), donkey anti-rabbit (Jackson ImmunoResearch Laboratories; West Grove, PA) 1:1000 or donkey anti-goat (Jackson) 1:1000. After washing in PBS, StreptABComplex/HRP peroxidase complexes (DAKO; Glostrup, Denmark) were added to sections following the manufacturer's instructions. To amplify the TGF-β stain, biotin-conjugated anti-streptavidin (Vector) 1:100 was applied for 1 hr, followed by a second round of ABC complexes. After further washing, staining was visualized with 1.67 mM di-aminobenzidene-4HCl in 0.05 M Tris-HCl, pH 7.5, plus 0.06% (v/v) H2O2. The slides were then either counter-stained with Mayer's hematoxylin (Merck; Poole, UK) and mounted in DPX or mounted without counterstain which tended to mask the delicate TGF-β stain. For double immunofluorescence staining, the above secondaries were used sequentially with an avidin-biotin blocking step. TGF-β was detected with streptavidin-FITC (Southern Biotechnology; Birmingham, AL) and collagen IV or laminin with streptavidin-Texas Red (Vector). Chicken IgY, goat IgG, or rabbit IgG isotype controls were applied to sections at the same concentration as primary antibodies. All reagents were from Sigma, (Poole, UK) except where indicated.
In Situ Hybridization
Plasmids containing cDNA for TGF-β1, -β2, and -β3 (Pelton et al. 1990) (a kind gift from Professor H.L. Moses; Vanderbilt University Medical School, Nashville, TN) were used to make digoxygenin (DIG)-labeled riboprobes using the DIG RNA labeling kit from Boehringer–Mannheim (Mannheim, Germany). The TGF-β1, -β2, and -β3 constructs consisted of nucleotides 421–1395, 1511–1953, and 831–1440 of the murine cDNAs, respectively. Sequences of all inserts were confirmed before probes were made and the probe size was verified by Northern blotting of DIG-labeled transcripts.
In situ hybridization and detection of hybrids were performed essentially as previously published (Williams et al. 1999). Control and inflamed sections were placed on one slide and were treated identically. Briefly, tissues were fixed in 4% paraformaldehyde in PBS, permeabilized in 0.4% Triton X-100 in PBS, and then immersed in 0.25% acetic anhydride in 0.1 M triethanolamine buffer at pH 7.8, with PBS washes between each step. Slides were dehydrated through an ethanol series and then air-dried. Sections were then covered with hybridization buffer (50% deionized formamide, 10% dextran sulfate, 100 mM Tris-HCl, pH7.5, 5 mM EDTA, 0.5 mg/ml tRNA, 1 × Denhardt's solution, and 1.21 × SSC to a final concentration of 0.2 M Na+) (1 × SSC = 0.15 M NaCl, 0.015 M trisodium citrate) containing the riboprobe (0.2 μg/ml). Slides were covered with Geneframes and coverslips (Advanced Biotechnologies; Epsom, UK) and placed on a Hybaid Omnislide thermal cycler. This was programmed to 80C for 5 min, falling to the final hybridization temperature of 57C over 20 min before overnight incubation at 57C.
Sections were desalted before the stringent wash in 0.1 × SSC at 65C for 60 min in a Hybaid Omnislide wash module. Slides were then cooled and stabilized in 2 × SSC for 5 min before treatment with RNase A (Boehringer-Mannheim) (1 μg/ml) at 37C for 30 min followed by extensive washing. DIG-labeled transcripts were then detected as previously (Williams et al. 1999) using alkaline phosphatase-conjugated goat anti-DIG and NBT/BCIP. The reaction was stopped by washing in distilled water and the slides were then mounted in Aquaperm (Life Sciences International; Basingstoke, UK) without counterstaining.
Microscopy and Image Analysis
Images for analysis were viewed through an F15 Panasonic CCD video camera, grabbed with Neotech software, and analyzed using GaugeGem software (both from ME Electronics; Reading, UK) or captured using a Colour CoolView camera (Photonic Science; Robertsbridge, E. Sussex, UK) and Image Pro Plus software (Media Cybernetics; Baltimore, MD). Crypt lengths were measured using the ×10 objective. Images were calibrated using a graticule.
