Structure of the archaeal head-tailed virus HSTV-1 completes the HK97 fold story

Abstract

It has been proposed that viruses can be divided into a small number of structure-based viral lineages. One of these lineages is exemplified by bacterial virus Hong Kong 97 (HK97), which represents the head-tailed dsDNA bacteriophages. Seemingly similar viruses also infect archaea. Here we demonstrate using genomic analysis, electron cryomicroscopy, and image reconstruction that the major coat protein fold of newly isolated archaeal Haloarcula sinaiiensis tailed virus 1 has the canonical coat protein fold of HK97. Although it has been anticipated previously, this is physical evidence that bacterial and archaeal head-tailed viruses share a common architectural principle. The HK97-like fold has previously been recognized also in herpesviruses, and this study expands the HK97-like lineage to viruses from all three domains of life. This is only the second established lineage to include archaeal, bacterial, and eukaryotic viruses. Thus, our findings support the hypothesis that the last common universal ancestor of cellular organisms was infected by a number of different viruses.

Every organism is constantly under the attack of viral infections because viruses outnumber cellular organisms by at least an order of magnitude. It has been estimated that viruses kill approximately one fifth of the marine biomass on a daily basis (1, 2). Furthermore, viruses are the leading cause of infectious diseases (3). Thus, viruses have a crucial role in controlling the evolution of cellular organisms owing to their high abundance and genetic diversity (1). However, cryoelectron microscopic and crystallographic studies have revealed that this genetic diversity may be reduced to a limited number of structure-based viral lineages (2, 4–6).

The discovery that a bacterial and an animal virus have the same major capsid protein (MCP) fold and virion architecture (5, 7) led to the hypothesis that viruses infecting cells that belong to different domains of life (in this case Bacteria and Eukarya) may have a common origin and be related, although there is no detectable sequence similarity. Further examination of viral capsid protein structures has revealed that this double-β-barrel MCP of tailless icosahedral viruses is also present in viruses infecting archaea (5, 8, 9). These observations have led to the hypothesis of viral structure-based lineages unifying viruses previously considered unrelated and leading to a paradigm shift in how we see the viral universe (2, 4, 5).

A second lineage has been proposed containing head-tailed bacterial viruses and herpesviruses sharing the so-called canonical Hong Kong 97 (HK97)-like MCP fold (4, 10). This fold was first determined for bacteriophage HK97 (10) and has since been discovered in seven additional phages and in herpes simplex virus type 1 (11-18). Head-tailed viruses morphologically similar to phages are also known to infect archaeal organisms (19, 20). Comparative genomics indicate that these archaeal viruses share common genes with bacteriophages for virion assembly and maturation, as well as genome packaging (21, 22). The fundamental issue is whether archaeal head-tailed viruses also share the canonical HK97-like MCP fold, as suggested by bioinformatics (21, 23). Interestingly, the HK97-like fold has been recognized in icosahedral particles produced by hyperthermophilic archaea (24, 25). However, these particles are most likely nanocompartments for enzyme storage, not viruses because they contain no nucleic acids (24, 25).

To date, only approximately 30 high-resolution protein structures from archaeal viruses have been determined, and only 5 are MCPs. However, none come from head-tailed viruses (26). All known archaeal head-tailed viruses infect either extreme halophiles or anaerobic methanogens belonging to the phylum Euryarchaeota (20). Here we set out to screen a previously isolated group of haloarchaeal head-tailed viruses (19) to find a suitable candidate for the determination of the MCP fold by electron cryomicroscopy (cryo-EM). A podovirus, Haloarcula sinaiiensis tailed virus 1 (HSTV-1) (19), was chosen because it was also stable at low salinity and could be readily purified. The genome of HSTV-1 was sequenced, and the major structural proteins were identified. The 3D reconstruction of the HSTV-1 capsid at subnanometer resolution revealed that the virus adopts the HK97-like fold. Thus, the HK97-like lineage can be extended to archaeal head-tailed viruses, suggesting a common ancestor for prokaryotic head-tailed viruses and eukaryotic herpesviruses.

Results

Virus Production and Morphology.

