Strains of Haloquadratum were isolated from salt water samples using the extinction dilution method described in (Burns et al., 2004; Dyall-Smith, 2009). Hqr. walsbyi strain LT53-19 was isolated from a LT brine sample (taken on January 4, 2009) provided to MDS by Dr. Karla Heidelberg (University of Southern California, United States). Strain LT53-19 was sent to, and is available from, Dr. Eric Allen (University of California, San Diego, United States). The Bajool9 strain of Haloquadratum was isolated in the MDS laboratory in 2007 as part of the study reported in (Oh et al., 2010), and was mentioned later in (Dyall-Smith et al., 2011) and deposited in the Japan Collection of Microorganisms (accession JCM 15065). It was recovered from a brine sample taken from a crystallizer pond of the Bajool saltern in Queensland, Australia. The coordinates of LT and the Bajool saltern are given in Table 1.

The sequences of plasmids PL6A and PL6B (accessions FR746101 and FR746102) were used to design consensus PCR primers for detecting and amplifying related plasmids in other strains of Haloquadratum. Amplified products were sequenced [ABI PRISM BigDye terminator method, performed at the Core Facility of the Max-Planck-Institute of Biochemistry (Dyall-Smith et al., 2011)] and the resulting sequences used to design primers for further rounds of PCR/sequencing or primer-walking on purified plasmid, until plasmid closure. The sequences have been deposited under accessions LT984489 and LT984491.

The LT metavirome data (accessions SRR402039 and SRR402041-47) were downloaded from the GenBank sequence read archive (SRA), imported into Geneious (v10.2.3) and the reads mapped to the pLT53-7 plasmid sequence (Geneious mapper). Reads mapping to that plasmid were then de novo reassembled (Geneious assembler) and the longest contig (which also had the greatest read coverage) was used for further rounds of mapping against the metavirome reads in order to extend the contig ends. The process was repeated for several rounds until closure. The final sequence was checked manually at every base against the mapped reads, in order to identify and correct errors. Annotation was performed in the Geneious (v10.2.3) environment using plasmids PL6A, PL6B and pLT53-7 for reference. The sequence has been deposited under accession LT991975.

Spacer matches to PL6-related plasmids were sought from multiple sources; (1) the CRISPR finder database of spacers¹; (2) the prokaryotic spacers collection produced as part of the study by (Shmakov et al., 2017); (3) spacers extracted from metagenomic data available at the SRA² using the crass v0.3.12 software (Skennerton et al., 2013), i.e., metagenomes were searched for using keywords (hypersaline, saltern, LT, crystallizer, and salt lake) and the appropriate data downloaded and processed for spacers on a laptop computer. The extracted spacers were then transformed into a BLAST database (in Geneious) and used to search for matches to PL6-family plasmid sequences. The crass software also outputs the direct repeats (DR) associated with the discovered spacers.

Two novel Australian isolates of Haloquadratum (Table 1) were each found to carry a PL6-like plasmid, designated pBAJ9-6 and pLT53-7. One isolate (LT53-19) was recovered in 2009 from LT (Victoria) and the other (Bajool9) in 2007 from a saltern crystallizer pond near Bajool (Queensland). An agarose gel profile of the Bajool9 plasmid has been previously reported [see Figure 2A of (Dyall-Smith et al., 2011)]. The sequences of both plasmids were determined and compared to the PL6 plasmids (PL6A, PL6B) of Hqr. walsbyi C23^T (Dyall-Smith et al., 2011), an isolate recovered in 2004 from the Cheetham saltern in Geelong, Victoria (Burns et al., 2004; Burns et al., 2007). The sites of isolation of all these strains are geographically well separated, e.g., Geelong, on the south coast of Victoria, is 348 km from LT, and 1727 km from Bajool in Queensland. The coordinates of all sites are given in Table 1.

The general properties of pBAJ9-6 and pLT53-7 are given in Table 1, along with the previously published details of plasmids PL6A and PL6B. Also shown in Table 1 is a PL6-like plasmid, pLTMV-6, that was reconstructed from LT metavirome data (see section “Materials and Methods” and below). The five plasmids were of comparable size (5.8 – 7.0 kb) and their GC contents spanned a narrow range (51–53%) but were significantly above that of the chromosome of Hqr. walsbyi (47.8% GC, (Dyall-Smith et al., 2011)). Multiple alignment using the MAUVE algorithm (Darling et al., 2004) (Figure 1) gave pairwise similarities of 61–79% (average, 69%), with the region from nt 1 to about 3.2 kb showing a higher average sequence similarity (84%; as indicated by the MAUVE plots), compared to the sequences from 3.2 kb to the right end (54%). This characteristic was noted previously for PL6A and PL6B (Dyall-Smith et al., 2011). The split into two subregions is also obvious from the coding potential. Two sets of protein-coding genes are transcribed in opposite orientation from an intergenic region which traverses the point where the plasmid sequences have been linearized. There are three strongly conserved and closely spaced protein coding genes encoded on the forward strand (F1–F3) in the first half of these plasmids. All plasmids have four protein coding genes encoded on the reverse strand (R4–R7), covering the second half of the plasmid.

Phylogenetic tree reconstructions (MrBayes) based on nucleotide alignments produced clades that corresponded to the host strain and/or geographical origin (Figure 1, upper left panel), i.e., plasmids PL6A/PL6B of Hqr. walsbyi C23^T formed one clade, and the LT plasmids formed another clade, while the plasmid from the Bajool saltern in Queensland (Oh et al., 2010) branched as a distinct lineage.

