ABSTRACT
Hepatitis B virus (HBV) encodes a multifunction reverse transcriptase or polymerase (P), which is composed of several domains. The terminal protein (TP) domain is unique to HBV and related hepadnaviruses and is required for specifically binding to the viral pregenomic RNA (pgRNA). Subsequently, the TP domain is necessary for pgRNA packaging into viral nucleocapsids and the initiation of viral reverse transcription for conversion of the pgRNA to viral DNA. Uniquely, the HBV P protein initiates reverse transcription via a protein priming mechanism using the TP domain as a primer. No structural homologs or high-resolution structure exists for the TP domain. Secondary structure prediction identified three disordered loops in TP with highly conserved sequences. A meta-analysis of mutagenesis studies indicated these predicted loops are almost exclusively where functionally important residues are located. Newly constructed TP mutations revealed a priming loop in TP which plays a specific role in protein-primed DNA synthesis beyond simply harboring the site of priming. Substitutions of potential sites of phosphorylation surrounding the priming site demonstrated that these residues are involved in interactions critical for priming but are unlikely to be phosphorylated during viral replication. Furthermore, the first 13 and 66 TP residues were shown to be dispensable for protein priming and pgRNA binding, respectively. Combining current and previous mutagenesis work with sequence analysis has increased our understanding of TP structure and functions by mapping specific functions to distinct predicted secondary structures and will facilitate antiviral targeting of this unique domain.
IMPORTANCE HBV is a major cause of viral hepatitis, liver cirrhosis, and hepatocellular carcinoma. One important feature of this virus is its polymerase, the enzyme used to create the DNA genome from a specific viral RNA by reverse transcription. One region of this polymerase, the TP domain, is required for association with the viral RNA and production of the DNA genome. Targeting the TP domain for antiviral development is difficult due to the lack of homology to other proteins and high-resolution structure. This study mapped the TP functions according to predicted secondary structure, where it folds into alpha helices or unstructured loops. Three predicted loops were found to be the most important regions functionally and the most conserved evolutionarily. Identification of these functional subdomains in TP will facilitate its targeting for antiviral development.
INTRODUCTION
Hepatitis B virus (HBV) remains a major cause of acute and chronic viral hepatitis, liver cirrhosis, and hepatocellular carcinoma. Although an effective human immune response can cure the infection, currently available antiviral therapeutics cannot. They are, however, effective at suppressing viral load in the blood. HBV is a member of the Hepadnaviridae family, which also includes animal viruses such as the duck hepatitis B virus (DHBV), and replicates through a complex strategy of reverse transcription (1–3). Despite its functional complexity, HBV has a small and deceptively simple genetic code, consisting of a 3.2-kb, partially double-stranded, relaxed circular DNA. This small DNA genome has only four genes and produces only one protein with enzymatic functions. This enzyme is the viral polymerase (P), a specialized reverse transcriptase (RT) (4–6).
The P protein synthesizes viral DNA, using as its initial template an RNA stem-loop structure, epsilon (ε), located near the 5′ end of the pregenomic RNA (pgRNA), which is specifically recognized and bound by the P protein (7–11). Subsequently, the P protein-pgRNA complex is specifically recognized by the viral capsid protein to facilitate its packaging into the assembling nucleocapsid, the site of DNA synthesis (12–14). Thus, the ε RNA also acts as an RNA packaging signal.
P is divided into four domains: the terminal protein (TP), spacer, RT, and RNase H domains. Although the catalytic activities of the P protein, DNA synthesis and RNA degradation, lie within the RT and RNase H domains (6, 15), respectively, the TP domain is important for several steps that are required to synthesize viral DNA. The TP domain contains the essential tyrosine residue Y63 (Y96 in DHBV), whose hydroxyl group provides the substrate for the covalent attachment of the first nucleotide of the viral minus-strand DNA to the P protein; that is, the TP domain acts as the primer for DNA synthesis (16–18). Additionally, the TP domain is needed for binding and packaging the pgRNA into nucleocapsids (7, 9, 10, 19–22). Each of these steps is a potential target for antiviral drug intervention. Thus, due to its multiple roles in viral replication and a lack of cellular homologs, TP is, in principle, an attractive therapeutic target.
Currently, no high-resolution structural data are available for the P protein, including the unique TP domain. Obtaining large quantities of active protein has been difficult due to the requirement for host cell chaperones in P protein folding (4, 6, 10, 23, 24), and there are several regions in TP that are predicted to be disordered. In contrast to the RT domain of the HBV P protein, which has been modeled using the homologous HIV RT domain (25), the sequence of the TP domain is unique to the Hepadnaviridae family, and thus there is no viable homology model for it.
On the other hand, some structural inferences can be made using functional homologs of the TP domain. For example, bacteriophages such as φ29, members of the Adenoviridae family, and bacteria of the genus Streptomyces all encode a TP for priming DNA synthesis (26–28). All known TP proteins use a Tyr, Ser, or Thr (which have a free hydroxyl group [–OH]) as the site of protein priming, and this amino acid in all cases is located in a disordered priming loop between alpha helices (29, 30). The lack of sequence similarity among these TP proteins and their priming loops indicates convergent evolution of this protein priming activity. Regardless of the TP sequence, all priming loops must access the polymerase active site together with the template DNA/RNA, which is likely facilitated by the flexibility of this loop.
The lack of a structural model for the HBV TP domain necessitates the use of other methods to identify functionally important regions, such as sequence conservation analysis, structural modeling, and directed mutational studies. Although several studies of these types have been performed, the current study collates these with a specific focus on the TP domain and includes the analysis of a number of novel TP mutants.
