XI. Novel mouse models to study pathogenic mechanisms of Crohn's disease

Theresa T. Pizarro, Kristen O. Arseneau, Fabio Cominelli

Abstract

Crohn's Disease (CD) affects more than 500,000 individuals in the United States and represents the second most common chronic inflammatory disorder after rheumatoid arthritis. Although major advances have been made in defining the basic mechanisms underlying chronic intestinal inflammation, the precise etiopathogenesis of CD remains unknown. We have recently characterized two novel mouse models of enteritis that express a CD-like phenotype, namely the TNF ΔARE model of tumor necrosis factor (TNF) overexpression and the SAMP1/Yit model of spontaneous ileitis. The unique feature of these models is that they closely resemble CD for location and histopathology. These genetically manipulated new models of intestinal inflammation offer a powerful tool to investigate potential causes of human disease and may allow the development of novel disease-modifying therapeutic modalities for the treatment of CD.

  • ileitis
  • cytokines
  • inflammation

crohn's disease (CD), one of the chronic inflammatory bowel diseases (IBD), is a debilitating disorder of unknown etiology whose incidence is on the rise worldwide (21). Although it is well recognized that the disease may involve genetic as well as environmental factors, the precise initiating events leading to CD are unknown. Therefore, therapeutic approaches for the treatment of CD are aimed at alleviating symptoms, are often nonspecific, and do not affect the natural history of the disease. One approach to investigating the pathogenic mechanisms involved in a complex disease such as CD is the use of appropriate animal models. Such an approach offers several advantages compared with direct clinical investigation of the affected patient population. For example, the time course of disease phenotype as well as specific pathophysiological events occurring before disease onset can be studied. In addition, modern technology has made it possible to perform genetic and immunologic manipulations of relevant genes that may be involved in the pathogenesis of the human disease (2). The relative contribution of bacterial and dietary antigens can also be assessed using germ-free animals, monoassociation studies with specific bacteria, as well as defined dietary regimens. Finally, specifically targeted interventions as well as experimental treatment modalities can be successfully tested in the appropriate animal model system. Ideally, an appropriate animal model for investigating IBD, particularly CD, should display an intestinal phenotype that develops spontaneously without gene targeting or immunologic manipulations (1) and should closely resemble the specific human disease. The present article briefly reviews currently available animal models of IBD and discusses in detail two novel genetically engineered murine models of experimental CD.

OVERVIEW OF ANIMAL MODELS OF IBD

Several animal models of intestinal inflammation/IBD have been developed and extensively studied (3). These models are based on chemical induction, gene targeting, or immune-cell transfer manipulations that all result in gut inflammation. Their characteristics are described in Table 1. One limiting factor with such models is that, with few exceptions, they do not possess the hallmark features of IBD in regards to disease location, histological features, or clinical course of the human condition. However, these models may prove useful in characterizing a specific biological mechanism of gut inflammation. Examples include the characterization of the interleukin (IL)-1/IL-1 receptor antagonist balance in the rabbit formalin-immune complex colitis model (5), the role of protective factors in response to intestinal epithelial injury in the dextran sodium sulfate (DSS) colitis model (16), the key function of regulatory T cells in the CD45RBhigh adoptive transfer model (6), and the importance of neurogenic gut inflammation in the dinitrobenzenesulfonic acid (DNBS) colitis model (23). These models of intestinal inflammation/IBD, however, may have little value in defining the precise etiopathogenesis of the human disease.

View this table:
Table 1.

Characteristics of animal models of intestinal inflammation/IBD

Chemically induced models.

This group of animal models requires the administration of an exogenous chemical agent for the induction of colitis. With few exceptions, the resulting intestinal inflammation is acute and persists as long as the chemical agent is present in the gut lumen. Examples include the acetic acid model of colitis (14), the 2,4,6-trinitrobenzenesulfonic acid (TNBS) model of colitis (19), the DSS model of colitis, and the formalin-immune complex colitis model (7). These models have each generated important information on specific biochemical pathways of inflammation and have provided proof of concept for specific therapeutic interventions for the treatment of IBD. However, they are primarily characterized by acute mucosal injury (29) and clearly differ from human IBD in many aspects, including the initiating event and the clinical course of gut inflammation.

Immunologic models.

