Preparation of total RNA and Northern blotting

Sulfolobus acidocaldarius cells were grown at 78 °C in complex medium containing 2% tryptone (Schleper et al. 1995). Total RNA was prepared from exponentially growing and stationary phasecells by the phenol-guanidium-thiocyanate-chloroform method (Sambrook and Russell 2001) with DNase I treatment. Twenty µg of total RNA was fractionated in an 8% polyacrylamide gel with 7 M urea, 90 mM Tris, 64.6 mM boric acid, 2.5 mM EDTA, pH 8.3, together with a 10–100 nt RNA ladder (Decade Marker System, Ambion, Huntingdon, U.K.) or a 0.1–2.0 kb RNA ladder (Invitrogen, Paisley, U.K.). The RNA was transferred onto nylon membranes (Hybond N+, Amersham Biosciences, Amersham, U.K.) using the Bio-Rad semi-dry blotting apparatus (Trans-blot SD, Bio-Rad, Hercules, CA). After immobilizing the RNAs using a Crosslinker (Stratagene), the nylon membranes were prehybridized for 1 h in 6 × SSPE (0.9 M NaCl, 60 mM NaH₂PO₄, 4.6 mM EDTA and pH 7.4), 0.5% SDS and 5 × Denhardt’s solution at 59 °C. Oligonucleotide primers (26-mers), complementary to each strand of a terminal spacer in saci-4 (5′-GATACGTTGCAGGCAGATGATGAGAG-3′, 5′-CTCTCATCATCTGCCTGCAACGTATC-3′), were end-labeled with [3²P]ATP and T4 polynucleotide kinase. Hybridization was carried out at 59 °C in 6 × SSPE, 0.5% SDS, 3 × Denhardt’s solution and 100 µg ml^–1 tRNA for 18 h. The sample was washed three times at room temperature in 6 × SSPE and 0.1% SDS for 15 min each and subsequently at 59 °C in the same buffer. Membranes were exposed to Ultra UV-G X-ray film (Dupharma, Kastrup, Denmark) for 3 days.

Properties of the sequence repeats

Repeat sequences vary in both length and sequence and are presented for each cluster in Table 2. The crenarchaeal genomes range in size from 24 to 26 bp, whereas the euryarchaea and the nanoarchaeon genomes vary from 26 to 37 bp. There is only limited conservation of many repeat sequences, with the left half showing the higher conservation, as was noted earlier (Peng et al. 2003). Nevertheless, some repeat sequences show major differences, especially in the right half of the sequence; compare, for example, the right halves of the repeats of plasmids pNOB8 and pKEF9 (Table 2). Most repeat sequences show some kind of weak, dyad symmetry generally in the form of interrupted and imperfect short inverted repeats and it has been shown, at least for the genus Sulfolobus, that this provides a recognition site for a repeat binding protein (Peng et al. 2003).

Within some clusters, the repeat sequence exhibits a little variation. For example, in the S. solfataricus cluster, ssol-95, the repeat sequence changes at the center of the cluster to another sequence (albeit with a single nucleotide change) (She et al. 2001). Therefore, we examined the constancy of repeat sequences within each genomic cluster and the results demonstrate that most clusters are not homogeneous in their repeat sequence (Table 2). Many carry one to four altered repeat sequences and two, the aforementioned ssol-95 cluster of S. solfataricus and stok-112 of S. tokodaii, show more dramatic changes. In addition, six of the 20 clusters in M. jannaschii carry two repeat sequences differing in their central regions (Table 2).

Uniqueness of the spacer sequences?

Spacers vary in size from 35 to 44 bp between clusters as well as between organisms, and tend to be conserved within a cluster (Table 2). However, occasional exceptions were detected. In the S. solfataricus cluster ssol-91, a half spacer precedes two atypical repeat sequences, the second of which is followed by a regular repeat sequence and not a spacer. In S. tokodaii cluster stok-121, two atypical repeats are 18 bp longer than the other repeats, but are followed by shorter spacers such that the repeat-spacer unit is conserved in size. In N. equitans, two spacers are 25/26 bp longer than the others. The 56 bp spacers of M. kandleri are more than 10 bp longer than those found in any other archaea.