Advances have been made only recently to allow the testing of HBV P protein functions biochemically, whereas assays for RNA binding and protein priming activities of the DHBV P protein were established much earlier (21, 31, 32). Studies using DHBV have led to the identification of functionally important regions, such as the T3 motif (residues 153 to 160 in HBV and 176 to 183 in DHBV), which is conserved in HBV and DHBV (33) and thought to be critical for binding ε RNA (34). Recently, we have developed an in vitro system for testing the RNA binding and protein priming activities of the HBV P protein in a chaperone- and ε RNA-dependent manner (mimicking the in vivo requirements) (9). In this system, the HBV P protein and ε RNA are coexpressed in cells that associate with host cell chaperone proteins in an active conformation. This complex is isolated by immunoaffinity purification of the P protein, which can be tested for RNA binding and protein priming activity levels.
Using this in vitro system, we also discovered a novel protein-primed terminal transferase activity of the HBV P protein that is independent of the ε RNA and stimulated by Mn2+ (35) instead of Mg2+ (the presumed physiological cofactor used by P to carry out DNA synthesis in vivo). No in vivo significance has been established for this transferase activity of the P protein; however, the P protein requirements for the transferase activity can be different from those of authentic protein priming in the presence of Mg2+ (22, 35).
The current study tested a number of new mutants using these assays, and the phenotypes of these mutants and those from previous studies in both HBV and DHBV were combined in a meta-analysis. Furthermore, by performing a sequence conservation analysis and secondary structure prediction of the HBV TP domain, we could assign a specific function to three highly conserved predicted loops within the TP domain. This study, in the absence of high-resolution structural data, allows for a functional determination of predicted secondary structures of this critical and unique domain of the HBV P protein. This greatly increases our knowledge of the TP domain and will facilitate the development of antivirals to target the TP.
RESULTS
Conserved regions align with secondary structure elements in the TP domain.To identify conserved and, thus, potentially functionally important regions in the TP domain, all known HBV (n = 583) and DHBV (n = 45) sequences with a complete TP domain were aligned to derive a consensus sequence for each virus. Using the consensus sequence, several algorithms (PsiPred, QUARK, SABLE, and I-TASSER) were used to predict the secondary structure of the TP domain of both HBV and DHBV P protein. As shown in Fig. 1, the structures predicted by the four different methods were in general agreement, revealing overlapping secondary structures of TP in both viruses. The predictions categorized the sequences into three main structure types, unstructured, alpha helix, and beta sheet, and indicated that the TP domain is largely helical. By overlaying these predictions onto aligned HBV and DHBV sequences, it appeared that the predicted helices and beta sheets were localized in equivalent regions of the TP domain despite limited sequence conservation (39% amino acid similarity; SIAS tool; SGCTI, Spain) (Fig. 1). It was notable that the predicted helix 6 spans the traditional boundary between the TP and spacer domains, which means that a more realistic boundary may be after this helix, with the TP domain ending at position 192, as has also been suggested recently (36).
Conservation in HBV and DHBV in the primary and predicted secondary structure of the TP domain of the P protein. For comparison between HBV and DHBV, the sequence for the TP domain is shown, divided into two rows. HBV (n = 583) and DHBV (n = 45) TP domain sequences were aligned and used to create a WebLogo, where the top letter is the most common, the height of the stack indicates relative sequence conservation at that position, and the height of individual letters indicates the proportion of sequences which use that amino acid. Several software programs (from top to bottom: PsiPred, SABLE, I-TASSER, and QUARK) were used to analyze the consensus sequence and predict secondary structure; predicted alpha helices and beta sheets are shown as cylinders and arrows, respectively. Between the two species, the majority of predicted secondary structure elements overlapped. Many of the conserved sequences (light gray boxes) are in loops between helices. Three disordered conserved loops are highlighted in boxes. Numbering for HBV is according to genotype D, and boundary numbers near the boxes indicate HBV amino acid positions.
In order to identify putative subdomains that may be important for protein structure and function, the regions of highest conservation between HBV and DHBV were mapped to determine their locations in the predicted secondary structures. There was moderate homology in both helices and loops (regions that may contain beta sheets but are between helices) among HBV isolates or among DHBV isolates (data not shown), indicating the conservation was not limited to specific secondary structures when evaluating TP sequences from a single species. However, when comparing HBV to DHBV, most of the sequence conservation was in loops and not within helices. Approximately 30 identical sites were found in regions identified as loops, while less than 10 were located in predicted helical regions (Fig. 1). We also performed a similar sequence analysis of another mammalian hepadnavirus, the woodchuck hepatitis virus, and obtained similar results (data not shown).
Mutant viral phenotypes map to loop regions in TP, demarcating functional subdomains.In addition to the above-described conservation analysis and sequence alignment, functionally important residues, as identified by mutational analyses, were mapped to specific predicted secondary structures in order to identify putative functional subdomains in TP. All previously published TP domain mutants assayed for one or more of the four activities of the HBV P protein outlined below were combined in a meta-analysis together with the findings of the current study (which will be described in detail below) (Fig. 2). Mutant P proteins were scored as showing either a wild-type (WT) or a defective phenotype in these four main activities of HBV P: ε RNA binding, RNA packaging, protein priming, and DNA synthesis. Mutational studies performed using the DHBV P protein were included in the meta-analysis only if the relevant amino acid residues matched those in the HBV P protein. Among the seven secondary structure elements predicted in TP containing either helices or loops (Fig. 2), mutants that showed a functional defect(s) were almost exclusively mapped to three major loops (with the exception of helix 5), similar to the finding that conservation between the HBV and DHBV P proteins is highest within these loops.