ELISA
Colons were taken from different groups of mice from those examined histologically, but transplanted mice showed similar levels of disease. Colons were frozen in liquid nitrogen after removal of fecal contents by extrusion under gentle pressure. Tissues were weighed while frozen, pulverized in a freezer mill, and then subjected to hypotonic lysis at a ratio of 100 mg tissue:500 μl lysis buffer (20 mM Tris buffer, pH 7.5) on ice, in the presence of protease inhibitors (2 mM PMSF, 1.66 μM aprotonin, 0.5 μM soybean trypsin inhibitor, 20 μM leupeptin, 10 mM EDTA). Samples were centrifuged at 20,000 × g for 20 min at 4C and lysates were used immediately. Protein concentration of the supernatant was determined by the bicinchoninic acid (BCA) method (Pierce; Rockford, IL) using BSA for standard curve preparation. Lysates were first diluted to an equal protein concentration of 6.8 mg/ml and then to 1.7 mg/ml in 0.2 × PBS, acidified to pH 2 with 0.5 M HCl, and neutralized with 0.5 M NaOH to pH 7.5 (pH was confirmed using pH-fix 0–14 dipsticks; Fisher, Loughborough, UK). PBS–Tween (0.05% v/v) (PBT) was then added to take samples to a dilution of 1.36 mg/ml and globulin-free BSA (Sigma) was added to a final concentration of 0.05% (w/v). The rabbit and chicken anti-TGF-β antibodies used in the ELISA had previously been shown to detect mouse splenic TGF-β by Western blotting (data not shown). Wells were coated with catching antibody (5 μg/ml rabbit anti-TGF-β; R&D) in carbonate buffer (pH 9.5) overnight at 4C. Plates were washed and then blocked with 0.1% BSA in PBT. After one wash, triplicate samples and TGF-β standards (R&D) were added to wells and incubated overnight at 4C. Plates were washed and then chicken anti-TGF-β (0.1 μg/ml; R&D) was added in 0.05% globulin-free BSA in PBT. Biotinylated goat anti-chicken (1:1000) was added after washes, then the complexes (1:50), followed by TMB substrate. Plates were read at 650 nm and concentrations of TGF-β calculated against a standard curve using Labsystems Genesis software (Life Sciences International). Positive controls included spleen lysates and mouse plasma (with and without acid treatment). Negative controls included omission of sample and replacement of chicken detection antibody with chicken IgY. The sensitivity of the assay was 10 pg/mg soluble protein. In some experiments, lysates or BSA (1.36 mg/ml) in the same buffer as added to wells were spiked with 100 pg TGF-β. Spiked samples were added to wells after 15-min incubation on ice. Results are expressed as expected TGF-β (100 pg spike plus any TGF-β present in unspiked sample) minus actual sample reading calculated from the standard curve.
Statistical Analysis
The Mann–Whitney U-test was used to determine the significance of differences between groups of data. An initial multiple regression analysis was done using a general linear model to determine if any disease score indicators correlated with TGF-β transcription. Final analysis was carried out using Pearson correlation.
Results
Disease Status
All T-cell-transplanted mice showed symptoms of advanced disease. All diseased colons were macroscopically thickened at autopsy. Histologically, the colon tissues exhibited various degrees of crypt hyperplasia and leukocyte infiltration. Infiltrating cells were principally CD3+ T-cells, macrophages, and polymorphonuclear cells (data not shown).
In Situ Hybridization
Distribution of TGF-β mRNA. The hybridization conditions used were sufficiently stringent to differentiate the three isoforms. All three TGF-β isoform probes gave strong hybridization signals using splenic megakaryocytes as a positive control (Figures 1K, 1L, and 1M).
Epithelium. TGF-β1 and TGF-β3 mRNA showed similar distribution in the epithelium. Transcription by crypt epithelial cells was observed in tissues from both control (Figures 1B and 1H), diseased (Figures 1A and 1G), and C.B-17+/+ (data not shown) animals. Expression in epithelial cells was clearly cytoplasmic (Figure 1J). No TGF-β2 mRNA was detected in either control or inflamed SCID mouse colon (Figures 1D and 1E) or C.B-17+/+ colon (data not shown).
