Assembly and budding of a hepatitis B virus is mediated by a novel type of intracellular vesicles


  • Mouna Mhamdi,

    1. Heinrich-Pette-Institut für experimentelle Virologie und Immunologie an der Universität Hamburg, Hamburg, Germany
    Search for more papers by this author
  • Anneke Funk,

    1. Heinrich-Pette-Institut für experimentelle Virologie und Immunologie an der Universität Hamburg, Hamburg, Germany
    Search for more papers by this author
  • Heinz Hohenberg,

    1. Heinrich-Pette-Institut für experimentelle Virologie und Immunologie an der Universität Hamburg, Hamburg, Germany
    Search for more papers by this author
  • Hans Will,

    1. Heinrich-Pette-Institut für experimentelle Virologie und Immunologie an der Universität Hamburg, Hamburg, Germany
    Search for more papers by this author
  • Hüseyin Sirma

    Corresponding author
    1. Heinrich-Pette-Institut für experimentelle Virologie und Immunologie an der Universität Hamburg, Hamburg, Germany
    • Department of General Virology, Heinrich-Pette-Institut für experimentelle Virologie und Immunologie, PO Box 201652, 20206 Hamburg, Germany
    Search for more papers by this author
    • fax: +49 (40) 48051-222

  • Potential conflict of interest: Nothing to report.


Formation of enveloped viruses involves assembly and budding at cellular membranes. In this study, we elucidated the morphogenesis of hepadnaviruses on the ultrastructural and biochemical level using duck hepatitis B virus (DHBV) as a model system. Formation of virus progeny initiates at the endoplasmic reticulum (ER) and is conserved both in vitro and in vivo. The morphogenesis proceeds via membrane-surrounded vesicles containing both virions and subviral particles, indicating a common morphogenetic pathway. The virus particle–containing vesicles (VCVs) are generated and maintained by reorganization of endomembranes accompanied by a striking disorganization of the rough ER (rER). VCVs are novel organelles with unique identity and properties of ER, intermediate compartment, endosomes, and multivesicular bodies. VCVs are dynamic structures whose size and shape are regulated by both membrane fusion and fission. Conclusion: Our data indicate a strong reorganization of endomembranes during DHBV infection, resulting in the biogenesis of novel organelles serving as multifunctional platforms for assembly and budding of virus progeny. (HEPATOLOGY 2007.)

Hepatitis B viruses (HBVs) are small, enveloped DNA viruses consisting of a membrane-surrounded nucleocapsid harboring the partially double-stranded viral genome and the covalently attached polymerase.1 The viral envelope has an unusually high protein content for a biological membrane and contains only a small amount of cellular lipids.2, 3

DHBV-infected hepatocytes produce not only complete virions, but also large quantities of subviral particles (SVPs). SVPs are devoid of the nucleocapsid and viral DNA,4, 5 but contain the envelope proteins. The extracellular DHB virion has a spherical structure of approximately 45 to 65 nm and contains an electron-dense nucleocapsid of about 27 nm.6 Little is known about the morphogenesis of this virus family, but several of the following features indicate that this process is likely to be unique and complex: the DHB viral envelope proteins are multispanning transmembrane proteins, encoded by a single open reading frame, whose differential transcription results in the large (L) and the small (S) surface protein.7 Both proteins share an identical carboxyterminal region, representing the S protein, with an additional N-terminal extension forming the preS-region of L. The ratio between the L and S proteins in the viral particles is about 1:4.8 As the major structural component of the envelope, the S protein determines envelope curvature and is indispensable for both budding and secretion of viral particles.9 Both surface proteins are synthesized at the rough endoplasmic reticulum (ER), or rER, and then bud into the ER lumen, forming SVPs. This budding event is autonomous and independent of all other viral components (e.g., nucleocapsid).10, 11 The intrinsic membrane-deforming activity of hepadnaviral envelope proteins is host cell–independent and conserved from yeast to human.12 Virion formation occurs by envelopment of nucleocapsids and is dependent on the tight interaction of preformed cytoplasmic nucleocapsids with preassembled surface proteins at ER membranes.9, 13 This envelopment is achieved by specific interactions between the nucleocapsid and the cytoplasmically-exposed preS-region of the L protein.14, 15 Cellular compartments exploited during viral morphogenesis remain as largely undefined as the host factors involved. Recently, some chaperones and γ2-adaptin were identified as host factors participating in HBV progeny formation.16–18

Viruses exploit a broad arsenal of strategies to assemble and bud within cells. For example, Pox-like, Irido-like, and Asfar-like viruses induce the formation of “aggresomes,” which serve as scaffolds to concentrate structural components.19, 20 The formation of replication complexes and assembly sites of different RNA viruses is frequently associated with the rearrangement of cellular membranes, usually from the secretory pathway.21, 22 Morphogenesis of HIV and some other viruses in macrophages occurs at the internal vesicles of multivesicular bodies (MVBs) or late endosomes.23 The presence of the endosomal sorting complex required for transport (ESCRT) machinery on endosomal membranes may trigger virus assembly and budding at these sites.23, 24

Despite recent advances in our knowledge about the morphogenesis of HBVs, several principal questions remain to be answered. Do virions and SVPs follow distinct assembly and budding pathways? Which cellular compartments and factors are exploited? To explore these and other emerging questions, we used DHBV and primary duck hepatocytes (PDHs), and duck liver biopsies as a convenient model system. The data obtained indicate that DHBV infection induces a strong reorganization of the endomembrane system, resulting in the formation of new membrane-surrounded organelles. These organelles are pivotal for the coordinated progress of the distinct and complex morphogenetic steps of hepadnaviruses.


