Spindle-shaped virus-like particles are abundant in extreme geothermal environments, from which five spindle-shaped viral species have been isolated to date. They infect members of the hyperthermophilic archaeal genus Sulfolobus, and constitute the Fuselloviridae, a family of double-stranded DNA viruses. Here we present four new members of this family, all from terrestrial acidic hot springs. Two of the new viruses exhibit a novel morphotype for their proposed attachment structures, and specific features of their genome sequences strongly suggest the identity of the host-attachment protein. All fuselloviral genomes are highly conserved at the nucleotide level, although the regions of conservation differ between virus-pairs, consistent with a high frequency of homologous recombination having occurred between them. We propose a fuselloviral specific mechanism for interviral recombination, and show that the spacers of the Sulfolobus CRISPR antiviral system are not biased to the highly similar regions of the fusellovirus genomes.
Five fuselloviruses have so far been isolated from acidic geothermal environments in different locations in Asia, Europe and North America, and they replicate in species of the hyperthermophilic archaeal genus Sulfolobus, which represents a significant percentage of the microbial population in most acidic terrestrial hot springs. Another major player in these environments is the genus Acidianus, from which several viruses have been isolated, including the linear filamentous and rod-shaped viruses AFV1 and ARV1, respectively, which have close viral relatives that also infect Sulfolobus (Prangishvili et al., 2006a; Snyder et al., 2007). Although the two genera coexist, no fusellovirus has yet been isolated from Acidianus, even though it appears to be the most predominant Sulfolobus viral type.
The circular dsDNA genomes of five known fuselloviruses are highly similar at both nucleotide and amino acid sequence levels, with the majority of gene products being of unknown function and lacking homologues in public sequence databases other than in other archaeal viruses (Wiedenheft et al., 2004). The viral DNA is protected against the harsh environment, at temperatures above 80°C and pH values below 2, within a spindle-shaped virion about 100 nm long and 60 nm wide, with a bunch of short, thin fibres at one of the pointed ends (Martin et al., 1984; Stedman et al., 2003; Wiedenheft et al., 2004; Peng, 2008). In the electron microscopy, the body is sometimes observed to be slightly elongated and more ‘cigar-shaped’, and the tail fibres appear to be quite sticky, readily attaching to cellular fragments, as well as linking virions to produce rosette-like aggregates (Fig. 1A– SSV7).
SSV1 is the best studied fusellovirus, and the virion has been shown to contain proteins VP1, VP2, VP3 and small amounts of SSV1_D244 and SSV1_C792 (Reiter et al., 1987a; Menon et al., 2008). VP1 and VP3 are thought to be capsid proteins, whereas VP2 has been assigned a DNA-binding role, organizing DNA, but it is not encoded by other fuselloviruses (Stedman et al., 2003; Wiedenheft et al., 2004). Four non-structural SSV1 proteins have been characterized. SSV1_D63 is considered to link two different protein complexes, while SSV1_F93 and SSV1_F112 are DNA binding proteins implicated in transcriptional regulation (Kraft et al., 2004a,b; Menon et al., 2008). The fourth protein is an integrase of the tyrosine recombinase family, which catalyses site-specific integration of the viral genome into the host chromosome. As the viral recombination site (attP) is located within the integrase gene, integration leads to gene partition (Palm et al., 1991; Muskhelishvili et al., 1993). Despite this highly specialized adaptation, the integrase was recently shown to be non-essential for virus replication and basic viral functions (Clore and Stedman, 2007).
Here we report on the isolation and properties of four novel members of the Fuselloviridae, infecting species of the hyperthermophilic archaeal genera Sulfolobus and Acidianus, almost doubling the number of known fuselloviruses and extending their host-range to a new genus, Acidianus. This merited a revised comparative genomic analysis of fuselloviruses, which provided insights into functions of some viral proteins and addressed general questions concerning the evolution of the viruses and interactions with their hosts.
