A filamentous archaeal virus is enveloped inside the cell and released through pyramidal portals

Significance

Egress of most eukaryotic enveloped viruses, including such human pathogens as HIV-1, Ebola, and coronaviruses, occurs via budding through cellular membranes, a process concomitant with virion assembly. Archaea are also infected by enveloped viruses, but how their virions are assembled and released from the cells remained largely unknown. We show that virions of Sulfolobus islandicus filamentous virus (SIFV) are assembled and enveloped in the cell cytoplasm. Instead of budding, SIFV induces the formation of pyramidal structures, which penetrate the cell envelope and serve as portals for virion release. Comparison of the infection cycles of evolutionarily related enveloped and nonenveloped filamentous archaeal viruses suggests that the primary role of the lipothrixvirus membrane is to protect the genome against extreme environmental conditions.

Abstract

The majority of viruses infecting hyperthermophilic archaea display unique virion architectures and are evolutionarily unrelated to viruses of bacteria and eukaryotes. The lack of relationships to other known viruses suggests that the mechanisms of virus–host interaction in Archaea are also likely to be distinct. To gain insights into archaeal virus–host interactions, we studied the life cycle of the enveloped, ∼2-μm-long Sulfolobus islandicus filamentous virus (SIFV), a member of the family Lipothrixviridae infecting a hyperthermophilic and acidophilic archaeon Saccharolobus islandicus LAL14/1. Using dual-axis electron tomography and convolutional neural network analysis, we characterize the life cycle of SIFV and show that the virions, which are nearly two times longer than the host cell diameter, are assembled in the cell cytoplasm, forming twisted virion bundles organized on a nonperfect hexagonal lattice. Remarkably, our results indicate that envelopment of the helical nucleocapsids takes place inside the cell rather than by budding as in the case of most other known enveloped viruses. The mature virions are released from the cell through large (up to 220 nm in diameter), six-sided pyramidal portals, which are built from multiple copies of a single 89-amino-acid-long viral protein gp43. The overexpression of this protein in Escherichia coli leads to pyramid formation in the bacterial membrane. Collectively, our results provide insights into the assembly and release of enveloped filamentous viruses and illuminate the evolution of virus–host interactions in Archaea.

Hyperthermophilic archaeal viruses are among the most enigmatic members of the virosphere, with many of them displaying unique virion architectures and genomic contents (1–5). The understanding on virus–host interactions in Archaea remains scarce when compared to bacterial or eukaryotic viruses. However, recent studies have provided first insights into different steps of the infection cycle for several model archaeal viruses, showing that some of the mechanisms used by archaeal viruses to interact with the hosts are similar to those of eukaryotic and/or bacterial viruses, whereas others are unique (6–13).

Two major strategies of virion assembly and release have been described for hyperthermophilic archaeal viruses (14). One strategy is exemplified by the Sulfolobus spindle-shaped virus 1 (SSV1), the prototypic member of the Fuselloviridae family, whereby virion assembly is concomitant with its release via budding through the host cell envelope, resembling the release of many eukaryotic enveloped viruses, such as HIV-1 and influenza (7). This release strategy typically does not result in the lysis of the infected cell and, following the eukaryotic virus paradigm, is expected to be common to other enveloped archaeal viruses. By contrast, viruses which assemble virions intracellularly employ egress strategy involving the disruption and death of the host cell. Archaeal viruses have evolved a unique cell lysis mechanism based on the formation of large pyramidal structures, dubbed virus-associated pyramids (VAPs), on the host cell surface (15). The VAPs protrude through the surface protein (S-) layer, the only component of the archaeal cell envelope besides the cytoplasmic membrane (16), and at the end of the infection cycle, the triangular facets of the VAP come apart as flower petals, producing apertures through which the mature virions exit the host cell (6, 17–19). Thus far, the VAP-based egress mechanism has been shown to be used by viruses belonging to three unrelated families, namely, Rudiviridae, Turriviridae, and Ovaliviridae, all infecting hyperthermophilic and acidophilic archaea of the order Sulfolobales. The VAPs formed by nonenveloped rod-shaped rudiviruses and icosahedral turriviruses are seven sided (i.e., the VAP has seven triangular facets) (17, 18, 20) and are built from homologous proteins, which, in all likelihood, have been exchanged between viruses from the two families by horizontal gene transfer (21, 22). By contrast, the VAPs built by the ovalivirus SEV1 are six sided, but the protein responsible for the VAP formation has not been identified (23). Notably, similar six-sided pyramids have been also observed on the surface of hyperthermophilic neutrophiles of the order Thermoproteales (24, 25), suggesting that VAP-based egress strategy is widespread among hyperthermophilic archaeal viruses.

