Innate immunity in inflammatory bowel disease: a disease hypothesis

Abstract

Crohn's disease arises from a defective interaction between the highly concentrated mass of bacteria in the gastrointestinal tract and the underlying tissues. It has generally been believed to result from an excessively exuberant inflammatory response or from ‘autoimmunity’. Recent evidence has emerged that the problem is instead a failure of the way in which the body responds to the penetration of bacteria and other bowel contents through the intestinal mucosal barrier. Rather than Crohn's disease being caused by excessive inflammation, the primary mechanism is actually that of an immunodeficiency. Failure of inflammatory mediator production leads to insufficient recruitment of neutrophils, resulting in inadequate removal of bacteria and other debris. This impairment of acute inflammation can be compensated in some circumstances by signalling through NOD2. If not cleared, the foreign material in the bowel wall is taken up within macrophages, eliciting a granulomatous reaction and the local and systemic sequelae so characteristic of Crohn's disease.

Introduction

Crohn's disease is a condition that manifests with ulceration and chronic inflammation of the gastrointestinal tract, accompanied by other systemic abnormalities [1]. In established lesions, there is an infiltration of inflammatory cells focally in the bowel, leading to the production of pro-inflammatory cytokines and propagation of the immune response [2]. This resembles the pathology of a number of other conditions, notably certain infections and autoimmune diseases, and for years these processes were believed to cause Crohn's disease. When it was not possible to consistently culture any specific normal bacteria that fulfilled Koch's postulates from these patients, more elaborate techniques were used to demonstrate the presence of cell wall-deficient atypical Mycobacteria paratuberculosis [3], L-form organisms [4], Listeria monocytogenes [5] and measles virus [6]. Although an effort was made to demonstrate their pathogenic relevance, the balance of evidence argues against them playing an important aetiological role [7]. Similarly, although findings of autoreactive T cells [8] and antibodies [9] are documented, these have not been mechanistically linked to disease causation and it is very difficult to determine whether they have arisen as primary or secondary phenomena. The situation is further compounded by the absence of an animal model that accurately reproduces all the features related to Crohn's disease [10].

The relationship between bowel contents and Crohn's lesions

What do we know for certain about Crohn's disease? Pathologically, the intestinal lesions are characterized by a transmural inflammatory infiltrate that is frequently accompanied by granuloma formation. To generate these lesions, there is an absolute requirement for the bowel luminal contents. This has been elegantly demonstrated in a series of studies in Crohn's patients in whom the fecal stream was surgically diverted away from distal bowel into an ileostomy [11]. This induced resolution of lesions in the defunctioned bowel, which recurred only when bowel continuity was restored [12] or effluent from the ileostomy was experimentally reintroduced [13]. It is also well recognized that surgical resection of Crohn's bowel is not curative, and that new lesions have a tendency to arise at the site of re-anastomosis [14]. Similarly, alterations to the bowel flora and antigenic determinants, induced by either antibiotics [15] or use of an elemental diet [16,17], can effect disease remission. Given the failure to find a definite pathological alteration in the bowel contents, it is likely to be the interaction between its constituents and the mucosal immune system that is abnormal.

NOD2 plays a supporting role

With the advent of genetic linkage analyses, a number of susceptibility loci for Crohn's disease have been identified [18]. The most well-defined was the CARD15 gene [19-21], which encodes the NOD2 protein, although other potentially strong candidates include CARD4 [22], TLR4 [23], OCTN1 and OCTN2 [24], IL23R [25] and ATG16L1 [26]. CARD15 polymorphisms are documented largely in Caucasian populations and are not found in Oriental populations [27,28]; they are principally associated with small bowel disease [29]. The presence of a single polymorphism is associated with an up to four-fold increase in the relative risk of developing Crohn's disease, whereas carriage of two variant alleles (homozygosity or compound heterozygosity) raises the odds ratio to 40 [19]. The NOD2 protein is predominantly expressed within mononuclear phagocytes [30], but has also been reported in intestinal epithelial cells [31], Paneth cells [32] and neutrophils [31]. It is a 115 kDa protein consisting of three domains [30]. All the polymorphisms associated with Crohn's disease occur in the region of its leucine-rich repeats (LRRs).

