Molecular epidemiology of hepatitis B virus genotypes circulating in acute hepatitis B patients in the Campania region

Patients

One-hundred and twenty-five consecutive patients with acute hepatitis B observed from September 1999 to July 2010 at one of the two participating Units of Infectious Diseases in southern Italy, one in Naples, and the other in Caserta, had been evaluated in a previous study performed to identify the lamivudine-resistant HBV strain rtM204V/I in acute hepatitis B [Coppola and Potenza, 2013]. These two Units had been using the same clinical and laboratory approach for years and had cooperated in other clinical investigations [Coppola et al., 2008, 2013b; Sagnelli et al., 2008; 2012]. Of these 125, 53 were included in the present study based on the availability of a plasma sample stored at −40°C and never thawed until used for this investigation.

The 53 patients with acute hepatitis B showed an increase in aspartate aminotransferase (AST) and alanine aminotransferase (ALT) serum levels of at least 10 times the upper normal limit and the presence in serum of hepatitis B surface antigen (HBsAg) and IgM to hepatitis B core antigen (anti-HBc IgM); no patient showed serum antibodies to hepatitis D virus (anti-HDV) or to human immunodeficiency virus (anti-HIV). A major risk factor for the acquisition of HBV infection was identified for 30 patients, 13 patients declared no risk factor and in one case the information was not available. The patients were considered to have developed severe hepatitis if they showed at least one of the following complications: porto-systemic encephalopathy, ascites or a progressive reduction in prothrombin activity to below 35% [Sagnelli et al., 2009].

All procedures used in the study were in accordance with the current international guidelines, with the standards on human experimentation of the Ethics Committee of the Azienda Ospedaliera Universitaria of the Second University of Naples, Italy, and with the Helsinki Declaration of 1975 and revised in 1983. In addition, at the time of the first observation all patients signed their informed consent for the collection and storage of biological samples and for the anonymous use of their data in clinical research, prepared according to the rules of the Ethics Committee of the Azienda Ospedaliera Universitaria of the Second University of Naples.

Routine Methods

HAV, HBV, HCV, HDV, and HIV serum markers were sought using a commercial immunoenzymatic assay [Abbott Laboratories, North Chicago, IL, for HBsAg, anti-HBs, anti-HBc (total and IgM), anti-HAV (total and IgM) and anti-HIV 1 and 2; DiaSorin, Saluggia, VC, Italy, for anti-HDV IgG and anti-HDV IgM; Ortho Diagnostic Systems, Neckargemund, Germany, for anti-HCV]. Liver function tests were performed applying routine methods.

HBV Molecular Methods

Plasma samples of all patients were tested for HBV DNA, HBV genotype, and mutants in the polymerase region. Plasma HBV DNA was sought by real-time polymerase chain reaction (PCR), as previously described [Coppola et al., 2011]. Hepatitis B virus genotypes were determined by phylogenetic analysis of sequences of 400 nt of the pre S–S region. The pre S–S gene was amplified in a nested PCR reaction (outer primers: forward 5′-CCTCATTTTGCGGGTCACC-3′ and reverse 5′-TTTGACATACTTTCCAATCAAT-3′; inner primers: forward 5′-TCACCATATTCTTGGGAAC-3′ and reverse 5′-AGGGTTTAAATGTATACCCA-3′) [Coppola et al., 2010]. The PCR products were then purified using the MinElute PCR purification kit (Qiagen, Qiagen GmbH, Hilden, Germany) according to the manufacturer's instructions; the sequencing reaction was performed using the primer 5′-CTAGGACCCCTGCTCGTGTT-3′ with the Big Dye Terminator v 1.1 Cycle Sequencing Kit (Applied Biosystems, Carlsbad, CA), according to the manufacturer's instructions, and run in an ABI 310 Genetic Analyzer (Applied Biosystems).

