Accelerated growth in the absence of DNA replication origins.

DNA replication initiates at defined sites called origins, which serve as binding sites for initiator proteins that recruit the replicative machinery. Origins differ in number and structure across the three domains of life and their properties determine the dynamics of chromosome replication. Bacteria and some archaea replicate from single origins, whereas most archaea and all eukaryotes replicate using multiple origins. Initiation mechanisms that rely on homologous recombination operate in some viruses. Here we show that such mechanisms also operate in archaea. We use deep sequencing to study replication in Haloferax volcanii and identify four chromosomal origins of differing activity. Deletion of individual origins results in perturbed replication dynamics and reduced growth. However, a strain lacking all origins has no apparent defects and grows significantly faster than wild type. Origin-less cells initiate replication at dispersed sites rather than at discrete origins and have an absolute requirement for the recombinase RadA, unlike strains lacking individual origins. Our results demonstrate that homologous recombination alone can efficiently initiate the replication of an entire cellular genome. This raises the question of what purpose replication origins serve and why they have evolved.

Haloferax volcanii is a genetically-tractable archaeon[2,3], its 2.85 Mb main chromosome is replicated from multiple origins[4] using machinery homologous to that found in eukaryotes[1]. To characterize replication dynamics in H. volcanii, we generated replication profiles by deep sequencing the wild isolate DS2 and laboratory strain H26 (Supplementary Table 1). Read counts from asynchronous replicating cells were normalized to non-replicating cells (Extended Data Fig. 1)[5]. Peaks in relative copy number correspond to sequences that are over-represented in replicating cells and therefore identify active origins (Fig. 1). In the wild isolate DS2, peaks at 0 and 1593 kb of the main chromosome co-localize with previously described origins (oriC1 and oriC2, respectively), as do peaks in the mega-plasmid profiles (Extended Data Fig. 2)[4]. The peak at 571 kb represents a third chromosomal origin, oriC3 (Extended Data Fig. 3). Unlike oriC1 and oriC2, oriC3 is not situated at a nucleotide skew inflection point[4]; in bacteria and archaea, such inflection points reflect origin usage over evolutionary timescales[6]. This is consistent with infrequent use of oriC3 or the recent acquisition of an origin at this location.

Figure 1

Replication profiles for H. volcanii wild isolate and laboratory strain

(a) Relative copy number plotted against chromosomal co-ordinate for the main chromosome and pHV4 of wild isolate DS2. Circular chromosomes are displayed linearized at position 0, vertical lines mark replication origins. (b) Sequence reads for laboratory strain H26 mapped to the reference genome (DS2). pHV4 shading reflects chromosomal co-ordinate. (c) Sequence reads for H26 mapped to a reconstructed assembly of the main chromosome with pHV4 integrated at ~249 kb. Grey shading from (b) indicates the orientation of pHV4 integration. (d) Integration of pHV4 into the main chromosome of H26.

The sharp peaks reflect discrete origins, whereas the smooth valleys represent broad zones of termination[7]. Broad termination zones (as opposed to specific termination sites) have been described in other archaea[8] and in eukaryotes[7,9], suggesting they are a feature of chromosomes with multiple origins. The variable peak heights indicate that the chromosomal origins differ in activity, this interpretation is supported by mathematical modeling (A. de Moura, personal communication) and plasmid-based assays (Extended Data Fig. 3c). Such a functional hierarchy of origins may be due to different usage and/or activation times[9,10].

Laboratory strain H26 shows discontinuities in the replication profiles of the main chromosome and mega-plasmid pHV4 (at 249 and 286 kb respectively; Fig. 1b). Discontinuities indicate substantial differences in replication time between adjacent regions and suggest genome rearrangements[11]. We determined this rearrangement to be integration of pHV4 into the main chromosome (Fig. 1c, 1d and Extended Data Fig. 4). Remapping the data to a reconstructed genome sequence results in a continuous profile (Fig. 1c and 3a). The additional peak at 535 kb corresponds to the integrated pHV4 origin, ori-pHV4.