A naturally occurring molecule and derivative of bacterial cell walls, muramyl dipeptide (MDP), has been documented as an agonist of the NOD2 system [33,34]. It was thought that MDP binds directly to the NOD2 LRR, although this seems unlikely as MDP is a tiny molecule of only 493 Da, whereas a similar LRR in ribonuclease inhibitor binds to a ligand of 13 kDa [35]. MDP can nonetheless activate the pathway, since when added to macrophages in culture it induces nuclear translocation of the transcription factor NF-κB and changes in gene expression [33]. These effects were abolished in the presence of the polymorphisms associated with Crohn's disease.

Although the discovery of NOD2 attracted a lot of attention, it was not clear how the wild-type protein and the polymorphisms might be functioning. On the one hand, initial experiments suggested that stimulation of cells containing wild-type NOD2 with MDP led to induction of pro-inflammatory interleukin-1β (IL-1β) expression [33]. This was attenuated in the presence of the polymorphisms. On the other hand, since Crohn's disease was thought to arise from an excessive inflammatory response, it was proposed that the normal function of NOD2 might be to damp inflammation down. Evidence in favour of this position included the observations that prolonged stimulation of macrophages with MDP led to secretion of anti-inflammatory interleukin-10 [36], that MDP down-regulated Toll-like Receptor-2 (TLR2) pathway signalling [37], and that different mutations in NOD2 give rise to the Blau syndrome, consisting of inflammatory arthritis, uveitis and skin lesions [38]. However, it should be noted that many pro-inflammatory stimuli will initiate negative feedback mechanisms after extended activation to limit their duration [39] and that the interaction with the TLR2 system has not been universally replicated [40,41]. Alternatively, NOD2 has been demonstrated within enterocytes and Paneth cells [32], in which the polymorphisms are associated with diminished production of anti-bacterial α-defensins [42]. This might be predicted to lead to bacterial overgrowth and subsequent infection and chronic inflammation; however, no such phenotype has been observed in mice lacking these molecules [43]. Finally, a toxic gain of function was implied by transgenic mice engineered to carry a nonsense mutation thought to be homologous to one of the human polymorphisms [44]. These mice demonstrated enhanced NF-κB activation and IL-1β processing, consistent with a pro-inflammatory phenotype, but contrasting with in vitro and animal data examining the naturalistic genetic variants. The explanation for this discrepancy has not been forthcoming, but it might relate to the background strain of mouse, or to the fact that that the effects of the engineered mutations on protein structure do not correspond to those of the human polymorphism as closely as predicted.

It is very important to realize that, despite the flurry of activity that followed the first clear description of a concrete genetic risk factor for this disease, the genes so far identified only play a very minor role in the pathogenesis of the disease. For example, the CARD15 gene has the strongest association with Crohn's. Polymorphisms occur predominantly in patients with small bowel disease, and then are restricted to certain racial groups. Second, they are of low penetrance. Although their carriage greatly enhances the relative risk of developing Crohn's disease, the absolute increase is far less dramatic. Based on the known epidemiology of this disease, we can calculate that for every affected patient carrying a single polymorphism there must be approximately 500 individuals with the same polymorphism that do not develop the disease. Even with patients possessing two polymorphisms there should be about 25 people with the same genotype who are entirely healthy. Finally, cells from mice in which the homologous gene has been knocked out demonstrate identical signalling abnormalities to those in cells from human patients with CARD15 mutations, but do not spontaneously develop intestinal pathology [40,45,46]. Therefore, some other factors must be in play.

Defective acute inflammation in Crohn's disease

Granulomatous inflammation is the hallmark of failure of the cellular inflammatory response to remove foreign material [47]. This may result because it is indigestible (as illustrated by berylliosis or silicosis) because an infective agent evades immune eradication (as seen in tuberculosis), or in cases in which the digestive function of neutrophils is defective. The latter include congenital genetic disorders, such as chronic granulomatous disease (CGD) [48] and glycogen storage disease-Ib (GSD-Ib) [49], in which molecular lesions impair the ability of neutrophils, not only to kill bacteria but also to digest phagocytosed contents. In all the above examples, the exogenous agents are subsequently enveloped and contained within macrophages, which progress to form giant cells and granulomata.