In inflamed tissues, transcription of TGF-β1 and TGF-β3 by crypt epithelium showed an extended distribution along the crypt axis (compare Figures 1A and 1G to Figures 1B and 1H). In some areas, message was observed in all crypt epithelial cells, including surface epithelium. By image analysis, the average length of crypts in control mice was 199 μm (range 180–214 μm) compared to 470 μm (range 356–633 μm) in diseased animals. In control mice, crypt epithelium expressed TGF-β1 mRNA over an average of 63% of crypt length before the signal became undetectable, whereas in tissues from inflamed mice transcription of TGF-β1 extended further along the crypt axis (82%) (Figure 2). Similarly, extension of TGF-β3 transcription along the crypt axis was also seen (control 53%, inflamed 73%) (Figure 2). The difference between control and inflamed tissue was significant for both TGF-β1 and TGF-β3 (p < 0.01). However, there was no correlation between degree of crypt hyperplasia (crypt length) and the distance along the crypt of epithelial cells expressing TGF-β1 or TGF-β3 mRNA. TGF-β1 transcription was significantly correlated with the sum of the scores for crypt branching and mononuclear cell infiltration (p < 0.02) (Table 1).
Lamina Propria. Colon lamina propria of control SCID mice showed low-level infiltration by mononuclear cells which were almost all F4/80+ (data not shown), with occasional clusters of presumed pre-B-cells in immature follicles. There were very few TGF-β mRNA-positive cells. However, inflamed colon was infiltrated at variable levels throughout mucosa and serosa by cells expressing both TGF-β1 and TGF-β3 mRNA (Figures 1A, 1G, and 1J).
Determination of TGF-β Protein
Immunohistochemistry. TGF-β is produced as a latent precursor protein, with the active molecule maintained in the latent state by electrostatic interaction with its latency-associated peptide (LAP). This study used goat anti-β1-LAP to detect latent protein and chicken anti-TGF-β antibody, which has been reported to detect only active TGF-β (Ehrhart et al. 1997). Our ELISA results confirmed this because acid activation of spleen or plasma samples led to elevated TGF-β levels compared to untreated samples, indicating that the chicken antibody used for detection failed to recognize latent TGF-β in the non-acid-treated samples. Using spleen from CD4+-transplanted mice as positive control tissue, both latent (Figure 3F) and active TGF-β (Figure 4F) protein were readily detected in megakaryocytes, whereas mononuclear cells expressed both forms to a variable extent. The megakaryocyte stain co-localized with TGF-β mRNA by in situ hybridization on serial sections. On a technical note, it was important to use affinity-purified antibodies rather than the IgG-precipitated forms for detection of both β1-LAP and TGF-β in mouse gut by immunohistology, otherwise the equivalent IgG controls yielded very high background. This was not the case in mouse spleen. β1-LAP was detected in all cells in the colon from all groups of mice (Figures 3A, 3B, and 3G). Expression was perinuclear in and confined to the basal cytoplasm of epithelial cells (Figure 3G). Epithelial β1-LAP was most intense in deep crypt epithelium, but cells in the surface epithelium also expressed LAP. No clear changes were seen between controls and colitis.
Immunohistology of colon tissues showed weak but distinct expression of active TGF-β1 at the junction between the epithelium and lamina propria in both non-transplanted (Figures 4A and 4C) and diseased mice (Figures 4B and 4D). This basal protein co-localized with both collagen IV (Figures 5A and 5D) and laminin (Figure 5B) in the basal lamina in both controls and colitis and was strongest on the epithelial side of the basement membrane. Although this basal lamina TGF-β was observed along the entire length of the crypt, expression was strongest beneath luminal epithelium. In controls, laminin (Figure 5B) and collagen IV (Figure 5D) showed a similar distribution and were restricted to basement membranes bordering crypts or lamina propria vessels. As can be seen in Figure 5A, there was disordered collagen IV deposition within the expanded lamina propria of diseased mice, and both matrix components were more widely distributed throughout the tissue. Within epithelial cells, TGF-β expression was generally weak, confined to the apical cytoplasm, and was expressed by cells throughout the length of the crypt. Only rarely were clusters of strongly stained epithelial cells seen in diseased mice in the deep crypts or near the lumen (Figure 4G). These clusters were always surrounded by infiltrating cells, but no conclusions could be drawn because of their infrequent occurrence. In the lamina propria, mononuclear cells were only weakly positive for active TGF-β, whereas endothelium on some large and small vascular structures was positive in both controls (Figures 4A and 4C) and colitis (data not shown).