Figure 3

Origin deletion strain replication profiles

Comparison of replication profiles for (a) laboratory strain (wild type), (b) ΔoriC1, (c) ΔoriC2, (d) ΔoriC3, (e) ΔoriC1,2,3 and (f) ΔoriC1,2,3,pHV4 mutants. Relative copy number for the main chromosome with integrated pHV4 was derived and displayed as in Fig. 1c, dashed lines mark the location of deleted origins.

If an origin is active in all cells and used only once per generation, the ratio of origin to terminus regions cannot exceed 2:1; values of >2:1 are only possible if concurrent rounds of replication are initiated. The ratio of the wild isolate is 2:1 (Fig. 1a), but exceeds 2:1 for H26 (Fig. 1c and 3a). This is consistent with concurrent rounds of replication and precludes the existence of alternating phases of replication and segregation in H. volcanii (in contrast to eukaryotes and crenarchaea such as Sulfolobus[12]). Therefore, regulated timing of origin activation is unlikely and the peak height differences we observe are probably due to differences in origin usage.

We tested the requirement for origins by chromosomal deletion in strain H26 (hereafter designated wild-type). All combinations of origin deletion resulted in viable strains, including a strain deleted for all four chromosomal origins (Fig. 2a). Deletion of individual origins led to minor changes in DNA content (notably ΔoriC3), but the strain lacking all chromosomal origins had a DNA content profile indistinguishable from wild-type (Fig. 2b). We used pairwise growth competition to quantify strain fitness (Fig. 2c). Single origin deletion strains grew slower than wild-type, with strains lacking oriC3 exhibiting the greatest growth defect. Surprisingly, the strain deleted for all four origins grew 7.5% faster than wild-type, and the strain lacking the three most active origins (oriC1,2,3) grew 5.5% faster (Fig. 2c). In fact, growth rate correlates inversely with the activity of remaining origins. For example, the ΔoriC2,3 strain retains the most active origin oriC1 and has a 0.8% growth defect, whereas the ΔoriC1,2 strain has lost the two most active origins and has a 2.3% growth advantage.

Figure 2

Characterization of origin deletion strains

(a) Deletion strains were confirmed by hybridization with origin-specific probes (“p” refers to ori-pHV4). (b) Flow cytometry was used to measure DNA content of origin deletion strains, biological replicates are shown; no differences in cell size were observed (data not shown). (c) Pairwise growth competition assays comparing wild-type (H54, bgaHa+) and origin deletion strains. The average and standard error of four independent replicates are plotted.

How could genome replication be maintained despite the deletion of all chromosomal origins? Deletion of known origins might reveal dormant origins as seen in yeast[13,14]. Alternatively, replication could initiate independently of canonical origins, with little or no site specificity. To distinguish between these possibilities, we profiled replication in the origin deletion strains (Fig. 3, Supplementary Table 1). The peaks associated with deleted origin(s) are no longer evident and there are no new discrete peaks. The minima have relocated indicating that there are no enforced termination sites. The profiles of strains deleted for all (or the three most active) origins show a zone of copy number enrichment near the ΔoriC2 locus (Fig. 3e and f; ~2230 kb). However, this does not resemble the sharp peaks associated with characterized origins (Fig. 1 and 3a). Therefore, we find no evidence for activation of dormant origins.

Instead, the profiles are consistent with origin-independent initiation. In contrast to the sharp peaks observed in the wild-type, profiles of the single origin deletion strains exhibit global flattening that has rounded the remaining peaks (Fig. 3b-d); the minima at termination zones are also shallower. Sharp peaks indicate discrete origin sites, therefore peak flattening is a consequence of replication initiation at dispersed sites. The profiles of strains deleted for all (or the most active) origins are largely flat, consistent with widespread origin-independent initiation (Fig. 3e and f).

