From genotypes to haplotypes in hepatobiliary diseases: One plus one equals (sometimes) more than two

Primary biliary cirrhosis (PBC) and primary sclerosing cholangitis (PSC) are chronic cholestatic liver diseases, which often lead to liver cirrhosis and its inherent complications, such as portal hypertension and liver failure. Although major advances in the understanding of the molecular pathomechanisms have been made, more studies are needed to define the immunologic and genetic factors involved in these potentially fatal disorders.1, 2 Evidence for a genetic susceptibility are the associations of both diseases with alleles of the major histocompatibility complex3, 4 and the raised familial risk.5 However, family clustering could also be due to shared environmental factors or gene × environment interactions.5 Until recently, no twin studies supporting the genetic component were available, but a preliminary report suggests a concordance rate of 75% for PBC in monozygotic twins, consistent with a significant genetic contribution to disease susceptibility.6 Because autoimmune mechanisms are thought to be crucial in the pathogenesis of both diseases, investigators have focused on genetic variants (polymorphisms) in non-HLA candidate genes involved in immune regulation.7-15 In addition, it has been reported that a gene variant of the matrix metalloproteinase 3, which is involved in extracellular matrix degradation and liver fibrogenesis, is associated with susceptibility and disease progression in PSC.16 These examples indicate that not only genes regulating innate and acquired immune functions play a role in the pathogenesis of PBC and PSC, but that the systematic identification of additional genes modulating disease susceptibility and/or progression can be awaited. For both diseases it is not yet clear whether immune-mediated, toxic and/or vascular impairment of the bile-secretory apparatus represents a prerequisite for liver injury. The secretion of bile normally depends on the integrity of an ensemble of membrane transport systems in hepatocytes and cholangiocytes.17 Of note, the transport of all three major biliary lipids (i.e., bile salts, phosphatidylcholin, and cholesterol) across the canalicular membrane is mediated by ATP-dependent export pumps, known as ATP-binding cassette (ABC) transporters.18 Two of the major transporters are the "multidrug resistance" P-glycoprotein 3 (MDR 3/ABCB4), which translocates phosphatidylcholine (lecithin), and the "bile salt export pump" (BSEP/ABCB11). Mutations in the genes encoding these hepatobiliary transporters have been identified as causative in progressive familial intrahepatic cholestasis in children.17 One study reported an association between PSC and functional variants of the cystic fibrosis gene (CFTR/ABCC7),19 yet another member of the ABC transporter family that is expressed on the apical membrane of biliary epithelial cells. Of note, mice with homozygous disruption of the MDR3 (ABCB4) ortholog (Mdr2) develop liver lesions that are characterized by segmental biliary strictures due to periductal fibrosis and obliteration of bile ducts, thus resembling sclerosing cholangitis.20, 21 The liver injury in these mice is generally attributed to the phospholipid deficiency of their bile, which does not contain mixed lecithin-cholesterol-bile salt micelles that normally protect the biliary epithelium against the detergent properties of bile salts.20, 21 In the light of these findings, ABCB4 and ABCB11 represent promising candidate genes, which could also modify the cholestatic injury in PBC and PSC. Therefore, the investigation of the role of BSEP (ABCB11) and MDR3 (ABCB4) gene variants in patients with PBC and PSC by Pauli-Magnus et al. in this issue of HEPATOLOGY22 is a timely candidate gene association study. The study design is based on the expectation that there will be higher frequencies of the contributing genetic components in patients than in a group of comparable subjects without the disease. Although association studies are prone to selection bias and spurious allelic association,23, 24 there is growing evidence that well-designed and well-analyzed studies with appropriate controls are an important tool to dissect the genetic basis of complex diseases. One of the important achievements of the investigation by Pauli-Magnus et al.22 is the definition of the complete haplotype structures of ABCB4 and ABCB11 by sequencing the promoters, the coding regions, and the flanking intronic parts of the corresponding genes in a large number of individuals. Of note, the use of haplotype analysis contrasts to the majority of non-HLA association studies performed in the field of cholestatic liver diseases so far. But why does analysis of haplotype structures seem to be so important? Nearly every genetic marker, i.e., a variable site within the genome, results from a single historical mutational event and is therefore initially associated with the other alleles nearby that happened to be present on the particular chromosomal segment. A haplotype is defined as such a specific set of alleles present on a single chromosome. Of note, in many chromosomal regions only few (<10) common haplotypes are observed, and these account for most of the genetic variation of this locus among individuals. In complex diseases, haplotype analysis can provide more information than the analysis of individual polymorphisms.25 In some cases an association of a complex phenotype might even be missed if only polymorphisms are analyzed instead of haplotypes.26 On the other hand current high-throughput genotyping methods, when applied to DNA from diploid individuals, determine only genotypes but do not provide haplotype information, i.e., the combinations of alleles that are present on one of the two chromosomes (Fig. 1). Haplotype information can be obtained experimentally with allele-specific PCR or through genotyping of family members. However, despite familial aggregation