Both hepatitis B virus (HBV) and hepatitis C virus (HCV) cause persistent liver infection, but the two viruses are notably different in terms of replication strategy and host interaction, as well as their global impact on public health [1], [2]. 170 million people are estimated to be infected worldwide with HCV, with 70–90% of infected individuals becoming chronically infected [3]. In contrast, more than 350 million people are estimated to be globally infected with HBV but more than 95% of cases will result in viral clearance [4]. The viruses also possess strikingly different genome arrangements. HCV, a member of the flaviviridae family, possesses a positive strand RNA genome of ∼9.6 Kb encoding a polyprotein that is co- and post-translationally processed to form three structural (core, envelope 1 (E1) and envelope 2 (E2/p7)) and six non-structural proteins (NS2, NS3, NS4A, NS4B, NS5A & NS5B) [5]. HBV, on the other hand, is the smallest known DNA virus with a genome only 3.2 kb in length. The genome comprises four open reading frames (ORF): core (C), polymerase (P), surface (S) and X. All four ORFs are overlapped completely or partially. Specifically, S is encompassed entirely by P. Gene overlapping is a common strategy adopted by many viruses to reduce their genome size and maximize their encoding capacity. However, this inevitably constrains the independent evolution of the individual reading frames as a mutation with little effect on one gene may cause severe or even fatal changes on the cognate overlapping gene. Thus, in an overlapping region, if one gene undergoes adaptive evolution (positive selection) with a high ratio of non-synonymous nucleotide mutations (d_n/d_s >1), the cognate gene often undergoes purifying selection (negative selection; d_n/d_s <1). This has been observed in many different viruses such as simian immunodeficiency virus [6], potato leaf roll virus [7] and human papilloma virus [8].

Studies on the variation in HCV sequences indicate that E1 & E2 (which interact with the host immune system) possess greater variation compared to the NS5B protein, which encodes the RNA-dependent RNA polymerase, and this disparity may reflect the differing functional roles of these two proteins [9]–[11]. In HBV, however, S and P shared an overlapping segment of DNA sequence, raising the question of whether these proteins face similar competing requirements and, if so, how they resolve this apparent conflict. From a structural perspective, the HBV P protein comprises a terminal protein (TP) domain, a reverse transcriptase (RT) domain, an RNase H (RH) domain and a spacer domain [12]. Of these structures, the TP, RT and RH domains are conserved, while the spacer domain is highly variable (for review see [13], [14]). Previous genetic studies have suggested that the spacer, which acts a tether between TP and RT, is dispensable as it has little effect on replication competence [12], [13]. The S ORF encodes three surface proteins termed large (L), middle (M), and small (S) protein. The M protein is comprised of the S and PreS2 domains. The L protein includes another N-terminus genotype-dependent domain termed PreS1. PreS1 is essential for viral entry and infection. In particular, amino acids 2–48 act as the recognition site for a hepatocyte-specific receptor [15]. The S protein is predicted to function as a membrane spanning protein and contains four trans-membrane (TM) regions, each consisting of an α-helix structure [16]. Compared with the variable PreS domain (including PreS1 and PreS2), the TM regions, which maintains the stability of protein structure, are conserved [16]. However, the hydrophilic loops between the α-helices harbor more variable amino acid residues. Moreover, variation in epitopes may aid virus escape from the host immune system. The T cell and B cell epitopes in the S protein (HBsAg) targeted by the host immune system are mainly concentrated in these loop regions, including the major hydrophilic region (MHR, amino acids 99–160) which contains a conformational B cell epitope cluster. In addition, the core of MHR contains the “a” determinant (residues 121–147) which is the region primarily associated with induction of a protective humoral immune response [17], [18].

The selection pressure exerted by the host immune system can be focused on the epitope regions, although this pressure varies over the course of an HBV infection [19], [20]. Specifically, the humoral immune (B cell mediated) response to the S protein plays a relevant role in the clearance of infectious HBV particles, whereas cellular immune (T cell mediated) responses contribute to the elimination of infected hepatocytes [19], [21]. The studies on immunopathogenesis of S are extensive, and hence there are large quantities of associated epitope data. In contrast, as a consequence of the fewer reports on both the humoral immune response and the cytotoxic T cell (CTL) response, there are relatively less data available on epitopes data in P [22]–[27]. Moreover, the higher levels of conservation in P may restrict viral escape via mutations in epitopes [24]. Thus, the correlation between epitope of P protein and selection pressure is not considered in this report.

