A Novel Model of Inflammatory Bowel Disease: Mice Deficient for the Multiple Drug Resistance Gene, mdr1a, Spontaneously Develop Colitis

The inflammatory bowel diseases (IBD)², Crohn’s disease and ulcerative colitis, are syndromes characterized by chronic inflammation of the gastrointestinal tract. The etiology of these disorders remains unknown despite many years of investigation. The current hypotheses suggest that persistent intestinal inflammation may be the result of either enhanced or aberrant immunologic responsiveness to normal constituents of the gut lumen (1, 2) or an overall autoimmune dysregulation and imbalance (3, 4, 5, 6, 7, 8, 9, 10, 11). The recent establishment of various animal models relevant for studying the pathogenesis of intestinal inflammation has provided some insight into potential disease mechanisms. In particular, the discovery that many animals with altered T cell populations or cytokine deficiencies spontaneously develop colitis has implicated an immune imbalance as one of the critical factors in the manifestation of IBD (3, 4, 5, 6, 7, 8, 9, 10, 11). Although the pathologic changes associated with IBD and the extensive lymphoid infiltrates into the inflammatory lesion suggest that there is some immunologic dysregulation, it is likely that many of the immunologic changes associated with IBD are probably not the primary cause of the disease process, but may be due to secondary nonspecific inflammation.

There is increasing evidence that the intestinal microflora is an important cofactor in the pathogenesis of intestinal inflammation. The development of spontaneous colitis in many mouse models can be prevented if mice are generated and maintained in a germfree environment or treated with oral antibiotics, highlighting the central role for bacterial colonization in the initiation and/or perpetuation of experimental IBD (4, 12, 13, 14, 15, 16). There is also evidence that human patients with IBD suffer from adverse and enhanced reactivity to their autologous resident intestinal flora (1, 2, 17). Considering that the cells of the mucosal immune system are protected from the large antigenic load in the gut lumen by a single layer of epithelial cells, it seems reasonable to assume that epithelial cells play a major role in providing a barrier that regulates contact between bacteria and immune cells. In support of this, transgenic mice expressing a mutant cadherin on their epithelial cells in the small intestine spontaneously develop IBD, presumably due to breaches occurring within the epithelial cell barrier (18). Thus, proteins expressed in epithelial cells may in some way contribute, directly or indirectly, to maintaining the protective barrier.

Intestinal epithelial cells and some lymphocyte subsets have been shown to express multiple drug resistance (MDR) genes (19, 20, 21, 22, 23, 24). These genes belong to a family of transporters known as ATP binding cassette transporters, which are characterized by their ability to pump small amphiphilic and hydrophobic molecules across membranes in an ATP-dependent manner (25, 26). MDR genes were first identified by their ability to confer resistance to chemotherapeutic agents in tumors and neoplastic cell lines (27). Three MDRgenes have been identified in rodents and two in humans, each of which has a restricted pattern of tissue expression and, most likely, a different function. Although it is well known that some of these gene products (MDR1 in humans and mdr1a and mdr3 in mice) can actively pump toxic drugs out of cells, their natural in vivo role has not been fully elucidated (28, 29). In mice, mdr1a is expressed on many tissues including intestinal epithelial cells, CD8⁺ T cells, a subset of CD4⁺ T cells, hematopoietic cells, and cells at the blood-brain barrier (19, 21, 23, 30, 31). The function of mdr1a in each of these different cell types and tissues is unknown.

Recently, mice with a targeted deletion of the mdr1a gene were generated (32). Analysis of these knockout mice has revealed that they have an increased sensitivity to certain drugs, but do not appear to show any constitutive abnormalities (31, 32, 33, 34). In the current study we observe that mdr1a^−/− mice housed under specific pathogen-free (SPF) conditions develop spontaneous intestinal inflammation. We describe the nature of the inflammatory cell infiltrate, describe a role of the intestinal microflora in the initiation and perpetuation of colitis in these mice, and provide evidence that the development of colitis appears to arise as a result of a defect in the intestinal epithelial barrier.

FVB control mice and mdr1a^−/− mice were initially purchased from Taconic Farms (Germantown, NY) and then bred and maintained in the SPF facility at Immunex (Seattle, WA) in accordance with approved ethical guidelines. mdr1a^−/− mice had been backcrossed onto the FVB strain for at least seven generations at Taconic Farms before shipping. Control and knockout mice used for experiments were age matched and received treatment simultaneously. Mice were monitored on alternate days for the presence of diarrhea and mucous discharge from the anus. All mice were bred and maintained under SPF conditions in Thoren isolation racks under positive pressure and fed autoclaved food and water ad libitum. The quality of the SPF facility is monitored through the Charles River Health Monitoring Plus program (Charles River, Wilmington MA) using the Charles River tracking profile. Sentinel mice receiving bedding from donor cages are tested for viral, parasite, and bacterial infection. There were no viruses, parasites, or pathogenic bacteria detected in our colony throughout the course of the study. Of particular note, we did not detect any evidence of Helicobacter infection (Helicobacter bilis, Helicobacter hepaticus, or Helicobacter spp.) in any mice.

