Introduction

Crohn's disease (CD) and ulcerative colitis (UC) are two related chronic remitting and relapsing inflammatory diseases of the gastrointestinal tract, commonly known as inflammatory bowel diseases (IBD), showing a peak incidence in early adulthood and affecting approximately 1–2 in about 1000 individuals. CD is characterized by a chronic discontinuous transmural inflammation, involving any portion of the gastrointestinal tract, but most commonly the terminal ileum. UC is characterized by inflammation of the colonic mucosa, extending in a continuous manner, and to a variable extent, from the rectum to the proximal colon.

While the causes of IBD are unknown, it is thought that inflammation results from an inappropriate chronic activation of the innate and adaptive mucosal immune systems in a genetically susceptible host 1 and that enteric microflora plays a central role in the initiation and maintenance of disease. The latter conclusion is drawn partly from the knowledge that reductions in luminal bacteria concentrations, through the use of antibiotics, can sometimes have a beneficial effect in IBD, and from the identification of an adherent‐invasive Escherichia coli (AIEC) strain specifically associated with the ileal mucosa of CD patients 2. Experimental models of colitis, most of which do not develop inflammation when kept under germ‐free (gnotobiotic) conditions, also support this conclusion 3.

The epidemiological data on IBD suggest a strong genetic contribution to disease pathogenesis, and a familial inheritance pattern that does not follow simple Mendelian models 4. Because of this, IBD has been classified as a complex disease. In contrast to single‐gene disorders, where the causal relationship between mutation and disease is clear, complex diseases result from the interaction of multiple genetic and non‐genetic factors. A strong familial clustering of disease has been reported in IBD, with up to 5%–10% of patients having at least one affected first‐degree relative, and with relative risk to siblings reported in the range of 13–36 and 7–17 for CD and UC, respectively 4–6. Twin studies strongly support a genetic contribution to IBD, with monozygotic and dizygotic twins showing 37%–58% and 4%–12% disease concordance for CD respectively, as opposed to 6%–17% and 0%–5% in UC 4,5, 7,8. Furthermore, it has been observed that age of diagnosis and disease phenotype are also inheritable in CD 9. Finally, some studies have shown that first‐degree relatives of CD and UC patients are at greater risk of developing either disease, indicating that while CD and UC have disease‐specific susceptibility loci, they appear to be genetically related and may share some susceptibility genes 4,5,10.

This review will offer an overview of the methods used in identifying novel risk loci in IBD and in characterizing their potential pathophysiological mechanisms. Based on current knowledge in IBD, many different etiologies have been proposed to explain the pathophysiological mechanisms of disease. Although much remains to be discovered, we will review the different loci implicated in disease risk in the context of three proposed mechanisms for uncontrolled inflammation of the gut mucosa (illustrated in Figure 1): 1) inappropriate regulation of innate immune response to enteric microflora or pathogens; 2) increased permeability across the epithelial barrier; or 3) defective regulation of the adaptive immune system. Finally, a review of the potential clinical applications of genetic findings, such as in the area of pharmacogenomics, is also presented.

Molecular pathogenesis of inflammatory bowel disease: Genotypes, phenotypes and personalized medicine

All authors

Philippe Goyette, Catherine Labbé, Truc T. Trinh, Ramnik J. Xavier & John D. Rioux

https://doi.org/10.1080/07853890701197615

Published online:

08 July 2009

Figure 1. Potential molecular pathways involved in inflammatory bowel disease(IBD) pathogenesis. In healthy individuals, bacterial invasions trigger cascades of immune events leading to chemokine secretion, bacterial clearance and apoptosis. Although there is much still to be discovered, genetic defects at three different levels are suspected to lead to chronic inflammation in IBD: 1) Mutations in environmental sensors and their signaling pathways, such as TLRs and CARD15, may impair the innate immune system and cause a decrease in NF‐kB activation and defensin production. 2) Mutations in transporters such as MDR1, OCTN1 and OCTN2, or in genes involved in epithelial integrity (DLG5, MYO9B) may affect the permeability of the epithelial barrier. 3) Mutations in genes involved in adaptive immunity, such as the human leukocyte antigen (HLA) genes and tumor necrosis factor (TNF‐) alpha within the major histocompatibility complex (MHC), as well as in genes coding for subunits of cytokines and cytokine receptors like IL‐12 and IL‐23R, may cause an imbalance between regulatory and effector cell immune responses important in the control of inflammatory reactions. IBD = inflammatory bowel disease; TLR = Toll‐like receptor; NF‐kB = nuclear factor‐kB; MDR1 = multidrug resistance gene; OCTN1 and 2 = organic cation transporter 1 and 2; DLG5 = drosophila discs large homolog 5.

Display full size

Key messages

• Significant advances have been made over the last decade in the understanding of the genetics of inflammatory bowel diseases (IBD) with several susceptibility loci identified through genome‐wide linkage studies, association mapping and candidate gene association studies, and confirmed by multiple replication studies.

• Currently known genetic risk factors support the role of at least three pathophysiological mechanisms in disease etiology: inappropriate regulation of innate immune response to enteric microflora or pathogens, increased permeability across the mucosal epithelial barrier and defective regulation of the adaptive immune system.

• The approaches and tools developed for the genetic studies of complex traits will undoubtedly identify additional pathophysiological mechanisms.

• These approaches and tools can now be applied to the field of pharmacogenetics with the goal of identifying genetic factors affecting variable drug responses and therefore provide the basis for a more personalized clinical management to patients with IBD.

Genetic approaches for identification of genetic risk

The two major statistical genetic approaches that are used to study genetic susceptibility in IBD are linkage and association analyses. The steps involved in identification of genetic factors are summarized in Figure 2. In linkage analysis, which makes use of family structure in order to follow the inheritance patterns of marker alleles and disease phenotype, a chromosomal region is identified as a potential locus when disease and marker allele are co‐inherited more often than would be expected by chance (Figure 3). While successful in identifying genes for diseases with classical Mendelian inheritance, linkage analysis has been less powerful in most complex diseases due to the lack of strong direct relationships between mutation and disease. In IBD however, linkage analysis has shown success in identifying some of the chromosomal regions conferring disease risk.