We considered two mechanisms for dispersed initiation. Origins are binding sites for the initiator protein ORC1; in the absence of origins, ORC1 could bind non-specifically throughout the genome[15]. Alternatively, dispersed initiation could rely upon homologous recombination. Origin-independent replication can occur when recombination (D-loop) or transcription (R-loop) intermediates are used to prime replication[16,17]. We note that in strains deleted for all or the most active origins, the zone of copy number enrichment at ~2230 kb is near the rrnB rRNA operon (Fig. 3e and f; 2234-2239 kb). Highly transcribed DNA is associated with elevated recombination levels[18], therefore D-loops and R-loops in the rrnB region could facilitate replication initiation. This is analogous to recombination-dependent replication in viruses[17] and to DNA damage-inducible replication in Escherichia coli. The latter is known as ‘stable DNA replication’ and occurs in the absence of oriC or the initiator protein DnaA; instead it relies on recombination catalyzed by RecA to initiate replication[16].

H. volcanii mutants lacking RadA (the archaeal RecA/Rad51 homologue) are viable but defective in recombination. Unlike RecA, RadA does not have a secondary role in activating an SOS response[19]. However, RadA is essential for the replication of pHV2-based plasmids, which do not use ORC-based initiation[20]. We attempted to delete radA from the origin deletion strains using established methods[21]. This was successful in the wild-type and single origin deletion strains, but only a single ΔoriC1,2,3 ΔradA isolate was recovered; this strain had undergone a chromosomal rearrangement involving ori-pHV4 (Extended Data Fig. 5). We were unable to delete radA from the strain lacking all four origins, indicating that recombination is essential in the absence of replication origins. To confirm this, we placed radA under control of a tryptophan-inducible promoter (Extended Data Fig. 6)[22]. In the absence of tryptophan, when this promoter is tightly repressed, wild-type cells with inducible radA are viable whereas origin-less cells fail to grow (Fig. 4).

Figure 4

RadA recombinase is essential in an ΔoriC1,2,3,pHV4 mutant

radA was placed under control of the tryptophan-inducible p.tnaA promoter, in oriC and ΔoriC1,2,3,pHV4 strains (H1637 and H1642). The former grows slowly in the absence of tryptophan while the latter is inviable. Absence of tryptophan does not affect the growth of oriC and ΔoriC1,2,3,pHV4 control strains (H26 and H1546); the ΔtrpA control strain (H53) is auxotrophic for tryptophan.

Work by Kogoma[16] showed that E. coli oriC mutants can use homologous recombination to initiate replication. However these cells exhibit profound growth defects[23]. In contrast, origin-less strains of H. volcanii grow faster than wild-type. Furthermore, recombination-dependent replication in E. coli is only possible in strains harboring suppressor mutations (e.g. sdrA, which stabilizes R-loops by inactivating RNaseHI[16]). We found no mutations in any of the four H. volcanii RNaseH genes. Only five single nucleotide polymorphisms (SNPs) were identified in the strain lacking all origins, and all of these SNPs are already present in the respective parent strains (Extended Data Table 1). Therefore, we find no evidence for suppressors, akin to those reported by Kogoma[23], which are required for growth in the absence of origins.

Our results indicate that it is possible to replicate an entire genome by recombination-dependent initiation alone, with no apparent cost to fitness. How might this be accomplished? In wild-type, binding of ORC1 at origins leads to recruitment of the replicative helicase MCM, which may be rate-limiting for initiation[1]. In the single origin deletion strains, liberation of MCM from deleted origins could stimulate recombination-dependent initiation, resulting in flattening of the replication profiles (Fig. 3b-d). We postulate that the activity of origins correlates with their affinity for MCM. Therefore, deleting an active origin (oriC1) liberates more MCM than deleting a weak origin (oriC3). This liberated MCM is recruited to D-loops and used to initiate recombination-dependent replication. Our observation that growth rate correlates inversely with the activity of remaining origins (Fig. 2c) suggests that recombination-dependent replication is more efficient than origin-dependent replication, but the former has a lower affinity for MCM. Consistent with this, Pyrococcus abyssi MCM is recruited to both the origin and a region containing rRNA and tRNA genes; the latter becomes the major binding site in stationary phase, suggesting liberation of MCM from the origin[24].