Isolation of virus–host systems
Three different methods were used to acquire the new viruses reported in this communication. SSV5 was discovered as an extrachromosomal element within cells of S. solfataricus P2 (DSM1617), infected as a result of mixing the cells with an icelandic HVE14 enrichment culture (see Experimental procedures). This traditional method of isolating new viruses allows a large number of viruses to be screened, but it restricts the search for specific virus–host systems.
A different approach was used for SSV6 and SSV7, where transmission electron microscopy analysis of the supernatant from an enrichment of the G4 site at Hveregedi, Iceland, revealed a large number of fuselloviruses. Attempts to isolate single virus–host systems by colony purification resulted in two pure strains, each harbouring a different fusellovirus. Strain G4T-1 was a host for Sulfolobus spindle-shaped virus 7 (SSV7) while strain G4ST-T-11 was the natural producer of a pleiomorphic virus named Sulfolobus spindle-shaped virus 6, SSV6. The former was found to be produced in very low amounts under normal growth conditions, but it was possible to increase SSV7 production about 10-fold (as estimated by counting viral particles in the electron microscope), either by shifting the culture to a medium with lower tryptone concentration, or by inducing with UV light. Strain G4T-1 and G4ST-T-11 may in fact be the same species, as their partial 16S rRNA sequences were identical to S. islandicus strain I7 (AY247894.1) with a single base substitution to distinguish them from S. solfataricus P2. This virus isolation approach did not impose any bias on the choice of viral host (except for choosing the growth conditions), and it provided a ‘natural’ virus–host system. However, a bias is imposed on the virus, which excludes the possibility of isolating single colonies of the host if the virus is highly lytic under the chosen conditions.
Finally, the fourth virus described here, Acidianus spindle-shaped virus 1, ASV1, was discovered as an extrachromosomal and integrated element in the course of sequencing the genome of Acidianus brierleyi DSM1651, and the production of virions was subsequently confirmed by electron microscopy (Fig. 1). While this method for isolating new viruses is not generally applicable, it is likely to become more common that extrachromosomal elements are detected while sequencing genomes from strain collections.
The spindle-shaped virion of SSV7, ∼90 nm long and ∼50 nm wide, resembles virions of all previously known fuselloviruses morphologically, as well as by its tendency to form ‘rosettes’ by sticking to neighbouring viral particles (Fig. 1). In contrast, negatively stained virions of SSV6 and ASV1, both appear much more pleiomorphic than the other fuselloviruses, and assume shapes ranging from thin cigar-like to pear-like, with tail fibres at the end corresponding to where the pear ‘stalk’ would be (Fig. 1).
The virion bodies, and tail fibres of ASV1 and SSV6, seem to differ from those of the other fuselloviruses. Instead of multiple thin fibres, these virions carry 3 or 4 thicker and slightly curved, fibres that appear to protrude sideways, not from the particle apex but from a point slightly more towards the body (Fig. 1B). Furthermore, the ASV1 and SSV6 fibres seem to be less ‘sticky’ than their thin counterparts, and the characteristic ‘rosettes’ were never observed for ASV1 and SSV6.
To exclude that the observed pleiomorphicity of SSV6 virions was an artifact caused by the purification process, or by uranyl-acetate staining, two control experiments were carried out. (i) The virions were analysed by EM directly after removal of host cells by mild centrifugation at 4000 r.p.m. (Jouan S40 rotor), and although omitting the concentration step yielded few virions, they exhibited the normal pleiomorphicity. (ii) The virion pleiomorphicity was also observed when we used phosphotungstenate as an alternative contrasting agent (not shown), confirming that the heterogeneity of the shape was an integral property of the virions rather than a result of the experimental treatment. Moreover, SSV7 virions, for which little to no pleiomorphicity was observed, were routinely treated in exactly the same manner as SSV6 and ASV1 virions (Fig. 1A).