HSTV-1 virus sloppy agar stocks had high titers, on average 3 × 10¹¹ pfu/mL. Turbidity measurements of the infected cultures demonstrated that the virus is lytic (Fig. S1). Virus titers in liquid cultures, however, reached only approximately 6 × 10⁹ pfu/mL, and for this reason the virus was purified directly from the stocks. The specific infectivity of the purified HSTV-1 virions was ∼9 × 10¹² pfu/mg of protein, and the density in CsCl was ∼1.45 g/mL, which is typical of head-tailed viruses (27). Cryo-EM micrographs of HSTV-1 virions at high and low salinity showed that the overall morphotype remained unaltered despite changes in salinity (Fig. 1A). HSTV-1 was inactivated at low salinity, but the infectivity was regained when the high salinity conditions were restored (Fig. 1B). HSTV-1 had an isometric, genome-filled capsid and a short, podovirus-like tail with at least five long fibers attached to the central tail structure (Fig. 1, Inset).

Fig. 1.

Podovirus HSTV-1 tolerates salinity changes. (A) Cryo-EMs of HSTV-1 at 4-μm underfocus. (I) Virions incubated in 9% (wt/vol) SW containing 1.2 M NaCl, (II) virions in 20 mM Tris·HCl (pH 7.2) buffer, and (III) virions first incubated in 20 mM Tris·HCl (pH 7.2) and then restored in 9% (wt/vol) SW. (Inset) (6-μm underfocus) A particle viewed along the tail and showing possible tail fibers. (Scale bars, 50 nm.) (B) Specific infectivity of HSTV-1 virions at a logarithmic scale. (I, II, and III) High and low salinity conditions as in A.

Genome Sequence and Structural Proteins.

The genome of HSTV-1 contained 53 putative open reading frames (ORFs) (Table S1). As has been seen in the few head-tailed archaeal viruses sequenced previously (21, 28, 29), the organization of the genome is reminiscent of the genomes of tailed bacteriophages: the predicted genes occupy 90% of the genome and are closely spaced and in some cases slightly overlapping (Fig. 2A). The majority of the predicted proteins of HSTV-1 make no sequence matches to the sequence databases in a BlastP search, and of the ones that do match, more than half match “hypothetical proteins” with unknown function. The predicted proteins that do match proteins of known function are also typical of the tailed bacteriophages, with functions such as DNA replication, nucleotide metabolism, and structural proteins of the virion. Some of the structural proteins can be identified by a simple BlastP search, and some are found with more extensive searching (PSI-Blast, HHpred). We confirmed a few of these identifications by N-terminal sequencing of virion components (Fig. 2 B and C). Strikingly, the order of the genes for structural proteins identified in this way is the same as the order almost always observed for the corresponding genes in the genomes of tailed bacteriophages.

Fig. 2.

Genome and proteins of HSTV-1. (A) Genome of HSTV-1 is presented with markers spaced at 1-kbp and 100-bp intervals. The predicted genes are shown as boxes either above or below the genome, depending on whether they are rightward or leftward transcribed, respectively. The gene numbers are shown for annotated genes, and putative functions are shown above the genes. Abbreviated functions are histidine-asparagine-histidine homing endonuclease (HNH), proliferating cell nuclear antigen family of sliding clamps (PCNA), and a minichromosome maintenance helicase (MCM). (B) Protein profile of the purified virions in a tricine-SDS-polyacrylamide gel stained with Coomassie blue. Numbers on the left indicate molecular mass markers. The identified proteins are indicated by the corresponding gp (gene product) number. (C) N-terminal sequences determined from the protein bands gp12, gp29, and gp14. The first two amino acids of the N-terminal sequence of gp12 and gp29 could not be identified. The theoretical molecular masses of the mature proteins are in parentheses.