DHBV, duck hepatitis B virus; EDTA, ethylene diamine tetraacetic acid; EEA1, early endosomal antigen 1; ER, endoplasmic reticulum; ESCRT, endosomal sorting complex required for transport; GFP, green fluorescent protein; HB, homogenization buffer; HEPES, 4-2-hydroxyethyl-1-piperazineethanesulfonic acid; IC, intermediate compartment; IP, immunoprecipitation; MTP, microsomal triglyceride transfer protein; MVB, multivesicular body; PDH, primary duck hepatocyte; PDI, protein disulfide isomerase; PNS, postnuclear supernatant; rER, rough ER; SVP, subviral particle; TEM, transmission electron microscopy; TSG, tumor susceptibility gene; VCV, virus particle–containing vesicle; YFP, yellow fluorescent protein.

Materials and Methods

Cell Culture

We prepared and cultivated congenitally DHBV-infected PDHs as described.25 The PDHs were seeded into 6-well or 12-well plates at a density of about 1 × 106 or 5 × 105 liver cells per well, respectively. We performed the experiments 3 to 7 days postplating if not otherwise indicated.

Antibodies and Plasmids

For immunoblot and immunofluorescence analyses, we used primary antibodies against the following antigens: DHBV-L KpnI26 and 1H.1,27 DHBV core (kindly provided by L. Cova, Lyon, France), calnexin and membrin (from StressGene), Rab5B (Santa Cruz Biotechnology), protein disulfide isomerase (PDI), and microsomal triglyceride transfer protein (MTP; both were a kind gift from M. Hermann, Vienna, Austria), and γ2-adaptin (kindly provided by R. Prange, Mainz, Germany). HRPO (horseradish peroxidase)-coupled and fluorochrome-coupled secondary antibodies were from Dianova and Molecular Probes, respectively. The following expression vectors were commercially acquired or kindly provided: pEYFP-Golgi (BD Biosciences; yellow fluorescent protein [YFP]), green fluorescent protein (GFP)-Rab7, GFP-Rab11, tumor susceptibility gene 101 (TSG101)-GFP, and red fluorescent protein (RFP)-CD63 (from W. Mothes, New Haven, CT).

Subcellular Fractionation and Iodixanol Gradient Ultracentrifugation

We adapted the protocol from Refs.28 and29. We cut congenitally DHBV-infected fetal livers into small pieces and resuspended them in 3 mL homogenization buffer (HB) (0.25 M sucrose, 1 mM ethylene diamine tetraacetic acid [EDTA], 60 mM 4-2-hydroxyethyl-1-piperazineethanesulfonic acid [HEPES] pH 7.4, and protease inhibitor cocktail; Roche). We centrifuged the suspension for 4 minutes at 100g and 4°C, then washed the pellet twice with HB. Afterwards, we resuspended the pellet in 0.5 mL HB and homogenized it in a glass homogenizer for 15 strokes. We centrifuged the homogenate for 10 minutes at 2,500g and 4°C. We transferred the postnuclear supernatant (PNS) into a new tube. The pellet was resuspended in 500 μL HB, recentrifuged and the obtained supernatant is the second PNS and was pooled with the first PNS. For a 0%-26% linear iodixanol gradient, we initially diluted a stock solution of 60% (wt/vol) Iodixanol (optiprep, Axis-shield) to 50% by adding 5 volumes Optiprep to 1 volume of the diluent solution (0.25 M sucrose, 6 mM EDTA, 60 mM HEPES, pH 7.4). We mixed equal volumes of this working solution and HB to make a 25% Iodixanol solution. We prepared a 12-mL continuous density gradient using a 2-chamber gradient maker, by mixing 6 mL 25% Iodixanol and 6 mL HB. We loaded the PNS on top of the gradient and centrifuged it for 1 hour and 55 minutes at 288,000 g and 4°C in a Beckman SW Ti 41 rotor. We collected 17 fractions of 700 μL each from the bottom, by tube puncture. We measured their refractive index and calculated the densities according to the formula: ϕ = 3333η − 3442.

SDS-PAGE and Immunoblot Analysis

We performed SDS-PAGE and immunoblot analysis as described.25 We quantified the immunoblot signals using a chemiluminescence imager (BioRad).

Immunoprecipitation of VCVs

Congenitally DHBV-infected PDHs cultures from a 6-well plate were scraped in phosphate buffered saline and pelleted. The pellet was dissolved in homogenization buffer (0.25 M sucrose, 1 mM EDTA, 10 mM Hepes, pH 7.4, and protease inhibitor cocktail) and cells were pottered 15 times. One-half of the homogenate was subjected to immunoprecipitation. For immunoprecipitation from subcellular fractions, 100 μL of fractions 8, 9, 10, and 11 were pooled, diluted to 1.5 mL with phosphate buffered saline, and incubated with the antibody-decorated beads overnight at 4°C under rotation. For immunoblot analysis, 5% of input, 5% of postimmunoprecipitation (post-IP) supernatant, and 20% of the IP pellet were loaded.