Genomic organization and comparison
Owing to their special structural properties, we originally suspected that ASV1 and SSV6 were representatives of a new spindle-shaped viral family. However, genome analyses revealed that they, and the SSV5 and SSV7 isolates, are all closely related to known members of the family Fuselloviridae, and we therefore assign the four newly isolated viruses to this family. The similarities are evident, both in terms of overall gene synteny and sequence similarity (Table 1, Figs 2 and 3), and also extends to the distribution of the cysteine codons in a manner that supports the findings of Menon and colleagues (2008).
Table 1. Genes in SSV5, SSV6, SSV7 and ASV1, as well as the homologues from other fuselloviruses.
SSV1 (15 465 bp)
SSV2 (14 796 bp)
SSV4 (15 135 bp)
SSV5 (15 330 bp)
SSVk1 (17 385 bp)
SSVrh (16 473 bp)
SSV6 (15 684 bp)
SSV7 (17 602 bp)
ASV1 (24 186 bp)
A, B and C indicate genes on the three reading frames of the plus-strand, and D, E and F indicate genes on the minus-strand. The number following the letter is the number of encoded amino acids. The 13 ‘core’ genes are in boldface, and proteins for which experimental data are available are underlined. The asterisk indicates an ad hoc ORF name for a gene which is not present in the NCBI annotation.
The genome of ASV1 carries 24 186 bp and is by far the largest of the fuselloviruses, and one or two gene duplications appear to have occurred (ASV1_B91 and ASV1_C137), as well as the acquisition of new genes. Most of the ASV1 genome is closely related to the other fuselloviruses, with several regions of more than 75% identity at the nucleotide level (Fig. 3). One 5.6 kb region that is similar to SSV6, starts in the middle of ASV1_C213 and ends in ASV1_B90 (Fig. 3D).
An extreme example of how closely related some fuselloviruses are, can be seen by comparing SSV4 and SSV5, where a 7.9 kb region is almost 100% identical (Fig. 3B), consistent with a recent recombination event having occurred between the viruses. Moreover, the junctions of nucleotide similarity regions are generally intragenic, such that sections of high sequence similarity are mostly short, distributed all over the genomes, and often start and stop in the middle of open reading frames (ORFs) (Fig. 3). These patterns of similarity raise interesting questions concerning interplay and recombination between fuselloviral genomes.
The presence of regions of nucleotide identity between the fuselloviruses raises the question as to how they avoid the extensive antiviral CRISPR systems present in all sequenced Sulfolobus genomes. Therefore, we analysed the correlation between sequence matching of CRISPR-spacers and fuselloviral genomes. A total of 3420 CRISPR spacer sequences were obtained from four complete and nine incomplete Sulfolobales genomes (after subtracting the 278 spacers which S. solfataricus P1 and P2 have in common). Ninety-one of these spacers match to one or more of the fuselloviruses on a nucleotide sequence level. An additional 101 spacers were found matching to one or more fuselloviruses when extending the search to the amino acid sequence level. Thus out of the 3420 Sulfolobales spacers, in total 192 spacers yield 436 significant matches to fuselloviral genomes. The latter number exceeds the former because many spacers, especially on the amino acid sequence level, yield matches to more than one virus, and because some spacers match to repeats within the same viral genome. We found no biased correlation between conserved regions and spacer matches, and it is possible that fuselloviruses recombine frequently enough to reduce the effectiveness of the CRISPR system. The results are summarized in Fig. 4, exemplified by SSV2 which has the highest number of spacer matches, and by the most distantly related fusellovirus, ASV1. The spacer matches occur on both strands of the viruses, consistent with DNA recognition by the spacer transcripts, as recently proposed (Marraffini and Sontheimer, 2008; Shah et al., 2009).
