T Cell–mediated Pathology in Two Models of Experimental Colitis Depends Predominantly on the Interleukin 12/Signal Transducer and Activator of Transcription (Stat)-4 Pathway, but Is Not Conditional on Interferon γ Expression by T Cells

Donor (C57BL/6 × CBA/J)F1, C57BL/6, C57BL/6/ Scid, and C57BL/6/IFN-γ^null mice were purchased from The Jackson Laboratory (Bar Harbor, ME). The Stat-4^null mice were a gift from Dr. J.N. Ihle (St. Jude's Children's Hospital, Memphis, TN). 129/SvEv RAG-2^null and wt mice were purchased from Taconic Farms (Germantown, NY). The Tgε26 recipient mice were generated as previously described by transgenic overexpression of a full-length human CD3ε gene (32) and bred on the original C57BL/6 × CBAJ background in the Beth Israel Deaconess Medical Center animal facility. All mice were kept under standard conditions in microisolator cages with autoclaved food, water, and bedding. In RAG^null and Scid reconstitution experiments, recipients were between 4 and 6 wk of age. Tgε26 recipient mice were between 5 and 10 wk of age.

The CD45RB^hi model was generated as described by Powrie et al. with minor modifications (37). CD4⁺ T cells obtained from the spleens of donor animals and were enriched by magnetic sorting. For cell purification, the following biotinylated anti–mouse antibodies were used to label non-CD4⁺ T cells isolated from the spleens of donor mice: B220 (RA3-6B2), Mac-1 (M1/70), Gr-1(RB6-8C5), and CD8α (53-6.7). Magnetically labeled streptavidin beads (Miltenyi Biotec Inc., Sunnyvale, CA) were used to bind the biotinylated antibodies. All antibodies were obtained from PharMingen (San Diego, CA). Negative selection was accomplished using a MACS magnetic cell sorter (Miltenyi Biotec Inc.). The enriched CD4⁺ cells were then labeled for cell sorting with FITC-conjugated CD45RB (16A) and PE-conjugated CD4 (Becton Dickinson, Mountain View, CA). Subsequently, cells were sorted under sterile conditions by flow cytometry for CD4⁺ CD45RB^hi cells, using criteria similar to that described in previous studies (37). This resulted in a >98% pure population of T cells. Harvested cells were resuspended at 10⁶/400 ml PBS (no FCS) and injected into the tail veins of recipient mice.

For the BM→ Tgε26 model, bone marrow was prepared as previously described (34). In brief, BMCs were depleted of T cells using two rounds of treatment with anti-Thy1.2 antibody (30-H12) followed by rabbit complement lysis (Cedarlane Labs, Westbury, NY). This procedure resulted in <0.1% CD4⁺ and CD8α⁺ cells within the inoculum, as determined by flow cytometry. Bone marrow recipients were total body irradiated with a sublethal dose (400 rads) 6–8 h before transplant or were treated with 5-Fluorouracil (75 mg/kg) 48 h before transplantation. 5 × 10⁶–10⁷ cells were resuspended in 400 ml PBS (no FCS) and transferred by tail vein injection.

To reduce the possibility of graft versus host disease (GVHD) in the CD45RB^hi model, cell transfers were made, where possible, between mice of the same backgrounds. Stat-4^null mice were of the mixed background C57BL/6 × 129/SvEv, making responses against segregated minor antigens from the 129/SvEv donor background a possibility. However, no evidence of GVHD, such as skin inflammation, was observed in any of the animals studied. In the Tgε26 bone marrow transfer model, significant differences were present between donor and recipient mice in some cases. However, preconditioning of bone marrow was effective in depleting the BMCs of T cells (see above). Inflammation of the skin and small bowel were not apparent in the mice included in this study.

The anti–IL-12 antibody (clone C 17.8; reference 16) was a gift from Dr. T. Veldman (Genetics Institute, Cambridge, MA). 260 μg of purified anti–IL-12 antibody (or vehicle control [PBS]) was injected intraperitoneally into mice on a weekly basis starting at the time of BMC transfer.