Meta-analysis of the phenotypes of known mutants within the TP domain of the HBV P protein. The consensus sequence for the TP domain is shown, divided into three rows, and numbered according to genotype D. Residues with homology to DHBV are highlighted (light gray boxes). Predicted alpha helices and beta sheets are shown as cylinders and arrows, respectively, as in Fig. 1. The TP domain is grouped (vertical lines) into seven subdomains according to predicted secondary structures. Known mutants in each of these subdomains are shown above the TP sequence. The filled shapes represent a defective phenotype (significant decrease or loss of function due to a mutation), and empty shapes represent WT phenotype (activity not significantly impaired by the mutation); circles are from studies with the HBV P, and squares are from DHBV studies. The novel mutants tested in the current study are circled with a gray line. The four commonly tested functions are shown by letter: B for RNA binding, R for RNA packaging, P for protein priming, and D for DNA synthesis. If a shape is absent, the mutant was not tested in that assay. The T3 motif, underlined on the right side of the 3rd row, is expanded on the left side of the 3rd row to show the many mutations that have been constructed in this subdomain. Except for helix 5, all defective mutants are found within the predicted loop subdomains.
Residues in the three predicted loops appeared to contribute to distinct steps of viral DNA synthesis. Loop one, from amino acids 41 to 81, was defined as the priming loop, not only because this subdomain contains the actual priming site, Y63, but also because a number of mutants constructed in this study (discussed below) were specifically defective in protein priming without affecting the preceding step of ε RNA binding or RNA packaging. Loop two (L2), from amino acids 97 to 134, includes a number of mutants with a defect in RNA packaging or DNA synthesis (15, 20, 22, 37–39). Loop three, also called the T3 loop, spans positions 153 to 174 and was so named because it contains the well-characterized T3 motif (positions 153 to 160). Although not targeted here for mutagenesis, it has already been analyzed fairly extensively (19, 20, 22, 33, 34, 37–40) and found to be critical for ε RNA binding (Fig. 2).
Few functionally important residues appear to reside outside the putative loops. Substitutions of residues in the three N-terminal helices, or in helix 4 or helix 6, do not appear to affect function (38, 39). Consistent with this, these helices are also the least conserved regions in the TP domain between HBV and DHBV (Fig. 1). Similarly, E31K, in the N-terminal helix subdomain, tested in the current study did not affect any P functions (also see below). In contrast, helix 5 does appear to be functionally important and is the most conserved helix between HBV and DHBV. This is especially true for residues near the T3 motif (Fig. 1). Specifically, helix 5 contains a conserved Tyr residue, Y147, which has been found to be important for RNA binding, priming, and DNA synthesis (19, 22, 34, 38).
Novel TP mutations constructed in the current study to further map functional sites.Taking into consideration sequence conservation, secondary structure prediction, and the meta-analysis of previous studies of mutant phenotypes, we constructed a number of novel TP mutants to further define functionally important regions of the TP domain. The phenotypes of these mutants were examined in detail to identify any specific functional defects in ε RNA binding, protein priming, and pgRNA packaging. Nineteen novel mutants (ten substitutions, three N-terminal truncations, and six internal deletions) were chosen to probe conserved residues not previously targeted in other studies and to define minimal domains required for specific P functions. In addition, near the Y63 priming site there are several conserved hydroxyl-containing residues, T60, S64, S65, and T66. In the aforementioned TP proteins from other viruses/organisms, hydroxyl-containing amino acid residues are also located near the priming site (29, 30), suggesting there is a conserved phosphorylation requirement. Substitutions of these residues thus were made to mimic either phosphorylation (with Glu) or nonphosphorylation (with Ala) in order to test the potential role of phosphorylation at these sites for TP functions.
To determine the phenotypes of these novel TP mutations, the WT or mutant HBV P protein was expressed in cells in one of two schemes. The first scheme is to coexpress the P protein with the ε RNA. This two-component assay allows testing of the RNA binding activity of the P protein and of its protein priming activity as a purified P protein-ε RNA complex. The second scheme is to cotransfect the P expression construct with a second one that does not express the P gene (Pol− or P−) but expresses the pgRNA and all other viral proteins. This complementation assay allows formation of the viral nucleocapsid for testing pgRNA encapsidation levels (RNA packaging). As a negative control, a previously characterized P mutant containing only amino acids 1 to 175 and 300 to 775 was used, which is inactive in all P functions tested to date (22).
RNA binding phenotypes of the new TP mutants.When the P protein and the ε RNA are coexpressed in cells, they form a ribonucleoprotein complex that can be isolated by immunoprecipitation using a FLAG epitope tag on the N terminus of P. The amount of ε RNA bound to the P protein was measured and normalized to the total amount of ε RNA expressed in the cells. Most of the TP mutants maintained ε RNA binding activity (Fig. 3), including Δ2-60 and Δ56-66 (lanes 13 and 23). In contrast, constructs with deletions in residues 67 to 80 (Δ2-75 and Δ70-80) were strongly impaired in ε RNA binding (lanes 14 and 24). This established a new minimal N-terminal boundary of TP for RNA binding by showing that the first 66 residues of TP were dispensable (Fig. 3; see also Fig. 6B). The internal deletion mutants that remove the two halves of the L2 loop, Δ101-115 and Δ116-130, were also defective in RNA binding (lanes 7 and 11). In addition, the charge-swap substitution R114E within this loop was partially defective in RNA binding (lane 5). These results highlighted a putative role for the L2 loop in pgRNA interactions in addition to the well-characterized T3 loop.