Immunologically mediated models are defined as models of adoptively transferred T cells or bone marrow precursors, which are introduced into immunodeficient recipients, usually mice. The two best examples are the CD45RBhigh transfer model (20) and the bone marrow chimera transfer model (8). This group of animal models has provided very important information on the role of regulatory T cells in controlling mucosal immunity and intestinal inflammation and offers strong evidence that Th1 polarization may play a key role in the pathogenesis of CD (22). However, the profound immunological abnormality in the recipient mice, who lack a competent immune system, make these models unlikely to be suitable for investigating the precise factors involved in human CD.

Genetic models.

The advent of transgenic and knockout methodologies have revolutionized the field of animal models of IBD after the discovery that a variety of genetically manipulated mouse models develop chronic intestinal inflammation. Examples include the IL-2 (26), T cell receptor (TCR)α/β (18), IL-10 (13), and Gi2α (25) knockout models. With the exception of IL-10-deficient mice, which possess some features of human CD (Table 2), the majority of these models display features more similar to ulcerative colitis than CD. These models have contributed greatly to our current understanding of the role of key immune-related molecules and genes, including specific cytokines, in the pathogenesis of chronic intestinal inflammation. Collectively, these models have clearly demonstrated the requirement for strict regulation of the mucosal immune response and have allowed identification of key components involved in gut immune regulation. However, it is unlikely that these mutations represent the underlying defect in human CD. Therefore, the use of these models is limited, particularly when they are directed at investigating causal factors of human CD.

View this table:
Table 2.

Characteristics of mouse models of human Crohn's disease

Spontaneous models.

This group represents the most attractive model system of intestinal inflammation/IBD because gut disease occurs without any apparent exogenous manipulations, similar to human disease. Two animal models have been investigated and characterized in this group. The first is represented by the cotton-top tamarin model, which shows features similar to human ulcerative colitis in regards to clinical course, location of disease, histological features, increased risk of colon cancer, and response to conventional treatments (15). The second model is the C3H/HeJBir model of colitis characterized by a spontaneous and chronic focal inflammation localized to the right colon and cecal region (27). Interestingly, the colitis occurs in young mice and tends to resolve with age, without recurrence. Similar to severely immunocompromised mice, as well as some of the knockout mice, there appears to be a correlation between Helicobacter infection and the onset of colitis in the C3H/HeJBir model. However, both the C3H/HeJBir and cotton-top tamarin models of gut inflammation have similarities to human ulcerative colitis and do not involve the small intestine. Until recently, no models of spontaneous ileitis were reported.

NOVEL MOUSE MODELS OF CROHN'S DISEASE

In the last three years, our group has characterized two new murine models of experimental CD that closely resemble many of the characteristics of the human disease. These models provide an exciting opportunity to attempt to define the precise etiology of CD and to begin to develop a cure for this devastating disease.

TNF ΔARE model of ileitis.

A large body of evidence supports the concept that tumor necrosis factor (TNF) may play a key role in the pathogenesis of chronic inflammatory diseases, including IBD (9). The most compelling evidence for the importance of TNF in mediating human CD is the ability of drugs that modulate TNF production to markedly decrease symptoms and severity of inflammatory lesions in patients with active CD (28). On the basis of these observations, we recently reported (10) a unique model of TNF overexpression that develops an impressive CD-like phenotype, which is localized primarily to the small intestine and is characterized by chronic transmural ileitis with features very similar to human CD (Table 2).

TNF mRNA contains a repeated octanucleotide AU-rich motif in its 3′-untranslated region that has been implicated in the posttranscriptional and translational regulation of TNF synthesis. Mice containing an endogenous 69-bp deletion of the 3′-AU-rich region in the TNF gene were generated by homologous recombination in embryonic stem cells and Cre-LoxP-mediated excision of the inserted Neogene. The resulting TNFΔARE [change in TNF AU-rich elements (ARE)] mutant mice possess high circulating levels of TNF protein that can increase up to threefold after lipopolysaccharide administration. In addition, peritoneal macrophages, as well as other hemopoietically derived cells isolated from TNFΔARE mice, show increased TNF mRNA stability that correlates with the enhanced production of TNF protein. Together, these data demonstrate an overall suppressive role of the ARE on TNF production that results in increased constitutive and inducible production of TNF in these animals.