While the average GC% for all plasmids was 52%, plots of GC% (Supplementary Figure S1, blue lines) revealed a consistently higher than average value (55%) across the first ∼3.2 kb, a region encompassing genes F1–F3. Immediately following this is a distinctly AT-rich central region from ∼3.2 kb to the end of R7 (41–43% GC), and from R7 to the start of R4 the values fluctuated but overall were around the average for the entire plasmid (51%). Of the various nucleotide skew plots, GC profile plots (Gao and Zhang, 2006) gave good discrimination between different plasmid regions (Supplementary Figure S1), with minima near or within the F1 gene rising steadily to maxima at about 3.1 kb followed by a steep drop in the AT-rich central, intergenic region between the C-termini of F3 and R7. After this, in most cases, there is a gentler drop to the end. The annotated gene diagrams below the x-axes show how these plots correspond to the major gene blocks. pLT53-7 differed slightly from the others in having a very regular rise and fall around the maximum followed by a gentle rise (rather than a drop) to the end. The altered pattern is at least in part due to an extended AT-rich central region (GC% = 37.8) that corresponds with a methylase gene found between F1 and R7 only in this plasmid, and transcribed in the same direction as the F1–F3 protein encoding genes.

Publicly available metagenomic sequences from hypersaline lakes were searched to detect sequences matching those of the PL6-family plasmids and were readily detected in the large datasets available for LT (Australia), Santa Pola (Spain) and Lake Meyghan (Iran) (Table 2). The highest counts were found in the LT data with LT metavirome sequences giving higher counts compared to LT metagenome data. However, it should be noted that this increase (2–3 fold) is paralleled by an even higher increase in total read number (∼14-fold). The plasmid pLT53-7 matched the most metavirome reads (>14,000), and this plasmid originated from a Haloquadratum strain isolated from the same lake. The abundance of metavirome reads allowed the assembly of a 5.88 kb, circular, PL6-family plasmid from this data (see section Materials and methods), that was designated pLTMV-6 and included in the following analyses.

To simplify description, the major protein coding genes shown in Figure 1 have been numbered F1–F3 for genes encoded on the forward strand from left to right, and R4–R7 for genes encoded on the reverse strand from right to left (Figure 1, bottom level). The characteristics of the predicted proteins are summarized in Table 3 and protein alignments are given in Supplementary Figures S2, S3.

The CRISPR spacers of available haloarchaeal genomes and metagenomes were searched (see section Materials and Methods) for matches to PL6-family plasmids. Three spacers were found that showed moderate nucleotide similarity (Table 4); one to each of the F2 and F3 genes, and the third overlapped the start of F3.2, a small gene downstream of F3. Two spacers were from metagenomic data of Lake Meyghan, and one was from the genome of Haloferax alexandrinus (Khelaifia et al., 2017). The direct repeats (DR) flanking the Lake Meyghan metagenome spacers were most similar to those found in Halogeometricum and Natrinema (not shown). Predicted translations of the spacers were compared with the predicted protein sequences of the putative targets of the PL6 plasmids, and the alignments (Table 4, right column) showed a good correspondence. In case 1, the proposed start and stop codons (shown as their complement) are underlined in the nucleotide alignment, and are identical, i.e., an amber (TAG) stop codon, indicated by •_A, two nucleotides from a GTG start codon.

Plasmid sequences were examined for ribosome binding site sequences but good matches to the consensus (GGAGGTGA) were not found near annotated start codons except for the F3 gene, where the sequence GGAGGcGA was consistently located 4 nt upstream of the F3 start codon (i.e., within F2). These results are consistent with previous studies showing that SD sequences are rarely used in Haloferax, and possibly restricted to internal genes of operons (Kramer et al., 2014). Haloarchaeal promoter motifs were also sought, particularly in the region between genes F1 and R4 where outward transcription of the F1–F3 genes and R4–R7 genes is likely to initiate. The consensus promoter motif SRnnRnnnTTWW (Babski et al., 2016) detected 9–13 potential sites per plasmid but only two were both intergenic and conserved in position across all plasmids. These two promoter motifs were overlapping and oriented in opposite directions, and found embedded in a highly conserved sequence (designated CIS I) within the F1 – R4 intergenic region (Supplementary Figure S5A). Nearby is another highly conserved intergenic sequence (CIS II), which could be involved in gene regulation of the F1–F3 operon.

At the other end of the F1 – R4 intergenic region, the R4 CDS was seen to begin with an ATG start codon in every case, and this was supported by LT metagenomic data that mapped to pLT53-7 or pLTMV-6 (data not shown). Leaderless transcripts are known to be very common in haloarchaea and the great majority (94%) of these use an ATG start codon (Brenneis et al., 2007; Hering et al., 2009; Babski et al., 2016), while GTG codons appear to be reserved for poorly expressed genes (Babski et al., 2016). Inspection of the upstream sequences of the R4 genes for potential promoter motifs centered at -27 to -28 (Babski et al., 2016) relative to the A of the initiator revealed inverted repeats centered around -16 to -25 and containing AT-motifs of 4–5 nt in length that were present in every plasmid (red arrows in Supplementary Figure S5A). The R4 proximal inverted repeats may represent regulatory sequences that encompass promoter motifs that are less closely similar to the Haloferax consensus. Additional conserved intergenic sequences (CIS III, IV, and V) were found midway between the F3 and R7 genes and immediately downstream of the R6 coding sequence (Supplementary Figures S5B,C).