A Structural Model for Duck Hepatitis B Virus Core Protein Derived by Extensive Mutagenesis

Abstract

Duck hepatitis B virus (DHBV) shares many fundamental features with human HBV. However, the DHBV core protein (DHBc), forming the nucleocapsid shell, is much larger than that of HBV (HBc) and, in contrast to HBc, there is little direct information on its structure. Here we applied an efficient expression system for recombinant DHBc particles to the biochemical analysis of a large panel of mutant DHBc proteins. By combining these data with primary sequence alignments, secondary structure prediction, and three-dimensional modeling, we propose a model for the fold of DHBc. Its major features are a HBc-like two-domain structure with an assembly domain comprising the first about 185 amino acids and a C-terminal nucleic acid binding domain (CTD), connected by a morphogenic linker region that is longer than in HBc and extends into the CTD. The assembly domain shares with HBc a framework of four major α-helices but is decorated at its tip with an extra element that contains at least one helix and that is made up only in part by the previously predicted insertion sequence. All subelements are interconnected, such that structural changes at one site are transmitted to others, resulting in an unexpected variability of particle morphologies. Key features of the model are independently supported by the accompanying epitope mapping study. These data should be valuable for functional studies on the impact of core protein structure on virus replication, and some of the mutant proteins may be particularly suitable for higher-resolution structural investigations.

Hepatitis B viruses (HBVs), or hepadnaviruses, comprise a family of small enveloped DNA-containing viruses that replicate through reverse transcription (2). HBV, the causative agent of B-type hepatitis in humans, is the prototype of the orthohepadnaviruses which infect selected mammals, while duck HBV (DHBV) is the prototype of the avihepadnaviruses, which are endemic in some bird species (16). Overall, genome organization and replication mechanism of both genera are closely related; hence, DHBV serves as an important model hepadnavirus (43).

However, although the DHBV genome is even smaller than that of HBV (3.0 kb versus 3.2 kb), its core protein (DHBc) is substantially larger (262 versus 183 or 185 amino acids) than that of HBV (HBc). Both core proteins are the sole building blocks for the viral capsid shell. The capsids are actively involved in reverse transcription (21, 33, 55) and genome trafficking (23); are the substrate for various phosphorylation and dephosphorylation events (1, 17, 25, 32, 37, 57); and provide interaction sites, regulated by the maturation state of the packaged genome (47), for envelopment by the surface proteins (9). Evidently, the short HBc sequence fully supports these multiple functions; hence, the biological reasons behind the larger size of the avihepadnavirus core proteins are enigmatic. Knowledge of the DHBc structure would be crucial to understand this unresolved issue, and it might help to exploit the experimental advantages of DHBV (43) for tackling the structural dynamics of the hepadnaviral nucleocapsid. Presently, however, such information is scarce.

In contrast, the structure of the HBc protein and of assembled HBV capsids is known in detail from biochemical (4, 26, 27, 36) and biophysical (46) investigations of recombinant HBV capsid-like particles (CLPs). The first about 140 amino acids (aa) constitute the assembly domain (4, 53); this is followed by a 9-aa morphogenic linker (53) that affects the distribution between a larger (triangulation number T = 4) and a smaller (T = 3) class of particles. The C-terminal domain (CTD) downstream of position 149 contains clusters of R residues that bind nucleic acid. Most of the CTD is required for pregenomic RNA encapsidation and reverse transcription (25, 30, 34); similarly, the RNA content of recombinant CLPs containing at least part of the CTD is much higher than if the CTD is deleted (4, 36). The T = 4 particles consist of 120 HBc dimers, and the T = 3 particles consist of 90 HBc dimers (14, 24). The HBc assembly domain (Fig. 1A) contains five α-helices (6, 12, 54), of which α3 and α4, composed of α4a and α4b, form a hairpin, which at its tip harbors the immunodominant c/e1 B-cell epitope (3, 11, 13). Association of two such hairpins into a four-helix-bundle, protruding as a spike from the capsid surface, provides for most intradimer contacts, with the N termini wrapping around the base of the spike. The interdimer contacts are mainly provided by the “hand region” (6) consisting of α5 (residues 112 to 127) onto which downstream residues to about position 140 fold back. Although the individual interdimer contacts are weak (58), the intact particles are so stable that even complete foreign proteins can be inserted into the c/e1 epitope (28, 35, 44); this is achieved by an inherent flexibility within the subunits, as well as in their arrangement on the icosahedral lattice (5, 7). Such structural plasticity may be crucial for the active role of the capsid in reverse transcription, although only subtle differences between HBV CLPs and genome-containing nucleocapsids were detected in a recent cryo-electron microscopic (cryo-EM) study (40).

FIG. 1.

(A) Structure of the HBc dimer. The schematic representation is based on the X-ray structure of C terminally truncated HBc (54). Two monomers (dark and light) forming the dimer are shown in top and side views. Helices are indicated as cylinders; α3 and α4a/α4b form the major intradimer interface; the hand region encompassing α5 and the downstream sequence to about position 140 provides, together with residues in the N-terminal arm wrapping around the spike base (omitted in the side-view of the right-hand monomer), the interdimer contacts. In the capsid, the C termini point toward the particle interior. (B) Primary sequence alignment of DHBc with HBc. The alignment is as predicted by the PFAM database routine. Amino acid positions for DHBc are indicated above; those for HBc are indicated below the sequences. Residues in DHBc that differ in the HHBV core protein are highlighted by lowercase lettering. Boxes in the lines designated α-PHD and α-xray indicate α-helices, for HBc as in the X-ray structure and for DHBc as predicted by the PHD program. Regions predicted with higher reliability (scores of 7 or higher, on a scale from 0 to 9) are indicated by darker shading; those with lower scores (5 to 7) are indicated by lighter shading. DHBc helices are numbered by homologous positions to HBc, those within the presumed DHBc insertion sequence by D_insα1 to 3. According to data from the present study, D_insα3 is, in fact, homologous to HBc α4a, as indicated by the double designation. The zig-zag lines represent regions that are strongly predicted not to have a defined structure. For HBc the position of the immunodominant c/e1 epitope is indicated. The invariant G residue (G111 in HBc and G157 in DHBc) is shown in boldface. The two highly conserved primary sequence regions I and II are marked by boxes.

