The study of viruses infecting archaea is a rapidly evolving field, with new viruses being isolated at an increasing frequency1. Approximately 50 archaeal viruses have been described so far (see Universal Virus Database of the International Committee on Taxonomy of Viruses), with approximately equal numbers of viruses isolated from the Euryarchaeota phyla (principally methanogens and halophiles) and from the Crenarchaeota phyla (mainly those from thermophilic archaea). So far, no viruses have been isolated from the Korarchaeota or the Nanoarchaeota, two recently proposed phyla of the Archaea. Although 14 of the thermophilic archaeal viruses2,3,4,5,6,7,8,9,10,11,12 have been isolated within the past 5 years, when compared with the approximately 5,100 viruses known to infect bacterial and eukaryotic hosts, the archaeal viruses are vastly under represented.

The study of archaeal viruses provides tools for the elucidation of the biochemistry, evolution and ecology of both the viruses and their hosts. As this field is relatively new, there is a paucity of information about these viruses, and there are several possible avenues of exploration, including the determination of the total diversity of archaeal viruses, the relationship among archaeal viruses and their relationship to viruses from other domains, and the fundamentals of archaeal virus replication cycles. There are several excellent reviews describing the diversity of viruses infecting archaea isolated from both thermal1,13,14,15,16 and non-thermal17 environments, so detailed descriptions of these viruses will not be a focus of this review. Instead, this review aims to summarize the overall field of thermophilic crenarchaeal virology through comparisons of virus isolates, analysis of structural and genetic features and integration of environmental studies.

Hot spring environments and archaeal hosts

Hot, acidic springs where temperatures are greater than 75°C and the pH is less than 4.0 are typically dominated by members of the Crenarchaeota. The dominance of archaea in these environments has been shown by culture-independent PCR amplification of 16S rDNA using universal 16S rDNA primers18 (M.Y., unpublished observations) and quantitative PCR studies19. This is in contrast to 16S rDNA surveys of hot, neutral or near-neutral springs, where Bacteria seem to dominate the microbial communities20. We do not understand the basis of this environmental partition between thermophilic archaea and bacteria, but it is probably a consequence of specific physiological adaptations to a combination of both high temperature and low pH.

A wide diversity of crenarchaeotes have been detected and cultured from terrestrial hot-spring environments18,20, including members of the genera Sulfolobus21, Acidianus22, Vulcanisaeta23, Caldococcus24, Stygiolobus25, Metallosphaera26, Desulfurococcus27, Sulfophobococcus28, Thermocladium29, Caldivirga30, Pyrobaculum31, Thermofilum32, Thermoproteus33 and Caldisphaera34. Phylogenetic studies of three hot springs in Yellowstone National Park (YNP) detected 16S rDNA sequences with similarity to nine cultured genera (Acidianus, Desulfurococcus, Caldococcus, Vulcanisaeta, Sulfolobus, Metallosphaera, Pyrodictium, Stygiolobus and Thermocladium) and to unknown clones from other thermal areas35. Epifluorescence microscopy of YNP acidic hot springs reveals that cell densities vary between 105–106 cells ml−1, which is similar to abundances found in neutral hot springs36. It is generally believed that, as an environment becomes more extreme (for example, hot and acidic), species richness will decrease, however final validation of this assumption awaits complete metacommunity analysis between comparable extreme and non-extreme environments.

The prevalence of crenarchaeal species in hot, acidic environments has contributed to their reputation as obligate extremophiles. This concept has been upended by recent molecular phylogenetic surveys showing that crenarchaeotes are diverse, abundant and ubiquitous in various moderate to low temperature habitats (−1.5°C to 37°C) (Refs 37–40). For instance, quantitative filter hybridization41 and real-time PCR quantification42 have estimated that crenarchaeotes represent approximately 0.5–3.0% of the total microbial community in mesophilic soils and a substantially larger portion, 10–30%, of the prokaryotic biomass in coastal marine waters43 and polar seas44.

Crenarchaeal viruses

Our current knowledge of thermal viruses comes from cultured isolates of archaeal hosts. Most of the 21 thermophilic archaeal viruses isolated so far have originated from cultures established from terrestrial hot springs, and infect members of the Crenarchaeota, except for one thermal archaeal virus that was isolated from a thermophilic marine euryarchaeote45.

