Consequences of producing DNA gyrase from a synthetic gyrBA operon in Salmonella enterica serovar Typhimurium

Abstract DNA gyrase is an essential type II topoisomerase that is composed of two subunits, GyrA and GyrB, and has an A2B2 structure. Although the A and B subunits are required in equal proportions to form DNA gyrase, the gyrA and gyrB genes that encode them in Salmonella (and in many other bacteria) are at separate locations on the chromosome, are under separate transcriptional control, and are present in different copy numbers in rapidly growing bacteria. In wild‐type Salmonella, gyrA is near the chromosome's replication terminus, while gyrB is near the origin. We generated a synthetic gyrBA operon at the oriC‐proximal location of gyrB to test the significance of the gyrase gene position for Salmonella physiology. Although the strain producing gyrase from an operon had a modest alteration to its DNA supercoiling set points, most housekeeping functions were unaffected. However, its SPI‐2 virulence genes were expressed at a reduced level and its survival was reduced in macrophage. Our data reveal that the horizontally acquired SPI‐2 genes have a greater sensitivity to disturbance of DNA topology than the core genome and we discuss its significance in the context of Salmonella genome evolution and the gyrA and gyrB gene arrangements found in other bacteria.


| INTRODUC TI ON
DNA gyrase is an essential type II topoisomerase that introduces negative supercoils into DNA through an ATP-dependent mechanism (Gellert et al., 1976a;Higgins et al., 1978;Nöllmann et al., 2007); it can also relax negatively supercoiled DNA via an ATPindependent mechanism (Gellert et al., 1977;Higgins et al., 1978;Williams & Maxwell, 1999). The enzyme is composed of two copies of two subunits, GyrA and GyrB, giving it an A 2 B 2 structure (Bates & Maxwell, 2005). Gyrase binds to DNA, makes a double-stranded cut, with 4-base overhangs, in the "Gate" (or G) segment of the DNA, and passes a nearby "Transported" (or T) segment of intact DNA through the gap, changing the linking number of the DNA. The GyrA subunits form covalent bonds to the single-stranded DNA overhangs via tyrosine amino acids in their active sites, while the GyrB subunits bind and hydrolyze ATP (Corbett & Berger, 2004).
Topoisomerase activity is required to eliminate the over-wound (positively supercoiled) and under-wound (negatively supercoiled) zones of the DNA template that are generated by transcription and DNA replication (Liu & Wang, 1987;Stracy et al., 2019). DNA gyrase relaxes the positively supercoiled DNA by introducing negative supercoils in an ATP-dependent manner (Ashley et al., 2017).
Transcription and the associated disturbance to local DNA topology contribute to the architecture of the bacterial nucleoid by influencing the probability of DNA-DNA contacts between parts of the genome that border long transcription units that are heavily transcribed (Le & | 1411POZDEEV Et al. Laub, 2016. The changes in local DNA supercoiling that are caused by transcription and DNA replication also affect the activities of some transcription promoters (Ahmed et al., 2016(Ahmed et al., , 2017Chong et al., 2014;Dorman, 2019;Higgins, 2014;Rahmouni & Wells, 1992;Rani & Nagaraja, 2019;Tobe et al., 1995;Wu et al., 1988).
The in vivo superhelical density of DNA in Escherichia coli is −0.025 (Bliska & Cozzarelli, 1987) and it has been estimated that the DNA of E. coli has 15% more supercoils than that of Salmonella enterica serovar Typhimurium (Champion & Higgins, 2007). Because a large subset of promoters is sensitive to alterations in DNA supercoiling and, to ensure appropriate gene expression, topoisomerases are thought to play an important role in maintaining supercoiling set points within a range that is tolerable by the DNA transactions of the cell (Cheung et al., 2003;Dorman & Dorman, 2016;Peter et al., 2004;Sutormin et al., 2019). The promoters of gyrA and gyrB, the genes that encode the A and B subunits, respectively, of DNA gyrase, are stimulated by DNA relaxation (Menzel & Gellert, 1983Straney et al., 1994;Unniraman & Nagaraja, 1999). This is part of a mechanism that maintains DNA supercoiling homeostasis, keeping average DNA supercoiling values within the tolerable range (DiNardo et al., 1982;Dorman et al., 1989;Pruss et al., 1982;Raji et al., 1985;Richardson et al., 1988;Steck et al., 1984). As a corollary to this, the transcription of topA, the gene encoding DNA topoisomerase I (Topo I), is stimulated by negative supercoiling (Ahmed et al., 2016;Tse-Dinh & Beran, 1988). Topo I relaxes negatively supercoiled DNA using an ATP-independent type I mechanism (Dekker et al., 2002).