As yet, mostly indirect evidence suggests that DHBc shares structural features with HBc. First, although the overall sequence homology is low, there are two highly conserved regions (Fig. 1B), one N proximal (HBc aa 19 to 37 and DHBc aa 15 to 34), which in HBc comprises mainly helix α2a, and a second one encompassing HBc residues 96 to 143, i.e., helix α4b, the kink at G111, and the hand region; the homologous segment in DHBc encompasses aa 142 to 189. Second, the DHBc sequence from about position 200 to the C terminus, although nearly twice as long as the HBc CTD, also contains many basic residues. Deletions downstream of about position 225, caused, in DHBV-transfected cells, similar replication defects (42, 56), as did CTD truncations in HBc (25, 30, 33), and recombinant DHBc truncated after positions 229 or 237 still formed particles (55). This suggests that DHBc has a CTD that is dispensable for assembly, but its borders, as well as the existence of a morphogenic linker as in HBc, are not known. A low-resolution cryo-EM reconstruction of DHBc CLPs showed T = 4 and T = 3 particles with a HBc-like clustered-dimer architecture (24) with, however, laterally wider surface spikes, suggesting that they comprise more residues than in HBc. According to an earlier alignment (8), DHBc contains an “insertion sequence” comprising residues P86 to K130. The alignment routine used in the PFAM database also predicts such an insertion (Fig. 1B), although with somewhat shifted borders. Finally, secondary structure algorithms such as PHD (41) and PSIPRED (22) predict several α-helices in DHBc, some of which could correspond to those in HBc, most significantly in the second highly conserved region (Fig. 1B). Accordingly, DHBc residues 142 to 189 might adopt a structure similar to that of HBc helices α4b and α5 kinked at G111 (43).

A recent PepScan (18) analysis identified six antigenic regions (AR1 to AR6) in DHBc that were recognized by sera from DHBV-infected and, in part, from DHBc-immunized ducks (49); of these, AR2 (aa 64 to 84), AR3 (aa 99 to 112), and AR5 (aa 183 to 210) were proposed to be surface exposed. However, the structural state of the antigen as encountered by the ducks' immune system is unclear, and isolated peptides may or may not mimic the authentic protein structure. Directly testing surface exposure was not possible due to the polyclonality of the antisera.

In the present study we used an extensive mutagenesis approach to identify primary sequence constraints for the ability of DHBc to assemble into particles. We generated a large panel of DHBc mutants, including C-terminal and internal deletions variants plus a transposon-derived library of variants containing 5 aa insertions throughout the protein's primary sequence. Exploiting an efficient Escherichia coli expression system, we determined their assembly properties by velocity sedimentation, native agarose gel electrophoresis, and negative-staining EM. This enabled us to define the domain structure of DHBc, to demonstrate the existence of a morphogenic linker region that is much more extended than in HBc and, eventually, to combine the data into a plausible model for the DHBc fold. Although structural evidence obtained via mutagenesis may still be considered indirect, key topological features predicted by the model were independently verified by the accompanying epitope mapping study.

MATERIALS AND METHODS

Alignments, secondary structure prediction, and three-dimensional modeling.

The alignment shown in Fig. 1B is derived from the PFAM database entry “hepatitis_core” available at http//:www.sanger.ac.uk/Software/PFAM/data/jtml/seed/PF00906.shtml, manually adjusted to the primary sequence of the actually used DHBV16 (GenBank accession no. K01834) core protein. Secondary structure prediction was performed using the PHD program (41); largely consistent results were obtained using PSIPRED (22). The HBc secondary structure elements indicated in Fig. 1B are as reported previously (54). Three-dimensional modeling was performed by using the Geno3D server (http://geno3d-pbil.ibcp.fr) with HBc (pdb entry: 1QGT) as a template.

Plasmid constructs.

The expression vector for DHBc, pET28a2-DHBc, was generated by replacing the HBc gene in plasmid pET28a2-HBc (52) by the DHBc gene from the DHBV16 genome (31) in plasmid pCD-16 (38); pET28a2-HHBc was obtained similarly using a plasmid encoding the genome of HHBV4 (45). A C-terminal His₇ tag was added, via PCR, to yield plasmid pET28a2-DHBc_H7. C terminally truncated DHBc variants and the point mutant DHBc_R124E with residue R124 replaced by glutamic acid were also generated via PCR; of the truncated variants, named DHBcn, with “n” indicating the position of the last authentic DHBc amino acid, DHBc230 contained a C-terminal unrelated peptide of the sequence YKGEPLKA, and DHBc195 contained a single nonauthentic L residue. Internal deletion variants were obtained by cutting plasmid pET28a2-DHBc at the unique EcoRI site overlapping the codons for R124, I125, and H126 (AGA ATT CAT; the EcoRI site is in italics), limited Bal31 nuclease digestion, and subsequent digestion with AlwNI in the vector part. Fragments of the appropriate size range were ligated either with the 5′-terminal or the 3′-terminal AlwNI-EcoRI fragment of the unmodified pET28a2-DHBc plasmid in which the EcoRI overhang had been blunted. This yielded a collection of plasmids lacking DHBc sequence upstream, or downstream of the I125 codon; in one clone, termed (I125)₂, this codon was fortuitously duplicated. The sequences of the proteins analyzed are indicated by “Δ” followed by the positions of the deleted residues; for instance, Δ121-124 lacks DHBc aa 121 to 124.

Transposon linker scanning mutagenesis.