Filamentous viruses of the family Lipothrixviridae are among the most broadly distributed archaeal viruses, with representatives being isolated from hot springs in Iceland, Italy, Russia, the United States, and Japan (24, 26–31). Lipothrixviruses have linear double-stranded (ds) DNA genomes, and based on genomic similarities, are divided into four genera, Alphalipothrixvirus, Betalipothrixvirus, Gammalipothrixvirus, and Deltalipothrixvirus. Structural studies have shown that all lipothrixviruses share the same virion organization; namely, linear dsDNA is complexed and condensed by two paralogous major capsid proteins (MCPs) into a helical nucleocapsid, which is further enveloped with a lipid membrane (30, 32, 33). Both ends of the virion are capped with terminal structures responsible for host recognition and binding (10, 27, 28). Similar to several other hyperthermophilic archaeal viruses (34–36), the dsDNA in the nucleocapsid of lipothrixviruses is stored in the A-form (30, 32, 33), which is believed to be one of the adaptations to high temperature environments. Structural studies have shown that lipothrixviruses are evolutionarily related to archaeal viruses of the families Rudiviridae and Tristromaviridae but distinct from all other known viruses. Accordingly, the three families have been recently unified into a realm Adnaviria (37).

Remarkably, the lipid envelope surrounding the nucleocapsid of lipothrixviruses is two times thinner than the cytoplasmic membrane of the host cell. It has been shown that gammalipothrixvirus AFV1 selectively recruits from the host those tetraether lipid species, which can be bent into a U-shaped horseshoe conformation, and molecular dynamics simulation has further suggested that these lipids form a thin monolayer membrane around the nucleocapsid (32). By contrast, the envelope of alphalipothrixvirus SFV1 is strongly enriched in archaeol, a short lipid molecule corresponding to ∼1% of lipids in the host membrane (30). However, whether the viral envelope is acquired during the budding process, as in the case of the majority of other enveloped viruses (38), remains unknown. Notably, previous studies have suggested that lipothrixviruses are released without causing host cell lysis (26–28), which would be consistent with the budding process, but the exact mechanism has not been investigated.

Here, we characterize the assembly and release of Sulfolobus islandicus filamentous virus (SIFV), the type member of the Betalipothrixvirus genus. The SIFV virions are enveloped, flexible, filamentous particles measuring ∼2 µm in length and 24 nm in width (Fig. 1A). At each end of the filament, the SIFV virions are decorated with terminal mop-like structures, which are thought to play a role in host recognition (26–28). Using dual-axis electron tomography, we show that the ∼2-μm-long SIFV virions are assembled in the cytoplasm of the infected cells, which have a diameter of 1 to 1.2 μm, and are released at the end of the infection cycle through six-sided VAPs. The 89-amino-acid (aa)-long SIFV protein gp43 is sufficient for VAP formation and its heterologous overexpression in Escherichia coli leads to formation of similar structures in the bacterial membrane. The VAP protein is conserved in all members of the Betalipothrixvirus and Deltalipothrixvirus genera but is unrelated to any of the previously characterized VAP proteins from other viruses. Unexpectedly, our results show that, differently from other characterized enveloped viruses, SIFV nucleocapsids are enveloped with a lipid membrane inside the host cell.

Fig. 1.

Characterization of the SIFV infection cycle in S. islandicus LAL14/1 cells. (A) Electron micrograph of purified SIFV particles negatively stained with 2% uranyl acetate. (Scale bar, 500 nm.) (B) Electron micrograph of S. islandicus LAL14/1 cells infected with SIFV. Sample was collected 2 min postinfection and negatively stained with 2% uranyl acetate. Arrows indicate the termini of one selected virion. (Scale bar, 500 nm.) (C) One-step growth curve (black) and adsorption kinetics (gray) of SIFV using as host S. islandicus LAL14/1. For the one-step growth curve and adsorption assay, the cells were infected with an MOI of 0.01 and 0.05, respectively, and the number of extracellular virions was estimated as described in Materials and Methods. (D) Optical density (OD) of S. islandicus LAL14/1 liquid cultures infected with SIFV using MOIs ranging from 0.01 to 10. (E) Number of viable cells of infected S. islandicus LAL14/1 liquid cultures at different MOIs (0.01 to 10). CFU, colony-forming units.

Results

SIFV Infection Cycle.

To obtain insights into the life cycle of liprothixviruses, we focused on SIFV (26) and its host Saccharolobus (formerly Sulfolobus) islandicus LAL14/1 (39). The S. islandicus LAL14/1 cells display an irregular coccoid morphology typical of Saccharolobus species with a diameter of ∼1 µm (40). Thus, SIFV virions are twice as long as the diameter of the host cell (Fig. 1B), indicating that strategies to overcome intracellular space limitation should be in place for efficient virion morphogenesis. We first determined the length of the infection cycle by performing a one-step growth experiment using a multiplicity of infection (MOI) of 0.01. A sharp increase in the extracellular virus titer at 11 h postinfection (hpi) signified the length of the latent period (Fig. 1C). With 26 ± 7 virions produced per cell, the burst size is comparable to that determined for the filamentous nonenveloped rudivirus SIRV2 (6). The adsorption assay showed that SIFV binding to the host cells is highly efficient, with nearly 70% of the virions being attached to the host cells within the first 2 min postinfection (Fig. 1C), ensuring nearly synchronous infection of the S. islandicus population. The adsorption rate constant calculated at 2 min postinfection was 5.8 × 10⁻⁹ mL min⁻¹, which is similar to those reported for the turrivirus STIV, rudivirus SIRV2, and bicaudavirus SMV1 (8, 9, 12).

