Structural Basis for Host Membrane Remodeling Induced by Protein 2B of Hepatitis A Virus

ABSTRACT

The complexity of viral RNA synthesis and the numerous participating factors require a mechanism to topologically coordinate and concentrate these multiple viral and cellular components, ensuring a concerted function. Similarly to all other positive-strand RNA viruses, picornaviruses induce rearrangements of host intracellular membranes to create structures that act as functional scaffolds for genome replication. The membrane-targeting proteins 2B and 2C, their precursor 2BC, and protein 3A appear to be primarily involved in membrane remodeling. Little is known about the structure of these proteins and the mechanisms by which they induce massive membrane remodeling. Here we report the crystal structure of the soluble region of hepatitis A virus (HAV) protein 2B, consisting of two domains: a C-terminal helical bundle preceded by an N-terminally curved five-stranded antiparallel β-sheet that displays striking structural similarity to the β-barrel domain of enteroviral 2A proteins. Moreover, the helicoidal arrangement of the protein molecules in the crystal provides a model for 2B-induced host membrane remodeling during HAV infection.

IMPORTANCE No structural information is currently available for the 2B protein of any picornavirus despite it being involved in a critical process in viral factory formation: the rearrangement of host intracellular membranes. Here we present the structure of the soluble domain of the 2B protein of hepatitis A virus (HAV). Its arrangement, both in crystals and in solution under physiological conditions, can help to understand its function and sheds some light on the membrane rearrangement process, a putative target of future antiviral drugs. Moreover, this first structure of a picornaviral 2B protein also unveils a closer evolutionary relationship between the hepatovirus and enterovirus genera within the Picornaviridae family.

INTRODUCTION

Due to their limited genome size, viruses are obligate intracellular parasites that recruit cell host components for their multiplication. All currently known positive-strand RNA viruses, as well as some DNA viruses, rearrange host intracellular membranes to create the so-called viral factories, vesicular structures that concentrate the viral and host proteins required for replication as well as the viral genomes and, at the same time, shield them from host antiviral defenses (1).

Despite the key role of viral factories in viral replication, the molecular processes leading to their formation are only partially understood. This is especially true for picornaviruses (PVs), one of the largest families of pathogens known, for which the lack of information on host membrane rearrangement processes contrasts with the highly detailed data available for other stages of the picornavirus viral cycle (e.g., receptor binding, viral entry, and RNA synthesis).

The Picornaviridae family includes many important human and animal pathogens such as polioviruses, rhinoviruses, foot-and-mouth disease virus, and hepatitis A virus (HAV). HAV is the only species and only serotype in the Hepatovirus genus, with three genotypes circulating in humans. The disease, characterized by acute liver inflammation, is endemic in developing countries with poor sanitation, where infections often occur in children. However, outbreaks also occur in developed countries, in both adults and children (2–4).

Similarly to all other picornaviruses, HAV has a nonenveloped icosahedral capsid with a ∼30-nm diameter that protects the positive-sense single-stranded RNA genome. The HAV genome (∼7.5 kb) contains a single open reading frame, which is translated as a large polyprotein (∼250 kDa) that undergoes self-cleavage, originating the mature viral proteins and their (sometimes rather stable) intermediates. The picornaviral polyprotein is organized into three regions, named P1, P2, and P3. The P1 region codes for the capsid proteins (VP1 to VP4), while the P2 and P3 regions contain the enzymes, precursors, and accessory proteins essential for viral replication and polyprotein processing (2A to 2C and 3A to 3D) (5). However, there are important differences in polyprotein processing among the different members of this family: for Enterovirus, the largest and most extensively studied picornavirus genus, primary cleavage takes place at the P1/P2 junction (VP1/2A) and is mediated by 2A, which functions as a cis-acting protease (6, 7). In contrast, primary polyprotein processing for Hepatovirus takes place between 2A and 2B and is carried out by the single viral proteinase 3C^pro, which displays a different substrate specificity from that of the enteroviral 3C proteinases, marked by radically different amino acid preferences at the P4 and P2′ positions of the substrate (8–11). HAV 3C^pro cleaves the 2A/2B junction between residues Gln836 and Ala837 (numbering according to the GenBank entry for HM175 [M14707.1; 12]), a region located 144 residues upstream of the enteroviral 2A/2B junction, resulting in a smaller 2A protein (71 to 73 amino acids long) and a considerably larger 2B protein (251 amino acids long) (11, 13–15). The functions of the 2A and 2B proteins are also different in distinct viruses. HAV 2A lacks the proteolytic activity displayed by enteroviral 2A proteins, which, in addition to their own cleavage, are responsible for shutting off host protein synthesis by cleaving eukaryotic translation initiation factor 4G (eIF4G) (16). HAV 2A is released from the polyprotein as the C-terminal region of the capsid protein precursor VP1-2A, being essential for the formation of the pentameric intermediates (17).