Several studies have shown that gene position on the chromosome is physiologically significant in bacteria (Bogue et al., 2020;Bryant et al., 2014;Gerganova et al., 2015;Scholz et al., 2019). For example, moving the gene that encodes the nucleoid-associated protein FIS (Factor for Inversion Stimulation) from its native position, proximal to the origin of chromosome replication in E. coli, to locations close to the replication terminus, impaired the competitive growth fitness and altered the overarching network regulating DNA topology, resistance to environmental stress, hazardous substances and antibiotics (Gerganova et al., 2015). Among the genes that FIS regulates in E. coli (Schneider et al., 1999) and S. Typhimurium (Keane & Dorman, 2003) are gyrA and gyrB.
In S. Typhimurium, the gyrA and gyrB genes are widely separated on the genetic map of the circular chromosome: the gyrB gene is located close to the origin of chromosome replication, oriC, while gyrA is located near to the terminus region, Ter (McClelland et al., 2001). This arrangement closely resembles that seen in the model organism, E.coli (Berlyn, 1998;Blattner et al., 1997). It has been proposed that the order of genes along each replichore in the bidirectionally replicated circular chromosome of E. coli correlates with the peak levels of expression of genes as a culture passes through each of the major stages of its growth cycle in batch culture (Sobetzko et al., 2012). DNA supercoiling plays an important role in the initiation of chromosome replication, so locating gyrB close to oriC is consistent with the gene location hypothesis.
Bacteria emerging from the lag phase and entering a period of rapid growth in the exponential phase, experience a build-up of negative DNA supercoiling that stimulates the transcription of genes whose products support rapid growth (Colgan et al., 2018;Conter et al., 1997). Rapidly growing bacteria undergo multiple rounds of the initiation of chromosome replication, so genes close to oriC are present in more copies per cell than those close to the terminus (Cooper & Helmstetter, 1968). GyrA and GyrB are required in equal amounts to form active DNA gyrase molecules, so the physical separation of gyrA from gyrB on the chromosome, and their organization as independent transcription units, seem counterintuitive. In particular, why is gyrA so far away from gyrB on the circular chromosome of S. Typhimurium? The GyrA dimer is a stable structure that lends stability to the tetrameric DNA gyrase (Klostermeier, 2018).
Perhaps placing gyrA close to the terminus of chromosome replication is well tolerated because a pool of GyrA is present in the cell throughout the growth cycle, available to interact with GyrB produced from the gyrB gene close to oriC. If this is so, why is this pattern not seen universally in bacteria? This may reflect species-specific differences in gyrase subunit stabilities and in the ways that the GyrA and GyrB subunits interact in different bacteria (Weidlich & Klostermeier, 2020).
The genetically separated pattern of gyrA and gyrB gene location seen in S. Typhimurium and E. coli is not found universally among bacteria: many possess a gyrBA operon, although none appears to have a gyrAB operon. Perhaps this is unsurprising given that the functional domains in eukaryotic type II topoisomerases that are equivalent to the A and B subunits of DNA gyrase are always arranged in the order BA (Berger, 1998;Forterre et al., 2007), suggesting that the ancestor of eukaryotic type II topoisomerases was an operon with a gyrBA structure. Examples of other bacteria with a gyrBA setup include, inter alia, Listeria monocytogenes (Glaser et al., 2001), Mycobacterium tuberculosis (Unniraman et al., 2002;Unniraman & Nagaraja, 1999), and Staphylococcus aureus (Baba et al., 2008). The operon arrangement appears to offer a number of advantages over the individual transcription unit model. Co-expression allows gyrA and gyrB to share the same promoter and the same transcription regulatory signals. Production of GyrA and GyrB from a common, bicistronic mRNA is likely to facilitate the establishment of equal amounts of each protein. The co-production of GyrA and GyrB might also be expected to enhance the efficiency of gyrase enzyme assembly. It should be noted that the gyrA and gyrB genes are only seen to be widely separated from one another on the unfolded, circular genetic map of S. Typhimurium: the genes may be brought into closer proximity in the folded chromosome within the nucleoid. Furthermore, in the tiny universe of the bacterium's interior, the problem of gyrase assembly from GyrA and GyrB subunits produced from spatially separated mRNA molecules may be an insignificant one (Moffitt et al., 2016). We investigated this issue by building a derivative of S. Typhimurium with a gyrBA operon and comparing its physiology with that of the wild type.