Many of the ORFs encoded on ASV1 yield no, or very weak, matches in public sequence databases, especially ORFs found in the ‘extra’∼6 kb that are not present in other fuselloviruses. Exceptions are ASV1_B276 and C106, which are homologous to genes from an integrated virus in the Sulfolobus tokodaii chromosome (ST1724 and ST1725), and ASV1_A59, which exhibits sequence similarity to CopG transcriptional regulators in M. sedula and S. acidocaldarius. Both SSV6 and ASV1 encode a homologue of the structural protein SSV1_VP2, which is absent from the other six fuselloviruses. Furthermore, SSV6 and ASV1 do not encode a full-length SSV1_C792 homologue, and SSV1_B78 homologue, as do all other fuselloviruses (Fig. 2B). Instead, they carry two other genes: a small gene (SSV6_C213 and ASV1_B208) homologous only to the C-terminal 170 aa of SSV1_C792, and following this gene, a large ORF (ASV1_A1231 and SSV6_B1232), which is similar to Saci_1002 from Sulfolobus acidocaldarius (49% identity, 65% similarity, for SSV6_B1232). No other sequence similarity is found in databases, but a clue to the function of both the SSV1_C792 and SSV6_B1232 homologues is given by the Phyre fold-prediction-server (Kelley and Sternberg, 2009), which suggest they both have a fold similar to the adsorption protein P2 from bacteriophage prd1 (E-value < 0.5, estimated precision 85%).
ASV1, SSV7 and SSVk1 differ from the other fuselloviruses by lacking all genes of the SSV1_T5 operon except the integrase and, for ASV1 and SSVk1, a predicted helix–turn–helix transcriptional regulator (Fig. 2). Instead, the three viruses carry a set of ORFs on the plus-strand, which encode a putative Rad3-like helicase, an Msed_2283 homologue (hypothetical protein) and a few small proteins (Fig. 2).
Beside these peculiarities of the individual genomes, analyses have revealed 13 genes that are conserved in all nine fuselloviruses. These ‘core’ genes include VP1 and VP3, the integrase and three putative transcriptional regulators, including one helix–turn–helix and two zinc-finger proteins (Fig. 2). The attP sites within the integrase genes all have their best hits to tRNAs from S. solfataricus, with Gln, Gln, Gly and Lys for SSV5, SSV6, SSV7 and ASV1 respectively. Table 1 shows an overview of the genes in SSV5, SSV6, SSV7 and ASV1, as well as the corresponding homologues in the other fuselloviruses.
In this paper we describe four new members of the family Fuselloviridae, SSV5, SSV6, SSV7 and ASV1, isolated from acidic hot springs of Iceland and USA, which infect members of the hyperthermophilic archaeal genera Sulfolobus and Acidianus.
Until now, fuselloviruses had only been found to replicate in Sulfolobus species. Our discovery of ASV1 in Acidianus brierleyi shows that fuselloviruses can propagate in both the major culturable genera from aerobic, acidic hot springs. Therefore, it is likely that fuselloviruses also infect other host species from these environments, such as Caldococcus, Vulcanisaeta and Stygiolobus (Snyder et al., 2007). Furthermore, the family Fuselloviridae presumably also extends its host range into the vast number of currently uncultured species found in other extreme environments, such as the acid mine drainage ecosystem, where a VP2 homologue, recently found by community genomics (Andersson and Banfield, 2008), indicate the presence of fuselloviruses.
By almost doubling the number of described fuselloviruses, we are refining the definion of ‘core’ genes of the family. The 18 conserved, or ‘core’ genes, that were defined for SSV1, SSV2, SSVk1 and SSVrh (Wiedenheft et al., 2004) can now be reduced to 13 (Table 1) and may have to be revised further as more fuselloviruses are sequenced, but our findings correlate well with a recent analysis of fuselloviral proviruses in S. islandicus strains (Held and Whitaker, 2009). We exclude the SSV1_C792 homologues from the list of ‘core’ genes, because we do not consider SSV6_C213 and ASV1_B208 to be able to fully complement the proteins found in other fuselloviruses, which are about four times larger (Table 1).