In picornaviruses, the nonstructural proteins 2B and 2C, their precursor 2BC, and protein 3A have sequences that integrate into the host membrane bilayer (18–20) and appear to be involved in both the rearrangement of the target membranes during infection and the tethering of the RNA replication complex to these membranes (21). Nevertheless, the mechanisms of action of these proteins vary enormously among the different picornavirus genera. In enteroviruses, viral infection fills the host cytoplasm with networks of densely packed double-membrane vesicles, mostly derived from the host endoplasmic reticulum (ER), Golgi apparatus, and lysosomes (22–26). In these vesicles, the 2BC, 2C, and 3A proteins are believed to anchor the viral replication complexes to the membranous structures. In all enteroviruses, 2B proteins display similar lengths (97 to 99 amino acids) and predicted secondary structures, containing two hydrophobic regions of ∼20 amino acids each. The first hydrophobic region appears to form a cationic amphipathic or amphiphilic α-helix, whereas the second one forms a completely hydrophobic helix. Upon insertion into the ER and Golgi complex membranes, the enterovirus 2B protein oligomerizes as a tetrameric bundle forming a pore with a diameter of ∼6 μm (27). These pores modify the host cell Ca²⁺ homeostasis and inhibit intracellular protein trafficking (18). In contrast, HAV 2B is ∼150 residues longer than enteroviral 2B proteins and shares little, if any, sequence and structural similarity with these proteins. HAV 2B is not predicted to associate in tetramers or to form pores, and accordingly, it does not seem to affect ion homeostasis or to interfere with host membrane trafficking (18). However, HAV 2B, 2C, and 2BC proteins expressed in cultured cells, either alone or in the context of other HAV proteins, induce important alterations in ER membranes. These rearrangements lead to the formation of compact clusters of tubular vesicles (18, 28) and a highly ordered crystalloid ER (29). In HAV-infected cells, in which 2B/2C/2BC proteins are expressed at much lower levels, membrane rearrangements are also thought to occur although at a much reduced scale. HAV 2B is predicted to have at least one hydrophobic transmembrane α-helix (30) and binds to ER membranes (18). In addition, it plays a key role in the establishment of HAV infections by preventing the synthesis of beta interferon (IFN-β) and consequently inhibiting the host innate immune response (31). It has also been suggested that HAV 2B might be involved in the release of viral particles, by weakening the cytoplasmic membrane (32).

Little is known about the structure of the proteins involved in the formation of viral factories for any of the picornaviruses: only a nuclear magnetic resonance (NMR) model of the soluble region of PV 3A has been reported (33). In particular, progress in the molecular characterization of HAV has been hampered by its low replication rate and difficulties in growing the wild-type virus in cellular cultures in vitro (34). Structural analysis of HAV proteins seems to be limited by their inefficient heterologous overexpression, probably due to the highly deoptimized codon usage and its concomitant effect on protein folding. So far, only the three-dimensional structure of one HAV protein, 3C proteinase, has been described (35).

Here we report the three-dimensional structure of the soluble domain of HAV 2B (2B_sol), determined at a 2.7-Å resolution. The extra N-terminal domain, apparently transferred from an ancestor 2A protein, is involved in a higher-order fibrillar structure, providing insights into host membrane rearrangements produced during HAV infection.

MATERIALS AND METHODS

Structure determination and refinement.

Expression, purification, and crystallization of HAV 2B, as well as crystallographic data collection and processing, were performed as previously reported (30). Briefly, a synthetic gene coding for a 217-residue protein (23 kDa), including HAV 2A and the soluble domain of the HAV 2B protein (2AB protein), was overexpressed in Escherichia coli and purified by affinity chromatography. Crystals were grown from native and selenomethionine-substituted 2AB protein and were used for X-ray diffraction data collection using synchrotron radiation at the European Synchrotron Radiation Facility (Grenoble, France), beamline ID23-EH2. The native data set contained information up to a 2.7-Å resolution, and the usable anomalous signal of the final derivative data set (obtained from two different crystals) extended to a 3.2-Å resolution.

