Mosaic hepatitis B virus core particles presenting the complete preS sequence of the viral envelope on their surface

The sequence of the preS domain of the hepatitis B virus (HBV, genotype D) envelope was inserted into the major immunodominant region (MIR) of the C-terminally truncated HBV core (HBc) protein. In Escherichia coli, the HBcpreS fusion protein was partially soluble and did not produce particles. Co-expression of the wild-type HBc as a helper protein along with the fusion protein led to the formation of mosaic HBc particles that exhibited HBc, preS1 and preS2 antigenicity. Two alternative combinations of medium- and high-copy plasmids were used for co-expression of fusion and helper proteins, in an attempt to improve mosaic particle production. However, the preS fusion content of the particles remained the same in both expression combinations. In a third co-expression in which the modified HBc helper lacked aa 7685 in the MIR, the incorporation level of HBcpreS fusion into the particles was noticeably lower. Purified chimeric particles were immunogenic in mice.

Recombinant virus-like particles (VLPs) are potent carriers of foreign epitopes for prospective use as vaccines and diagnostic and gene therapy tools. Among the large number of existing candidate VLPs, hepatitis B virus (HBV) core protein (HBc) VLPs have been well-characterized and have been used as carriers for over 100 different foreign sequences [for reviews, see Ulrich et al. (1998); Pumpens & Grens (2001, 2002)].

Natural HBc particles exist in two forms of icosahedral shells with triangulation numbers T=3 or 4, which contain 90 or 120 dimers, respectively (Crowther et al., 1994). Amino acids 7882 present the major immunodominant region (MIR) of the HBc molecule on surface-oriented protruding spikes that are formed by HBc dimers (Böttcher et al., 1997; Wynne et al., 1999).

In line with the latest X-ray structure data (Wynne et al., 1999), better antigenicity and immunogenicity of foreign insertions are observed at the MIR and N-terminus of the HBc molecule. The capacity of the MIR appears to be especially high: incorporation of 120 aa of the hantavirus nucleoprotein (Koletzki et al., 1999) and even the 238 aa green fluorescent protein (Kratz et al., 1999) does not interfere with particle-forming ability. At the same time, rather short insertions into the MIR can disturb self-assembly of the HBc molecules significantly, demonstrating the importance of factors such as hydrophobicity, volume and β-strand index of the sequence to be inserted (Karpenko et al., 2000; G. Borisova, V. Ose & P. Pumpens, unpublished data).

HBV surface proteins remain the first insertion candidates for construction of novel, multi-targeted HBV vaccines. All HBV surface proteins large (L), middle (M) and short (S), encoded by a single ORF share the common, 226 aa S domain. In addition, the M protein contains the preS2 sequence as a 55 aa, N-terminal extension of S and the L protein contains an additional 119 or 108 aa (for HBV genotypes A and D, respectively) as a preS1 sequence, in addition to the preS2 sequence (Heermann & Gerlich, 1991). The preS sequence (preS1+preS2) plays an important role in HBV infection and induction of immunological responses (Kann & Gerlich, 1998). The immunodominant epitope of the HBV preS1 region, which is recognized by the mAb MA18/7 (Heermann et al., 1984) and mapped at aa 31DPAF34 (Sominskaya et al., 1992a), is regarded as the central part of a possible site of attachment of HBV to hepatocytes (Pontisso et al., 1989a, b).

Insertions of DPAF (Borisova et al., 1999) and of longer (up to 27 aa) preS fragments (Schödel et al., 1992; Makeeva et al., 1995; Borisova et al., 1996, 1997) into the MIR do not disturb the particle-forming ability of the HBc derivatives. However, our attempts to construct HBc VLPs carrying preS or separate preS1 or preS2 sequences were not successful for either the MIR or for C-terminal insertions.