Transposon mutagenesis was performed with plasmid pCD16 as a template using the GPS-LS linker scanning system as recommended by the vendor (New England Biolabs). Transformants were selected on kanamycin containing agar, and plasmid DNA was isolated from the pooled colonies. Next, the DHBc gene was amplified by PCR; the inserted transposon increases its size from 0.8 kb to about 2.5 kb. The 2.5-kp products were cut with either NcoI (overlapping the DHBc start codon) plus EcoRI (overlapping codons 124 to 126), or EcoRI plus AvrII (overlapping codons 260 and 261), or NcoI plus AvrII. The corresponding restriction fragments were cloned into the appropriately cut pET28a2-DHBc_H7 vector. DNAs from randomly picked colonies showed restriction patterns indicative of random insertions. The body of the transposon sequence was removed by PmeI digestion and religation of plasmid DNA from about 2,000 pooled colonies, yielding about 1,200 colonies after religation. The positions of the insertions sites were determined by DNA sequencing. Individual constructs with in-frame insertions, and their encoded proteins, are designated by the prefix “i” followed by the position of the insertion site. C-terminal deletion variants, arising from about one-third of the integration events that introduce a premature translational stop, are correspondingly designated by the prefix “st.” Inherently, these stop codons are preceded by a transposon-encoded V residue.

Recombinant expression of DHBc proteins.

The pET28a2 plasmids were transformed into E. coli BL21 Codonplus cells (Stratagene), and protein expression and purification were performed essentially as described for HBc (52). In brief, for expression screening 2-ml cultures of the transformed bacteria were grown at 25°C in the presence of 100 μM IPTG (isopropyl-β-d-thiogalactopyranoside) for about 5 h, and then the pelleted cells were boiled in sodium dodecyl sulfate (SDS) sample buffer (100 μl per ml of culture). Aliquots (5 μl) of these SDS lysates were analyzed by SDS-polyacrylamide gel electrophoresis (PAGE) and Coomassie blue staining. For large-scale preparations (200-ml cultures, induced at 25°C for 12 to 16 h), 5 ml of cleared lysates (52) was subjected to sedimentation in 38-ml sucrose gradients (in steps of 10, 20, 30, 40, 50, and 60% sucrose (wt/vol) in TN300 buffer (50 mM Tris-HCl, 300 mM NaCl [pH 7.5]) in an SW28 rotor run for 4 h at 20°C and 28,000 rpm. For small-scale preparations (20-ml cultures), 200 μl of 1 ml of cleared lysate was sedimented through 1.4-ml 10 to 60% sucrose gradients in a TST-55 rotor (45 min at 20°C; 55,000 rpm). Gradients were harvested in 14 fractions from the top, and the distribution of the recombinant proteins was determined by SDS-PAGE analysis of 5-μl aliquots from each gradient fraction followed by Coomassie blue staining.

Native agarose gel electrophoresis.

Electrophoresis was performed as previously described (51, 52) in 1% agarose gels containing 0.5 μg of ethidium bromide/ml to visualize encapsidated RNA. Protein was subsequently detected by staining with Coomassie blue.

EM.

Aliquots (2 μl) from the relevant gradient fractions were applied to glow-discharged carbon-coated EM grids, incubated for 2 min to allow specimen adsorption, washed with water, blotted, and stained with 2% uranyl acetate for 3 min. Excess liquid was removed by blotting, and the grids were air dried. Micrographs were recorded on a Morgagni 268 instrument and run at an acceleration voltage of 100 kV at nominal magnifications of 71,000- or 140,000-fold, respectively. Approximate particle size distributions were manually determined by measuring the diameters of between 100 and 150 well-formed particles from the topmost three gradient fractions containing the bulk of the corresponding protein. For class 4 mutants, characterized by the formation of two distinct peaks in the gradients, both peak fractions were analyzed accordingly.

RESULTS

Efficient bacterial expression system for recombinant DHBV CLPs.

Recombinant DHBc was poorly expressed from earlier lambda pL promoter-based vectors (24). The use of a T7 promoter-based pET vector derivative (51, 52) led to only a slight improvement (data not shown); however, when combined with an E. coli strain providing extra copies of rare tRNAs (such as BL21 Codonplus), expression of even full-length DHBc increased to levels of 20 to 50 mg per liter of culture. Likely, rare codons in the DHBc sequence caused inefficient translation in the absence of sufficient amounts of the cognate tRNAs, as also observed with HBc. Sedimentation in sucrose gradients served as a first test for the assembly status of the recombinant DHBc. Under the conditions used, HBc CLPs sediment as a relatively sharp zone into fractions 7 to 9 in the gradient center (52). For DHBc, SDS-PAGE analysis of aliquots from the 14 gradient fractions showed a broader distribution, starting in fractions 7 to 8 and extending to nearly the bottom fractions. Negative-staining EM revealed abundant particles (Fig. 2A) of two major size classes, with approximate diameters of 37 and 32 nm, respectively (see below). The faster-sedimenting material in fraction 9 and below also consisted mainly of particles which, however, and in contrast to HBc CLPs, had a pronounced tendency to aggregate (Fig. 2A). Congruently, in native agarose gel electrophoresis the DHBc CLPs, even from fraction 8, appeared as a less distinct band than HBc CLPs (Fig. 2B), and the material from further-down fractions remained mostly in the loading slot (not shown). Particle aggregation increased further upon prolonged storage, leading to precipitation of nearly pure DHBc protein, which was used to generate the monoclonal anti-DHBc antibodies described in the accompanying study (52a).

FIG. 2.

Sedimentation behavior of, and particle formation by, C terminally truncated DHBc proteins. (A) Sedimentation profiles in preparative sucrose gradients. Cleared bacterial lysates containing the indicated DHBc proteins were sedimented through sucrose gradients. Aliquots from each of the 14 fractions were analyzed by SDS-PAGE and Coomassie blue staining. The top three fractions containing the recombinant protein were inspected by negative-staining EM (right panels). For DHBc226, the rightmost panel shows a magnified view from fraction 8 to better visualize the spectrum of particle sizes. Note the particle aggregates in fractions 9 from wt-DHBc and DHBc230. (B) Native agarose gel electrophoresis. Samples from the indicated gradient fractions were separated in a 1% agarose gel and stained for nucleic acid by using ethidium bromide (upper panel) and subsequently for protein by using Coomassie blue (lower panel). Lane M contained DNA marker fragments; the sizes in kilobases of some fragments are marked on the left. HBc refers to recombinant full-length HBc CLPs. (C) Typical morphologies of differently sized particles from wt-DHBc and truncated variants. Particles on micrographs, like those shown in panel B, were manually sorted into the indicated size classes. Two representative particles each are shown. Particle types constituting the majority (>70%) in a given population are marked by a dashed frame.

