DNA methylation

Using the PacBio reads and the assembled genome sequence, base modifications were analyzed using the SMRT^® Analysis software (Basemods tool) (Chin et al., 2013). All replicons contained methylated C residues (^m4C) at position 1 of the tetranucleotide sequence CTAG, on both strands. Methylation was estimated to be present at >90% of sites. The CTAG motif was significantly under-represented in all three replicons; a feature that is commonly found in many haloarchaeal genomes (Fullmer, Ouellette, Louyakis, Papke, & Gogarten, 2019). For example, there were only 1,430 sites on the chromosome (odds ratio = 0.37). Methylation is probably carried out by the Zim CTAG modification methylase (HBSAL_08190), a homolog of the methylase (HVO_0794) described for Hfx. volcanii (Hartman et al., 2010; Ouellette, Gogarten, Lajoie, Makkay, & Papke, 2018). The distribution of CTAG motifs around the chromosome and plasmids of strain 91-R6 is indicated in Figure 1 (see below).

Overall structure of the replicons

For the chromosome, a cumulative GC-skew plot (Figure 1, inner-most ring) shows an overall trend of increasing GC-skew while moving clockwise from the top, around the circle, and back to the top, with a strong inflection near the canonical replication origin (point of ring opening). This general pattern is similar to many bacterial genomes, where the major inflection point indicates the position of the replication origin (Lobry & Louarn, 2003). Variations in GC content (7th level, black plot) often coincide with disturbances of the GC-skew, as is seen across the single rRNA operon found close to and pointing away from the ori. Other, more extended regions of lower GC show higher densities of both MGEs (4th level, gray arrows) and CTAG motifs (3rd level, blue lines). A BLASTn comparison to strain R1 (6th level, pink) highlights the close similarity between the two strains, with only three large interruptions (labeled divSEGs 04, 12 and 18). Predicted genomic islands (GIs; 5th level, brown) are correlated with these divSEGs and represent likely regions of horizontally acquired DNA, and show the typical features of high levels of MGEs and lower than average GC. They also have a higher density of CTAG motifs. In summary, the chromosome appears to have an underlying organization, as evidenced by the cumulative GC-skew, interspersed by large genomic islands (HGT) and smaller indels.

Both plasmids (Figure 1, right side) have a reduced GC content compared to the chromosome; 6.5% less for pHSAL1 and 10.6% for pHSAL2 (Tables 2 and 3). The BLASTn rings of each map (4th level, pink) display the sequence similarity to the other plasmid, clearly revealing the 39.2 kb of sequence that they share in common. The unique region of plasmid pHSAL1 (107 kb) is near-identical to part of R1 plasmid pHS3 (see Figure A1 in Appendix 2).

Ribosomal RNA and tRNA genes

Strain 91-R6 has a single rRNA operon and 48 tRNA genes, all carried on the main chromosome. The rRNA operon has the typical bacterial gene order (Hui & Dennis, 1985): 16S–tRNA^Ala(UGC)–23S–5S–tRNA^Cys(GCA), an arrangement noted previously in strains R1 and NRC-1 (Ng et al., 2000; Pfeiffer, Schuster, et al., 2008). The 16S and 5S rRNA sequences are identical to those of the R1 strain, while the 23S rRNA sequence differs by a single base change (nt 2,890, C/T). There are tRNAs for all 20 amino acids. Three tRNA genes contain predicted introns: tRNA^Ile(CAU), tRNA^Trp(CCA), and tRNA^Met(CAU). The only tRNA difference between strains 91-R6 and R1 is that strain 91-R6 carries an extra (although partial) copy of tRNA^Gly(GCC) at nt 1,621,908–1,621,851, adjacent to a 7.5-kb indel (divSEG30, see below, Section 3.2).