| Constructing a derivative of S. Typhimurium with a synthetic gyrBA operon
A kanamycin resistance cassette, kan, was inserted adjacent to the gyrA gene in S. Typhimurium strain SL1344 to serve as a selectable marker (Experimental procedures). This gyrA-kan combination was amplified by PCR, leaving behind the transcription control signals of gyrA, and inserted immediately downstream of the gyrB gene, creating a gyrBA operon with an adjacent kan gene that was bordered by directly repeated copies of the FRT sequence; the kan gene was then deleted by FLP-mediated site-specific recombination at the frt sites. A kan gene cassette, flanked by directly repeated frt sites, was used to replace the gyrA gene at the native gyrA location in the gyrBA-operon-containing strain; this kan cassette was then excised by FLP-mediated recombination. This process produced a derivative of SL1344 that had a gyrBA operon at the chromosomal position that is normally occupied by only gyrB and had no gyrA gene at the chromosomal site where this gene normally resides (Figure 1).
The whole-genome sequence of this new strain was determined to ensure that no genetic changes, other than the desired ones, were present; none was detected.

| The growth characteristics of the gyrBA operon strain
The growth kinetics of the wild type and the strain with the gyrBA operon were compared in batch liquid culture. Cultures grown in Miller's lysogeny broth (LB) (Miller, 1972) had identical growth curves when measured by plating and colony counting or by optical density measurements ( Figure S1a,b). Growth was also assessed in a minimal medium in an experiment that included low magnesium stress, an important environmental challenge that S. Typhimurium encounters in the macrophage vacuole during infection (Colgan et al., 2018).
The wild-type and the gyrBA-operon strains were grown in minimal medium N (Nelson & Kennedy, 1971) with either 10 µM (low magnesium) or 10 mM (high magnesium) MgCl 2 . Once again, the two strains had identical growth kinetics ( Figure S1c).

| Morphology of the strain with the gyrBA operon
The identical growth characteristics of the strain with the gyrBA operon and the wild type, both in LB and in minimal medium, showed that producing DNA gyrase from an operon made no difference to the growth cycle and suggested that the cell cycle was unlikely to be altered either. Interference with the timing of major events in the cell cycle (initiation, replication fork passage, and termination) can lead to delays in cell division, resulting in the filamentation of the bacterial cell (Martín et al., 2020;Sharma & Hill, 1995). When we compared the morphologies of mid-exponential-phase cultures of the wild type and the gyrBA operon strain by light microscopy, no differences in the shapes of the cells or the frequency of cell filamentation were detected ( Figure S2). Taken together with the growth kinetic data, these findings showed that the operonic arrangement of gyrB and gyrA is well tolerated by S. Typhimurium.

| Sensitivity to gyrase-inhibiting antibiotics
The minimum inhibitory concentrations of gyrase-inhibiting antibiotics were compared for wild-type SL1344 and SL1344 gyrBA ( Figure 2). Four drugs were tested: coumermycin and novobiocin are coumarins that target the GyrB subunit of DNA gyrase (Lewis et al., 1996), while nalidixic acid and ciprofloxacin are quinolones F I G U R E 1 Construction of a derivative of S.Typhimurium strain SL1344 with a gyrBA operon. Chromosomal maps of the WT SL1344 and SL1344 gyrBA strains. Positions of oriC, dif, and chromosome macrodomains are shown. Promoter (angled arrow), protein-coding region (open reading frame, ORF), and the terminator (stem-loop structure) of the genes of interest are shown and color coded. The gyrA ORF is green and the gyrB promoter and ORF are red. Not to scale [Colour figure can be viewed at wileyonlinelibrary.com] that target GyrA (Drlica & Zhao, 1997). Quinolones also target GyrB and coumarins and quinolones inhibit topoisomerase IV, the second type II topoisomerase found in Salmonella and related bacteria (Bush et al., 2020). The two strains were equally sensitive to the quinolones, but the SL1344 gyrBA strain was more resistant than SL1344 to novobiocin, while SL1344 was more resistant than SL1344 gyrBA to coumermycin (Figure 2). The reasons for the differential sensitivity patterns of the strains to the two classes of antibiotics, and for the opposing patterns of resistance to the two coumarins were not determined. Keeping in mind that the coumarins also target topoisomerase IV, we cannot be sure that the differences we observed do not reflect differences in the response of this second drug target in the SL1344 and SL1344 gyrBA strain. However, the results indicated that producing the subunits of gyrase from a gyrBA operon resulted in coumarin MIC data that were not equivalent to those measured for the strain producing the subunits from physically separate genes.