The HSTV-1 genome is 32,189 bp long. The sequence assembles as a circle, and there is no indication in the sequence data for discrete ends, which argues that the genome is circularly permuted across the population of virions. As base pair 1 we have chosen the first base pair of the large terminase subunit gene. We identified four genes encoding structural proteins and two more as genes encoding enzymes (terminase and protease) required for capsid assembly and maturation: the terminase large subunit gene (gene 1) and the portal gene (gene 3) are identified directly by the sequence similarity of their protein products to the corresponding proteins of tailed bacteriophages. The portal similarity extends only through the first two-thirds of the sequence; the last one-third is homologous to a minor head protein found in some bacteriophages (gpF in phage Mu; gp7 in phage SPP1), which is encoded by a gene immediately following the portal gene. The two genes are evidently fused in HSTV-1. The major capsid gene (gene 14) was identified by N-terminal sequencing of the most prominent virion protein revealed by SDS-gel electrophoresis (Fig. 2B). This sequence starts at amino acid 107 of the predicted protein, indicating that the major capsid protein is cleaved at this position, presumably during capsid maturation. The protease is probably the product of gene 13; the sequence-based evidence for this identification is not strong, but the location of the gene just upstream from the gene encoding its putative target is consistent with the arrangement found in tailed bacteriophages. Two additional genes (genes 12 and 29) are identified as encoding virion components on the basis of the N-terminal sequences of their proteins (Fig. 2 B and C). The sequence of gene 12 does not identify any known viral proteins, but sequence-based structural predictions show a high probability for a protein with a β-helix structure. Gene 12 is unusual in having a distinctly different G+C content from most of the rest of the genome (46% G+C for gene 12, compared with 60% for the genome as a whole). Gene 29 also encodes a moderately abundant component of the virion. Because the tail fibers are the most prominent feature of the virions not otherwise accounted for, we suggest that they are composed of the protein encoded by gene 29.

There are five additional plausible functional identifications in the remainder of the HSTV-1 genome. Gene 48 encodes a putative cytosine methyl transferase and gene 52 a putative adenine methyl transferase. We find two likely DNA replication functions, a member of the proliferating cell nuclear antigen family of sliding clamps encoded by gene 40 and a member of the minichromosome maintenance family of replication helicases encoded by gene 45. These are the archaeal versions of replication functions that are frequently found in bacteriophage genomes. Gene 19 encodes a putative homing endonuclease. Such endonuclease genes are usually contained within an intron, which typically disrupts an essential gene. It seems likely that gene 19 is in fact part of an intron, inasmuch as it is within an island of low G+C content that extends from nucleotide ∼12,650 to ∼13,510.

3D Structure of the Capsid.

We solved the structure of the DNA-filled capsid of HSTV-1 using cryo-EM and 3D image reconstruction. High salt conditions lower the contrast and consequently inhibit high-resolution reconstruction (30). To avoid this, we imaged the virions in low salt conditions, which they could still tolerate, allowing high-resolution structure determination to 8.9 Å (Fig. 3) (31).

Fig. 3.

HSTV-1 capsid structures. (A) A 2.26-Å-thick central section of the HSTV-1 icosahedral reconstruction viewed along an icosahedral twofold axis. Twofold (ellipse), threefold (triangle), and fivefold (pentagon) symmetry axes are indicated. (Scale bar, 10 nm.) (B–D) Capsid isosurface representations along an icosahedral twofold axis. The HSTV-1 capsid is displayed at σ = 0.9 (B) and σ = 3.0 (C and D) threshold. The empty HSTV-1 (D) capsid is shown sectioned with the front half of the capsid removed. The empty capsid was generated by manually removing most of the genome density in UCSF Chimera. The symmetry axes are indicated as in A. The models were colored using radial-depth cueing (bars, 210- to 360-Å radius) in UCSF Chimera (56).

The virion has several layers of highly ordered, concentrically packed dsDNA with a spacing of ∼23.6 Å (Fig. 3A). The genome packaging density is 0.46 bp/nm³. The diameter of the HSTV-1 capsid is 560 Å facet to facet and 624 Å vertex to vertex (Fig. 3A). The shell is only 20 Å thick. On the virion surface, HSTV-1 has 20 cone-shaped towers (Fig. 3 A and B). These towers are at least 138 Å long and 74 Å wide, located at positions of threefold symmetry at the center of each facet. The tower density in the reconstruction is considerably lower than that of the rest of the capsid, yet the structure is well defined, suggesting that not all tower sites are occupied (Fig. 3 B and C).