Localization of TGF-β1, TGF-β2, and TGF-β3 mRNA in colon mucosa by in situ hybridization. Representative sections from a CD4+ T cell-transplanted SCID mouse with colitis (
Transcription of TGF-β1 and TGF-β3 is extended along the crypt axis in the epithelial compartment in colitis. Twenty-five crypts were measured for each animal and the percentage of the crypt epithelium positive for TGF-β transcription was determined by length. Each point represents the mean positive crypt length for a single mouse. p < 0.01 inflamed vs non-transplanted SCID control for TGF-β1 and TGF-β3. Bars represent mean values.
ELISA
Increased TGF-β transcription in disease did not lead to obvious differences in the levels of TGF-β protein by immunohistology. Therefore, to determine changes in protein levels, tissue lysates were first examined by Western blotting, but no protein was detected in colon tissue, whereas it was readily detected in spleen (data not shown). Therefore, an ELISA was developed that detected mouse TGF-β at a sensitivity of 10 pg/mg (700-fold more sensitive than Western blotting). Acid-activated spleen lysates and mouse plasma contained 830 pg/mg and 5220 pg/ml, respectively, with much lower levels (44 pg/mg and 180 pg/ml) in non-activated samples (plasma data not shown). In lysates of control and diseased colon (Figure 6), levels of TGF-β were near the limits of detection of the assay and many samples were negative. There was no difference in either acid-activated or non-activated samples between controls and diseased mice. To investigate possible sequestration of TGF-β by soluble matrix components, colon lysates were spiked with 100 pg TGF-β. Detection of the TGF-β spike in lysis buffer containing BSA as a protein source at an equivalent protein concentration to cell extracts gave 102.4 ± 4.6 pg recovered (mean ± SD of three determinations). In contrast, only 30% of the TGF-β spike was detectable when spiked colon lysates were used in the ELISA (mean 28.7 ± 6.8 pg recovered).
Discussion
This study has shown that, in severe colon inflammation, TGF-β1 and TGF-β3 transcription by cells in all layers of the intestine is increased, but particularly in the epithelium. Although mRNA was translated to β1-LAP, this did not result in a concomitant elevation of active TGF-β protein. To our knowledge, there have been no other studies in the rodent large intestine of TGF-β transcription detected by in situ hybridization in combination with the determination of latent and active protein by immunohistology. First, with regard to the epithelium, we have shown that, in control SCID mice (and in C.B-17+/+ mice), TGF-β1 and TGF-β3 mRNA were readily detected in colon lower crypt epithelial cells. This agrees with the majority of reports on rodent small intestine (Koyama and Podolsky 1989; Wang et al. 1998). In diseased tissue, transcription extended along the crypt axis and in some specimens was expressed by most surface crypt epithelial cells. Similarly, in radiation-induced enteropathy, small intestinal epithelial TGF-β1, but not TGF-β3, transcription was extended along the crypt axis (Wang et al. 1998). This contrasts with the single report of TGF-β transcription in human IBD (Babyatsky et al. 1996) in which, surprisingly, no TGF-β mRNA was observed in the epithelium in colon biopsy tissue. In our study, a significant correlation was shown between the disease scores for crypt branching and cell infiltration and the extended transcription of TGF-β1 in the crypt epithelium. TGF-β1 has many promoter sequences activated in disease (McCartney–Francis and Wahl 1994), and it is therefore highly likely that increased epithelial TGF-β1 transcription resulted from inflammatory signals from the underlying lamina propria. Inflammatory cytokines have been shown to increase epithelial permeability (Planchon et al. 1994), and the increased epithelial TGF-β transcription may indicate that epithelial integrity has been compromised and that restitution, known to involve TGF-β protein (Dignass and Podolsky 1993), has been signaled. TGF-β2 mRNA was not observed in any colon tissue. This agrees with data from other rodent studies by Barnard et al. (1989, 1993). Moreover, using the same sequence of DNA to generate probes as in our study, TGF-β2 mRNA has been found in the submucosa, but not the epithelium, of embryonic gut by in situ hybridization (Pelton et al. 1989). Wang et al. (1998) have shown TGF-β2 mRNA in rat small intestine, but the hybridizations in their study were carried out using less stringent conditions and on paraffin-embedded tissue. In our study, stringent conditions were employed to reduce sense background, but because this would also have reduced specific hybridization it is possible that TGF-β2 is transcribed but at a much lower level than either TGF-β1 or TGF-β3.