| Motility and competitive fitness measurements
The gyrBA operon strain was compared with the wild type to assess relative motility on agar plates and competitive fitness in liquid co-culture. The operonic strain showed a small, but statistically significant, reduction in motility compared to the wild type ( Figure 3a).
The reasons for this were not determined and may reflect changes at one or more levels in the production and operation of the complex motility machinery of the bacterium. In contrast, the two strains were equally competitive when growing in LB (Figure 3b). To perform the competition, the two strains were each marked genetically by insertion on a chloramphenicol resistance (cat) cassette that is located in the pseudogene SL1483. This cat insertion has a neutral effect on fitness and serves simply to allow the competing bacte-

| Transcription of the separate and the operonic gyr genes
The output of mRNA from the gyrA and gyrB genes was measured by quantitative PCR in wild-type SL1344 and in SL1344 gyrBA, using the transcript of the hemX gene as a reference. (Expression of the hemX gene does not change under the growth conditions used here [Kröger et al., 2013]). Gyrase gene transcription in both strains was found to vary with the growth cycle stage, with mRNA outputs being highest in the early exponential phase (2-hr time point) and lowest in the stationary phase ( Figure 4). In the wild type, the gyrA gene (located near Ter, the terminus of chromosome replication) was expressed to a significantly higher level than gyrB (located close to oriC) at 2 hr. This was interpreted as a reflection of the need to compensate for the effect of increased gyrB gene dosage relative to that of gyrA in rapidly growing cells (Cooper & Helmstetter, 1968).
As the culture approached the stationary phase, the levels of gyrA and gyrB transcripts equalized, in line with the convergence of oriCproximal and Ter-proximal gene dosages ( Figure 4). The formation of the gyrBA operon eliminated the difference in gyrB and gyrA mRNA levels because each became part of the same bicistronic transcript and has adopted the expression profile of gyrB ( Figure 4).

| DNA supercoiling in the strain with a gyrBA operon
The distributions of the topoisomers of the pUC18 reporter plasmid isolated from the wild type and the gyrBA operon strain were com- In LB, the reporter plasmid was more relaxed in the gyrBA strain than in the wild type (Figures 5a, S3b). Low-magnesium growth was used to mimic one of the stresses experienced by Salmonella in the macrophage vacuole. In the high MgCl 2 control, the wild type and F I G U R E 3 Motility and competitive fitness of strain SL1344 gyrBA. (a). Diameters of swimming motility were measured after 5 hr incubation at 37°C on soft 0.3% LB agar. The gyrBA strain is slightly, but statistically significantly, less motile than the WT. Values below 1 indicate that the strain is less motile than the WT. (b) Fitness of the gyrBA strain was compared to the WT SL1344 in LB broth grown for 24 hr with aeration at 37°C. Fitness index (f.i.) = 1 means that the competed strains were equally fit, f.i. ˂ 1 indicates that the competitor strain is less fit than the WT, f.i. ˃ 1 indicates that the competitor is more fit than the WT. The gyrBA and the WT were equally fit. Significance was determined by one-sample T-test, where p < .05 [Colour figure can be viewed at wileyonlinelibrary.com] F I G U R E 4 Expression of the gyrA and gyrB genes in wild-type SL1344 (WT) and SL1344 gyrBA during growth in liquid culture. Cells were grown in LB broth at 37°C with aeration and samples were taken at 2 hr, 3.5 hr, 5 hr, and 7 hr representing the lag, exponential, exponential-stationary transition, and early stationary phases of growth, respectively. Transcription of gyrA and gyrB was measured and is reported relative to that of the hemX reference gene. Three biological replicates were used. Statistical significance was found by unpaired Student's T-test, where p < .05 [Colour figure can be viewed at wileyonlinelibrary.com] the gyrBA operon strain differed in their reporter plasmid distributions: DNA from the gyrBA strain was more negatively supercoiled than that from the wild type and showed a linking number difference (ΔLk) of −3 (Figures 5b, S3b). At the low MgCl 2 concentration, the topoisomer distributions were more relaxed in both strains than in the high MgCl 2 controls (ΔLk = +3). The reporter plasmid from the gyrBA operon strain was also more negatively supercoiled than that from the wild type, with the peak in its topoisomer distribution being approximately two linking numbers below that of the wild type We, therefore, monitored SPI-2 gene transcription in our two strains.

