Distinct MHC class I and II alleles are associated with hepatitis C viral clearance, originating from a single source
The role of cytotoxic T lymphocyte responses, restricted by human leukocyte antigen (HLA) class I alleles, is recognized as highly significant in the successful clearance of hepatitis C virus (HCV). The frequency of class I alleles in females inoculated with HCV genotype 1b from a single source was examined for an association with outcome. Class I typing was performed using polymerase chain reaction sequence-specific primers in 227 female subjects: 141 had chronic infection and 86 had viral clearance. Statistical analysis included χ2 testing and multiple logistic regression analysis. A*03, B*27, and Cw*01 occurred more frequently in those with viral clearance (39.5%, 14%, and 9.3%, respectively) compared with those with chronic infection (19.1%, 2.1%, and 1.4%, respectively; P ≤ .005). B*08 occurred more often in those with chronic infection compared with viral clearance (39.7% vs. 19.8%; P = .002). In combination with previously reported class II allele associations, over 75% that successfully eliminate HCV carry either A*03, DRB1*0101, or *0401, compared with only 37% of those with chronic infection (P < .0001). The haplotypes A*03-B*07-DRB1*15-DQB1*0602 and A*02-B*27-Cw*01-DRB1*0101-DQB1*0501 are associated with viral clearance (P = .004 and .01, respectively). By multiple logistic regression analysis, the alleles A*03, B*27, DRB1*0101, *0401, and *15 are associated with viral clearance, and B*27 has the strongest association (odds ratio [OR] 7.99). The haplotype A*01-B*08-Cw*07-DRB1*03011-DQB1*0201 is associated with chronic infection (P = .002), being independent for DQB1*0201 (OR 0.27). In conclusion, certain class I alleles are associated with outcome in this homogenous cohort. More significantly, either HLA-A*03, -DRB1*0101, or -*0401 are carried by an overwhelming majority of those subjects who successfully clear HCV. (HEPATOLOGY 2004;40:108–114.)
Hepatitis C virus (HCV) is a hepatotropic virus with a high rate of chronic infection. It is known that progression to cirrhosis and hepatocellular carcinoma may take up to 20 years in individuals who are chronically infected and thus represents a leading cause of hepatocellular morbidity and mortality.1–4 To date, viral factors (e.g., genotype) and host factors (e.g., age of acquisition, male sex, and alcohol consumption) are known to alter both the natural history of the disease and treatment outcomes.2, 4–6 As yet, no in vivo models of infection are available in HCV infection, therefore the pathogenic mechanism of disease remains unclear.
In HCV infection, recognition and elimination of infected cells by cytotoxic T lymphocytes (CTLs) require the presentation of specific HCV antigens on the membrane of hepatocytes in the context of HLA-A, -B and -C antigens. In acute HCV infection, animal models suggest that viral clearance appears to be dependent on an intrahepatic CTL response directed against multiple HCV antigens restricted by several class I molecules.7 Human studies on peripheral blood similarly report that viral clearance is associated with a strong initial HCV-specific CD8+ T cell response that may be followed by a strong CD4+ T cell response.8–10
In chronic HCV infection, it is proposed that the HLA class I–restricted CTL responses may be responsible for controlling viral replication and may be directly responsible for the histological injury sustained in chronic infection, rather than a direct hepatotoxic effect of the virus. Intrahepatic and peripheral blood CTL responsiveness to viral epitopes is inversely related to viral titre.11–14 In chronic infection, HCV-specific CTLs localize within the liver to areas of piecemeal and lobular necrosis in association with increased expression of HLA-A, -B, and -C antigens15, 16 and at titres up to 30-fold higher than in the peripheral blood.17
The major histocompatibility complex (MHC) plays a critical role in the immune response, because these gene products present viral antigen to cytotoxic and helper T cells. We have previously reported on the association of class II alleles with viral clearance (DRB1*0101-DQB1*0501) and chronic infection (DRB1*03011-DQB1*0201).18 However, understanding the role of the class I alleles remains crucial to elucidating the host response to HCV. Presently, the relationship between class I alleles and outcome in HCV infection is poorly described.19–23 Compared with other studies, our subject population is a single-sex, single-genotype cohort with the same ethnic background and mode of infection inoculated at a reasonably young age and within a documented time frame. This cohort lacks any confounding variable previously identified to alter outcome in HCV infection and thus represents an ideal population in which to test the association of class I alleles with outcome.