The pol region mutants were assessed by direct sequencing of the region by nested PCR using the outer primers F-out-76-5′-CCTGCTGGTTCCAGTTC-3′ and R-out-1178—5′-GGT TGCGTCAGCAAACACTTG-3′ and the inner primers F-in-230—5′-CCTCACAATACCTGCAGAGTCTAGACT-3′ and R-in-997-5′-AAAGCCCAAAAGACCCACAAT-3′. The PCR products were then purified using the MinElute PCR purification kit (Qiagen GmbH, Hilden, Germany) according to the manufacturer's instructions; the sequencing reaction was performed using the primers F-in-230 and R-in-997 with the Big Dye Terminator v 1.1 Cycle Sequencing Kit (Applied Biosystems), according to the manufacturer's instructions, and run in an ABI 310 Genetic Analyzer (Applied Biosystems) [Coppola et al., 2013c].

Hepatitis B Virus Data Sets

Two different sequence data sets were analyzed in the present study. The first data set consisted of 128 HBV polymerase gene sequences, of which 53 were isolates and 75 were genotype and subgenotype specific reference sequences. All reference sequences were downloaded from the National Centre for Biotechnology Information [http://www.ncbi.nlm.nih.gov/].

The second data set included only D3 subgenotype isolates (N° = 24) and was used to estimate the evolutionary rate and time of the most recent common ancestor (tMRCA).

Phylogenetic Analysis

All data sets were aligned using CLUSTAL X software [Thompson et al., 1994], then manually edited with the Bioedit software [Hall et al., 1999]. The ModelTest program v 3.7 [Posada and Buckley, 2004] was used to select the simplest evolutionary model that adequately fitted the sequence data. A maximum likelihood (ML) phylogenetic tree was inferred with the PhyML program [http://www.atgc-montpellier.fr/phyml/] using the GTR + I + G nucleotide substitution model for the first data set. The tree was rooted using a midpoint rooting. The statistical robustness and reliability of the branching order within each phylogenetic tree were confirmed with a bootstrap analysis.

Evolutionary Rate Estimate, Time-Scaled Phylogeny, and Phylodynamics

To estimate the evolutionary rate on the second data set both a strict and relaxed clock with an uncorrelated log normal rate distribution were compared. A Bayesian Markov chain Monte Carlo (MCMC) method implemented in the BEAST software package 1.7.4 [Ciccozzi et al., 2011; Zehender et al., 2013] was used. As coalescent priors, three parametric demographic models of population growth (constant size and exponential and logistic growth) and a Bayesian skyline plot (BSP, a piecewise constant non-parametric model) were compared.

The MCMC chains were run for at least 50 million generations, and sampled every 5,000 steps.

Convergence of the MCMC was assessed by calculating the ESS (effective sample size) for each parameter. Only parameter estimates with ESS's of >200 were accepted. Uncertainty in the estimates was indicated by 95% highest posterior density (HPD) intervals, and the best fitting models were selected using a Bayes factor (BF) applying marginal likelihoods implemented in BEAST software [Ciccozzi et al., 2011].

In accordance with [Kass and Raftery, 1995], the strength of the evidence against H0 (Null Hypothesis) was evaluated as follows: 2 ln BF < 2, no evidence; 2–6, weak evidence; 6–10, strong evidence; >10, very strong evidence. A negative value indicated evidence in favor of H0. Only values of ≥6 were considered significant. To reconstruct the time-scaled phylogeny of the HBV-D3 subgenotype (the second data set), a logistic demographic model with relaxed molecular clock was used, assuming the HKY + I + G (Hasegawa, Kashino, Yano, invariable sites, gamma distribution) model of nucleotide substitution (selected by ModelTest). Statistical support for specific clades was obtained by calculating the posterior probability of each monophyletic clade.

Population dynamics was also analyzed in the HBV-D3 isolates, implementing a relaxed molecular clock under the Bayesian skyline plot.