These observations generated the suggestion that the chronic granulomatous inflammation that characterizes Crohn's disease, which as described above depends upon the presence of fecal contents, could result from the failure of clearance of this material after it gains access to the bowel wall. This led to an assessment of the adequacy of the processes normally associated with clearance of such material, namely the acute inflammatory response and recruitment of neutrophils; the latter engulf, digest and expel such material from the body. Subsequent investigation of neutrophils at newly traumatized sites in the skin did reveal a major defect of their accumulation [50], which was due not to an inherent defect in neutrophil motility but to a general failure of the acute inflammatory response [51]. Although this finding was replicated [52], it was later dismissed, based largely on the identification of a serum factor that inhibited neutrophil chemotaxis in vitro [53]. This inhibitory factor effect was, however, found in serum from patients with both Crohn's disease and ulcerative colitis, and correlated with disease activity. In contrast, the impaired neutrophil migration into skin windows in vivo was exclusively seen in Crohn's patients, including those in complete remission.

This issue was recently readdressed by Marks et al [54]. In view of the absence of an accurate animal model of Crohn's disease, all these experiments were performed on human subjects. The majority of patients were in complete remission and were not receiving medication.

Profound and consistent abnormalities of acute inflammation were found in all patients with Crohn's disease. These consisted of a failure of neutrophil accumulation and interleukin-8 (IL-8) production at sites of trauma in the bowel, and in the skin. These findings were not present in healthy subjects or patients with other chronic inflammatory conditions, including ulcerative colitis and rheumatoid arthritis. This diminished recruitment of neutrophils into traumatized skin was not due to inherent abnormalities in their motility, as it could be reversed by the application of exogenous IL-8 or MDP directly onto the skin window. As expected, MDP was ineffective in the presence of Crohn's-associated mutations in CARD15. The basic problem with neutrophil accumulation in Crohn's was shown to be due to the reduced secretion of IL-8 by macrophages, an effect that was replicated in vitro by placing monocyte-derived macrophages into culture and exposing them to acute wound fluid.

To test the hypothesis that there was a major defect in the handling of bowel flora in Crohn's disease in the body, heat-killed Escherichia coli were injected directly into the subcutaneous tissue of the forearm. Once again, Crohn's patients showed a consistent failure of the resulting acute inflammation. Whereas blood flow increased approximately nine-fold by 24 h in healthy controls, the increase was only four-fold in Crohn's patients with small bowel disease and two-fold in those with large bowel disease. Importantly, there was no overlap in the magnitude of the responses between any of the 13 healthy controls and 12 Crohn's patients studied.

Despite the apparently muted response to the bacteria in the Crohn's tissues, other immunological events were occurring. The levels of interleukin-6 (IL-6) and of acute phase proteins in the blood of these patients rose much higher than in healthy subjects. This suggests that the bacteria were taken up instead by macrophages, which produce IL-6 [55], a cytokine that induces acute phase protein production by the liver. These cytokine and acute phase protein changes consequent upon bacterial injection accurately mirror those observed in Crohn's patients in relapse, providing additional support for the concept that this condition results from failed clearance of bacteria and other foreign material from the tissues.

The cause of impaired acute inflammation in Crohn's disease

The acute inflammatory response resulting in the generation of IL-8 and the accumulation of neutrophils is likely to be regulated by a large number of genes. The fact that the strongest genetic linkage in Crohn's disease was to the CARD15 gene, for which penetrance is so low, indicates that the much stronger phenotype observed by Marks et al. [54] must be controlled by the compound effect of a large number of different genes. These are likely to each possess low penetrance or great variability amongst the general population.

The suggestion is that the impaired inflammation is simply a consequence of the Normal distribution of any biological phenotype that shows a bell-shaped Gaussian distribution. Any individual falling into the extreme lower tail of the response will be strongly predisposed to the development of Crohn's disease (Figure 1). This is exemplified by the syndromes of CGD [56] and GSD-1b [49], in which there is a failure of neutrophil killing and also digestion, where the incidence of non-infectious granulomatous inflammation of the bowel is in the region of 30%.

Figure 1.