Several studies have investigated the effects of these functional constraints in HBV [28]–[30]. Although sequence evolution in the overlapped P and S regions is constrained [28], Zaaijer et al. (2007) [29] proposed that HBV is able to use the degeneracy in the genetic code to overcome these restraints. Due to the frame shift between the coding regions for P and S, the first position in the P codon corresponds to the third position in the S codon (P₁S₃), the second position in the P codon corresponds to the first position in the S codon (P₂S₁), and the third position in the P codon correspond to the second position in the S codon (P₃S₂). Thus, a synonymous mutation in P is able to produce a corresponding non-synonymous mutation in S (P₃S₂), which can produce an amino acid change in S but conserve the corresponding site in P, satisfying the constraints on both genes. Furthermore, the study found that the most of changes take place in P₁S₃ (P) or P₃S₂ (S) and the nucleotide mutations in P₂S₁ are rare. In a more recent study, negative selection was simultaneously detected in both the overlapped P and S genes [30]. However, the dataset was composed of a single study set of 33 patients and hence involved analysis of a relatively small number of sequences.

In this study, we reinvestigated the overlapping P and S regions using a dataset, significantly larger compared to previous studies and considered how the observed variation in this region was related to what is known about the respective functions of the two proteins[28]–[30]. We examined the correlation between conservation and structured protein domains, as well as the relationship between positive selection sites and epitope regions and consider how these competing requirements help to shape the HBV genome and impact the life cycle of the virus.

Gene overlapping is a common occurrence in viruses. In this way, a virus can minimize its genomic size, effecting a more economical replication cycle. The trade-off is that within an overlapping region, nucleotide substitutions may result in simultaneous amino acid mutations in the two distinct proteins encoded by the same nucleotide sequence. Consequently, this restricts the independent evolution of overlapped genes. How natural selection acts on the different viral proteins within an overlapping region continues to interest evolutionary biologists and virologists. Adaptive (positive) selection in one protein contrasted by purifying (negative) selection in the other overlapping protein has been observed in several viruses including simian immunodeficiency virus, potato leaf roll virus and human papilloma virus [6]–[8].

HBV is the infectious pathogen responsible for hepatitis B and possesses a highly compact DNA virus with half of its genome overlapped. The first identification of distinctive evolutionary constraints acting on the HBV genome was reported in 1997 by Mizokami et al. who analyzed 27 HBV strains [28]. They found a lower rate of nonsynonymous substitutions to synonymous substitutions (d_n/d_s) in the non-overlapped region compared to the overlapped region. Subsequently, Zaaijer et al. showed the overlapping polymerase and surface protein were undergoing adaptive selection but proposed they were nevertheless, to a certain degree, evolving independently [29]. Most recently, based on a different dataset, van der Klundert et al. detected negative selection acting on the HBV genome [30]. In this study, we reexamined the adaptive evolution acting in the overlapping P and S regions using a significantly larger dataset compared to earlier studies. In addition to investigating d_n/d_s we also examined the association between conserved sites and structured protein domains, and between sites under positive selection and epitope regions. Although the crystal structure of both proteins remains to be determined, in this study we primarily focused on the secondary structure which can be predicted with higher confidence (∼80%) [44]–[46]. Furthermore, our structure prediction for P is based on the tertiary RT structure of the HIV RT, which contains many regions that are well conserved between the two viruses, further increasing the accuracy of our prediction.

As a DNA-dependent DNA polymerase (DDDP) and an RNA-dependent DNA polymerase (RDDP), the HBV polymerase requires sequence conservation to implement its normal catalytic function. In particular, a highly conserved catalytic core in the proximity of the YMDD motif is crucial for replication competence, even a single amino acid mutation may severely decrease the catalytic activity of polymerase. Long-term antiviral treatment with nucleoside analogues (NA) such as Lamivudine and Entecavir often results in YMDD mutations and mutations have occasionally been reported in Tenofovir treatment [47], [48]. Although they can tolerate antivirals, the mutants may reduce their replication competence in comparison to the wild type virus [49]. To remove any bias that might result from anthropogenic selection so as to focus on investigating the evolutionary pattern under natural selection pressure, sequences with antiviral therapy were excluded.

In the HBV P and S overlapping region, several interrelated factors including structural, immunological and evolutionary constraints affect relative gene domain arrangement and genetic variation. The requirement for maintaining structural protein elements (α-helix and β-sheet) appears to govern the conservation of residues and restrict the virus variation within the genome. In particular, as a stable and important structural element, few nonsynonymous substitutions exist in the α-helices. In contrast, the β-sheets can accommodate more variation which is consistent with results from an analysis of structural restraints in HIV [50].