DHBc has a two-domain structure similar to that of HBc.

For a coarse mapping of the putative assembly and CTD domains in DHBc, we analyzed a set of C terminally truncated variants, ending after positions 230, 226, 218, and 195. In the alignment (Fig. 1B) V195 corresponds to HBc aa V149; HBc1-149 is efficiently expressed in E. coli and forms mostly T=4 particles (24, 53, 59). Sedimentation analysis suggested that all variants formed particles, although with distinct distributions in the gradient (Fig. 2A). DHBc230 displayed a broad distribution similar to that of wt-DHBc. Variants DHBc226 and DHBc218 appeared already in fraction 5, with the bulk of the protein sedimenting into fractions 6 to 8. DHBc195 peaked in fractions 5 to 7, with almost no material in fractions 9 and below. For all proteins, the presence of particles, though with various morphologies (see below), was confirmed by negative-staining EM (Fig. 2A). Aliquots from the topmost peak fractions, containing about 3 to 4 μg of protein per sample, were analyzed by native agarose gel electrophoresis side by side with recombinant HBc CLPs (Fig. 2B). All DHBc mutants except DHBc195 migrated as distinct bands slightly ahead of the wt-DHBc CLPs and yet slower than HBc CLPs, and all contained comparable amounts of RNA as estimated by comparing the ethidium bromide versus Coomassie blue staining intensities. DHBc195 appeared as a more diffuse band that was visible only after protein staining. These data are consistent with the presence in DHBc of an assembly domain, comprised within the first 195 aa, and a C-terminal nucleic acid binding domain. As in HBc (4, 36, 40), part of the CTD appears sufficient for packaging of E. coli RNA. A major difference from HBc, however, was the distinct shift in sedimentation velocity between the differently truncated DHBc variants; under the gradient conditions used, CLPs from HBc1-140, which forms mainly T=3 particles (36, 53, 59), still sedimented in one major band into fractions 7 and 8 (not shown).

DHBc contains an extended morphogenic linker region that encompasses part of the basic CTD.

To correlate the different sedimentation profiles of the truncated DHBc variants with potential differences in particle morphology, we next used the electron micrographs to determine the approximate diameters of between 100 and 150 individual particles per construct. The accuracy of these measurements is limited by the staining procedure, potential flattening of particles as they stick to the grid, by the different particle orientations that sometimes displayed prominent surface spikes and sometimes did not, and by the number of particles analyzed. However, wt-DHBc clearly produced two main classes of particles with diameters of about 37 and 32 nm, plus some smaller particles (ca. 10%) about 28 nm in diameter; representative examples are shown in Fig. 2C. DHBc230 produced wt-like 37- and 32-nm particles, but an increased fraction of smaller particles with mean diameters of about 26 to 28 nm and 22 to 24 nm were observed. DHBc226 and DHBc218 contained a further increased proportion of particles in the 22- to 28-nm range, plus even smaller particles with diameters of about 17 to 19 nm. DHBc195, finally, essentially lacked 37-nm particles; besides a few 32-nm particles, the majority had diameters between 22 and 28 nm, and particles <20 nm in diameter were also abundant. Despite the limitations of the negative-staining method, particles with distinctly different diameters were frequently seen side by side on one micrograph, as is particularly evident on the enlarged view shown for DHBc226 (Fig. 2A, rightmost panel). Together with the distinct sedimentation profiles, these data strongly suggest that DHBc can assemble into a larger range of differently sized particles than HBc and that a much larger sequence in DHBc (residues 195 to 226) than the morphogenic linker peptide in HBc (residues 141 to 149) affects particle morphology.

Linker scanning precisely maps the C-terminal border of the DHBc assembly domain and identifies internal regions important for folding and assembly.

Transposon mutagenesis provides a means to randomly integrate short peptide sequences into a target gene. The system used here produces eventually a 15-nucleotide (nt) insertion; 10 nt are derived from the transposon and contain the recognition sequence of the restriction enzyme PmeI preceded by an A and followed by a T (aGTTTAAACt), or vice versa (tGTTTAAACa), depending on the insert orientation; additional five nt are duplicated from the insertion site. Two out of six possible insertion events introduce a premature translational stop (GTT TAA; the stop codon is underlined); the other four create 5 aa insertions containing a limited set of different amino acids. Individual mutants were derived from three plasmid pools containing the insertions in the N-terminal half or C-terminal half of DHBc or throughout the entire sequence (see Materials and Methods for details). Constructs containing an insertion in the full-length DHBc context are designated by the prefix “i” plus the amino acid position of the insertion, constructs with a premature stop codon by the prefix “st” plus the position after which the stop codon was introduced.

Of about 150 clones sequenced more than 90% showed insertions at differing positions and in different orientation. Few sequences occurred twice, and very few occurred thrice. About 100 of the plasmids were transformed into the E. coli expression strain, and aliquots from 2-ml induction cultures were analyzed by SDS-PAGE (results not shown). A major fraction of plasmids led to the expression of proteins that comigrated with wt-DHBc or its His-tagged derivative DHBc_H7; another fraction produced distinctly smaller proteins, a finding indicative of premature translational stops.