Mice were weighed twice a week and monitored for appearance and signs of loose stool and diarrhea. In the BM→ Tgε26 model, the level of wasting was determined by percentage of loss of weight from the starting weight, measured at the time of transplant. Weights were subsequently ranked 1 point for every 10% of body weight (or part thereof) lost: 0 = no weight loss; 1 = 1–10% weight loss; 2 = 11–20% weight loss; and 3 = ⩾21% weight loss. At necropsy animals were assessed for the level of colitis, which they displayed using two parameters: colon thickening and stool consistency, which were subsequently ranked as follows: for colon enlargement, 0 = no colon thickening; 1 = mild thickening; 2 = moderate thickening; 3 = extensive thickening; and for stool consistency, 0 = normal beaded stool; 1 = soft stool; 2 = diarrhea; and an additional point was added if gross blood was noted. The total score given for disease was the combined scores for colon thickening, stool consistency, and weight loss, divided by three.

Colon tissue samples were fixed in 10% buffered formalin and embedded in paraffin. Sections (4 μm) were cut and placed on gelatin-coated microscope slides for staining with hematoxylin and eosin using standard techniques as previously described (34). The severity of colitis was determined based on histological examination of the distal colon, whereby the extent of colonic inflammation was graded on a scale of 0–3 in each of three criteria: cell infiltration, thickening of the bowel wall, and the number of crypt abscesses. Histological grades were assigned in a blinded fashion by the same pathologist (A.K. Bhan).

Mesenteric lymphocytes were harvested and single cell suspensions prepared in PBS with 2% FCS (Intergen, Purchase, NY). Viable cells were counted and determined by trypan blue exclusion. Surface and cytoplasmic staining and FACS^® (Becton Dickinson, Indianapolis, IN) analysis were performed as previously described (40). All antibodies were purchased from PharMingen and were conjugated to PE, FITC, or biotin. In the case of biotin-conjugated antibodies, labeling was accomplished by secondary staining with streptavidin-RED670 (GIBCO BRL). Cytoplasmic staining for IFN-γ (clone XMG1.2) and TNF-α (clone MP6-XT22) was done as previously described (40). In brief, freshly isolated cells were stimulated overnight with plate-bound anti-CD3 (clone 145-2C11). Brefeldin A was added for the final 2 h of incubation at a final concentration of 10 mg/ ml. Cells were stained with anti-CD4 (clone RM4-5) and anti– CD3-biotin (145-2C11), washed, and fixed with 2% paraformaldehyde (Sigma Chemical Co., St. Louis, MO). Cells were permeabilized with Saponin (0.2% in PBS) to facilitate cytoplasmic cytokine staining. Cells were analyzed by flow cytometry using a FACscan^® flow cytometer in conjunction with FACScan^® software (Becton Dickinson) and CellQuest Software (Becton Dickinson).

To examine the role of IL-12 in the BM→ Tgε26 colitis model, Tgε26 mice were transplanted with BMCs from (C57BL/6 × CBA)F1 donors as previously described (34) and given a weekly dose of anti–IL-12 antibody, or vehicle control (PBS). The requirement for Stat-4 expression in T cells was tested by comparing disease in Tgε26 mice transplanted with BMCs from Stat-4^null mice or wt control animals of the same genetic background (129/SvEv). Using a similar approach in the CD45RB^hi model, we used 129/ SvEv RAG^null mice, rather than CB.17/SCID mice, as recipients, allowing us to maintain an equivalent genetic background to the Stat-4^null donor animals (see Materials and Methods). RAG^null mice were thus reconstituted with FACS^®-sorted splenic CD4⁺ CD45RB^hi T cells from Stat-4^null mice (Stat-4^null→ RAG^null). Control RAG^null mice in these experiments were transplanted with wt 129/SvEv CD4⁺ CD45RB^hi T cells (129/wt→ RAG^null).