Patients and Methods
The subject population is derived from a cohort of females inoculated with anti-D immunoglobulin in 1977 that had been contaminated with HCV from a single source. The genotype was identified as 1b. These subjects were identified in 1994 during a “look-back” program and offered referral to any specialist center. This cohort is described in detail in a previous publication.24 Notably, only 55% of those subjects who were antibody-positive 17 years post inoculation had evidence of chronic infection (HCV RNA reverse-transcriptase polymerase chain reaction [RT-PCR]–positive and biopsy proven chronic active hepatitis), and despite being a genotype 1b infection there is a uniquely slow rate of progression of disease (2% had probable or definite cirrhosis 17 years postinoculation). The subjects in this study—who were derived but not selected from this cohort—are those females who are attending two centers in Dublin by their own choice: St. James's and St. Vincent's hospitals. Patients with no other risk factors for the acquisition of viral hepatitis, no history of alcohol excess, and no concomitant liver disease were asked to participate over a 6-month period. Two hundred forty-three females gave written consent to participate in this study, which has received ethical approval from the Research and Ethics Committees at both institutions and conforms to the guidelines of the 1975 Declaration of Helsinki.
The age range of this population at the time of infection was 27.35 ± 5.52 years. There was no difference in age between those subjects who had persistent infection compared with those who had viral clearance (48.57 ± 5.66 years vs. 48.01 ± 5.32 years; P = .4).
Spontaneous viral clearance had occurred in 95 subjects. These subjects were documented recipients of contaminated anti-D immunoglobulin in 1977.25 All had evidence of prior HCV infection, and all were anti-HCV–positive as determined by both third-generation enzyme-linked immunosorbent assay (EIA, Abbott, Abbott Diagnostics, Germany) and recombinant immunoblot assays (RIBA-3, Chiron Corporation, Emeryville, CA). All 95 subjects had constantly normal serum alanine aminotransferase values and negative HCV RNA RT-PCR (Amplicor, Roche Diagnostic Systems Inc., NJ) tests since 1994 on repeated testing. Chronic HCV infection was present in 148 subjects. All were persistently HCV RNA RT-PCR–positive and had biopsy-proven chronic active hepatitis.
Class I Typing.
DNA was extracted from whole blood using a salting out technique, as described previously.26 During this process, HCV RNA was removed by incubating the digested preparation with 1.5 μL RNase (Boehringer Mannheim UK Ltd., East Sussex, UK) per 400 μL of nuclear lysate according to the manufacturer's instructions.
HLA-A, -B, and -C loci were typed according to the DNA typing method described by Bunce et al.27 This polymerase chain reaction (PCR) technique uses sequence-specific primer (SSP) reactions to simultaneously identify alleles of the class I locus in an allele-specific or group-specific manner. Briefly, to identify class I alleles each PCR reaction consisted of, in final concentration, 0.02 μg DNA, 0.42 units Taq DNA polymerase (Advanced Biotechnologies, Surrey, UK), 1 μL of 10X Buffer IV (Advanced Biotechnologies), 3.4 mM MgCl2, 250 μM of each dNTP, 0.54 μL glycerol, 0.01 μL 100 mg/mL Cresol Red, and 5 μL of the allele-specific and control primer mixes at a concentration of 1–4 μM each (United Kingdom Transplant Support Service Authority, Bristol, UK). The PCR reaction mix was plated out into 96-well plates, and amplifications were carried out in an MJ Research PTC-225 DNA engine tetrad. The cycling parameters used were as follows: 96°C for 1 minute, followed by 14 cycles of 96°C for 30 seconds, 65°C for 50 seconds, and 72°C for 20 seconds, followed by 25 cycles of 96°C for 30 seconds, 62°C for 50 seconds, 72°C for 20 seconds, followed by 72°C for 4 minutes. The PCR products were electrophoresed in a 1.5% agarose gel for approximately 35 minutes at 150 V in 0.5 Tris-borate-EDTA (TBE) and visualized using UV illumination.