All spacer sequences within a cluster and within a chromosome are generally different, but a systematic search revealed several exceptions. Spacers are occasionally repeated, sometimes more than once within a cluster, and can appear in different clusters within the same chromosome. The results demonstrate that 12 of the 28 chromosomes investigated carry one or more repeated spacers, which tend to be located in the larger clusters. The distributions of duplicated spacers are indicated in Table 3. There are no clear patterns for the arrangement of duplicated spacers, as some are arranged consecutively, others are located in different parts of a given cluster and a few occur in different clusters. The greatest number of duplicated spacers (36) occurs in the two clusters of M. thermautotrophicus (Table 3) and their distribution is illustrated for the larger mthe-124 cluster in Figure 1A. Although identical groups of spacer-repeat units have been observed in closely related strains, they have not been detected in different species.

Open in a separate window

Figure 1.

(A) A map of mthe-124 from M. thermoautotrophicus showing the locations of duplicated spacer-repeat units, or groups of units, labeled 1–5. Each triangle represents a spacer-repeat unit. Duplicated spacer-repeat units are shaded in gray. (B) Schematic representation of the six clusters which occur in the genome of S. solfataricus P2B (She et al. 2001). Each triangle represents a spacer-repeat unit. The colored triangles are coded to indicate the archaeal viruses, plasmids or chromosomes (S. solfataricus or A. fulgidus) that yield good matches with the spacer sequence. (C) A corresponding scheme is presented for three of the six clusters (ssol-95, ssol-32 and ssol-7) that are present in the chromosome of the closely related strain S. solfataricus P1 (Accession numbers: DQ831675, DQ831676 and DQ831677). Open triangles denote those spacer-repeat units that differ in sequence between the strains P1 and P2A. Abbreviation: del. indicates the two sites where deletion of several spacer-repeat units has occurred in strain P1 (or where insertions have occurred in strain P2B).

Integrity of the clusters

Repeat clusters are generally highly conserved in their repeat sequences and in the sizes of their repeats and spacers. Such integrity extends to an almost complete lack of insertion sequences (ISs) or other mobile elements. This is surprising, given the large number of mobile elements in some archaeal genomes and the large variety of spacer sequences that could provide potential target sites for the insertion of mobile elements (Brügger et al. 2002). Nevertheless, one IS element (ISH4) was identified within the repeat cluster, hmar-57, of the H. marismortui megaplasmid, pNG400 (Table 2). Moreover, a miniature inverted-repeat transposable element (MITE) occurs in the mace-33 cluster of M. acetivorans (Table 2), located at the center of a repeat sequence close to the end carrying the flanking sequence, but not in any of the other identical repeat sequences. The 132 bp MITE shares a 14 bp inverted terminal repeat and a 3 bp direct repeat with the IS element, ISMac11, which exists in the same chromosome and encodes a transposase likely responsible for the MITE transposition (Brügger et al. 2002).

Properties of the flanking sequence

Many chromosomal clusters carry a flanking sequence at one end, which is sometimes referred to as a “leader,” although its function is unknown (see below). These sequences tend to be rich in short homopolynucleotide sequences and AT-rich regions, and they lack open reading frames (Jansen et al. 2002, Tang et al. 2002). Archaeal flanking sequences range in size from 132 bp for H. marismortui to 564 bp for M. kandleri (Table 2), and they directly adjoin the first repeat sequence of the cluster. Moreover, they invariably occur at the same end of the cluster, with respect to the strand orientation of the repeat sequence. There is an approximate direct correlation between the sequence length and the optimal growth temperature of the organism (Tables 1 and ​and2). 2).

For chromosomes carrying multiple clusters with identical repeats, the flanking sequence (if present) is often conserved (Table 2). However, there is generally no conservation of the flanking sequence between organisms, with the exception of three Methanosarcina species, which share five highly similar 171 bp sequences with > 87% sequence identity (Figure 2A).

Open in a separate window

Figure 2.

(A) Alignment of the five conserved flanking sequences adjoining the first repeat (in bold type) of five clusters of the three Methanosarcina strains. A putative TATA-like box is outlined. (B) Alignment of two flanking regions of the P. abyssi genome where paby-1 appears to exhibit a defective start site.

Several clusters lack a flanking sequence, including those with different repeat sequences found in P. aerophilum and A. pernix (Table 2) and some with identical repeats found in S. solfataricus P2B (see ssol-7 and ssol-91 in Figure 1B).