ε RNA binding activities of HBV TP mutants. The HBV P, either WT or mutated, was expressed in HEK293T cells together with the ε RNA and purified by immunoprecipitation. RNA was extracted from both the immunopurified P (top) and cytoplasmic lysates (bottom) and resolved by urea-PAGE gel, followed by detection using an ε RNA-specific radiolabeled riboprobe. A control with no ε RNA expression was included (lane 2). The negative control (lane 25) expressed a mutant P containing only amino acids 1 to 175 and 300 to 775, known to be inactive in all P functions. This experiment was repeated at least three times; representative images are shown. Mutants with a defective phenotype are highlighted in boldface.
All TP mutants constructed here, including the deletion/truncation mutants, were expressed at WT levels (Fig. 4A, bottom), suggesting that they were unlikely to be grossly misfolded. This is also consistent with the expectation that deletion of flexible loops is unlikely to grossly alter protein structure. However, for those mutants that showed a defect in ε RNA binding, the first functional step that we could monitor, we could not formally exclude the possibility that the observed effects of the mutations were due, in part, to protein misfolding. On the other hand, for the mutants that were active in ε RNA binding, global misfolding could be excluded.
In vitro protein priming activities of HBV TP mutants. (A) The HBV P, either WT or mutated, was expressed in HEK293T cells together with the ε RNA and purified by immunoprecipitation. Radiolabeled dGTP and either Mg2+ (top) or Mn2+ (middle) were added to allow protein priming, during which the nucleotide(s) become covalently attached to the P protein. Levels of radiolabeled P proteins were detected by autoradiography (top and middle) following SDS-PAGE resolution of the priming reactions. Levels of total P proteins were detected by Western blotting (bottom) using an antibody against the triple-FLAG tag at the N terminus of each P construct. A control with no ε RNA expression was included (lane 2). The negative control (lane 25) expressed a P mutant containing only amino acids 1 to 175 and 300 to 775, known to be inactive in all P functions. This experiment was repeated at least three times; representative images are shown. Mutants with a defective phenotype are highlighted in boldface. (B) Analysis of DNA synthesis products by the WT HBV P protein and the H140A mutant in the protein-primed terminal transferase reaction. WT or H140A mutant P was expressed in HEK293T cells and purified by immunoprecipitation. Priming assays were performed with Mn2+ and radiolabeled TTP. Priming products were then mock treated (lanes 2 and 4) or treated with Tdp2 (lanes 1 and 3) in order to release the nucleotides/DNA strands attached to the P protein. Samples were resolved by urea-PAGE and detected by autoradiography. The lengths of the nucleotides/DNA strands are indicated.
Protein priming phenotypes of the new TP mutants.The isolated P protein-ε RNA complex was tested for in vitro protein priming activity as described in Materials and Methods. A radiolabeled nucleotide substrate was included along with a buffer containing either Mg2+ or Mn2+ to test whether the complex possesses authentic ε RNA-dependent protein priming or ε RNA-independent terminal transferase activity. For the Mg2+ priming reaction, labeled dGTP was used as the substrate, and if the complex was active, it would become covalently attached as dGMP (representing the 1st nucleotide of the viral minus-strand DNA) to Y63 in TP. Labeled dGTP was also used for the Mn2+ priming reactions unless noted otherwise.
As expected, all mutants that were defective in ε RNA binding were defective in the Mg2+-dependent priming reaction, which depends on P-ε interaction (Fig. 4A). These included R114E (lane 5), the two L2 loop deletions Δ101-115 and Δ116-130 (lanes 7 and 11), and the long N-terminal truncation Δ2-75 (lane 14). Furthermore, the four small internal deletions in the priming loop (Δ39-47, Δ48-55, Δ56-60, and Δ70-80) were all severely defective in priming with Mg2+ (Fig. 4A, lanes 21 to 24; see also Fig. 6B), providing additional support for the proposed role of the priming loop. Importantly, all these deletions, except for Δ70-80, maintained WT levels of ε RNA binding, indicating a specific role of the priming loop in protein priming distinct from the involvement of TP in ε RNA binding.
A priming-specific defect was also observed in substitutions of the potential phosphorylation sites near the priming site. The phosphomimetic Glu substitutions T60E and S64E/S65E/T66E were defective in protein priming but maintained ε binding activity (Fig. 4A, lanes 16 and 19; see also Fig. 6A). The triple Ala substitution S64A/S65A/T66A also showed a severe defect in priming (and again no defect in RNA binding), whereas the single Ala substitution T60A showed WT priming levels (Fig. 4A, lanes 15 and 18; see also Fig. 6A).
N-terminal truncations also helped to define the functional domain boundary necessary for priming. An N-terminal truncation of 59 residues (Δ2-60), extending into the priming loop, caused a loss of priming activity, while truncation of 12 residues (Δ2-13) still retained a weak level of priming (Fig. 4A, lane 12). Because both of these truncations retained WT levels of ε RNA binding, a specific role in protein priming could be assigned to these N-terminal residues in addition to the priming loop.