Infection of S. islandicus cells using an MOI as low as 0.01 resulted in growth retardation of the culture (Fig. 1D), whereas at MOIs >1 there was a significant decrease in the number of colony forming units (Fig. 1E). The effects were more pronounced upon infection with higher MOIs, suggesting that SIFV infection leads to cell death in an MOI-dependent manner. The impact of SIFV on the growth dynamics and viability of the infected cells is reminiscent of those reported for the rod-shaped lytic virus SIRV2 (6), which is structurally and evolutionarily related to SIFV but is not enveloped (33, 36). Notably, SIRV2 causes massive degradation of the host chromosome upon infection (6). To determine whether this is also the case during SIFV infection, the intracellular DNA content of noninfected and SIFV-infected cultures was monitored over time by flow cytometry. Unlike for SIRV2, there was no host DNA degradation in the case of SIFV-infected cells (SI Appendix, Fig. S1). Instead, we noted that upon SIFV infection the fraction of cells with a single chromosome copy, especially after 6 hpi, was diminished. The fact that the majority of infected cells contained two chromosome copies suggests that SIFV infection affects cell-cycle control and/or cell division (SI Appendix, Fig. S1), consistent with the growth retardation of the infected culture (Fig. 1D).

Envelopment of SIFV Virions Occurs in the Host Cytoplasm.

To gain insights into SIFV virion assembly and envelopment, the infected cells were analyzed using dual-axis electron tomography at 10 and 12 hpi. The reconstructed tomographic volumes were analyzed using convolutional neural networks (CNN) (41) to annotate virions, envelopes, ribosomes, and S-layer. At 12 hpi, bundles of filamentous particles resembling SIFV virions were observed (Fig. 2 A–F), whereas at 10 hpi similar structures were barely detectable (SI Appendix, Fig. S2 A and B). Reconstructed electron tomograms showed that virion-like particles are organized into bundles, which were bent to follow the membrane plane (Fig. 2 A–C), explaining how the long SIFV virions are spatially accommodated within the host cells. In cross-sections of the infected cells, we could trace up to 70% of the total virion length (i.e., 1.4 µm out of 2 µm), with the remaining portion of the virion bundles being invisible, likely due further bending out of the visible plane. Cross-sections showed the presence of 86 ± 15 virions per infected cell (n = 7), which is about three times higher than the burst size estimated using the plaque test, suggesting that only one-third of the assembled virions are infectious (i.e., able to form plaques). The CNN analysis revealed numerous ribosomes, which were distributed evenly in the infected cells (Fig. 2), although occasionally, ribosomes were ordered along the viral particles (SI Appendix, Fig. S3), suggesting either their active role during virion morphogenesis or steric exclusion by the forming virion bundles.

Fig. 2.

Assembly of SIFV virions in the cytoplasm of the host cell. (A) A slice through a reconstructed tomogram of a sectioned sample of SIFV-infected cells at 12 hpi. (B and C) A segmented and surface-rendering displays of the tomogram in A, including various viral and cellular components: S-layer (dark salmon), membranes (green), nucleoprotein cores (blue), and ribosomes (magenta). (D) A slice through a reconstructed tomogram of a sectioned sample of SIFV-infected cells at 12 hpi, displaying a transversal view of the virions assembled in the host cytoplasm. (E and F) A segmented and surface-rendering displays of the tomogram in D: S-layer (dark salmon), pyramid in a closed conformation (yellow), membranes (green), nucleoprotein cores (blue), and ribosomes (light magenta). (G and H) Linear density profiles of four nonenveloped (G) and enveloped (H) nucleocapsids located adjacent to each other. C, nucleoprotein core, M, membrane. (I) Measurement of the diameter (nm) of the nucleoprotein cores of enveloped and nonenveloped virions. (J) Measurement of the distances (nm) between contiguous virions in clusters of enveloped and nonenveloped nucleocapsids. The distance was measured between the centers of adjacent nucleoprotein cores. (K) Distribution of lipid species identified in S. islandicus LAL14/1 cells and highly purified SIFV virions. (Scale bars: A–C, 100 nm; D–F, 50 nm.)

