Host genetic factors and vaccine-induced immunity to hepatitis B virus infection

Background: Vaccination against hepatitis B virus infection (HBV) is safe and effective; however, vaccine-induced antibody level wanes over time. Peak vaccine-induced anti-HBs level is directly related to antibody decay, as well as risk of infection and persistent carriage despite vacination. We investigated the role of host genetic factors in long-term immunity against HBV infection based on peak anti-HBs level and seroconversion to anti-HBc. Methods: We analyzed 715 SNP across 133 candidate genes in 662 infant vaccinees from The Gambia, assessing peak vaccine-induced anti-HBs level and core antibody (anti-HBc) status, whilst adjusting for covariates. A replication study comprised 43 SNPs in a further 393 individuals. Results: In our initial screen we found variation in IFNG, MAPK8, and IL10RA to affect peak anti-HBs level (GMTratio of <0.6 or >1.5 and P≤0.001) and lesser associations in other genes. Odds of core-conversion was associated with variation in CD163. A coding change in ITGAL (R719V) with likely functional relevance showed evidence of association with increased peak anti-HBs level in both screens (1st screen: s545_22 GMTratio 1.71, P=0.013; 2nd screen: s595_22 GMTratio 2.15, P:= 0.011). Conclusion: This is to our knowledge the largest study to date assessing genetic determinants of HBV vaccine induced immunity. We report on associations with anti-HBs level, which is directly related to durability of antibody level and predictive of vaccine efficacy long-term. A coding change in ITGAL, which plays a central role in immune cell interaction, was shown to exert beneficial effects on induction of peak antibody level in response to HBV vaccination. Variation in this gene does not appear to have been studied in relation to immune responses to viral or vaccine challenges previously. Our findings suggest that genetic variation in loci other than the HLA region affect immunity induced by HBV vaccination. © 2008 Hennig et al. Recruitment was carried out in 2003 in the West Kiang region in The Gambia as part of the survey to determine HBV vaccine efficacy 19 years post vaccination, and in particular to evaluate of the magnitude and duration of protective antibody responses induced by infant HBV vaccination [9]. Findings from earlier studies in the same region (1985, 1989, 1993, and 1998) including information on demographics, medical background and epidemiological data have been published previously [2], [5]–[9]. Briefly, children born within the area covered by the MRC Unit in Keneba (West-Kiang, The Gambia) vaccinated within the HBV vaccination programme were considered eligible for the present study. The recruitment was restricted to non-infected individuals, only confirmed non-immune children <5 years old had been recruited prior to 1985 and from there onwards all study participants were vaccinated at birth, i.e. are deemed anti-HBc negative (perinatal transmission of HBV infection is very rare in this population[32]). This genetic study was carried out in two stages: The first screen study participants comprised all individuals identified as anti-HBc positive on at least one survey (i.e. ever positive) and, where possible, two randomly selected age-group matched consistently anti-HBc negative individuals. All other individuals, who by definition were anti-HBc negative, were included in the second stage, which was designed with the aim to replicate the ‘top’ results from the first screen. Exclusion criteria were: no information on peak anti-HBs level or date, peak anti-HBs measurement time less than one month after the last vaccination (anti-HBs is thought to peak at 1 month after the last dose), no information on number of doses of vaccine and age >5 yrs at time of last vaccination. Study participants originated from three main villages Keneba, Manduar and Kantonkunda and a small number of individuals from other villages or visitors to the region, all of whom had been included in the vaccine programme. Vaccination groups up to 2002 reflect type of vaccine, administration route and time period during which a regime was used (see Table 1); accordingly there is a direct relationship between vaccine group and date of birth (i.e. age of vaccinee). The number of individual belonging to vaccine group 6 was small (n = 2) and these were thus excluded from the analysis. The number of vaccine doses received was grouped into ≤3 (containing a small number of two and one dose recipients) versus 4 doses. An overview of selection/exclusion criteria is given in Supplementary Figure S1. We studied two outcome measures in the 1st screen (i) peak anti-HBs level as quantitative variable and (ii) anti-HBs status as binary variable, and for the 2nd screen peak anti-HBs level only. Additional screen specific exclusion of samples and covariate frequency distribution with corresponding geometric mean titer (GMT) of peak anti-HBs for each of the two sample sets are described below and are summarized in Table 1. This project was approved by the joint Gambia Government/MRC Ethics Committee, as well as LSHTM and Oxford University (OXTREC) Ethics Committees. All subjects and/or legal guardians provided written, informed consent. The total available sample for the first screen with complete data on covariates was 662, these were included in the assessment of peak anti-HBs level. The number of individuals included in the analysis of anti-HBc status was reduced to 594, because we decided to confine the analysis to a clear-cut binary outcome of consistently anti-HBc positive versus consistently negative individuals. Thus, we excluded individuals who were reconverters (n = 61; 9.3%), i.e. those shown to have lost core-antibody positivity over the course of follow-up, and those who presented with variable anti-HBc status over time (n = 7). The group of individuals with less than 3 doses of vaccine included two individuals with two doses only. The mean age of the whole 1st screen sample was 13.4 years (range 1.2 to 22.9 years) and consisted of 377 sibships with up to seven siblings, 27.1% of these sibships comprised just one individual; relatedness was accounted for in the analysis. The sample available for analysis for the 2nd screen was 393 individuals with complete data on covariates, the mean age was 7.4 years (range 1.1 to 16.2 years). These individuals tended to be younger and all were consistently anti-HBc negative, due to the initial selection procedure applied for the 1st screen (see above). The group of individuals with less than 3 doses of vaccine included six with two doses and one with one dose. The 2nd screen sample consisted of 277 sibships with up to five