Despite repeated attempts, it was not possible to determine class I alleles on 16 subjects; this was due either to poor-quality DNA or a PCR reaction failure. These subjects were therefore not included in the analysis. The subject population for analysis included 141 subjects with chronic infection and 86 subjects with spontaneous viral clearance.
HLA class II typing has been previously performed on this cohort.18 Class I and II haplotypes were generated based on the allele frequencies in this cohort.
Comparisons of the MHC class I phenotypic frequencies were made between those subjects who were chronically infected and those who had sustained viral clearance. Differences in frequencies between the groups were analyzed by Pearson's χ2 and Fisher's exact tests where appropriate. A t test was used to identify differences in ages between the two groups. Multiple logistic regression analysis was used to generate a model of the data that would predict an outcome based on the frequencies of HLA alleles. A forward conditional stepwise method was used. In this model, the probability for entry was .05 and removal was .10. The statistical package used was SPSS (version 10.1 for Windows, SPSS Inc, Chicago, IL). In all tests, a P value of ≤ .05 was considered to be statistically significant.
Class I Alleles and Viral Outcome.
The phenotypic frequency of each allele was compared in subjects who had spontaneous clearance with those who had chronic infection (Table 1). A*03 was present in 39.5% of subjects who had viral clearance compared with 19.1% of those who had chronic infection (P = .001; odds ratio [OR] 2.8). Similarly, B*27 and B*07 were also more frequent in those who had viral clearance (14% and 40.7%, respectively) compared with those who had chronic infection (2.1%, P = .001, OR 7.5 and 25.5%, P = .017, OR 2.0, respectively). Cw*01 was found more frequently in subjects who had viral clearance (9.3%) compared with those who had chronic infection (1.4%, P = .005, OR 7.1). B*08 occurred more frequently in subjects who had chronic infection compared with those who had viral clearance (39.7% vs. 19.8%, P = .002, OR 0.4). B*18 was also more frequently observed in subjects who had chronic infection (11.3%) compared with those who had viral clearance (2.3%, P = .015, OR 0.2).
Table 1. MHC Class I Allele Phenotype Frequencies in Descending Order
|A*01||0.42||63 (44.7)||32 (37.2)||.3||0.7||0.4–1.3|
|A*02||0.49||70 (49.6)||42 (48.8)||.9||0.9||0.6–1.7|
|A*03||0.27||27 (19.1)||34 (39.5)||.001||2.8||1.5–5.0|
|A*11||0.17||23 (16.3)||15 (17.4)||.8||1.1||0.5–2.2|
|A*23||0.06||9 (6.4)||4 (4.7)||.6||0.7||0.2–2.4|
|A*24||0.07||11 (7.8)||5 (5.8)||.6||0.7||0.2–2.2|
|A*25||0.04||8 (5.7)||1 (1.2)||.09||0.2||0.02–1.6|
|A*26||0.02||3 (2.1)||1 (1.2)||.6||0.5||0.05–5.3|
|A*29||0.1||14 (9.9)||9 (10.5)||.9||1.1||0.4–2.6|