Six of the ‘core’ genes have no discernible function based on their primary sequence, except for some of them carrying predicted transmembrane segments (Table 1), and experimental data will be needed to determine their functional roles. Of the remaining seven, the integrase function was characterized experimentally (Muskhelishvili et al., 1993; Muskhelishvili, 1994; Serre et al., 2002; Letzelter et al., 2004; Clore and Stedman, 2007). Moreover, VP1 and VP3 are virion components in SSV1 virions, and VP1 is processed from the N-terminus in SSV1, to a length of 73 aa (Reiter et al., 1987a), which may explain the significant size difference we observe among the VP1 genes (Table 1). The remaining C-terminus of VP1 is similar in both length and sequence to the VP3 protein, and their roles might be partially interchangeable in the virion matrix. Bioinformatical analyses predict DnaA-like activity for SSV1_B251 homologues (Koonin, 1992) and transcriptional regulation activity for three other ‘core’ genes: SSV1_A79 and SSV1_B129, which are transcribed early, during infection and are probably involved in controlling the hosts transcriptional apparatus, and SSV1_B115, which is co-transcribed together with VP1, VP2, VP3 and SSV1_C792, later in infection, and may be involved in controlling the assembly and/or packaging of virions.
Genes that are highly conserved but present in a subset of the fuselloviruses could provide a possible way of classifying the fuselloviruses into subgroups, albeit subgroups that overlap.
Thus, ASV1, SSV6 and SSV1, encode a VP2 homologue, indicating that they all share a DNA packaging system. However, the difference to the SSV1 protein in the C-terminus may indicate an alternative mode of interaction of the protein and viral DNA with the major virion proteins, VP1 and VP3.
Another subgroup would be the SSVs, which all encode a highly conserved homologue of SSV1_C80, a protein containing the RHH 1 CopG domain. ASV1 does not encode any gene with obvious sequence similarity to SSV1_C80. However, ASV1_A59 also has an RHH 1 CopG domain, although it groups with other RHH1-containing genes, including a few Sulfolobus chromosomal genes (e.g. Saci_0942). Furthermore, ASV1_A59 occupies the same genomic position as the SSV1_C80 homologues do in the SSV genomes, and it very likely acts as a functional homologue of SSV1_C80.
A third subgroup consists of ASV1, SSV7 and SSVk1, which all encode the Rad3-like helicase protein and the neighbouring Msed_2283 homologue (Fig. 2). The presence of the helicase strongly suggests that these two proteins are involved in DNA replication or recombination, and it is possible that the other fuselloviruses recruit host proteins to fulfill the same function.
A possible filament protein
The most striking genomic difference among the fuselloviruses is the ‘replacement’ of the SSV1_C792 module with the SSV6_B1232 module (Fig. 2B). It seems the C-terminal 170 aa from SSV1_C792 are essential, since they are retained as a small separate gene in both the ASV1 and SSV6 genomes; however, the remaining ∼620 aa of SSV1_C792 and the whole of SSV1_B78 are substituted by SSV6_B1232. The presence of the SSV6_B1232 module correlates with a difference in the number and structure of the sticky terminal filaments of the SSV6 and ASV1 virions, when compared with the SSV1_C792 module viruses (Fig. 1B). Possibly, there is a phenotype–genotype link, with the SSV1_C792 module being responsible for the multiple, thin, sticky filaments and the SSV6_B1232 module for the few, thick, less sticky filaments. In support of this hypothesis, small amounts of SSV1_C792 were recently found by mass-spectrometry in SSV1 virions (Menon et al., 2008). Moreover, the Phyre prediction tool suggested that both SSV1_C792 and SSV6_B1232 had a similar fold to the P2 receptor binding protein prd1, and it was recently shown that a large protein is responsible for the sticky end-fibres in the rudivirus SIRV2 (Steinmetz et al., 2008). Nevertheless, further studies will be needed to determine the exact functions of the SSV1_C792 and SSV6_B1232 modules in fuselloviruses.
Fuselloviral nucleotide similarity and a putative mechanism for interviral recombination
The multiple regions of high nucleotide similarity, or even identity, between the fuselloviral genomes do not represent a ‘core’ fusello-genome, since the regions of similarity differ between the various pairs of viruses, and often do not include the ‘core’ genes (Fig. 3). Instead, the pattern of similar and non-similar sections of DNA indicates frequent recombination events between fuselloviruses, similar to that observed for some bacteriophages (Hendrix et al., 1999). Possibly this occurs between pairs of fuselloviruses, present in the same host; however, we do not see a similar pattern of sequence similarity for the linear non-integrating archaeal viruses (Vestergaard et al., 2008a). Therefore, we suggest that a different mechanism is more likely.