| SPI-1 and SPI-2 gene expression in the gyrba operon strain
The SPI-1 and SPI-2 pathogenicity islands encode distinct type 3 secretion systems and effector proteins that are used to invade epithelial cells (SPI-1) or to survive in the intracellular vacuole (SPI-2) (Figueira & Holden, 2012;van der Heijden & Finlay, 2012;Hensel, 2000). Transcription of SPI-1 genes was monitored using a gfp + reporter fusion under the control of the prgH promoter, P prgH , while a gfp + fusion to the ssaG promoter (P ssaG ) was used to monitor SPI-2 gene transcription. Wild type and gyrBA operon strains harboring these fusions were grown in LB medium (Figure 6a However, SPI-2 transcription occurred at reduced levels in the gyrBA strain in the later stages of the growth cycle under low magnesium conditions (7.8% lower between 800-and 1460-min time points in Figure 6f). These findings showed that when the subunits of gyrase are produced from an operon, rather than from separate gyrA and gyrB genes in their native chromosome locations, the normal expression profile of the SPI-2 virulence gene cluster is disrupted, but that this is conditional on growth in a low magnesium medium.

| Impact of the gyrase operon on cell infection by Salmonella
The abilities of the wild type and the gyrBA operon strains to invade and to replicate in cultured mammalian cells were compared.
Bacteria, grown to the mid-exponential phase to promote SPI-1 gene expression, were used to infect RAW264.7 macrophage. When intracellular bacteria were then enumerated post-invasion, fewer of the gyrBA operon strain cells were detected than wild-type cells at and after the 16-hr time point (Figure 7). This reduction in bacterial survival correlated with the diminished SPI-2 expression detected in the gyrBA operon strain in low Mg 2+ , a macrophage relevant condition.