The structure was determined by single-wavelength anomalous dispersion (SAD) from selenomethionine derivative data. Twelve selenium sites were located and refined by using autoSHARP (36) and were used to estimate experimental phases. The initial maps allowed automatic tracing of 230 residues, using ARP/wARP (37). Two independent molecules were identified in the asymmetric unit. Using the positions of selenium atoms as a guide for sequence assignment, the initial model was extended manually with COOT (38) and refined against the 2.7-Å native data by using REFMAC5 (39). Additional rounds of manual model rebuilding alternated with cycles of automatic refinement using BUSTER (40). The final refinement statistics are summarized in Table 1.

TABLE 1.

Crystallographic data quality, phasing, and refinement statistics

Parameter	Value forg:
	Selenomethionine derivativef	Native crystal
Data quality statistics
Space group	P43	P43
Unit cell (Å)	a = b = 90.3, c = 73.2	a = b = 90.4, c = 73.4
Resolution range (Å)	63.9–3.20 (3.37–3.20)	57.0–2.70 (2.85–2.70)
Mean 〈I/σ〉	3.8 (1.6)	8.9 (2.0)
Rp.i.m.a	0.04 (0.15)	0.07 (0.27)
Multiplicity	38.9 (18.6)	2.4 (1.3)
Completeness (%)	100 (100)	81 (62.9)
Anomalous multiplicity	19.9 (9.4)
Anomalous completeness (%)	100 (100)
Phasing statistics
No. of sites	12
Phasing powerb	1.99
Rcullisc	0.58
FOMh (centric/acentric)	0.22/0.41
Refinement statistics
Resolution (Å)		48.2–2.7
No. of reflections used in refinement		13,239
Rworkd/Rfreee		18.4/21.5
No. of atoms
Protein		2,263
Ligand		10
Water		25
B-factors (Å2)
Protein		69.39
Water + ligands		65.51
RMSDs
Bond lengths (Å)		0.010
Bond angles (°)		1.140
Ramachandran plot
No. (%) of residues in preferred regions		271 (96.8)
No. (%) of residues in allowed regions		9 (3.2)

^aR_p.i.m. = ∑_hkl∑_i|I_i(hkl) − 〈I_(hkl)〉|/∑_hkl∑_iI_i(hkl), where I_i(hkl) is the observed intensity and 〈I_(hkl)〉 is the average intensity of multiple observations of symmetry-related reflections. p.i.m., precision-indicating merging.

^bPhasing power = RMS (DANO_calc)/RMS (DANO_calc − DANO_obs), where RMS is the root mean square and DANO is the anomalous amplitude.

^cR_cullis = RMS (DANO_calc − DANO_obs)/RMS (DANO_obs), where RMS is the root mean square and DANO is the anomalous amplitude.

^dR_work = ∑_hkl||F_obs(hkl)| − |F_calc(hkl)||/∑_hkl|F_obs(hkl)|, where F_obs and F_calc are the structure factors deduced from measured intensities and calculated from the model, respectively.

^eR_free is as defined as described above for R_work but for 5% of the total reflections chosen at random and omitted from the refinement.

^fMerged data from two selenomethionine derivative crystals. Statistics referring to the individual data sets were reported previously (1).

^gValues in parentheses refer to the outermost-resolution shell unless otherwise indicated.

^hFOM, figure of merit.

Data analysis.

Amino acid sequence alignments were performed with T-Coffee (41). Disorder of the 2A region was predicted by using Foldindex (42), DisEMBL (43), Disprot (44), and PrDOS (45). The programs FoldAmyloid (46) and Aggrescan (47) were used to calculate the propensity of the sequence included in the β2-β3 hairpin (IMKFSWRG) to form amyloid fibrils. Structural alignments were performed with SHP (48). Illustrations were prepared with PyMOL (http://www.pymol.org/).

Thioflavin-T assay.

Thioflavin-T (ThT) fluorescence measurements were performed in 50-μl reaction mixtures containing 30 μM ThT, 20 μM protein, and one of three different buffers: 50 mM HEPES (pH 7.5), 100 mM NaCl, 1 mM EDTA, and 1 mM dithiothreitol (DTT) (buffer A); 50 mM HEPES (pH 7.5), 1 mM EDTA, and 1 mM DTT (buffer B); or 0.1 M morpholineethanesulfonic acid (MES) (pH 6.0), 1.26 M (NH₄)₂SO₄, and 5 mM DTT (buffer C). The reaction mixtures were overlaid with 20 μl of paraffin oil to prevent evaporation. The ThT fluorescence intensity of each sample was recorded every 30 min for 60 h at 37°C in 96-microwell plates on a FluoDia T70 microplate fluorimeter (Photon Technology International). The data presented are the averages of data from five independent experiments and were normalized by using Prism 5 (GraphPad Software).