Viruses that infect only four genera of crenarchaeal hosts (Acidianus, Sulfolobus, Thermoproteus and Pyrobaculum) have been isolated, once again reflecting our limited knowledge of the total diversity of archaeal viruses present in the environment.

Isolated crenarchaeal viruses have been classified into five approved families (Fuselloviridae, Rudiviridae, Lipothrixviridae, Guttaviridae and Globuloviridae) and two proposed families (Ampullaviridae and Bicaudaviridae) at this time, although two other families will probably be proposed soon1. The families are based on morphology of virus particles and genome similarities. The number of isolates in each family range from one (Guttaviridae, Ampullaviridae and Bicaudaviridae) to six in the Lipothrixviridae, which is divided into four subfamilies. Currently, only the Lipothrixviridae include viruses that infect both acidophilic thermophiles (Sulfolobus spp. and Acidianus spp.) and thermophiles from more neutral environments (Thermoproteus tenax). The Globuloviridae includes viruses that infect T . tenax and Pyrobaculum spp., both of which occur at a more neutral pH. The remaining families include viruses that infect only Sulfolobus spp. (Fuselloviridae and Guttaviridae), Acidianus spp. (Ampullaviridae and Bicaudaviridae), or both (Rudiviridae).

One of the most striking characteristics of crenarchaeal viruses is that they have unusual and diverse particle morphologies (Fig. 1). This is in contrast to the prevalence of the classic head and tail morphology observed in many of the characterized viruses that infect the Bacteria and Euryarchaeota46. So far, the head and tail morphology has not been detected in viruses that infect the Crenarchaeota. As described below, it is becoming evident that some virus morphologies are unique to the Archaea whereas others span all three domains of life.

Figure 1: Typical virus morphologies observed in both environmental samples and enrichment cultures established from hot, acidic environments.
figure 1

a | Spindle-shaped virus morphology is restricted to the Archaea. b | Rod and filamentous morphologies are common in both bacteria and archaea. c | Icosahedral viruses are common in all three domains of life. Scale bars denote 100 nm.

At least 12 different virus morphologies have been observed in enrichment cultures established from YNP hot springs that range in temperature from 70°C to 93°C and pH values from 1.5 to 4.5 (Refs 47,48). Many of these morphologies are found in both Sulfolobus and Acidianus spp. enrichment cultures, and enrichment cultures from Pozzuoli, Italy3. Although culture-based methods will undoubtedly remain central to the identification and characterization of thermal viruses, they are not without limitations. Culture biases of both the archaeal hosts and their viruses have been observed49, and are of particular concern in acidic thermal environments given our limited understanding of the metabolic capabilities of hyperthermophiles3,48 (Box 1).

Several virus morphotypes have been isolated and described from thermophilic archaea, including 400 × 40 nm flexuous filaments (Lipothrixviridae)50, 600–900 × 23 nm stiff rod-shaped particles that resemble some RNA plant viruses (Rudiviridae)51, a 100–185 × 70–95 nm droplet-shaped particle with fibres that form a beard-like structure (Guttaviridae)2, a 230 × 75 nm bottle-shaped particle (Ampullaviridae)3 and spherical particle morphologies (Globuloviridae)13,16. Two other morphotypes, the spindle-shaped viruses (Fuselloviridae and Bicaudaviridae)7,12,52 and the icosahedral Sulfolobus turreted icosahedral virus (STIV)8, have also been described. The phylogenetic distribution of the latter two virus morphologies indicates different evolutionary histories.

The Sulfolobus spindle-shaped virus 1 (SSV1) was the first crenarchaeal virus to be isolated52. This virus and other members of the Fuselloviridae, including SSV2 (Ref. 53), SSVK1 and a YNP isolate SSVRH7, all share a 60 × 90 nm spindle-shaped morphology with flexuous tail fibres that extend from one end. All isolated spindle-shaped viruses (SSVs) infect Sulfolobus spp. Although all SSV isolates share the same morphology, variations of this morphotype have been reported. An unusually large (107 × 230 nm) SSV-like virus, the Sulfolobus tengchongensis spindle virus 1 (STSV1) was recently isolated from Sulfolobus tengchongensis 9. This virus is significantly larger than members of the Fuselloviridae and probably represents a new viral family. The 75-kb genome shares little sequence similarity with any genes in the public database and is significantly larger (approximately 5-times larger) than genomes of the Fuselloviridae9. Spindle-like morphologies have been reported outside of the Sulfolobus genus. Acidianus two-tailed virus (ATV) of the Bicaudaviridae family is a roughly spindle-shaped virus (100 × 150 nm) that undergoes a novel cell-independent structural transition12. On release from its host, the virus particle develops long tails at each end, but only at high temperatures (75–90°C). At lower temperatures, the virus retains its spindle-shaped, tailless morphology. The extracellular growth of tails might represent a survival strategy for the virus in its harsh environment by promoting host recognition and attachment, especially where host availability might be limited.