Virus defence systems and prophage genes

No CRISPR regions or cas genes were detected in strain 91-R6. The R1 and NRC-1 strains also lack CRISPR-Cas genes (Ng et al., 2000; Pfeiffer, Schuster, et al., 2008). A search for other species of this genus that have sequenced genomes found complete CRISPR-Cas regions in two (Hbt. hubeiense and Halobacterium sp. DL1), a partial (and nonfunctional) system in one (Halobacterium jilantaiense), and none in Halobacterium noricense CBA1132. Recently, a distinct virus defense system has been identified in bacteria, the BREX system (Goldfarb et al., 2015), which is also present in many haloarchaea, including strain R1, where it is located on plasmid pHS3. Goldfarb et al. classified the haloarchaeal BREX system as “type 5.” Among the variations specific for this type, they identified a helicase domain gene, denoted as brxHII. While they were able to identify a helicase BrxHII in Haloarcula hispanica (HAH_4399), they did not identify this gene in Hbt. salinarum strain R1 (Goldfarb et al., 2015). The reason is that the gene (OE_5343R) is disrupted by transposon targeting and thus is not included in the protein sequence databases. BrxHII disruption may render the BREX system of Halobacterium nonfunctional, and this may be the reason why strain R1 (and its derivative S9) is susceptible to attack by viruses like phiH1 or ChaoS9 (Dyall-Smith et al., 2019; Dyall-Smith, Pfeifer, Witte, Oesterhelt, & Pfeiffer, 2018). This region of pHS3 is missing in strain NRC-1, which thus is devoid of a BREX system.

In strain 91-R6, distant homologs of the strain R1 BREX system proteins were identified, encoded by a cluster of closely spaced genes (brxABC and pglXZ; HBSAL_05050 to HBSAL_05080) on a strain-specific sequence of the chromosome (divSEG12, see below, Section 3.2). In strain 91-R6, no homolog to OE_5343R could be identified and no other helicase domain protein is encoded in the genomic vicinity to the BREX system. The pglX gene (DNA methyltransferase) is disrupted, and methylation analysis of the SMRT data did not indicate any motifs with methylation of A residues. At present, it is unclear whether the BREX system in this strain is functional.

Prophage prediction tools did not identify any integrated proviruses, but several strain-specific regions have characteristics which are typical for integrative elements (strain-specific regions with integrase genes in close vicinity to tRNA genes or having targeted a protein-coding gene and being bounded by a direct repeat) (see below, divSEG14, the divSEG15/16/17 trio, divSEG30, and divSEG31).

Opsin genes

Strain 91-R6 carries one bacteriorhodopsin (bop) gene, one halorhodopsin gene (hop), and two sensory rhodopsins (sopI, sopII). All are carried on the main chromosome along with their associated and regulatory genes (e.g. bat, bap, blp), and all are present in genomic regions strongly related to strain R1. The bop gene has a short insertion and may not be functional (see later, Section 3.2, divSEG27).

Motility genes

Archaellins (flagellins), the structural genes of the Halobacterium archaellum (flagellum), are encoded by a multigene family, and while the archaellin (flagellin) genes arlB1-B3 (previously flgB1-B3) are encoded in the type and both laboratory strains in immediate genomic vicinity to the motility (Arl, previously Fla) and chemotaxis (Che) clusters, the arlA1A2 (flgA1A2) gene pair of strain R1 is not associated with other motility or chemotaxis genes. Instead, this gene pair is encoded on a strain-specific sequence, as is a single arlA gene in strain 91-R6. Both arlA loci occur on divSEG18 (see below, Section 3.2). The protein sequence of ArlA is distinctly different from the homologs of other sequenced species, and by BLASTp was most similar (89% protein sequence identity) to ArlA (FlaA) of Hbt. jilantaiense (accession SEV92461.1).

N-glycosylation

In addition to the S-layer glycoprotein, there are many other haloarchaeal proteins that are known to be N-glycosylated, such as archaellins and some pilins (Jarrell et al., 2014). The pathway of N-glycosylation in Hbt. salinarum has also been studied (Kandiba & Eichler, 2015). Several enzymes of the N-glycosylation pathway (aglF, aglG, aglJ, aglM, aglR) are encoded as distant homologs on strain-specific regions (on divSEG18, see above “Motility genes” and below, Section 3.2). Strain 91-R6 lacks a close homolog of aglE. Additional glycosyltransferases are encoded in both strain-specific regions. The last bases of divSEG18 code for the N-terminal 18 codons of aglB (44% protein sequence identity), while the remainder of the protein is encoded on the subsequent matchSEG (98% protein sequence identity).

Biofilm formation

Strain 91-R6 is known to display a strong ability to form biofilms (Fröls et al., 2012; Losensky et al., 2017; Losensky, Vidakovic, Klingl, Pfeifer, & Frols, 2015). By comparison, strain R1 is nearly as proficient while strain NRC-1 shows negligible ability under the laboratory conditions tested. The close similarity of the genome sequences of these strains and their wide difference in biofilm phenotype attracted our attention, providing a basis for speculating on the genetic basis of biofilm formation in this species.