Transmission electron microscopy.

A purified protein sample (100 μM) was incubated at 37°C in buffer A or C for 48 or 24 h, respectively. Ten microliters from each reaction mixture was collected and centrifuged. The supernatant was placed onto a Formvar- and carbon-coated nickel grid and blotted off after 2 min. The sample was then stained with 1% (wt/vol) uranyl acetate, dried, and observed at a magnification of ×80,000 by using a JEM-1400 transmission electron microscope (JEOL) equipped with an Orius Sc1000 digital camera (Gatan).

Protein structure accession number.

The refined structure of the HAV 2B protein was deposited in the Protein Data Bank under accession number 4WZN.

RESULTS

Overall structure of 2B_sol.

A synthetic gene coding for residues Ser765 to Gly981 of the HAV polyprotein (GenBank entry for HM175 [12]), corresponding to the 2A protein plus the soluble region of 2B, was expressed in a prokaryotic system and crystallized (30). The structure was determined at a 3.2-Å resolution by SAD methods from a selenomethionine derivative and refined against the 2.7-Å native data (Table 1). The electron density maps (Fig. 1A) allowed the tracing of two independent molecules in the asymmetric unit (named molecules 1 and 2), for which 138 residues (from Lys838 to Gln975) (molecule 1) and 142 residues (from Ser835 to Glu976) (molecule 2) were placed unambiguously. The modeled region corresponds almost exclusively to the soluble domain of protein 2B (2B_sol). The 70-amino-acid-long N-terminal segment, corresponding to the 2A region, is disordered in both molecules and is not visible in the electron density maps, although PAGE analysis of the crystals confirmed that the crystallized sample corresponds to intact HAV 2AB. The flexibility observed for the 2A region is in agreement with the results from several different disorder prediction algorithms (e.g., Foldindex [49], DisEMBL [50], Disprot [51], and PrDOS [52]).

FIG 1.

Overall structure of HAV 2B_sol. (A) Stereo view of a weighted 2F_o-F_c map of the β2-β3 loop region. The electron density (contoured at 1.5 σ) is represented as a blue mesh. The model is in a stick representation colored in atom-type code, with the carbon atoms of molecules 1 and 2 shown in yellow and blue, respectively. The N- and C-terminal residues of each loop are labeled. (B) Ribbon diagram of an HAV 2B_sol monomer, colored from blue (N terminus) to red (C terminus). Secondary structural elements are explicitly labeled. (C) Topology diagram of HAV 2B_sol, colored as described above for panel B. Residue numbers are indicated for the first and the last positions of each secondary structure element. (D) Superposition of the two 2B_sol molecules in the crystal asymmetric unit. Molecules 1 (yellow) and 2 (blue) differ mainly in the position of the β2-β3 hairpin. Subtle differences are also observed in the positions of strand β8 and helix α5.

2B_sol is folded into 5 α-helices and 8 β-strands, organized in two subdomains (Fig. 1B and C). The N-terminal subdomain consists of a pseudo-β-barrel, involving 5 β-strands (β1, β4, β5, β6, and β7) arranged in a curved antiparallel β-sheet. The two longest strands, β1 and β4, are prolonged by a β-hairpin formed by the short strands β2 and β3. The second subdomain is folded into a helical bundle formed by 4 α-helices (α1 to α4) (Fig. 1B and C). The C-terminal tail of 2B_sol is arranged in a short β-strand (β8) that is stabilized by an antiparallel interaction with the N-terminal portion of strand β4. Strand β8 is followed by the short helix α5, which projects perpendicularly from the main body of the protein (Fig. 1B and C).

The two independent molecules in the asymmetric unit are nearly identical, with a root mean square deviation (RMSD) of 0.51 Å for the superposition of 122 equivalent residues, differing almost exclusively in the conformation and disposition of the β2-β3 hairpin loops (Fig. 1D): compared to molecule 1, the β2-β3 hairpin of molecule 2 is rotated by 140° along the hairpin axis and is tilted by 40° toward helix α5. As a consequence of this rearrangement, strand β4 of molecule 2 appears shorter at its N terminus, thus preventing the interaction with β8, which in turn loses the standard β conformation. Helix α5 is also slightly shifted in molecule 2 (Fig. 1D).