The handedness of the capsid was determined using tilt experiments and HK97 proheads [triangulation number (T-number) = 7 laevo] as a control. Capsomers of the virion are arranged on a T = 7 laevo lattice, with the capsid protein clustered into flat hexamers and angular pentamers, similar to those seen in HK97. Thus, there are seven nonequivalent gp14 subunits, organized into an ABCDEF hexamer and a G₅ pentamer using the established HK97 nomenclature (Fig. 3 B–D) (10). In total there are 415 copies of gp14 in the capsid, because one pentamer is replaced by a portal vertex to which the tail is attached (Fig. 1). There are three main types of contacts between capsomers: the basal domains interact via a strict threefold contact on the threefold axis of symmetry (Fig. 3), a quasi-threefold contact between two hexamers and a pentamer, and a quasi-threefold contact between hexamers connecting near the twofold axis of symmetry.

Close inspection of the capsid revealed that the overall topology of the MCP seemed to be similar to that of HK97. It was possible to directly place the asymmetric unit from the atomic model of HK97 Head II (32) into the reconstruction (Fig. 4A and Movie S1), resulting in the superposition of the major features of each monomer (Fig. S2). We then averaged the monomers within one hexamer together to improve the definition of the monomer (Fig. 4B). The G subunit was noticeably too different to the A–F subunits to be used in the averaging. Independently we generated a homology model of gp14, identified as the MCP (Fig. 2), using the I-TASSER server (33, 34). The best model had an I-TASSER C-score of only −2.81 which is a low confidence value (33). All of the top five models were very similar to the HK97 fold. We took the top model and used rigid body fitting to place the model within the averaged electron density of one monomer. The superposition of the homology model (Fig. 4B, blue ribbon) within a single monomer was as good as that with the HK97 MCP (Fig. 4B, red ribbon), thus showing that the homology model is indeed validated by the experimental results, despite the low C-score. According to the structural alignment of HK97 gp5* and the proteolytically cleaved form of gp14, only 33 of 307 amino acids are identical. Gp14 has the two typical mixed α-β domains found in this protein fold, the axial (A) domain and the peripheral (P) domain, along with an extended N terminus and the E loop (Fig. 4 and Fig. S2). The A domains interact around the five- and quasi-sixfold axes within the capsomers, whereas the P domains meet at the three- and quasi-threefold axes (Fig. 4) between capsomers (Fig. 3). The N terminus of one subunit lies parallel to the E-loop of the adjacent subunit.

Fig. 4.

HSTV-1 gp14 has the HK97 fold. (Left) Viewed from the capsid interior; (Right) viewed from the capsid exterior. (A) Cryo-EM density of an HSTV-1 asymmetric unit drawn at σ = 3.0 threshold (gray transparent). The atomic model of HK97 asymmetric unit (PDB ID 1OHG) (32) is shown in red ribbon after rigid body fitting into the density in UCSF Chimera (56). (B) Averaged density of the six segmented subunits from one hexamer shown at σ = 3.5 threshold (gray transparent). HK97 chain E in red ribbon (PDB ID 1OHG) (32) and homology model of gp14 in blue ribbon (33, 34) after rigid body fitting into the density in UCSF Chimera (56). (C) Red surface represents a hexamer built from six copies of the averaged subunit. Green wireframe shows the position of the pentameric subunit that has a significantly different conformation than the hexameric units and thus was not included in the averaging.

Discussion

Our study shows that the podovirus HSTV-1 infecting an extremely halophilic archaeon, H. sinaiiensis (19), has the same canonical HK97-like fold of the MCP as bacterial head-tailed viruses and eukaryotic herpesviruses. Thus, we evolutionarily connect these viruses from all three domains of life. Hence we provide structural and genomic evidence that the head-tailed viruses were evolving before the archaea, bacteria, and eukaryotic domains of life split approximately 2 to 3 billion years ago (2, 5, 7).