Protein genes associated with clusters

So far, the only protein that has been shown to interact directly with a cluster is the genus-specific SSO0454 from S. solfataricus P2, which binds and distorts the repeat sequence (Peng et al. 2003). However, superoperons containing > 20 genes often flank one or more repeat clusters within archaeal and bacterial chromosomes (Jansen et al. 2002). Because they are absent from bacterial genomes that lack repeat clusters, they were inferred to be co-functional with the repeat clusters (Jansen et al. 2002). Although some of these genes (now denoted cas or csa genes) were earlier annotated as unusual DNA repair enzymes (Makarova et al. 2002), they have recently been reassigned to the regulation and processing of the repeat clusters and to a putative role in piRNA function (Makarova et al. 2006). Predicted functions of the more common gene products are listed in Table 4. Some of the genes show a degree of specificity for different archaeal or bacterial phyla including five archaea-specific genes (denoted csa1 to csa5) (Bolotin et al. 2005, Haft et al. 2005), although we find csa2 homologs in some bacterial genomes. Cas5 and cas6 should not be confused with identically named genes exclusive to bacterial genomes lacking cas2, cas3 and cas4 (Bolotin et al. 2005). An overview of the genes commonly found in archaeal genomes is presented in Table 5.

Typical examples of fairly conserved gene orders present in superoperons of crenarchaea and euryarchaea are presented in Figure 3 for the most commonly occurring genes. All euryarchaeal genomes lack Csa1, except for A. fulgidus. Each gene generally occurs once per superoperon, although duplicate copiesoccur in M. jannaschii and T. kodakaraensis. A few genes, including cas3, cas5, cas6, csa2, csa3 and COG2254, are sometimes located distantly from the repeat clusters, most commonly inthe genomes of the three Pyrococcus species and A. fulgidus.

Open in a separate window

Figure 3.

Typical composition and orientation of the major gene components of superoperons which are closely linked to repeat clusters in crenarchaea and euryarchaea. The cas genes and their COG identities are given in Tables 4 and ​and5. 5.

Organisms with multiple clusters carrying the same repeat sequence exhibit only a single superoperon adjacent to one cluster (Jansen et al. 2002). In some genomes, the superoperons physically link two, or even three, clusters with identical repeat sequences (e.g., in P. aerophilum, S. solfataricus, S. tokodaii and H. marismortui). In contrast, organisms containing two or more clusters with different repeat sequences generally have two or more superoperons (Jansen et al. 2002), except for P. aerophilum, A. pernix and S. tokodaii, which each have one. Exceptionally, superoperons are absent from the chromosomes of T. acidophilum and P. abyssi and plasmids pNOB8 and pKEF9, which all carry repeat clusters (Table 5).

In M. acetivorans and M. barkeri, a superoperon links clusters with dissimilar repeats and both the gene order (cas6-cas4-cas1-cas2-Cluster1-csa2-cas5-cas3-Cluster2) and sequence (> 94% identity) are conserved. Since the repeat sequences of the corresponding pairs of clusters are identical between these two organisms, this suggests that a lateral transfer event of both superoperon and clusters has occurred. However, such an event must have happened at an early stage of cluster development, because no similarities were detected between the spacer sequences of the two organisms.

Development of the clusters

The only relevant study of cluster development was performed on related bacterial strains of M. tuberculosis (van Embden et al. 2000) and of Yersinia (Pourcel et al. 2005). Examination of 26 strains of M. tuberculosis revealed several differences in a large repeat cluster consistent with insertions/deletions of internal repeat-spacer units having occurred. For the related Yersinia strains, a similar phenomenon was observed, but it was also inferred that repeat-spacer units could have been added at the cluster end adjoining the conserved flanking sequence.

In order to shed some insight on how clusters may develop and change in archaea, we completely sequenced three of the six large cluster regions of S. solfataricus P1 and compared the spacer content with those of strain P2B. The two highly similar strains were originally sampled about one meter apart from a small stream at Pisciarelli, Naples (W. Zillig, deceased, personal communication). A comparative repeat-spacer alignment of the two strains (Figures 1B and ​and1C)1C) demonstrates the following: (1) new repeat-spacer units are added at, or near, the end of clusters b and c, which both exhibit an adjoining flanking sequence; (2) the large b clusters are more active in adding new repeat-spacer units than the smaller c clusters; (3) no new repeat-spacer units are added to the e clusters, which lack a flanking sequence, but the clusters remain highly conserved in sequence; (4) deletion and/or insertion of single, or multiple, repeat-spacer units can occur; and (5) deletion/insertion of repeat-spacer units occurs precisely, reinforcing the idea that the structural integrity of the clusters is important for their function.

