A dysregulated immune response towards intestinal bacteria has been associated with inflammatory bowel disease (IBD). This group of chronic relapsing inflammatory diseases of the gastrointestinal tract comprises Crohn's disease (CD) and ulcerative colitis (UC) 1, 2. Although CD and UC share several genetic risk factors, clear clinical, histological and genetic differences suggest a distinct ethiology 1–3. Recent genome-wide association studies revealed over 30 susceptibility loci for CD and/or UC 3–5. These studies point to gene–gene and gene–environment interactions that are key to the complex immunological disorder seen in patients with IBD.
Functional identification of triggering events and characterisation of primary immune pathological pathways in the inflamed intestine of IBD patients are extremely difficult. Humans harbour a complex and highly individual intestinal microbiota making it difficult to judge the pathogenic or anti-inflammatory potential of single bacteria within this complex community. As patients only present with clinical symptoms, it is nearly impossible to study early stages of the disease in order to distinguish primary and secondary events. In addition, characterisation of the molecular pathology of disease is hampered by the large clinical and genetic heterogeneity of IBD patients as well as the effects of multiple drug treatment regimes. The study of more defined cohorts of patients by means of clear phenotypic, genetic and co-treatment stratification within large multicenter clinical trials may address some but not all of these problems.
Although current anti-inflammatory, immunosuppressive and immunomodulatory therapies have been successful in inducing and maintaining remission, a proportion of patients do not respond to these therapies and others suffer deleterious side effects 6, 7. Furthermore, current drug therapies do not cure IBD in the long-term.
In order to understand the complexity of IBD pathogenesis and to discover therapeutic targets, many investigators have turned to mouse model systems of intestinal inflammation. Although often criticised for not replicating all of the features of IBD, it is striking that the results from the mouse models have converged with those from human genetic studies to identify key pathways involved in intestinal inflammation. In this viewpoint, we discuss key themes that have arisen from mouse models (Fig. 1). Like the human disease, the results are consistent with a multistep model in which innate recognition of facultatively colitogenic intestinal microbiota and deviation of the innate and adaptive immune system act as dominant mechanisms for development of intestinal inflammation.
Mouse models of intestinal inflammation
The homology of basic immunological mechanisms between mammals and conservation of major cellular and intercellular signalling pathways is the basis for choosing mouse models of intestinal inflammation. A good animal model should allow a simplified view of the complex pathology found in the human disease, provide a tractable and reproducible system to identify inflammatory pathways, and test therapeutic interventions. Other considerations include simplicity of the experimental design combined with short duration of the experiment, low costs and minimal harm to animal welfare. Naturally, there is not a single model that fulfils all these demands. Nevertheless, as a group, mouse models have shaped and refined our understanding of host microbial interactions in the intestine and have been successfully applied to dissect immune-mediated pathways of intestinal inflammation over the past two decades.
Currently there are well over 50 mouse models to study intestinal inflammation. Traditionally, mouse colitis models have been divided into those with spontaneous development of intestinal inflammation, those that are induced by genetic manipulation, chemically induced colitis models and those that depend on the transfer of cell populations into immunodeficient hosts (for review see 1, 2). Development of colitis in such a large number of genetically modified mice that lack or overexpress distinct molecules was unexpected and has revealed completely unknown functions of gene products in the intestine. These results also raise the question of why so many distinct genetic manipulations in mice result in intestinal inflammation? The most likely explanation is that the dynamic relationship between the intestinal immune system and the diverse intestinal microbiota is controlled at multiple levels. Thus, manipulation of intestinal barrier function, innate host defence and adaptive immunity give rise to intestinal inflammation in mouse models. Many of the models reflect a process that involves the presence of triggering stimuli such as particular bacteria and/or epithelial cell stress leading to innate immune activation and altered adaptive immune responses. It is important to note that although the histopathological features of inflammation in different models may be similar, investigators need to consider the primary initiating events that lead to disease when interpreting results. For example, inflammatory mediators such as IL-23 that promote epithelial barrier function and innate host defence may be protective in barrier disruption models where there is increased penetrance of intestinal bacteria but the same molecule may be deleterious and drive the inflammatory response in models induced by a defect in negative regulatory pathways. Although some models have a very strong resemblance to a particular form IBD such as CD-like fistulizing terminal ileitis in the SAMP/YitFc strain 8, many models have histological and immunological features that are not easy to define as either CD or UC. Thus, in the G-α-i2-deficient mouse model of colitis histological analysis revealed several similarities to UC in humans despite being associated with a Th1 immune response 9, 10. We propose that instead of being “the” model of CD or UC, different mouse models represent a particular stage of disease (acute versus chronic for example) or mimic pathology in a particular subgroup of patients.