PCR Analysis

We diluted the subcellular fractions 1:20 in PCR lysis buffer, processed them, and analyzed them after PCR as described.25

Transient Transfection, Immunofluorescence, and Confocal Microscopic Analysis

We grew congenitally DHBV-infected PDHs on coverslips in 24-well tissue culture plates. At 1 day postplating, we transfected the cells with 3 μg DNA/well of the indicated plasmids using JetPEI (Polyplus transfection) according to the manufacturer's instructions. The next day, we changed the medium and estimated the transfection efficiency by epifluorescence microscopy. We cultivated the cells for a further 24 to 48 hours prior to fixation and immunostaining. The cells were fixed using either 3.7% paraformaldehyde or ice-cold methanol-acetone (1:1) for 10 minutes at room temperature. We then permeabilized the paraformaldehyde-fixed cells with 0.1% Triton-X-100 for 10 minutes prior to immunofluorescence analysis.25 We analyzed the staining with a confocal laser scanning microscope (LSM 510 META; Zeiss) and processed the images using Zeiss LSM Image Browser.

Electron Microscopy

We processed the noninfected or congenitally DHBV-infected PDHs and the liver biopsies from DHBV-infected ducks as described.25, 30 We performed Immunogold staining on paraformaldehyde-fixed sections of PDHs using core-specific and L-specific antisera as described.30 We acquired and processed the images using DigitalMicrograph (Gatan).


Ultrastructural Characteristics of Viral Assembly, Envelopment, and Budding Sites, Both In Vitro and In Vivo

To analyze the subcellular sites of viral morphogenesis, we analyzed ultrathin sections of primary liver cell cultures from congenitally DHBV-infected ducks by transmission electron microscopy (TEM). Figure 1A shows a section of the cytoplasm from a noninfected hepatocyte. The extensive tubuloreticular ER network was the most prominent organelle in the cytoplasm of noninfected hepatocytes (Fig. 1B, arrows). The strong decoration of these endomembranes with ribosomes identifies them as rER (Fig. 1B, inset). In contrast, the cytoplasm of DHBV-infected hepatocytes lacked the impressive rER network and was instead full of vesicular structures (Fig. 1C and D, arrows). The lumen of these vesicles contained spheroid-shaped particles, which corresponded in size and morphology to viral particles.5, 6 Most of the cytoplasmic vesicles in infected cells seemed to be without connection to other membrane-surrounded compartments but a few were constricted at one end. These constrictions may represent connections to the endomembrane system from which they were presumably derived. These vesicles, which we will refer to as VCVs, were absent in noninfected hepatocytes (Fig. 1A and B).

Figure 1.

Ultrastructural analysis of noninfected and congenitally DHBV-infected duck hepatocytes in vitro and in vivo by ultrathin section TEM. (A) The cytoplasm of noninfected hepatocytes shows the typical distribution of ER membranes throughout the cytoplasm. (B) Rough ER (rER) structures (arrows) in the cytoplasm of noninfected hepatocytes. The inlet shows a magnification to visualize ribosomal structures on the cytosolic surface of the rER membrane. (C) The cytoplasm of infected hepatocytes lacks the typical rER structures and is full of vesicular structures. (D) The numerous vesicular structures (arrows) in infected hepatocytes contain viral particles. (E) The cytoplasm of infected hepatocytes from liver samples contains similar vesicular structures as shown above in vitro. (F) The vesicles contain viral particles (arrows). The black bars indicate the size.

To obtain evidence that the morphological features of DHBV morphogenesis in vivo are similar to those seen in vitro, we analyzed congenitally DHBV-infected duck livers by electron microscopy. The cytoplasm of infected hepatocytes was filled with vesicles of different size and shape (Fig. 1E). Higher magnification revealed that they contain viral particles with SVPs being the dominant species (Fig. 1F, arrows). Overall, these data show that the principal morphological features of DHBV morphogenesis are conserved both in primary cultures of liver and in vivo. Furthermore, DHBV morphogenesis was restricted to hepatocytes, since nonparenchymal cells present in the primary cultures and liver were devoid of any VCVs and viral particles (data not shown).

To study the subcellular distribution of viral antigens in more detail, we performed immunogold staining of paraformaldehyde-fixed ultrathin sections of DHBV-infected PDHs using an L-protein–specific and core protein–specific antiserum (Fig. 2). L-staining was associated with intracytoplasmic vesicles and showed strong labeling of viral particles in the lumen of VCVs (Fig. 2A and B). The pattern of anti-core labeling was different from that of L and was often seen in the cytosol or close to the limiting membrane of VCVs and less frequently inside the vesicles (Fig. 2C and D).

Figure 2.

Transmission electron microscopy of immunogold-labeled congenitally DHBV-infected duck hepatocytes. Sections of hepatocytes were stained for viral L (A,B) and core protein (C,D). Micrograph in (A) shows that L staining was virtually restricted to VCVs which were distributed throughout the cytoplasm (arrows). (B) Immunogold staining was mainly found on the viral particles residing in the lumen of the VCVs. (C,D) Immunogold particles were predominantly scattered in the cytosol (arrow heads) and at the limiting membrane of VCVs (black arrows), and rarely found inside the vesicles (white arrows).