Streptomycin sulfate (1.25% w/v), neomycin (1% w/v), bacitracin (1% w/v), and amphotericin (0.25% w/v) were dissolved in autoclaved drinking water (pH 2.5–2.8) supplemented with saccharine (1% w/v) and then filtered through a 0.22 μm Nalgene filter (Nalgene, Milwaukee, WI). As a control, mice received saccharine alone (1% w/v) in the drinking water. All reagents were obtained from Sigma (St. Louis, MO). Regular drinking water was replaced with antibiotic or saccharine drinking water during treatment.

The following primary Abs were used for flow cytometric analysis and immunoperoxidase histochemistry: anti-CD4 (rat IgG; L3T4), anti-CD8α (rat IgG; 53-6.7), anti-TCRαβ (hamster IgG; H57-597), anti-TCRγδ (hamster IgG; GL3), anti-GR1 (rat IgG; RB6–8C5), anti-B220 (CD45R) (rat IgG; RA3–6B2), and anti-CD3ε (hamster IgG; 500A2; Immunex). The primary Abs were conjugated either with biotin or R-phycoerythrin (PE). All Abs were obtained from PharMingen (San Diego, CA) except where indicated. The following secondary Abs were used for immunoperoxidase histochemistry: biotinylated sheep anti-rat IgG (Amersham, Arlington Heights, IL) and biotinylated goat anti-hamster IgG (Caltag, South San Francisco, CA).

Isolated cells were incubated with the appropriate PE- or biotin-conjugated primary Ab for 30 min on ice in PBS supplemented with 5% BSA and 0.02% sodium azide essentially as described previously (35). Stained cells were washed two times, incubated with streptavidin-allophycocyanin (Molecular Probes, Eugene, OR), and then fixed in 1% paraformaldahyde before analysis. Stained cells were analyzed on a Becton Dickinson FACScan flow cytometer (Immunocytochemistry Systems, San Jose, CA) using Becton Dickinson Cellquest software.

Lymphocytes were isolated from mice and incubated with 0.4 μg/ml rhodamine 123 (Sigma) for 15 min at 37°C in complete RPMI 1640 supplemented with 10% FCS. For two-color analysis, cells were simultaneously stained with PE-conjugated primary Abs. The stained, rhodamine-loaded cells were then washed two times and incubated in complete medium without rhodamine 123 for an additional 2 h at 37°C. Cells were then washed two times in complete medium and analyzed on a Becton Dickinson FACScan flow cytometer (Immunocytochemistry Systems) using Becton Dickinson Cellquest software.

Large intestines were removed, embedded in OCT compound (Tissue-Tek, Torrence, CA), frozen in a dry ice/ethanol bath, and stored at −80°C until use. Frozen sections (5 μm) of intestine were cut, fixed for 15 min in acetone, and stored at −80°C until staining. Peroxidase staining was performed using the Vectastain ABC system (Vector Laboratories, Burlingame, CA) according to specified directions as described previously (35). Briefly, tissue sections were preincubated in 5% goat serum, 5% bovine serum, and 1% mouse serum for 30 min at room temperature and then incubated with biotin-conjugated primary Ab for 1 h at room temperature. Slides were washed three times in Tris-buffered saline, further incubated with a biotin-conjugated secondary Ab for 30 min at room temperature, and washed again three times in Tris-buffered saline. Sections were then incubated with avidin peroxidase (Vectastain ABC system), washed, and finally incubated with diaminobenzidine substrate (Sigma) before counterstaining in hematoxylin. Positive cells were identified by the presence of a brown reaction product.

Tissues were frozen in liquid nitrogen as described above, and 5-μm cryosections were cut, air dried, and acetone fixed before staining. Tissue sections from both mdr1a^−/− and control FVB mice were stained with either 20 μg/ml FITC-conjugated anti-MDR mAb (C219; Signet Laboratories, Dedham, MA) or FITC-conjugated IgG2a isotype control mAb (PharMingen) in the presence of 2 mM TOTO-3 (Molecular Probes) and 30% mouse serum. Sections were visualized using a confocal laser scanning microscope (Molecular Dynamics, Sunnyvale, CA).