Role of intestinal microbiota
Study of the host microbiota and its influence on the immune system influences our understanding of concepts such as commensalism, mutualism, pathogens or probiotics as much as it facilitates the understanding of gene–environment interactions. In the majority of mouse models, intestinal inflammation is dependent on the presence of the intestinal microbiota since colitis does not develop in mice housed under germ-free conditions (for review see 1, 11). Some model systems require the presence of specific (facultatively) pathogenic intestinal bacteria such as Citrobacter spp. or Helicobacter spp. As illustrated by colitis induction after Bacteriodes vulgatus colonisation of IL-10-deficient mice, commensal bacteria can induce colitis in mice with a dysregulated immune system 12. In some model situations, the selection of a colitogenic microbiota is secondary to immune defects resulting in an intestinal flora that can induce colitis in wild-type hosts 13. There are also protective microbiota–host interactions since B. vulgatus inhibits Escherichia coli -induced colitis in IL-2-deficient mice 14. Similarly, the human symbiont B. fragilis inhibits Helicobacter hepaticus-induced experimental colitis via a polysaccharide A-dependent mechanism 15. Bacterial strains such as E. coli Nissle 1917 or bacterial cocktails (such as VSL♯3, Lactobacillus acidophilus, L. casei, L. plantarum, L. bulgaricum, Bifidobakterium longum, B. breve, B. infantis and Streptococcus thermophilus) with established therapeutic effects in human UC patients have been shown to have some efficacy in mouse models of colitis 16–18. Current studies suggest that bacterial recognition via pathogen recognition receptors such as TLR-2 and TLR-4 as well as stimulation of regulatory T cells is among the mechanisms involved in the in vivo anti-inflammatory activity of probiotic strains.
It should also be noted that although most colitis models are performed under specific pathogen free conditions, the composition of the microbiota itself is largely unknown. Differences in the flora composition of different animal facilities can therefore substantially influence the outcome of experiments.
Epithelial homeostasis, microbial recognition and autophagy
The epithelial barrier and epithelial cell products such as mucus and defensins are innate mechanisms that control the intestinal microbiota by preventing translocation as well as shaping the composition of the microbiota. The important role of the epithelial barrier has been demonstrated by the development of colitis in mice with defects in mucus production or epithelial cell interaction, homeostasis and repair. Examples include dextran sodium sulphate-induced colitis as well as mice deficient in Muc2, XBP-1, Mdr1a, intestinal trefoil factor or epithelial cell specific deletion of NF-κB components (for review see 1, 2, 19, 20). Defensins have also been shown to play an important role in intestinal immunity 21, and alterations of defensin secretion have been described in CD patients as well as in mice with experimental colitis 22; however, mice deficient in defensins do not develop spontaneous intestinal inflammation. This may be due to overlapping redundant functions of different antimicrobial molecules, which represent a failsafe mechanism of the intestinal immune system.
Polymorphisms in the NOD2 protein (nucleotide-binding oligomerisation domain protein 2) have been associated with susceptibility to ileal inflammation in patients with CD 23, 24. NOD2 recognises muramyl dipeptide, a peptidoglycan component of gram-positive and gram-negative bacteria. Consequently, it seems likely that a defect in recognition of (intracellular) bacteria is central to the immunopathology of CD; however, NOD2-deficient mice do not develop spontaneous colitis. Gain and loss of function mechanisms have been proposed to explain how NOD2 might influence susceptibility to colitis since NOD2 is expressed in epithelial cells such as Paneth cells as well as by immune cells such as macrophages. It has been suggested that activated NOD2 signalling counterbalances TLR-induced proinflammatory immune responses in dendritic cells/macrophages or Paneth cells 25. The fact that not all individuals with NOD2 variants develop CD suggests a strong environmental component to NOD2-dependent disease or/and additional genetic predisposition, which may be missing in mouse models.