Mutant mice carrying the deletion in a homozygous state (TNFΔARE/ΔARE), compared with heterozygous TNFΔARE/+ and normal littermate controls, display a severe wasting syndrome and increased mortality, succumbing to disease between 4 and 12 wk of age. Phenotypically, TNFΔARE/ΔARE mice are runted, possess a ruffled coat with alopecia-like lesions, often have a cataract appearance in one or both eyes, and develop a chronic inflammatory arthritis with swelling and distortion of paw joints leading to severe impairment in mobility. Systematic histological analysis of the gastrointestinal tract of TNFΔARE mice has revealed profound inflammatory changes consistent with a CD-like phenotype. These alterations are first detected between 2 and 4 wk of age and are localized primarily to the terminal ileum, and occasionally to the proximal colon, of both TNFΔARE/ΔARE and TNFΔARE/+ mice. The initial lesions consist of mucosal abnormalities with intestinal villous blunting and broadening. These changes are associated with mucosal and submucosal infiltration of chronic as well as acute inflammatory cells, including mononuclear leukocytes, plasma cells, and scattered neutrophils. Severe intestinal inflammation is usually observed by 3 wk of age in TNFΔARE/ΔARE mice and 8 wk in TNFΔARE/+ mice, with an increased number of submucosal lymphoid aggregates and follicles becoming evident at this time. Progressive diffuse inflammation with patchy variability develops with age. The inflammatory infiltrate extends deep into the muscular layers of the bowel wall, displaying characteristics typical of transmural inflammation. In older mice (4–7 mo), complete loss of villous structures as well as rudimental granulomata, resembling noncaseating granulomas with multinucleated giant cells, have been observed.

To analyze the contribution of the two TNF receptors (TNFRI and TNFRII), as well as mature T and B cells in the ileitis characteristic of the TNF ΔARE model of TNF overexpression, the mutation can be introduced onto TNFRI-, TNFRI-, or recombinase-activating gene (RAG) 1-deficient backgrounds. Histological analysis of TNFΔARE/ΔARE mice at 1, 3, 6, and 8 mo shows complete alleviation of gut pathology (as well as joint pathology) when they are back-crossed onto a TNFRI-deficient background. In these mice, no evidence of transmural inflammation, increased number of lymphoid aggregates, or granulomatous structures is detected. In comparison, a significant attenuation in disease severity, with only mild inflammatory changes localized to the mucosa and submucosa, is observed when TNFΔARE/ΔARE mice are back-crossed onto a TNFRII-deficient background. Interestingly, TNFΔARE/ΔARE mice back-crossed onto a TNFRII-, but not TNFRI-, deficient background demonstrate a much more aggressive and exacerbated arthritis. Similarly, when the TNF ΔARE mutation is introduced into a RAG1-deficient background, which lacks both mature T and B lymphocytes, the CD-like phenotype is virtually prevented, with only minimal villous blunting and mild inflammatory changes limited to the intestinal mucosa. These mice, however, display a full arthritis phenotype. Together, these results indicate a dominant role for TNFRI compared with a differential role for TNFRII, as well as T and B lymphocytes, in regulating the pathogenic mechanisms mediating CD-like ileitis vs. the chronic inflammatory arthritis characteristic of the TNF ΔARE model (10).

Overall, deregulation of TNF gene expression by genetically mutating elements responsible for the regulation and function of its AU-rich region, as in TNFΔARE mutant mice, has resulted in the spontaneous development of a chronic ileitis similar to that observed in patients with CD. Interestingly, there appears to be a gene dosage effect of the mutated TNF allele because TNFΔARE/ΔARE mice develop the clinical phenotype much faster than the heterozygotes for this mutation. However, although it is unlikely that a mutation in the 3′-AU-rich region of the TNF gene occurs in patients with CD, this model offers a unique opportunity to study the precise mechanisms underlying TNF-induced CD. This may have remarkable implications for understanding how TNF mediates small intestinal inflammation as well as for developing novel therapies aimed at reducing the activity of TNF in human CD.

SAMP1/Yit model of ileitis.