Most of the TP mutants that were deficient in Mg2+ priming also had defects, albeit to a lesser extent, in Mn2+-dependent priming. This is consistent with our previous observation that the Mn2+ priming (transferase) activity is more refractory to mutations in the P protein than the Mg2+ priming reaction (22), likely a reflection of less stringent requirements for the former compared to the latter (for example, ε RNA is dispensable for Mn2+ priming but required for Mg2+ priming) (35). In contrast, one mutant, H140A, which showed only a moderate defect in Mg2+ priming, showed a severe defect in Mn2+ priming (Fig. 4A, lane 6), the only mutant identified to date with this phenotype. To further define the consequences of this variant, the DNA synthesis (transferase) products produced during priming with Mn2+ were released from the P protein using the enzyme tyrosyl-DNA phosphodiesterase 2 (Tdp2). Here, labeled TTP was used as the substrate, which is known to be preferred for the transferase activity of P, and released DNA products were resolved by urea-PAGE and identified as described previously (35). Whereas the WT P protein synthesized poly(T) DNA strands both short (<30 nucleotides [nt]) and long (30 to >200 nt) in relatively equal amounts (Fig. 4B, lane 1), the H140A mutant preferentially synthesized poly(T) strands of 10 to 20 nt in length during the priming reaction (Fig. 4B, lane 3). The H140A mutation thus appeared to affect the processivity of the P protein during Mn2+ priming. Also of note in reactions with Mn2+, the Δ56-66 mutant, which no longer contains the primer Y63 residue, was still able to prime DNA synthesis with Mn2+ at very low levels (Fig. 4A, lane 23). This was likely due to cryptic site priming, which is known to occur when the normal priming site is removed, especially in the presence of Mn2+ (35, 41, 42). It is possible that the priming signal produced by the SST64-66AAA and SST64-66EEE mutations, which are directly adjacent to the authentic priming site (Y63), also could have initiated from cryptic priming sites, instead of Y63, if these mutations in close proximity to Y63 affected the use of Y63 to prime DNA synthesis. However, we consider this unlikely, as we have never detected cryptic site priming by the WT HBV polymerase in the presence of Mg2+ (cryptic site priming by the HBV polymerase has been detected so far only in the presence of Mn2+), and even with Mn2+, the priming activity from cryptic sites is ca. 1% or less of that from Y63.
RNA packaging phenotype of the new TP mutants.The extent of pgRNA encapsidation, mediated by the WT or mutant P proteins in the complementation assay, was measured after native agarose gel electrophoresis. Here, cytoplasmic HBV nucleocapsids were resolved as particles and their pgRNA content quantified following transfer of the nucleocapsids to nitrocellulose membrane. The amount of total capsids, with or without packaged pgRNA, was then determined by Western blot analysis and used to normalize RNA packaging levels.
As shown in Fig. 5A, several mutants were found to be defective in RNA packaging, including the deletions Δ101-115 and Δ116-130 in the L2 loop (lanes 6 and 10) as well as the N-terminal truncations Δ2-60 and Δ2-75 (lanes 16 and 17). Loss of ε RNA binding likely accounts for the absence of RNA packaging in the two deletion mutants in the L2 loop and the Δ2-75 N-terminal truncation. In contrast, Δ2-60 (lane 16) was fully functional in ε RNA binding but defective in RNA packaging. Also, the shorter N-terminal truncation, Δ2-13, showed a moderate RNA packaging defect (lane 15), indicating that the N-terminal portion of TP contributes to pgRNA packaging (as well as protein priming, as described above) despite its dispensability in ε RNA binding. On the other hand, of the four internal deletions covering the priming loop, only the Δ70-80 mutant showed a strong defect in RNA packaging (lanes 18 to 21), suggesting that most of the priming loop was not involved in either ε binding or RNA packaging. Similarly, the substitution mutants at the putative phosphorylation sites, T60A and T60E as well as S64A/S65A/T66A and S64E/S65E/T66E, in the priming loop were all functional in pgRNA packaging (lanes 11 to 14).
RNA packaging activities of the HBV TP mutants. (A) Nucleocapsids were purified from HEK293T cells which were cotransfected with the indicated HBV P construct and pCMVHBV-Pol−, which expresses the pgRNA and all HBV proteins except P. (Top) Levels of RNA packaging activity were measured by probing a nitrocellulose membrane with a plus-strand-specific probe following resolution of nucleocapsids on an agarose gel. (Bottom) Levels of capsid proteins were measured on the same membrane by Western blotting with an anti-HBV core antibody. The negative control (lane 22) expressed a P mutant containing only amino acids 1 to 175 and 300 to 775, known to be inactive in all P functions. Mutants with a defective phenotype are highlighted in boldface. (B) Effect on pgRNA packaging by expressing various levels of the P and capsid proteins in the trans-complementation assay. HBV nucleocapsids were purified from HEK293T cells that were cotransfected with the indicated P:P− ratios and analyzed as described for panel A. Samples with less RNA packaging at the 1:9 ratio of P:P− are highlighted in boldface. These experiments were repeated at least three times; representative images are shown.