Autophagy represents a cellular mechanism that controls degradation of cytoplasmic constituents. This process influences not only insoluble cytoplasmatic proteins but also the survival of, and immunity towards, invading intracellular microbes. In several genome-wide association studies, autophagy-associated genes such as ATG16L1 and IRGM were associated with increased risk of CD but not UC in humans 3. Results from recent studies suggest that autophagic pathways may contribute to the development of intestinal inflammation at several different levels. In the thymus autophagy is involved in selection of the T-cell repertoire. Mice with altered autophagy in thymic epithelial cells due to ATG5 deficiency develop wasting disease, multi-organ inflammation and severe colitis 26. Additionally, mice that are deficient in ATG16L1 in haematopoietic cells show increased susceptibility to dextran sodium sulphate-induced acute colitis 27. Finally, autophagy is also involved in the activity of intestinal Paneth cells 28. Paneth cells from ATG16L1- and ATG5-deficient mice exhibit defective granule exocytosis and increased expression of genes involved in injury responses. Via this mechanism, autophagy may influence the indigenous microbiota and could predispose the animal to colitis. The functional effect of this latter mechanism on experimental colitis, however, has not been formally shown.
Inflammatory pathways – The IL-23/Th17 and the IL-13 axis of intestinal inflammation
The impaired recognition and handling of intestinal microbiota by innate mechanisms set the threshold for intestinal inflammatory immune responses. There is evidence from several model systems that innate immune cell activation itself may be sufficient for colitis 13, 29, 30. Nevertheless, the majority of models depend on the proinflammatory activity of the adaptive immune system. In most models, CD4+ T cells play a dominant role but proinflammatory activities of CD8+ lymphocytes, NK cells, NKT cells or B cells have also been described. One of the early findings in mouse models of intestinal inflammation was the pro-inflammatory activity of TNF-a 31. The anti-inflammatory effect of anti-TNF-a blockade has since been successfully translated into clinical practice 7.
Recent studies have identified the IL-23/Th17 signalling pathway as a key player in a number of models of experimental colitis (for review see 32). IL-23 is essential for development of colitis in the T-cell transfer model, in IL-10-deficient mice as well as in a T-cell-dependent model of H. hepaticus infection. In addition, IL-23 drives innate colitis induced by H. hepaticus or anti-CD40 injection 30, 33–35. RORgt, a transcription factor required for Th17 cell development, as well as cytokines and their receptors involved in Th17 cell responses such as IL-6, IL-21 and IL-17 are required for intestinal inflammation in mouse models 36–39. IL-23 may primarily drive chronic colitis as acute inflammation such as oxazolone-induced colitis and 2,4,6-trinitro benzene sulphonic acid-induced colitis appear to be IL-23 independent 40, 41. There is also evidence for increases in Th17-associated cytokines in IBD patients 42, 43. Perhaps most persuasively, a number of genes in the IL-23/Th17 pathway including the IL-23 receptor, the common p40 subunit of IL-12 and IL-23, as well as signalling molecules downstream of the IL-23 receptor such as STAT3 and JAK2 have been identified in genome-wide association studies in both CD and UC.
Besides model systems of intestinal inflammation that are associated with Th1 and Th17 immune responses, there are also mouse models with Th2-like immunopathology. For example, the experimental colitis induced by rectal administration of oxazolone exhibits several features similar to UC in humans 19. In this model, NKT cells rather than conventional CD4+ T cells mediate the immunopathology, and neutralization of IL-13 prevents colitis in this model 44. Recently, it has been shown that mice deficient in the nuclear factor of activated T cells c2 do not develop oxazolone-induced colitis and show reduced IL-6 and IL-13 cytokine production by mucosal T lymphocytes 41. Similar to the mouse model, it has been proposed that IL-13 and unconventional NKT cells are involved in the development of UC 45.
Negative regulation of inflammation
Anti-inflammatory and anti-proliferative pathways counterbalance proinflammatory immune responses. This allows down-regulation of inflammation after successful immune responses following acute infection, avoids tissue destruction in the case of chronic infection and helps to prevent autoimmunity. Consequently, a lack of innate or adaptive immunoregulatory activity shifts the immune balance towards inflammation.