Although the spindle-shaped morphology seems to be prevalent among thermophilic crenarchaeotes, there are examples of morphologically similar viruses that infect euryarchaeotes, including one thermophile, the Pyrococcus abyssi virus 1 (PAV1), which was isolated from a marine thermophilic euryarchaeote collected from a deep-sea hydrothermal vent45. PAV1 particles are 80 × 120 nm, with tail fibres that extend from one end. In addition to the close morphological resemblance of the PAV1 particle to that of the Fuselloviridae, PAV1 packages a similarly organized genome (ccds DNA) of approximately the same size (17.5 kb). Spindle-shaped virus morphologies are not limited to thermophilic archaea, as exemplified by the 44 × 74 nm spindle-shaped His1 virus isolated from the extreme halophile Haloarcula hispanica54 and the spindle-shaped particle isolated from Methanococcus voltae A3 (Ref. 55). It is worth noting that this unique virus morphology seems to be restricted to the archaeal domain. One explanation for this observation is that the SSV morphology evolved after the Archaea separated from the other two domains of life (Fig. 2).

Figure 2: Comparison of the evolutionary history of different types of virus compared with the 16S rDNA-based tree of life.
figure 2

Icosahedral viruses, similar to Sulfolobus turreted icosahedral virus (STIV), have been detected in all domains of life, and probably evolved before the first branching event. The head and tail morphology is detected in the Euryarchaeota and the Bacteria, indicating that this lineage also has an ancient origin. Recent evidence indicates that this lineage is also related to the herpesviruses of the Eukarya, however no representative has yet to be detected that infects the Crenarchaeota. The spindle-shaped morphology has only been detected within the Archaea and seems to have arisen more recently than the other morphologies.

In contrast to the domain-restricted distribution of spindle-shaped morphologies, recent structural studies focusing on the major capsid protein (MCP) of STIV8 have identified a fold (double-barrel jelly roll) also used in bacteriophage and eukaryotic viral architectures56. The MCP of STIV is structurally homologous to the MCPs of the algal virus Paramecium bursaria Chlorella virus 1 (PBCV1), mammalian adenovirus and the bacteriophage PRD1. Additional support for this evolutionary link has been provided by protein and lipid analysis of STIV particles57. This analysis showed that STIV has an internal lipid membrane composed of a subset of the host's lipids, as seen in PRD1 (Ref. 57). More recently, a halophilic virus, SH1 (Ref. 58), has been isolated which seems to have a similar structural architecture, including the icosahedral shape, the internal lipid membrane and a protein with an ATPase motif that is similar to one occurring in PRD1 (Ref. 59), and homolgous to one predicted to exist in STIV57. Although these viruses have different genome structures and encode for structural genes with low primary sequence similarity, their evolutionary relationship is preserved in the 3D structures of their MCPs. The structural similarity between MCPs and the overall virion architecture indicates that these viral capsid proteins share an ancient ancestor. Once separated, the individual viruses evolved independently. This is the first example of a virus architecture that has been found in all three domains of life and suggests that this virus architecture might have existed prior to the separation of these three domains, over 3 billion years ago (Fig. 2).

The shared ancestry among icosahedral viruses from the three domains is mirrored by the likelihood of shared ancestry between tailed phage and herpesviruses. Recent comparative analysis of the capsid proteins of tailed phages infecting bacteria and herpesviruses infecting Eukarya has revealed a common fold in the capsid protein60. This fold provides stability for the capsid and allows for the conformational flexibility seen in the assembly and maturation of these viruses. Further comparisons between the cryo-electron microscopy (cryo-EM) reconstructions of SPO1, a tailed phage, and HSV-1, a herpesvirus, have strengthened the argument for shared ancestry61. The tailed phage has been detected in bacteria and euryarchaeotes, but so far has not been shown to infect crenarchaeotes. If the tailed phage shares a common ancestry with the herpesviruses, it is possible that there is either a tailed phage or a herpes-like virus that infects members of the Crenarchaeota.