|A*30||0.02||5 (3.5)||0 (0)||.08||NA|| |
|A*31||0.04||6 (4.3)||2 (2.3)||.4||0.5||0.1–2.7|
|A*32||0.07||10 (7.1)||6 (7.0)||.9||0.9||0.3–2.8|
|A*33||0.004||1 (0.7)||0 (0)||.4||NA|| |
|A*68||0.06||8 (5.7)||6 (7.0)||.7||1.2||0.4–3.7|
|B*07||0.31||36 (25.5)||35 (40.7)||.017||2.0||1.1–3.6|
|B*08||0.32||56 (39.7)||17 (19.8)||.002||0.4||0.2–0.7|
|B*13||0.01||3 (2.1)||0 (0)||.2||NA|| |
|B*14||0.13||18 (12.8)||11 (12.8)||.9||1.0||0.4–2.2|
|B*15||0.05||6 (4.3)||5 (5.8)||.6||1.4||0.4–4.7|
|B*18||0.08||16 (11.3)||2 (2.3)||.015||0.2||0.04–0.8|
|B*27||0.07||3 (2.1)||12 (14.0)||.001||7.5||2.0–27.3|
|B*35||0.15||22 (15.6)||11 (12.8)||.6||0.8||0.4–1.7|
|B*37||0.04||6 (4.3)||4 (4.7)||.9||1.1||0.3–4.0|
|B*38||0.01||2 (1.4)||1 (1.2)||.9||0.8||0.07–9.2|
|B*39||0.02||3 (2.1)||2 (2.3)||.9||1.1||0.2–6.7|
|B*40||0.01||2 (1.4)||1 (1.2)||.9||0.8||0.07–9.2|
|B*41||0.004||0 (0)||1 (1.2)||.2||NA|| |
|B*44||0.39||50 (35.5)||40 (46.5)||.09||1.6||0.9–2.7|
|B*45||0.004||1 (0.7)||0 (0)||.4||NA|| |
|B*49||0.02||2 (1.4)||3 (3.5)||.3||2.5||0.4–15.3|
|B*50||0.009||1 (0.7)||1 (1.2)||.7||1.6||0.1–26.7|
|B*51||0.04||7 (5.0)||2 (2.3)||.3||0.5||0.09–2.2|
|B*53||0.004||1 (0.7)||0 (0)||.4||NA|| |
|B*55||0.03||6 (4.3)||0 (0)||.05||NA|| |
|B*57||0.09||10 (7.1)||10 (11.6)||.2||1.7||0.7–4.3|
|B*60||0.08||11 (7.8)||6 (7.0)||.8||0.9||0.3–2.5|
|B*63||0.004||1 (0.7)||0 (0)||.4||NA|| |
|B*70||0.009||1 (0.7)||1 (1.2)||.7||1.6||0.1–26.7|
|Cw*01||0.04||2 (1.4)||8 (9.3)||.005||7.1||1.5–34.4|
|Cw*02||0.03||3 (2.1)||3 (3.5)||.5||1.7||0.3–8.4|
|Cw*04||0.18||28 (19.9)||12 (14.0)||.3||0.7||0.3–1.4|
|Cw*05||0.27||33 (23.4)||29 (33.7)||.09||1.7||0.9–3.0|
|Cw*06||0.15||21 (14.9)||14 (16.3)||.8||1.1||0.5–2.2|
|Cw*07||0.65||93 (66.0)||54 (62.8)||.6||0.9||0.5–1.5|
|Cw*08||0.13||19 (13.5)||11 (12.8)||.9||0.9||0.4–2.09|
|Cw*09||0.05||10 (7.1)||2 (2.3)||.1||0.3||0.07–1.5|
|Cw*10||0.09||12 (8.5)||9 (10.5)||.6||1.3||0.5–3.1|
|Cw*12||0.05||7 (5.0)||4 (4.7)||.9||0.9||0.3–3.3|
|Cw*14||0.02||4 (2.8)||1 (1.2)||.4||0.4||0.04–3.7|
|Cw*15||0.02||2 (1.4)||2 (2.3)||.6||1.7||0.2–11.9|
|Cw*16||0.08||12 (8.5)||6 (7.0)||.7||0.8||0.3–2.2|
|Cw*17||0.01||1 (0.7)||1 (1.2)||.7||1.6||0.1–26.7|
MHC Class I and II Alleles and Viral Outcome.
We have described previously an association between class II alleles and viral outcome in this cohort. Of the 227 females who had successful typing of the class I alleles, DRB1*0101 was present in 32.6% of those who had viral clearance compared with only 8.5% of those who were chronically infected (P < .001, OR 5.2), and DRB1*0401 is present in 29.1% of those who had viral clearance compared with 15.6% of those who had chronic infection (P = .015, OR 2.3). Overall, 75.6% of those subjects who successfully cleared the HCV carry either an A*03, DRB1*0101, or DRB1*0401 allele compared with 37.6% of those who failed to clear HCV and developed chronic infection (P < .001, OR 5.1).