In Hfx. volcanii, PilA pilins are required for surface adhesion (Esquivel, Xu, & Pohlschroder, 2013). Several pilins of Haloferax are N-glycosylated, and interference with glycosylation has been shown to modify pilus assembly and function (Esquivel, Schulze, Xu, Hippler, & Pohlschroder, 2016). The six characterized PilA proteins of Haloferax share an identical 30 amino acid H-domain of their type III signal sequence. This represents a specific subtype of the more general Pilin_N (previously DUF1628) domain (PFAM:PF07790).

Strain 91-R6 has four proteins with an assigned DUF1628 domain, each with an ortholog in strains R1 and NRC-1. Strains R1 and NRC-1 have one additional, plasmid-encoded paralog. Curiously, only one DUF1628 domain protein (HBSAL_01455) has a type III signal sequence H-domain that is highly similar to Haloferax PilA. There are 22 strictly conserved residues, followed by relaxed similarity (three point mutations in eight residues). In the nonadhesive strain NRC-1, the ortholog is disrupted by transposon targeting (VNG_0110d + VNG_0112a), while the corresponding genes are intact in the adhesive strains R1 (OE_1186A1F) and 91-R6 (HBSAL_01455). The proteins from strain R1 and 91-R6 show 93% protein sequence identity, are identical in length (122 aa), and have four potential N-glycosylation sites. In Haloferax, the pilB3C3 gene pair has been identified as the PilA pilus assembly machinery (Esquivel & Pohlschroder, 2014). This assembly machinery is not clustered with its target genes, which in turn are not clustered with any assembly machinery. Most other pilBC assembly genes are in operons which also code for proteins with a type III signal sequence. For Halobacterium strain R1, it was shown that cells displayed a ten-fold reduction in glass adherence when the pilB1 gene was deleted (Losensky et al., 2015). Halobacterium pilB1 (OE_2215R, HBSAL_04190) is the ortholog of Haloferax pilB3 (HVO_1034) (same for Hbt. pilC1, OE_2212R, HBSAL_04185, vs. Hfx. pilC3, HVO_1033). From these analyses, we conclude that Halobacterium pilB1C1 is the assembly machinery for a nonclustered PilA pilin and that this PilA pilin mediates cell adhesion and the biofilm phenotype.

The enhanced biofilm formation properties of strain 91-R6 compared to R1 may be mediated either by protein sequence differences or by alterations in their N-glycosylation pathways (see above, N-glycosylation).

Amino acid biosynthesis genes

Halobacterium salinarum strain R1 (and NRC-1) is reported to be auxotrophic for several amino acids, including leucine and isoleucine (Falb et al., 2008; Gonzalez et al., 2009). However, strain 91-R6 codes for several genes of leucine and isoleucine/valine biosynthesis, specifically, leuABCD and ilvBCDN. The four genes ilvBCDN (within divSEG18, see below, Section 3.2) code for three enzymes with relaxed substrate specificity that catalyze equivalent reactions within the biosynthetic pathways of both isoleucine and valine. Consistent with bioinformatic reconstruction, strain 91-R6 grows well in the absence of leucine, isoleucine, and valine (Figure A3 in Appendix 2). While strain R1 did not grow in the absence of leucine, we observed growth in the absence of isoleucine and valine. This discrepancy between bioinformatic reconstruction and experimental results is yet unresolved. Besides the differences in isoleucine/valine and leucine biosynthesis genes, we did not detect any other differences in amino acid metabolism.

Overall similarity between the chromosomes from the three strains of Hbt. salinarum and other species from the genus Halobacterium

The similarity between the chromosome of the type strain and the two laboratory strains was examined by in silico DNA–DNA hybridization (DDH) and average nucleotide identity (ANI), and the results are summarized in Tables A1 and A2 (Appendix 1). The type strain showed DDH values of 95% and ANI values of 98% to the laboratory strains, well above the accepted thresholds for membership of the same species (70% DDH; 95%–96% ANI) (Chun et al., 2018; Oren, Ventosa, & Grant, 1997). The ANI values also indicated a high level of sequence conservation between the strains. When compared to other recognized species of the genus Halobacterium, the type strain exhibited far lower DDH (<25%) and ANI (<81%) values, consistent with the current classification.