Not surprisingly, all virus particles isolated from high-temperature environments are thermostable, some withstanding near-boiling, acidic conditions. It has been suggested that the unusual morphologies detected in these environments are adaptations for survival of free viruses12. However, it is unclear how these morphologies contribute to the stability of virus particles in hot, acidic environments. Recent atomic-resolution structural studies of one crenarchaeal virus, STIV, indicates that the thermal stability of the particle is a function of formation of the capsid with increased buried surface area between subunits, a tightly packed subunit structure with reduced cavity volume, higher protein proline content, short solvent-exposed protein loops, an increase in electrostatic interactions and an increased polar surface area61 (Fig. 3). The occurrence of intracellular disulphide bonds might also aid in the stabilization of the virus-encoded proteins. Atomic-resolution structural analysis of several small SSV and STIV proteins revealed disulphide bond formation that probably functions to stabilize these proteins (M. Lawrence, personal communication). It is speculated that intracellular disulphide bond formation might also be a general feature of proteins of thermophilic archaea to enhance thermal stability62. The same general principles of stabilization of non-viral thermophilic proteins seem to apply to viral subunits that comprise the virus particle.

Figure 3: Cryo-transmission-electron microscopy image reconstruction of Sulfolobus turreted icosahedral virus.
figure 3

A cut-away shows the internal lipid layer (yellow) and the integration of the turret structures into the main capsid8,56. Image courtesy of J. Hilmer, Montana State University, using Chimera.

The nature of genome structure and content

All archaeal viruses isolated so far package circular or linear double-stranded DNA genomes, two of which are highly modified1. The genome of Sulfolobus neozealandicus droplet-shaped virus (SNDV) has been shown to have Dam-like methylation, possibly owing to a virus-encoded methylase2, whereas STSV1 seems to have several different modifications, including specific modification of cytosine residues, although Dam-like methylation does not seem to occur9. Some of the linear genomes have covalently closed ends51, whereas others have unidentified modifications6. These modifications, like those used by many DNA bacteriophage, are expected to aid in chromosome replication and in preventing degradation by host nucleases. The SSV1 genome was the first example of positively supercoiled topography63. This helped lead to the discovery that reverse gyrase, the enzyme responsible for positive supercoiling of DNA, is both specific and common to all hyperthermophiles64.

The genomes of most isolated crenarchaeal viruses have been sequenced, providing a wealth of information about the genetic diversity of these viruses3,6,7,8,9,10,12,65. One surprise revealed by these viral genomes, and the genomes of their thermophilic hosts, has been their low GC content, typically around 40%. It is unclear why thermophiles have not exploited the increased thermal stability provided by GC base pairing. However, several other strategies have been implicated in maintaining nucleic acid stability in thermophiles, including increased intracellular ionic concentrations, binding of polyvalent cations or polyamines and nucleoside modifications such as methylation66. Not surprisingly, codon usage in these viral genomes tends to follow the codon bias found in the host organisms, and, similar to most viral genomes, the crenarchaeal viruses have genes that are tightly arranged on their genome. On average, the viral genes tend to encode proteins that are smaller than their host proteins.

Sequence comparisons of crenarchaeal viral genomes using standard alignment algorithms such as BLAST have revealed that few of the putative viral genes share significant similarity with other sequences in the public databases. Where sequence similarity is detected, it tends to be with genes of unknown function in other archaeal viruses or archaeal genomes. However, threading algorithms such as 3D-PSSM, phyre and FUGUE, which help relate sequence data to secondary and tertiary structures, do identify some similarities between putative crenarchaeal virus proteins and proteins in the structural databases. For example, structural analysis shows both a ROP-like protein and a winged-helix protein encoded by SSV67,68 and a winged-helix protein in STIV (E. Larson, unpublished observations). Although these structural predictions hint at protein function, biochemical and genetic studies are required to accurately determine biological functions. Structural and proteomic analysis of purified virus particles has led to the identification of several structural proteins4,8,9,61 and identified virion-associated proteins encoded on the host chromosome57. At least nine virus-encoded gene products and one host protein are found within the STIV particle.