At the outer nuclear membrane of infected hepatocytes, we frequently observed membrane dilatations of different extent, which contained one or few viral particles (Fig. 3A and inset). The viral cargo was identical to that observed within cytoplasmic vesicles. It is conceivable that after reaching a certain critical size, the membranous sacs extend into the cytoplasm and finally segregate from the outer nuclear membrane. An intermediate state of this process is ultrastructurally documented in Figure 3A. As the outer nuclear membrane is part of the ER, the observed dilatations may imply that the formation of viral particles actually initiates at this compartment. Consistent with this proposal is the observation of viral particles in the lumen of tubulovesicular extensions of rER (Fig. 3B). However, these types of vesicles represented only a minor subpopulation of VCVs. The great majority was devoid of ribosomes (Fig. 3C).

Figure 3.

Ultrastructural features of virus particle-containing vesicles and different stages of virus assembly. (A) The outer nuclear membrane (ONM) of infected hepatocytes shows dilatations containing viral particles in the perinuclear space. INM, inner nuclear membrane. An early stage of this dilatation is shown in the inlet. (B) Viral particles were also found in the lumen of rER (white arrow) decorated with ribosomes (arrowheads) and smooth ER (sER, black arrow) structures indicating that assembly and budding process can occur at these compartments. (C) VCVs (arrows) are variable in size and shape. The asterisk indicates an electron-light structure presumably corresponding to a lipid droplet or apolipoprotein accumulation. (D-F) Fission or fusion events (arrows) with membrane indentations between VCVs can be observed. Bars indicate the size.

VCVs were heterogenous in size, ranging in diameter from 100 to 800 nm (Figs. 1D and F, 3B–F). The large variability in vesicle size could be the result of the pinch-off of variable membrane sacks from the outer nuclear membrane or rER. However, we found no ultrastructural evidence for this. Another explanation for the bigger and smaller vesicles could be that they were formed through homotypic and/or heterotypic fusion and fission. Ultrastructurally, we could indeed observe advanced stages of either fusion or fission events between intermediate vesicles (Fig. 3D–F). These findings are suggestive for a scenario in which DHBV-associated vesicles can associate with other cellular vesicles.

Closer examination of the vesicular content revealed 2 types of particle entities (Fig. 4A and F). The vast majority of viral particles consists of membrane-surrounded “empty” spheroids with a mean diameter of about 50 nm representing SVPs. A few contained, in addition, “filled” particles with an electron-dense nucleocapsid corresponding to virions with a mean diameter of about 60 nm (Fig. 4A and F, black arrows). Pleomorphic filamentous or tubular particles were not observed. The overall ultrastructure of the intracellular viral particles is similar, if not identical, to extracellular progeny as reported.31 Furthermore, concurrence of both particle types in the same compartment is strongly suggestive for a common morphogenetic pathway. Free viral particles in the cytoplasm outside these vesicles were not observed.

Figure 4.

Ultrastructural features of the early steps of DHBV budding. (A) A VCV showing slight membrane indentations (white arrow) as an early step of viral budding. The arrowhead indicates a SVP, the arrow a virion. (B) Advanced membrane indentation (arrow) indicating a progressed step of viral morphogenesis/budding. (C,D) Strong membrane indentations (white arrow) in VCVs already containing viral particles (black arrows). (E) Nucleocapsids with different electron densities (arrowheads) can be observed in close proximity to VCVs. (F) Interaction of nucleocapsids (arrowheads) with VCVs on a vesicle already containing virions (black arrows). Bars indicate the size.

Ultrastructural Features of the Early Steps of DHB Viral Budding

Ultrastructural analysis revealed that not all VCVs were spherical in shape. Some vesicles showed slight or strong inward indentations, respectively (Fig. 4A,B, white arrows). These different invaginations presumably represent early and intermediate stages of SVP budding. The later steps of budding such as pinch-off and release of viral particles into the lumen of VCVs were rarely observed (Fig. 4C and D, white arrows).

Nucleocapsids destined to be enveloped and secreted have an intrinsic membrane affinity.32 Figure 4E shows particles of different densities measuring 22 to 27 nm in diameter in the cytosol (Fig. 4E, arrowheads). Based on their morphology and size, we identified these particles as nucleocapsids. In addition, these nucleocapsids were observed both in close proximity to and interacting with the membrane of VCVs (Fig. 4E and F, arrowheads). The nucleocapsids observed outside the vesicular membrane were morphologically similar to those observed within enveloped virions (Fig. 4F, arrows). The interaction of nucleocapsids with vesicular membranes was exclusively observed at membrane indentations and probably represents the early step of virion budding (Fig. 4F, arrowheads). Unfortunately, the later steps of virion budding into the vesicles were not observed under our experimental conditions.