We first exploited such truncated variants to more precisely map the C-terminal border of the DHBc assembly domain by sedimentation analysis. Five variants with translational stops after position 232 showed a wt-DHBc-like distribution in the gradient, and variants st213 and st210 produced profiles similar to those of DHBc226 and DHBc220 (data not shown). Most informative were variants st203 (not shown), st199, st187, st183, and st167. The first three all produced gradient profiles like variant DHBc195, with a distinct peak in fractions 5 to 7 (Fig. 3); the further truncated variants st183 and st167 did not form any fast-sedimenting material. Hence, the C-terminal border of the DHBc assembly domain is located between aa 183 and 187. Notably, I186 of DHBc is homologous to HBc aa L140 (Fig. 1B), which must be present to allow for HBc assembly (53, 59). Upon gel filtration on Superdex 75, most of the material from variant st183 present in the top gradient fractions eluted in the void volume, indicative of larger aggregates, but a detectable amount eluted at about the same volume as the 44-kDa marker protein ovalbumin (not shown), as expected for a dimer with a calculated mass of 43 kDa. Notably, a pronounced aggregation tendency was also observed for the assembly incompetent HBc1-139 (53).

FIG. 3.

Fine mapping the C-terminal border of the DHBc assembly domain. Cleared bacterial lysates containing the indicated proteins, derived from transposon insertions causing premature translational stops, were subjected to sedimentation as in Fig. 2A. M, protein size markers; L, SDS lysates from the same cultures that the cleared lysates were derived from. The band marked CAT represents chloramphenicol acetyltransferase (25 kDa), which is abundantly expressed from the rare tRNA plasmid present in the expression strain. Note the complete absence of material sedimenting into fractions 5 and lower for variant st183.

Next, a large number of nontruncated DHBc variants with 5 aa insertions scattered over the entire protein sequence were evaluated. Such insertions could be structurally neutral at sites that are not crucial for folding, such as external loops; in contrast, insertions into important secondary structure elements could prevent the formation of the proper tertiary structure or affect quaternary interactions. Variants i22 and i92 were poorly expressed, and four other variants, i113, i124, i128, and i168, although well expressed, were largely insoluble. Hence, at these sites the insertions appeared to cause severe folding defects. Essentially all other variants generated one of four distinct classes of sedimentation profiles. Class 1 mutants produced a profile similar to wt-DHBc, with a broad peak starting at about fraction 8 and extending to fraction 12. Class 2 mutants displayed a relatively narrow peak in fractions 8 to 10, suggesting the formation of wt-DHBc-like particles but with a reduced tendency to aggregate (which for some of the mutants was directly confirmed by EM; see below). Class 3 mutants sedimented like variant DHBc195, with one major peak around fractions 5 and 6. Finally, class 4 mutants showed two distinct peaks, one around fraction 5 and a second one around fractions 9 and 10, which sometimes extended further down, probably due to particle aggregation. Typical examples are shown in Fig. 4. EM inspection of the peak gradient fractions of several class 3 mutants revealed that, in contrast to the C terminally truncated DHBc195, the proteins formed few regular particles but mostly subparticle-sized aggregates that tended to precipitate upon removal of the sucrose by dialysis (representative examples are shown in Fig. S1 in the supplemental material). The same was true for the upper peak material of class 4 mutants; however, the faster-sedimenting material was largely present as regular particles, although with substantial variations in size (examples are shown in Fig. S2 in the supplemental material). Hence, class 3 mutations caused severe folding and/or assembly defects, and class 4 mutations caused at least partial folding and/or assembly defects. A summary correlating sedimentation profiles and insertion positions is given in Fig. 5.

FIG. 4.

Peptide insertions in the DHBc primary sequence cause four distinct sedimentation profiles. Cleared lysates from bacteria expressing the indicated insertion variants, based on wt-DHBc_H7, were sedimented through sucrose minigradients and analyzed as in Fig. 2A. All soluble variants generated one of four distinct profiles; representative examples for each class are shown. Class 1 mutants sedimented like wt-DHBc or its His-tagged variant wt-DHBc_H7 (upper left panel). Class 2 mutants produced one distinct peak around fractions 9 to 10; class 3 mutants produced mostly one peak in fraction 4 or fractions 5 to 7. Class 4 mutants generated two distinct peaks, one around fraction 4 or fractions 5 to 7 and the second around fractions 9 to 11. EM data for class 3 and class 4 mutants are shown in Fig. S1 and S2 in the supplemental material. A schematic summary of all variants tested with respect to the DHBc primary sequence is shown in Fig. 5.

FIG. 5.

Correlation between sedimentation classes of individual DHBc insertion mutants and insert location in the DHBc primary sequence. The DHBc sequence and predicted secondary structure elements are shown as in Fig. 1B. Diamonds represent insertion sites that caused poor expression (black) or insolubility (gray). Circle and square symbols with numbers indicate the respective sedimentation profile classes.

The wt-like class 1 mutants clustered at the very N terminus (i6, i8, and i12) and in the C-terminal region past position 192 (variants i192, i227, i232, i242, and i248). Class 2 mutants occurred exclusively within the first 100 aa, often at neighboring sites (e.g., i11, i34, i35, i37, i48, i72, i181, i95, and i97). Class 3 mutations, except for a single N proximal site (i28), were narrowly clustered between positions 137 and 169 (i137, i142, i144, i151, i154, i156, i160, i169), i.e., in the highly conserved region II (Fig. 1B). The same region also harbored several of the class 4 mutants (i139, i141, i148, and i158), which otherwise occurred at selected N proximal sites (i14, i53, i54, i64, and i75) and clustered in the supposed insertion sequence (i105, i110, i114, and i123). An interpretation of these data in relation to the proposed DHBc structural model is presented in the Discussion.

The presumed insertion sequence is partly unstructured but contains structured elements that are important for folding and multimerization.