IEL were isolated according to a modification of published procedures (35, 36). Briefly, the small and large intestines were removed and washed in Ca²⁺/Mg²⁺-free HBSS (Life Technologies, Bethesda, MD). Peyer’s patches were removed, and the intestines were opened longitudinally, cut into 0.5-cm pieces, and placed in 1 mM EDTA in Ca²⁺/Mg²⁺-free HBSS at 37°C for three sequential 15-min incubations with intermittent vortexing to remove epithelial cells and IEL. Supernatants were collected and pooled for IEL purification. The remaining tissues were then washed in PBS and incubated in trypsin-EDTA two times to remove any residual epithelial cells, and the denuded tissues were digested with 90 U/ml collagenase type VIII (Sigma) in RPMI 1640/10% FCS for three sequential 20-min incubations at 37°C with intermittent vortex mixing to release LPL. Supernatants were collected and pooled for LPL purification. IEL and LPL were purified by passing cells over a prewet glass wool column before spinning them through a 40%/80% discontinuous Percoll gradient (Pharmacia, Uppsala, Sweden) as previously described (35). The cells at the 40%/80% interface were collected, washed, and resuspended in complete medium for analysis.

Intestinal bacterial Ags were prepared from fecal contents harvested from the cecum according to a modification of published procedures (37). Briefly, individual preparations were processed from three different mouse sources: 1) an FVB control mouse, 2) an mdr1a^−/− mouse with no signs of inflammation, and 3) an mdr1a^−/− mouse with active colitis. The cecal contents from the individual mice were placed in a 2-ml bead shaker tube half-filled with 2-μm zirconium beads suspended in PBS. The mixture was vortexed in a bead beater (Biospec Products, Bartlesville, OK) for 30 s at 5000 rpm and placed on ice. The intestinal bacterial slurry was then microfuged at 14,000 × g, and the supernatant was harvested and filtered through a 0.22 μm filter. The relative protein content in each sample was estimated using a Bradford assay (Bio-Rad, Hercules, CA).

Cell suspensions were prepared by gently mashing mesenteric lymph nodes (MLN) between ground glass slides and passing them through nylon mesh. MLN cells were cultured at a density of 2 × 10⁵ cells/well in 96-well flat-bottom plates in complete RPMI supplemented with 10% FCS in a final volume of 200 μl in a humidified 6% CO₂ incubator at 37°C for 48 h. Cells were cultured either in the presence of titrating amounts of intestinal bacterial Ag preparations, in the presence of 1 μg/ml Con A, or in anti-CD3 mAb-coated wells (1 μg/ml). Proliferation was assessed by addition of 1 μCi/well [³H]TdR 18 h before harvesting using a Matrix-96 cell harvester (Inotech, Lansing, MI). The amount of radioactivity incorporated into DNA was measured using a direct β counter (Packard, Meriden, CT). All cultures were performed in triplicate. Data are reported as mean cpm ± 1 SD of triplicate wells.

Mice were exsanguinated, and the serum was collected. Polyclonal serum Ab titers were determined by isotype-specific ELISAs for IgA, IgE, IgG1, IgG2a, IgG2b, IgG3, and IgM as previously described (38). Briefly, 96-well Nunc Maxisorp plates (Nunc, Naperville, IL) were coated overnight at 4°C with isotype-specific anti-mouse Ig Abs (0.2 mg/well), blocked with PBS/10% FCS for 2 h at room temperature, and washed with PBS/0.1% Tween-20. Serum samples and control samples were serially diluted in PBS/10% FCS starting at 1:100. Plates were incubated for 2 h at room temperature, washed, and incubated with the appropriate peroxidase-conjugated specific anti-isotype detecting Ab for an additional 2 h at room temperature. Plates were washed again, and enzyme activity was detected using TMB microwell peroxidase substrate reagents (Kirkegaard & Perry Laboratories, Gaithersburg, MD) and 2 N H₂SO₄. The amount of reaction product was assessed on an ELISA plate reader at OD 450 nm using the Deltasoft program (DeltaPoint, Monterey, CA). The affinity-purified Abs used for determining isotype-specific serum Ab titers by ELISA were obtained from Southern Biotechnology Associates (Birmingham, AL), except for anti-IgE, which was obtained from Serotec (Oxford, U.K.), and antiIgG2a, which was obtained from PharMingen. Ig concentrations were determined by comparing test sample dilution curves with known concentrations of isotype controls. Results are indicated as mean concentration ± 1 SD of two separate mice for each group.

mdr1a^−/− mice and FVB mice were irradiated with 10 Gy using a ¹³⁷Cs source, rested for 4–6 h, and then reconstituted by i.v. injection with 2 × 10⁶ bone marrow cells collected from the femurs of appropriate donors.