| D ISCUSS I ON
The genes in Salmonella that encode DNA gyrase, gyrA, and gyrB, are located at the opposite ends of the left replichore of the F I G U R E 5 Reporter plasmid DNA supercoiling in SL1344 gyrBA. The pUC18 reporter plasmid was extracted from the WT and the SL1344 gyrBA strains at the stationary phase of growth and electrophoresed on a 0.8% agarose gel containing 2.5 μg/ ml of chloroquine. The arrow shows the direction of migration, with the more supercoiled plasmid topoisomers at the right of the gel. (a). Global DNA supercoiling pattern of the WT and the gyrBA strain when grown in LB. (b). Global DNA supercoiling pattern of the WT and the gyrBA strain when grown in minimal medium N with high (10 mM) MgCl 2 or low (10 μM) MgCl 2 . Sample lanes are supplemented with densitometry profiles that were generated with ImageJ. The analysis is representative of four biological replicates chromosome ( Figure 1). In contrast, many other bacteria, such as Listeria monocytogenes, Staphylococcus aureus, and Mycobacterium tuberculosis, possess a gyrBA operon (Baba et al., 2008;Glaser et al., 2001;Unniraman et al., 2002;Unniraman & Nagaraja, 1999).
As a first step in assessing the significance of the stand-alone gyrase gene arrangement versus the operon model, we constructed a derivative of S. Typhimurium with a gyrBA operon at the chromosomal location that is normally occupied by gyrB alone, while removing the individual gyrA gene from its native position in the genome. This strain, with the gyrB and gyrA genes transcribed from a common promoter (P gyrB ) and located close to the origin of chromosomal replication, had normal growth characteristics ( Figure S1) and cell morphology ( Figure S2).
Although the production of DNA gyrase from an operon was well tolerated by S. Typhimurium, the operon strain differed from the wild type in a number of phenotypic characteristics. These included a modest decrease in competitive fitness in the operon strain ( Figure 3) that hinted at generalized impacts on physiology. There were subtle differences in sensitivities to coumarin antibiotics, but not quinolones, that distinguished the operon strain from the wild type ( Figure 2). In E. coli, the gyrA and gyrB genes respond differently to treatment with DNA gyrase inhibitors: while coumarins and F I G U R E 6 Expression of genes in the SPI-1 and SPI-2 pathogenicity islands in wild-type SL1344 (WT) and SL1344 gyrBA. Expression of gfp + reporter gene fusions was measured in the wild-type and SL1344 gyrBA strains every 20 min over a 24-hr. period. (a). SPI-1 expression in the gyrBA strain was identical to that in the WT in LB. (b). SPI-2 expression in the gyrBA strain was identical to that of the WT in LB. (c). SPI-1 expression in minimal medium N with high MgCl 2 concentration (10 mM) was repressed in both the WT and the gyrBA strain. (d). SPI-2 expression in minimal medium N with high MgCl 2 concentration was repressed in both the WT and the gyrBA strains. (e). SPI-1 expression in minimal medium N with a low MgCl 2 concentration (10 µM) was repressed in both the WT and the gyrBA strains. (F) SPI-2 expression in minimal medium N with low MgCl 2 concentration was lower in the gyrBA strain than in the WT at the stationary phase of growth. All plots are the results of at least three biological replicates; error bars represent the standard deviation. Statistical significance was found by Student's unpaired T-test, where p < .05 [Colour figure can be viewed at wileyonlinelibrary.com] quinolones both increase the expression of gyrA, the expression of only gyrB is induced by coumarins (Neumann & Quiñones, 1997). In our operon strain, the coumarin-sensitive gyrB promoter drives the transcription of both gyrB and gyrA, and this may have contributed to the difference between the operon strain and the wild type in responding to the coumarin challenge. It is also possible that the differences in coumarin sensitivity may have involved indirect effects of operonic gyrase production on processes involved in drug uptake and/or be due to differential effects on the activity of the bacterium's second coumarin target, topoisomerase IV, in the wild type and gyrase operon strains.
The stand-alone gyrA gene is expressed to a higher level than gyrB in the wild type, albeit from a distant location on the chromosome ( Figure 4). Also, the stable GyrA protein seems to play a foundational role in the assembly of DNA gyrase (Klostermeier, 2018;Weidlich & Klostermeier, 2020) The Salmonella-containing vacuole of the macrophage is a stressful, low magnesium environment where the SPI-2-encoded type 3 secretion system and its associated effector proteins, play a key protective role (Figueira & Holden, 2012;van der Heijden & Finlay, 2012;Hensel, 2000). DNA relaxation contributes to the full expression of SPI-2 genes (Cameron & Dorman, 2012;Quinn et al., 2014) and DNA in S. Typhimurium becomes relaxed when the bacterium is in the macrophage (O Cróinín et al., 2006), The gyrBA operon strain maintained its DNA in a less relaxed state in a low-magnesium environment ( Figure 5) and this may explain the poorer transcription of SPI-2 that was seen in low magnesium growth ( Figure 6).