Which part of the central DHBc region actually represents the insertion cannot reliably be predicted from alignments because of the low sequence homology between HBc and DHBc upstream of the highly conserved region II (Fig. 1B). The previous experiments showed that the region between positions 80 to about 100 was largely tolerant toward peptide insertions, whereas the following 30-aa segment was not (Fig. 5). To further modify its primary sequence, we used recombinant HHBV core protein (HHBc) as a natural DHBc variant with several amino acid exchanges in this region (Fig. 6A), and we generated a collection of internal DHBc deletion variants upstream or downstream of I125. HHBc formed CLPs with sedimentation properties similar to those of wt-DHBc (Fig. 6E). Two large deletions upstream I125 (variants Δ86-124 and Δ82-124) were still competent for multimer formation, however, of the mixed class 4 sedimentation profile. Smaller deletions of 4, 7, and 13 aa (variants Δ121-124, Δ118-124, and Δ112-124) all produced clear class 2 profiles (Fig. 6B); hence, deletion of this predicted unstructured part had no negative impact on particle formation. In contrast, even very small deletions on the C-terminal side of I125 (Δ126-127, Δ126-128, Δ126-130, and Δ126-136) and the duplication of a single amino acid [I125 in (I125)₂], all produced class 4 profiles (Fig. 6C), often with a large proportion of the protein in the bottom fractions, a finding indicative of heavy aggregation. Hence, this region contributes importantly to overall folding. All of the class 2 deletion variants lacked R124, the only charged residue between positions 112 and 129, whereas R124 was present in all mutants with deletions downstream of I125. To test for a potential structural influence of this residue, we finally exchanged it against a negatively charged E residue. The corresponding variant DHBc_R124E was well expressed and generated a clear-cut class 2 phenotype (Fig. 6D).

FIG. 6.

Mutational analysis of the presumed DHBc insertion sequence. (A) DHBc variants used. The DHBc sequence from positions 80 to 140, containing the presumed insertion sequence is shown; boxes indicate the positions of the predicted helices. The sequence below is that of HHBc. Identical amino acids are symbolized by a dot. Mutants lacking the indicated amino acids upstream and downstream of I25 are collectively designated Δ< and Δ>, respectively. Mutant (I125)₂ contained a duplication of I125, and R124E is a point mutant with the indicated amino acid exchange. Numbered squares indicate the sedimentation classes. (B to E) Sedimentation profiles of individual mutants with internal deletions upstream of I125 (B) and downstream of I125 (C), DHBc-R124E (D), and HHBc (E).

Direct EM evidence for increased particle homogeneity in class 2 mutants.

To directly corroborate the class 1 versus class 2 distinction, we compared, by negative-staining EM, particles from HHBc as an additional authentic avihepadnavirus core protein with class 2 mutants that resulted from either 5 aa insertions (variants i72, i95, and i97), internal deletion (Δ121-124), or the single R124E exchange. Like wt-DHBc CLPs, most HHBc CLPs (>70%) had diameters of around 37 and 32 nm; in addition, particles about 28 nm in diameter plus a possibly distinct 25-nm species, not seen for DHBc, were observed (Fig. 7A). In contrast, all of the class 2 variants analyzed, regardless of the underlying mutations, produced regular CLPs consisting mainly (>70%) of the largest (about 37 nm in diameter) and a smaller fraction of the second largest (about 32 nm) size class. Particles with diameters of less than 30 nm were essentially absent (Fig. 7B). This increased homogeneity was also reflected in native agarose gel electrophoresis, where the class 2 mutants migrated as more distinct bands than wt-DHBc (52a). Thus, the class 2 mutations do cause a distinct structural phenotype characterized by a reduced particle size heterogeneity and by a reduced tendency of the particles to aggregate.

FIG. 7.

Particle morphologies of HHBc and class 2 mutants. (A) HHBc. Negatively stained HHBc particles showed limited size heterogeneity (upper panel). The majority of the particles had, like wt-DHBc CLPs, apparent diameters of approximately 37 and 32 nm; in addition, some 28-nm and, possibly distinct, 25-nm particles were observed. (B) Class 2 mutants. Five class 2 mutants, originating from either transposon insertions, internal deletion, or point mutation, were analyzed by negative-staining EM. All appeared more regular in size and shape than HHBc and wt-DHBc CLPs, and all displayed much less size variability than the other mutant DHBc proteins, congruent with their narrow sedimentation profiles.

DISCUSSION

This study provides an extensive mutational data set on DHBc which, combined with the locally restricted sequence homologies to HBc and secondary and tertiary structure predictions, can be merged into a plausible model for the DHBc fold. The key features of this model (Fig. 8C), independently confirmed by the accompanying epitope mapping study, are the preservation of a structural framework similar to that of HBc, although with various additions originating from the extra sequences in the central and C-terminal part of DHBc.

FIG. 8.

DHBc structural features. (A) Linear comparison of the domain structures of DHBc and HBc. The two open boxes per protein symbolize the assembly domains and the CTDs; the thick black line indicates the morphogenic linkers. Gray boxes indicate the regions involved in forming the spikes (helices α3 to α4b), and the hatched box in DHBc indicates the insertion sequence. Numbers are approximate amino acid positions. (B) The highly conserved regions I and II from DHBc can substitute for the authentic HBc sequences. Shown are two views of the modeled three-dimensional structure of a hypothetical HBc protein carrying the DHBc sequences from conserved regions I (aqua) and II (red); authentic HBc sequence is shown in gray, the conserved G residue between α4b and α5 is shown in green. (C) Proposed model for the fold of DHBc. Helices are shown as cylinders, with darker and lighter shading referring to higher and lower reliabilities of the predictions, as in Fig. 1B. Conserved regions I and II are shown in aqua and red, the morphogenic linker region is indicated as a thick black line, and the rest of the CTD is shown as a dashed gray line; CTD-bound RNA may influence particle morphology. Insertion and deletion mutations and sedimentation classes are indicated as in Fig. 5. The previously proposed insertion sequence is underlaid in light yellow. However, according to the present study it rather comprises amino acids from about positions 75 to 120. The spatial arrangement in this part is not explicitly predicted, but helix D_insα2 could contribute to the lateral extensions on the DHBc spikes (24). Superimposed on the model are the locations of various antibody epitopes (see the accompanying study for details). Linear epitopes of monoclonal antibodies (MAbs) d1, d2, and d3 are symbolized by green arrowheads, and the bipartite epitope of MAb d4 is indicated by two connected arrowheads pointing to the positions where insertions impair antibody binding. Surrounding sequences may contribute to epitope formation (dashed extensions). For the conformational epitopes of the anti-native DHBc MAbs (n1, n2, and n3), sites where mutations reduce recognition but not particle formation are indicated by lightning symbols. Linear epitopes recognized by polyclonal sera (pc), as defined by the PepScan technique, are represented by irregular green bars; dashed extensions indicate weakly reactive sequences.