Mice were killed under sterile conditions, and aerobe/anaerobe bacterial culture swabs (Medical Wire and Equipment, Corsham, Wilts, U.K.) were used to collect cecal and colonic contents. Samples were shipped to Phoenix Laboratories for identification of aerobic and anaerobic bacterial outgrowth. At Phoenix Laboratories, swabs were streaked and bacterial species were identified based on nutrient requirements, growth, morphology, and staining properties.

Mice were killed under sterile conditions, and colonic washout samples were harvested and shipped to Microcheck for bacterial identification by microbial cellular fatty acid analysis. Samples were subcultured at Microcheck, and microbial fatty acid profiles were determined by high-resolution gas chromatography. Identification of bacteria in individual samples was determined by comparing the fatty acid profiles of the unknown sample with known profiles from microbial fatty acid libraries.

Efflux of rhodamine 123 from preloaded lymphocytes has been used to demonstrate the presence of mdr activity in peripheral T cells (19, 30). However, the assay does not distinguish whether rhodamine efflux is due to mdr1a activity, mdr3 activity, or perhaps even another unidentified transporter. We therefore analyzed rhodamine 123 efflux in lymphocytes from normal FVB mice and mdr1a^−/− mice to determine whether rhodamine 123 efflux was associated with a functional mdr1a gene and to identify which mucosal and peripheral lymphocyte subsets might exhibit the pump activity. mdr1a-associated functional activity was detected in MLN cells from normal FVB mice but not from mdr1a^−/− mice, as determined by ability to efflux rhodamine 123 (Fig. 1,A). These data indicate that a functional mdr1a gene product is necessary for rhodamine 123 efflux. To determine which subsets of mucosal lymphocytes have the mdr1a activity, we performed the rhodamine efflux assay on IEL and LPL isolated from the large intestines of FVB mice. We simultaneously stained these cells with mAbs against CD3, CD4, CD8, TCRαβ, and TCRγδ (Fig. 1 B). The individual CD3, CD4, CD8, TCRαβ, and TCRγδ subsets within the IEL and LPL compartments had similar efflux profiles. Virtually all of the CD8⁺ cells and TCRγδ⁺ cells appeared to exhibit mdr activity and efflux rhodamine 123, whereas only half of the CD4⁺ cells and half of the TCRαβ⁺ cells were able to efflux the dye. None of the mucosal lymphocyte populations from the mdr1a^−/− mice were able to efflux the rhodamine 123 (data not shown), further suggesting that mdr1a is necessary for rhodamine 123 efflux activity in mucosal lymphocytes.

Immunofluorescent staining was used to localize mdr1a expression in the colon. Fig. 2 shows mdr1a staining (green) on transverse sections of colons from FVB control (+/+) and mdr1a-deficient (−/−) mice. The sections were counterstained with TOTO-3 (red) to visualize gut architecture. mdr1a staining appears to be restricted to the apical surface of epithelial cells in FVB mice. No mdr1a staining was apparent on sections from mdr1a^−/− mice or on sections from FVB mice stained with an isotype control (data not shown). We were also unable to detect any mdr1a staining on LPL or on IEL in FVB mice, despite functional evidence that these cells express mdr1a (see above). It is unclear whether this was due to low expression levels of mdr1a on mucosal lymphocytes or to the relatively low frequency of lymphocytes within each section.

Approximately 20–25% of the 169 mdr1a knockout mice included in this part of the study appeared to develop loose stools and anal mucous discharge by 1 year of age (Table I). Histologic analysis of mice with clinical signs of colitis confirmed the presence of active intestinal inflammation that was primarily restricted to the large intestine. The difference in incidence between males (20%) and females (26%) was not significant. The average age at onset was approximately 20 wk, although the first signs of colitis development could be seen in mice ranging between 8 and 36 wk of age. Although some mice with colitis developed a wasting-type disease, the majority of colitic animals were maintained for up to 3 mo without signs of cachexia. Disease was nontransmissible, since mdr1a^−/− mice both with and without colitis were housed in the same cage. We observed no clinical or histologic signs of colitis in any FVB control mice, heterozygous mdr1a^+/− mice, or other immunocompetent mouse strains of any age housed within the facility.