Our data reveal a distinction between the sensitivity of genes encoding housekeeping functions and genes in the horizontally acquired accessory genome to the production of DNA gyrase from an operon. Genes that have been acquired by Salmonella via horizontal gene transfer (HGT) are more A + T-rich than core genome members and are subject to multifactorial control that includes a prominent role for nucleoid-associated proteins such as H-NS, FIS, IHF, and HU (Banda et al., 2019;Cameron & Dorman, 2012;Dillon & Dorman, 2010;Fass & Groisman, 2009;Mangan et al., 2006Mangan et al., , 2011Quinn et al., 2014). These genes also display sensitivity to changes in DNA topology (Cameron & Dorman, 2012;O Cróinín et al., 2006;Quinn et al., 2014). SPI-2 expression is held in check in Salmonella until it is required during adaptation to the macrophage vacuole and a shift in global DNA supercoiling levels is a component of the activation signal (Cameron & Dorman, 2012;O Cróinín et al., 2006;Quinn et al., 2014). The shift in global DNA supercoiling values that we see in Salmonella, growing in a minimal medium that mimics the intravacuolar environment, dysregulates SPI-2 transcription and compromises Salmonella infectivity. This finding is consistent with proposals that Salmonella maintains a level of DNA topology that is optimized to control the activities, and the expression, of mobile genetic elements that include pathogenicity islands, bacteriophage, and transposons (Cameron et al., 2011;Champion & Higgins, 2007;Higgins, 2016).
We have shown experimentally that there is no absolute barrier to the organization of the gyrB and gyrA genes as a gyrBA operon in Salmonella. Furthermore, there are many examples of naturally occurring gyrBA operons among bacterial species. Why is this arrangement not found universally? Sharing a common promoter, common transcriptional regulatory features, and a common chromosomal location would appear to offer the advantages of coordinated gene expression (Price et al., 2005) and physical co-production of protein products that will need to combine with a fixed stoichiometry to form an active product (Dandekar et al., 1998;Pal & Hurst, 2004;Swain, 2004). Indeed, the coupling of transcription and translation in prokaryotes may aid the production of operon-encoded proteins that are required in stoichiometric amounts (Li et al., 2014;Rocha, 2008).
Often, but not invariably, operons are composed of genes that contribute to a common pathway (de Daruvar et al., 2002;Lawrence & Roth, 1996;Price et al., 2006;Rogozin et al., 2002) and that is true of the gyrBA operon. Colocation of genes within an operon facilitates their collective translocation via horizontal gene transfer, allowing them to replace lost or mutated copies in the recipient cell (Lawrence  , 1996). According to this "selfish operon" hypothesis, this creates selective pressure for the maintenance of an operon structure.
However, since the loss of either gyrA or gyrB is lethal, the gyrase operon may be one of the exceptions to the selfish operon rule, because gyrase-deficient recipients cannot exist.
We conducted a survey of gyrase gene locations in bacteria to assess the frequency of the stand-alone arrangement seen in Salmonella and the gyrBA operon arrangement seen in other species (Experimental procedures). We were unable to find any example of a bacterium with a gyrAB operon. It should be noted that the functional domains corresponding to GyrA and GyrB in eukaryotic type II topoisomerases are found in the order BA (Berger, 1998;Forterre et al., 2007), suggesting that their ancestor may have been encoded by an operon with a gyrBA structure. The results of the survey are shown in Table 1, where bacteria are grouped according to their gyrase gene arrangement, using oriC as a reference point. Figure 8 shows a phylogenetic tree summarizing the occurrence of different gyrase gene arrangements among bacteria from the four groups listed in Table 1.
Inversions of DNA between the left and the right replichores were seen frequently and these followed no obvious patterns. This is in agreement with a previous finding that, while the distance to the origin is highly conserved, inversions of genes around the Ter region of a chromosome are frequent and well tolerated in E. coli and Salmonella (Alokam et al., 2002). Various relative arrangements of gyrA and gyrB were observed and subdivided into four groups: There is perhaps more variation within Group 4, but this was not detected using the method employed here. Mycoplasma is an anomaly of Group 3, since not all its species clearly belong to this group.
Some Mycoplasma possess the expected conserved genes 5' to gyrB, but not in its immediate vicinity. However, the orientation of genes 5' to gyrB remains favorable for the initiation of its transcription, therefore, Mycoplasma is placed in Group 3. It was clear from the analysis that members of the same taxonomic rank do not necessarily have to belong to the same Group, especially in diverse phyla.
For example, both Group 2 and Group 3 arrangements are present within the Firmicutes. Moreover, both arrangements are present within the order Lactobacillales alone. Some less diverse phyla such as Fusobacteria and Chlamydiae belong to only one Group. No variation was found within families.
It was difficult to determine whether given taxons were enriched in particular groups in Table 1, so a phylogenetic tree was plotted that included all of the bacteria in the table (Figure 8).
The tree was constructed using the phylogenetic tree generator phyloT, based on NCBI taxonomy (Letunic & Bork, 2019)  When the immediate genetic environment of both genes in bacteria listed in Table 1 was studied, one distinct pattern was foundhomologs of dnaA (encoding chromosomal replication initiation protein DnaA), dnaN (encoding the beta subunit of DNA polymerase III), and recF (encoding the DNA repair protein RecF) or at least one of these three genes, are found directly upstream of gyrB gene in all bacteria in which gyrB is located in the immediate vicinity of oriC, such as most bacteria of Groups 1 and 3 (Table 1). Transcription from these co-oriented neighboring genes provides a strong input of DNA relaxation (Sobetzko, 2016) that stimulates the transcription of the supercoiling-sensitive gyrB promoter, P gyrB (Menzel & Gellert, 1987). This is true of most bacteria where gyrB is in the immediate vicin-