Bundle cross-sections showed that filamentous virions are present in two forms (Fig. 2 A and D and SI Appendix, Fig. S2 B and C): 1) nucleocapsids surrounded by lower density rings that presumably represent the viral envelopes, and 2) nucleocapsids devoid of any visible envelope (Fig. 2 D–F and SI Appendix, Fig. S2C). Linear density profiles measured across four nonenveloped and enveloped nucleocapsids located adjacent to each other showed that the pixel intensities were the same in the region corresponding to the nucleocapsid, whereas profiles of enveloped virions had additional intensities at each side of the cross-section profile, representing the viral envelope (Fig. 2 G and H). Consistent with this interpretation, no significant difference was found between the diameters of the enveloped and nonenveloped nucleocapsid cores (Fig. 2I). However, as expected, the center-to-center distances measured between the enveloped nucleocapsids were significantly larger than those between the nonenveloped nucleocapsids, consistent with the additional spacing contributed by the envelopes (Fig. 2J).

To further characterize the virion bundles inside the cells, we performed three-dimensional (3D) reconstruction of the tomographic data (Fig. 3A). Top and lateral views of the 3D models showed that the virion bundles are organized on a nonperfect hexagonal lattice (Fig. 3 B and E) and are slightly twisted (Fig. 3 B–D). In particular, virions in the periphery of the bundle twist around the virions in the center at an 8° angle (Fig. 3 C and D). Notably, in certain tomograms, the enveloped and nonenveloped virions were parallel to each other and seemingly belonged to the same virion bundle (Fig. 2 A–C). Collectively, our findings indicate that virion assembly and maturation take place in the cytoplasm of infected cells and proceed through the initial formation of naked nucleocapsids, which undergo sequential envelopment. Intriguingly, this means that envelopment itself occurs in the cytoplasm rather than by extrusion of the naked nucleocapsids through the cellular membrane.

Fig. 3.

SIFV virions organized into twisted filament bundles. (A) A slice through a reconstructed tomogram of a cluster of enveloped virions observed at 12 hpi. (B) Top view of the array of enveloped virions. Virions located at the center (dark blue) and periphery (purple) of the bundle were colored to mark them as positional references. (C) Cross-section through the middle of the array displayed in B brought to a vertical orientation. (D) Visualization of the total array displayed in B. Virions located at the periphery were used to calculate the twist angle θ (θ = 8°). (E) Black lines trace the nonperfect hexagonal lattice on which the SIFV virions are organized within the bundle shown in B.

Characterization of the SIFV Envelope.

Recent structural characterization of the SIFV virions has revealed that the envelope surrounding the nucleocapsid is twice as thin as the cytoplasmic membrane of the host, as observed for lipothrixviruses from other genera (30, 32, 33). To determine and compare the lipid compositions of the viral and cellular membranes, we performed liquid chromatography with time of flight mass spectrometry (LC-ToFMS) analysis of the mature SIFV virions and S. islandicus LAL14/1 cells. The lipid composition of the viral envelope was found to be quantitatively very different from that of the host membrane (Fig. 2K). The S. islandicus LAL14/1 membrane nearly exclusively contains glycerol dibiphytanyl glycerol tetraether (GDGT) lipid species, long molecules spanning the entire thickness of the membrane, which is effectively a monolayer of GDGT lipids (42). The dominant lipid species (∼60% of all lipids) in the host membrane is GDGT-4 carrying four cyclopentane rings (SI Appendix, Fig. S4). By contrast, the envelope of SIFV is strongly enriched in C₂₀ sn-2,3-glycerol diphytanyl ether lipid, known as archaeol (SI Appendix, Fig. S4), and C₄₀ glycerol trialkyl glycerol tetraether GTGT-0, a tetraether lipid with one C40 biphytanyl and two C20 phytanyl moieties. These two lipid species together account for less than 1% in the host membrane but reach over 40% of lipids in the viral envelope (Fig. 2K). Thus, similar to some other archaeal viruses (30, 32, 43, 44), the lipids are incorporated into the SIFV envelope in a highly selective manner, in line with the observation that the envelope is not acquired through the budding process.

SIFV Is Released from the Cell through Hexagonal VAPs.

The intracellular envelopment of SIFV virions raises questions regarding the mechanism of their egress from the host cell. Thus, to better understand this last stage of the SIFV life cycle, infected cells were observed by transmission electron microscopy (TEM) at different time points after infection. The TEM analysis at 12 and 24 hpi revealed the presence of six-sided (hexagonal) apertures on the cell surface (Fig. 4A). No such structures were observed on the surface of noninfected cells. Scanning electron microscopy (SEM) confirmed the presence of perforations in the envelope of infected cells at 12 hpi (Fig. 4B). The hexagonal apertures closely resemble the opened VAPs previously observed in Pyrobaculum oguniense cells (25). Thin section TEM imaging of infected cells revealed the presence of pyramidal structures in SIFV-infected cells (Fig. 4C), similar to those previously described for lytic viruses of the families Rudiviridae, Turriviridae, and Ovaliviridae (6, 17, 23).

Fig. 4.