Structural relationship between HAV 2B_sol and enterovirus 2A^pro.

In hepatoviruses, protein 2B is >2-fold larger than enteroviral 2B proteins. As a result of a change in the substrate specificity of the HAV 3C^pro protease compared to enterovirus 3C^pro (8–11), the N-terminal domain of HAV 2B, corresponding to the structure described here, is located upstream of the residues homologous to the 2A/2B junction in enteroviruses (Fig. 2A and B). Thus, in the enterovirus polyprotein, the region equivalent to HAV 2B_sol would correspond to the carboxy terminus of 2A^pro (Fig. 2A and C).

FIG 2.

Structural relationship between the N-terminal domains of HAV 2B and the enteroviral 2A proteins. (A) Schematic representation of the 2A/2B junction in different picornavirus genera: the enteroviruses coxsackievirus B4 (CVB4), enterovirus 71 (EV71), and human rhinovirus 2 (HRV2); the hepatovirus hepatitis A virus (HAV); and the parechovirus human parechovirus 1 (HPEV1). The polyprotein cleavage sites are indicated with vertical lines, colored in red when cleavage is performed by 3C^pro and in turquoise when it is catalyzed by 2A^pro. The region corresponding to the 2A/2B cleavage site in enterovirus, which is conserved in all five viruses, is highlighted with a blue box. The catalytic residues from enterovirus 2A proteins are indicated with cyan star. The green box indicates the structurally superposable regions between HAV 2B and enteroviral 2A proteins. The dotted line in parentheses in the HAV polyprotein represents 75 residues, which have been compacted to simplify the comparison with the other polyproteins. (B) Amino acid sequence alignment of the 2A/2B junction region of the picornaviruses indicated in panel A. For the alignment, the sequences under GenBank accession numbers ABF19105 (CVB4), Q66478 (EV71), P04936 (HRV2), P08617 (HAV), and CAQ76820 (human parechovirus 1) were used. Strictly conserved residues in the five viruses are shown as red blocks, and similar residues are shown in blue boxes. Residues are numbered according to their position in the respective polyprotein. (C) Structural comparison between HAV 2B_sol (left) and HRV2 2A^pro (PDB accession number 2HRV [50]) (right). Both proteins are displayed as gray cartoons, with the structurally equivalent domains shown in red and the equivalent residues in the SHP superposition shown in dark green. The catalytic residues in HRV2 2A^pro are represented as sticks and colored in cyan.

Structural comparisons of HAV 2B_sol and the existing enteroviral 2A^pro structures (50–52) show striking similarities between the β-region of HAV 2B_sol and the β-barrel of human rhinovirus 2 (HRV2) 2A^pro. In fact, superposition of these two regions by SHP resulted in 28 equivalences (representing >50% of the HRV2 2A β-barrel) with an RMSD of 2.23 Å and an average distance of 1.94 Å. The equivalent residues are located within four of the five β-strands of the pseudo-β-barrel of HAV 2B_sol (β1, β4, β5, β6, and β7), which align with four of the six β-strands of the HRV2 2A β-barrel (Fig. 2C).

Quaternary organization of 2B_sol.

Crystal packing analysis revealed that HAV 2B_sol molecules are organized as helical fibers with a diameter of ∼92 Å, growing along the crystal c axis (Fig. 3). The contacts between the different molecules along this axis are established through the alternate concatenation of β2-β3 hairpin loops, which are oriented orthogonally to the fiber longitudinal axis and organized in a very thin (∼25-Å) but endlessly long antiparallel β-sheet, which we named the “β-spine” (Fig. 3). This peculiar arrangement is facilitated by the different β2-β3 hairpin conformations in the two independent molecules of the crystal asymmetric unit. The β2-β3 hairpin is stabilized by a strong cation-π interaction between the side chains of Trp854 and Arg855 and by a hydrophobic cluster formed by Met850, Phe852, and Val857. The other face of the β-hairpin is polar and exposes the side chains of Lys851 and Thr858 (Fig. 3A), resulting in a highly polarized β-spine: while the hydrophobic face involves extensive contacts of hydrophobic clusters on adjacent β-hairpins, the hydrophilic face is characterized by the formation of hydrogen bonds between the side chains of Thr858 residues on opposing β-strands (Fig. 3A).