Mosaic technology for HBc VLPs is elaborated thoroughly for C-terminal insertions that are separated from the gene for the truncated HBcΔ protein (aa 1144) by a UGA stop codon. Expression of these constructs under conditions of UGA suppression allows concomitant synthesis of the wild-type (wt) HBc protein as an assembly helper and prolonged HBc fusion as a readthrough chimeric protein (Koletzki et al., 1997). In this way, fragments of up to 213 aa from the hantavirus nucleoprotein have been incorporated into mosaic particles (Kazaks et al., 2002). Alternatively, mosaic HBc particles carrying Staphylococcus aureus nuclease have been obtained with the wt HBc and HBc fusion proteins synthesized from separate plasmids (Beterams et al., 2000). An M13mp10- and plasmid pUC-derived two-vector system was used for the construction of mosaic HBc particles with an 8 aa epitope insertion into the HBc MIR (Loktev et al., 1996). Two plasmid-mediated co-expression of the wt HBc along with its naturally occurring MIR deletion variants also results in formation of mosaic HBc particles (Preikschat et al., 2000), which are composed of homo- and heterodimers (Kazaks et al., 2003).

Here, we inserted the complete, 163 aa HBV genotype D preS sequence into the MIR of the HBcΔ molecule. Due to its length and the presence of hydrophobic moieties, this sequence was considered as a model to investigate the potential of mosaic VLPs for incorporation of problematic sequences in their active, functional forms. Also, the preS sequence contains a set of well-characterized epitopes, allowing the detection of surface exposure of both preS1 and preS2 domains on the chimeric particles. The expression constructs are described in Fig. 1a. C-terminally truncated HBcΔ was used as both carrier and helper, as it showed a remarkably high expression level in Escherichia coli, compared to the full-length HBc. Target genes were placed under the control of the tryptophan operon promoter (P_trp). The preS-encoding sequence was PCR-amplified from the pHB320 plasmid, which encodes the entire HBV genome (genotype D, subtype ayw) (Pumpens et al., 1981), with the following primers: 5'-TTCACGTGATGGGGCAGAATCTTTCCACCAGC-3' and 5'-TTTACGTAGTTCAGCGCAGGGTCCCCAATC-3'. The appropriate PCR fragment, after restriction with Eco72I and Eco105I endonucleases, was cloned into the HBcΔ gene in the high-copy, Ap^R vector p2-19 (Borisova et al., 1999) and treated with the same enzymes, resulting in the p2-19preS plasmid [Fig. 1a (i), (ii)]. For lower expression, the Klenow-filled PvuII/NdeI fragment, which contained the expression cassette (along with the P_trp) from the p2-19preS plasmid, was cloned into the SalI-restricted and Klenow-filled, medium-copy, Km^R vector pREP4 (Qiagen), resulting in the pREP2-19preS plasmid [Fig. 1a (iii), (iv)]. Similarly, expression of the wt HBcΔ as a helper was performed either from the Km^R, medium-copy plasmid pREPpT [Preikschat et al., 2000; Fig. 1a (ii)] or from the Ap^R, high-copy plasmid pT31 [Borisova et al., 1988; Fig. 1a (iii)]. As an alternative helper, we used an artificial, self assembly-competent HBcΔ variant with a deleted MIR: the appropriate pT31 derivative, plasmid pHBcΔ7685, was constructed by PCR mutagenesis with the introduction of the amino acids D, E and L to generate an endonuclease Ecl136II cleavage site [Fig. 1a (iv)].

(31K): Fig. 1. Expression of the recombinant HBcpreS gene. (a) Schematic presentation of constructions for single expression (i) and co-expression (iiiv) of the recombinant HBcpreS gene: (ii) HBcpreS from high-copy and HBcΔ helper from medium-copy vector; (iii) HBcpreS from medium-copy and HBcΔ helper from high-copy vector; (iv) HBcpreS from medium-copy and deletion variant HBcΔ7685 helper from high-copy vector. The appropriate vectors are shown on the right. Amino acid positions of the HBc protein are given under the gene boxes. (b) Western blotting of soluble expression products from variants (iiv) shown in part (a). Anti-HBc mAb 14E11, which recognizes HBc aa residues 136144 (Pushko et al., 1994), was used for protein detection.

The E. coli strain K802 was used for expression. To ensure co-expression of the helper HBcΔ protein along with the HBcpreS fusion protein, appropriate vector combinations [Fig. 1a (iiiv)] were used for co-transformation and double Ap^RKm^R transformants were selected. To induce the P_trp, cells were grown in tryptophan-free minimal M9 medium, supplemented with 1 % Casamino acids (Difco) and 0·2 % glucose, on a shaker at 37 °C for 1820 h. Cell lysis and protein detection by SDS-PAGE were performed according to standard protocols and essentially as described previously (Borisova et al., 1999).