Molecular pathogenesis of inflammatory bowel disease: Genotypes, phenotypes and personalized medicine

All authors

Philippe Goyette, Catherine Labbé, Truc T. Trinh, Ramnik J. Xavier & John D. Rioux

https://doi.org/10.1080/07853890701197615

Published online:

08 July 2009

Figure 2. Schematic representation of the major steps towards the identification of genetic risk factors in disease. Once epidemiological studies have established a genetic link to disease, the search for genetic risk factors may progress through two major statistical approaches. Genome‐wide linkage analyses, which typically tests 300 simple sequence length polymorphism (SSLP) markers or 3000 single nucleotide polymorphisms (SNPs), can be used to identify chromosomal regions which putatively contain risk factors. Association analyses can be used to study these chromosomal regions or candidate genes identified through biological function. Recently, genome‐wide association analyses, which can test 100,000 to 500,000 SNPs, have also been used in the identification of risk loci. Association mapping narrows the search for risk factors to a particular genomic region, a specific gene or a single SNP. Once a putative association has been identified, and replicated in independent samples, the associated allele can then be tested for interaction with other identified genetic and non‐genetic risk factors. Re‐sequencing allows the identification of all variants in the region, in an attempt to identify the causal allele. The final, and often the most challenging, step in genetic studies is to link the genetic variant to a pathophysiological mechanism of underlying disease. The associated variant may be located within a specific gene and/or suggest a potential functional effect in disease etiology, but will also require functional studies in cellular and animal models, as well as evaluation in patient samples for final link to disease. The confirmation of novel disease risk factors should allow the identification of novel biological pathways to study in genetic association analysis, the generation of a molecular classification system for patients, and the discovery of potential drug targets for treatment.

Display full size

Molecular pathogenesis of inflammatory bowel disease: Genotypes, phenotypes and personalized medicine

All authors

Philippe Goyette, Catherine Labbé, Truc T. Trinh, Ramnik J. Xavier & John D. Rioux

https://doi.org/10.1080/07853890701197615

Published online:

08 July 2009

Figure 3. Classical representation of linkage and association studies. Linkage studies attempt to localize chromosomal regions containing disease genes by employing indirect statistical tests that examine for co‐inheritance of genetic markers and (disease) trait within families. Linkage analyses are very powerful to study highly penetrant genetic factors which are characteristically rare in a population. Typically, these studies look at many families and compare allele sharing between affected family members. For example, panel A depicts two parents and two affected children. Affected children are expected to share 0, 1, or 2 alleles in a proportion of, respectively, 0.25/0.50/0.25. If the affected siblings share alleles at a specific marker more often than they would by chance, then the marker is likely to be linked to the disease. Although the alleles shared between siblings have to be the same, the shared alleles across families can be different and still be linked to the disease. Association studies, on the other hand, test a specific allele across all individuals. There are two common designs for association studies: cases/controls (B), and trios consisting of two parents and one affected child (C). Case/control association studies test whether an allele is correlated with a disease by determining if allele frequencies differ between cohorts of unrelated cases and controls. Hundreds to thousands of samples are tested in cases/controls studies. Panel B illustrates an obvious difference between the frequencies of allele C in a study of six cases, 66%, and six controls, 33%, ‘suggesting’ association of this allele with the disease phenotype. In trios, the transmitted allele is equivalent in a case/control study to the ‘case’ and the untransmitted allele to the ‘control’. Panel C demonstrates the principle using a single trio, but hundreds of trios are usually tested. In this illustration, allele C is transmitted twice (100%) to the affected child ‘suggesting’ an association. Case/control studies are more efficient than family‐based designs because they require fewer samples (one case, one control as opposed to three samples in trios) to yield the same amount of information. However, in population‐based designs there are sometimes differences in allele frequencies in cases and controls not related to the trait under study. This phenomenon is called stratification. Trios are not prone to stratification since the perfectly matched control is the untransmitted chromosome. Linkage and association studies can be genome‐wide or targeted to a specific chromosomal region depending on the choice of markers. Fc = frequency of allele C; T = transmitted; U = untransmitted.

Display full size

While early studies focused on linkage analysis, more emphasis has recently been placed on association studies for the identification of causative alleles. An association study compares the allele frequency of a selected marker, most often a bi‐allelic single nucleotide polymorphism (SNP), for differences between case and control populations (Figure 3). SNPs represent most of the common genetic variation, with an estimated 10 million SNPs found in the human genome (International HapMap Consortium 2003) 11. It has recently been shown that genetic variation in the human genome exists as simple patterns of highly correlated SNP alleles, defining haplotypes within which little recombination occurs 12–14. It is therefore possible to study the entire common variation of a region through the use of a limited number of carefully selected SNPs capable of capturing highly correlated SNPs alleles. Association studies have shown higher statistical power to detect causal alleles in complex disease than linkage studies. The statistical power of an association study is dependent upon the sample size, the frequency and strength of the disease allele, the capacity of selected SNPs to capture the causative allele, as well as frequency of the disease in the population. Association studies have been performed on candidate genes located within linkage regions or selected based on function (Figure 1). Currently, whole‐genome association studies are being performed on both simple and complex diseases, and hold great promise for the search for genetic risk factors in complex diseases. It should be noted that putative association results from an association study need to be replicated in order to distinguish between true and false‐positive associations. As final proof of disease causality, an associated variant should be evaluated for functional effect in patient samples and shown to affect mechanisms leading to disease pathogenesis in a biologically relevant system. This final link of genetic variation to disease is often the most challenging part of identifying a genetic risk variant.

In order to get around some of the difficulties encountered in the genetic studies of IBD patients, several research groups have adopted the complementary approach of studying experimental models. There are in excess of 30 mouse models of IBD, both spontaneous (from mutation or transgene expression) and induced (chemically or through immune‐cell transfer), which can be classified into different categories depending on the system they affect: disruption of epithelial barrier, inappropriate regulation of T cell function and cytokine synthesis, and aberrant response to luminal bacteria. While a thorough review of the different experimental models of IBD is beyond the scope of this article, several recent reviews have been published on the subject 1,15. Animal models have been invaluable in better understanding some of the major pathophysiological mechanisms leading to IBD; however, most loci identified in mice as causing an IBD‐like phenotype have not been shown to have genetic variation that leads to disease in human. Nonetheless, animal models will continue to play an essential role in IBD research, through the functional evaluation of novel risk alleles identified by genetic analysis of IBD patients.

Identification of IBD risk loci

Genome‐wide linkage mapping studies in IBD have met with great success with over ten chromosomal regions showing evidence of linkage and replication in IBD, some showing specificity to CD, to UC, or to both 16–22. That no single locus was identified in every study illustrates a limitation of linkage analysis when using relatively small cohorts in complex diseases. Subsequent fine‐mapping of these regions and candidate gene association studies have identified several susceptibility loci, although the reproducibility of the evidence and the strength of the association signals for these vary greatly.

The genes discovered to date offer insight into some of the pathophysiological mechanisms leading to IBD. They include firmly established regions where a specific locus or specific alleles have been identified, as with the CARD15 gene 17,19,23, regions of clearly replicated association where no single locus or allele has been implicated, such as IBD5 19,24–29 and the human leukocyte antigen (HLA) 17,30–37 within the major histocompatibility complex (MHC), and putative susceptibility genes where further replication is needed to provide definitive proof of association including DLG5 38–41, TLR4 42–46 and others. These loci, however, explain only a fraction of the genetic risk associated with IBD. Because of this, there has been great effort invested in the identification of additional genetic loci involved in risk to IBD through further positional mapping, as well as candidate gene and genome‐wide association (GWA) studies.