Visualization of infected S. islandicus LAL14/1 cells by TEM and SEM. (A and B) Infected S. islandicus LAL/14 cells collected at 12 hpi, negatively stained with 2% uranyl acetate, and visualized under TEM. (C and D) Infected S. islandicus LAL/14 cells collected at 12 hpi and visualized by SEM. (E and F) Thin sections (70 nm) of infected S. islandicus LAL/14 cells collected at 10 hpi and visualized by TEM. Arrows indicate VAPs at different stages. (Scale bars: A, 200 nm; B, 100 nm; C, 200 nm; D, 200 nm; E, 200 nm; and F, 100 nm.)

To model the 3D shape of the SIFV-induced VAPs, we used dual-axis electron tomography. The VAPs displayed considerable variation in size: the height (measured from the base to the tip of the VAP) ranged from ∼38 to 124 nm, and the diameter (measured between the opposite sides of the VAP base) varied from ∼48 to 220 nm (n = 22). The SIFV VAPs grow outwards from the cell membrane, penetrating and disrupting the S-layer of the host cell (Fig. 5A and SI Appendix, Fig. S5). At 12 hpi, when the virions are being released, open VAPs were also detected; VAP opening leads to the loss of intracellular content, including the virions, yielding empty “ghost cells” (SI Appendix, Fig. S5). Occasionally, SIFV VAPs were associated with dense spherical bodies (Fig. 5A), likely representing storage granules (45), which were also observed in the case of SIRV2- and STIV-infected cells (18, 20). The relevance of these structures for VAP formation and/or virus release remains unknown.

Fig. 5.

Visualization of the pyramidal structures formed in the SIFV-infected Saccharolobus cells and E. coli. (A) XY slice through a reconstructed tomogram showing the closed conformation of an SIFV VAP. (Scale bar, 50 nm.) (B) Computed slice view of the same SIFV VAP aligned to the pyramidal base reveals its hexagonal shape. (Scale bar, 50 nm.) (C) Lateral view of a 3D map in solid representation of an SIFV VAP. (D) Bottom view of a 3D map in solid representation of an SIFV VAP. (E) Thin section electron micrographs of E. coli cells overexpressing SIFV gp43. I, inner membrane, P, periplasmic space, O, outer membrane. (Scale bars: 200 nm; in Insets, 100 nm.)

Computational reslicing of the tomographic volume clearly revealed that the VAPs have a hexagonal base (Fig. 5B), consistent with the six-sided apertures observed in the cell envelope by negative stain TEM (Fig. 4 A and B). A 3D model of an SIFV VAP in closed conformation was obtained by manual segmentation (Fig. 5 C and D). The reconstruction shows that SIFV pyramids are baseless hollow structures consisting of six triangular sides. Given the presence of VAPs of highly variable sizes, including small VAPs located beneath the S-layer, it is likely that VAP formation is nucleated by a hexameric assembly, which develops into the six-sided VAP by gradual growth of the triangular facets.

SIFV gp43 Is Sufficient for VAP Formation.

SIFV does not encode identifiable homologs of the previously reported VAP proteins P98 and C92 of rudivirus SIRV2 and turrivirus STIV, respectively (21, 22). Given that the VAPs of rudiviruses and turrivirus are seven sided (18, 20, 21), whereas those of SIFV are six sided, it is conceivable that the proteins forming the two types of VAPs might be unrelated. Thus, to identify the protein responsible for formation of the SIFV VAPs, the proteins enriched in the membrane fraction of SIFV-infected cells at 12 hpi were analyzed by sodium dodecyl sulfate polyacrylamide gel electrophoresis and liquid chromatography with tandem mass spectrometry (LC-MS/MS). Five protein bands (B1 to B5) appeared or grew in intensity in the membrane fraction of infected cells at 12 hpi, compared to earlier time points postinfection or the noninfected control (SI Appendix, Fig. S6). The upper bands B1 and B2 (with molecular masses of ∼20 and ∼24 kDa, respectively), also visible in the membrane fraction of infected cells at 10 hpi, were identified as the two MCPs of SIFV. The bands B3 (∼15 kDa), B4 (∼12 kDa), and B5 (∼9 kDa) were detected exclusively in the membrane fraction of infected cells at 12 hpi (SI Appendix, Fig. S6). Whereas B3 contained no identifiable virus proteins, LC-MS/MS analysis has shown that bands B4 and B5 contain several viral proteins of unknown function, namely, SIFV gp15, gp43, gp20, and gp71 (SI Appendix, Table S1).

The four proteins were analyzed for the presence of predicted N-terminal transmembrane domain (TMD), a feature found in other VAP proteins (21, 22). Only gp43 fulfilled this requirement. To investigate if SIFV gp43 is involved in VAP formation, the corresponding gene was cloned and expressed in E. coli Rosetta (DE3) pLys. Protein expression was confirmed by Western blot analysis with anti–6× His antibodies (SI Appendix, Fig. S7). Electron microscopy analysis of thin sections of gp43-expressing cells 4 h after induction showed the presence of multiple VAP-like structures on the cytoplasmic membrane of the bacterial cells protruding toward the periplasmic space (Fig. 5E). The VAPs were always found in the closed conformation, suggesting that the signal triggering the VAP opening is archaea specific and might require the presence of additional viral factors. Nevertheless, at 4 hpi, the optical density of induced cell culture was significantly lower compared to the noninduced control, suggesting that the protein expression and VAP formation are toxic to bacteria. Collectively, these results indicate that gp43 is the only essential structural component of VAPs.