Histological scores and percent of crypt axis epithelium transcribing TGF-β
Detection of β1-LAP in the mucosa of control SCID and transplanted SCID mouse colon by immunohistology. Representative sections from control (
Localization of active TGF-β by immunohistology. Active TGF-β protein is principally subepithelial (large arrows) in colon tissues or associated with lamina propria vessels (small arrows). Lamina propria cells are weakly positive (filled arrowheads). Splenic megakaryocytes are indicated by open arrowheads. Control SCID colon (
Co-localization (open arrowheads) of TGF-β (FITC) with basement membrane collagen IV (Texas Red) (
In summary, we suggest that, in the normal adult rodent, concordant transcription of TGF-β1 and TGF-β3 takes place in colon immature crypt epithelial cells and that, as differentiation takes place, transcription is suppressed. This is consistent with the known role of TGF-β in enterocyte differentiation and with TGF-β2 being essential for epithelial–mesenchymal interactions during embryogenesis, but not for epithelial differentiation in the adult. In inflammation-induced crypt hyperplasia, as the proliferative zone extends and the proportion of immature enterocytes in the crypt increases, so the transcription of TGF-β1 and TGF-β3 is extended in distribution, with no effect on TGF-β2 transcription. Indeed, no change was found in epithelial TGF-β2 transcription in radiation-induced enteropathy (Wang et al. 1998). It is also possible that transcription is not suppressed in mature enterocytes in disease.
Second, with regard to the inflammatory infiltrate, many cells were observed that transcribed both TGF-β1 and TGF-β3. They were seen in all layers of the intestine, including the muscle wall. Similarly, in human IBD (Babyatsky et al. 1996), there were increased numbers of TGF-β-transcribing lamina propria cells, and in radiation-induced enteropathy TGF-β transcription was increased in many cell types in all layers of the gut (Wang et al. 1998).
ELISA for TGF-β in colon lysates of control and diseased mice. Levels of TGF-β in colon tissue were similar and near the limits of detection (dotted line) for all groups of mice and were 2 logs lower than in spleen (SP). C.B-17 (CB), non-transplanted (NT), diseased SCID (IBD) mice.
In contrast to transcription, little evidence for elevated levels of latent or active TGF-β protein was found in this study by immunohistology, Western blotting, or ELISA. All cells produced perinuclear β1-LAP, with strongest epithelial expression in deep crypt epithelium. However, there was no obvious increase in translation of LAP in colitis, even though transcription of TGF-β1 was greatly increased, most notably in the epithelium. There may be post-transcriptional controls that prevent increased translation to LAP.
Active TGF-β1 protein was clearly demonstrated at the interface between the epithelium and the lamina propria. In both controls and inflamed tissue, it co-localized with basement membrane collagen IV and laminin. Collagen IV has been shown to bind TGF-β (Paralkar et al. 1991) with greater affinity than laminin or collagen I. Nakajima et al. (1999) have also demonstrated one of the latent TGF-β-binding proteins, LTBP-1, at the base of epithelial cells in the embryonic mouse intestine and, using electron microscopy, localized LTBP to microfibrils in the basement membrane. It has been proposed that LTBP is involved in targeting TGF-β to the cell surface before its activation (Flaumenhaft et al. 1993), and also that matrix represents a reservoir for various growth factors (Taipale et al. 1996). The chicken antibody used in our study has been reported to recognize only active TGF-β (Ehrhart et al. 1997), and this is substantiated by our ELISA data. It is plausible that latent TGF-β secreted by epithelial cells is sequestered in the basement membrane and undergoes activation by the interaction between the epithelium and either lamina propria cells or components in the extracellular matrix, e.g., thrombospondin (Crawford et al. 1998) or plasminogen (Vassalli et al. 1991). This locally generated active TGF-β could then regulate epithelial cell growth and differentiation, or it might be sequestered and thus be protected from degradation (Paralkar et al. 1991). The distribution of active TGF-β was similar in control and diseased SCID mice along the length of the crypt, but the strongest signal was detected at the neck of the crypts and at the luminal surface, where cells are maximally differentiated. Investigation of receptor and signaling molecule distribution should clarify those cells that respond to TGF-β (Zhao et al. 1998). Epithelial intracellular expression of active TGF-β was generally weak and was observed along the length of the crypt. However, occasional clusters of epithelial cells expressed high levels of active TGF-β, and these were surrounded by accumulations of infiltrating cells. This could represent clusters of cells either producing large amounts of TGF-β or responding to TGF-β where the intracellular stain represents receptor-bound/internalized TGF-β, and their infrequent observation indicates the tight regulation of the short-lived mature dimer.
