Potential conflict of interest: Nothing to report.
Twin and family studies suggest there is a significant genetic component to primary biliary cirrhosis (PBC). However, the inability to replicate reported associations has been a recurring problem, with the only consistently reported genetic association that between PBC and HLA-DRB1*0801. However, recently even this has been questioned, and a number of novel associations have also been reported. We reinvestigated HLA class II DRB1, DQA1, and DQB1 alleles and haplotypes in a total of 492 well-characterized PBC patients, 412 from the United Kingdom and an additional 80 patients from northern Italy. There was a clear and significant association with HLA-DRB1*0801 in both groups of patients compared to population-specific healthy controls (12% versus 4% in the UK patients, P = .00087, OR = 3.05; and 18% versus 6% in the Italian patients, P = .021, OR = 3.15). There were also significant protective associations with DRB1*11 in the Italian patients (28% versus 47%, P = .0071, OR = 0.42), but not in the UK patients (8% versus 8%) and a protective association with DRB1*13 in both series (14% versus 20%, P = .042, OR = 0.65 in the UK patients; and 10% versus 31%, P = .00092, OR = 0.25 in the Italian patients). In conclusion, a complex relationship exists between HLA and PBC, and some genetic associations may be population specific. (HEPATOLOGY 2006;44:667–674.)
Primary biliary cirrhosis (PBC) is an autoimmune liver disease characterized by immune-mediated damage to the biliary epithelial cells lining the small intrahepatic bile ducts.1 Clustering of disease in geographic regions2 and within families,3–6 together with the recent reports of 75% concordance in monozygotic twins7 and a sibling relative risk (λs) of 10.5 for PBC,3 all suggest there may be a significant heritable component to the disease.8–10
The role of host genes in the pathogenesis of PBC remains poorly understood, though female sex, specific HLA alleles,11–23 and other immunoregulatory genes8, 12, 24–27 are all thought to be important. However, as with many non-Mendelian “complex” diseases, unraveling the genetic basis of PBC has proved very difficult, with failure to replicate reported associations a recurring problem. The most consistent reports refer to the associations with HLA, yet even these are controversial.11–23
The most frequently described HLA association in PBC is that with the DRB1*08 family of alleles; DRB1*0801 in Europeans12–14, 22, 23 and North Americans of European descent,11, 15, 20 and DRB1*0803 in Japanese.16–19 In Europeans the association, though strong in relative terms, accounts for a minority of patients.11–15, 20, 22, 23 More recent studies have described novel associations in PBC, and one report suggested the DRB1*0801 association, reported in North America (United States and Canada),15, 20 England,13, 14, 23 and Germany,12 was not present in northern Italy.21 Among the novel findings reported recently are a strong protective association with DRB1*1121, 28 and DRB1*13.28 These latter findings have yet to be replicated in PBC patients from northern Europe.
A second controversy surrounds our own observations that the HLA-DRB1*08 allele is more common in patients with late-stage disease than in patients with early-stage disease.23 PBC is a phenotypically heterogeneous disease29; some patients remain asymptomatic, with only mild histological changes,29, 30 whereas others progress rapidly to cirrhosis and end-stage disease.29 We previously hypothesized that case mix could be a factor in the variation in reported strength of this association,23 further compounded by the effects of treatment on disease progression and the small sample sizes in subgroup analyses.
In the present study we present data for a total of 492 PBC patients, 412 from the United Kingdom (UK) and a smaller group from northern Italy. The aims of the study were to investigate the recent reports of associations with DRB1*11 and DRB1*13 from Italy in UK patients and to confirm or refute the reported absence of a DRB1*0801 association in Italian patients. Further, we hoped to both identify population-specific differences in the distribution of HLA DRB1, DQA1 and DQB1 alleles in PBC and reexamine the relationship between specific HLA alleles, genotypes, and haplotypes and disease progression in PBC. The data presented illustrate the value of looking at different populations to identify and confirm genetic associations, the importance of large series in such analyses, and the implications for understanding investigations of HLA in other genetically complex liver diseases.
PBC, primary biliary cirrhosis.
Patients and Methods
Subjects in UK Series.