Integrated fusellovirus genomes have been found in the Sulfolobus solfataricus P2 and in four S. islandicus chromosomes, where no trace of the covalently closed circular DNA (cccDNA) form was detected (Stedman et al., 2003; Held and Whitaker, 2009). Once a virus has been ‘caught’, a second, slightly different, fusellovirus might infect the same host, and insert itself into the same tRNA gene, resulting in a concatamer of the two fuselloviruses in the host chromosome (Fig. 5). This structure might be maintained for a couple of generations, but it would be inherently unstable if the two viral genomes are reasonably similar, as there would be a high chance of homologous recombination between the two integrated viruses. Such a recombination event would lead to the formation of one cccDNA virus and one inserted virus, both of which would consist of a part of each of the original two viruses (Fig. 5). Owing to the very short sequence similarity required for homologous recombination in Sulfolobus (Grogan, 2009), the cross-over point could potentially be in many different places, and each of these recombination events would form a unique mixture of the two viruses, similar to meiosis in eukaryotes. Thus, this offers a mechanism for rapidly generating a large number of diverse viral offspring. Our model does not exclude direct recombination between the cccDNA forms of fuselloviruses, but we propose that this type of ‘tandem insertion’ event happens frequently (on an evolutionary scale) in nature, and that repeated events, each involving a different pair of ‘parent’ fuselloviruses, would eventually produce the patchwork viral genomes we see today (Fig. 3).
Our model also serves to explain why fuselloviruses have developed an integrase that is inactivated upon integration. The integrase is not essential for viral propagation (Clore and Stedman, 2007) but if the proposed recombination mechanism is correct, then the unique SSV-type integrase will help the virus in the long term, by promoting recombination with closely related viruses, since the inactivation provides a high chance of the viral genome being ‘caught’ in an integrated form in the chromosome. Nevertheless, inactivation of the integrase is not required for recombination between tandem insertions. Studies of the Sulfolobus plasmids pARN3 and pARN4 reveal stretches of nucleotide identity, which might have been generated by tandem insertions, even though these plasmids carry non-inactivatable integrases (Greve et al., 2004).
The inherent instability of a tandem insertion makes it difficult, if not impossible, to detect in nature. However, a concatamer of inserted viral genomes, similar to the one proposed in our model, was recently discovered in the chromosome of Methanococcus voltae A3. There, the two viral genomes integrated into the same tRNA gene are very different, preventing homologous recombination, thus ‘trapping’ the viral concatamer in the host chromosome (Krupovic and Bamford, 2008). The attP sites of SSV2 and SSV7 as well as SSV5 and SSV6 match the same tRNA in S. solfataricus P2 (Table 1), making it likely that fuselloviruses are also able to integrate into the same tRNA, forming concatamers, which are unstable due to the similarity between the fuselloviruses. Moreover, it was shown that SSVk1 is able to integrate into several different sites in the host genome (Wiedenheft et al., 2004), increasing the likelihood of finding a ‘partner’ for recombination. Finally, examples of related viruses infecting the same host at the same time are known for Sulfolobales, such as AFV6, AFV7 and AFV8 in Acidianus convivator (Vestergaard et al., 2008b).
If the ‘tandem insertion’ model is correct, then an evolutionary tree of an entire viral genome has no meaning, nor would that from individual ‘core’ genes (since two halves of the same gene might originate from different ‘parent’ viruses). One might instead analyse genes, described in the previous section, that are not shared by all fuselloviruses, since these genes cannot serve as cross-over points for homologous recombination. Although for the moment, the data set is too small for a phylogenetic analysis based on these genes, the presence or absence of certain genes in a subset of the viruses, has provided important clues to understanding protein functions in the fuselloviruses, including the putative filament proteins SSV1_C792 and SSV6_B1232.