Outline of the procedure for detailed comparison of the chromosomes

The chromosome comparison strategy used here is the same as previously developed and applied to strains of P. laumondii (Zamora-Lagos et al., 2018). The sequence alignment program MAFFT (Katoh & Standley, 2013) was used to delineate matching segments (matchSEGs) which are common to both strains and divergent segments (divSEGs) which represent strain-specific genome regions (see the legend to Table 4 for details). In this way, genome sequences can be partitioned so that consecutive regions toggle between matchSEGs and divSEGs.

Matching genome segments between the chromosomes of strain 91-R6 and strain R1

Alignment of the chromosomes of strains 91-R6 and R1 revealed they are highly similar and completely colinear (Figure 1, Table 4), with an overall sequence identity of 99.63%. There are 39 matching segments (matchSEGs) that together cover the majority of both chromosomes (1.85 Mb; 84.9% for the Hbt. salinarum strain 91-R6 chromosome and 92.5% for strain R1), and between these are 38 strain-specific sequences. Thirty of the 39 matchSEGs show <1% sequence divergence, while the remaining nine matchSEGs have more than 1% sequence difference (average 1.47%) but are relatively short (90 kb total). Overall, 6,719 point mutations and 87 small indels were detected in the 39 matchSEGs (65 indels < 20 nt, longest indel 79 nt).

Strain-specific regions in the chromosomes of strains 91-R6 and R1

The strain-specific sequences (referred to as divergent segments, divSEGs) sum up to 328,119 bp for Hbt. salinarum strain 91-R6 (15.1% of its genome) and 150,261 bp for strain R1 (7.5% of its genome).

DivSEGs were classified into two categories, indels and replacements (Table 4). Indels refer to sequences that are contiguous in one genome while the other has an insertion of additional sequence, and where this insertion can be pinpointed to an exact position. There are 18 insertions in strain 91-R6 and 10 insertions in strain R1. Several of the insertions are MGEs, which are described in more detail below (Section 3.3). Replacements are where the two strains have dissimilar sequences located at an equivalent position, and the borders can be discerned with 1-base resolution. A total of 10 replacements were detected. Most of the sequences in replacements were completely unrelated between the strains. For sequence regions longer than 1 kb, we found an upper limit of 80% DNA sequence identity, which indicates independent sequences in a genome context with >99% DNA sequence identity. The locations of most divSEGs are visible in Figure 1 as white gaps in the BLASTn ring of the chromosome. Only three divSEGs exceed 10 kb, and these represent the three large GIs detected by Island Viewer (see below and Figure 1).

Correlation of protein-coding genes among the three analyzed strains

As an annotation principle, every gene encoding a protein on a matchSEG in one strain must have a correlated gene in the other strain. The gene sets of the three strains have been correlated in detail (see Section 2 and Appendix 4 and Appendix 5). Proteins are classified as strain-specific only after validation by tBLASTn that they are not mere missing gene calls. Correlated proteins encoded on the chromosome of strains 91-R6 and R1 are listed in Table S1 (1,986 proteins) (via Zenodo; https://doi.org/10.5281/zenodo.3528126). The corresponding proteins from strain NRC-1 are also listed. In addition, there are regions of very high similarity between the plasmids from strains 91-R6 and R1 (see below), and the resulting plasmid-encoded correlated proteins are listed in Table S2 (via Zenodo). Furthermore, Table S2 lists proteins which are encoded on a plasmid in strain R1 but in a strain-specific region of the chromosome from strain 91-R6 (see below). Chromosomally encoded proteins from strain-specific regions of 91-R6 are listed in Table S3 (via Zenodo). In several cases, a homolog exists in R1, but the genes are not positionally correlated. Such homologs are also listed in Table S3. Residual proteins which are specific for the chromosome of strain R1 are listed in Table S4 (via Zenodo). Plasmid-encoded proteins specific for strain 91-R6 are listed in Table S5 (via Zenodo), while plasmid-encoded proteins specific for strain R1 are listed in Table S6 (via Zenodo). Some of the strain-specific plasmid proteins are discussed in more detail (see below). A few protein-coding genes exist in strain NRC-1 but are absent in both strains 91-R6 and R1. These are listed in Table S7 (via Zenodo). Finally, Table S8 (via Zenodo) lists ORFs which are annotated in the current version of the NRC-1 genome (AE004437, AE004438, AF016485) but which are considered not to code for a protein (spurious ORFs; for definition, see Appendix 4).