FIG 3.

The 2B_sol protein forms a fiber along the crystal c axis. Shown is an overall view of the fiber generated along the crystal c axis (middle), with the two molecules in the asymmetric unit shown as yellow (molecule 1) and blue (molecule 2) ribbons. Panels A to D show details of the interactions established between protomers. Hydrogen bonds are represented as dotted black lines. (A) Closeup view of two consecutive β-hairpins, highlighting the polarization of the fiber. The hydrophobic cluster is shown at the right, while a 180° rotation view, showing the hydrophilic face of the β-hairpin, is displayed at the left. (B) Interactions established between the β2-β3 loop of molecule 1 and the main body of molecule 2. (C) Interaction between the β2-β3 loop of molecule 2 and the main body of molecule 1. (D) Helix α5 from molecule 1 interacts with both helix α5 and strand β2 of molecule 2. (E) Crystal packing interactions in the a-b plane. The reference HAV 2B fiber is shown in yellow and blue, and the symmetry-related molecules are shown in gray. (F) Closeup view of the interactions established between neighboring fibers in the crystal. These interactions involve residues Pro943, Glu947, and Thr948 within loops α3 and α4 from the two pseudo-2-fold-related molecules 1 and 2. The location of the pseudo-2-fold axis is indicated by a black diamond.

Additional interactions, involving Trp854 (within the loop in the β2-β3 hairpin) of molecule 1 and Ala884 (loop β5-β6) of molecule 2, together with those established between Ser853, Trp854, and Arg855 of molecule 2, contacting Glu848 (loop β1-β2), Tyr843 (strand β1), and Thr861 (loop β3-β4) of molecule 1, respectively, connect the β-spine to the main body of 2B (Fig. 3B to D).

The two alternate conformations of 2B_sol facilitate the helicoidal arrangement of the α5 helices around the β-spine, while the major bodies of the 2B_sol molecules are disposed in a cross-like organization, consistent with P4₃ crystal packing (Fig. 3E). In the crystal context, the 2B_sol fibers are packed through interactions between the α3-α4 loops of a local pseudo-2-fold symmetry axis that relates the main body of molecules 1 and 2 (Fig. 3F). The tetragonal packing leads to the formation of large cavities of a square section (∼76 Å by 76 Å) with one fiber at each vertex (Fig. 3E). The 70 disordered amino acids at the N terminus of the 2AB construct (see above) probably occupy these seemingly empty spaces.

2B_sol forms fibers in solution.

The assembly of HAV 2B_sol into a helicoidal β-spine along the crystal c axis is reminiscent of the structures observed in amyloid fibers (52). In agreement, different algorithms (FoldAmyloid [46] and Aggrescan [47]) predict that the 2B_sol region forming the β-spine, the β2-β3 hairpin (IMKFSWRG), is involved in aggregation and/or amyloid formation (Fig. 4A).

FIG 4.

HAV 2B_sol forms amyloid-like fibers in vitro. (A) Propensity of the sequence included in the β2-β3 hairpin (IMKFSWRG) to form amyloid fibrils, as estimated by FoldAmyloid (40) (gray line, left axis) and Aggrescan (41) (black line, right axis). A black horizontal line indicates the threshold above which the sequence is considered amyloidogenic. The sequence and secondary structure of the hairpin region are represented below the graph. (B) HAV 2B_sol aggregation kinetics monitored by thioflavin-T fluorescence at 37°C under near-physiological conditions (buffers A and B) or in crystallization buffer (buffer C). Under near-physiological conditions (buffer A), the kinetics of aggregation can be clearly divided into three phases: an ∼13-h lag phase with a slight upward slope followed by a steep increase and a final slow increase in fluorescence intensity. This behavior is characteristic of a nucleated polymerization reaction. The lag phase is shortened when a low-ionic-strength buffer is used (buffer B), and the slope of the curve becomes steeper in the presence of the buffer with a lower pH used in the crystallization assays (buffer C). Buffer A contained 50 mM HEPES (pH 7.5), 100 mM NaCl, 1 mM EDTA, and 1 mM DTT; buffer B contained 50 mM HEPES (pH 7.5), 1 mM EDTA, and 1 mM DTT; and buffer C contained 0.1 M MES (pH 6.0), 1.26 M (NH₄)₂SO₄, and 5 mM DTT. (C and D) Electron micrographs of negatively stained HAV 2B_sol end-stage fibrils grown at 37°C in buffer A (physiological conditions for 48 h) (C) or buffer C (crystallization conditions for 24 h) (D). The boxed insets are magnifications of representative areas of the main images. Bar: 200 nm. The end-stage HAV 2B_sol fibrils have an average diameter of ∼14 nm, compatible, in general terms, with that of the crystal 2B_sol fiber (∼9 nm) plus the two N-terminal 2A regions attached to the outer surface of the 2B_sol main body (70 amino acids each; not seen in the electron density maps).