siblings, 42.2% of these contained only one individual. There were no individuals representative of vaccine group 1 and 2 in this sample (due to the younger mean age) and because groups 3 and 6 consisted of very few individuals, the five participants belonging to these vaccine groups were excluded. Vaccination types and regimes changed over time (as shown in Table 1). The median measurement time between last vaccination and peak antibody level assessment was 9 weeks (range 5.0 to 57.2 weeks). At each survey time point, concentrations of core antibody (anti-HBc) and, if found to be positive, hepatitis B surface antigen (HBsAg) and HBeAg were assessed, as described previously [9]. Infection was defined as the presence of anti-HBc with at least 30% inhibition, as stipulated by the manufacturer (Sorin Biochemica or DiaSorin). If infection was found only during a single follow-up time point and was not confirmed subsequently, the person was considered to have had a transient infection. Carriage was defined as the detection of HBsAg on two separate occasions at least six months apart. Primary vaccine failure was defined as a peak anti-HBs response of <10 mIU/mL; a responder was defined as a vaccinee with a peak anti-HBs response of ≥10 mIU/mL. The proportion of HBsAg positive (n = 21), anti-HBeAg positive (n = 7) and persistent carriers (n = 7) in the total sample was too small for a subgroup analysis of these parameters. DNA extraction from whole blood or PBMCs was carried out using a standard salting-out method [33]. Genotyping assays for the initial screen were run on the Illumina BeadArray platform (www.illumina.com). Our candidate gene selection criteria were based on literature searches, previous reported genetic associations with vaccine-induced antibody level (in particular with anti-HBs), possible biological implication in immune-regulatory processes, inclusion of gene families and coverage of regulatory pathways. The selection of 768 single nucleotide polymorphism (SNPs) for genotyping was based on HapMap frequency data, validation, anticipated success rate on the Illumina platform, gene size, distribution across loci, enrichment in coding changes, inclusion of sequence 500 bp up- and down-stream of each gene. Just over 95% of our candidate polymorphisms were SNPs genotyped in both Yoruba (YRI) and Caucasians (CEU) as part of the HapMap project (http://www.hapmap.org/), many of these variants are tag SNPs in either or both of these populations. Linkage disequilibrium (LD) and haplotype blocks were assessed using the Haploview programme (http://www.broad.mit.edu/mpg/haploview/) applying default parameters according to Gabriel et al [34]. To account for markers with low minor allele frequency (MAF) we employed r2 as measure of LD throughout and considered an r2 value of 0 to 0.5 as low, 0.5 to 0.75 as intermediate and >0.75 as strong LD. The FASTSNP programme (http://fastsnp.ibms.sinica.edu.tw/pages/input_CandidateGeneSearch.jsp) was employed to establish whether SNPs were of functional relevance [35]. The 2nd screen genotyping was carried out using the Sequenom platform (iPlex and hME; http://www.sequenom.de/). The aim was to replicate the results from the first screen by concentrating on the ‘top’ markers in the remainder of the sample set. Validated demographic and serological data was available as an Access 2000 database from the MRC’s long-term hepatitis B vaccination programme [9]. All statistical analysis was conducted using Stata software (version 9; StataCorp) and employed forward regression models to determine the role of genetic markers on outcomes, with inclusion of individual SNPs into the (multi-marker) gene/locus models according to LD structure. No adjustment for multiple comparisons was made. Almost 95% of individuals were Mandinka, ethnicity was thus not considered informative enough for stratification of the study cohort by ethnic groups. This society is polygamous, thus resulting in a complex pedigree structure, e.g. presence of a large proportion of half-sibs. Consequently, clustering by sibship was determined by maternal ID in order to account for relatedness. For polymorphisms with a low MAF heterozygotes and homozygotes were grouped for the analysis if there were less than 10 individuals in the latter genotype category by outcome measure. Genetic variation was entered in the analysis as categorical genotype data with 11 representing ancestral homozygotes, 12 representing heterozygotes and 22 representing variant homozygotes. Two outcome measures were assessed: (i) peak anti-HBs level and (ii) anti-HBc status. From the peak anti-HBs level we can predict the persistence of antibody level at any time point afterwards, peak anti-HBs is also a predictor of risk of infection and persistent carriage. Peak anti-HBs was not normally distributed and thus log-transformed for the analysis. Multiple linear regression, including robust standard errors to account for relatedness by sibship, was used to identify covariates associated with log peak anti-HBs level. The model included as covariates the logarithm of measurement time (i.e. duration of time between last vaccination and peak anti-HBs measurement to account for the exponential decay of anti-HBs over time), vaccine group (i.e. type, regime, administration route, year), number of doses (≤3 versus 4), age group at recruitment (5 year intervals), village (Keneba, Manduar, Kantonkunda+others), gender and relatedness. Results are summarized as the ratio of the geometric mean titer (GMT), associated 95% confidence intervals (CI) and P-values. Consistent core antibody positive individuals were defined as those who had been identified as anti-HBc positive at any time point and remained positive at subsequent surveys and compared to subjects who were consistently negative for anti-HBc over the course of time; reconverters and variables were excluded (see above). Multiple conditional logistic regression, including robust standard errors to account for relatedness by sibship, was used to identify covariates associated with being core antibody positive. The model for anti-HBc status accounted for peak anti-HBs level (as surrogate measure for vaccine group, anti-HBs measurement time, number of vaccine doses), age group at recruitment (5 year intervals), village (Keneba, Manduar, Kantonkunda+others), and gender as covariates. Results are reported as odds ratios (OR), with associated 95% CI and P-values. The statistical analysis for the 2nd screen was carried out in an identical manner to that of the 1st screen for anti-HBs level as the outcome. The distribution of covariates for the 2nd screen was slightly different in terms of age and vaccine group distribution (see below).