Proteins encoded on strain-specific chromosomal regions

The characteristics of the longest divSEGs are described here. For an analysis of other divSEGs see Appendix 7. The three longest divSEGs (04, 12, 18) are GIs (GI-1, GI-2, GI-3) (as indicated in Figure 1).

DivSEG04 is a replacement where the R1 sequence is 61,595 bp long and represents the well-known “AT-rich island” (GC content reduced to 56.1%) (Joshi, Guild, & Handler, 1963; Moore & McCarthy, 1969; Ng et al., 2000; Pfeifer & Betlach, 1985). At the equivalent position in strain 91-R6 is a 47,062 bp region, which also has a reduced GC content (56.3%), and both are likely to represent horizontally transferred DNA. Both regions carry many mobile genetic elements, at least one Orc paralog, and are rich in glycosyltransferases and other sugar metabolism related genes. The DNA sequences are mostly unrelated (only two BLASTn matches exceed 1 kb), but eight encoded proteins are homologous and show up to 86% protein sequence identity. Nearby and upstream of this region are genes for the S-layer glycoprotein (HBSAL_01075), secreted glycoproteins (HBSAL_01070, 01065), and sugar nucleotidyltransferases (HBSAL_01110, 01105), and a potential role for this replacement region is to provide an altered repertoire of sugars for modifying secreted glycoproteins (e.g. S-layer) and extracellular polysaccharides (EPS), perhaps to avoid virus predation. We propose that this replacement region be called genomic island 1 (GI-1) (Figure 1).

DivSEG12, which corresponds to genomic island GI-2, is the longest strain-specific sequence in strain 91-R6 (164,295 bp). In R1, there is a 2,306 bp region at the same genome position. The R1 sequence codes for most of the alpha subunit of dimethylsulfoxide reductase (dmsA, codons 69–836 of 837), while the N-terminal 68 residues are encoded on the preceding matchSEG. The termini of the 164,295 bp region in strain 91-R6 code for a close homolog of DmsA (57% protein sequence identity) which has been disrupted due to targeting by MITEHsal3. The integration point corresponds to codon 622 of R1 DmsA. The concatenated protein sequence was most similar (78% amino acid identity; BLASTp) to a homolog from Halostella sp. DLLS-108 (accession WP_135820841.1). The long N-terminal fragment (HBSAL_04215) covers the 4Fe-4S (IPR006963) and catalytic (IPR006656) domains. The C-terminal fragment (HBSAL_05135) covers the molybdopterin dinucleotide-binding domain (IPR006657). The 91-R6 specific 164 kb region has a GC content below 60% and carries several Orc paralogs and multiple MGEs, thus showing characteristics of an integrated plasmid. The full copy of a MITEHsal3 at one end and a partial copy (truncated due to MGE targeting) at the other suggests that, after the initial MITE insertion into dmsA, there were further integration events that initially targeted the MITE. This is further supported by a hybrid TSD (TATGACA) around these copies of MTEHsal3. Among the proteins encoded in the 91-R6 specific region are multiple paralogs of TATA-binding transcription factors. Several of the encoded proteins are close homologs of proteins encoded on the plasmids from strain R1 (see below and Table S2 (via Zenodo)). Four sequences, totaling 42.5 kb, show a close but complex relationship to R1 plasmid pHS3 (see below, Figure 2, and Tables 5 and 6).