The gp43 of SIFV is 89-aa-long and is the shortest VAP protein identified to date (SI Appendix, Fig. S8). To study the distribution of SIFV gp43 homologs, we performed Position-Specific Iterative Basic Local Alignment Search Tool searches against the viral nonredundant protein database at the National Center for Biotechnology Information. SIFV gp43 homologs were found to be conserved in all characterized members of the Betalipothrixvirus and Deltalipothrixvirus genera of the Lipothrixviridae family but have no identifiable homologs in viruses from other families. Thus, SIFV gp43-like proteins form a family of VAP proteins, distinct from that including other known VAP proteins from rudiviruses and turriviruses (21, 22). The two protein families display similar features, including the N-terminal TMDs and extensive α-helical content (SI Appendix, Fig. S8). However, the pattern of amino acid conservation is distinct in the two families, suggesting that the VAPs have evolved in archaea on at least two independent occasions and that these complex structures can be built from proteins with highly different sequences.

Discussion

In this study, we explored the assembly and egress mechanisms of SIFV, a representative of the family Lipothrixviridae. Structural studies have shown that lipothrixvirus virions consist of a helical nucleocapsid enveloped with a thin lipid membrane (30, 32, 33). The nucleocapsid of lipothrixviruses is homologous to the helical capsid of nonenveloped viruses of the Rudiviridae family (33, 37), indicating that viruses from the two families have evolved from a common ancestor. Based on phylogenomic and structural studies, it has been suggested that this ancestor was an enveloped virus, resembling lipothrixviruses, and that rudiviruses have emerged by shedding the lipid membrane (33). Given that in most viruses, virion envelope plays key roles during different stages of virus–host interaction, such as genome delivery or virion egress, functional comparison between rudiviruses and lipothrixviruses offers a unique opportunity to study the evolution of virus–host interactions.

In many ways, the SIFV infection process resembles that of rudiviruses. The SIFV infection cycle starts with rapid virion adsorption to the host cell surface. The high rate of adsorption, similar to that documented for other hyperthermophilic archaeal viruses (8, 9, 12, 46), is likely to be important for limiting the exposure of the viral particles to extreme environmental conditions. Notably, however, unlike many other hyperthermophilic archaeal viruses, which recognize their hosts through pili (12, 47–49), SIFV has been suggested to bind the receptor located directly within the cellular membrane (26). The latent period of SIFV is rather long (∼11 h), which is typical of many other archaeal viruses, and might signify the general adaptation of hyperthermophilic archaeal viruses to spending more time within, rather than outside of the cell. The burst size of SIFV is also similar to that of rudivirus SIRV2 (6). Notably, the particle-to-PFU (plaque-forming unit) ratio determined for SIFV is 3:1, meaning that only every third virion forms a plaque. A similar particle-to-PFU ratio, 5:1, was determined for fusellovirus SSV1 (43). More generally, particle-to-PFU ratios are virus-specific and, for eukaryotic viruses, vary from 1:1 to more than 1,000:1 (50). The discrepancy between the total number and infectious virions may be due to the presence of immature or defective virions, accumulation of detrimental mutations in the viral genome, or activity of antiviral defense systems, which abort a substantial fraction of virus infections.

Electron tomography analysis has provided insights into the intracellular assembly of the SIFV virions. The formation of SIFV nucleocapsids is highly reminiscent of the assembly of mature rudivirus virions (6, 18). In the case of both viruses, filamentous (nucleo)capsids are assembled in the cell interior forming bundles containing multiple virions. The SIFV virions in the bundles are arranged on a hexagonal lattice, resembling the tendency of many icosahedral virions to form crystalline-like arrays within the cell cytoplasm (51). The 3D reconstruction has shown that the bundles are twisted at an 8° angle. Although the biological relevance of the SIFV bundle twisting is unclear, a similar behavior has been characterized for many biological filaments and artificial materials, such as carbon nanotube ropes and micropatterned filament arrays (52). Interestingly, it has been concluded that the lowest energy state for a bundle of sufficiently flexible and long filaments is achieved when the bundle is twisted (52). Accordingly, twisting of the SIFV bundles might derive from the geometric frustration of the bulk virion packing and surface energy of noncontacting virions at the boundary of the bundle. In addition to twisting, the bundles undergo a pronounced bending to fit within the cell. Indeed, the bundles of ∼2-μm-long SIFV virions follow the inner outline of cytoplasmic membrane. Such bending is not observed in the case of rigid SIRV2 virions, which are ∼0.9-μm long and span nearly the entire width of the infected cell. It is tempting to speculate that virion rigidity and dimensions of the host cell limit the genome length of rudiviruses.