R114E, which showed a defect in ε RNA binding, appeared to package almost WT levels of pgRNA into nucleocapsids in initial experiments where the P and P− (or Pol−) constructs were used at a 1:1 ratio (Fig. 5A, lane 4 versus 1, and B, lane 3 versus 1). As ε RNA binding is needed for pgRNA packaging, this result was quite unexpected. To resolve this apparent discrepancy, we considered the experimental conditions for the trans-complementation assay used to measure RNA packaging. Given that the levels of P protein active in pgRNA packaging are thought to be much lower than those of the capsid protein during HBV infection, the 1:1 ratio used to express P and the P− genomic construct would lead to artificially high (saturating) P protein concentration relative to the capsid concentration. Consequently, RNA packaging activity of WT P might be underestimated under these conditions. Therefore, we lowered the amount of the P expression construct and increased the P− construct (to a ratio of 1:9) in the complementation assay for both the WT and the R114E mutant. The WT P supported higher levels of RNA packaging at the 1:9 ratio, where the P levels would be lower but the capsid levels higher (Fig. 5B, lane 2), indicating that the P protein was indeed saturating and the capsid level was limiting for RNA packaging with the 1:1 ratio of transfection (Fig. 5B, lane 1). In contrast, for the R114E mutant, a decrease in P expression resulted in reduction of RNA packaging despite an increase in capsid levels (Fig. 5B, lane 4 versus 3). At these more physiological (lower) levels of P, the pgRNA packaging levels by R114E were ca. 2- to 3-fold lower than those of WT P protein (Fig. 5B, lane 4 versus 2), consistent with a defect in ε RNA binding. We also repeated the RNA packaging assay for all other P mutants at the 1:9 (P:P−) ratio. The results obtained were generally consistent with those obtained at the 1:1 ratio (Fig. 5B, lanes 5 to 8, 11, and 12 and data not shown). However, the internal TP deletion mutant, Δ70-80, behaved similarly to the R114E mutant in that it also showed a more severe defect in pgRNA packaging at the 1:9 ratio (Fig. 5B, lanes 9 and 10). These results indicate that caution must be exercised in trans-complementation experiments to maintain the physiologically relevant stoichiometry of the components, especially for mutants with a partial functional defect, so as to avoid artificial results. This RNA binding-negative but packaging-positive discrepancy has been witnessed in other TP domain mutants before (19) (Fig. 2), which might have also been an artifact of the experimental conditions.
We noted that the levels of capsid protein varied greatly (Fig. 5A, bottom), but in general these levels were inversely correlated with the RNA binding activity of the P protein (Fig. 3, top). For example, the highest level of capsid protein expression (Fig. 5A, lanes 6, 10, and 22) was associated with a loss of ε RNA binding by the P protein mutants (Fig. 3, lanes 7, 11, and 25). This result was consistent with previous reports (22, 43, 44) and could be explained by the known inhibitory effect of P on pgRNA translation due to the occlusion of ribosomes from pgRNA upon P protein-ε RNA interaction. Other than the effects of the polymerase mutations on core protein expression, some variations in transfection efficiency might have contributed to the variation of core expression. However, by normalizing the pgRNA packaging efficiency to the capsid levels, the potential influence of transfection efficiency on pgRNA packaging efficiency was effectively removed. We also measured the levels of total core protein (including free core protein subunits as well as capsids) by denaturing SDS-PAGE/Western blot analysis and found that total core protein levels were directly correlated with levels of assembled capsids (data not shown). This indicated that neither the WT nor mutant polymerase proteins affected capsid assembly, consistent with a general lack of evidence for a role of the polymerase in capsid assembly per se.
DISCUSSION
In the absence of a high-resolution structure for the HBV P protein, our analysis of the TP domain, employing multiple computational and biochemical approaches, adds to our understanding of P structure and function. One interesting finding was that, despite significant amino acid sequence differences, the HBV and DHBV P proteins are predicted to share core secondary structural elements throughout the TP domain. Furthermore, the majority of conserved amino acid residues were localized to regions outside predicted helices in the putative flexible loop regions of the TP (Fig. 1). To carry out its diverse roles in viral replication, the TP domain is known to interact with multiple factors, including the pgRNA, the capsid protein, other domains within the P protein, and host factors (6, 45), these highly conserved and presumably flexible loops are likely to be involved in these interactions. Consistent with this proposition, the conserved loops are predicted to be surface exposed in a three-dimensional model of TP (QUARK) (Fig. 6C).
Summary of TP mutant phenotypes. The activity levels of the substitution mutants (A) and N-terminal truncation and internal deletion mutants (B) are represented by symbols (>67%, +++; 34 to 66%, ++; 10 to 33%, +; <10%, −). Results represent averages from at least three independent experiments. *, RNA packaging levels of these mutants were based on comparison with the WT at a 1:9 ratio of P:P− construct in transfection, which were significantly lower relative to the WT than those measured at the 1:1 ratio (Fig. 5B). All other mutants showed similar levels of RNA packaging relative to the WT at either 1:1 or 1:9 P:P−. §, More severe defect in priming activity with Mn2+ than Mg2+ (Fig. 4). (B) The priming loop and L2 loop (gray boxes) are shown. For the mutant constructs, solid bars represent the sequences that are retained, and thin dipped lines represent deleted sequences. The protein priming results were the same using both Mn2+ and Mg2+ and are shown in a single column for clarity. At the bottom, the minimal necessary domains (thick bars) for ε RNA binding (downstream of position 66) and protein priming (downstream of position 13) are shown. Only the first 140 positions are shown (the remainder of TP is indicated by “…”). The image is drawn to scale. (C) The three-dimensional structure of TP as predicted using QUARK. The seven putative secondary structural elements of TP are indicated and shown in different shades. The N and C termini of TP are labeled, as is the Y63 priming site. The three loops, extending outward, are predicted to interact with the viral RNA or other factors during the various steps of viral DNA synthesis (see the text for details).
Our analysis also revealed that the predicted loops in TP contain almost all of the residues whose substitutions caused a defect in P functions, consistent with the predicted functional importance of the TP loops based on their sequence conservation. In contrast, most of the predicted helical regions in TP were devoid of mutants with defective phenotypes, consistent with a relative lack of sequence conservation in these helices. The predicted helices likely represent a “folding core” which is not affected significantly by the sequence variations, whereas the presumably surface-exposed loops provide sequence-specific interaction surfaces for RNA binding, positioning of the primer residue Y63, and interfacing with the catalytic core in the RT domain.