Selection Pressure Analysis

The dN/dS rate (ω; non-synonymous to synonymous) was also estimated by the ML approach implemented in the program HyPhy [Pond and Muse, 2005] using the second data set. Site-specific positive and negative selection were estimated by two different algorithms: the fixed-effects likelihood (FEL), which fits a ω rate to every site and uses the likelihood ratio to test if dN ≠ dS, and the random-effects likelihood (REL), a variant of the Nielsen–Yang approach [Nielsen and Yang, 1998], which assumes that a discrete distribution of rates exists across sites and allows both dS and dN to vary independently from site to site. The two methods have been described in more detail elsewhere [Lo Presti et al., 2011]. In order to select sites under selective pressure and keep our test conservative, a P value of ≤0.1 or a posterior probability of ≥0.9 as relaxed critical values [Kosakovsky Pond and Frost, 2005] were assumed. Part of the analysis was conducted using the web-based interface Datamonkey [http://www.datamonkey.org/]. For the evolutionary analysis, the reference sequences of HBV-D3 complete genome (EU594382; protein polymerase id ACF95157.1) were used to trace the exact position of the amino acids found under positive selection. The HKY + I + G evolutionary model was chosen as the best-fitting model for the ML tree, as the input tree in the program HyPhy and to investigate the presence of codons under positive selection in the second data set.

Viral Gene Flow Analysis

The MacClade version 4 program (Sinauer Associates, Sunderland, MA) was used to test viral gene out/in flow among HBV-D3 subgenotype-infected subjects (second data set) with different risk factors using a modified version of the Slatkin and Maddison test, as already described [Ciccozzi et al., 2013]. A one-character data matrix was obtained from the original data set by assigning to each taxon (viral sequence) in the tree a one-letter code indicating the risk factor of the patient infected with that specific HBV-D3 strain. The putative origin of each ancestral sequence (i.e., internal node) in the tree was inferred by finding the most parsimonious reconstruction (MPR) of the ancestral character. The final tree length, that is, the number of viral gene flow events observed in the genealogy, can be computed easily and compared to the tree-length distribution of 10,000 trees obtained by random joining-splitting (null distribution). Observed genealogies significantly shorter than the random trees indicate the presence of subdivided populations with a restricted gene flow. The viral gene flow among the different risk factors (character states) was traced with the State changes and stasis tool (MacClade software), which counts the number of changes in a tree for each pair-wise character state.

When multiple MPRs were present (as in our data sets), the algorithm calculated the average migration count over all possible MPRs for each pair. The resulting pair-wise migration matrix was then normalized and a randomization test with 10,000 matrices obtained from 10,000 random trees (by random joining-splitting of the original tree) was performed to assess the statistical significance of the migration counts observed.

Phylogenetic Analysis

The ML tree of the first data set (Fig. 1) shows the presence of different statistically supported clades (bootstrap values > 70%). HBV genotype D was the most common HBV genotype in the 53 patients (34 cases, 64.1%). Of the remaining 19 patients, 14 (26.4%) had HBV genotype A (all A2 subgenotype), 2 (3.8%) HBV genotype E, and 3 (5.7%) HBV genotype F (Table I).

Of the 34 patients with genotype D sequences, 24 (70.6%) had subgenotype D3, 7 (20.6%) subgenotype D1, 2 (5.9%) subgenotype D2, and 1 (2.9%) subgenotype D4 (Table I, Fig. 1). Of these 34 patients 5 were born abroad, 3 with HBV genotype D1, one with D2, and one with D3 (Table II).

All five patients with severe acute hepatitis showed subgenotype D3 (Table II).

We also evaluated the demographics and initial clinical, biochemical and virological characteristics of the 34 patients with HBV genotype D and the 14 patients with HBV genotype A (Table II). No significant differences were found regarding age, country of origin or gender, although 13 of the 14 (92.9%) individuals infected by genotype A were males (Table II). Unsafe sexual intercourse and intravenous drug use were the most frequent (86.4%) risk factors for acquiring genotype D, while several risk factors (surgery, endoscopy, dental care, and sexual intercourse with an infected partner) were reported by patients with genotype A (Table II).