The severe inflammation of the large intestine that develops in mdr1a^−/− mice bears resemblance to the human IBD, ulcerative colitis. Histopathologic analysis of multiple colitic mdr1a^−/− animals revealed inflammation spread along the entire length of the colon. The intestinal inflammation was characterized by a massive thickening of the mucosa and evidence of inflammatory cell infiltrates into the lamina propria in mdr1a^−/− mice with active colitis (Fig. 3,a) compared with mdr1a^−/− mice without colitis (Fig. 3,b). Occasional crypt abscesses and ulcerations extending through the mucosa to the muscular layer were also often noted in the colitic animals (Fig. 3,c). Crypt length was greatly increased in colitic animals, with approximately three- to fivefold more epithelial cells per crypt in colitic mice (Fig. 3,a) compared with noncolitic mice (Fig. 3,b). There also appeared to be some evidence of dysregulated epithelial cell growth in colitic mdr1a^−/− mice in areas of inflammation (Fig. 3 d).

Immunohistochemical analysis of the gastrointestinal tissues of animals with prevalent colitis or animals with early stages of colitis was used to further characterize the lymphocytic infiltrate into the mucosa of the large intestine compared with control FVB mice (Fig. 4). There was an increased number of infiltrating CD3⁺ T cells scattered diffusely throughout the lamina propria of inflamed colonic tissue from mice with active colitis (Fig. 4,a) compared with colonic tissue from control FVB mice (Fig. 4,c). A less dramatic CD3⁺ T cell infiltrate was observed in mice with early colitis (Fig. 4 b).

FACS analysis of the large-intestine IEL and LPL isolated from colitic animals revealed that the majority of the infiltrating T cells were CD4⁺TCRαβ⁺ (Table II). Although there was an apparent decrease in the relative proportion of CD8⁺ T cells in the lamina propria of colitic animals, the absolute number of CD8⁺TCRαβ⁺ T cells was not diminished, probably owing to the fourfold increase in the total number of LPL isolated from colitic animals (data not shown). There was no apparent increase in TCRγδ⁺ T cells in the inflamed tissue (data not shown).

In addition to T cell infiltrates, immunohistochemical analysis also revealed the presence of numerous B220⁺ B cell clusters/follicles within the lamina propria of animals with active colitis (Fig. 4,d) compared with control FVB animals (Fig. 4,f). Whereas the T cells were diffusely scattered throughout the lamina propria, the B cell infiltrates appeared to be localized to follicular clusters within the mucosa. The follicular clusters varied in size and were only observed in the colon. In contrast to animals with fully developed colitis, animals with intermediate colitis mostly had individual B220⁺ cells scattered throughout the lamina propria and the few clusters that were present were reduced in size (Fig. 4 e).

Gr1⁺ cellular infiltrates were also prominent throughout the lamina propria of the large intestine in mice with active colitis (Fig. 4,g) compared with control FVB mice, which appeared devoid of Gr1⁺ cells (Fig. 4,i). The Gr1⁺ infiltrates were uniformly present throughout the lamina propria, although an increased number of cells did appear to be migrating across the epithelium into the gut lumen in regions where the epithelialbarrier may have been compromised (Fig. 4,g). Mice with early colitis had relatively few Gr1⁺ infiltrating cells, and these were restricted to small focal regions in the lamina propria (Fig. 4 h).

It has been well established in a majority of mouse models that spontaneous colitis can be prevented if mice are housed under germfree conditions, suggesting that the presence of the gut flora is necessary for the development of colitis (4, 12, 14). It has also been previously demonstrated in another model of spontaneous colitis that treating mice with antibiotics can prevent inflammation (13). In the first part of the study, we treated mdr1a^−/− mice with broad-spectrum oral antibiotics to eliminate the commensal gut flora and then monitored for the development of colitis. Control mice were given water treated with saccharine alone. None of the mice had detectable colitis at the beginning of the study.

The incidence of colitis in mdr1a^−/− mice treated with antibiotics was greatly reduced compared with mdr1a^−/− mice treated with saccharine alone (Fig. 5). In the 16-wk treatment, 14 of 31 saccharine-treated mdr1a^−/− mice developed colitis (loose stools and mucous discharge), whereas only 1 of 39 antibiotic-treated mdr1a^−/− mice developed any signs of inflammation (no clinical signs, histopathologic evidence only). The one antibiotic-treated mdr1a^−/− mouse that showed signs of colitis had only one minor focal region of inflammatory cell accumulation (data not shown) and did not show any other overt signs of intestinal disease.