It is proposed that the extent of the acute inflammatory response follows a Normal Gaussian distribution in the population and that those individuals at the extreme lower end of the response are susceptible to Crohn's disease. Inflammation can be enhanced, and the curve shifted to the right, taking those liable to Crohn's disease into a resistant phenotype, by secondary compensatory systems that react to stimulation by agonists from the gut content. These include MDP, signalling through NOD2

The action of NOD2 can be integrated into this mechanism by suggesting that it, and other molecules, can react to bacteria or their products by amplifying the inflammatory response. This could therefore shift the inflammatory response to the ‘right’ and out of the at-risk population into the normal range (Figure 1). Thus, the development of Crohn's lesions would require a ‘two-hit’ pathogenesis, the first being an inherently impaired acute inflammatory response and the second a defect in one or other compensatory pathways (Figure 2). NOD2 provides one example of an agonist of such a pathway. Crohn's lesions associated with mutations in CARD15 principally affect the small bowel, where the contents are fluid and small diffusible bacterial products, such as MDP, are present at high concentrations [57]. In the large bowel other agonists, including bacteria themselves, possibly acting through Toll-like-receptors (TLRs), might be important activating mechanisms.

Figure 2.

Schematic representation of mechanisms outlined in the legend to Figure 1

Relationship between impaired inflammation and other associations with Crohn's disease

The immunodeficiency hypothesis of Crohn's disease is consistent with many of the established epidemiological risk factors. It fits in well with the ‘hygiene hypothesis’, under which the increased incidence of the disease in the developed world has been attributed to enhanced sanitation [58]. In situations in which individuals have chronic intestinal inflammation due to repeated gastrointestinal infections and infestations, including therapeutic infestations with helminths [59], the mucosal immune system would be primed and capable of mounting a much more effective acute inflammatory response to a secondary insult.

Smoking also clearly predisposes to the development of Crohn's lesions [60]. This is both directly immunosuppressive, reducing concentrations of IL-8 within the intestinal mucosa [61], and vasoconstrictive, diminishing mucosal blood flow [62]. Both IL-8 and blood flow are closely correlated to rates of neutrophil emigration into the tissues [63]. Finally, the failure of clearance under the mechanism proposed is consistent with a number of otherwise unexplained observations from Crohn's tissues. These include the unusually low levels of IL-8 production in healing bowel following intestinal resection that preceded anastomotic recurrence [64], and the demonstration of bacterial DNA within Crohn's granulomata [65].

Implications for other granulomatous diseases

The proposed model of the development of disease through defective removal of exogenous material might well also apply to other granulomatous disorders of unknown aetiology. These could include sarcoidosis [66], where the predilection to respiratory involvement might indicate impaired clearance of a foreign (possibly inhaled) antigen, as well as granulomatous hepatitis, giant cell arteritis and Wegener's granulomatosis.

Therapeutic implications

Current effective treatments of Crohn's disease, and many potential future therapies, revolve around the use of drugs or biological agents that suppress the immune system [67]. Under the immunodeficiency hypothesis, these damp down the symptoms and signs of the chronic immune response to foreign material retained in the bowel in the secondary phase of the disease, whilst possibly exacerbating the underlying abnormality. This view is indirectly supported by the poor efficacy of corticosteroids in maintaining disease remission [68] and by the surprising potency of elemental diets [17]. There are some positive data from the use of granulocyte monocyte colony-stimulating factor (GM-CSF) as an immunostimulant [69], although this was trialled in active disease rather than for maintenance of remission, where the greatest benefits might be expected. The apparent efficacy of helminth infestation might also relate to the ability of these worms to provoke an immune response, in conjunction with their effects on T lymphocyte phenotype [59].

Alternative approaches might be to redress the failure of chemotactic cytokine production locally in the gastrointestinal tract [54]. This could be achieved by topical application of IL-8 or other agonists that stimulate its endogenous production, either directly or by use of bacteria genetically modified to secrete the appropriate mediator [70]. Such agents would be expected to penetrate the mucosa maximally at sites of developing or established ulceration, thereby limiting systemic absorption and resultant side-effects. The vasculature provides another potential target. Although the failure of blood vessel responses identified experimentally [54] may represent a marker of the underlying pathogenic process, as opposed to a primary abnormality, it could theoretically compound the problem, as leukocyte delivery depends on blood flow to the region [63]. The normal response to bacteria had a nitric oxide-mediated component, which acts through cyclic GMP (cGMP). In Crohn's patients, blood flow was transiently normalized by treatment with oral sildenafil citrate, a phosphodiesterase-5 inhibitor that augments cGMP concentrations [54]. Drugs of this class might therefore also prove useful in healing or preventing lesions.