To further assess the propensity of HAV 2B_sol to self-assemble into ordered β-fibrillar structures in solution, the appearance of amyloid protofibrils was monitored in solution by using the fluorescent dye thioflavin-T. Thioflavin-T stacks onto the characteristic β-sheet structure of the growing fibril, allowing real-time monitoring of earlier stages of the aggregation process (53, 54). Thioflavin-T binding kinetics assays were performed at 37°C either in the presence of the protein storage buffer (Fig. 4B), which mimics near-physiological conditions, or upon incubation with the crystallization buffer (Fig. 4B). All aggregation assays showed the characteristic behavior of a nucleated polymerization reaction, independently of the buffer system used, revealing that self-assembly into β-fibrillar structures is not induced by the crystallization conditions and can also occur in solution under near-physiological conditions (Fig. 4B). Furthermore, electron micrographs of the end-stage polymerization products revealed that they possess a twisted fibrillar morphology, independently of the buffer system used (Fig. 4C and D). These short protofibrillar structures, compatible with the helicoidal β-spine observed along the crystal c axis, are often observed in prefibrillar wormlike amyloid oligomers (54, 55).

Taken together, these results indicate that, in solution, HAV 2B_sol spontaneously organizes into fibrillar higher-order structures that have structural features characteristic of amyloid-like protein self-assemblies.

DISCUSSION

A shift in the HAV 2A/2B cleavage site results in the fusion of a new domain to the N terminus of 2B.

The replication strategy of HAV, typically establishing persistent infections without cytopathic effects when grown in vitro in cell cultures, is in contrast to that of cytolytic picornaviruses, which block host protein translation to enhance viral protein synthesis. In enteroviruses, host shutoff is induced by the 2A proteinase, which, in addition to being responsible for primary polyprotein processing, cleaves eukaryotic translation initiation factor 4G (eIF4G) (16), inhibiting the recruitment of capped mRNA to the ribosomes by the eIF4F initiation complex. In this way, the internal ribosome entry site (IRES)-dependent initiation of the picornaviral mRNA is enhanced (56, 57). The three-dimensional structures of the enteroviral 2A proteins from HRV2, coxsackievirus B4 (CVB4), and enterovirus 71 (EV71) (50–52) reveal the presence of a small N-terminal domain composed of a four-stranded antiparallel β-sheet and a larger C-terminal domain containing a six-stranded antiparallel β-barrel, with the active site located in the cleft between the two domains (50, 51).

The structure of 2B_sol determined in this work shows that the N-terminal region of 2B is organized into a pseudo-β-barrel subdomain, which appears to be closely related to the C-terminal β-barrel of the enteroviral 2A^pro protein (Fig. 2C). The observed structural relationship between the N terminus of HAV 2B and the C-terminal β-barrel of the enteroviral 2A^pro protein, together with the high level of sequence conservation of the region corresponding to the enteroviral 2A/2B junction (in contrast to the high level of sequence divergence for the rest of the polyprotein), suggests that this genomic region of HAV may have evolved from an enterovirus-like ancestor, whose 3C protease acquired subtle changes in substrate specificity. The different specificity of 3C would have resulted in the displacement of the 2A/2B junction (13) (Fig. 2A) and the concomitant truncation of protein 2A, thus leaving its C-terminal β-barrel linked to 2B.

The shift of the cleavage site would have disrupted the 2A^pro catalytic triad, leaving the key cysteine residue (the nucleophilic residue that performs catalysis on the substrate) on the 2B side of the cleavage, while the other two catalytic residues (histidine and aspartic acid) would remain on a different polypeptide chain, the truncated 2A protein (Fig. 2). Alignment of the HAV 2A sequence with that of the N-terminal domain of enterovirus 2A^pro aligns HAV residue His793 with the catalytic histidine of the enterovirus proteinases (His18 in HRV2 2A^pro and His21 in CBV4 2A^pro and EV71 2A^pro) (data not shown). The aspartic acid and the cysteine residues, on the other hand, do not seem to be conserved and could not be identified in either the HAV 2A or 2B protein sequence. This is in accordance with the very low level of sequence conservation between enteroviruses and hepatoviruses in this region, consistent with a reduction of sequence mutation restriction and an increase of functional sequence space that would follow the proposed shift of the cleavage site and the loss of catalytic activity (58).