Only weak expression of TGF-β by infiltrating mononuclear cells was seen. These findings of low levels of active TGF-β expression in both the epithelium and the inflammatory infiltrate suggest that levels are tightly controlled in the gut. The absorption of active TGF-β by colon lysates suggests the presence of an inactivating protein such as decorin, which has been shown to tightly bind TGF-β and block its biological activity (Ständer et al. 1999). Unfortunately, antibodies to mouse decorin were not available to test this possibility.
Other studies of the distribution of TGF-β protein (Barnard et al. 1993), using different antibodies and using paraffin-embedded tissue, have demonstrated all three isoforms in the small intestinal upper villus and colon surface epithelial cells. It is noteworthy that although transcription of TGF-β2 was barely detectable, TGF-β2 protein was reported in these studies, although it is possible that serum may have been the source of this TGF-β2 protein. These studies did not attempt to differentiate active from latent TGF-β. In preliminary studies using paraffin-embedded tissue, we found that the anti-β1-LAP antibody yielded a similar staining pattern to that of frozen tissue but that the chicken antibody resulted in staining equivalent to background (data not shown). TGF-β protein levels are elevated in human IBD tissue samples (Babyatsky et al. 1996) by bioassay, and in a rat model of colitis (Mourelle et al. 1998) by ELISA. These authors prepared tissue homogenates using similar methods to those used in our study but found much higher levels (0.3–7 ng/mg), although they used different assays and antibodies. Our study used spleen tissues from colitic mice as positive control tissue for all assays. Immunohistology on spleen tissues showed that both β1-LAP and active TGF-β1 proteins co-localized with transcription for all three isoforms in megakaryocytes. This concurs with the observed role of these cells in the production of platelet precursors, known to contain high levels of TGF-β. Because of this, we feel that the spleen was an appropriate positive control for mouse TGF-β and that colon tissue contained only low levels of soluble TGF-β. Most previous studies did not attempt to access matrix-bound TGF-β in tissues. Although matrix/TGF-β interactions may be different in gut and spleen, our immunohistochemical data suggest that the low level of active TGF-β in normal colon tissue is only marginally increased in colitis, despite abundant transcription and translation of β1-LAP in the mucosa. It is possible that all colon TGF-β is tightly matrix-bound or that the binding protein that assists assembly and secretion (Miyazono et al. 1991) is ineffective, or that molecules such as decorin (Ständer et al. 1999) are upregulated. Alternatively, TGF-β activation mechanisms may be downregulated, or short-lived mature TGF-β dimers may be degraded in the inflammatory lesion. All these scenarios imply a failure to generate sufficient active TGF-β to downregulate the mucosal inflammation induced by CD4+ T-cell transfer into SCID mice. Treatment with mucosa-derived TGF-β, produced either by gene therapy (Giladi et al. 1995) or by TGF-β-secreting T-cells generated during oral tolerance (Groux et al. 1997), may ameliorate colitis if delivered early in disease (Powrie et al. 1996). However, TGF-β bound in the basal lamina after release by MMPs from, e.g., decorin (Imai et al. 1997), or after activation may contribute to increased matrix deposition by epithelial and mesenchymal cells, leading to fibrotic pathologies. Taken together, these studies highlight the need for a full understanding of the regulation of this multifunctional molecule during mucosal inflammation.
Footnotes
Acknowledgements
Supported by grants from the European Union no. BMH4-96-0612 and QLG1-CT-1999–00050.