The role of innate immunity in IBD

The first proposed mechanism in IBD involves the inappropriate regulation of the innate immune response at the level of the intestinal mucosa (Figure 1). The innate immune system is the first line of defense against resident luminal microflora and invading pathogens, and can respond to a wide variety of microorganisms. The innate immune system has evolved to monitor the resident microflora and relay danger signals in response to infection by invasive organisms. This response is mediated through the recognition by specific pathogen recognition receptors (PRR) of microbial components, known as pathogen‐associated molecular patterns (PAMPs). The PRRs include the members of the Toll‐like receptor (TLR) family, which are predominantly cell surface receptors, and the cytosolic Caterpillar‐(CARD)/NOD intracellular receptors 47,48. The binding of PAMP ligands to specific PRRs leads to the activation of several intracellular signaling pathways, which include the nuclear factor‐kB (NF‐kB) and mitogen‐activated protein kinase (MAPK) pathways for TLRs and predominantly the NFκB pathway for CARD/NOD receptors. These pathways in turn lead to the activation of transcriptional programs resulting in the broad‐spectrum non‐specific killing mechanisms of innate immunity. These mechanisms include synthesis of reactive oxygen species, activation of the complement protein system, secretion of chemokine and cytokines for chemotaxis of phagocytotic macrophages, and secretion of antimicrobial proteins by Paneth cells. Paneth cells are specialized epithelial cells located at the base of small intestinal crypts, which monitor the intestinal lumen and are considered important mediators of mucosal innate immune defense. They contribute to host defense and maintenance of the gastrointestinal barrier through the luminal secretion of a number of antibacterial peptides (defensins, lysozyme and secretory phospholipase A2) 49–51, which protect nearby intestinal stem cells and control microbial density.

Recent studies in experimental models estimate that close to 40 different genes (including known TLR family members) may play a role in innate immune response to bacterial invasion, many of which have yet to be identified 52. Given the importance of innate immunity and pattern recognition receptors in host defense, and following the identification of the CARD15 as an important risk locus for IBD (described below), several laboratories have focused their attention on the role of PRRs, such as the CARD/NOD and TLR receptor families, and their downstream signaling pathways in innate immunity and the development of IBD. The results of these studies are described below and summarized in Table I.

The first identified, and most strongly established risk locus for IBD is the CARD15/NOD2 gene, identified by Hugot et al. 23 using an association‐based fine‐mapping approach of the CD‐specific IBD1 linkage region. In this study, the authors observed a strong association signal centered on a single region which decayed rapidly on either side of the CARD15 locus, thus implicating this gene in the risk for CD. In parallel, two groups 53,54 reported an association of CD to the CARD15 locus using a candidate gene approach. These studies identified three common CARD15 coding variants (Leu1007insC, Arg702Trp, and Gly908Arg) associated to CD but not UC, all located within the leucine‐rich repeat (LRR) domain of the protein involved in ligand recognition. The association of the three common CARD15 variants with CD has been replicated in a number of independent reports 32,55–57, and therefore these can be considered as confirmed CD susceptibility alleles. While the common variants account for approximately 20% of the genetic risk in IBD 58, several additional CARD15 variants have been identified in CD patients, suggesting that additional CARD15 variants may have an impact on risk in CD 57. However, due to the low prevalence of these, it has been difficult to evaluate which of these truly impact on disease susceptibility.

CARD15, a member of the PRRs, is believed to play an important role in innate immunity. This cytosolic receptor activates NFκB signaling following binding of its ligand muramyl dipeptide (MDP), a peptidoglycan (PGN) derived peptide from both Gram‐positive and ‐negative bacteria. CARD15 is mainly expressed by epithelial cells and Paneth cells, and by antigen presenting cells (APC) such as macrophages, dendritic cells and monocytes.

While its link to CD is firmly established, the suggested effects of CARD15 variants on bacterial sensing, and in particular on NFκB signaling, remain seemingly contradictive to the clinical phenotype of increased NFκB activity and pro‐inflammatory T‐helper type 1 (Th1) response. Specifically, all three common CARD15 variants exhibit decreased NFκB activation following MDP stimulation in vitro54 and were suggested to cause increased susceptibility to infections. Watanabe et al. 59 observed that CARD15‐signalling inhibited TLR2‐mediated pro‐inflammatory cytokine production in wild‐type mouse splenocytes, and that CARD15 deficiency led to a loss of this inhibition, resulting in increased of NFκB activity and Th1 response upon stimulation of TLR2. Moreover, a recent report from the same group showed that the development of colitis in CARD15‐deficient mice, induced through the adoptive transfer of ovalbumin‐specific CD4+T cells followed by intrarectal challenge with ovalbumin, was abolished in CARD15‐TLR2 double mutants 60. This observation further supports the role of an uncontrolled TLR2 response in CARD15 deficiency, leading to the development of an exaggerated Th1 response and colitis in this system. However, the inhibition of TLR2 signaling by CARD15 was not observed by van Heel and colleagues 61, who showed instead that CARD15 stimulation, with low doses of MDP, caused a synergistic enhancement of signaling through TLRs in human peripheral blood mononuclear cells. They also demonstrated a generalized defect in TLR‐mediated responses in human mononuclear cells homozygous or compound heterozygous for CARD15 mutations. This loss of synergy between CARD15 and TLRs signaling was also observed by Kobayashi et al. 62 in cells from an independent knock‐out mouse model. This group observed that CARD15‐deficient mice also showed increased susceptibility following oral, but not intraperitoneal or intravenous, challenge with Listeria monocytogenes, a Gram‐positive intracellular bacterium, as well as decreased production of bactericidal defensins by Paneth cells, indicating that CARD15 may play a central role in mucosal immunity. It has recently been reported that CD patients with ileal disease show significantly reduced levels of defensin production by Paneth cells compared to unaffected individuals or to CD patients with colonic disease, and that carriers of CARD15 mutations show an even more pronounced reduction 63,64. Interestingly, studies investigating the association of CARD15 variants with disease phenotype have shown a strong association to ileal or ileo‐colonic CD 32,44,53,65,66, but not with colonic CD or with UC. Studies using CARD15 knock‐out models have not demonstrated spontaneous intestinal inflammation, nor have they showed increased susceptibility to chemically induced inflammation. Maeda et al. 67 examined a mouse knock‐in model for the common human frameshift mutation (CARD15^2939iC) and observed increased NFκB and downstream pro‐inflammatory cytokine production in macrophages stimulated with MDP. These animals also showed increased susceptibility to intestinal inflammation in response to chemical injury. The disparity of observations among studies on CARD15 demonstrates the challenges in deciphering functional consequences in a complex disease, where a single genetic or environmental factor is neither necessary nor sufficient for expression of the final clinical phenotype.

A second member of the NOD receptor family, NOD1, has also been evaluated in IBD. NOD1 is very similar in structure and gene expression pattern to CARD15, and can also activate NFκB in response to ligand. NOD1 is the intracellular receptor for a peptidoglycan‐derived peptide, g‐D‐glutamyl‐meso‐diaminopimelic acid (iE‐DAP), mainly associated to Gram‐negative bacteria 68,69. A study by Zouali et al. examined a single non‐conservative polymorphism identified from IBD patients and failed to identify association between NOD1 and IBD 68. A more recent association study of the allelic variation at the NOD1 locus in IBD patients has reported the potential association of an intronic insertion/deletion polymorphism in two independent cohorts 70. It is important to point out that, although interesting, these results in IBD are preliminary and will require further replication.