Hepatocellular Distribution of DHB Viral Structural Proteins L and Core

To characterize the origin and nature of VCVs, we performed extensive immunofluorescence analysis of congenitally DHBV-infected PDHs using antibodies against marker proteins for different cytoplasmic compartments. In the following experiments, the intracellular distribution of L was used as an indicator for virions and SVPs since both viral particle entities were cocompartmentalizing. The staining of L revealed 3 easily distinguishable staining patterns: a reticular one that most likely corresponds to nonparticulate surface proteins, a vesicular one with vesicles of different sizes, and a plasma membrane-associated one as shown by the visible contour of the cell (Fig. 5A). Core protein was detected as a fine punctuate cytoplasmic staining (Fig. 5B) where it partially colocalized with the L protein (Fig. 5C). This colocalization was confirmed by cytometric analysis of the immunofluorescence data (Supplementary Fig. 1{FIGS1}). A nuclear dot-like staining for core protein was also observed (Fig. 5B) and presumably corresponds to nuclear core bodies previously reported.33 To precisely define the nature of VCVs, DHBV-infected cells were fixed and stained for L and cellular marker proteins. Anti-calnexin and anti-PDI antibodies were used as markers for the ER. As judged from the immunofluorescence staining, the distribution of L overlapped almost completely with the ER staining (Fig. 5D–F, and data not shown, respectively). Membrin, an ER-to-Golgi SNARE that mediates the transport between both compartments, is mainly located in the intermediate compartment (IC) and to a minor extent in the cis-Golgi.34 This marker protein was found in variously sized and shaped cytoplasmic foci indicative of peripheral, vesicular, and tubular clusters of the IC (Fig. 5H). The membrin staining overlapped only to a minor extent with that of L in infected cells (Fig. 5G–I). These data imply that VCVs partially overlap with the IC and that this compartment is reorganized during DHBV infection in comparison to noninfected cells (data not shown). The ectopically-expressed Golgi marker protein YFP-β-galactosyltransferase, which showed a typical juxtanuclear Golgi distribution (Fig. 5K), did not colocalize with L-positive structures (Fig. 5J–L). Immunofluorescence analysis of Rab5B, a marker protein for early endosomes, revealed a partial colocalization of Rab5B with L-positive vesicles, although part of the Rab5B-positive structures were devoid of any L (Fig. 5M–O). However, the staining pattern for early endosomal adaptor protein EEA1 did not match that of L (Fig. 5R). Further cytometric analysis of the immunofluorescence data confirmed this absence of colocalization (data not shown). This indicates that L is not enriched in early endosomes, but that a subpopulation of Rab5B is present on VCV membranes.

Figure 5.

Intracellular distribution of the viral envelope protein L in correlation to protein markers of organelles in infected PDHs. Cells were indirectly coimmunostained for L (parts A, D, G, J, M, and P) and the following marker proteins: viral core protein (part B), calnexin (part E) as ER marker, membrin as a marker for the intermediate compartment (part H), ectopically expressed β-galactosyltransferase (β-GT) as a marker for the Golgi (part K), and finally Rab5B and EEA1 as early endosome markers (parts N and Q, respectively). Merged signals of L and marker proteins together with the counterstained nuclei are shown in panels C, F, I, L, O, and R. Bars correspond to 5 μm.

These colocalization studies were extended by overexpression of fluorescently-tagged compartment-specific marker proteins in DHBV-infected PDHs. The results are summarized in Table 1. To test whether DHBV exploits endosomal compartment for its budding, we performed colocalization analysis of VCVs with Rab7, a marker for late endosomes,35 and Rab11, a marker for both recycling endosomes and the trans-Golgi network.36 L did not colocalize with any of these marker proteins (Table 1). This indicates that VCVs are distinct from late and recycling endosomes. In addition, Rab7 and Rab11 are not recruited to L-positive cellular compartments. To test whether DHBV assembly involves MVBs, we overexpressed CD63/lamp-3, a tetraspannin found in late endosomes and MVBs,37, 38 and TSG101, a member of the vacuolar protein sorting machinery that is known to play an essential role in formation and sorting of cargo into MVBs/late endosomes in a wide range of eukaryotic cells.39, 40 We observed a partial colocalization of L-positive vesicles with CD63, but not with TSG101-positive compartments (Table 1). This indicates that VCVs are distinct from MVBs and that DHBV probably recruits proteins like CD63 from the multivesicular machinery to its assembly and budding sites.

Table 1. Cellular Compartments Analyzed and Corresponding Markers Used
Cellular CompartmentMarkerColocalization with L
  1. Abbreviations:–, no colocalization;–/+, partial colocalization; +, colocalization; IC, ER-golgi intermediate compartment; RFP, red fluorescent protein; VPS, vacuolar protein sorting.

Early endosomesRab 5−/+
Late endosomesGFP-Rab7
Recycling endosomesGFP-Rab11
Multivesicular bodiesRFP-CD63−/+
VPS machineryTsg101-GFP

In conclusion, we showed that the majority of L protein in infected cells is located in cytoplasmic vesicular structures positive for the ER marker proteins calnexin and PDI. Only a small fraction of L protein colocalized with the IC and the early endosome marker Rab5B. In addition, VCVs containing L are clearly distinct from late endosomes and Golgi. Besides, we have shown that some cellular proteins like Rab5B and CD63 are present on VCV membranes.