To ascertain whether the intestinal inflammation in mice with active colitis could be reversed by antibiotic treatment, mice with active colitis (as determined by evidence of loose stools and mucous discharge from the anus) were therapeutically treated with antibiotics in the drinking water. Within 3 wk of continuous antibiotic therapy, the previously colitic mice had no residual visible signs of anal discharge or loose stools. To determine whether the pathology associated with colitis could also be reversed along with the gross symptoms, mice treated with antibiotics for 10 wk were killed and the colons analyzed histopathologically. As expected, the colons of untreated colitic animals were grossly thickened (Fig. 6,a), compared with control animals (Fig. 6,b). Interestingly, the gross thickening of the colon associated with active IBD was significantly reduced in mice that had received therapeutic antibiotic treatment (Fig. 6,c). On further histopathologic examination of these intestines, we observed little evidence of inflammation, and the general gut architecture appeared similar to that of control mice (data not shown). Immunohistochemical analysis revealed that the mice treated with antibiotics had no evidence of persisting Gr1⁺ infiltrates or B cell follicles (data not shown). However, despite the apparent reduction in inflammation arising from antibiotic treatment, there still appeared to be an increased number of CD3⁺ T cells in the lamina propria of these animals (Fig. 6,d) compared with control mice (Fig. 4 c).

Although the commensal bacterial flora resident in the intestinal lumen of the gut appears to play a major role in the induction of colitis, little is known about the role of individual bacteria in the initiation and perpetuation of intestinal inflammation. In this study we analyzed the bacterial flora present in the cecum and colon of mdr1a^−/− mice with colitis, mdr1a^−/− mice with no evidence of active colitis, and control FVB mice. Table III lists the major bacterial species recovered from the animals. All of the mice appeared to have a gut bacterial profile devoid of any pathogenic organisms.

Recently, it has been shown that Helicobacter infection in immunocompromised mice can give rise to a pathology resembling colitis (39, 40, 41). To ensure that colitis in the mdr1a^−/− mice was not due to Helicobacter infection, samples from colitic, noncolitic, and control mice were analyzed by PCR for the presence of H. bilis, H. hepaticus, and Helicobacter spp. All mice were negative (data not shown).

Since bacterial colonization of the intestine appears to be crucial for the initiation of spontaneous colitis in mdr1a^−/− mice, we analyzed the in vitro proliferative response of cells isolated from the MLNs following stimulation with intestinal Ags (Fig. 7). Intestinal Ags were prepared from cecal contents obtained from FVB control mice, mdr1a^−/− mice with no overt signs of colitis, and mdr1a^−/− mice with active colitis. MLN were removed from the same animals, and cell suspensions were tested for proliferative reactivity against each of the intestinal bacterial Ag preparations.

We observed a significant increase in proliferation in response to bacterial Ags when MLN from mdr1a^−/− mice with active colitis were compared with those of other mice (p < 0.05). In contrast, there was no statistically significant difference in the response to bacterial Ag stimulation when MLN from FVB mice were compared with MLN from noncolitic mdr1a^−/− mice (p > 0.1). Reactivity profiles of each source of MLN were similar regardless of the source of bacterial Ags (data not shown). No significant differences were noted in the proliferative responses to Con A or anti-CD3 between any of the three groups of mice analyzed in these experiments (p > 0.1). The data shown are from one experiment and are representative of three similar experiments.

The increased frequency of B cell clusters in the lamina propria of colitic animals suggests that there may be a dysregulated humoral response in these animals. To test this, we measured the serum Ab titers of colitic, noncolitic, and control animals. Each of the Ig isotype levels was assessed by ELISA. All of the serum Ig isotype levels were elevated in colitic animals compared with control FVB mice or mdr1a^−/− mice without colitis, suggesting polyclonal activation of B cells associated with colitis. Serum IgA and IgG1 levels appeared most elevated in the mice with active colitis, although IgG2a, IgG2b, IgM, and IgE levels were also increased over control mice (Fig. 8).

Given that mdr1a is expressed on both lymphocytes and epithelial cells in the large intestine, we sought to determine whether the spontaneous development of colitis was the result of an mdr1a defect within the lymphoid or within the epithelial cell population. We generated irradiation bone marrow chimeras in which irradiated FVB animals were reconstituted with bone marrow from mdr1a^−/− or FVB donors, or irradiated mdr1a^−/− mice were reconstituted with bone marrow from FVB or mdr1a^−/− donors. As controls, additional groups of mice were generated in which FVB mice were reconstituted with FVB bone marrow and mdr1a^−/− mice were reconstituted with mdr1a^−/− bone marrow. The incidences of colitis in the groups are listed in Table IV. mdr1a^−/− mice that were reconstituted with wild-type bone marrow developed signs of colitis within 12 wk after reconstitution. After being maintained for 52 wk, 25% of the mice had developed active colitis. In contrast, two groups of FVB mice, which were reconstituted with mdr1a^−/− bone marrow, were maintained for up to 38 wk without any overt signs of colitis. As expected, the control FVB mice reconstituted with FVB bone marrow did not develop colitis. Surprisingly, 100% of the mdr^−/− mice reconstituted with mdr1a^−/− bone marrow developed colitis. The reason for the increased incidence in the latter group compared with unmanipulated mdr1a^−/− mice is unclear, although it is possible that radiation damage to the mucosa sufficiently enhanced susceptibility to developing colitis.