The Toll‐like receptors (TLRs) are another important family of PRRs whose function has been studied in IBD. TLR4 is expressed in intestinal epithelium, where it functions as the sensor for lipopolysaccharide (LPS) from Gram‐negative bacteria. TLR4 is an interesting candidate for IBD since increased protein levels were observed in the intestinal epithelium of IBD patients 71. Furthermore, genetic variants and biological antagonists of this receptor have been reported to modulate severity of inflammation in experimental models of colitis 72,73. A recent report by Franchimont et al. 43 described an association of the D299G variant with CD and UC in two case/control cohorts. This variant is located in the ligand binding leucine‐rich repeat (LRR) domain and shows LPS hyporesponsiveness both in vivo and in vitro74. While this association was replicated by several groups 75–77, there have been some conflicting reports as well 42,44–46. In parallel, studies of the bacterial receptor for the LPS/LBP complex, CD14, important in the presentation of LPS to TLR4, have also yielded interesting results. Klein et al. 78 reported the significant association of a functional promoter polymorphism (−159T/C) with CD but not UC, an observation that was replicated in an independent study 76; however, several independent studies do not support these findings 42,46,79–83. Given the modest effect on any one genetic locus in a complex trait, individual studies may often be underpowered to detect a given locus and therefore lead to seemingly conflicting results and inconsistent replication studies. Lohmueller et al. suggested that a meta‐analysis approach, which combines the data from multiple similarly designed studies, may help resolve this issue and distinguish between false‐positive and false‐negative results 84.

Variants in two additional TLRs have been identified as possible risk factors in IBD, but require replication for definitive proof of association. The first, TLR5, mediates the innate immune response to bacterial flagellin resulting in the induction of epithelial pro‐inflammatory gene expression 85. A recent study by Gewirtz et al. 86 has putatively identified a protective effect of a dominant‐negative allele of TLR5 (TLR5‐stop) on CD, where carriers of this allele have a reduced risk of developing CD. Flagellin has been identified as a dominant antigen in several models of experimental colitis (87), as well as a target for CD‐associated immune response (88,89). In fact, the antiflagellin antibodies seem independently associated to CD, and more specifically with complicated CD (88). The other TLR family member implicated in IBD risk, TLR9, maps to chromosomal region 3p21.3 (IBD9), a region previously identified by several linkage studies 16,20,90 as a susceptibility locus for Crohn's disease and ulcerative colitis. Torok et al. 91 reported the putative association of a promoter polymorphism (−1237C/T) in TLR9 to Crohn's disease. TLR9 is strongly expressed in Paneth cells, where it recognizes unmethylated CpG dinucleotides contained within consensus sequences common in bacterial DNA (CpG‐motifs). Rachmilewitz et al. 92 suggested that the beneficial effect of probiotic bacteria occurs through TLR9 signaling by these DNA motifs, and that CpG motifs can ameliorate colitis in experimental models and inhibit pro‐inflammatory cytokine production in ex vivo analysis of intestinal mucosa samples from UC patients 93. Probiotics have classically been defined as live microorganisms that benefit human health. However, the literature on the subject of probiotics and gastrointestinal health is contradictory, with several reports indicating that live bacteria are necessary for probiotic effect and that bacterial DNA can exacerbate inflammation in experimental models, while others indicate that non‐viable microorganisms and their components can promote the probiotic effect.

A recent study by Pierik et al. 94, focusing on the association of 36 non‐synonymous SNPs within TLR1‐10 with IBD, failed to identify SNPs associated with disease susceptibility in these genes. However, the study did find a positive association of SNP R80T in TLR1 as well as SNP R753G in TLR2 with the pancolitis phenotype in UC, a negative association of TLR6 with proctitis in UC, and a negative association of TLR1 with ileal disease involvement in CD. TLR1 and TLR6 are co‐factors to TLR2 in the recognition of triacyl and diacyl lipopeptides respectively. Further replication is needed to confirm these observations.

As mentioned previously, one of the primary signaling pathways activated upon recognition of specific ligands by both the NOD and TLR family members is the NFκB pathway. NF‐kB1, a member of the nuclear factor‐kB (NF‐kB) family of transcription factors, is involved in the regulation of innate and adaptive immune responses, acute inflammatory response and apoptosis 95. An increased NFκB activity was identified in the intestinal lamina propria cells of CD and UC patients 96. In addition, NF‐kB1 is located in chromosomal region 4q24 previously linked to IBD susceptibility, but more specifically associated to UC 17,18,20,97. Karban et al. identified a common promoter polymorphism in the NF‐kB1 gene (−94delATTG) which they reported was significantly associated to UC 98. Functional studies of this variant showed reduced binding to colonic nuclear extracts and reduced activity of the mutated promoter in reporter assays. The association of the −94delATTG promoter variant with UC was recently replicated by Borm et al. 99, but not by Oliver et al. 100 or by Glas et al. 101.

The role of mucosal integrity and transepithelial transport in IBD

The second proposed mechanism for the development of IBD involves the increased permeability across the mucosal epithelial barrier due to loss of structural integrity and/or abnormal transepithelial transport (Figure 1). While innate immunity is a key mediator of mucosal immune defense, the epithelial mucosa is the first line of physical defense against invading microorganisms. The intestinal epithelium represents a unique challenge for the organism, as it must balance the need for an extensive surface area permitting effective absorption of nutrients, with the need of defending against the intrusions of microorganisms which colonize the gastrointestinal (GI) tract 49,50. Unlike other epithelia, the intestinal mucosa is composed of a single layer of polarized intestinal epithelial cells which protects against direct contact of enteric antigens, bacteria or other pathogens with the underlying gut‐associated lymphoid tissue (GALT). The integrity of the epithelium is maintained mostly through a combination of intercellular adhesion structures and specialized junctions, which also define cellular polarity. In addition, the presence of mucins and trefoil peptides, the rapid turnover of epithelial cells, and the peristaltic movement of the GI tract all help to protect against colonization and invasion of the intestinal mucosa by pathogens 50. The role of increased epithelial permeability across the gut epithelial barrier (leaky gut) has gained increasing support in IBD pathogenesis, particularly as this epithelium represents an interface for genetic and environmental influences 102–110. Evidence of a genetic contribution to increased intestinal permeability has long been suspected in IBD patients 111–114 and their unaffected relatives. Interestingly, some of the loci identified through linkage and association analysis have been suggested to be involved in epithelial integrity, differentiation and transepithelial transport, although a causal link of these to modulation of permeability remains speculative at this time 27,40.