Purification of VLPs by Sepharose CL-4B chromatography was according to the protocol used for the previous HBcpreS1 chimeras (Borisova et al., 1999), with some modifications. Before low-speed centrifugation, lysates were adjusted to 0·45 M urea and after precipitation with ammonium sulfate, proteins were resuspended in 40 mM phosphate buffer (PB) that contained 1·5 M urea, 50 µg PMSF ml¹ and 0·1 % Triton X-100. Urea and Triton X-100 were included to increase the solubility of particles. After chromatography and rechromatography on Sepharose CL-4B, the appropriate VLP fractions were pooled, protein was precipitated with ammonium sulfate and dissolved in PB that contained 1·5 M urea without Triton X-100, dialysed against PB without urea and stored at 20 °C in 50 % glycerol with the addition of NaCl to 150 mM.

To characterize VLPs morphologically, samples were adsorbed onto carbonFormvar-coated grids, stained with 2 % phosphotungstic acid and subjected to electron microscopy. As expected, only a very low amount of the HBcpreS fusion protein was found to be soluble when expressed alone [Fig. 1b (i)]. No VLPs, but only aggregates were detected in the supernatant of lysed cells [Fig. 2a (i)]. In the presence of either HBcΔ or HBcΔ7685 as a helper, solubility of HBcpreS increased up to tenfold, correlating with the amount of wt HBc in the soluble fractions [Fig. 1b (iiiv)].

(83K): Fig. 2. Structural properties of mosaic HBcpreS VLPs. (a) Electron microscopy of soluble cell fraction: (i) expression of HBcpreS without a helper; (iiiv) purified VLPs from co-expression variants (iiiv), respectively (Fig. 1a). (b) Silver-staining of peak fractions of purified VLPs, showing the proportion of co-expressed proteins in variants (iiiv). (c) Western blotting of the same fractions with anti-preS1 mAb MA18/7. (d) Immunogold labelling electron micrograph of the VLPs from co-expression variant (ii) treated with mAb MA18/7 according to Louro & Lesemann (1984). Both of the other variants gave similar pictures. Bars, 50 nm.

For all three co-expression variants, electron microscopy revealed HBc-like particles in the supernatants (not shown), as well as after purification [Fig. 2a (iiiv)]. SDS treatment of particles from CL-4B column fractions revealed the presence of both co-expressed proteins, demonstrating the mosaic structure of particles (Fig. 2b, c). Evidence of the presence of preS epitopes on the surface of VLPs came from colloidal gold immunoelectron microscopy using the anti-preS1 mAb MA18/7 (Fig. 2d).

Figs 1b and 2b show that the relative amount of HBcpreS protein in purified particles is much lower than in cell lysates, indicating a limitation in the incorporation of HBcpreS into stable VLPs. Although the level of incorporated HBcpreS was estimated as roughly 46 % for vector combinations (ii) and (iii) [Fig. 2b (ii) and (iii)], helper from high-copy vector ensured an approximately eightfold better yield of purified mosaic VLPs. The HBc deletion variant was less effective as a helper than wt HBc; only about 12 % of fusion protein was incorporated during co-expression of HBcpreS with HBcΔ7685 [Fig. 2b (iv)].

To test the surface accessibility of different parts of the preS sequence, mosaic HBcpreS VLPs were subjected to competition with the preS1 or preS2 peptides (containing preS aa 2047 and 120145, respectively) for mAbs MA18/7 and S26 (Sominskaya et al., 1992b), which recognize preS aa 3135 and 132135, respectively. Competitive ELISA was performed as described previously (Borisova et al., 1999). MaxiSorp immunoplates (Nunc) were coated with 100 µl peptides (10 µg ml¹) and serial dilutions of VLPs in 0·5 % BSA were added. Dilutions of mAbs that were used were 1 : 3000 for MA18/7 and 1 : 200 000 for S26, as calculated from direct ELISA data (not shown). Competition was estimated as percentage signal decrease compared to the negative control, to which 0·5 % BSA was added instead of VLP samples. In competition with the preS1 peptide for MA18/7, all three types of mosaic particle showed similar behaviour: inhibition reached 7586 % (Fig. 3a). Despite the lower content of HBcpreS in the mosaics obtained in co-expression with the HBcΔ7685 helper, these particles competed for MA18/7 slightly better than particles with the wt HBcΔ helper. In competition with the preS2 peptide, inhibition reached 5363 % for mosaics with the wt helper, whereas the HBcΔ7685-based mosaic particles reacted very poorly with mAb S26 (Fig. 3b). Therefore the preS1 domain or at least, its epitope 31DPAF34 is better surface-exposed and available to antibodies than the preS2 epitope 132QDPR135 on VLPs from co-expression variants (ii) and (iii), whereas deletion within the MIR of the helper leads to conformational changes that hide the preS2 epitope.