Furthermore, we could assign a specific function to each of the loops in TP (Fig. 2). The T3 loop (positions 153 to 174) contains the T3 motif, which has been mutated extensively (19, 20, 22, 33, 34, 37–40) (Fig. 2). It is highly conserved among mammalian and avian HBV and critical for interactions with pgRNA. Residues that lie inside the T3 loop but outside the T3 motif itself (KAGILYKR; positions 153 to 160) have not yet been shown to be important for RNA binding, but some of those residues, such as T162 and Y173, have been shown to be necessary for RNA packaging (22, 38, 46). The L2 loop (positions 97 to 134) includes several mutants with a defective RNA packaging phenotype (15, 20, 22, 37–39) (Fig. 2). This predicted loop may also play a role in the preceding step of pgRNA binding, as large internal deletions Δ101-115 and Δ116-130 and the substitution R114E in this subdomain showed defects in RNA binding (Fig. 6). Another secondary structure region that showed defective phenotypes is the priming loop (positions 41 to 81) (Fig. 2). All of the defective substitution mutants in the priming loop which have been tested affected priming but had no or only modest effects on ε RNA binding or pgRNA packaging (22, 38), including several internal deletions which removed ∼10 amino acid regions (Fig. 6B).
The minimal region necessary for RNA binding was previously shown to span residue 42 to the end of the TP domain, as well as most of the RT domain (11). We have shown here that the Δ2-60 truncation retained a moderate level of RNA binding but Δ2-75 was almost totally defective in RNA binding. In addition, the internal deletion mutant Δ56-66 was able to bind ε RNA but Δ70-80 was not. Together, our results here have localized the N-terminal boundary of the minimal RNA binding domain around positions 67 to 70 in TP (Fig. 6B). Regarding protein priming, our previous study showed that removing just the first 20 amino acids from the N-terminal end of TP (or from 175 to 199 at its C-terminal end) resulted in a total loss of priming activity (22), suggesting that the entire TP domain was required for priming. Since our secondary structure prediction analysis localized the first helix in TP to amino acids 5 to 13, we constructed the mutant Δ2-13 to remove this entire helix. This mutant was still able to carry out protein priming (and subsequent DNA synthesis; data not shown), albeit at only ∼20% of WT levels. Thus, the first predicted helix in the N-terminal region of TP is not essential for priming, although it can contribute to it (Fig. 6B).
Generally, our analysis suggests that the N-terminal part of TP primarily plays a structural role, since substitutions within this region are generally tolerated for TP functions but larger deletions impair RNA packaging and/or protein priming (Fig. 2 and 6). For example, the truncation mutants Δ2-13 and Δ2-60 both showed a defect in protein priming and RNA packaging; however, multiple substitutions (E31K, N42A, H54A, and K55E) or small deletions (Δ39-47, Δ48-55, and Δ56-66) in this region remained functional in RNA packaging and/or protein priming (Fig. 6). This is also in agreement with results obtained using several previously tested substitution mutants in the N-terminal helix subdomain, which also did not affect function (Fig. 2).
Several conserved Ser/Thr residues are located near the Y63 priming site, which potentially can be phosphorylated. However, the similar (inhibitory) effect on authentic protein priming (i.e., Mg2+ and ε dependent) of the phosphomimetic (Glu) and nonphosphorylatable (Ala) substitutions of SST at positions 64 to 66 suggests that their functional role in priming is not phosphorylation dependent. The conserved hydroxyl groups in the side chains of these residues may be involved in direct interactions, such as hydrogen bonding, that are critical for protein priming instead of serving as acceptor sites for phosphorylation. On the other hand, only the Glu, but not the Ala, substitution at T60 was defective in Mg2+ priming. This indicates that the conserved hydroxyl group at T60, in contrast to those at 64 to 66, is not functionally required for priming. Furthermore, it suggests that phosphorylation at T60 is unlikely to occur during viral replication, but if it should occur, it could inhibit protein priming. The HBV P protein has been shown to be phosphorylated in vivo on at least two sites, although the precise sites have never been defined (47). Several Ser/Thr residues in the TP domain, downstream of those targeted here, were predicted as potential phosphorylation sites (NetPhos 2.0; data not shown). Future studies may be warranted to further determine the role of phosphorylation of TP (and P in general) in HBV replication.
In summary, by dividing the TP domain into subdomains based on their predicted secondary structure and by assigning a specific function(s) to each subdomain through mutational analyses, we have identified three functionally critical and evolutionarily conserved loops in TP. These flexible loops are likely surface exposed, poised for interactions with other factors such as the pgRNA and the other regions of the P protein, while the helices may form a folding core, serving a mostly structural role. Given the functional importance of the TP domain in multiple stages of the viral replication cycle and the lack of high-resolution structural data for TP or any other domain of the P protein, our analysis provides a useful map of putative TP structure-function relationships until further structural data come to light. The newly defined subdomains, with their assigned function, also provide an excellent foundation for developing therapeutics targeting this unique and multifunctional domain of the HBV P protein.
MATERIALS AND METHODS
Sequence analysis and quantification of results.The program MEGA 4.0 was used for sequence viewing and alignment (48). Several secondary structure prediction software programs were used: PsiPred (49), QUARK (50), SABLE (51), and I-TASSER (52). The maximum input sequence for QUARK is 200 amino acids; therefore, only amino acids 1 to 200 of the DHBV TP sequence were included for analysis with QUARK. Phosphorylation site prediction was performed using NetPhos 2.0 (53). Phosphorimager screens were scanned using a Typhoon 9400 imager and quantified by Quantity One software. The chemiluminescent signal of FLAG-tagged polymerase or core protein on Western blots was detected by a ChemiDoc MP system and analyzed by BioLab software. All experiments were repeated at least three times, and representative blots are shown for each set of experiments.