Four hundred and twelve well-characterized PBC patients from 2 major UK centers and 236 geographically and racially matched controls were studied. All patients and controls were of northern European ancestry. Thirty-three were male (8%) and 379 female (92%).
Subjects in Italian Series.
For comparison, 80 well-characterized PBC patients and 95 geographically and racially matched controls from a single center in northern Italy were also studied. All were of European ancestry and resided in the Padova area of northern Italy. Eight patients were male (10%) and 72 female (90%).
All patients had definite disease defined as all of the 3 standard criteria: (1) liver histology diagnostic of or compatible with PBC, (2) cholestatic liver function tests, and (3) positive serum antimitochondrial antibody titer ≥ 1:40 detected by immunofluorescence. Subjects were excluded from the study if their biopsy (or any other clinical data) suggested additional, potentially confounding causes of liver pathology.
The liver biopsies of 246 patients (173 UK patients and 73 Italian patients) were reviewed to confirm diagnosis in order to determine stage of disease. Patients were classified as having advanced-stage (late) disease, that is, Scheuer stage III or IV, or early-stage disease, that is, Scheuer stage I or II.31 The most recent liver biopsies of 163 patients (67%) showed histologically advanced disease (Scheuer stage III or IV).
All subjects and controls gave informed consent, and the study was cleared by the relevant local hospital ethics committees. DNA samples were labeled and stored by code only and analyzed without prior knowledge of individual identities.
Determination of HLA DRB1, DQA1, and DQB1 Genotypes.
HLA genotyping of all 492 PBC patients and 331 controls was performed by standard polymerase chain reaction protocol for a total of 32 HLA DRB and DQB alleles or groups of alleles. In addition, HLA genotyping of 255 PBC patients (175 from the United Kingdom) and all 331 controls was performed for 10 DQA1 alleles or groups of alleles. For each locus a pair of DRB-, DQA-, or DQB-specific primers corresponding to the second exon sequence was used (Table 1) to amplify approximately 100 ng of genomic DNA in a 50-μL reaction mix comprising 200 μmol/L each of dATP, dCTP, dGTP, and dTTP (Amersham Pharmacia-Biotech, St. Albans, UK); 1.5 mmol/L MgCl2; 10 mmol/L Tris-HCl (pH 8.3); 50 mmol/L KCl; 0.01% gelatin; 1 μmol/L of each primer; and 2-2.5 U Taq polymerase (Perkin Elmer, Norwalk, CT) on a Perkin-Elmer GeneAmp 9600. PCR cycling parameters were as follows: 94°C for 120 seconds, 94°C for 10 seconds, 56°C for 60 seconds (10 cycles), 94°C for 30 seconds, 52°C for 30 seconds, 72°C for 45 seconds (23 cycles), and a final extension at 72°C for 300 seconds.
Table 1. Sequences of HLA Class II Primers
5′-CCC CAC AgC ACg TTT CTT g-3′
5′-CCg CTg CAC TgT gAA gCT CT-3′
5′-CAT gTg CTA CTT CAC CAA Cgg-3′
5′-CTg gTA gTT gTg TCT gCA CAC-3′
5′-ATg gTg TAA ACT TgT ACC AgT-3′
5′-TTg gTA gCA gCg gTA gAg TTg-3′
Following amplification, PCR amplicons were denatured and dot-blotted on a series of positively charged nylon membranes (20 membranes for DRB1, 16 for DQB1, and 10 for DQA1). Each membrane was hybridized with one of a series of digoxigenin-labeled allele- and sequence-specific oligonucleotide probes (SSO). Alleles were detected by chemiluminescence and assigned by 2 trained individuals according to probe specificity tables supplied by the British Society of Histocompatibility and Immunogenetics adapted from the 11th International Histocompatibility Workshop and Conference.32
Allele and genotype distributions were compared with the χ2 test using the EPISTAT statistical analysis program (Centers for Disease Control, Atlanta, GA) as appropriate. No correction factor was necessary as only a priori-identified alleles/haplotypes were found to be significant. Associations with alleles were tested by counting individuals positive for each allele as suggested by Svejgaard and Ryder,33 and associations with haplotypes by counting the number of chromosomes. Odds ratios (ORs) are given instead of relative risk as the accepted standard.