DivSEG18 corresponds to a replacement where the R1 sequence is 44,146 bp long and the 91-R6 sequence at the equivalent genome position has 78,224 bp. This corresponds to genomic island GI-3. While both sequences have more than 60% GC, it is slightly reduced from the genome average due to the presence of several MGEs. The region from strain 91-R6 contains several transposons which are specific for this strain (canonical transposons ISHsal1, ISHsal2, ISNpe16, and HsIRS45; noncanonical transposons ISHsal5, ISHsal12, and ISHsal14; see below, Section 3.3, Table 7 and Appendix 11). The 2nd ORF in the R1 region is an integrase domain protein, and it may be no accident that just upstream of divSEG18 is a tRNA-Met gene (a typical arrangement for integrative elements). At the genome level, the two sequences show restricted sequence similarity (up to 80% DNA sequence identity for regions up to 3 kb). Although these sequences are strain-specific, they code for distant homologs, and even retain partial gene synteny. In this context, homologs are considered distant even at 85% protein sequence identity, as orthologs within matchSEGs show at least 98% protein sequence identity (with very few exceptions). Remarkably, the only gene copies of triosephosphate isomerase (tpiA), histidinol-phosphate aminotransferase (hisC), and archaetidylglycerolphosphate synthase (agsA) are encoded on divSEG18. Also encoded is GTP cyclohydrolase 3 (IIa), which is involved in riboflavin biosynthesis (arfA1 in both strains, an additional arfA2 paralog only in strain R1). DNA polymerase Y is encoded in this region, but is disrupted in strain 91-R6. Some proteins which are physiologically relevant are encoded by more than one paralogous gene, of which one is located on divSEG18, including a probable adenylate kinase (adk2) and arlA, the gene coding for one of the archaellins (see above, Section 3.1). In R1, htr13, one of the methyl-accepting chemotaxis proteins (haloarchaeal transducers), is encoded close to the arlA locus while a transducer is not encoded on divSEG18 in strain 91-R6. Several enzymes of the N-glycosylation pathway are encoded on divSEG18 (see above, Section 3.1). The last bases of divSEG18 code for the N-terminal 18 codons of aglB, while the remainder of the protein is encoded on the subsequent matchSEG.

Other divSEGs which are briefly described in Appendix 7 are divSEG05, the divSEG22/23 pair, divSEG27, divSEG32, divSEG37, and divSEG39. Several divSEGs code for integrase domain proteins, and some of those have a tRNA gene at or close to the integration point (an arrangement typical of integrative mobile elements). This applies to the divSEG15/16/17 trio and divSEG30. DivSEG30 looks suspiciously like a provirus (7.5 kb long; targets a tRNA; one integrase family gene; 10 other genes, none of them well characterized) but does not match any virus in GenBank. Notably, divSEG14 and divSEG31 also code for an integrase domain protein but have targeted a protein-coding gene and are flanked by direct repeats (8 bp, CTGGCACA and 13 bp, GAACATGGTGTTC, respectively). This is reminiscent of the 10,007 bp insertion in strain NRC-1 compared to R1, which also codes for an integrase domain protein and shows an 8 bp direct repeat.

The unique region of strain 91-R6 plasmid pHSAL1 corresponds to R1 plasmid pHS3

The unique region (107 kb) of plasmid pHSAL1 is exceedingly similar to part of pHS3 from strain R1. Most of this common sequence also occurs on pNRC200 from strain NRC-1 (Tables 9 and 5, tag pp11-pp14). The main differences are due to targeting by MGEs (three events in strain 91-R6, two in R1, and two in NRC-1). Leaving aside MGE targeting (and the deletion of 1.7 kb at the 3′ end in NRC-1), the common region covers 107,860 bp and is almost identical in all three strains, except that the type strain has one point mutation and one indel (four bases, intergenic). Such an extreme conservation is atypical for plasmids. The region is GC-rich (see Figure 1, Figure A1 in Appendix 2, and Table 9), shows a high protein coverage upon proteome analysis in strain R1 (Tebbe et al., 2005), and encodes several essential genes. Accordingly, the megaplasmids carrying this extended region can be considered minichromosomes (Pfeiffer, Schuster, et al., 2008). Among the genes which are encoded in this region, HBSAL_12005 to HBSAL_12615 sequence in strain 91-R6 is the only arginine–tRNA ligase (argS, HBSAL_12475) in each of the three strains. Adjacently encoded is the arginine fermentation cluster (arcDBCAR) which has been characterized in strain R1 (Ruepp & Soppa, 1996; Wimmer, Oberwinkler, Bisle, Tittor, & Oesterhelt, 2008) and is required for Halobacterium bioenergetics (Gonzalez et al., 2008). The chemotactic arginine sensor Car is encoded just beyond this shared region, in a 93.5 kb sequence which is unique to pHS3 (Table 5, tag p3F). The first step of arginine fermentation is the cleavage of arginine into ornithine and carbamoylphosphate. The latter compound is one of the substrates of the enzyme aspartate carbamoyltransferase (pyrBI), which catalyzes the first committed step of pyrimidine biosynthesis and is encoded immediately upstream of argS. This region also codes for other metabolically important enzymes, two of which have been characterized in Halobacterium: alkaline phosphatase, aph, (Wende et al., 2010) and catalase-peroxidase, katG (Long & Salin, 2001). Other important genes are glycerol dehydrogenase (gldA1), the queCED genes required for biosynthesis of the hypermethylated modified tRNA base archaeosine, and a probable siderophore biosynthesis cluster (iucABCD). Finally, this region codes for the so-called “chromosomal” gas vesicle cluster (gvpACDEFGHIJKLMNO) (Surek, Pillay, Rdest, Beyreuther, & Goebel, 1988). The assignment as “chromosomal” was a prediction based on the high GC content and was made before the genome structure had been resolved. The “plasmid” gas vesicle cluster is present in strains R1 and NRC-1 but not encoded in strain 91-R6 (see below). The last part of the 107 kb region shared between strains 91-R6 and R1 has been deleted from the plasmid in NRC-1 (see Table 9, tag pp14). However, upstream of this shared region, R1 and NRC-1 have a long region of 31 kb in common, which is lacking in strain 91-R6 (see Table 5, tag p3A). This region encodes the kdpFABCQ cluster for a potassium uptake system (see below). This region may have been lost in strain 91-R6 during the event which transferred the 3′ terminal part of the 39.2 kb duplication from pHSAL2 to pHSAL1 (Figure A2, junction JA2) (see Appendix 8).