It is clear that viruses with similar morphologies and genome structures share more ORFs than viruses with different morphologies7,65. For example, the comparison of four SSV-like isolates from different hot springs worldwide revealed that out of the approximately 31–35 ORFs encoded by each genome, a common set of 18 ORFs are shared by all four sequenced isolates7. When the genomes of two different Icelandic isolates of the rudivirus SIRV were compared, a similar situation was observed65. The SIRV genomes share clusters of ORFs with high similarity, interspersed with small regions with little or no sequence similarities. The unique genes observed among related viruses might reflect different evolutionary histories or unique adaptations required for a particular host within a particular hot spring environment. As has been suggested for other viruses, it seems that crenarchaeal viral genomes might be composed of a mosaic of genes, probably obtained from different sources at different times69.

As more viral genome sequences become available, it is evident that there are subsets of gene families that are common among different types of crenarchaeal viruses. Some shared gene families have been identified from diverse groups of these viruses (Fig. 4). Based on sequence similarity, ORFs that code for glycosyltransferases have been detected in STIV (E. Larson, personal communication), ATV12, and many viruses of the Rudiviridae65 and Lipothrixviridae5 families. Another protein family shared by these viruses and by the Fuselloviridae, is the CopG family of DNA-binding proteins. These transcriptional-repressor proteins have been implicated in controlling plasmid copy number70 and the apparent conservation of these genes in archaeal viruses might imply an ancient origin. A third example of related genes shared by diverse thermal virus families is ORF A247 from SSVRH, which is similar to genes found in the three other fuselloviruses, and STSV1 and ATV, but has an unknown function. This is by no means a complete list of the similarities among the different isolates of viruses that infect thermophilic crenarchaeotes. There are many ORFs that are shared between the Lipothrixviridae and the Rudiviridae which, along with morphological similarities, has led to the suggestion that the Lipothrixviridae and the Rudiviridae might form a new virus superfamily65. All of the above examples compare viruses isolated from two closely related genera, Sulfolobus and Acidianus. The relationship between these viruses might be a result of the close evolutionary history of their hosts. This suggestion is supported by the fact that the two spherical viruses that infect the Thermoproteales, which occur in more neutral-pH environments, share several ORFs, but have no significant similarities to any of the ORFs from viruses infecting the Sulfolobales6. However, horizontal exchange of viral genes across diverse phylogenic boundaries, as has been proposed for bacteriophage69, might become evident as more sequences become available. Further analysis of viral genes and their products will provide insights into the past histories and current lifestyles of crenarchaeal viruses.

Figure 4: Schematic representations of selected crenarchaeal viral genomes illustrating examples of shared gene families.
figure 4

ORFs representing putative glycosyl transferases (blue), putative CopG-family DNA-binding proteins (yellow), ORFs with unknown function (red) and putative DNA-binding proteins (green) shared by spindle-shaped viruses are indicated. Not all of the shared ORFs have been indicated in the figure, as Acidianus filamentous virus 1 (AFV1), Sulfolobus islandicus filamentous virus (SIFV) and Sulfolobus islandicus rod-shaped virus 2 (SIRV2) have large numbers of ORFs in common. Acidianus two-tailed virus (ATV), Sulfolobus spindle-shaped virus Ragged Hills (SSVRH), Sulfolobus turreted icosahedral virus (STIV) and Sulfolobus tengchongensis spindle virus 1 (STSV1) have circular genomes, whereas the other three genomes are linear. Genomes are not scaled to each other.

Viral replication cycles

Although the basic replication cycle (adsorption, replication and release of progeny virus) for any crenarchaeal virus remains to be determined, several trends are emerging.

Entry and uptake. Viruses of the Fuselloviridae, Lipothrixviridae4,11, Rudiviridae5 and Guttaviridae2 are thought to associate with their host through tail fibres that are present at one or both ends of the virus particle. It is probable that these tail fibres are involved with attachment of the virus particle to components in the host S-layer or the underlying membrane. Similarly, the turret-like projections on the surface of the STIV virion are thought to be involved with host cell recognition and attachment8. The mechanisms of virus uptake are not known for any crenarchaeal virus. Presumably, a specific virus enters through a pathway in which either the virion is taken up into the cell or the viral genome is delivered into the host cell from the virion, which is attached on the exterior surface of the cell. Whereas some viral infections can be cured from cultured hosts11, others seem to be stably associated with the host cells51.