Evolutionary Rate Estimate, Time-Scaled Phylogeny, and Phylodynamics

The mean evolutionary rate of the HBV-D3 subgenotype polymerase sequences, the most frequent subgenotype in the patients enrolled, was evaluated on the second data set including 24 isolates with known sampling dates.

A comparison of the strict and relaxed clock models using the BF test showed that the relaxed clock fitted the data significantly better than the strict clock (2 ln BF = 26.45 in favor of the relaxed clock).

A comparison of the coalescent priors showed that the logistic model was better than the other models already described in the methods section (2 ln BF > 6). Under this model, the estimated mean value of the evolutionary rate of the HBV-D3 subgenotype polymerase gene fragment analyzed was 6.002 × 10⁻⁴ sub/site/year (95% HPD: 2.81 × 10⁻⁵–1.47 × 10⁻³).

Figure 2 shows the reconstruction of the time-scaled Bayesian tree of the 24 HBV-D3 isolates. The estimation of the time of the tree's root gave a mean value of 23 years ago, corresponding to 1987 (95% HPD: 1870–1997). The Bayesian tree showed two main clades (I and II). Clade I, which included a total of six isolates, had a tMRCA estimation of 16 years, corresponding to the year 1994 (95% HPD: 1941–1999). Interestingly, one of the six sequences from acutely infected subjects, labeled as 61IT@@01, was female; two were intravenous drug users and four belonged to the sexual risk group. Clade II dated back to the year 1994 (95% HPD 1941–1999) and is further divided into two subclades (IIa and IIb). Subclade IIa included two statistically supported clusters. The first, including three isolates, dated back to the year 2002 (95% HPD: 1948–2005) and the second, including three isolates (two female intravenous drug users and one male belonging to the sexual risk group), dated back to the year 1999 (95% HPD: 1939–2001).

Subclade IIb dated back to the year 1997 and included nine sequences, of which three were included in a cluster dating back to the year 2001 (95% HPD: 1947–2002) and two were included in another cluster dating back to the year 2002 (95% HPD: 1952–2006), both of them statistically supported; the remaining sequences were not statistically significant.

The BSP showed that the effective number of infections grew exponentially between 1987 and 1995 and then the curve reached a plateau approximately around the year 2000 (Fig. 3).

Viral Gene Flow Analysis

HBV-D3 subgenotype sequences were assigned to six risk groups: sexual, intravenous drug use, surgery-endoscopy-dental care, tattoo-acupuncture, unknown and household contact, and the gene flow among these groups were estimated (Fig. 4). The null hypothesis of panmixia (i.e., no population subdivision or complete intermixing of sequences from different geographical areas) was rejected by the randomization test (P < 0.0001).

The most frequent (40%) viral gene flow events observed occurred from intravenous drug use to sexual intercourse, 25% of the gene flow was from sexual to intravenous drug use and to tattoo-acupuncture, and 20% was from intravenous drug use to tattoo-acupuncture, suggesting the central role of these risk groups as a potential drive of the HBV-D3 epidemic.

Selection Pressure Analysis

The selection pressure analysis for the polymerase region of the HBV-D3 subgenotype second data set, through computation of the ratio of dN/dS (ω) with Hyphy programs, revealed limited positive selection and abundant negative selection for the polymerase proteins, the dN/dS rate (ω) being around 2.55.

FEL analysis revealed 15 codons under statistically significant negative pressure and REL analysis identified three sites under significant positive selection (Table III). Positive and negative sites were confirmed by FEL and REL analysis and also by Datamonkey (Table III).

The positively selected sites were not located in a specific domain of the polymerase protein, whereas some of the negatively selected sites were located in the reverse transcriptase RNA-dependent DNA polymerase domain (RT-like superfamily, RVT.1) and others were located in the carboxy terminal domain of the viral DNA polymerase (DNA poly-viral-C).