Whereas the nonenveloped capsids represent mature virions primed for egress in the case of rudiviruses, the SIFV nucleocapsids have to be further enveloped. Most of the studied enveloped viruses, including filamentous Ebola viruses, escape from their host cells by budding (38, 53, 54). Thus, budding is often considered to be the default mechanism of envelope acquisition in enveloped viruses. Indeed, virion morphogenesis and egress of the archaeal lemon-shaped fusellovirus SSV1 are concomitant and occur at the cellular cytoplasmic membrane via a mechanism highly reminiscent of the budding of enveloped eukaryotic viruses (7). Similarly, archaeal pleolipoviruses and bicaudaviruses have been proposed to use budding as an exit mechanism (13, 55). Hence, the finding that SIFV virions are enveloped inside the cell cytoplasm was unexpected. Using electron tomography, we observed both nonenveloped and enveloped SIFV virions within the same cell and sometimes as part of the same bundle (Fig. 2 A–C), suggesting an order of events from nonenveloped nucleocapsids, resembling mature rudivirus virions, to mature, enveloped SIFV virions (SI Appendix, Fig. S9 and Video S1).

In eukaryotes, some viruses acquire envelopes inside the cell by budding through organelles, such as endoplasmic reticulum, nuclear envelope, or Golgi complex (56–59). However, internal, membrane-bound compartments have never been observed in Saccharolobus or any other archaeal cells, rendering the possibility that SIFV virions are enveloped by budding through intracellular membranes highly unlikely. Recently, it has been shown that insect viruses of the Nudiviridae family are enveloped by a distinct mechanism inside the cell, involving extensive remodeling of the nuclear membrane (60). However, unlike in the case of SIFV, nucleocapsids of nudiviruses are enveloped simultaneously with the genome packaging. Consequently, the envelopment of SIFV might occur by a novel mechanism, involving either de novo membrane formation or trafficking of lipids from the cytoplasmic membrane to the virion assembly centers—neither mechanism has been demonstrated for other prokaryotic viruses. Additionally, differences in the composition and thickness of the viral and cellular membranes indicate that the incorporation of lipids into the viral membrane occurs in a highly selective manner. Similarly, selective lipid acquisition has been demonstrated for the lytic turrivirus STIV, which, unlike lipothrixviruses, has an internal membrane sandwiched between the icosahedral protein capsid and the dsDNA genome (61). The mechanism of the STIV membrane recruitment remains unclear but has been suggested to involve the archaeal ESCRT-based cell division machinery (62). Recently, it has been suggested that ovoid-shaped archaeal virus SEV1 also acquires its envelope intracellularly (23). Thus, the mechanism of membrane remodeling and envelopment employed by lipothrixviruses might be widespread among evolutionarily unrelated archaeal viruses.

TEM and SEM analyses showed that SIFV induces formation of VAPs on the surface of infected cells, which gradually grow in size (SI Appendix, Fig. S9 and Video S1). A similar mechanism of virion release has been described for archaeal viruses from families Rudiviridae, Turriviridae, and Ovaliviridae (6, 17, 23). Among these, SIFV VAPs more closely resemble VAPs formed by ovalivirus SEV1 (23), because VAPs of both viruses are six sided rather than seven sided as observed for rudiviruses and turriviruses (6, 17). Notably, whereas VAP proteins of rudiviruses and turriviruses share relatively high (55.4%) sequence identity (21, 22), gp43 of SIFV has no homologs in ovalivirus SEV1 and appears to be unrelated to the VAP proteins of rudiviruses and turriviruses (SI Appendix, Fig. S8). The protein responsible for VAP formation during the ovalivirus SEV1 infection remains unknown, but it is likely to represent yet another protein family. Furthermore, gp43 homologs could not be identified in lipothrixviruses of the Alphalipothrixvirus and Gammalipothrixvirus genera, suggesting that a considerable diversity of protein families capable of VAP formation remains to be discovered in the archaeal virosphere.

Our results show that expression of gp43 in E. coli leads to VAP formation in the bacterial membrane. It should be noted that bacterial and archaeal membranes consist of unrelated lipids; whereas bacterial membranes are bilayers containing phospholipids (fatty acids linked to glycerol moieties by ester linkages), the membrane of S. islandicus is largely a monolayer of tetraether lipids (long isoprenoid chains capped on both ends by glycerol moieties through ether linkages). Thus, the inherent ability of the two proteins, lacking any recognizable sequence similarity, to form VAPs in both bacterial and archaeal membranes is remarkable. Whether proteins from the two families have diverged from a common ancestor or have originated independently remains unclear. Regardless, the general replication cycle of enveloped lipothrixviruses and rudiviruses appears to be closely similar, involving formation of helical nucleocapsids, which are released through VAPs, suggesting that evolutionary transition from a postulated enveloped lipothrixvirus-like ancestor to the nonenveloped rudivirus-like ancestor did not entail any major adaptations in the mechanisms underlying the virus–host interactions. This finding raises the question regarding the function of the membrane in lipothrixviruses. We hypothesize that the primary role of the lipothrixvirus envelope is protection of the viral genome in a hot and acidic environment. Indeed, structural studies have shown that MCP packing in rigid rod-shaped rudiviruses is tighter than in flexible lipothrixviruses (33). Thus, once the ancestral MCP has evolved toward forming a more robust virus particle, impermeable to harmful extracellular solutes (36), the membrane layer might have become dispensable and was shed.