Mucosal immunology has advanced in recent years with the discovery of several different mouse models of IBD. Most of the mouse models described involve a disruption or modification of the immune system. Spontaneous colitis occurs in mice genetically deficient for TCRαβ⁺ cells (5) or the cytokines IL-2 (6) or IL-10 (4) and in mice deficient for the Gα2_i signaling protein (42). In addition, IBD occurs in human CD3ε-transgenic mice reconstituted with normal bone marrow (7), in HLA-B27 transgenic rats (43), and in immunodeficient SCID mice reconstituted with CD4⁺CD45RB^high cells (9, 12, 44). The great diversity of these models implies that there are complex mechanisms regulating the various components of the mucosal immune system that create and maintain a homeostatic balance. Disruption of this balance in any number of ways can obviously lead to the development of an adverse inflammatory response.

Here we describe a novel model of spontaneous colitis occurring in mdr1a^−/− mice that is characterized by a broad inflammatory response that appears restricted to the entire length of the colon. The histopathology of the inflammation is most consistent with features associated with ulcerative colitis in humans and is characterized by dysregulated epithelial cell growth, ulcerations, crypt abscesses, and lymphocytic infiltration into the mucosa. Given the large number and diverse nature of the IBD models, it is not surprising that there is some controversy regarding the precise nature of the cells responsible for initiating and perpetuating IBD. The diffuse lymphocytic infiltrates in the mdr1a^−/− mice are primarily CD4⁺TCRαβ⁺ T cells and Gr1⁺ cells. In addition, follicular B cell clusters are prominent throughout the mucosa.

mdr1a^−/− mice require a commensal bacterial flora for the development of colitis. Colitis can be prevented in mdr1a^−/− animals by oral treatment with broad-spectrum antibiotics to eliminate intestinal bacteria. The requirement for bacterial flora in the induction of colitis in mdr1a^−/− mice is consistent with many of the other IBD models. IBD does not develop in the TCRα or IL-10 knockout animals maintained under germfree conditions (4, 14), and IBD can be prevented in the CD4⁺CD45RB^high SCID transfer model if the animals have a reduced number of enteric flora or are treated with broad-spectrum antibiotics (12, 13). In addition to preventing the initiation of IBD, we show here that antibiotic treatment can therapeutically reverse colitis in mdr1a^−/− mice with active inflammation. This indicates that intestinal flora is not only necessary for initiating inflammation but is also required for sustaining the inflammatory response. Interestingly, despite the reversal of inflammation, a large number of CD3⁺ T cells remain present in the mucosa. The functional state of these T cells is unclear.

In the course of this study, we observed a relatively high incidence of colitis in mdr1a^−/− mice treated with saccharine compared with unmanipulated knockout mice in the colony. We have no definitive explanation for this, since there is no a priori reason why saccharine should make animals more susceptible to developing colitis. However, it is possible that changes do occur in the commensal flora when saccharine is included in the drinking water, perhaps permitting outgrowth of certain bacterial species. Despite the increased incidence of colitis in the saccharine-treated mice, the addition of antibiotics was sufficient to eliminate the problem and prevent both the development of colitis and reverse ongoing intestinal inflammation.

It is possible that an individual bacterial species that colonizes the intestine might be responsible for initiating inflammation in these murine IBD models. To determine whether there might be a specific bacterial pathogen associated with disease in the mdr1a^−/− mice, we analyzed the bacterial composition of the resident cecal and colonic flora in colitic and noncolitic mdr1a^−/− mice and control FVB mice. We were not able to identify a single bacterial species that could be solely responsible for triggering intestinal disease. There did appear to be variations in the relative abundance of certain bacterial species in the different samples analyzed, and while it remains possible that an excess of a singular species might be responsible for the initiation of colitis, this may simply reflect a selective advantage for certain bacteria in the local microenvironment during intestinal inflammation.

Another potential explanation for the requirement of intestinal flora in the induction of IBD is that cross-reactive epitopes may exist between intestinal flora and enterocytes, such that the enterocytes become the target of an autoimmune inflammatory response. In support of this, IBD in TCRα knockout mice is accompanied by the appearance of autoAbs, including those directed against epithelial cell proteins (8, 45). mdr1a^−/− mice with active colitis have large B cell follicles in the lamina propria. These mice also have a massive increase in serum Ab titers of all isotypes, suggesting a polyclonal activation of B cells; however, it is unclear whether the Ab response is directed against intestinal tissues. We are currently breeding the mdr1a^−/− mice with B cell-deficient mice to determine whether B cell activation is a critical component in the initiation of colitis.