(19K): Fig. 3. Antigenicity and immunogenicity of mosaic HBcpreS VLPs. Competition of mosaic particles with: (a) the preS1 peptide for mAb MA18/7; and (b) the preS2 peptide for mAb S26. Co-expression variants (iiiv) (Fig. 1a) are represented: , variant (ii); , variant (iii); , variant (iii). (c) Immunogenicity of mosaic VLPs in mice. Antibody titres in murine sera after immunization with mosaic VLPs from co-expression variants (iiiv) are shown as mean values of five sera: black bars, with adjuvant; grey bars, without adjuvant. Titres are expressed as decimal logarithms from the reciprocal of the highest serum dilution that is required to yield an optical density value three times that of non-immunized mice. HBcΔ, control sera from mice immunized with HBcΔ particles.

To study the immunogenic properties of the purified mosaic VLPs, BALB/c female mice (five per group) were immunized intraperitoneally/subcutaneously (25/25 µg) on day 0 with 50 µg VLPs in complete Freund's adjuvant (Sigma) and boosted in the same way on days 10 and 24 with particles in incomplete Freund's adjuvant. For all VLPs, the immune response was also tested in the absence of adjuvant. Sera were collected on day 32 after immunization and their anti-HBc, anti-preS1 and anti-preS2 antibody responses were tested by direct ELISA. For anti-HBc titres, 96-well PolySorp plates (Nunc) were coated with 100 µl full-length purified HBc protein at 10 µg ml¹. For anti-preS1 and anti-preS2 responses, plates were coated with the appropriate preS1 and preS2 peptides (see above). Plates were incubated overnight and blocked as described previously (Borisova et al., 1999), then serially diluted murine sera were added and the reactions were processed with secondary antibodies and developed as described previously (Borisova et al., 1999). For controls, sera from mice that had been immunized with HBcΔ only and from non-immunized mice were used. The results of immunization with mosaic VLPs are summarized in Fig. 3c. A strong anti-HBc response was observed for mosaics that were purified from all three co-expression variants, which is comparable to the immunogenicity of the wt HBcΔ itself. Surprisingly, the MIR deletion of the helper did not significantly reduce the strong anti-HBc response of mosaic particles. In standard procedure using the adjuvant, the anti-preS1 response of mosaics was slightly higher than the anti-preS2 response in all cases, in line with the antigenicity data from competitive ELISA. However, without the adjuvant, anti-preS1 and anti-preS2 responses were practically equal. The same could be said about the adjuvant effect on anti-HBc activity in all constructs, as well as anti-preS1 and anti-preS2 activity in variants (ii) and (iii). Concerning variant (iv), where the helper with a deletion was used, the anti-preS response to VLPs was comparable with that to VLPs from variants (ii) and (iii), despite the lower content of preS sequence in this type of particle. However, this remains true only for protocols without adjuvant, demonstrating the possibility that the effect of the adjuvant could be contrary to that expected in a particular case. In general, anti-preS antibody titres in sera were medium-range and reached 3·2x10³ for both anti-preS1 and anti-preS2, reflecting the relatively low level of incorporation of HBcpreS fusion protein into the mosaic particles. However, even in the envelope of native HBV particles, the full-length preS sequence is presented in only one molecule out of five as the large surface protein L (Heermann et al., 1984).

In conclusion, incorporation of the full-length preS sequence, carrying hydrophobic stretches, demonstrates the high potential of the mosaic-particle approach for exposure of long or problematic sequences on HBc VLPs. This also opens up a way of engineering VLPs that harbour a set of different epitopes for multivalent vaccines and/or gene therapy tools.