Distribution of HLA-DRB1 Alleles.
There were 3 significant differences in DRB1 allele distribution (Table 2). First, both patient populations had significantly higher DRB1*0801 frequencies than their population-specific healthy controls: 12% of UK patients versus 4% of controls (P = .00087, OR = 3.05); 18% of Italian patients versus 6% of controls (P = .021, OR = 3.15). Second, both patients populations had significantly lower frequencies of DRB1*13 alleles compared to controls, though the difference in UK patients was only of borderline significance: 14% of UK patients versus 20% of controls (P = .042, OR = 0.65); 10% of Italian patients versus 31% of controls (P = .00092, OR = 0.25). Third, and in contrast to the observations above, a strong protective association with DRB1*11 was noted but only in the Italian patients (28% versus 47%, P = .0071, OR = 0.42).
Table 2. Distribution of DRB1, DQA1, and DQB1 Genotypes in Those with PBC Versus Controls for Both UK and Italy
Allele Family and Population
Patients N (%)
Controls N (%)
Probability and Odds Ratio (95% Confidence Interval)
Overall, there was only one novel association at either the DQA1 or DQB1 locus. In both the UK and Italian populations the DQB1*0301 allele appeared to encode a reduced risk of PBC (OR = 0.7 and 0.49, respectively, Table 2). This allele is very commonly but not exclusively found in DRB1*11 haplotypes.
The only other significant associations at DQA1 or DQB1 were with DQA1*0401 and DQB1*0402, both of which preferentially cosegregate with DRB1*0801. These were significant in both patient series (Table 2, DQA1*0401: OR = 3.59 in UK patients, 2.67 in Italian patients; DQB1*0402: OR = 3.02 in UK patients, 3.42 in Italian patients).
Interestingly there were no associations with DQA1*0102 or any other DQA or DQB alleles that preferentially cosegregate with DRB1*13.
Distribution of HLA DRB1, DQA1, and DQB1 Haplotypes in Patients and Controls.
In the UK series there was only one significant haplotype (Table 3): DRB1*0801-DQA1*0401-DQB1*0402 [44 of 824 (5.3%) patient chromosomes versus 9 of 472 (1.9%) control chromosomes, P = .0027, OR = 2.9]. Though there was a weak protective effect of DRB1*13 alleles in UK patients, no particular DRB1*13 haplotype could account for this. In contrast, in the smaller Italian series there were 3 significant haplotypes (Table 3): DRB1*0801-DQA1*0401-DQB1*0402 [16 of 160 (10%) patient chromosomes versus 6 of 190 (3%) control chromosomes, P = .0086, OR = 3.41], DRB1*13-DQA1*0103-DQB1*0603 [3 of 160 (2%) versus 12 of 190 (6%), P = .0041, OR = 0.28], and DRB1*11-DQA1*0501-DQB1*0301 [21 of 160 (13%) versus 46 of 190 (24%), P = .0086, OR = 0.47]. The first and last haplotypes are highly conserved in European populations, and these findings therefore are not surprising. However, the second haplotype is only one of several possible DRB1*13 haplotypes.
Table 3. Haplotype Distribution in PBC Patients and Controls
Allele at Each Locus
Probability and Odds Ratio
UK (n = 824) Italian (n = 160)
UK (n = 472) Italian (n = 190)
P = .0027, OR = 2.9
P = .0086, OR = 3.41
P = .0086, OR = 0.47
P = .0041, OR = 0.28
Distribution of DRB1 Amino Acid Residues.