Materials and Methods

Propagation and Purification of Virus Particles.

Saccharolobus islandicus LAL14/1 and S. islandicus HVE10/4 (63, 64) were grown aerobically at 75 °C, pH 3.5 in rich medium containing 0.2% (weight/volume [wt/vol]) tryptone, 0.1% (wt/vol) sucrose, 0.1% (wt/vol) yeast extract, and mineral salt solution, as described previously (64).

TEM.

For negative-staining TEM analysis, 5 µL of the samples were applied to carbon‐coated copper grids, negatively stained with 2% uranyl acetate (wt/vol) and imaged with the transmission electron microscope FEI Spirit Tecnai Biotwin operated at 120 kV.

Lipid Analysis.

Lipids were analyzed by LC-ToFMS at the Royal Netherlands Institute for Sea Research using an Agilent 1290 Infinity II ultrahigh-performance LC coupled to a 6230 Agilent TOF MS as described by Besseling et al. (65).

Heterologous Expression of SIFV gp43.

The SIFV ORF43 was amplified from a pure SIFV stock and cloned into the pETM-11 plasmid. The vector contains an isopropyl β-D-1-thiogalactopyranoside–inducible promoter that was used for the expression of the His-tagged protein. E. coli Rosetta (DE3) pLys (Novagen, Merck) cells were transformed with the construct; liquid cultures were grown in 2YS medium and induced with 1 mM IPTG at OD₆₀₀ of 0.4 to 0.6 for 4 h. The noninduced cell culture was used as a control. Thin sections of E. coli cells were prepared as described in SI Appendix, SI Methods. The expression of gp43 was detected by Western blot (SI Appendix, SI Methods)

Sample Preparation for Electron Tomography.

Cultures at 10 and 12 hpi were pelleted by low-speed centrifugation and resuspended in a minimal volume of rich medium. Samples were taken up into cellulose capillary tubes (Engineering Office M. Wohlwend GmbH) as described previously (66), transferred into the 0.2-mm cavity of a type B sample holder filled with hexadecen, and frozen with a high-pressure freezing machine (HPM100, Leica). The samples were subsequently freeze substituted with 1% OsO4 in acetone according to the following schedule: −90 °C for 24 h, 5 °C/h for 12 h, −30 °C for 12 h, 10 °C/h for 3 h, and 0 °C for 1 h in a Leica AFS2 (Leica Microsystems). Cells were rinsed at room temperature within acetone and slowly infiltrated with low viscosity resin (Electron Microscopy Sciences). After heat polymerization, embedded cells were cut into 70-nm thin sections with an Ultracut R microtome (Leica) and collected on Formvar-coated copper grids. Thin sections (70 nm) were poststained with 4% uranyl acetate for 45 min and Reynold’s lead citrate staining for 5 min. Samples were imaged with the transmission electron microscope FEI Spirit Tecnai Biotwin operated at 120 kV. For electron tomography, embedded cells were cut into 200-nm-thick sections with an Ultracut R microtome (Leica) and collected on Formvar-coated copper grids. Protein A-gold particles of 10 nm were added on both sides of the sections and stained with 4% uranyl acetate (wt/vol) and Reynold’s lead citrate.

Dual-Axis Electron Tomography.

Grids were loaded on a dual-axis tomography holder (Model 2040, Fischione) and observed with a TECNAI F20 Transmission Electron Microscope (FEI) operating at 200 kV and equipped with a 4,000 × 4,000 charge-coupled device camera (Ultrascan 4000, Gatan). Micrographs, tilt series, and maps, in low and middle magnifications, were acquired using SerialEM (67, 68). After identifying areas of interest on middle magnification maps, the areas were baked using a total dose of 1,500 e⁻/A². The continuous tilt scheme was used for the automatic acquisition of micrographs every 1° over a ±55° range at higher magnification (usually 29,000 or 50,000). After the acquisition of tilt series in all areas of interest, the grid was manually rotated by 90° to acquire the second orthogonal tilt axis series in the same areas of interest.

Initial image shifts of the tilt series were estimated using the function tiltxcorr from the IMOD software package (69). Alignments were further optimized in IMOD using the tracing of gold fiducials across the tilt series. 3D reconstructions were calculated in IMOD by weighted back projection using the SIRT-like radial filter to enhance contrast and facilitate subsequent segmentation analysis. The volumes from the two tilt axes were combined to one using fiducials present in IMOD (70).

Data Availability

All study data are included in the article and/or supporting information.