In this cohort, we found that certain haplotypes were inherited in linkage disequilibrium. HLA A*01-B*08-Cw*07 was found more frequently in those subjects who had chronic infection (P = .009, OR 0.4) and was linked with the class II haplotype DRB1*03011-DQB1*0201. In its entirety, the haplotype was associated with chronic infection (P = .002, OR 0.3). Multiple logistic regression analysis was performed, including in the analysis all alleles noted to be associated with chronic infection either independently or through linkage disequilibrium, regardless of their statistical significance, to determine which allele was associated more closely with chronic infection. We found that HLA DQB1*0201 was most significantly associated with chronic infection (P < .001, OR 0.27, 95% CI 0.142–0.516, Cox & Snell R square = 0.108).
HLA A*03 and B*07, both associated with viral clearance, were inherited in linkage disequilibrium forming the haplotype A*03-B*07-DRB1*15-DQB1*0602. Although the haplotype remained associated with viral clearance (P = .004, OR 3.5), the individual class II alleles were not associated with viral outcome by χ2 analysis (P = .10) (Table 2).
Table 2. Comparison of Allele and Haplotype Frequencies According to Viral Status
HLA A*02-B*27-Cw*01 was inherited in linkage disequilibrium with DRB1*0101-DQB1*0501, and this haplotype was associated with viral clearance (P = .01). Both class II alleles were associated with viral clearance as well as B*27 and Cw*01 (see Table 2). However, only eight subjects carried HLA A*02-B*27-Cw*01 compared with 38 who carried DRB1*0101-DQB1*0501. The OR for viral clearance was strongest for B*27 (7.46) individually and for DRB1*0101-DQB1*0501 (5.41) as a haplotype.
Multiple logistic regression analysis of all alleles associated either directly or indirectly with viral clearance demonstrated that A*03, B*27, DRB1*0101, DRB1*0401, and DRB1*15 were all associated with viral clearance, and the strongest OR was for B*27 at 7.99 (Table 3).
Table 3. Multiple Logistic Regression Analysis of Alleles Associated with HCV Viral Clearance
In this unique cohort of females inoculated with HCV genotype 1b from a single source, we have clearly identified class I alleles that are associated with outcome: A*03, B*07, B*27, and Cw*01 occur more frequently in individuals who have viral clearance, and B*08 and B*18 occur more frequently in those who have chronic infection. This study population is a single genotype cohort that is relatively homogenous for factors that contribute to the natural history of HCV infection and hence is ideal for examining purely the genetic contribution to outcome.
This study was undertaken 20 years following inoculation with HCV. Only females with positive anti-HCV antibodies were included. Females who were exposed but were anti-HCV–negative were omitted, because confirmation of acute infection could not be obtained in those who had possibly lost anti-HCV antibodies with time. Therefore, this study does not address potential differences in HLA alleles between those who clear HCV at differing rates.
In many laboratories PCR-SSP is now the method of choice for medium to high resolution HLA typing,28 but it does not identify the greater than 700 class I (A, B, and C) alleles now described.29 However, the patient cohort in this study is of Irish descent, with historically limited ethnic diversity, and the antigen frequencies identified in this study by PCR-SSP are comparable to those published previously.30–32
The A*03 allele was associated with viral clearance by direct analysis and confirmed with multiple logistic regression analysis (OR 2.4). A*03 is carried by 27% of this population overall and is the third most common HLA-A allele in this study population. More than 75% of subjects with successful viral clearance carried either HLA-A*3, -DRB1*0101, or -0401. In this study, HLA-B*27 was not a particularly common allele, but by statistical analysis it holds the strongest association with viral clearance. Further studies are required in this cohort to identify the HCV peptides presented by the widely prevalent A*03 allele or the highly associated B*27 allele that are capable of generating an adequate CTL response required for HCV clearance. In view of the extensive sequence diversity within genotype 1b and between genotypes, detection of similar class I associations in more heterogeneous populations may be extremely difficult. Nonetheless, we believe the strong association of HCV clearance with HLA alleles in this highly defined population suggests a major role for the MHC in HCV clearance.