The IBD5 locus represents an example of this possible link between modulation of transepithelial transport and IBD. The IBD5 region was first reported by Rioux et al. 20 as a significant linkage signal for CD within chromosomal region 5q31, and showing the strongest signal in early onset disease. Association mapping of this region identified a common, highly conserved, 250‐kb risk haplotype for CD 29, extending uninterrupted across five known genes (IRF1, OCTN2, OCTN1, PDLIM4, and P4HA2). This haplotype could be defined equally well by any one of 11 unique SNP alleles in strong linkage disequilibrium (LD) across the region. LD occurs when the alleles at neighboring markers (observed on the same chromosome) are associated within a population more often than if they were unlinked and therefore are co‐inherited more often than expected by chance. Because of this strong LD, the narrowing down of the region to a single locus and the identification of the primary causal allele from nearby associated variants have been problematic. The association of the IBD5 risk haplotype to CD has been replicated by several other groups 25,26,28,115 in European cohorts, using both case/control and trio analyses (P‐values ranging from 0.016 to 2×10⁻⁷).

Recently, Peltokova et al. 27 described two potentially functional variants for Crohn's disease located within the IBD5 risk haplotype. The first is a leucine to phenylalanine substitution at amino acid position 503 (L503F) of the OCTN1 (organic cation transporter 1) protein. The second is a G to C transversion 207 bases upstream of the OCTN2 start codon (−207G/C) which has been proposed to disrupt a putative regulatory element. OCTN1, a sodium‐independent organic cation transporter, and OCTN2, a sodium‐dependent high‐affinity carnitine transporter, are believed to play critical roles in the elimination of endogenous small organic cations, drugs and environmental toxins. This group investigated the impact of the L503F OCTN1 variant in fibroblasts and found carnitine uptake to be 2.7 times lower in cells expressing 503F than in cells expressing 503L. In parallel, a functional study of the ‐207C OCTN2 risk allele suggested this variant may lead to a decrease in OCTN2 expression. The authors described a stronger association of these alleles to Crohn's disease than the SNPs originally defining the IBD5 haplotype, suggesting these may be responsible for the observed IBD5 association and may contribute directly to CD susceptibility. More recently, reports have been published which indicate that the OCTN variants are not better predictors of CD risk than other IBD5‐specific SNPs 41,116–119 and that CD susceptibility cannot be attributed exclusively to the two OCTN variants. Given the current difficulty in conclusively identifying the causal variants within IBD5, future progress will likely require supportive evidence from functional studies, linking identified variants with phenotype. The link between the OCTN variants and disease pathogenesis remains unclear since currently no investigation of functional differences in mucosal samples from IBD patients and healthy controls has been reported. Ultimately, mouse knock‐in models for the CD‐associated OCTN1 and OCTN2 variants showing an effect on permeability or any other mechanism leading to disease pathogenesis would strengthen the role of these variants in IBD. Finally, there are several other genes within the region which may represent equally good candidates for playing a role in CD pathogenesis. These, however, are not related to permeability, and therefore a different mechanism by which IBD5 haplotype increases risk may yet be determined.

A second locus, with a speculated role in epithelial integrity, was identified by association mapping of chromosomal region 10q23 40, a region previously identified in genome‐wide linkage analyses 17. In this study, Stoll et al. reported that the ‘A’ allele of a G/A polymorphism at position 113 in the drosophila homolog discs large gene 5 (DLG5) was overtransmitted to IBD patients. The 113A allele results in an arginine to glutamine substitution at amino acid position 30 (R30Q) of the DLG5 protein. This change was hypothesized to impair the suggested scaffolding function of this protein and affect the epithelial barrier, suggesting a potential role for this variant in epithelial integrity. A subsequent association study by Daly et al., comprised of case‐control and family based designs, confirmed the R30Q association in two out of three independent cohorts 38. Several other studies, performed in different populations, could not confirm this association signal 41,44,118,120–126. The contradictory results for DLG5 seem to indicate that its effects on IBD development are modest and might be population‐ or phenotype‐specific. Furthermore, Friedrichs et al. 127 found that the association of the R30Q variant with CD was restricted to male patients, a possible explanation for the discrepancies in the results published on DLG5. Further efforts to clarify the function of this protein in the intestinal epithelium will be particularly pertinent in light of present challenges of replication with IBD.

A third locus, myosin 9B (MYO9B), has recently been associated to UC through a candidate gene approach 128. MYO9B is a member of the class IX myosin family, which contain a Rho‐GTPase activation domain 129,130. Rho‐GTPases have previously been shown to be involved in actin filament remodeling and regulation of tight junction assembly 131,132. This activity of MYO9B in enterocytes has been suggested to play a role in the regulation of epithelial permeability. MYO9B was originally identified as a risk factor in celiac disease, a distinct intestinal inflammatory disease, through a candidate gene association search within a linkage peak on the short arm of chromosome 19 133,134. The Chr19p region (IBD6) was also identified as a locus showing genome‐wide significance in a linkage scan of IBD 20. Based on its position and on its suggested function in epithelial integrity, this locus was selected as a strong candidate for evaluation in IBD. In a recent study, Van Bodegraven et al. 128 reported that a non‐synonymous variant, Ala1011Ser, in one of the calmodulin domains of MYO9B, showed a strong association to disease risk in UC, with CD also showing a significant but weaker signal. As part of this study, the association results were replicated in three independent cohorts, and a highly significant signal (OR 1.2; P‐value = 1.9×10⁻⁶) was detected in a combined meta‐analysis of results from these three cohorts (a total of 2500 cases studied: CD n = 1200, and UC n = 1300). An independent study of MYO9B, in a smaller Norwegian IBD cohort (including 457 cases: CD n = 149, and UC n = 308), failed to detect association at this locus 135. However, this analysis did not investigate the most associated variant (Ala1011Ser) described in the former study, and the authors suggest that based on its design the study is underpowered to detect signals with OR below 1.5.

The fourth and final locus described in this section is involved in epithelial transport. The multidrug resistance gene (MDR1) encodes p‐glycoprotein 170, an adenosine triphosphate (ATP)‐dependent efflux pump expressed on the epithelial surface of the gut, which has been suggested to be implicated in the elimination of xenobiotics such as bacterial products. MDR1 is located on chromosome 7q, a region previously identified through a genome‐wide linkage search 16. MDR1 knock‐out mice develop spontaneous intestinal inflammation resembling human IBD with abnormal epithelial cell growth and leukocytic infiltration into the lamina propria 136. Several association studies investigating the role of MDR1 in IBD have been performed, focusing on two polymorphisms found to alter MDR1 expression: C3435T in exon 26 137, and G2677T in exon 21 138. Specifically, a study by Schwab et al. reported association between the 3435T allele (and 3435TT genotype) and UC 139, while subsequent association studies on this polymorphism yielded variable results in different populations and on different clinical phenotypes 140–142. A recent meta‐analysis of MDR1 variants in IBD has proposed a weak but significant association between the 3435T allele (and 3435TT genotype) and UC 143. Finally, a recent study by Ho et al. of the allelic variation at the MDR1 locus found a significant association between UC and common MDR1 haplotypes 144, bringing further support to the possible implication of the MDR1 locus in UC susceptibility.