A strain-specific chromosomal region from strain 91-R6 (divSEG12) shows a close but complex relationship to plasmid pHS3 from strain R1

A near-identical sequence of 42.5 kb (having only two point mutations) is shared between the 164 kb strain-specific chromosomal sequence divSEG12 and R1 plasmid pHS3. However, this sequence has become “scrambled” into four fragments by MGE targeting, genome inversions, and strain-specific deletions (Table 5, tag p3I, J, K, L; Table 6, tag c10,c11,c13,c16, Figure 2). The underlying evolutionary history and processes could be discerned by junction analysis, taking into account targeted or truncated genes as well as “hybrid TSDs” (target site duplications which became disjunct by a subsequent genome rearrangement). This analysis also uncovered a 3 kb extension of the shared sequence which, however, has been shifted from chromosomal divSEG12 to the duplicated part of pHSAL1/pHSAL2 (Table 5, tag p3G, H, see below). Full details are provided in Appendix 8 (junctions JC1 and JC2, see also Figures 3 and 4).

In Figure 2, the first two fragments of 2.1 kb and 6.1 kb are contiguous in pHS3 (Table 5, tag p3I, p3J) but dislocated and inverted in divSEG12 (Table 6, tag c10,c16) (Figure 2, junction JB1). Disconnection in divSEG12 is attributed to MGE targeting (ISH3C) with a subsequent MGE-triggered genome inversion. This attribution is supported by a hybrid TSD. PacBio reads reveal heterogeneity with respect to this inversion, both orientations being frequent in the population with support from at least 70 reads (see also below, Section 3.3, Figure 5, and Appendix 8 and Appendix 10, case D). (b) On divSEG12, the 6.1 kb fragment is contiguous with a 24.6 kb matching sequence (Table 6, tag c11) which is dislocated and inverted on pHS3 (Table 5, tag p3L; Figure 2, junction JB2, Figure 6). Disconnection and inversion in pHS3 is attributed to MGE targeting (ISH8B) with a subsequent MGE-triggered genome inversion. This attribution is supported by (i) two hybrid TSDs and (ii) one pair of disrupted proteins. OE_5405F and OE_5013R together correspond to HBSAL_04690 and are a full-length homolog of Halxa_0005. The N-terminal fragment, corresponding to OE_5405F, has been lost from strain NRC-1, while the C-terminal fragment, corresponding to OE_5013R, has been retained (VNG_6145a) (see below; for further details see Appendix 8). Finally, (iii) there is a disrupted transposon ISH32 where the two disconnected fragments together represent one complete MGE. (c) The 24.6 kb sequence p3L traverses the point of ring opening in pHS3. The ring opening point is associated with a discontiguity between R1 plasmid pHS3 and NRC-1 plasmids pNRC100 and pNRC200. While regions p3L and c11 are colinear between strains 91-R6 and R1, part of this sequence has been lost from NRC-1 (1,065 bp; including the equivalent to OE_5405F) and part has been shifted to pNRC100 (16,511 bp; reverse orientation). The region shared between divSEG12, pHS3, and pNRC100 codes for an arsenic resistance cluster which has been characterized (Wang, Kennedy, Fasiludeen, Rensing, & Dassarma, 2004). (d) The next and last common fragment has 6.3 kb (Table 5, tag p3K; Table 6, tag c13) but is still inverted on pHS3. In divSEG12, these fragments are separated by a 8.2 kb strain-specific sequence (Table 6, tag c12) with just one ISH2 element at the corresponding position in pHS3 (Figure 2, junction JB3). This is attributed to MGE targeting (integration of two copies of ISH2) with subsequent recombination of these MGEs, resulting in deletion of the intervening sequence. This attribution is supported by the absence of a TSD around the ISH2 and by completeness of HBSAL_04810 while its R1 equivalent OE_5019R is truncated at the ISH2 element and does not continue on the other side.