The A*03 allele occurred in linkage with A*03-B*07- DRB1*15-DQB1*0602. Interestingly, from this haplotype, multiple logistic regression analysis identified both A*03 (OR 2.4) and DRB1*15 (OR 2.2) to be associated with HCV clearance, but not B*07. Similarly, in the haplotype A*02-B*27-Cw*01-DRB1*0101-DQB1*0501, both B*27 and DRB1*0101 were identified by multiple logistic regression analysis to be strongly associated with HCV clearance, but not DQB1*0501. This suggests that further mapping of the MHC is required to clarify the relative contributions of each allele to HCV clearance and to identify whether or not other immune modulating genes inherited in linkage are of greater significance.
Multiple logistic regression analysis estimates the measure of association (rather than effect) of one variable across a population defined simultaneously by levels of several other variables. In this study, this analysis was used to generate a model that would predict those alleles most significantly associated with viral clearance or persistence. Included in each analysis were all alleles associated with viral outcome and those not significantly associated on single hypothesis testing but occurring in linkage. The results generated by the forward stepwise conditional method were identical to those when the analysis was performed using a backward technique. There was no significant correlation between those alleles included in this analysis; therefore, collinearity was not determined to be an issue. Although the model generated is a product of the data input and hence is potentially limited by that selection, this model was chosen over pairwise association testing with correction for multiple comparisons (Bonferroni) because of the large numbers of P values computed (n = 52); furthermore, the latter method was believed to be overly conservative and would require imprecise confidence intervals.33
Thio et al. have documented an association between A*1101, B*57, and Cw*0102 and viral clearance, and A*2301 and Cw*04 and chronic infection.23 Although the study population is larger—including 231 individuals with viral clearance and 444 with chronic infection—the baseline characteristics are heterogeneous in terms of race, sex, and viral genotype. The resulting variance in antigenic specificities may account for some differences between the two studies. Significantly, more subjects with viral clearance had chronic hepatitis B virus infection, and although inclusion of hepatitis B surface antigen status did not alter allelic associations with outcome by multivariate analysis, this does not rule out a modulating effect of coinfection on the immune response resulting in viral clearance or persistence. In this same study population, similar class II associations with outcome were reported in the Caucasian subgroup (DRB1*0101-DQB1*0501 and viral clearance and DRB1*0301-DQB1*0201 and chronic infection),34 as were observed in our cohort. In the current study, the authors do not report on the class I alleles that occur in linkage disequilibrium with these class II alleles. Certainly multiple logistic regression analysis suggests that class I alleles may be more influential in viral clearance than the class II alleles, but Thio et al. are unable to confirm our findings. These differences also illustrate the potential limitations in performing association studies in a homogenous cohort and the need for testing these associations across viral genotypes and HLA phenotypes to identify those epitopes most relevant with regard to understanding the mechanisms of HCV clearance and vaccine development.
HLA-B*08 occurs at a frequency of 32%, as would be expected in a Northern European Caucasoids (NEC) population in which one allele *0801 has been described. It is inherited in tight linkage disequilibrium with HLA-A*01-B*08-Cw*07-DRB1*03011-DQB1*0201. Statistically, B*08 is associated with chronic infection in this cohort when tested independently, but multiple logistic regression analysis demonstrates that it is only DQB1*0201 that is associated with chronic infection when the codependencies of other alleles are taken into account. Although the mechanism for this is unclear, it is possible that HCV persistence is mediated by the emergence of viral escape mutants specific for DQB1*0201. Some diversity has been observed in viral sequences from this population.35 However, large-scale sequencing studies on stored sera would be required to test this hypothesis.
In conclusion, we have confirmed that the HLA associations with the natural outcome of HCV infection involves both class I and II antigens. Our data support the concept of the involvement of both CD4+ and CD8+ T cells in the clearance of acute HCV infection. Whether outcome is dependent on these loci alone or is in synergy with other immunomodulatory genes occurring in linkage disequilibrium requires further investigation.