VCVs Cofractionate with Microsomes

To confirm and extend the results of the colocalization studies, we used an independent biochemical approach and performed subcellular fractionation assays. These membrane flotation experiments involved 0%-26% iodixanol-based, linear density gradients and dounce-homogenates of infected PDHs. To determine the subcellular distribution of DHBV structural proteins core and L, as well as viral DNA, fractions were first analyzed by immunoblot and PCR, respectively. L protein was detected in fractions 6 to 14, with a major peak in fractions 9 to 11, while core protein was found in the same fractions, with a peak in fractions 7 and 8 (Fig. 6). PCR analysis of the same fractions showed enrichment of the viral DNA in fractions 6 to 11 (Fig. 6). Considering the coincidence of the viral structural proteins with viral DNA in the same fractions, we concluded that VCVs harboring viral cargo were mainly present in fractions 6 to 12. In addition, immunoblot analysis of the same fractions for ER marker proteins calnexin and PDI showed that the ER was mostly enriched in fractions 6 to 11, although a small amount of PDI was also observed in the last two fractions, 16 and 17. Thus, the ER fractions strongly overlapped with fractions containing viral markers, which confirms the colocalization studies above.

Figure 6.

Analysis of subcellular fractions from DHBV-infected PDHs for viral and cellular markers. Homogenates of congenitally DHBV-infected PDHs were subfractionated using a 0%-26% iodixanol-based linear density gradient and 17 fractions were recovered from bottom to top. Aliquots of each fraction were separated by 5%-20% gradient SDS-PAGE and analyzed for viral L, core protein, and organelle marker proteins calnexin and PDI (ER), membrin (IC), γ2-adaptin (trans-Golgi network), and Rab5B (early endosomes). Viral rcDNA in the same fractions was analyzed by PCR.

Immunoblot analysis of the same fractions for the IC and Golgi marker proteins membrin and γ2-adaptin41 showed that fractions 6 and 10 to 14 contained the IC as indicated by enrichment of membrin, whereas γ2-adaptin and thus the Golgi was present in fractions 6, 16, and 17. The presence of the Golgi marker in fractions 16 and 17 explained the result that PDI was also present in these last fractions. PDI is known to form a complex with the MTP mainly in the ER. But in addition, the PDI/MTP complex mediates transfer of membrane triglycerides to nascent apolipoproteins in the ER and then shuttles them to the Golgi, where the assembly of the apolipoprotein particle is completed.42 Thus, PDI in fractions 16 and 17 corresponds to the Golgi-associated protein. Immunoblotting of the subcellular fractions for the early endosomal marker protein Rab5B showed that endosomes were mainly present in the lighter fractions of the gradient (fractions 14-17) (Fig. 6). But a small fraction of Rab5B was also detectable in fractions 6, 8, and 10-13, where it overlapped with that of L. This subpopulation of Rab5B presumably corresponds to the fraction that colocalized with VCVs as observed in the immunofluorescence analysis (Fig. 5M–O). Fraction 6 was positive for all tested cellular and viral proteins and most likely contains aggregated and thus inseparable material.

Taken together, VCVs were highly and partially enriched in fractions containing ER (microsomes) and the IC, respectively. Furthermore, they were excluded from fractions containing Golgi membranes.

VCVs Are Novel Organelles with Mixed Properties of ER, IC, and Early Endosomes

Enrichment of VCVs in microsomal and IC fractions is indicative for an association of VCVs with and their derivation from these cellular compartments. To show a direct association of VCVs with the above mentioned endomembranes, native VCVs were isolated from the cytoplasm of infected hepatocytes. This assay was based on the assumption that VCVs contain nonparticulated viral surface proteins inserted into their membranes. If this is true, it should be possible to immunocapture VCVs from dounce homogenates of infected hepatocytes using an L-specific antiserum. To exclude a significant alteration of VCV integrity that may occur during both the homogenization and immunoprecipitation procedure, material bound to the protein A-beads was first analyzed by TEM. As shown in Figure 7A, we were able to isolate native and intact vesicles of different size from the cell homogenates (arrows). Such vesicles were not immunoprecipitated with a nonspecific antiserum (data not shown). As expected, VCVs harbored viral particles (Fig. 7B and C, arrows). These data also imply that VCV membranes indeed contain nonparticulate, cytosolically-accessible envelope proteins allowing their immunoisolation. Next, we analyzed these IPs by immunoblot for the presence of cellular marker proteins such as calnexin, membrin, Rab5B, and others. The IP was enriched for calnexin, MTP, membrin, and Rab5B (Fig. 7D–G, lane 3) compared to the immunoglobulin G (IgG) control (Fig. 7D–G, lane 2). To test whether the association of VCVs with Rab5B was specific and significant, we analyzed whether EEA1 was associated with VCVs. As predicted from the immunofluorescence experiments, EEA1 was not present in the IP pellets and was exclusively detected in the post-IP supernatant (data not shown). This underscores the specificity and selectivity of Rab5B recruitment to VCV membranes.

Figure 7.