The sequence between the last common fragment and the dislocated and inverted first fragment on divSEG12 is separated by 12 kb (Table 6, tag c14 + c15 + MGE:ISH3). This region consists of a 6,038 bp sequence with 87% DNA sequence identity to part of the duplication between R1 plasmids pHS1 and pHS2. The adjacent 4,509 bp are specific for strain 91-R6, followed by the MGE of subtype ISH3C which is involved in the inversion.

Targeted pseudogenes start in the strain-specific chromosomal region from strain 91-R6 (divSEG12) but continue in the region duplicated between pHSAL1 and pHSAL2

Strain 91-R6 plasmid pHSAL2 shows partial matches to R1 plasmid pHS1

The 102 kb plasmid pHSAL2 consists of a unique region (63.4 kb) and shares a 39.2 kb duplication with pHSAL1. Based on junction analysis, the 3′ end of the duplicated sequence belonged originally to pHSAL2 and has been transferred to pHSAL1 (see Appendix 8, junction JA2, and Figure A2 in Appendix 2). In the duplicated region is a 3.1 kb match (100% DNA sequence identity) to R1 plasmid pHS3 (see above). Even though duplications occur in the plasmids from all three analyzed Halobacterium strains, the duplications from the strain 91-R6 plasmids do not overlap with those from the R1/NRC-1 plasmids. At less than 90% DNA sequence identity, regions have to be considered independent. There is only one such homolog (2.3 kb, 83% identity) which occurs on duplicated regions of all three strains and codes for subunits of cytochrome bd ubiquinol oxidase (cydAB). The residual 28.2 kb of the duplicated as well as 45.4 kb of the pHSAL2 unique region are restricted to strain 91-R6. pHSAL2 has a 13.3 kb match to pHS1 (duplicated on pHS4; Table 8, p205 and p206) which consists of a 8.3 kb region with just three point mutations and one inserted MGE, while the remaining 5.0 kb have only 91% DNA sequence identity. This is reminiscent of a 30 kb duplication with sequence identity between R1 plasmids pHS1 and pHS4 and an adjacent 10 kb duplication which is much more dissimilar (98.5% DNA sequence identity). There is one additional 3.8 kb match (96% DNA sequence identity) between pHSAL2 and pHS1, but regionally disconnected in both strains.

Strain-specific proteins from plasmids of strain R1

Strain-specific regions from plasmids of strain R1 (and NRC-1) code for several Orc paralogs and also contribute to the multiplicity of basic transcription factors (several paralogs of tfb and tpb genes). The “plasmid” gas vesicle cluster is specific to strains R1 and NRC-1. This cluster has been extensively characterized in strain NRC-1 (DasSarma, 1989, 1993; DasSarma, Damerval, Jones, & Tandeau, 1987; DasSarma et al., 1988, 2013; Halladay, Jones, Lin, Macdonald, & Dassarma, 1993; Halladay, Wai-Lap, & Dassarma, 1992; Jones et al., 1989; Jones, Young, & Dassarma, 1991; Tavlaridou, Faist, Weitzel, & Pfeifer, 2013; Tavlaridou, Winter, & Pfeifer, 2014; Winter, Born, & Pfeifer, 2018). Also, the kdp-type potassium-transporting ATPase (kdpABCF) is encoded on a plasmid region which is present only in R1 and NRC-1 (Kixmuller & Greie, 2012; Kixmuller, Strahl, Wende, & Greie, 2011; Strahl & Greie, 2008). Chemotactic sensing of arginine, which is mediated by the transducer encoded by car (Storch, Rudolph, & Oesterhelt, 1999), is exclusive to strain R1 as it is encoded on a region of pHS3 which is neither represented in strain 91-R6 nor in strain NRC-1.