Comparing the amino acid sequences of the DRB1*08 family of alleles with those of the DRB1*11 and DRB1*13 families showed few amino acid substitutions that could be invoked to create a model of susceptibility and resistance at the molecular level. These are: glycine for serine at position 13, tyrosine for histidine at position 16, tyrosine for phenylalanine at position 47, and leucine for alanine at position 74. Considering each possibility in turn: (1) Glycine-13 is found with all members of the DRB1*08 family, whereas serine-13 is present in most, but not all DRB1*11 and DRB1*13 family members. Other DRB1 alleles encode mostly histidine, arginine, and phenylalanine at position 13, though a few of the less common alleles do encode glycine (including DRB1*1105, DRB1*1201, DRB1*1202, DRB1*1404, and DRB1*1411). Glycine and serine are both small amino acids. However, glycine is nonpolar/hydrophobic, and serine is polar/hydrophilic. (2) At position 16 DRB1*08 alleles encode tyrosine, whereas members of the DRB1*11 and DRB1*13 families and most other DRβ-polypeptides encode histidine. Tyrosine is an aromatic relatively hydrophobic amino acid, but histidine (with its imidazole group), carries a positive charge, a feature associated with the most hydrophilic amino acids. (3) In common with most DRB1 alleles, DRB1*08 alleles encode tyrosine at position 47. The exceptions to this rule are the DRB1*11 and DRB1*13 alleles, most of which encode another aromatic hydrophobic amino acid, namely, phenylalanine. Both phenylalanine and tyrosine are aromatic amino acids and relatively nonpolar, but tyrosine is significantly more polar than phenylalanine. (4) At position 74 most DRB1*08 alleles encode leucine, whereas most DRB1*11 and DRB1*13 and most other DRB1 alleles encode alanine (exceptions to this rule are the less common alleles DRB1*0805 and DRB1*0818, which both have alanine-74, and DRB1*1313, DRB1*1318, DRB1*1123 and DRB1*1125, which all have leucine-74). Leucine and alanine are both aliphatic hydrophobic amino acids; however, leucine is a significantly bigger molecule, with 6 carbon atoms compared with alanine which has only 3. Taken together, these data suggest that amino acid residues glycine-13, tyrosine-16, and leucine-74 may all contribute to DRB1*08-encoded susceptibility to PBC, whereas the amino acid residues serine-13 and phenyalanine-47 may both contribute to DRB1*11- and DRB1*13-induced resistance to PBC.
Relationship With Disease Progression.
In the UK patient series there is a modest relationship between possession of DRB1*0801 and histological stage of disease. Thus, of a total of 175 patients for whom data were available, 4 of 51 (8%) patients with early-stage disease had DRB1*0801 compared with 23 of 122 (19%) patients with late-stage disease (OR = 2.73, P = .069). However, using the same parameters to measure progression, namely, histological score, no such relationship was apparent in the Italian patients. Thus, 7 of 35 (20%) patients with early-stage disease had DRB1*0801 compared with 7 of 41 (17%) patients with late-stage disease. A comparison of the distribution of DRB1*11 and DRB1*13 alleles in the Italian series in those with early-stage disease (stages I and II) versus those with late-stage disease (stages III and IV) also showed no significant differences (DRB1*11, 23% versus 34%; DRB1*13, 14% versus 7%). Overall, this suggests that DRB1*0801, DRB1*1301, and DRB1*1101 are not predictive of disease progression in PBC, and such reported relationships may be the result of other factors such as case ascertainment bias.