What then is the purpose of replication origins? It is assumed that regularly-spaced origins ensure genome duplication in the shortest possible time[25]. This assumption is challenged by our data showing that origin-less cells grow faster than wild-type. Alternatively, defined origins can be used to co-ordinate the direction of replication with the orientation of highly-expressed genes. Collisions between replication and transcription machineries can stall DNA replication, and restarting stalled forks by recombination entails a risk of genome rearrangements. We did not observe any such rearrangements, except when the ΔoriC1,2,3 strain was challenged with inactivation of recombination (Extended Data Fig. 5). Moreover, the rapid growth of origin-less mutants suggests that collisions between replication and transcription are less problematic than assumed.

Regulated initiation at origins allows for coordination of genome replication with segregation. This is critical in organisms with tightly-regulated ploidy, such as E. coli, Sulfolobus and most eukaryotes[12]. However, H. volcanii is highly polyploid, tolerates variation in genome copy number[26] and there is no evidence for a regulated cell cycle (I. Duggin, personal communication). We suggest that the high ploidy of H. volcanii enables the accelerated growth of origin-less strains, in contrast to the growth defects observed in E. coli[23]. With a ploidy of 20, H. volcanii can rely on stochastic partitioning to ensure that daughter cells inherit a genome complement. However, it is vital that these 20 genome sequences are equalized to prevent the accumulation of recessive mutations, and this requires efficient recombination. In yeast, a screen for gene deletions that are lethal in polyploid cells found that almost all such mutations affect genomic stability, notably by impairing recombination[27]. Therefore, polyploidy creates a situation (in yeast) where homologous recombination becomes essential; it follows that naturally polyploid organisms such as H. volcanii are heavily reliant upon recombination. Indeed, radA mutants of H. volcanii suffer a more severe growth defect than recA mutants of E. coli[20].

In H. volcanii, origin-dependent initiation of replication appears to offer no demonstrable advantage; however, cells lacking individual origins are disadvantaged. We propose that origins are selfish genetic elements that ensure their own replication. Over time, origins become integrated with cellular processes such as the cell cycle, to coordinate genome duplication, segregation and cell division; ultimately this results in reduced ploidy. Propagation of selfish elements within a population requires a sexual process and lateral gene transfer by cell mating has been observed in H. volcanii[28]. It is notable that most archaeal origins are adjacent to the gene for their cognate initiator protein ORC1[1]. Such tight linkage, which is typical of selfish elements, ensures that origins acquired by lateral gene transfer can successfully subvert the replicative machinery of their host. This is known as the replicon takeover hypothesis, where the host cell chromosome becomes dependent on extra-chromosomal elements for its propagation[29]. The replicon takeover hypothesis has until now focused on the DNA replication apparatus, but our findings suggest that origins can also behave as selfish genetic elements.

Methods

Reagents

Strains, plasmids, oligonucleotides and probes are given in Extended Data Tables 2-4. H. volcanii was grown as described previously[30]. Pairwise growth competition assays were carried out as described previously[21], except that wild-type and mutant strains were mixed in a 1:1 ratio (for further details see Source Data for Fig. 2c). Tryptophan gradient agar plates[31] were cast from a tapered wedge of Hv-Ca agar[30] containing 0.25 mM tryptophan, which was overlaid with a converse wedge of Hv-Ca agar.

Molecular genetic methods

Transformation of H. volcanii and genomic deletions were carried out as described previously[30]. Standard molecular techniques were used, pulsed field gels were carried out as described previously[21]. Genomic DNA for deep sequencing was isolated from 100 ml Hv-YPC culture in stationary phase (A650 >1) or 1 L in exponential phase (A650 ~0.1) as described previously[30], followed by phenol:chloroform extraction. For flow cytometry, live cells in exponential phase (A650 ~0.1) were stained with acridine orange and immediately analyzed using an Apogee A40 as described previously[26,32]; 50000 cells were counted, doublet signals were removed by gating on peak/area plots and data analyzed using FlowJo (TreeStar Inc.).