Treatment of BM→ Tgε26 mice with anti–IL-12 antibody resulted in a significant reduction in disease compared with control (untreated) mice. A gross clinical disease activity score, which measures stool consistency, gross colon enlargement, and weight loss, is shown in Fig. 1. The difference in the extent of disease was most apparent between 4 and 5 wk after BMC transfer, where control animals showed signs of severe cachexia, and in many cases diarrhea which was occasionally accompanied by visible blood. By comparison, anti–IL-12–treated animals showed reduced wasting and generally did not develop diarrhea. At necropsy, the colons of treated animals were also visibly less thickened than those of control mice. Similarly, the majority of Tgε26 reconstituted with Stat-4^null BMCs showed milder disease compared with control animals that had received wt BMC (Fig. 1). 129/wt→ RAG^null mice developed wasting and colitis 7–10 wk after cell transfer. In contrast, Stat^null→ RAG^null retained a healthy appearance over the same period of time and showed mild colitis at necroscopy, similar to that seen in RAG^null mice that had received sorted CD4⁺ CD45RB^lo T cells (36, 37).

The reduced severity of disease found in each experimental group of mice was reflected both in the extent of loss of body mass (Fig. 2,A) and for the most part in the level of mucosal inflammation in the large bowel, determined by examination of histological tissue sections (Fig. 2,B). Thus, the extent of mucosal inflammation in all anti–IL-12–treated F1→ Tgε26 mice, and the majority of Stat-4^null reconstituted mice, was lower than in control animals. Nevertheless, significant histological colitis could be detected in a proportion of Stat-4^null recipient animals, even in cases where severe gross colitis (extensive colon thickening and diarrhea) was not apparent. In Stat-4^null→ RAG^null experiments, a large majority of animals showed considerably lower grades of colitis than did control mice. Representative histological sections from colons of CD45RB^hi→ RAG^null and BM→ Tgε26 mice from different groups are shown in Fig. 3, illustrating the principal differences between mild and severe forms of colitis in each model. The limited extent of mucosal inflammation in most of the Stat-4^null-recipients and in those subjected to anti–IL-12 antibody treatment was evident from the nominal level of cellular infiltration, crypt cell hyperplasia, and the paucity of crypt cell abscesses formation. However, it was noted, in the colons of some Stat-4^null→ RAG^null mice, that even in the absence of continuous mucosal inflammation there was a moderate frequency of focal inflammatory involvement. In some cases this was accompanied by the appearance of granulomas, a characteristic feature of the CD45RB^hi transfer model (Fig. 3 D). Collectively, the use of anti–IL-12 antibodies and Stat-4^null mutant mice as donors revealed a predominant role for IL-12 and Stat-4 proteins in the development of both the wasting syndrome and colon inflammation in two distinct models of IBD.

To examine whether disease correlated directly with the level of Th1 involvement in the two colitis models, we examined the development of Th1-type T cells in each system. Since previous studies have shown a direct effect of IL-12 and Stat-4 on the production of IFN-γ by T cells, we used cytoplasmic staining of IFN-γ and FACS^® analysis to determine the frequencies of T cells capable of IFN-γ expression. As shown in Fig. 4, the frequency of IFN-γ–producing CD4⁺ T cells in the mesenteric lymph nodes (MLNs) in anti–IL-12–treated F1→ Tgε26 mice and in Stat-4^null→ Tgε26 and Stat-4^null→ RAG^null mice was 60– 80% lower than in control animals. Additionally, numbers of colon lamina propria CD4⁺ T cells capable of IFN-γ production were also markedly reduced in protected animals (data not shown). Previous reports have suggested that in the absence of Stat-4 signaling T cells develop a Th2-like phenotype (29). Examination of IL-4 and IL-10 expression by MLN T cells in Stat-4^null recipients revealed no detectable expression of either cytokine (data not shown).

Together, the above data implicate a causal relationship between the numbers of T cells capable of IFN-γ production and the extent of wasting and colitis. We reasoned that if induction of IFN-γ gene expression did in fact represent a critical step in the development of pathogenic Th1 cells, then the ablation of this, specifically in T cells, should be sufficient to prevent disease. To test this, we reconstituted the T cell compartment of Tgε26 and Scid or RAG^null mice using IFN-γ–deficient donors (IFN-γ^null mice; reference 41).