An innate negative regulatory mechanism is mediated via the protein A20. A20 inhibits the proinflammatory TNF-receptor- as well as TLR- and NOD2 signalling 46, 47. A20 deficient mice develop systemic inflammation including colitis 46. This systemic immune pathology is initiated by the resident intestinal microbiota via MyD88-dependent TLR signalling 48.
Several cell types of the adaptive immune system contribute to immune regulation within the intestine including CD4+CD25+Foxp3+ regulatory T cells (Treg) and Tr1 cells 49. Treg are involved in the regulation of intestinal inflammation, given that these cells prevent and cure experimental colitis 49. Naturally occurring Foxp3+ Treg develop in the thymus. The intestine is also a site for the development of induced Tregs that can also control colitis in model systems. The anti-inflammatory and immunosuppressive activity of CD4+CD25+ Treg is mediated via IL-10 and TGF-β signalling pathways (for review see 49). Foxp3+ Treg have been shown to accumulate in secondary lymphoid organs (spleen, MLN) as well as the intestinal lamina propria. IL-10 producing regulatory T cells accumulate in the intestine 50, 51. Recent data suggest that CD4+CD25+ Treg may control intestinal inflammation within the lamina propria, as these cells inhibit inflammation even in splenectomized lymphotoxin-α-deficient mice that lack secondary lymphoid organs 52.
Despite the compelling data from mouse models, the role of Foxp3+ Treg in human IBD is not clear. Foxp3+ T cells are abundant in the inflamed intestine in IBD and these cells have immunosuppressive activity in vitro53. These results prompt the question of why Treg that are present in the inflamed intestines of patients with IBD are not able to overcome the inflammatory response? It remains to be determined whether (i) there are intrinsic defects in immunosuppressive mechanisms, or (ii) whether the inflammatory environment in the intestine inhibits the activity of Foxp3+ Treg or (iii) whether effector cells are resistant to Treg-mediated control. In contrast to IBD, in patients with immunodysregulation polyendocrinopathy enteropathy X-linked syndrome there is a clear link between the primary defect in Treg function and colitis 54, 55. Evidence for IL-10-mediated immune suppression in UC comes from genome-wide association studies that showed sequence variants of the anti-inflammatory cytokine IL-10 are associated with increased risk for UC 5.
Limitations of the model systems and problems to solve
The mammalian gut microbiome has been shaped by co-evolution with its hosts 56–58. A general problem associated with studying human intestinal commensals, pathogens or probiotic bacteria in mouse models is associated with this strong adaptation. Indeed, bacterial colonisation, adherence, translocation as well as metabolic and immunological interactions may have completely different characteristics between human and mice and may lead to biased results.
Recently it has been estimated that the human gut microbiome consists of at least 1800 genera and 15 000–36 000 species of bacteria 59. The effort towards global scale analysis of the human and mouse microbiome will greatly enhance our knowledge of the composition of the microbiota and its genetic structure; however, new technologies will be required to investigate molecular interactions in vivo between hundreds of bacteria themselves as well as that between the microbiota and the host.
The human genome-wide association studies point to several genes that have not been tested in mouse models of intestinal inflammation. It will therefore be of great interest to investigate the functions of these target genes. So far, most experimental studies have analysed the functional consequence of targeting single genes. In patients with IBD, several homozygous and heterozygous polymorphisms and mutations may contribute to disease susceptibility. The functional analysis of gene combinations will therefore be informative to understand gene dosage effects and gene–gene interactions. As there are clear differences in colitis susceptibility between different mouse strains 60, 61, a careful analysis of mouse genetics might similarly contribute to the understanding of intestinal inflammation.
There are several characteristics of IBD that are currently not adequately illustrated by animal models. For instance, most of the mouse models of intestinal inflammation resemble characteristics of colitis, whereas the ileum is not the prominent localisation of inflammation. In contrast, ileitis is typical for a majority of CD patients. The characteristic discontinuous phenotype of inflammation and fistula formation experienced by CD patients is rarely seen in mouse model systems 8. In mouse models, the inflammation is either self-limiting or chronic but typically not of a chronic and relapsing nature.
Although we have primarily discussed flow of information from models to the clinic, it is evident that further understanding of the environmental and genetic factors that affect clinical subtypes of CD and UC will allow the design of improved models that more closely resemble the spectrum of disease phenotypes.