“Return to sender, address unknown.…” Are variant addressins involved in PSC?
Article Outline
The past decades have witnessed an explosion in the identification of genes associated with relatively rare mendelian diseases. With the current DNA amplification and sequence technology and the availability of sufficient affected families it is feasible to find the responsible gene provided that the disease is monogenic (i.e., caused by a single gene) and that its phenotype is unambiguously defined (i.e., the patient group is really homogeneous and not composed of patients with similar but genetically different diseases). The principles of this successful strategy are also applied to more complex and more common diseases, based on the idea that such diseases also have a (partial) genetic background. However, there are many pitfalls in such screening strategies. In the present issue Bowlus et al. [1] report negative results that contrast a previously described association between addressin genes and the occurrence of PSC.
In genetic association studies screening for SNPs (single nucleotide polymorphisms) has become a highly valuable tool. The human genome contains about 6–12 million SNPs [2], which are essentially single nucleotide changes in the genome that occur with a certain frequency (>1%) in the population. The great majority of these SNPs have no effect on health because they are not in the coding region of a gene or, if they are present in the coding sequence they may not alter the function of the encoded protein. Yet, these SNPs are useful because they can be used as markers for a potential disease gene. A set of SNPs in a certain region of the chromosome constitute a haplotype which can be used as a “barcode” for a certain (unknown) mutation in that region. Analysis of the linkage between the haplotype frequency in a disease group vs. controls can elucidate whether a gene in that region plays a role in this particular disease or not. A minority of SNPs represent changes that directly affect function of a gene product. As a consequence such SNPs potentially cause or aggravate a disease and these SNPs can be used directly in genetic screening studies.
The search for an association between a disease phenotype and certain SNPs or haplotypes is complicated if several or even many loci play a role in disease initiation or development. The higher the number of loci, the more patients and individuals will have to be screened for a polymorphism or haplotype to detect a significant association. Such searches are even more hampered if a disease phenotype is not unequivocally defined and the patient group in fact consist of patients with similar symptoms but a different disease etiology.
Primary sclerosing cholangitis is a disease of intra- and extrahepatic bile ducts, characterized by concentric obliterative fibrosis and bile duct strictures. It is an immune-mediated disease, rather than a classical autoimmune disorder. There is a clear association with certain HLA haplotypes, but no specific autoantigens are observed. Rather, a wide range of autoantibodies are observed of which anti-neutrophil antibodies are the most frequent. This contrasts for example with Primary Biliary Cirrhosis in which a specific autoantigen (pyruvate dehydrogenase complex, PDC) appears to play a key role.
The primary trigger in the pathogenesis of PSC is unknown, but it is clear that the affected bile ducts have a T-cell predominant infiltrate [3]. PSC is strongly associated with inflammatory bowel disease (IBD) which has a similar immune-mediated background. More than 70% of patients with PSC also suffer from IBD. Any model of the development of PSC needs to account for the fact that the liver disease usually progresses independently from inflammation of the bowel. A common underlying derangement of the immune system would provide such an explanation.
Grant et al. [4] proposed a model in which effector T-cells, generated in lymphoid tissues of the gut during IBD, persist as long-lived memory cells that recirculate through the liver and, under the appropriate conditions, trigger hepatic inflammation even in the absence of simultaneous inflammation of the gut. For this to happen effector T-cells not only need to recognize a common (auto)antigen but also need to be able to home to the liver, as they do in the gut. Hence, SNPs in genes that play a role in recruitment of T-cells to both tissues may influence the progression of the disease.
Infiltration of lymphocytes is a complex process that involves many steps including the release of chemokines, rolling of lymphocytes on the endothelium and transendothelial migration. The initial recognition of the appropriate endothelium depends on tissue-specific expression of receptors that bind ligands on the lymphocyte. Thus, homing of activated T-cells to the gut appears to depend on expression of α4β7 integrin on the T-cell membrane, which interacts with MadCam-1 on the gut endothelium. This is highlighted by the fact that homing of T-cells to gut lymphoid tissue is strongly reduced in β7 knockout mice [5]. Although it was thought that α4β7 integrin expression was specific for the gut, Grant et al. [6] demonstrated that this “tissue address” is also present on the endothelium of chronically inflamed livers (in autoimmune hepatitis, PSC and PBC).
ICAM-1 is a less tissue-specific receptor for αLβ2 integrin (LFA-1) on lymphocytes, but its expression is induced during inflammation. Indeed, ICAM-1 was shown to be expressed on the biliary epithelium of PSC patients [7]. These observations indicate that both ICAM-1 and MadCam-1 potentially play a role in homing of T-cells to hepatic and intestinal sites of inflammation. Even if expression of these addressins is secondary to the inflammatory process, alterations in the genes encoding these proteins may influence the extent of the T-cell recruitment and therefore the progress of the disease. This justifies a search for SNPs in these genes and their association with diseases like PSC.
Against this theoretical background Yang et al. [8] studied the occurrence of two genetic variants of ICAM-1 in a group of 104 PSC patients and 213 healthy controls (matched for age and ethnic background). The variants under study were the G241R (in which a glycine at position 241 is replaced by arginine) and K469E (replacement of lysine by glutamate at position 469). Both residues are present in domains of the proteins that are thought to be important for binding to their ligands, but, importantly, biological effects of neither of the SNPs have yet been directly analyzed. Within the group of PSC patients 12% were homozygous for 469E variant while this was 24% in controls and this difference was considered to be significant (p
=0.02). From this observation the authors suggested that the 469E variant provides a protective effect against PSC. The authors found no significant association between progression of the disease and the occurrence of the 469E variant. There was also no association of the genotype with the type of coexistent IBD. This is striking because the same group observed an association between this SNP and CD [9]. However, in another study [10] the frequency of the same variant was studied in a group of IBD patients vs. controls and no association was found.