Immunocapture of virus particle–containing vesicles. (A) Ultrastructural analysis of immunoprecipitated VCVs. L-containing cellular vesicles and membranes were immunocaptured from a pool of fractions 8, 9, 10, and 11 after subcellular fractionation shown above. Samples were fixed by glutaraldehyde and subjected to transmission electron microscopic analysis. (A) Pansorbin bead (white asterisk) decorated with vesicular structures (black arrows). (B,C) Higher magnification of VCVs harboring SVPs (black arrows). (D-G) VCVs have mixed properties of ER, IC, and early endosomes. VCVs were immunocaptured from cell homogenates of infected PDHs using an L-specific antiserum (α-L) or nonrelated antiserum (α-IgG). The subsequent immunoblot analysis identified calnexin, MTP, membrin, and Rab5B (D-G, lane 3) as with viral L protein coimmunoprecipitating cellular proteins in comparison to the controls (D-G, lane 2).

In summary, we isolated VCVs from cell homogenates of infected hepatocytes by IP using an L-specific antiserum. These IPs were enriched for calnexin, MTP, membrin, and Rab5B, but not for EEA1. The heterogenous mixture of proteins, which are known to be marker proteins for different subcellular compartments, indicates that VCVs are generated during virus replication through the reorganization of ER membranes and recruitment of specific cellular proteins.


The morphogenesis of DHBV and related viruses is largely unknown. Using a combination of biochemical, cell biological, and ultrastructural approaches, we obtained evidence that the formation of viral progeny has unique aspects that distinguish DHBV from all other known animal viruses. Noteworthy, and probably unique for hepadnaviral morphogenesis, is the peculiar mixture of strategies and elements, which are not only characteristic for DNA viruses, but also for RNA viruses.20, 43

The ultrastructural analysis of DHBV-infected hepatocytes in vitro and in vivo shows a strong reorganization of endomembranes, resulting in membrane-surrounded structures occupying large regions of the cytoplasm. This finding is consistent with seminal observations reported both for DHBV and HBV in previous ultrastructural studies. Consistent with our study, DHBV SVPs and virions were found in hypertrophied cisternae of the ER.5 In agreement with and similar to our study, naked core particles were observed both free in the cytosol and close or at cisternal ER membranes. However, unlike in our study, the identity of the particles was not confirmed by immunoelectron microscopy. Taken together, the previous studies and our current studies strongly suggest that SVPs and virions assemble at and bud into ER-derived membranes. Visualization of the different stages of the budding process as demonstrated for the first time in our study strongly corroborates this interpretation. The same may apply for HBV, because the different virus particle types were found within the cisternae of ER.44–46 The excessive formation of a novel cellular compartment during the course of a DHBV infection is strong evidence for a virus-induced process. In line with this notion is that the ectopic expression of the HBV envelope protein L alone induces extensive reorganization of the hepatocellular endomembranes in transgenic mice, retention and accumulation of subviral particles, and cytotoxic demise of hepatocytes, leading to formation of hepatocellular carcinoma.47 Thus, reorganization of cellular endomembranes is mediated by L protein and presumably involves both membrane remodeling or neogenesis. The upregulation of cellular genes governing lipid biosynthesis that are expected to be altered by viral secretion from cellular membranes in the liver of HBV transgenic mice supports the membrane neogenesis hypothesis.48

Both electron microscopy and confocal analysis of L-stained infected hepatocytes showed that the hepadnaviral morphogenetic “centers” are vesicular structures heterogeneous in size and morphology. The appearance of VCVs is characterized by the impressive disorganization of the rER network and formation of virus-filled vesicles and small tubules during virus replication. The membrane of some vesicles was decorated with ribosomes, indicating that they are derived from the rER. The decoration of VCV membranes with ribosomes should allow the synthesis of both cellular and viral proteins. However, the majority of vesicles showed a smooth cytosolic membrane surface, which either hints at a different origin of these vesicles or at the loss of ribosomes during viral morphogenesis.

The large heterogeneity in the size of vesicles was impressive and could be the result of the pinch-off of variable membrane sacks from the outer nuclear membrane or rER. However, this seems unlikely, because we found no ultrastructural evidence for this event. Alternatively, one could assume that the vesicles grow in size over time. However, this growth would require an increased de novo membrane synthesis. Because such a membrane growth is likely to be limited, it can only account, in the best case, for part of the vesicular growth. Alternatively, but not exclusively, the bigger vesicles could be formed through homotypic or heterotypic fusion of smaller ones. Ultrastructurally, we could indeed observe advanced stages of membrane fusion between intermediate vesicles. Furthermore, and in favor of this hypothesis, dual immunofluorescence analyses and subcellular fractionation showed that DHBV-associated vesicles contained the endosomal marker Rab5B. This is consistent with a heterotypic fusion between DHBV-associated and endocytic vesicles.

The ultrastructural analysis of L-containing vesicles showed that they were filled with viral particles of rather homogenous appearance. These were exclusively found in the vesicular lumen, but not free in the cytosol. This result is in agreement with the biochemical analysis of subcellular fractions from infected PDHs, which showed that viral structural proteins were only present in fractions containing membrane-surrounded cellular compartments. The coincidence of both viral particle entities shows that virions and SVPs exploit a common morphogenetic pathway. Virions were always numerically much less than SVPs irrespective of the size of VCVs. Because the viral envelope proteins are cotranslationally inserted into ER membranes,6, 9 we can anticipate that the membranes of VCVs also contain not yet particulate envelope proteins. Consistent with this notion, native VCVs could be immunocaptured using L-specific antiserum. Unlike SVPs, the formation of virions requires interaction of preformed mature nucleocapsids with the surface proteins because the large envelope protein L is essential for both virion and SVP formation.10, 14 This result shows for the first time that the excessive formation of SVPs as a unique feature of hepatitis B viruses is already determined at the assembly and budding stage at least for DHBV. This may apply to all other hepadnaviruses as well.