Once inside the cell, some viruses can integrate their genome into the host chromosome, whereas others are maintained as extrachromosomal elements. In the case of SSVs, the full-length genome is integrated site-specifically into the host chromosome. Viral integration is mediated by a virus-encoded integrase of the tyrosine recombinase family71. A sequence in the integrase gene has been shown to target viral integration by complimentary base-pairing between the virus and host chromosomes. The four SSV viruses have been shown to integrate into different locations, commonly exploiting the tRNA genes of their host7. The biological significance of SSV integration is not currently understood, however other viruses, such as lambda phage, typically do not require the integrated copy to complete the viral replication cycle. It is not known if integration of the SSV genome is required to complete its viral replication cycle. The SSV genome is also maintained as a negatively supercoiled episomal plasmid72. Given what is known about the role of the integrated form of other viruses, it is probable that the integrated form of the SSV genome is used as a mechanism by which the viral genome can be maintained during cycles of host replication.

A tyrosine recombinase gene has also been identified in ATV. This virus seems to integrate into the host genome and induction results in a lytic cycle that culminates in lysis of the host12. Induction of ATV resulted when cultures were treated with either mitomycin C or ultraviolet (UV) light — two well known inducing agents. Interestingly, virus production was not detected when cells were grown at the optimum growth temperature for the host, 85°C, but production was triggered by lowering the temperature of the cultures to 75°C (Ref. 12).

Two other virus types have been shown to encode integrase-like genes. STSV1, which shares some genome similarity with ATV, has a putative, truncated integrase gene, however, Southern analysis of infected cells indicates that STSV1 does not integrate9. A putative tyrosine recombinase-like gene has also been identified in STIV8, however, an integrated form of the STIV genome has not yet been detected by Southern analysis (H. Þórisdóttir, unpublished observations). It is important to note that members of the tyrosine recombinase gene family have a broad range of biological functions in addition to integration of phage into host genomes73, including resolving concatamers of the viral chromosome formed during replication, which might be important for correct packaging of virus particles. Tyrosine recombinases might also aid in the partitioning of plasmids or phage genomes into daughter cells during cell division, a mechanism that might assist in the spread of viruses without requiring exposure to the extracellular environment.

Transcription and genome replication. Many crenarchaeal viruses produce transcripts that seem to be polycistronic, a common virus strategy for co-expressing functionally related genes. An exception to this is STSV1, which seems to transcribe monocistronic RNAs, as 76% of the ORFs are preceded by putative promoter regions9. In two studies of crenarchaeal virus transcription, polycistronic transcripts were detected74,75. So far, a high degree of differential viral gene expression has not been detected, although studies have not been carried out with synchronized cultures, which could identify temporal patterns of transcription. Virus production in SSV1 can be induced in cells by UV irradiation52,76, mitomycin C77 or infection with SIRV1 (Ref. 78). Eight transcripts (T1–T8) were identified in asynchronous cultures of SSV1, whereas a ninth transcript (Tind) was only expressed after UV induction, and corresponds to increased virus production74. In cells infected with SIRV1 or SIRV2, transcripts covering all but one ORF were detected 30 minutes after infection. Transcription of some ORFs seems to vary during the infection cycle, including the ORF coding for the coat protein, which was initially transcribed as a polycistronic message, but later detected as a monocistronic message75. Although multiple transcriptional promoters containing the TATA-like core are present in many of the crenarchaeal virus genomes4,6,8,11,79 it is not clear how many of these elements are functionally relevant and archaeal virus promoters remain poorly characterized. In general, crenarchaeal virus transcription is expected to mimic that of its host. Currently it is thought that archaeal gene transcription uses a eukaryotic pol II-like basal transcription apparatus that is regulated by bacterial-like systems in which sequence-specific DNA-binding proteins compete for, or prohibit, extension of 3′ ends by the transcriptional machinery.

Although there are few studies of replication strategies for crenarchaeal viruses, replication schemes from viruses of bacteria and eukaryotes might provide some insights. In the case of the Rudiviridae, it is postulated that DNA replication is similar to that shown for the eukaryal poxviruses65. In support of this model, the SIRV genome is structurally similar to the vaccinia virus genome, with both having linear, double-stranded DNA that is covalently closed by terminal hairpin structures. The hairpin structures in both SIRV and vaccinia virus genomes are preceded by inverted terminal repeats (called 'flip' and 'flop' in vaccinia virus). Replication is initiated by the introduction of a terminal nick, which exposes a free 3′-hydroxl and allows the free end to fold back on itself forming a duplex with its complementary sequence in the inverted repeat region80. An evolutionary link between these two apparently distant viruses is further supported by the relatively high sequence similarity of homologous ORFs, such as the putative Holliday junction resolvases detected in all three rudiviruses and required for this method of replication5,65. Further discussion about the replication of SIRV is provided in Ref. 1.