In the BM→ Tgε26 model, animals injected with BMCs from IFN-γ^null donors backcrossed to the C57BL/6 background were compared with controls injected with C57BL/ 6/wt BMCs. In the CD45RB^hi transfer model, CD45RB^hi CD4⁺ T cells from IFN-γ^null mice were transferred into C57BL/6/Scid or RAG^null mice. Scid and RAG^null mice transplanted with C57BL/6/wt CD4⁺ CD45RB^hi T cells served as controls in these experiments. Fig. 5 shows the cumulative results from these experiments, which compare the mean disease activity scores from each group. In both BM→ Tgε26 and CD45RB^hi→ Scid RAG^null animals little difference was apparent between the overall gross disease in each group of mice (Fig. 5). However, in some cases, weight loss and/or histological colitis were milder in IFN-γ^null–recipient mice (Fig. 6, A and B). A large number of IFN-γ^null→ Tgε26 mice showed a range of histological scores and, in some cases, despite showing clinical evidence of significant colitis, revealed comparatively mild inflammation of the mucosa by histological examination (Fig. 6,B). In CD45RB^hi→ Scid/ RAG^null experiments, most animals showed visible signs of wasting. However, due to their young age at the time of cell transfer, many of these mice increased in body mass due to compensatory growth. The extent of colon involvement observed in the large majority of CD45RB^hi→ Scid/ RAG^null mice was equivalent whether or not animals had received IFN-γ^null or C57BL/6/wt T cells. Representative histological sections from mice that developed colitis in the presence of IFN-γ^null T cells are shown in Fig. 7. In each case the salient features of pathology were similar to those in animals reconstituted with wt T cells.

Finally, we observed that large numbers of T cells from IFN-γ–reconstituted animals expressed a pattern of cell surface markers consistent with an activation or memory status, including low CD45RB and L-selectin expression and elevated CD69 expression (data not shown). To examine whether, in the absence of IFN-γ expression, these T cells retained other characteristics consistent with a pathogenic phenotype, their ability to produce TNF-α was tested. As shown in Fig. 8, equivalent frequencies of CD4⁺ T cells from IFN-γ^null and wt recipients in each model expressed TNF-α. Taken together, these data demonstrate that without producing IFN-γ themselves, T cells in each of the IBD models were nevertheless capable of mediating significant levels of wasting and colitis that correlated with the retained capacity of these cells to produce a proinflammatory cytokine.

In our study we used a novel experimental approach to examine the mechanisms leading to development and action of pathogenic Th1-type T cells in colitis. The significant effect of Stat-4 deficiency in limiting progression of disease in both the BM→ Tgε26 and the CD45RB^hi transfer systems establishes a principal role for this pathway in the development of pathogenic Th1-type T cells in colitis. Furthermore, the similar attenuation of both disease and T cell–specific IFN-γ expression seen in anti–IL-12–treated BM→ Tgε26 mice and Stat-4^null→ Tgε26 mice underscores the link between the actions of IL-12 and Stat-4. In the two colitis models studied here, inhibition of IL-12 and/or Stat-4 signaling affected two divisible aspects of disease, colitis and wasting, implicating involvement of Th1 cells in both components of pathology.

To an extent, our experimental results appear consistent with those obtained in the more acute TNBS-induced model of colitis (21). However, we observed that although disease was for the most part attenuated in the absence of IL-12/Stat-4 signaling, many animals nevertheless revealed evidence of pathology. Furthermore, this correlated with the continued (although strongly reduced) presence of T cells capable of IFN-γ expression. These data differ from those in the TNBS colitis model, where an almost complete inhibition of both disease and IFN-γ production were observed after anti–IL-12 treatment. These differences make apparent that the mechanisms of pathology in colitis models studied here and that induced by TNBS are not equivalent, and they suggest that factors other than IL-12 contribute to the development of more chronic forms of intestinal inflammation. In this capacity, the recently described IFN-γ–inducing factor (IL-18) is a potential candidate, since this cytokine has been shown to have a significant effect on the long-term ability of T cells to produce IFN-γ (42–44). By inference it is likely that this could significantly affect the development of pathogenic Th1-like T cells.