In this issue of the Journal, Bowlus et al. [1] report on analysis of a much larger group of 365 PSC patients, 327 patients with ulcerative colitis and 368 healthy controls. They studied the frequency of both ICAM-1 variants mentioned above, as well as 7 polymorphisms in MadCAM-1. They found no difference in the frequency of any of these alleles between any of these groups. Hence, they conclude that the neither the G241R nor the K469E variants of ICAM-1 play a role in PSC development and this also holds for the SNPs MadCAM-1. With the 7 SNPs the authors also analyzed haplotype association of MadCAM-1 with PSC (and UC); they found no association suggesting that no polymorphism in this gene is associated with these two diseases.
Genetic association studies are plagued by lack of reproducibility. Hirschhorn et al. [11] performed a comprehensive review on this subject and observed that out of 166 putative associations that had been studied three or more times, only six could be replicated at least 75% of the time. This worrisome lack of reproducibility can be caused by false-negative and false-positive associations, the latter being the most frequent. False-positive associations can be caused by hidden stratification within the case or control group, by statistical fluctuation, or by the inappropriate use of p-values of 0.05 as a threshold criterion for success. Newton-Cheh and Hirschhorn [12] have argued that association studies should utilize prior probability estimates on basis of the type of association that is studied. This involves estimation of the number of candidate genes involved in a certain disease, the odds ratio of an involved gene and whether a gene represents a major or minor determinant. They calculated that in the best case scenario (a top candidate gene in a polygenic disease with an odds ratio lower than 1.5) one should use a p<0.003 rather than 0.05. On the basis of this approach they also calculated that when a p<0.05 is used in such a study with 200 cases and 200 controls, the chance of finding a true-positive association is 1.7%. If a p<0.001 is applied this chance increases to 26%. If the same study involves 2000 cases and 2000 controls the chance of true positivity increases to 4.9% (with p<0.05%) and 60% (with p<0.001). These calculations demonstrate that conclusions from small association studies are much more difficult to make than generally accepted.
As far as the association of ICAM-1 variants with PSC is concerned, the study of Yang et al. [8] is discrepant with that of Bowlus et al. [1]. As discussed above studies of this type often tend to produce false positive results. Stratification may not be a cause of false positivity in this case as both studies were performed with Caucasians exclusively, although hidden stratification may still be an issue. Other possible causes are improper case control matching, multiple testing and, last but not least, mere chance. The latter is the most important and seems a possible factor in the study of Yang et al. [8] as well. Their positive association was based on a relatively high p-value (0.02), which is obvious because of the limited number of individuals enrolled. This leaves the considerable possibility that the association was found by chance. The present report by Bowlus et al. [1] did not observe an association and although the number of patients was still limited, it was larger than that in the study of Yang et al. [8].
It may be clear that the finding of an association between a genetic variant and a disease phenotype must be interpreted with great caution and must always be replicated dependent and sufficiently large patient groups in order to rule out the possibility that an association was found by chance. In this context it must be advocated that reports on replication of a reported genetic association are considered for publication even if their results are negative.
References
- Bowlus CL, Karlsen TH, Broomé U, Thorsby E, Vatn M, and the IBSEN study group, et al. Analysis of MAdCAM-1 and ICAM-1 polymorphisms in 365 Scandinavian patients with primary sclerosing cholangitis. J Hepatol 2006; 45:704–10.
- . Definition and clinical importance of haplotypes. Annu Rev Med. 2005;56:303–320
- . Primary sclerosing cholangitis. Autoimmun Rev. 2003;2:305–312
- . Homing of mucosal lymphocytes to the liver in the pathogenesis of hepatic complications of inflammatory bowel disease. Lancet. 2002;359:150–157
- Critical role for beta7 integrins in formation of the gut-associated lymphoid tissue. Nature. 1996;382:366–370
- . MAdCAM-1 expressed in chronic inflammatory liver disease supports mucosal lymphocyte adhesion to hepatic endothelium (MAdCAM-1 in chronic inflammatory liver disease). Hepatology. 2001;33:1065–1072
- . Adhesion molecule expression in primary sclerosing cholangitis and primary biliary cirrhosis. Gut. 1995;36:604–609
- . Susceptibility to primary sclerosing cholangitis is associated with polymorphisms of intercellular adhesion molecule-1. J Hepatol. 2004;40:375–379
- Inflammatory bowel disease is linked to 19p13 and associated with ICAM-1. Inflamm Bowel Dis. 2004;10:173–181
- . Intercellular adhesion molecule 1 gene associations with immunologic subsets of inflammatory bowel disease. Gastroenterology. 1995;109:440–448
- . A comprehensive review of genetic association studies. Genet Med. 2002;4:45–61
- . Genetic association studies of complex traits: design and analysis issues. Mutat Res. 2005;573:54–69
PII: S0168-8278(06)00441-7
doi:10.1016/j.jhep.2006.08.003
© 2006 Published by Elsevier Inc.