A potential origin of replication has been identified in STSV1 (Ref. 9). Analysis of the genome has identified a 1.4-kb intergenic region with a low GC content between ORF1 and ORF74. Within this region are two tandem repeats, TR1 (25 bp) with 5.5 copies and TR2 (40 bp) with 2.5 copies, and two sets of inverted repeats which could form stem-loop structures. These characteristics are consistent with this region being the origin of replication.

Assembly and release. The assembly and release of most crenarchaeal viruses known so far does not require cell lysis; the one known exception is the release of ATV, which is mediated by lysis12. Preliminary data from a time course of Sulfolobus spp. cells infected with SSV indicate that virus particles assemble at the host membrane where particles incorporate host lipids, and viral progeny are released without causing cell lysis. By contrast, many of the viruses infecting bacteria and euryarchaeotes tend to lyse cells on release81. Most crenarchaeal viruses seem to maintain a chronic infection, in which particles are produced either continually or during short events resulting in growth inhibition. Long infections with extended release of virus progeny have been proposed to be an adaptation to the extreme external environment16. The stability of a virus in hot, acidic environments might be reduced, therefore continuous release of virus might minimize the time a single virus is exposed to the extracellular environment.

Biogeographical isolation of crenarchaeal viruses

Related crenarchaeal viruses can be isolated from hot springs around the world. The occurrence of morphologically similar viruses from geographically isolated hot springs has raised questions regarding their genetic relationship. It has been suggested that genomic variability among these viruses might be related to their long-term geographical isolation from one another and/or differences in biogeochemical composition in each geographical location. Interestingly, studies examining diversity within the host genera, Sulfolobus, indicate biogeographical isolation82. If thermophilic crenarchaeotes and their viruses are geographically isolated, it would counter the argument that 'everything (all microbial species) is everywhere' and the environment determines which species are successful and dominant. Ongoing studies indicate high rates of viral migration between different hot springs within YNP. Understanding the mechanism(s), frequency and rates of viral migrations between distant habitats will be essential to understand the evolution and ecological significance of these and other widely distributed viruses.

Determining the total diversity of crenarchaeal viruses is the first step to understanding their distribution around the world. From the study of three hot springs in YNP it is clear that the community size for a single virus type is extremely large, as sequence accumulation curves show that the total diversity was not sampled in the approximately 2,000 clones sequenced for SIRV-like viruses35. These results indicate that the total virus metacommunity size probably exceeds 109 sequence types within the SIRV-like viruses in a hot spring environment, underscoring the substantial global biodiversity represented by viruses. The possibility that migration might have an important role in the distribution of viruses within a geothermal area leads to the question of how viruses migrate. Possibilities include movement of either virus particles or infected hosts through underground water connections, through the atmosphere by steam, or by movement of animals between sites. It is probable that there are several pathways of virus migration, but at this time we can only speculate which is the dominant route.

Future directions and questions

Although more crenarchaeal viruses have been isolated and examined in recent years, we still have only a rudimentary understanding of this group of unusual viruses. Current and future efforts will be directed at gaining a better understanding of their fundamental biology; including mechanisms of attachment, transcription, genome replication and viral release. These studies will probably contribute to the development of robust genetic systems, providing archaeal virologists with the fundamental genetic tools that have been historically offered by bacteriophage.

The total diversity of the crenarchaeal viruses has yet to be determined and remains an important avenue of research. This question is being studied by further isolation and characterization of viruses, and by culture-free techniques. The application of culture-free techniques might enable the study of non-thermophilic crenarchaeal viruses, where no host systems exist. An extension of the question of total diversity is whether there is a thermophilic crenarchaeal virus with an RNA genome. The isolation of an RNA virus that replicates in a deeply branching hyperthermophilic archaeon could lead to new insights into the role of RNA in the evolution of life. This would support the theory that primordial life was RNA based.

The challenge is to understand the ecological role these viruses have in their environments and their role in the evolution of life. Although the study of crenarchaeal viruses is still in its infancy, it is a field that is rapidly advancing, and the future holds potential for significant discovery.