Analysis of the HSTV-1 genome showed conservation with archaeal and bacterial head-tailed viruses in four groups: genes encoding virion structural components (MCPs and portal proteins), those involved in virion assembly (prohead proteases and terminase proteins), nucleotide metabolism, and DNA replication. Because the genome seems to be circularly permutated, and we identified both potential portal and terminase genes, HSTV-1 most likely uses a headful-DNA packaging mechanism. Furthermore, because the HSTV-1 MCP is proteolytically processed (Fig. 2 B and C), the head may undergo maturation whereby the N-terminal domain of the MCP is proteolytically removed and the capsid undergoes expansion and stabilization as the DNA is packaged, thus mimicking the HK97 assembly pathway (35). However, one of the crucial steps in HK97 assembly is the formation of cross-links within the capsid through covalent isopeptide bonds between Lys169 and Asn356 on adjacent subunits (10, 36). Despite the similar overall conformation of the MCP of HK97 and HSTV-1, these residues are not conserved when the HSTV-1 homology model and HK97 gp5 are aligned (Fig. 4B), and there is no evidence for such cross-linking to stabilize the HSTV-1 capsid. One of the very few residues that is conserved is gp5* Arg294 (gp14 Arg311). Arg311 from adjacent subunits form a ring around both the quasi-sixfold axis of symmetry and the fivefold axis of symmetry, which could coordinate anions (32).

We found that HSTV-1 was stable but reversibly noninfectious at low salinity. Hence HSTV-1 requires high salinity for infection, most probably for effective adsorption on to the host, the extremely halophilic archaeon, H. sinaiiensis (19). We identified base plate tail fibers, most likely built of gp29, that are probably involved in the initial stages of this interaction. Both haloarchaeal myo- and siphoviruses have previously been shown to have a similar dependency on high salinity for infection (37).

HSTV-1 has large, cone-shaped tower structures extending from the surface (Fig. 3). On the basis of its abundance in the virion, we suggest that gp12 is the main component of these towers. Most of the genome has a much higher G+C content than gene 12, thus we argue that gene 12 joined the genome at an evolutionarily recent time. The recent tower addition may have helped the virus to adapt to new conditions, such as a new host strain, and may also help to stabilize the capsid.

In the homology modeling described here, the sequence similarities were too low to give a reliable structural prediction without the experimental verification that we provided. Viruses evolve rapidly, and thus the genome or protein sequence similarity may be detected only between closely related viruses (38). Consequently, structure-based approaches offer an opportunity to reach over longer evolutionary distances and to detect connections in structure and assembly pathways even between viruses infecting organisms from different domains of life (2, 4, 5). The final question remains how many structure-based viral lineages exist. It has been estimated that despite the immense number of viruses, there might be only a low number of unique viral coat protein folds owing to the limited protein fold space (4). However, the only way to test the hypothesis is to determine more coat protein folds and viral structures, especially from archaeal viruses with novel morphotypes.

Materials and Methods

Virus Production and Purification.

H. sinaiiensis strain ATCC 33800 (39) was used as a host for HSTV-1 (19). Cultures were grown aerobically at 37 °C in modified growth media containing artificial salt water (SW) (40). A 30% (wt/vol) stock solution of SW (http://www.haloarchaea.com/resources/halohandbook/Halohandbook_2008_v7.pdf) contains 240 g NaCl, 30 g MgCl₂ · 6H₂O, 35 g MgSO₄ · 7H₂O, 7 g KCl, 5 mL of 1 M CaCl₂ · 2H₂O, and 80 mL of 1 M Tris·HCl (pH 7.2) per liter of water. Virus stocks were prepared as previously described (40), except that semiconfluent top-layer agar plates were incubated for 5 d.

For viral infection studies, cells of middle and late exponential cultures of H. sinaiiensis were infected using a multiplicity of infection of 15, incubated 30 min at 37 °C, and then grown aerobically at 37 °C. Turbidity (A₅₅₀) was followed, and after cell lysis, the number of viruses in the growth medium was determined using a plaque assay.

HSTV-1 particles were precipitated from the virus stocks using 10% (wt/vol) polyethylene glycol 6000 and purified as previously described for Halorubrum sodomense tailed virus 2 (37). The density of HSTV-1 in CsCl was determined as previously reported (37). To determine the virus stability at low salinity, the virus pellets were treated as previously described (37).

Genome Sequencing and Protein Analysis.