Plasmids.Wild-type (WT) HBV P protein was expressed using pcDNA-3FHP, which contains the HBV P gene (genotype D, strain ayw, GenBank accession number U95551.1 ) with three copies of a FLAG epitope tag at the N terminus under the cytomegalovirus (CMV) promoter (9). A short ε RNA construct was coexpressed using pCMV-HE as previously described (9, 35), but in this study the native HBV polyadenylation site and several bases downstream were removed by replacing the sequence from the ATAAA polyadenylation site to the XbaI site (nt position 1919 to 1992 in U95551.1 ) with two restriction enzyme sites, BamHI and NotI. In so doing, an RNA of 275 bases is expected, not including the poly(A) tail specified by the bovine growth hormone polyadenylation site on the vector. This epsilon RNA expression plasmid is called pCMV-HE-ΔpA. PCR-mediated mutagenesis was used to produce N-terminal truncations, internal deletions, and amino acid substitutions in P, using the pCDNA-3FHP background. To facilitate PCR-mediated manipulations, an AgeI restriction site was introduced using the New England BioLabs Q5 site-directed mutagenesis kit. This is a silent mutation, T2512G in U95551.1 , which maintained the original amino acid sequence of the awy strain at V68 in the P gene and did not affect P expression (data not shown). Pfx polymerase (Thermo Fisher) was used for PCR according to the supplier's directions with an annealing temperature of 55°C. Primer sequences are available upon request.
Protein expression and purification and in vivo ε RNA binding.HEK293T cells were maintained in DMEM-F12 medium at 37°C and 5% CO2. These cell culture experiments are designated in vivo for the purpose of this study, and cell-free experiments are designated in vitro. The CalPhos transfection kit (Clontech) was used to transfect either WT or mutant P protein expression constructs and either pCDNA3-HE-ΔpA or pCDNA3 (vector control) at a 1:1 ratio. At 2 days posttransfection, the FLAG-tagged P protein, bound to epsilon RNA, was purified using protein A/G beads coupled to anti-FLAG antibody as previously described (9). RNA was extracted from purified HP proteins or cytoplasmic lysates and detected by Northern blot analysis following resolution by urea-PAGE, as previously described (9). After band quantification, the amount of bound RNA was normalized to the amount of RNA in the lysate, and the WT was set to 100%. Final values are the averages from three repeated experiments.
In vitro protein priming assay.The purified WT or mutant HBV P protein, which was coexpressed with epsilon RNA and bound to agarose beads as described above, was analyzed for protein priming activity in vitro as previously described (9, 22, 35). Briefly, under RNase-free conditions, [α-32P]dGTP (10 mCi/ml, 3,000 Ci/mmol; PerkinElmer) was applied in either a magnesium- or manganese-containing buffer and incubated at 25°C for 4 h with shaking to allow protein-primed DNA synthesis, which resulted in radiolabeling of the P protein. After washing to remove unincorporated radionucleotides, immunopurified P protein aliquots were subsequently boiled in 2× SDS sample buffer for 10 min to elute the bound proteins. The radiolabeled HBV polymerase proteins were resolved by running the eluate on a 9% polyacrylamide SDS-PAGE gel and detected by phosphorimaging. After band quantification, the amount of radiolabeled P protein was normalized to the amount of immunopurified P protein as determined by Western blotting, and the WT level was set to 100%. Final values are the averages from three repeated experiments.
RNA packaging assay.HEK293T cells were cotransfected with a construct expressing either WT or mutant HBV polymerase and a construct expressing pgRNA and all viral genes except that for P (pCMVHBV-Pol−) at a 1:1 (5 μg to 5 μg per 6-cm dish) or 1:9 (1 μg to 9 μg per 6-cm dish) ratio as noted and lysed 5 days posttransfection, as previously described (9, 22). Briefly, cells were lysed with NP-40 lysis buffer (50 mm Tris-HCl [pH 8.0], 1 mM EDTA, 1% Nonidet P-40, 1× Complete inhibitor [Roche]) and cleared by centrifugation. The cytoplasmic fraction (supernatant) was treated with 15 U/μl micrococcal nuclease (Roche), after the addition of CaCl2 to 5 mM concentration, for 45 min at 37°C, and then a further 15 U/μl was added and samples were incubated for another 45 min. After clearing by centrifugation, the capsids in the nuclease-treated lysate were resolved on 1% agarose gels and the pgRNA packaged into nucleocapsids was detected with a plus-strand-specific radiolabeled probe after blotting to a nitrocellulose membrane. Although the RNA probe could, in principle, also detect the plus-strand DNA in nucleocapsids, we have shown previously that under our assay conditions, the plus-strand DNA is not detected, likely due to the fact that the plus-strand DNA is always hybridized to the minus-strand DNA in nucleocapsids and thus is unavailable for hybridization to the RNA probe (10, 54). The same membranes were subsequently reprobed with an anti-HBV core antibody to detect the capsid protein (Dako). After band quantification, the amount of packaged RNA was normalized to the amount of capsid protein, and the WT was set to 100%. Final values are the averages from three repeated experiments.
Tdp2-mediated release of DNA from HBV polymerase.For some samples, following in vitro priming, 32P-labeled DNA products attached to the HBV P protein were released with tyrosyl-DNA phosphodiesterase 2 (Tdp2; Abnova), which cleaves the tyrosyl-DNA linkage between the 5′ end of DNA and the P protein as previously described (9, 55). The released DNA was collected and resolved on an 8 M urea–20% polyacrylamide sequencing gel (9).
FOOTNOTES
- Received 6 September 2016.
- Accepted 12 November 2016.
- Accepted manuscript posted online 16 November 2016.
- Copyright © 2017 American Society for Microbiology.