Since data from many of the other IBD models indicate that T cells are a major component in the induction of IBD, and much of the intestinal damage in IBD appears to be the result of T cell-mediated injury (46, 47), we decided to analyze the functional activity of T cells isolated from mdr1a^−/− mice with active colitis. Lymphocytes isolated from animals with active inflammation appeared to be significantly more responsive to intestinal bacterial Ags compared with control animals. Lymphocytes isolated from mdr1a^−/− mice without active inflammation had a similar proliferative reactivity compared with control FVB animals. This indicates that the mdr1a defect does not inherently cause hyperresponsiveness to bacterial Ags. Rather, it appears that lymphocytes become hyperresponsive to bacterial stimulation only after the onset of inflammation. The finding that bacterial Ags prepared from animals with colitis and animals without colitis were equally potent in stimulating proliferation further suggests that colitis is probably not caused by a single pathogenic organism.

A central question in the mdr1a^−/− mouse IBD model is, where does the defect that leads to colitis reside? mdr genes are expressed on a number of immune and nonimmune cell types in mice and humans (19, 20, 21, 23). Thus, it is possible that mdr1a may have several different functional activities depending on the site of expression. We show here that mdr1a is expressed on murine peripheral and mucosal T lymphocytes including all TCRγδ⁺ and CD8⁺ cells and a proportion of TCRαβ⁺ and CD4⁺ cells. Despite the lack of mdr1a expression on immune cells in the mdr1a^−/− mice, the immune system does not appear to have any gross phenotypic or functional defects. It is known that mdr1a is also expressed on the apical surface of colonic epithelial cells, where it has been shown to actively pump orally administered toxins into the lumen of the gut (48). There is also evidence that mdr1a may play a role in Cl⁻ channel regulation in epithelial cells (49, 50, 51). Thus, we thought it possible that mdr1a^−/− mice have functional defects within the intestinal epithelial barrier that might ultimately lead to colitis. To assess whether immune or epithelial cell defects were responsible for the susceptibility to colitis in mdr1a^−/− mice, we generated a series of chimeric mice. Irradiated mdr1a^−/− mice reconstituted with wild-type bone marrow spontaneously developed colitis. In contrast, irradiated wild-type mice reconstituted with mdr1a-deficient bone marrow cells did not develop colitis. These data clearly suggest that the defect that leads to the initiation of colitis in mdr1a-deficient mice is concordant with defective mdr1a expression on a radiation-resistant element within the host, most likely the epithelial cells lining the gut. These data also suggest that the colitis in mdr1a^−/− mice is not likely to be associated with immune cell deficiencies.

The precise mechanistic defect causing colitis in the mdr1a^−/− mice is unclear. One possibility is that bacterial products/toxins become sequestrated within epithelial cells, since small amphiphilic, hydrophobic molecules are less likely to be secreted in the absence of a functional mdr pump. It is not unreasonable to assume that a buildup of bacterial products within mdr1a-defective epithelial cells could subsequently initiate an inflammatory response, either by direct toxicity to the epithelial cell or indirectly by altering the nature of the Ag-presenting capability of the epithelial cell. The demonstrated requirement for bacterial flora in the initiation and progression of the inflammation, and the association of disease with defective mdr1a expression in epithelial cells, suggests that one of these scenarios might be true. We are currently attempting to ascertain the precise function of mdr1a in colonic epithelial cells.

With all of the different immune models of colitis, it has become apparent that the intestinal microenvironment is an exquisitely regulated system and that perturbation of individual components within this environment can lead to inflammation. The model described here has many similarities with existing murine models of spontaneous colitis, including an inflamed pathology restricted to the colon and a dependence on intestinal flora. In contrast to most existing IBD models, the immune system of mdr1a^−/− mice does not have any gross abnormalities. Furthermore, this model reveals that disruption of a nonimmune component of the mucosal environment can also be responsible for the initiation of colitis. The mdr1a knockout mouse model of spontaneous colitis should provide us with insight into the role of the epithelial barrier in preventing aberrant immune responses to commensal bacterial, and possibly food, Ags. Moreover, it predicts a possible in vivo functional role for mdr1a in the maintenance of the intestinal epithelial barrier.

We thank Neil Fanger and Ken Schooley for assistance with the immunofluorescence analysis of frozen sections; Jamie O’Malley for technical help; Richard Aranda, Scott Binder, and JoAnn Schuh for histopathologic analysis of gut sections; and Anne Bannister for editorial assistance.