Purified viral genomic DNA was sequenced by the Pittsburgh Bacteriophage Institute at the University of Pittsburgh’s Genomics and Proteomics Core Laboratories to a depth of 40-fold coverage using 454 technologies. Raw reads were assembled using GS De Novo Assembler (Roche version 1.1); assemblies were then quality controlled using Consed, version 16 (41). Finished sequences were analyzed and annotated in genome editors, including DNAMaster (http://cobamide2.bio.pitt.edu), Glimmer (42), GeneMark (43), tRNA ScanSE (44), and Aragorn (45), to identify genome features. The nucleotide sequence has been deposited in GenBank under accession number KC117378.

Protein concentrations were determined using the Coomassie blue method (46) and bovine serum albumin as a standard. The purified virions were analyzed by modified tricine-SDS/PAGE with 4% (wt/vol) and 14% (wt/vol) acrylamide concentration in the stacking and separation gels, respectively (47). Protein N-terminal sequences were determined as previously described (40) in the Protein Chemistry Core Facility of the Institute of Biotechnology, University of Helsinki.

Cryo-EM and Image Processing.

Freshly purified virus samples were applied on Quantifoil R 2/2 grids, vitrified in liquid ethane as described previously (48), and transferred to a GATAN 626 cryo-holder for observation in an FEI Tecnai F20 microscope operating at 200 keV at liquid nitrogen temperature. Images from the virus particles in 20 mM Tris·HCl (pH 7.2) were collected under low-dose conditions on Kodak SO163 film at 62,000× magnification, with a range of underfocii of 0.6–3.2 µm. Images from the virus particles in 9% (wt/vol) SW or 20 mM Tris·Cl (pH 7.2) were collected on a GATAN Ultrascan 4000 charge-coupled device camera at 68,000× magnification at 4.0- or 6.0-µm underfocus.

The negatives were scanned using a PhotoScan TD scanner (Z/I Imaging) with 7-μm sampling and 12-bit grayscale readout. The digitized micrographs were processed, and viral particles were picked as previously described (49). The initial models for the viruses at approximately 30-Å resolution were generated using a random model computation method with a subset of 150 far from focus particles (50). These models were then used as starting maps for iterative full orientation searches and origin determinations of all of the particles using AUTO3DEM (51) including periodic recentering of the particles using RobEM. Full contrast transfer function correction was performed in AUTO3DEM. The resolution was determined using a Fourier shell correlation cutoff of 0.5 (31). A total of 7,115 particles from 164 micrographs were included in the final reconstruction.

The final reconstruction was sharpened using the program EM-Bfactor as previously described (52, 53). The B-factor applied was 450.17. The handedness was determined using tilt experiments as described previously, with HK97 proheads as a reference (54, 55). Visualization was done using University of California, San Francisco (UCSF) Chimera (56). The empty capsid was generated erasing most of the genome density using the “Volume Eraser” tool in UCSF Chimera, and the capsid volume was measured by generating an “Icosahedron Surface” inside the empty capsid and measuring the volume of this surface using the “Measure Volume and Area” tool (56).

An asymmetric unit from the HK97 Head II atomic model [Protein Data Bank (PDB) ID 1OHG] (32) was fitted into the EM density using the “fit-in-map” tool of UCSF Chimera. One density corresponding to each of the hexamer subunits was then written out using the “Zone” tool to a radius of 6 Å around the fitted HK97 molecule. The resulting densities were overlaid using the “Fit in map” tool and combined into an averaged density using the “Vop” command. UCSF Chimera’s “volume eraser” tool was used to delete outlier density (56). Finally a refined model of the asymmetric unit was constructed by replacing each hexameric subunit by the averaged density. Independently the structure of HSTV-1 gp14 was predicted using the homology modeling tool I-TASSER (33, 34). The top 10 template structures that were used by the program were ε15 major capsid protein GP7 (PDB ID 3C5B) (57), bacteriophage HK97 Head II (PDB ID 1OHG, PDB ID 1FH6) (10, 32), bacteriophage HK97 K169Y Head I (PDB ID 2FS3) (58), and bacteriophage HK97 Prohead I (PDB ID 3P8Q) (59). The best I-TASSER model was fitted in to the average hexamer density using the “Fit-in-map” tool of UCSF Chimera (56).

The reconstruction has been deposited in the Electron Microscopy Data Bank, accession code EMD-2279.