Because virions and SVPs are formed through the same morphogenetic pathway, the question arises why the intracellular formation of SVPs exceeds the formation of virions. A plausible answer would be that the formation kinetics of SVPs is significantly faster than that of virions. The formation of virions is a multistep process and requires maturation of capsids through reverse transcription of the viral genome. Thus, the availability of mature nucleocapsids is an important and possibly rate-limiting step of virion formation. In contrast, the formation of SVPs requires many fewer steps and is primarily dependent on the amount and ratio of viral envelope proteins S and L as well as the availability of cellular membranes. This relatively low abundance of mature nucleocapsids, combined with the autonomous budding activity of the surface proteins, provides reasonable explanation for the excessive formation of SVPs.

Nucleocapsids in different maturation states, as indicated by their different density, could be observed free in the cytosol in contrast to complete virions. In addition, the electron microscopy pictures showed that only the presumably mature, electron-dense nucleocapsids tethered to the membrane of VCVs. This is consistent with previously published flotation experiments, which showed that only mature nucleocapsids have the intrinsic affinity and ability to interact with intracellular membranes.32 We provide ultrastructural evidence that electron-dense nucleocapsids dock to the membrane of VCVs, leading to its deformation in the sense of an extrusion into the vesicle lumen (Fig. 4F). Unfortunately, the last phase of the budding, the pinch-off and release of newly-enveloped virions, could not be pictured.

Biochemical analysis of subcellular fractions from infected hepatocytes showed that the structural virus components L protein and core protein are mainly excluded from fractions containing Golgi membranes. This finding is further supported by the morphological observation that the intracellular distribution of viral L protein and a cotransfected YFP reporter protein for the Golgi compartment show no overlap. Moreover, our data indicate that assembly and budding of DHBV rather take place in pre-Golgi compartments, namely the ER and IC. This is in agreement with a study which proposed that HBV surface antigens assemble and bud in a post-ER and pre-Golgi compartment.49 Association with the rER suggests that VCVs are morphogenetic centers but also sites for translation of structural proteins prior to their assembly and budding. In addition, a number of host-cell proteins (calnexin, MTP, PDI, Rab5B, and CD63) have been identified as components of VCVs. They are probably recruited to VCVs in which the vesicle membrane could provide a structural framework for viral assembly and budding. However, TSG101, the entry component of the ESCRT-I for endosomal membranes, was largely absent from L-positive VCVs. This indicates that budding requirements for DHBV appear to be distinct from the ones of HIV, which depends on the ESCRT-1 complex for budding into endosomes.23, 24, 50 Overall, our data implicate that assembly and budding of DHBV involves the formation of novel organelles, which have mixed properties of ER, endosomes, IC, and MVBs. These findings support and extend the data of a recent work which showed that human HBV budding involves an endosomal compartment.18 Why do some viruses bud into endosome-like vesicles? The endosomal system clearly plays a significant role in the assembly of many retroviruses,24 preformed HIV particles from late endosomes are infectious.23 Viruses might also be able to hide in these late endosomes, sequestered away and protected from the entry route and degradative environment, respectively.

As depicted in the model (Fig. 8), VCVs seem to be morphogenetic centers for DHBV. It is tempting to assume that VCVs are multifunctional platforms at which the different steps of viral morphogenesis are efficiently executed and coordinated: (1) protein biosynthesis and cotranslational insertion of the surface proteins into the membrane; (2) recruitment of cellular adaptors and formation of special membrane microdomains; (3) assembly; (4) membrane deformation; and finally (5) viral budding. Additionally, VCVs could have a maturation and storage function. A further consequence of the compartmentalization of viral morphogenesis in VCVs might be their exocytic release as the mode of secretion. Molecular elucidation of signals, factors, and mechanisms that are involved in the generation of DHBV-induced vesicles and secretion require further studies to identify which viral and cellular factors play a key role.

Figure 8.

Assembly and budding model for DHBV. Envelope proteins assemble and bud following their synthesis on the rER membranes. These morphogenetic steps result in membrane dilatations and finally release of VCVs into the cytoplasm. In the absence of mature nucleocapsids, the ongoing envelope protein synthesis and their assembly on VCV membranes trigger budding of SVPs. The far less efficient virion formation initiates with the tethering of mature nucleocapsids to VCV membranes and results in envelopment and budding into the vesicular lumen. VCVs are dynamic structures and their size as well as shape is regulated by fusion and fission events.


We are indebted to L. Cova, INSERM, Lyon, France for providing the DHBV core antibody, to M. Hermann, Vienna, Austria for the anti-PDI and anti-MTP antibodies, and to R. Prange, Mainz, Germany for the γ2-adaptin antibody. We acknowledge the excellent technical support from N. Lohrengel for immunoblot analysis and B. Holstermann for the electron microscopy.