Domain structure of DHBc.

More than 70 residues from the C-terminal DHBc sequence could be deleted without compromising the protein's principal ability to form particles. Hence, like HBc, DHBc contains an N-proximal assembly domain, with a defined C-terminal border between residues 183 and 187 (Fig. 3). Notably, the in-between residue I186 aligns with HBc L140 (Fig. 1B), the last residue absolutely required for HBc particle formation (53, 59). At least a fraction of the nonassembling DHBc variant ending after position 183 behaved as a dimer, but not as a monomer, during gel filtration, in accord with the reported clustered-dimer architecture of DHBc capsids (24). Hence, these data strongly support the accuracy of the DHBc versus HBc alignment in the highly conserved region II.

Recombinant DHBc CLPs contained comparable amounts of RNA as HBc CLPs (4, 36, 40); the ratio of ethidium bromide versus Coomassie blue staining in native agarose gels did not significantly change by truncation to position 218 (Fig. 2B). Hence, a fraction of the basic CTD residues is sufficient for RNA packaging, as with HBc (4, 21). Whether the single basic CTD residue in DHBc195, R191, is sufficient for RNA encapsidation is not clear; although essentially no ethidium bromide staining was seen, the more diffuse protein band could also indicate instability of the DHBc195 particles. Certainly, however, and in accord with previous cell culture data (42, 56), the DHBc CTD serves as a nucleic acid binding domain.

A marked difference from HBc was the strong impact of C-terminal truncations on particle morphology. In HBc the morphogenic linker encompasses just the nine residues (positions 141 to 149) from the end of the assembly domain to the first R cluster, and its absence versus presence affects the ratio between two distinct particle forms, i.e., T=3, predominating in HBc1-140, and T=4, predominating in HBc1-149 (36, 53, 59). Other forms have not been observed. In contrast, DHBc truncations from positions 230 to 195 led to gradual shifts toward a spectrum of smaller particle sizes as determined by negative-staining EM (Fig. 2C), and this was paralleled by a corresponding upward shift in the sucrose gradients. By analogy to HBc, we hence refer to this part of the DHBc sequence as the morphogenic linker region, although it encompasses part of the CTD (Fig. 8A). Notably, the removal of various basic amino acid residues in the truncated variants could affect the way RNA is packaged and this, in turn, could influence particle morphology.

Model for the fold of DHBc.

The N-terminal about 185 aa of DHBc are functionally equivalent to the HBc assembly domain, and most of the predicted DHBc helices would have direct counterparts in HBc, providing the potential for an HBc-like framework that somehow accommodates the presumed insertion sequence. The most likely candidate segments for actually folding into a HBc-like structure are the conserved regions I and II. As a starting point, we first modeled the structure of a hypothetical chimeric HBc protein in which these two regions (HBc positions L19 to Y38 and positions K96 to L143) were exchanged for the homologous DHBc sequences (l14 to Y37 and R142 to I189). As shown in Fig. 8B, the N-proximal DHBc segment can indeed substitute for part of the arm around the base of the spike, including helix α2, and the second segment can substitute for helices α4b and α5, including the kink around the conserved G111 residue (G157 in DHBc). The addition of appropriate DHBc counterparts to HBc helices α3 and α4a would then generate an HBc-like architecture, with the extra DHBc sequence at a location equivalent to the c/e1 epitope in HBc. This is the basic concept of the model in Fig. 8C onto which the mutational data, shown in linear form in Fig. 5, are superimposed, together with a summary of the epitope mapping data derived as reported in the accompanying study (52a).

For the sake of simplicity, we consider sites where mutations caused class 1 or class 2 phenotypes as tolerant and those causing either insolubility, or a class 3 or 4 phenotype, as sensitive toward tertiary or quaternary structure perturbation. Class 1 and class 2 insertion mutations were located in three distinct segments, i.e., at the very N terminus, between positions 34 and 48, and interspersed between positions 72 and 97; small internal deletions upstream of I125 and the R124E replacement also generated class 2 profiles. In the model, nearly all of these insertions are placed in loops or at the very ends of predicted helices. The only exception is the predicted helix Dα1; hence, either this segment is not helical or perturbing its structure is innocuous. Both interpretations are compatible with an HBc-like arrangement wherein the very N-terminal residues can be deleted or replaced without affecting particle formation (39, 50). Furthermore, the first 15 aa of HBc were identified as one of two regions (“domain I”) where various mutations did not negatively affect expression in E. coli; about one-half of the mutants remained assembly competent (27). In the second such domain (residues F24 to P50) various mutations between amino acids L31 and E46 allowed particle formation, again in accord with our data for the DHBc segment from aa 34 to 48. Lastly, few mutations between HBc positions 15 and 30 allowed stable expression, and all prevented assembly; congruently, DHBc was sensitive to insertions at positions 14, 22, and 28. This strikingly similar pattern strongly supports that the first about 50 aa of DHBc adopt a structure similar to that in HBc.

The next DHBc segment from about positions 50 to 76 should contain a long helix homologous to HBc α3; indeed, after a class 2 insertion site at position 48, class 4 mutations occurred at positions 53, 54, 65, and 76, with a single class 2 mutation at position 72. This indicates that the predicted helix Dα3 does exist but, possibly, is not contiguous to the very end. Alternatively, the part around position 72 may not be structurally crucial. Apart from the small helix D_insα1, the following segment down to position 103 is predicted not to have a defined structure, and indeed it harbored three class 2 mutations. Thereafter followed, densely clustered, highly sensitive sites until the end of the assembly domain. Such a crucial structural role is fully compatible with DHBc residues 137 and 185 being structurally homologous to HBc helices α4b and α5. Completion of this framework structure requires one more helix as a counterpart to HBc α4a; this missing link must be provided by either D_insα2 or D_insα3 in the presumed insertion sequence.

DHBc insertion sequence.