Notwithstanding the above arguments, the results from our study suggest a degree of independence between the actions of IL-12 and/or other Th1-promoting cytokines and the specific expression of IFN-γ by T cells. Thus, although the absence of IFN-γ expression by T cells altered the severity of colitis in many cases, the overall effects on disease were less pronounced than might have been anticipated. On first assessment this is surprising, principally for three reasons. First, the dependence of IFN-γ expression on IL-12 signaling suggests that the effects of IL-12 (via Stat-4) result from its ability to induce IFN-γ transcription in T cells (22). Second, IFN-γ has been shown to be required for efficient IL-12 priming of Th1 cells, presumably through autocrine expression (45–47). Principally, this appears to be achieved through the upregulation of the IL-12 receptor (48–49). Third, a general consensus exists that the well-documented proinflammatory effects of IFN-γ are central to pathology in IBD (4, 7, 13). The validity of this last assumption in experimental colitis has been strengthened by a previous study by Powrie et al., who demonstrated that anti–IFN-γ treatment was effective in reducing disease in the CD4⁺ CD45RB^hi transfer model (38). It is possible that in this latter case anti–IFN-γ antibody treatment might block all available IFN-γ. By contrast, in our experiments non-T cells, resident in the host animals, might remain a sufficient source of IFN-γ to promote effects on both T cells and macrophages. However, the development of colitis in the absence of IFN-γ expression is not unprecedented, since IL-2^null mice, which succumb to a Th1-type colitis similar to that seen in BM→ Tgε26 mice, also develop disease after import of the IFN-γ^null mutation (Zand, M., C. Stevens, and T. Strom, personal communication). In addition, a recent study by Berg et al. revealed that antibody neutralization of IFN-γ in IL-10^null mice was sufficient to prevent disease only when administered to animals at 3 wk of age and not in animals aged 3 mo or older (50). Thus, although IFN-γ was required to establish colitis in IL-10^null mice, other inflammatory mediators were clearly sufficient to mediate disease once a pathogenic T cell phenotype had been established.

In two autoimmune models classically associated with Th1-type T cell responses, it has been observed that Th2 cells are also capable of mediating disease (51, 52). These studies support the contention that, in some cases, immunopathology normally associated with Th1-type responses can also be attributable to cells that have undergone immune deviation, leading to the expression of Th2-associated cytokines. However, in our studies we were unable to detect any IL-4 or IL-10 expression in the absence of either Stat-4^null or IFN-γ expression by T cells. Although these findings do not unequivocally demonstrate lack of immune deviation, they do argue that this is not predominant in the colitis models studied here.

Collectively, the results offered here lend credence to the argument that Th1-type cells develop pathogenicity via a complex range of mechanisms, not all of which fall under the influence of IL-12 or their ability to produce IFN-γ. Ultimately, this is most likely achieved by the concerted expression of a range of cytokines and cytotoxic molecules which could act directly or indirectly to promote inflammation. TNF-α, which is produced by T cells as well as macrophages, is a cytokine that encompasses these characteristics and is known to induce potent inflammatory effects. Consistent with this, we have shown that in colitic mice T cells expressed TNF-α at a similar frequency whether IFN-γ was expressed by the same cells or not. Furthermore, we have found that anti–TNF-α antibody treatment is highly effective in reducing disease in the BM→ Tgε26 model.² Similarly, Powrie et al. demonstrated an effect of anti–TNF-α treatment in the CD45RB^hi → Scid model (38). These data are consolidated by the finding that anti–TNF-α antibody therapy has been shown to be efficacious in treating Crohn's disease (53).

In summary, our study offers insights into the requirements for both the development and pathogenic activity of Th1-type cells in two distinct models of inflammatory colitis. The observations that IL-12 is upregulated in human IBD (7) and that interruption of the TNF-α pathway inhibits Crohn's disease correlate well with our observations in the two models studied here (38).² Further dissection of the pathways which lead to aberrant Th1 pathology in animal models of IBD will undoubtedly provide a useful source of information for future therapy in IBD.

We thank Dr. James Ihle for providing us with the Stat-4^null mice and Dr. T. Veldman for providing anti– IL-12 antibody.

This work was supported by grants from the Crohn's and Colitis Foundation of America (C. Terhorst) and from the National Institutes of Health (P30 DK-43551 and RO1 DK-47677 to A.K. Bhan). Samir A. Shah is a Howard Hughes Medical Institute Physician Postdoctoral Fellow. Stephen J. Simpson was supported by a Research Fellowship Award from the Crohn's and Colitis Foundation of America.