SOLiD sequencing and data analysis

Library preparation and sequencing was performed by DeepSeq (University of Nottingham) according to SOLiD instructions. Sequence reads were mapped to the H. volcanii genome (accession numbers: CP001953-CP001957) using BioScope (version 1.3.1). Custom Perl scripts were used to calculate and plot replication profiles as described previously[5]. Deep sequencing data are available here: http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE41961

Identification of SNPs

Single nucleotide polymorphisms (SNPs) present at >50% prevalence and with a coverage of >10x (mean genome-wide coverage is ~150x) are shown in Extended Data Table 1. Five SNPs were identified in the strain lacking all chromosomal origins (H1546) and four of these SNPs are already present in the oriC parent strain (H53). The remaining SNP leads to a predicted glycine to valine change in the hypothetical protein HVO_A0627, this mutation is already present in the ΔoriC1,2 parent (H1340). However, it is absent from its ΔoriC1::trpA+ ΔoriC2 parent (H1293) and both these strains grow at near-identical rates (Source Data for Fig. 2c).

Isolation of oriC3 by genetic screen for autonomously replicating sequences

We previously showed that genetic screens in H. volcanii isolate a single origin at a time, and that this can be circumvented by using origin deletion mutants[4]. Therefore, we deleted oriC1 in a Δori-pHV1 background, so that genetic screens would not be dominated by these two origins. Note that ori-pHV1 was previously named ori-pHV1/4; this sequence hybridizes to two bands of ~690 kb and 86 kb on a pulsed field gel, which correspond in size to pHV4 and pHV1, respectively[4]. However, it is now clear that pHV4 has integrated on the main chromosome in laboratory strains, therefore ori-pHV1/4 cannot be present on pHV4. Consequently, we have renamed this origin as ori-pHV1 and renamed ori-pHV4-2 as ori-pHV4. To delete oriC1, the EcoRI-BspEI oriC1 duplex unwinding element (DUE) fragment of pTA441 was replaced with the BamHI-XbaI hdrB selectable marker from pTA187 to generate the oriC1 deletion construct pTA946. Strain H300 was transformed with pTA946 as described previously[30], to generate the ΔoriC1::hdrB mutant H1023. Genomic DNA from strain H1023 was prepared as described previously[30], 25 μg was partially digested with 0.5 units/μg of AciI for 30 minutes and fragments of 4-8 kb were ligated in the ClaI site of plasmid pTA131. One μg of this genomic library was used to transform the recombination-deficient strain H112, plasmid DNAs from six transformants were passaged through E. coli and sequenced. All six clones contained the autonomously replicating oriC3 region (pTA1100; Extended Data Fig. 3).

Identifying the integration of pHV4 into the main chromosome

Genomic DNA from wild isolate DS2 and laboratory strain H26 was digested with ClaI, KpnI or NarI (Extended Data Fig. 4). A Southern blot was probed with PCR products upstream (US; primers RFB5F and RFB3R) and downstream (DS; primers RFBF and RFBR) of the H26 profile discontinuity (Fig. 1b). The upstream 3646 bp NarI fragment and downstream 7478 bp KpnI fragment were isolated from H26 genomic libraries and cloned in pBluescript II SK+. The upstream clone pTA1238 and downstream clone pTA1236 contained chromosomal and pHV4 sequences (shown in Extended Data Fig. 4b), indicating that the entire 690 kb pHV4 had integrated into the main chromosome, by recombination between ISH18 insertion sequence elements HVO_0278 (chromosome) and HVO_A0279 (pHV4), as shown in Fig. 1d.

Deletion of radA

Deletion of radA was carried out as described previously[21]. Briefly, the ΔradA::trpA construct pTA324 was used for chromosomal deletion of radA as described previously[30], but in the presence of pTA411 for in trans complementation of radA to facilitate efficient homologous recombination. Deletion of radA results in slow growth, the fraction of slow-growing colonies (ΔradA candidates) that proved to be ΔradA was 94%, 100%, 43% and 12% for the wild-type, ΔoriC1, ΔoriC2 and ΔoriC3 strains, respectively. Only a single ΔradA ΔoriC1 ΔoriC2 ΔoriC3 colony (1 of 70 screened) was recovered, this strain had undergone a chromosomal rearrangement involving the part of integrated pHV4 containing ori-pHV4 (Extended Data Fig. 5). We were unable to delete radA from the strain lacking all chromosomal origins (0 of 455 slow-growing colonies screened).