We made several significant and important observations in the present study. First, there was an association with the HLA DRB1*0801-DQA1*0401-DQB1*0402 haplotype in both the UK and the Italian PBC patients. This observation contrasts with a recent report suggesting this genetic association does not exist in Italy.21 Our data showed similarity in the distribution of most but not all HLA alleles in patients and controls from the United Kingdom and Italy, contradicting the idea Italians are “genetically different” from northern Europeans. Clearly, there are some differences, but these may not be as marked as implied by the authors of the earlier study.21 The observation of a very strong genetic association with DRB1*08, albeit a different allele (DRB1*0803), in Japanese patients with PBC is also strong evidence against the former suggestion regarding PBC.16–19
Second, our data did agree with the observation suggesting DRB1*11 may protect from the development of PBC.21 We found a significant reduction in the frequency of DRB1*11 in Italian patients but this was not found in the UK series. This failure is not attributable to small sample size or inadequate statistical power. The size of our study population (412) was sufficient to detect relatively small effects on disease risk with a high level of statistical confidence. Interestingly, DRB1*11 is much more common in southern Europe (Table 4), suggesting population differences may account for some of the reported differences in the incidence and prevalence of PBC in different geographic locations.2 Historically the possibility of a DRB1*11 association in PBC was first suggested by Gores et al.,11 who found reduced frequency of the DR5 serotype (9.6% of patients versus 25.2% of controls) in their 1987 study. DRB1*11 is the major subfamily of DR5. However, until recently this observation had not been replicated. Thus, DRB1*11 was reported in 7% of 159 UK patients compared to 12% of 162 controls,13 studies in California revealed 16% of patients with DRB1*11 versus 17% of controls,15 and a recent large series from Toronto20 reported 12.3% versus 14.4% of controls. However, with three independent reports of the same protective association—Gores et al.,11 Invernizzi et al.,21 and the present study—we need to reconsider the relationship between DRB1*11 and PBC. There are two possible explanations for the DRB1*11 findings. Either DRB1*1101 is simply a marker for a true MHC-encoded protective allele in PBC, or there are genuine population-specific differences in these genetic associations with PBC. Table 4 shows the frequencies of several PBC-related HLA haplotypes and alleles in several populations where PBC occurs. Note the major differences between northern and southern Europe relate only to haplotypes 1 and 3 (DRB1*4 and DRB1*11), whereas haplotypes 2 and 4 (DRB1*08 and DRB1*13) are relatively evenly distributed between these populations.
Table 4. Distribution of PBC-Related Haplotypes in Northern Versus Southern Europeans (U.S. and Japanese Data Shown for Comparison)
Number of chromosomes
NOTE. Data for all DRB1*04 haplotypes not available. DRB1*0801 haplotypes mostly carry DQA1*0401-DQB1*0402, DRB1*11 haplotypes carry DQA1*03-DQB1*0301, DRB1*13 cosegregates with a variety of different DQA1 and DQB1 alleles, and haplotype data are not generally available. Data for Japan46 are included for reference only. However, note that at the level of resolution applied in the current study the only major difference between northern Europe and Japan is in the much higher frequency of DRB1*08 alleles in Japan; all other allele families appear to be at similar frequencies. However, this similarity may be misleading because higher-resolution genotyping reveals more significant differences between Europeans and Japanese than is apparent at low resolution. In addition, comparisons show that although DRB1-DQA1-DQB1 haplotypes are mostly conserved (and therefore predictable) within European subpopulations, they are most frequently different between Europeans and Japanese. Data for United States are based predominantly (though not exclusively) on northern Europeans.
The possibility of genuine population-specific differences in the genetic basis of PBC is not without precedent. There are reproducible population-specific differences in the HLA class II haplotypes associated with autoimmune hepatitis,9, 34, 35 primary sclerosing cholangitis,36 and responses to viral liver diseases, especially the hepatitis B and C viruses.9 It is possible these differences are related to geographic differences in exposure to potential environmental triggers, especially but not exclusively infectious agents, several of which have been implicated in the pathogenesis of PBC.37–40
Alternatively, if the DRB1 alleles discussed here are simply markers for nearby “true” susceptibility alleles in PBC, there are 2 immediate candidates to consider: the DQB1 locus and the MHC-encoded C4 loci. It is difficult to unravel the relative contribution of DQB1 to the increased risk of PBC because the DRB1*0801 haplotype almost always carries DQB1*0402. However, when considering HLA haplotypes associated with a reduced risk of PBC, the weak association with DQB1*0301 reported here is of considerable interest. This allele is almost always found in combination with the DRB1*11 (DR11) family and is occasionally found on DRB1*13 haplotypes. In addition, DQB1*0301 is very commonly found in combination with the DRB1*04 (DR4) family. In European populations specific differences in the distribution of the predominant DQB1*0301 haplotypes may lead to different genetic associations at DRB1, all of which arise as a result of linkage disequilibrium with DQB1*0301. Interestingly, in hepatitis C virus infection the dominant protective HLA haplotype appears to be DRB1*04-DQB1*0301 in some UK studies and DRB1*11-DQB1*0301 in studies from France and Italy, with DQB1*0301 the common element in the two haplotypes.