Both data found in the literature (Pourcel et al. 2005) and the results presented in Figures 1B and ​and1C1C are consistent with new spacer-repeat units being added exclusively adjacent to the flanking sequence. Moreover, since no new units were added to the repeat clusters lacking a flanking sequence (cluster ssol-32 in Figures 1B and ​and1C),1C), this region is likely to provide a binding site for the cas genes involved in the copying-insertion events. Although there is no insight into the mechanism by which repeat-spacer units are added to a cluster, the process is likely to involve reverse transcription of mRNAs and recombination (Makarova et al. 2006).

There are several instances of a flanking sequence being followed by a single repeat or repeat-spacer unit (Table 2), consistent with the clusters developing from one end, but possibly also indicating a defective start site. One such region from P. abyssi is shown where the flanking sequence adjoins a one half repeat and truncated spacer, followed by a full repeat, and it is aligned with the start of paby-23 from the same genome (Figure 2B). There are also a few examples of single repeats that lack flanking sequences, mainly from the genera Pyrococcus and Methanosarcina (Table 2).

The preceding data suggest that development of clusters is primarily dependent on a combination of cas genes and the flanking sequence. An alternative hypothesis is that repeat clusters are spread intercellularly by plasmids (Godde and Bickerton 2006), and, at least for the Methanosarcina species, there is clear sequence evidence for the lateral transfer of a superoperon of cas genes and all its flanking sequences (Figure 2). Moreover, the presence of repeat clusters in conjugative plasmids provides a possible mechanism for intercellular transfer. However, the repeat cluster of pNOB8 does not appear in the S. tokodaii chromosome, where a closely similar plasmid sequence is encaptured (Kawarabayasi et al. 2001).

The role of RNA

Experimental studies have demonstrated that RNA is transcribed from the repeat-clusters of the euryarchaeon, A. fulgidus, and the crenarchaeon, S. solfataricus (Tang et al. 2002, 2005). Examination of clone libraries of reverse transcripts prepared from total cellular RNA (< 500 nt) yielded sequences of a series of cluster-encoded small RNAs (22 from A. fulgidus and one from Sulfolobus). Since the 5′-terminal sequence of each RNA was lacking (by about 10–25 nt), the start position could only be assigned approximately within a repeat sequence, while the 3′-terminus was located at or near the center of the following repeat, yielding an estimated average size of 50–70 nt (Tang et al. 2002, 2005).

Northern blotting, using a probe against the repeat sequence of the clusters, revealed a series of discrete RNA products that were multiples of ~68 nt (68, 136, 204, 272, 340 and 408 nt) for A. fulgidus (Tang et al. 2002) and of ~60 nt (60, 180, 360 and 540 nt) for S. solfataricus (Tang et al. 2005). In both studies, the smallest RNA detected corresponded approximately to the estimated sizes of most of the cloned RNAs. The RNA size distribution is consistent with the processing of larger transcripts at regular spatial intervals and the sequencing data support, but do not establish, that processing occurs at or near the center of each repeat.

To gain further insight into the RNA products, transcription was examined from a small cluster of S. acidocaldarius (saci-4) (Chen et al. 2005) using a Northern blotting approach. Probes were prepared against complementary sequences of the terminal spacer sequence adjacent to the flanking sequence. Total RNA was extracted from exponentially growing and stationary phase cell cultures. The results revealed a series of bands that correspond to transcription from the direction of the flanking sequence (Figure 4). The pattern of larger discrete bands is quite similar to that obtained earlier for S. solfataricus P1 (Tang et al. 2005). What is different, however, is that diffuse bands range in size from 52 to 35 nt for the exponentially growing sample, and from 52 to 30 nt for the stationary phase sample (Figure 4A). This suggests that progressive trimming of the smallest discrete band of 58 nt has occurred, possibly by exonucleases. Clearly, the yields of these smaller products increase, and their average size decreases, on going from the exponential to stationary growth phase. The largest band (~420 nt) exceeds the total size of the cluster (maximum 260 bp), which is consistent with the transcript extending beyond the cluster limits. Evidence is also provided for larger transcripts being produced by the complementary DNA strand in stationary phase cells (Figure 4B).