With the current understanding of the CRISPR antiviral system, high nucleotide similarity between viruses should be disadvantageous, since a single spacer, matching a conserved region, will provide a host with immunity to several virus strains (Lillestøl et al., 2009). Nevertheless, the puzzling fact remains that fuselloviruses do possess highly similar, sometimes identical, nucleotide regions, and it is possible that the integration and/or the frequent recombination somehow provide the fuselloviruses with the means to evade the CRISPR system in their hosts.
It has been proposed that thermoacidophilic archaeal viruses are highly mobile, even between distant hot springs in the same geothermal area, and that different fuselloviruses continuously infect a more-or-less stable population of host species (Snyder et al., 2007). The high nucleotide similarity we have found, even between fuselloviruses isolated on different continents, seems to confirm that they do manage to exchange genetic material over the intercontinental distances that separate some of the geothermal ‘islands’ in the cold ‘ocean’.
Sulfolobus and Acidianus medium
Z medium: 25 mM (NH4)2SO4, 3 mM K2SO4, 1.5 mM KCl, 20 mM glycine, 4.0 μM MnCl2, 10.4 μM Na2B4O7, 0.38 μM ZnSO4, 0.13 μM CuSO4, 62 nM Na2MoO4, 59 nM VOSO4, 18 nM CoSO4, 19 nM NiSO4, 0.1 mM HCl, 1 mM MgCl2, 0.3 mM Ca(NO3)2 adjusted to pH 3.5 with H2SO4. T medium: Identical to Z medium, but with 0.2% Tryptone added. ST medium: Identical to T medium, but with small amounts of elemental sulphur added.
Isolation and purification of hosts and viruses
Samples were collected from the Hveragerdi hot-spring area in south-western Iceland and 1 ml was used to establish an enrichment culture, by incubating in 50 ml ST medium for 9 days at 80°C, after which 1 ml was of the enrichment was transferred to fresh ST medium and incubated for a further 4 days. Four millilitres of the enrichment (designated G4ST) was then centrifuged at 4000 r.p.m. for 20 min (Jouan S40 rotor) to remove cells, whereupon the supernatant was spun further at 38 000 r.p.m. for 3 h to pellet virions (Beckman SW60 rotor). Finally, the pellet was resuspended in 50 μl of the supernatant. The resuspension was then examined by electron microscopy, and several different morphotypes of virus-like particles were in evidence. Among these was a group of fusellovirus-like particles, but with different filament structures at the end, and a large diversity in their morphotypes, ranging from sausage-shaped to an almost spindle-like pear-shape (Fig. 1).
To isolate single host–virus systems, G4ST was spread on a plate containing ST medium and solidified with Gel-rite (Sigma-Aldrich, St Louis, USA). After 10 days of incubation at 80°C, 30 colonies of representative sizes, shapes and colours were transferred to 5 ml liquid ST medium and incubated with vigorous shaking for 4 days. Each of the growing strains was examined for virus in the electron microscope, and the SSV6 and SSV7 were detected in the supernatant of strain G4ST-T-11 and G4T-1 respectively. The 16S rRNA genes of G4T-1 and G4ST-T-11 were amplified using the primers 8aF: TCYGGTTGATCCTGCC and 1512uR: ACGGHTACCTTGTTACGACTT (Accession number FJ870913 for G4ST-T-11 and FJ870914 for G4T-1).
SSV5 was present in HVE14, an enrichment culture, established from a natural sample collected near the G4 site, but 10 years previously (Zillig et al., 1996). It was propagated in S. solfataricus P2, by mixing a small amount of HVE14 with a well-grown S. solfataricus P2 culture (1:1000), which was then harvested and used for DNA isolation of extrachromosomal elements using plasmid miniprep kit from Qiagen. Acidianus brierleyi were cultured at 70°C in ST medium and ASV1 was recovered from the supernatant by ultracentrifugation (38 000 r.p.m. for 3 h in a Beckman SW41 rotor).