of PBC and PSC, systematic linkage (family) studies for these diseases have yet to be performed. Recently, algorithms have been developed that are able to statistically infer haplotypes from genotype data in such association studies (Fig. 1). With the current algorithms, as the Bayesian approach employed in the study by Pauli-Magnus et al.,22 inference of haplotype phase from population samples can be done with reasonable accuracy, provided the sample size is sufficiently large and the rate of intraregional recombination is sufficiently low.27 After statistical resolution of haplotypes, their frequencies in healthy and diseased individuals are compared with conventional contingency-table statistics, or preferably with permutation tests.27 From genotypes to haplotypes. In genetic association studies, patients are genotyped for gene variants (polymorphisms). The left panel of this schematic diagram displays the genotypes of five individuals with respect to four polymorphisms. Blue and red boxes indicate the two different alleles of the polymorphisms; yellow boxes represent heterozygous genotypes. The diagram illustrates that the reconstruction of haplotypes (right panel) increases the statistical power of an association study by doubling the sample size and simplifying the data structure. Haplotypes represent specific allele combinations on a single chromosome; thus, the combination of two haplotypes generates the genotypes of an individual. In this example, only two different allele combinations (right panel) underlie the complex genotype pattern (left panel). In their work, Pauli-Magnus et al.22 were able to identify 45 variant sites within the ABCB4 gene and 46 ABCB11 variant sites, all of which were in Hardy-Weinberg equilibrium. The overall allele frequencies of these gene variants were similar in healthy individuals, 76 subjects with PBC, and 46 patients with PSC. This is consistent with a recent analysis by Rosmorduc et al.,28 who determined 8 individual SNPs in 34 French PSC patients. In both the French study and the study by Pauli-Magnus et al., a few single nucleotide polymorphisms (SNPs) leading to nonsynonymous amino acid substitutions were specifically found in one or the other groups, but due to the limited number of patients the importance of these markers cannot be finally assessed. Although the study by Pauli-Magnus et al.22 does not intend to address higher-order risks, i.e., epistasis (gene × gene) or gene × environment interaction, two haplotypes were significantly associated with disease phenotypes: MDR3_3 (denoting the third most common MDR3 haplotype) was found more frequently in patients with PSC, while the rare haplotype BSEP_10 was detected at a higher frequency in PBC as compared to control subjects. Furthermore, the authors could link quantitative traits, like total bilirubin levels or the Mayo risk scores, to the presence of a distinct BSEP haplotype.22 It is worth noting that these haplotypes are not obligatory disease-associated or causative, but they could represent modifiers of cholestatic liver injury whose influence depends on the genetic background and environmental factors. Albeit the contribution of a specific haplotype to multifactorial and complex traits is expected to be small, the findings might still be pathophysiologically relevant and guide future research to elucidate the genetic component of hepatobiliary liver diseases. Would it be necessary to genotype all polymorphisms included in a specific haplotype to replicate the findings by Pauli-Magnus et al.22 or to extend their results to a larger cohort of patients? Fortunately, this does not seem to be the case. Due to linkage disequilibrium,29, 30 the number of distinct combinations of SNP alleles encountered in human DNA samples is only a small fraction of the theoretical number of haplotypes that would be possible if all alleles were distributed randomly. This haplotype structure of the genome is the basis for selection of a parsimonious set of SNPs, called "haplotype-tagging" SNPs (htSNPs), which represent the haplotype variation in a population. Such htSNPs can either be identified by eye or preferentially with validated computer algorithms.31 Thus, the haplotype blocks of the genome significantly reduce the number of SNPs required to extrapolate the genetic information in a given sample.30 The identification of htSNPs renders large-scale association studies with whole genome approaches (as opposed to candidate gene approaches) possible in the near future. This rationale is behind the International "HapMap" Project29 that produces a genome-wide haplotype map that can be used to streamline association studies in different populations with ancestry from Africa, Asia, and Europe. In summary, the study by Pauli-Magnus et al.22 provides the framework for further investigation of the genetic basis of polygenic hepatobiliary diseases. Although variations in ABCB4 and ABCB11 do not appear to be causative in PBC or PSC, the identified haplotypes might modify the pathogenesis and the clinical course of hepatobiliary liver diseases. The completion of the HapMap project,29 combined with the ongoing large-scale SNP identification projects, will provide the tools for genome-wide association studies and fine mapping of susceptibility loci. However, at this stage it becomes obvious that the bottleneck of future association studies (i.e., "translational research") is the recruitment of larger numbers of clinically accurately characterized patients. Specific MDR3 (and BSEP) haplotypes could help to characterize clinical subtypes, just as they have proven to be valuable for defining the low phospholipid-associated cholelithiasis (LPAC) syndrome as a subgroup of cholelithiasis in general.21, 32 As genotyping studies become more comprehensive and include larger numbers of polymorphisms, they call for more clinical studies to provide resources to define the genetics of hepatobiliary diseases like PBC and PSC.