Open in a separate window

Figure 4.

Northern blotting analyses of RNA transcripts obtained from the cluster saci-4 of S. acidocaldarius. (A) Transcripts from one strand were detected in total RNA extracted from cells grown to exponential (expon.) or stationary (stat.) phase. (B) Transcripts were observed from the complementary DNA strand at stationary phase. The approximate estimated nucleotide lengths of the RNA products are given. The minimal detection limit, using 26-nt. probes, was estimated at 15–16 nucleotides. (C) Denotes the location of the primers “a” and “b,” and “L” indicates the location of the putative transcriptional leader within the flanking sequence.

The observed sizes of the larger transcripts obtained from different clusters are consistent with transcription being initiated in a leader sequence within the flanking sequence adjoining the first repeat (Figure 2). The majority of the flanking sequences (20 out of 26) carry a putative TATA-like motif immediately upstream from the first repeat which, if active, would produce an initation start at the center of this repeat. However, this property is also shared by repeat clusters lacking a flanking sequence and one of these, from the plasmid pKEF, produces transcripts (R.K. Lillestøl, unpublished data), suggesting that the flanking sequence is not primarily involved in transcriptional initiation.

Derivation of the spacer sequences from extrachromosomal elements

Recently, evidence has been presented for both archaea and bacteria that some spacer sequences correspond closely to sequences occurring in extrachromosomal elements some of which lie within ORFs. For the archaea, sequence similarities were found with Sulfolobus fuselloviruses and rudiviruses as well as with the conjugative plasmid pNOB8 (Mojica et al. 2005), whereas, for the bacteria, they were found for bacteriophages of Streptococcus and a prophage of Yersinia (Bolotin et al. 2005, Pourcel et al. 2005).

We tested all the archaeal spacer sequences for matches against the GenBank database and against our in-house archaeal genome database, which contains all the available archaeal viral, plasmid and chromosomal sequences (Brügger et al. 2003). Several sequence matches were found, showing 88–100% identity with viruses, plasmids and chromosomes. They are listed in Table 6 and include those reported earlier (Mojica et al. 2005). Most matches were to ORFs; the few exceptions (Table 6) mainly correspond to predicted noncoding regions of chromosomes.

The majority of the positive archaeal matches are between spacer sequences of S. solfataricus and extrachromosomal elements of the related genera Sulfolobus and Acidianus. This strong bias probably reflects: (1) that Sulfolobus species are especially rich in repeat-spacer units (Table 1); and (2) that most sequenced archaeal viruses and plasmids derive from the related genera, Sulfolobus and Acidianus. Of the other positive matches, four were crenarchaeal (all to chromosomal sequences), eight were methanoarchaeal (mainly to phages or prophages), one was haloarchaeal and none were found to the haloarchaeal viruses, including SH1, His1 and His2 (Bamford et al. 2005, Bath et al. 2006) (Table 6).

The positions of the matching spacers for S. solfataricus are indicated in Figures 1B and ​and1C1C and the identities and predicted functions of the matching ORFs are presented in Table 7. Most positive matches are to viruses, in particular to ORFs of the Acidianus bicaudavirus ATV (62,730 bp) (Häring et al. 2005, Prangishvili et al. 2006). Matches also occurred to the similar rudiviruses SIRV1 and SIRV2 (Peng et al. 2001), the betalipothrixviruses AFV6, AFV7 and AFV8 (G. Vestergaard, University of Copenhagen, Denmark, unpublished data), the fusellovirus SSV4 (X. Peng, University of Copenhagen, Denmark, unpublished data) and the icosahedral virus STIV (Rice et al. 2004). Intriguingly, the cluster within the plasmid pKEF9 yielded matches with the rudivirus SIRV (Peng et al. 2001) and fusellovirus SSV5 (B. Greve, University of Copenhagen, Denmark, unpublished data; Table 7).

We thank Jan Christiansen, Elfar Torarinsson, Qunxin She, Xu Peng and Gisle Vestergaard for helpful discussions and Hien Phan for help with DNA sequencing. The research was supported by grants from the Danish Research Council for Natural Science. Grants from Copenhagen University supported K. Brügger and P. Redder, with K. Brügger also receiving grants from the Danish Science Research Council.