As a consequence of the catalytic-site disruption, HAV 2A lacks any proteolytic activity and cannot induce host shutoff in HAV-infected cells (59). On the other hand, the displacement of the 2A/2B junction also leads to the acquisition of an additional domain (a β-barrel) by protein 2B.

In light of the structural data presented here, the presence of the additional N-terminal domain is what provides the HAV 2B protein its distinctive ability to self-assemble into stable fibrillar structures with amyloid-like properties. A number of proteins acquiring different activities after the addition or loss of one domain have been described in the literature. One illustrating example is the regulation of human Fas receptor apoptosis-signaling activity by alternative splicing of its pre-mRNA: Fas (Apo-1/CD95) gene exon 6, coding for a transmembrane domain, can be included or skipped to generate mRNAs encoding a membrane-bound form of the receptor that promotes apoptosis or a soluble isoform that can act as an inhibitor of programmed cell death, respectively (60, 61).

Role of HAV 2B fibers in host intracellular membrane rearrangement.

Remodeling of host intracellular membranes to form structures associated with viral RNA synthesis is a characteristic feature of picornavirus infections. The nonstructural proteins 2B and 2C, as well as their precursor 2BC, play a central role in these membrane rearrangements through a mechanism that remains poorly understood (reviewed in reference 62).

Although HAV replication in cultured cells occurs at much lower rate than with most other picornaviruses, similar mechanisms appear to be at work. The HAV 2B, 2BC, and 2C proteins demonstrate efficient membrane association properties causing major rearrangements of ER membranes. These rearrangements display variable morphologies, as a highly ordered crystalloid ER (29) or packs of associated tube-shaped vesicles (28), when overexpressed in HeLa or FRhK-4 cell lines. Moreover, unique membranous structures that may serve as sites of viral replication were also observed after adaptation of HAV to growth in cell culture (28).

The crystallographic structure of HAV 2B_sol presented here sheds some light on the mechanism by which this protein can induce such dramatic membrane remodeling. The fiber-like arrangement adopted by HAV 2B_sol in crystals (Fig. 3) is suggestive of an amyloid-like structure. The present work also shows that this protein is able to spontaneously form thioflavin-T binding fibrils in vitro, under near-physiological conditions (Fig. 4). Therefore, the β-spine arrangement observed in the crystal packing of 2B_sol molecules likely reflects the organization adopted by 2B proteins in HAV-infected cells, adding to the growing number of proteins with the ability to self-assemble into functional amyloid-like structures (35, 52, 55, 63, 64). Moreover, the ability of the HAV 2B protein to bind microsomal membranes seems to be related to the presence of a putative transmembrane helix located at the C-terminal end of the protein, which is predicted to start 30 residues downstream of the sequence of the present crystallographic structure (30). The interaction established between strands β4 and β8 (Fig. 1) places the C terminus of 2B_sol close to the β2-β3 hairpin and, thus, correctly positioned to assist in the anchoring of 2B fibers on the proximal membranes through the transmembrane helices (Fig. 5). Reorganization of the host membranes around the 2B fibers could lead to both the tightly looping crystalloid ER and the tubular network observed in vitro in HAV-infected cells after adaptation of HAV to growth in cell culture (28, 29).

FIG 5.

Role of HAV 2B protein in membrane rearrangement. An HAV 2B membrane binding model is shown. The HAV 2B_sol β-spine is displayed as a cartoon in yellow and blue. The predicted transmembrane helix has been modeled and is shown in gray. Two membrane bilayers are displayed as gray sticks, with the polar heads in atom-type color (blue for nitrogen, red for oxygen, and orange for phosphorus). For clarity, only the β2-β3 hairpin is shown for some HAV 2B_sol monomers.

The fibrillar aggregation of the 2B protein (either the mature form or the 2BC precursor) may act as a recruitment platform, gathering all the viral and cellular components required for HAV replication. In HAV infections, in which the viral components are expressed at unusually low rates, the formation of such recruitment platforms might play a key role in ensuring that all the newly synthesized viral factors accumulate in the replication and assembly factories.