IBD and adaptive immunity

The third proposed mechanism for the development of IBD involves the deregulation of the adaptive immune system stemming from an imbalance between regulatory and effector‐cell immune responses to luminal antigens or other antigen (e.g. self‐antigens) (Figure 1). In contrast to innate immunity, adaptive immunity generates a slow and more targeted response involving antigen‐specific recognition and immune memory. The GALT represents the largest part of the body's immune system, and given the large surface area of the mucosal epithelium, the immune system encounters more antigens in the gut than any other location in the body. In addition, since most of the antigens encountered by the mucosal immune system are derived from food proteins and commensal bacteria, the immune system must remain relatively unresponsive to avoid responses to harmless antigens and maintain epithelial integrity. It has been proposed that tolerance to these luminal antigens, also known as oral tolerance, occurs through a state of active cellular suppression or clonal anergy of immune reactive cells induced by specialized regulatory T cells 145. The proposed role of adaptive immunity in IBD is derived from genetic studies indicating a central role of the MHC in the development of IBD and also from in vivo observations in IBD patients of abnormal patterns of cytokine production and immune cell responses (excessive Th1 response in CD and Th2 response in UC), of modulations in regulatory T and B cell functions 146, and of antibodies to luminal antigens.

The MHC locus, located on the short arm of chromosome 6, is a highly polymorphic and gene‐dense region that contains genes with essential roles in adaptive immunity. This region contains the classical HLA class I and class II genes involved in antigen presentation, as well as over 200 additional genes within the extended MHC locus, many with immune functions 147. Association to the MHC region has been observed in many immune‐mediated complex diseases, such as systemic lupus erythematosus, rheumatoid arthritis, type 1 diabetes, and multiple sclerosis 58.

Several genome‐wide linkage studies have identified a significant signal within chromosomal region 6p and suggested a role for the MHC locus in IBD. From linkage studies, it was observed that the MHC locus appears to carry a significant part of the genetic contribution to CD and UC susceptibility. Over the last 30 years, a number of association studies have reported involvement of the HLA genes with both UC and CD, although the findings have been inconsistent 35, with different HLA loci or alleles showing association to IBD between the multiple studies. This apparent inconsistency can be attributed to studies with small sample size, to the limited number of loci and alleles evaluated in each study (mostly restricted to classical HLA alleles), and to the extensive LD in the region. Despite this, convincing evidence exists that the HLA genes, or other MHC loci in LD with them, are likely both susceptibility genes contributing to disease pathogenesis, and modifier genes contributing to specific disease phenotypes 36,148,149.

In 1999, Stokkers et al. performed a meta‐analysis of 19 UC and 17 CD association studies published between 1980 and 1998 35. Their results are consistent with the previously described positive association results between UC and classical HLA alleles DRB1*1502, DR9, and DRB1*0103, and between CD and HLA alleles DR7, DQ4, and DRB3*0301. Since then, several additional association studies investigating the role of alleles in classical HLA genes have been published supporting some of these observations, and further describing novel association of specific alleles with subclinical phenotypes. In UC, associations to disease risk have been reported reproducibly mainly for two alleles: the uncommon HLA DRB1*0103 allele (showing a prevalence of less than 5%) and the HLA DRB1*1502 allele 33,36,37,150. In contrast, reproducible associations to disease risk in CD have been reported mainly for four alleles: HLA‐DRB1*07, HLA‐DRB1*0103, HLA‐DRB1*04, and HLA‐DRB3*0301. In contrast to CARD15 and IBD5, the association findings of HLA in IBD have been rather inconsistent across studies and await future confirmation.

Among the other genes implicated in immune response in the extended MHC region, the tumor necrosis factor (TNF)‐alpha gene makes an attractive candidate. This gene, located in the class III region, plays a role as a pro‐inflammatory cytokine, and increased levels of TNF‐alpha expression have been observed in inflamed mucosa of IBD patients 151. A recent biologic therapy, which uses recombinant antibodies to target TNF‐alpha, has been shown to be effective in treating CD 152, thus supporting the role of TNF‐alpha in the inflammatory process of IBD. Studies on the TNF‐alpha locus have centered mostly on a limited number of promoter polymorphisms. Several studies have reported association of the C allele of the TNF −857C/T polymorphism to IBD, with some showing association specifically to CD 123,153–156. In addition, the A allele (or TNF2 allele) of the TNF −308G/A variant has been associated to severe stenosing and penetrating UC but not to CD 157–161. This variant has been reported to be associated with increased levels of TNF production, although whether this variant is itself functional or in LD with the actual causative allele is still unclear 162.

Extensive LD within the MHC region has made the identification of a specific susceptibility locus difficult. Over the last few years, several studies have attempted to study the variability of this region using a more systematic SNP‐based approach 163,164. A recent report from de Bakker et al. 165 describes the creation of a high‐density map for studying the variability of the MHC region, and demonstrates that it is possible to not only capture the common variability from classical HLA alleles but of surrounding loci as well. Clearly, applying such a systematic approach in the study of the MHC region in IBD will likely help the identification of underlying causal variations in UC and CD.

Immune‐related loci located outside the MHC region have also been implicated in the development of IBD. Until recently, it was believed that the heterodimeric pro‐inflammatory cytokine IL‐12 played an important role in the development of IBD. This hypothesis was supported by observations of increased levels of this cytokine in the lamina propria of patients with IBD 166,167 and by the suggested role of this cytokine in other chronic inflammatory disorders. In addition, the inhibition of IL‐12 activity through the use of a monoclonal antibody against the IL‐12p40 subunit (IL‐12B), in both experimental models of colitis 168,169 and clinical trials 170, also supported this notion. It has been reported that the IL‐12B subunit is in fact shared with IL‐23 171, a heterodimeric cytokine composed of the IL‐23A (p19) and the shared IL‐12B (p40) subunit. This observation has prompted researchers to re‐evaluate the specific roles of these cytokines in inflammatory diseases, as well as the results obtained from the therapeutic anti‐p40 antibody. Recently, two groups have evaluated the role of IL‐12 and IL‐23 in experimental models of colitis, and found IL‐23 to be an essential effector in the development of chronic intestinal inflammation 172,173. Recently, the first GWA study in IBD, presented by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) IBD Genetics Consortium 174, identified variants in the IL‐23 receptor (IL‐23R) gene as influencing risk to CD. In this report, a highly significant association was detected between an uncommon glutamine variant at amino acid position 381 (R381Q) within IL‐23R and protection from ileal CD. This protective effect was observed and replicated in two independent ileal CD case‐control cohorts, as well as in a familial IBD cohort consisting of mother‐father‐affected child trios. Importantly, the latter study demonstrated that the 381Q allele also protects from development of UC. Further analysis of the association data also supported the existence of multiple additional risk variants in the region, independent of R381Q. This study also examined variants in the genes encoding for the different subunits of IL‐23 (IL‐23A and IL‐12B) and IL‐23R (IL‐12RB1), and found no evidence of association to IBD, suggesting that IL‐23R should be a focus for the development of novel targeted therapies.