The start of the second highly conserved region around DHBc position 135, and the experimentally supported existence of helix Dα3, confine the presumed insertion sequence to somewhere between positions 77 and 135. However, an exact assignment solely based on primary sequence is not possible. In fact, the alignment of DHBc aa 80 to 90 to HBc α4a, as in Fig. 1B, appears highly unlikely because the entire DHBc segment between aa 77 and 100 bears hallmarks of surface-exposed loops, such as a high frequency of P, and of polar (T) and charged (E) residues. In contrast, the DHBc sequence from 125 to 134 produces very high scores (8 or 9 for all seven central positions) in the PHD α-helix prediction, and it is nearly contiguous with the DHBc equivalent, Dα4b, to HBc helix α4b. We therefore strongly favor that helix D_insα3 is structurally equivalent to HBc α4a and consequently should be named Dα4a, as in Fig. 8C. A key structural role for Dα4a is supported by the strong negative impact of even small sequence modifications downstream, but not upstream, of I125. We therefore propose that the segment between the end of helix Dα3 around position 77, and the beginning of Dα4a around position 122, constitutes the actual DHBc insertion sequence.

Within this segment two stretches, positions 81 to 97 and positions 112 to 124, tolerated insertions or deletions, in accord with the predicted lack of defined structure. No definite statements are possible for the putative short helix D_insα1. In contrast, the existence of helix D_insα2 between positions 105 and 114 is strongly supported by three class 4 plus an insolubility-causing insertion. This is in line with a recent cell culture study of a fortuitously isolated core protein variant lacking the codon for H107 (20). This protein was unstable and did not form detectable nucleocapsids, a prerequisite for virus replication. Replication was, however, rescued by reintroduction of helix-compatible amino acids but not proline.

Implications for DHBV nucleocapsid assembly and replication.

A similar structural framework in DHBc and HBc implies similar assembly properties. For wild-type DHBc and HHBc, this was largely confirmed by the predominant formation of two major capsid size classes as with HBc. The largest, ∼37-nm-diameter DHBc particles definitely conform to T=4 symmetry (B. Böttcher and M. Nassal, unpublished data), and the ∼32-nm-diameter particles are compatible with T=3 symmetry (see below). However, we also found two striking differences, namely, a pronounced tendency of wild-type and various mutant DHBc particles to aggregate and an apparently much wider spectrum of particle sizes. Wild-type DHBc already contained a significant fraction (ca. 10%) of smaller (∼28-nm) particles, and wild-type HHBc seemed to produce an additional, possibly distinct, class of 25-nm particles. This size variability was further extended by C-terminal truncations and also occurred in the fast-sedimenting material from class 4 mutants (see Fig. S2 in the supplemental material). Both size variability and aggregation were strongly reduced in class 2 mutants.

HBc with its ability to form T=3 and T=4 particles, also in vivo (15, 24, 40), represents one of few examples of T number polymorphism (29). Hence, the wide size range of the DHBc particles is unusual. Despite the limitations of the negative-staining approach (see above), we consider it highly unlikely that the apparently different particle sizes are a mere artifact of the procedure. First, differently sized particles were frequently seen side by side on one micrograph (Fig. 2C). Second, smaller-appearing particles were predominantly seen for mutants which also displayed slower sedimentation profiles. Finally, very little size variation was observed for the class 2 mutants using the same technique. Quasi-equivalence theory (10) predicts that icosahedral capsids should conform to one of the allowed T numbers, defined by the equation T=(h² + hk + k²), with h and k being integers. In the simplest, T=1 form, 60 subunits with identical conformations make up the capsid. For the next allowed numbers, T=3 and T=4, the constituent capsid protein subunits must adopt three and four, respectively, similar but nonidentical conformations. Because the surface (F) of a sphere is defined as F = 4π × r², where r is the radius, and assuming each subunit contributes an equal proportion of surface area, the ratio of the diameters of a T=4 to a T=3 particle is √4/3 to 1, or 1.15:1, and that of a T=4 to a T=1 particle is √4/1 to 1, or 2:1. Accordingly, the T = 3 form of a 37-nm T=4 DHBc capsid should be around 32 nm in diameter, as observed, whereas T=1 particles are expected to be in the 18.5-nm diameter range, but not between 25 and 28 nm as observed. Nominally, this size could correspond to the nonallowed triangulation number T=2 (calculated diameter of 26 nm). Recently, the in vitro formation of T=2 particles has been shown for Brome mosaic virus capsids (29, 48). Another example is the VP3 subcore of bluetongue virus (19). Such nonquasiequivalent assemblies require substantial distortions within the subunits. Moreover, the nearly continuous size range of some of the mutant DHBc particles (Fig. 2C) suggests that not all can conform to strict icosahedral symmetry. Hence, DHBc can apparently use a whole variety of different interdimer contacts, indicating an inherent structural flexibility exceeding that of HBc (5, 7). In addition, the degree of flexibility in wt-DHBc appears to be intermediate between that of class 4 mutants and that of class 2 mutants.

In HBc, most interdimer contacts are provided by the hand region, plus by small N-proximal segments encompassing aa 14 to 17 and aa 29 to 36 (54). For various DHBc intra-assembly domain mutations, their impact on the interdimer contacts is easily imagined by their location in the model. However, mutations at nearly all other sites, including the remote D_insα2, also caused changes in particle morphology. Hence, structural alterations at any one site are conveyed through the body of the protein, as in a mechanical system of interconnected rods.

This invokes two tempting speculations. First, the flexibility of the wild-type DHBc structure may be an adaptation to the multiple core protein functions in viral replication. The panel of DHBc mutants causing distinct assembly properties will now allow us to systematically address this aspect in cell culture and even in vivo. Second, structural alterations at the inner capsid face could cause corresponding alterations on the capsid surface, such as those proposed to trigger selective envelopment of mature hepadnaviral nucleocapsids (47). Magnified by to the ordered extra elements such as helix D_insα2, they may be more easily detectable than in HBV capsids (40). Lastly, although the flexibility and variable particle morphologies disfavor high-resolution structural analyses of wt-DHBc, this problem would be alleviated by using class 2 mutants.