The second series of alternative candidates for MHC-encoded susceptibility and resistance to PBC that must be considered is MHC-encoded complement C4 genes. To date, very few studies have investigated C4 alleles in PBC. However, both studies that have looked at C4 reported significant associations, first with C4B*2,24 and later with both C4B*2 and C4A*Q0.12 It is possible that different complement alleles carried on the DRB1*0801, DB1*11, and DRB1*13 haplotypes determine subtle but significant differences in complement activity that are sufficient to explain the relative difference in disease risk associated with these haplotypes. If this hypothesis were true, then the population-specific differences discussed here would be coincidental.
The third observation, that of a significant association with DRB1*1301, has not been universally reported in PBC. In the present study this association, though relatively strong in Italian patients, was at best weak in the UK series, just reaching significance in the 412 patients studied. Additional, even bigger investigations are needed in both populations to determine whether this association applies to both populations or is only relevant to Italians. Interestingly, Begovitch et al.,15 in a study of 51 patients, reported a lower than expected frequency of the HLA DRB1*1302-DQA1*0102-DQB1*0604 haplotype in PBC. This association was not replicated here, nor was the hypothesis that the association may have been a result of the protective effect of the DQA1*0102 allele. This latter allele is also carried on the most common DRB1*1501 haplotype, but neither DQA1*0102 nor DRB1*1501 was found to be associated with PBC in the UK or the Italian patients (Table 2). The DRB1*1301-DQA1*0103-DQB1*0603 haplotype has been identified as a genetic determinant of susceptibility to PSC,36 and DR13 is associated with susceptibility to persistent hepatitis B virus infection,41 fibrosis due to schistosoma infection,42 type 1 AIH in children,43 and, in South America, persistent HAV infection.44
The fourth observation made in the present study relates to the commonly used explanation for genetic associations between disease and specific HLA alleles, that is, the associated alleles have specific antigen presentation profiles. Antigen presentation is the key event in initiation of the immune response, and it is a reasonable hypothesis that genetic associations with HLA occur because HLA alleles determine the relative dynamics of an immune response. Analysis of the amino acid residues encoded by the different associated alleles discussed here revealed 4 major differences, all of which have potential and are worthy of further investigation. These are glycine-13, tyrosine-16, and leucine-74 (encoded by DRB1*0801) versus serine-13 and phenylalanine-47 (encoded by most DRB1*11 and DRB1*13 alleles, respectively). Three of these 4 positions are of potential functional importance. The amino acid residue at position 13 affects the binding of antigen side chains associated with both the fourth and seventh pockets of the expressed DR molecule (Fig. 1).45 The amino acid residues at positions 47 and 74 influence the binding of antigenic side chains associated with the sixth and fourth binding pockets, respectively. Changes in hydrophobicity and hydrophilicity at positions 13, 16, and 47 and a major change in the size (molecular weight) of the residue at position 74 may influence the binding characteristics of the MHC molecule and promote or restrain the genesis of PBC (Table 5).
Table 5. Summary of Major Biochemical Differences Between HLA DRB1 Alleles Associated With an Increased Risk of PBC and Those Associated With a Reduced Risk
NOTE. Tyrosine and phenylalanine, though essentially aromatic and hydrophilic, do have regions that are both nonpolar and polar and as such may be classified as amphiphatic. However, as neither amino acid carries a charge, they are most often classified as hydrophobic, especially when compared with highly charged amino acids such as lysine, arginine, or histidine.
Histidine Alanine Arginine
4th and 6th
Size—large (Mr = 131) (Hydrophobic)
Size—small(Mr = 89) (Hydrophobic)
The last observation of the present study is that the proposed relationship between HLA class II alleles and disease progression/severity as measured by the histological index23 has not been upheld 5 years on. The current observation regarding the UK patients is in keeping with published data from Toronto20 and also our own observations of the Italian patients in this study.
The present study has confirmed the association with HLA-DRB1*0801 in PBC in patients from the United Kingdom and Italy and has stimulated new interest in the role of non-DRB1*08 haplotypes in PBC. DRB1*0801 is relatively rare in our population, and therefore identifying other “protective” haplotypes may be particularly useful in the process of mapping MHC-encoded susceptibility and resistance to PBC and thus contributing to the debate on PBC pathogenesis.