Generating tryptophan-inducible radA strains

Plasmid pTA1343 carries a recombinant radA allele under control of the tryptophan-inducible p.tnaA promoter[22]. The radA gene was cloned downstream of the p.tnaA promoter in pTA927[33], from which a cassette comprising the t.L11e terminator, p.tnaA promoter, radA and t.Syn terminator was excised and linked to the hdrB marker from pTA187[30], whereupon it was inserted between the upstream and downstream flanking regions of radA in pTA131[30] to generate pTA1343 (Extended Data Fig. 6a). Further details are available upon request. pTA1343 was used to replace the native radA gene in H98 (wild-type, generating H1637) and H1608 (ΔoriC1,2,3,pHV4, generating H1642) as described previously[30], except that transformants were plated on Hv-Ca+5-FOA containing 0.25 mM tryptophan to ensure expression of the p.tnaA-radA gene.

Extended Data Figure 1 | Correcting for GC-bias in deep sequencing data

Sequence composition has previously been reported to influence the depth of sequence coverage[34], therefore we investigated whether GC-content contributes to the noise in our data. Sequence reads from the wild isolate (DS2) stationary phase sample were analyzed with respect to GC-content. (a) For each 1 kb window of unique sequence the number of mapped reads was plotted against the GC-content of the window. We find a significant reduction in mapped sequence reads at elevated GC-content. A polynomial equation (inset and solid line) was fitted to the data. (b) For each 1 kb window of unique sequence, the read counts were plotted against chromosome position. (c) Using the method of Alkan et al.[34], we corrected for GC-bias using the polynomial equation shown in (a) and then plotted the corrected sequence reads against GC-content. (d) GC-bias corrected sequence reads are shown plotted against chromosomal position. With no substantial on-going replication in the stationary phase sample, we can justify using this dataset to normalize the exponential phase data. Both normalization methods result in low noise when compared to studies that do not employ a normalization step[35].

Extended Data Figure 2 | Replication profiles and copy numbers of mega-plasmids

Relative copy number plotted against chromosomal co-ordinate (kb) for pHV1 and pHV3 of (a) wild isolate DS2, (b) laboratory strain H26, (c) ΔoriC1 H1269, (d) ΔoriC2 H1267, (e) ΔoriC3 H1371, (f) ΔoriC1,2,3 H1374 and (g) ΔoriC1,2,3,pHV4 H1546. Each mega-plasmid is displayed linearized at position 0, the location of previously described origins[4]. The 6 kb pHV2 plasmid is not shown in the wild isolate due to the scarcity of data points; pHV2 is not present in laboratory strains (b-g). The pHV4 data for DS2 is shown in Fig. 1. Separate pHV4 data for laboratory strains (b-g) are excluded, since pHV4 is incorporated into the main chromosome in these strains (Figs. 1 and 3). (h) Relative copy number for each mega-plasmid was calculated using the GC-content normalized sequence counts from the stationary phase data for laboratory strain H26.

Extended Data Figure 3 | Characterization of oriC3

(a) Sequence features of oriC3. Double-headed arrow indicates the autonomously replicating fragment recovered from a genomic library of H1023 (pTA1100; see Methods for details), solid arrows represent open reading frames and triangles represent repeats. The intergenic region upstream of orc2 is typical of archaeal origins, it is enlarged to show the sequence features of oriC3 (DUE: duplex unwinding element), HVO_0635 encodes a conserved hypothetical protein. (b) Sequence of intergenic repeats upstream of orc2 (numbered in panel (a), triangles show repeat orientation). Dark grey shading indicates match to consensus origin recognition box (ORB), bases conserved between repeats are indicated by light grey shading. (c) Plasmid-based assays for the three chromosomal origins. Recombination-deficient strain H112 was transformed with 1 μg of pTA441 (oriC1), pTA612 (oriC2) or pTA1100 (oriC3). Transformants were plated with 100-fold dilution on Hv-Ca and incubated at 45°C for 14 days. Numbers indicate transformation efficiency in colony forming units (CFU) /μg DNA. (d) GC-disparity of main chromosome in wild isolate DS2 (adapted from Norais et al.[4]), positions of orc genes and replication origins are shown. The lack of a nucleotide disparity inflection point at oriC3 suggests that this origin has been acquired recently or is used infrequently, consistent with the replication profile (Fig. 1a) and plasmid-based assay (Extended Data Fig. 3c).