Many sites under positive selection were detected in both genes in this study, indicating they are both undergoing adaptive evolution and suggesting that they each play important roles in HBV survival. The adaptive evolution identified in S was almost exclusively located within the epitope regions, suggesting a role associated with evasion of host immune system. However, we identified additional sites outside the epitopes with weaker statistical support, that may also be associated with alternative roles within the virus life cycle and which would warrant further investigation. [51], [52]. On the other hand, the positive selection sites in P were mainly located in the spacer domain which corresponds to PreS in S, suggesting that this “dispensable” spacer domain may in fact be important for the HBV life cycle. For instance, amino acid substitutions in this region may affect the catalytic activity of the polymerase which is essential in the earliest steps in the HBV life cycle. Furthermore, recognition of P antigens may limit early HBV spread and its high degree of conservation may prevent viral escape via mutations in T cell epitopes ([22], [23], see [53] for review). In addition, two studies report the discovery of CD8+ and CD4+ T cell epitopes in the polymerase, although the regions are less than other HBV antigens [23], [24] suggesting that the variation in P may also be associated with limited immune escape.

While it is tempting to associate the observed correlation between positive selection and epitopes with response to host immune pressure as observed in HCV [10], [11], it is important to acknowledge the differences between the HBV and HCV virus life cycles [2] as well as considering the complex interplay between virus and host in HBV necessary for establishing a chronic infection. Multiple factors, including immune evasion, persistence of cccDNA and infection of immunologically privileged sites, contribute to HBV persistence [54]. Chronic infection in HBV is characterized by a weaker immune response [1] and production of excess HBsAg that captures most antibodies and HBV-specific immunosuppression play critical roles in immune evasion [55]. With regards to HBV-specific immunosuppression, from a viral perspective, the various proteins contribute in different ways towards achieving this modulated response. For example, HBeAg can induce tolerance in core protein (HBcAg) specific T cells, reducing their efficacy to kill infected cells [56], [57] and the X protein appears capable of inhibiting antigen processing and presentation, reducing the visibility of infected cells to the immune system [58]. Conversely, from the host perspective, a number of factors have been proposed or demonstrated to be associated with the inadequate cellular immune response including: deficient antigen presentation [59], a limited range of virus-specific T cells [60], anergy or exhaustion of rapid onset of T cell response due to antigen overload and T cell overstimulation [61], induction of regulatory T cells and ramping up of negative regulatory signals such as regulatory T cell mediated immunosuppression [62]. Thus, in a chronic infection, these factors will suppress the host immune response and, consequently, the selection pressure acting on the virus, complicating the interpretation of our results and the significance of our association between positive sites and epitopes.

Interpretation of our results are further confounded by the fact that our dataset represents a broad cross-section of the HBV patient population (comprising chronic infection, acute infection and HBV carriers) and, as such represents an average across multiple patients over the course of an HBV infection. Furthermore, the identified epitope/positive selection signal superimposed over the background noise generated by random mutations due to relaxed selection pressures on flexible loop regions that are free of secondary structure constraints (i.e. α-helix and β-sheet). Finally, it has been proposed that coevolutionary relationship may exist between sites in PreS (corresponding to spacer in P) and sites elsewhere in the genome that are expressed as compensatory mutations. [63].

The variable region (nucleotides 1–498 in the large surface protein reading frame) is essential both for P and S as it encodes the spacer in the P ORF and the PreS region in the S ORF. On one hand, this region provides the flexibility for changes in protein conformation. In P, some residues in TP and RT are believed to be associated with P-ε binding which triggers pregenomic RNA (pgRNA) encapsidation and DNA synthetic priming [64], [65]. This step can proceed only if the P protein changes its conformation from the stable state to the “active” state in the presence of chaperons such as hsp40 and hsp70 [66]–[69]. On the other hand, in S, besides its function in viral entry and infection [70], [71], the peptide from residues 2–48 is the target sequence for the hepatocyte-specific receptor [15]. Also, the flexibility of the Pre-S variable region allows mobilization of this region between the inner and outer face of the virus membrane [72], [73]. This region also has the flexibility to accumulate variation for adapting to immune pressure. PreS harbors many T cell and B cell epitopes for the large surface protein. Hence, adaptive evolution in this region may play a role in HBV escape from the host immune system.

The domain arrangement in P and S further demonstrates the compromise necessary to fulfill the distinct requirements of each protein. The spacer domain in the polymerase (corresponding to PreS in large surface protein) seems to have effectively balanced the distinctive conservation and variation requirements occurring within the overlapping region. Moreover, the four membrane-spanning regions composed of the conserved α-helices contribute to the stability of the S protein, while the intervening variable major hydrophobic regions and C-terminus exposed to the outer face of viral envelope may contribute towards helping HBV evade attack from the host immune system.

In conclusion, our findings indicate the spacer domain, which corresponds to PreS in S, provides an important function by serving as a harbor for maintaining heterogeneity for environmental adaptation as well as providing flexibility for conformational changes and response to immune selective pressure.