Another model of experimental CD that shows great promise to identify characteristics of the human disease is the SAMP1/Yit model of ileitis. This mouse line was generated by genetic manipulation involving brother-sister mating for >20 generations of the original senescence-accelerated mouse (SAM) line (17). This particular model has the unique feature of developing spontaneously without either gene targeting or immunologic manipulations. The spontaneous ileitis develops at ∼20 wk of age and reaches virtually 100% penetrance after 30 wk of age. Interestingly, the ileitis characteristic of this model does not resolve with age but tends to progress in severity, up to 80 wk (unpublished results). The histological features of this model closely resemble human terminal ileitis, with segmental inflammation localized primarily to the terminal ileum. In addition, transmural inflammation, presence of mucosal and submucosal granulomas, changes in epithelial cell morphology and architecture, and muscular and neural hypertrophy are all pathologies observed in the SAMP1/Yit model (Table2). Therefore, one of the most attractive features of this model is that all the aforementioned characteristics are hallmark events typical of human CD. Minor differences do exist, however, including the presence of sporadic macroscopic ulceration as well as the absence of pyloric metaplasia, which are usually included in the histopathological diagnosis of the human disease. Similar to other models of IBD, the chronic ileitis can be adoptively transferred by CD4+ T cells into severe combined immunodeficient (SCID) mice (11). Interestingly, the resulting disease is once again localized to the terminal ileum, suggesting that the activated CD4+ T cells may recognize antigens specifically localized to the terminal ileum.

Another feature of the SAMP1/Yit model is that Th1-polarizing cytokines, including TNF and IL-12, appear to be key mediators of the chronic gut inflammation. In fact, administration of monoclonal antibodies against both TNF and IL-12 has the ability to suppress markedly the severity of disease compared with treatment with appropriate isotype control antibodies (24). In addition, adhesion molecules, which have been implicated in T-cell homing and neutrophil trafficking, including E- and P-selectin as well as the α4-integrin, also appear to play a key role in the initiation of ileitis in SAMP1/Yit mice. Similar antibody treatments against the aforementioned adhesion molecules have been shown to decrease chronic ileitis found in SAMP1/Yit mice (4). Finally, the spontaneous nature of this model offers a unique opportunity to perform a complete analysis of the predisposing genes involved in disease susceptibility (12).

The SAMP1/Yit model of experimental CD has allowed us to formulate a working hypothesis for the pathogenesis of human CD. This hypothesis postulates that CD is caused by a dysregulated immune response to an unknown antigen (most likely bacterial in origin) in a genetically susceptible host. Adhesion molecules, including selectins and integrins, may play a key role in mediating T cell and other inflammatory cell trafficking into the intestinal mucosa. In this context, the intestinal epithelium may also play a critical role in modulating inflammatory and immunologic responses by producing regulatory cytokines as well as mediating antigen presentation. This overall hypothesis for the etiopathogenesis of CD is illustrated in Fig. 1. Future experiments will test the validity of the proposed hypothesis in patients with CD.

Fig. 1.

Working hypothesis for pathogenesis of Crohn's disease (CD). The disease is initiated and perpetuated by a dysregulated immune response to unknown antigens in a genetically susceptible host. CD4+T cells and Th1 polarizing cytokines, such as tumor necrosis factor (TNF), interleukin (IL)-12, and IL-18, may play an important role in this process. Leukocyte adhesion and trafficking is mediated by specific adhesion molecules, including intercellular adhesion molecule 1 (ICAM-1) and vascular cell adhesion molecule 1 (VCAM-1), which are expressed on endothelial cells. In this context, intestinal epithelial cells may play a key role in regulating pro- and anti-inflammatory events characteristic of CD. IFN-γ, interferon-γ; Mφ, macrophage; ROS, reactive oxygen species; APC, antigen-presenting cell.

In summary, the availability of recently developed genetically engineered mouse models of CD offers the opportunity to define the complex mechanism(s) involved in the pathogenesis of human CD. Results generated from these studies will have potential for the development of a cure for this devastating disease.

Acknowledgments

We wish to acknowledge Drs. G. Kollias and S. Matsumoto, who developed TNFΔARE and SAMP1/Yit mice, respectively. We also thank Drs. D. Kontoyiannis, S. M. Cohn, M. McDuffie, M. M. Kosiewicz, K. F. Ley, C. C. Nast, and C. A. Moskaluk for their contribution to these studies.

Footnotes

  • Address for reprint requests and other correspondence: F. Cominelli, Div. of Gastroenterology and Hepatology, Univ. of Virginia Health System, PO Box 800708, Charlottesville, VA 22908 (E-mail:fc4q{at}virginia.edu).

  • * Eleventh in a series of invited articles on Lessons From Genetically Engineered Animal Models.

  • The studies described in this article were supported by National Institute of Diabetes and Digestive and Kidney Diseases Grants DK-42191, DK-55812, and DK-07769 to F. Cominelli.

REFERENCES

View Abstract