Extended Data Figure 4 | Identifying integration of pHV4 into the main chromosome

(a) Map of region around ISH18 insertion sequence element HVO_0278 on the main chromosome of wild isolate DS2, showing restriction sites and probes used to determine the integration of pHV4. (b) Map illustrating integration of pHV4 into the main chromosome of laboratory strain H26, by recombination between ISH18 HVO_0278 (chromosome) and ISH18 HVO_A0279 (pHV4). Regions upstream and downstream of the integration are depicted with the same restriction sites and probes shown in Extended Data Fig. 4a, in addition to the genomic fragments cloned in pTA1238 and pTA1236. (c) Restriction fragment length polymorphisms in the main chromosome of laboratory strain H26. Genomic DNA from wild isolate DS2 and laboratory strain H26 was digested with KpnI, ClaI or NarI, and probed with sequences upstream (US) and downstream (DS) of ISH18 insertion sequence element HVO_0278. The upstream 3646 bp NarI fragment of H26 was cloned in pTA1238, and the downstream 7478 bp KpnI fragment of H26 was cloned in pTA1236. See Methods for details.

Extended Data Figure 5 | Identifying chromosomal rearrangement in ΔoriC1,2,3 ΔradA strain H1553

(a) Map of SfaAI restriction sites on the main chromosome of wild isolate DS2. The region around ISH18 HVO_0278 is shown with additional restriction sites and the probe. (b) Map of SfaAI restriction sites on the main chromosome of laboratory strain H26. The region downstream of integrated pHV4 is shown with the same restriction sites as in Extended Data Fig. 5a, and two additional probes (ori-pHV4 and bgaH) that hybridize to pHV4. (c) Map of SfaAI restriction sites on the main chromosome of ΔoriC1,2,3 ΔradA strain H1553. The region downstream of integrated pHV4 is shown as in Extended Data Fig. 5b. H1553 has undergone a chromosomal rearrangement involving part of pHV4 between ISH18 HVO_A0014 and ISH18 HVO_0278. These ISH18 elements are identical in sequence but in an inverted orientation (bold arrows), recombination between them results in inversion of the intervening sequence. (d) Restriction fragment length polymorphisms in H26 and H1553. Genomic DNA from wild isolate DS2, laboratory strain H26, ΔoriC1,2,3 strain H1501 and ΔoriC1,2,3 ΔradA strain H1553 was digested with SfaAI and displayed on a pulsed field gel. Southern blots were probed with the ori-pHV4 origin, bgaH gene (located on pHV4[21]), and sequences downstream (DS) of ISH18 element HVO_0278. (e) Confirmation of restriction fragment length polymorphisms by ClaI, KpnI and NarI digests, probed with sequences downstream (DS) of ISH18 element HVO_0278; see also Extended Data Fig. 4.

Extended Data Figure 6 | Generating tryptophan-inducible radA strains

(a) Map of p.tnaA-radA gene replacement plasmid pTA1343. (b) Map of region around radA, showing NspI restriction sites and the probe used to determine replacement of the native radA gene with the tryptophan-inducible radA allele. (c) Confirmation of radA replacement by p.tnaA-radA. Genomic DNA from laboratory strain H26, ΔoriC1,2,3,pHV4 strain H1546, p.tnaA-radA strain H1637 and ΔoriC1,2,3,pHV4 p.tnaA-radA strain H1642 was digested with NspI and probed with the radA region. See Methods for further details.