Ten microlitres of the samples was deposited on a carbon and formvar coated grid (Ted Pella, Redding, CA, USA) and left for 2 min before removing excess fluid. Ten microlitres of 2% Uranyl-acetate or phosphotunstenate (Sigma-Aldrich) was allowed to stain the samples negatively for 10 s. Images were taken on a JEOL1200EXII microscope with an 80 kV beam, using a CCD camera.
DNA isolation and sequencing
Six litres of G4ST-T-11 was grown in a fermentor, and after removing cells by centrifuging twice at 4000 r.p.m. for 20 min (Sorvall GS-3 rotor), the virions in the supernatant were concentrated using a Sartorius Vivaflow 200 filter cartridge (Sartorius, Goettingen Germany). The resulting 15 ml was further concentrated by spinning at 38 000 r.p.m. for 3 h using a SW41 Beckman rotor, and finally the virions were treated with Protease K and the DNA was extracted with Phenol, Phenol/Chloroform and Chloroform extraction. The SSV6 DNA was then treated as described below for SSV7.
In order to sequence SSV7, 5 ml of an exponential G4T-1-culture was pelleted by centrifugation, and resuspended in Z medium. The SSV7 production was induced by 50 J cm−2 UV radiation (254 nm) under constant mild agitation, and the cells were then transferred to 45 ml T medium for over-night incubation. Five millilitres was used for a miniprep (QIAprep Spin Miniprep Kit, QIAGEN SA, Courtaboeuf, France), which was used for amplification and subsequent library construction based on the Linker Amplified Shotgun Library method described at http://www.sci.sdsu.edu/PHAGE/LASL/index.htm.
Shot-gun library construction of SSV5 and SSV6, as well as ASV1-containing A. brierleyi total DNA, was performed as described previously using SmaI digested pUC18 as cloning vector (Peng, 2008). Plasmid DNA of clones, from all four libraries, were purified using a Model 8000 Biorobot (Qiagen, Westburg, Germany) and sequenced in MegaBACE 1000 Sequenators (Amersham Biotech, Amersham, UK). Sequences were assembled using Sequencher 4.5 (http://www.genecodes.com). Genome annotations and comparisons were done using the MUTAGEN software (Brügger et al., 2003) with a minimum ORF-length set to 50 aa and allowing AUG, GUG and UUG as possible start codons. Accession numbers are EU030939, FJ870915, FJ870916 and FJ870917 for SSV5, SSV6, SSV7 and ASV1 respectively.
CRISPR spacer analysis
To obtain a list of spacer sequences from Sulfolobales, the following partial or full genomes were used: S. solfataricus P2, S. tokodaii 7, S. acidocaldarius DSM 639, Metallosphaera sedula DSM5348 from GenBank (http://www.ncbi.nlm.nih.gov/Genbank/), Sulfolobus islandicus strains LD85, YG5714, YN1551, M164 and U328 from JGI (http://www.jgi.doe.gov/genome-projects/), and S. islandicus strains HVE10/4 and REY15A and Acidianus brierleyi (K. Brügger and Q. She, unpubl. data). CRISPRs were identified using publicly available software (Edgar, 2007; Bland et al., 2007). Spacer sequences from each repeat-cluster were aligned (Sæbøet al., 2005) against the fuselloviral genomes at a nucleotide level (Shah et al., 2009). Additionally, spacers were aligned against amino acid sequences of annotated ORFs of the Fuselloviruses, at an amino acid level (Vestergaard et al., 2008a; Shah et al., 2009). Significance cut-offs were determined for both alignment types by using the genome sequence of Saccharomyces cerevisiae as a negative control.
P.R. was funded by grant VIRAR (NT05-2_41674) from the Agence Nationale de la Recherche, France. The research in Copenhagen was supported by grants from the Grundforskningsfond and the Reseach Council for Natural Sciences. We would also like to thank the Electron Microscopy Platform at Institut Pasteur for helpful advice and use of their JEOL1200EXII microscope.