Finally, a number of other attractive candidates, involved in immune regulation, have been evaluated in IBD. Association signals have been detected for IL‐1B and IL‐1RA in UC 175–177, IL‐18 160,178 in CD, and for ICAM‐1 179–181 and IL‐10 124,182,183 in IBD, although the significance of these in disease risk has not yet been clearly established, as these have yet to be significantly replicated in independent studies. As mentioned previously, the inconsistency in results obtained from association studies may be a reflection of underpowered studies due to small sample size, alleles of modest effect, clinically and ethnically heterogeneous study populations, and variable prevalence of the alleles in different populations. Some of these limitations can be overcome through the study larger well characterized sample populations.

Potential clinical implications

Current therapies in IBD aim at controlling acute symptoms and maintaining remission in patients. Given the great heterogeneity of disease expression, clinical therapy should ideally be matched to fit each person's needs, based on disease phenotype severity, pathophysiological mechanism underlying disease etiology, success of previous therapy, and observed negative side effects of the medication. It is hoped that, beyond defining disease risk, genetics may provide a window into the understanding of these differences between individual patients.

One potential approach to addressing this question is a direct outcome of disease association studies. Specifically, given the fact that IBD is a complex genetic trait, where different patients will have different subsets of disease alleles, grouping patients based on the presence or absence of specific alleles may help to define categories reflecting the underlying molecular mechanisms. This molecular classification could potentially be less prone to confounding factors, such as treatment regimen and disease duration, that can in some cases plague clinical classification. The ultimate goal of such a classification scheme would be to understand how various genetic and non‐genetic risk factors lead to disease. Currently, this approach is limited given that the full complement of disease genes has not thus far been identified. However, in the interim, alternate approaches, such as genotype‐phenotype analyses can provide some useful information. For example, in the case disease expression, the clinical phenotypes have primarily been limited to age of disease diagnosis (as a proxy for disease onset), location, and behavior. While several reports have suggested associations of allelic variants at different loci with specific disease expression in IBD, only a few have been consistently replicated. The strongest and most consistently replicated genotype‐phenotype correlation in IBD is between the CARD15 variants and early age of diagnosis, stricturing lesions, and ileal involvement 32,55–57. On the other hand, the results regarding the effect of the IBD5 risk haplotype on predisposition to specific clinical phenotypes in CD have been more of a challenge. While some groups have reported a lack of association between IBD5 and specific disease sites 24–26, 28,29, others have reported an association of the IBD5 haplotype with both perianal and/or ileal CD 28,125,184,185. In contrast, Torok et al. 41 reported a novel phenotypic association of the IBD5/OCTN‐TC haplotype with colonic CD, and with non‐fistulizing and non‐stricturing behavior. In the case of the MHC, a strong association has been observed in UC between the HLA DRB1*0103 allele and both extensive and aggressive disease requiring colectomy 33,36,37,150. In addition, an association with DRB1*0103 has also been observed in CD patients with pure colonic disease 32,186,187. The shared association of DRB1*0103 with UC and colonic CD supports the notion of a common molecular basis for the colonic IBD phenotype. Although these studies are mostly observational, looking for a correlation between a gene and a given phenotype, they propose a potential use for genetic information in clinical management of IBD. For example, it has been suggested that CD patients carrying the CARD15 variants could benefit from early therapeutic strategies, while UC patients carrying the HLA allele DRB1*0103 may require a more aggressive management including colectomy. It is clear, however, that more work is necessary, in order to identify additional susceptibility loci and their effect on disease expression, before genetic information can be used routinely to help in the management of IBD.

Another potential clinical application of this genetic knowledge is in directing the development of novel therapies, through the identification of new therapeutic targets. This is illustrated by the recent discovery of the important role of IL‐23 in experimental models of IBD 172,173 and of IL‐23R genetic variants in modifying IBD risk 174. As mentioned previously, a therapeutic antibody against IL‐12B (p40 subunit of IL‐23 and IL‐12) has already shown great promise in controlling inflammation in experimental models 168,169 but ultimately targets both IL‐12 and IL‐23 through their common p40 subunit. A more specific inhibition of IL‐23 signaling in patients could potentially be achieved through the development of a therapeutic antibody targeting the p19 subunit unique to IL‐23. In addition, since genetic studies have suggested the existence of both protective and risk alleles at the IL‐23R locus, it will be interesting to better understand the mechanisms leading to apparently opposite functional effects of different changes in the same molecule. This knowledge could lead to the development of novel therapies, to either mimic the effect of the protective allele or to block the effect of the risk alleles, and to a matching of therapy to genetic status.

In broader terms, it is possible to use the same genetic approaches that are used for disease gene discovery described earlier, in the identification of genetic factors influencing response to treatment in IBD patients, thus potentially allowing for a better use of current therapies. Several classes of drugs are currently used to treat IBD, including aminosalicylates, antibiotics, chemotherapeutic drugs, corticosteroids, anti‐inflammatory drugs, immunosuppressant agents, immunomodulators, and novel biologic agents. As with any other treatment, drug performance and pharmacological activity in IBD are regulated by the principles of ADME (absorption, distribution, metabolism, and excretion), which influence drug levels, kinetics as well as toxicity. The field of pharmacogenomics studies the genetically encoded variability that can influence these parameters, such as modifications in membrane permeability and metabolic enzyme activity, as well as other factors which may influence drug activity, such as modification of drug target expression levels and availability. The ultimate goal of pharmacogenomics is to offer a more personalized approach to disease treatment which takes into account the patient's genetic makeup, subdividing patients into molecular subtypes for therapy. Pharmacogenomics should allow the tailoring of drug regimen and doses in a predictive fashion, rather than by trial and error, in order to increase efficacy, decrease costs, and reduce side effects.

The best example of the relevance of pharmacogenomics in IBD is for the enzyme thiopurine S‐methyltransferase (TPMT) and its influence on the metabolism of thiopurine drugs, such as azathiopurine (AZA). The prodrug azathiopurine (AZA) has been used as an immunosuppressant for many years in the treatment of IBD, where it induces and maintains remission of CD and UC 188. However, toxicity to AZA is a problem. AZA is non‐enzymatically converted to 6‐mercaptopurine (6‐MP), leading to the formation of active thioguanine nucleotides (TGNs). TGNs act as purine antagonists and induce cytotoxicity and immunosuppression by disrupting nucleic acid metabolism and purine synthesis. The enzyme thiopurine methyl transferase (TPMT) is crucial in the metabolic pathway of the drug, where it competes with TGN synthesis for 6‐MP, and catalyses the formation of the inactive methylation product 6‐methylmercaptopurine 189. TPMT activity is regulated genetically 190, with more than 20 variant alleles showing variable frequencies between populations having been associated to decreased enzymatic activity 191–194. The TPMT *3A and TPMT *3C alleles are the most common alleles in Caucasian populations with a carrier frequency of approximately 11% 195. Decreased activity of TPMT leads to toxicity due to increased concentration of TGNs, causing bone marrow suppression and decreased leukocyte counts 196. Current Food and Drug Administration (FDA) recommendations suggest that individuals initiating acute treatment with AZA or 6‐MP, as part of a new treatment regimen, should have their TPMT genotype or phenotype assessed. This test should be performed in an attempt to identify individuals deficient in enzyme activity or homozygous deficiency alleles, in order for these patients to avoid thiopurine drug treatment and its potential adverse effects.

In addition, different loci previously associated with IBD risk have been evaluated for their influence on response to therapy, but their role in modulating IBD treatment remains controversial. For example, glucocorticoids, used in the treatment of IBD, are known substrates for MDR1, the multidrug resistance gene. Elevated levels of MDR1 expression in peripheral blood lymphocytes and intestinal epithelial cells of CD patients who required bowel resection and of UC patients who required proctocolectomy following failure to respond to therapy have been reported 197. Also, it was reported that carriage of the 113A allele of DLG5 was independently associated with resistance to steroid treatment, where a direct influence of the DLG5 variant on intestinal permeability was suggested 198. Finally, Urcelay et al. studied the possible effect of IBD5 on the response to a monoclonal antibody against tumor necrosis factor‐alpha (TNF‐alpha) in 40 CD patients 124. Homozygosity for IBD5 risk haplotype was significantly associated with lack of response to treatment compared to the other genotypic groups. Replication of these various preliminary studies in larger cohorts will have to be performed before any conclusions can be drawn.

While the application of genetics discoveries to the clinical management of IBD is still in its early stages, the future is looking positive. Currently, most studies have been limited to examining the effects of identified alleles on clinical expression of disease, as well as response to treatments. However, the genetic approaches and knowledge, developed for the identification of genetic factors associated to disease risk, are only now being applied to pharmacogenomics. The availability of large collections of relevant patient populations for such studies will likely yield many more true associations between novel genetic variants, clinical expression of disease and response to treatment.

Discussion

Genetic research of complex traits, such as IBD, aims at identifying the genetic factors underlying disease risk, in order to better understand the molecular pathways and pathophysiological mechanisms involved in disease etiology. One of the major goals of such research is to devise a molecular classification system for defining biologically homogeneous subgroups of patients to help in predicting clinical expression of disease. Ultimately, such a classification system would help clinicians in the management of disease, directing the course of clinical treatment through a better match of current therapies to patients, as well as through the identification of new therapeutic targets and the subsequent development of novel therapies.

In the last few years, great advances have been made in the understanding of the genetics of IBD. Genome‐wide linkage studies, as well as follow‐up association mapping and candidate gene association studies have identified a number of susceptibility loci, some of which have been confirmed by multiple replication studies. While the advances in genetic risk factor identification in IBD have been impressive, work still remains to explain a large proportion of the genetic risk associated to disease. So far, most of the association findings have been related to CD, with only the MHC region showing strong reproducible associations to UC. Examination of the currently known genetic risk factors supports the existence of at least three different pathophysiological mechanisms. As described, these mechanisms include inappropriate regulation of innate immune response to enteric microflora or pathogens, increased permeability across the mucosal epithelial barrier and defective regulation of the adaptive immune system. However, it is clear that a vast amount of work still remains to be done in order to establish the link between any individual risk factor and a given pathophysiological mechanism, as well as to identify the remaining genetic and non‐genetic risk factors influencing disease risk. Nonetheless, the early work to link genetic risk factors to clinical expression of disease has revealed some potential molecular subgroups of disease. For example, CARD15 carriers have been more associated with the development of ileal disease in CD.

The identification and replication of multiple risk loci for IBD has placed the genetic analysis of these diseases in a unique position to evaluate the difficulties in mapping loci to complex disease and to illustrate some of the current limitations of existing genetic studies of complex traits. These limitations can be broadly divided up into the difficulty of detecting any individual risk allele, the difficulty in establishing the difference between true and false association results, and the challenge of determining the link between the biological variations, under control of the genetic variation, and the disease mechanism.

The difficulties of the early genetic studies of complex traits in detecting risk alleles have stemmed from the lack of power to examine the entire genome and from the limitations in our knowledge of common genetic variation. More specifically, genome‐wide linkage studies do not have the resolution or statistical power to identify all loci, and the more powerful association studies performed have had a limited focus in terms of the chromosomal regions and candidate genes studied, mostly restricted by the identified genetic risk regions from linkage studies and by our current knowledge on disease mechanisms. In addition, a majority of the candidate gene association analyses have not evaluated the complete common genetic variation of the locus or region under study, but only a few variants suspected to be of functional importance. The advent of the International HapMap Project (which has collected information about greater than 5 million common genetic variants across the genome) and the creation of appropriate analytical tools for selecting informative SNPs enable us to take a more comprehensive look at any given region of the genome. Furthermore, recent advances in genotyping technologies have now made it possible to perform GWA studies involving 100,000 to 500,000 SNPs and providing the potential to query the entire genome in a powerful manner 199. Because, GWA studies are not hypothesis‐driven, they offer an unbiased look at the genome and will most likely identify novel pathways and mechanisms involved in disease pathogenesis. While genome‐wide association studies offer a wider coverage of the common genetic variation, for equal sample size they are less powerful than the more targeted studies (due to the difference in statistical correction for multiple testing e.g. 500,000 tests in GWA versus a few dozen in targeted studies). In the end, it will likely be necessary to use a combination of both genome‐wide and targeted approaches in order to identify the majority of IBD risk alleles.

The capacity to distinguish between true and false‐positive associations is a challenge for both targeted and genome‐wide association studies of complex diseases, especially when dealing with alleles of modest genetic effect. Part of this challenge can be met by replication studies performed in sufficiently large and well phenotyped patient cohorts, and properly matched controls. The need for such large patient cohorts has been the motivation for the initiation of numerous national and international collaborative efforts to pool together available resources. The study of these large cohorts holds great promise in distinguishing true associations from spurious results, as well as in evaluating the possible interactions between these risk alleles, and between the genetic and non‐genetic risk factors on clinical expression of disease.

Furthermore, the final link between genetic variation and disease susceptibility will need to be met by innovative and comprehensive functional screens, taking into consideration not only the link between genetic variation and biological variation, but also linking this biological variation to a mechanism leading to disease susceptibility. To meet these goals, it will be necessary to use a combination of cellular and animal models, as well as to identify a clinically relevant link to disease pathophysiology in patients.

The discoveries made through genetic studies will undoubtedly identify other pathophysiological mechanisms than those described herein, providing hope of an improved understanding of disease etiology. Ultimately, the application of the current genetic knowledge, analytical approaches, and technology, developed for the study of genetic risk factors, to the field of pharmacogenetics will allow for the identification of the different genetic factors affecting variable drug responses in IBD patients. The identification of genetic determinants affecting disease risk, as well as clinical course and response to treatment will greatly help in disease clinical management and patient care, allowing the introduction of personalized approaches to treatment of patients with IBD based on genetic profile.