F I G U R E 8
Phylogenetic tree of bacteria that belong to different groups based on their gyrA and gyrB arrangement. The phylogenetic tree was built in phyloT, a phylogenetic tree generator based on NCBI taxonomy (Letunic & Bork, 2019). Each of the four groups (see Table 1) of gyrA and gyrB arrangements is indicated by color. Group 1, blue: gyrA and gyrB are at separate locations, with a conserved genetic environment 5' to gyrB. Group 2, orange: gyrA and gyrB are at separate locations, with a non-conserved genetic environment 5' to gyrB. Group 3, green: gyrBA operon, conserved genetic environment 5' to gyrB. Group 4, red: gyrBA operon, non-conserved genetic environment 5' to gyrB. Phyla names are indicated [Colour figure can be viewed at wileyonlinelibrary.com]

| Bacterial strains and culture conditions
The bacterial strains used in this study were the derivatives of S. Typhimurium strain SL1344 and their details are listed in Table 2. Bacterial cultures were grown routinely either in Miller's lysogeny broth (LB) (Miller, 1972) or in minimal medium N (Nelson & Kennedy, 1971). Bacteriophage P22 HT 105/1 int-201 was used for generalized transduction during strain construction (Schmieger, 1972 One milliliter of overnight culture was washed three times with a minimal medium N of the required MgCl 2 concentration to remove nutrients, sub-cultured into minimal medium of the corresponding MgCl 2 concentration in a total volume of 25 ml and grown for 24 hr to pre-condition the bacteria. The pre-conditioned culture was sub-cultured into 25 ml of fresh minimal medium N adjusted to an OD 600 of 0.03 and grown for a further 24 hr to obtain a culture in the stationary phase of growth.
To measure growth characteristics of a bacterial culture, an overnight culture was adjusted to an OD 600 of 0.003 in 25 ml of fresh LB broth and grown at the standard conditions for 24 hr in the appropriate liquid medium. The optical density of the culture at OD 600 was measured at 1-hr intervals for the first 3 hr and then every 30 min until 8 hr; the last reading was taken at 24 hr. Measurements were taken using a Thermo Scientific BioMate 3S spectrophotometer with liquid cultures in plastic cuvettes. To measure the growth characteristics of bacterial culture in the minimal medium with altered Mg 2+ concentration, an overnight bacterial culture was washed in the minimal medium with an appropriate concentration of MgCl 2 and pre-conditioned for 24 hr. The pre-conditioned culture was adjusted to an OD 600 of 0.03 in 25 ml of fresh medium in two flasks and the OD 600 was measured every hour beginning from 2 hr post-time zero to 8 hr. Separate cultures were set up similarly to measure OD 600 every hour from 8 hr to 15 hr. In this way, the number of times each flask was opened and sampled was minimized to yield reliable and reproducible measurements.
The growth characteristics of bacterial cultures in LB broth were also measured by viable counts. The culture was grown in the same way as for spectrophotometry, and an aliquot was taken at 2 hr, 4 hr,

| Bacterial motility assays
Assays were carried out precisely as described to achieve agreement between biological replicates. 0.3% LB agar was melted in a 100 ml bottle in a Tyndall steamer for 50 min, allowed to cool in a 55°C water bath for 20 min, six plates were poured and left to dry near a Bunsen flame for 25 min. One microliter of the bacterial overnight culture was pipetted under the agar surface with two inocula per plate. Plates were placed in a 37°C incubator without stacking to ensure equal oxygen access. After 5 hr, the diameters of the resulting swarm zones were measured and expressed as the ratio of the WT zone to that of the mutant. Competitor is a strain other than the WT; f.i. ˂ 1 means that the competitor is less fit than the WT; f.i. ˃ 1 indicates the opposite.

| Construction of a gyrBA operon strain
A derivative of S. Typhimurium with an artificial gyrBA operon was constructed by Lambda-Red homologous recombination (Datsenko & Wanner, 2000). Briefly, a kan cassette was amplified from plasmid pKD4 with primers Kan ins gyrA F and Kan ins gyrA R, (Table S1) using Phusion high-fidelity DNA polymerase. The amplicon, with overhangs homologous to a region immediately downstream of gyrA, was transformed into the WT strain harboring plasmid pKD46 (Table 3), then grown in the presence of arabinose to activate the Lambda Red system in order to tag gyrA with the kan cassette.
The gyrA::kan construct, including 20 nucleotides upstream from the gyrA translational start codon, was amplified using primers gyrB.int.
gyrA::kan_Pf and gyrB.int.gyrA::kan_Prev (Table S1). The amplicon had overhangs that were homologous to sequences immediately downstream of gyrB. This allowed the translation of the GyrA protein from the bicistronic gyrBA mRNA because several sequences closely matching to a consensus ribosome binding site (5′-AGGAGG-3′) were located in this 20 bp region. The gyrA::kan amplicon was inserted by Lambda Red-mediated recombination immediately downstream of the gyrB protein-coding region to construct the gyrBA operon.
The original gyrA gene was deleted by an in-frame insertion of a kan cassette (Baba et al., 2006). The kan resistance cassettes were subsequently eliminated via FLP-mediated site-specific recombination (Cherepanov & Wackernagel, 1995). The resulting gyrBA strain had the genes that encode both subunits of DNA gyrase arranged as a bicistronic operon under the control of a common promoter, P gyrB (Table 2).

| Whole-genome sequencing
Whole-genome sequencing was performed on final versions of the constructed strains to ensure that no compensatory mutations were introduced into their genomes. The sequencing was performed by MicrobesNG (Birmingham, UK) using Illumina next-generation sequencing technology. The output reads were assembled using Velvet (Zerbino, 2010) and aligned to the reference SL1344 sequence NC_016810.1 Breseq software (Deatherage & Barrick, 2014

| Minimum inhibitory concentration (MIC) of antibiotics determination
MIC90 of antibiotics (a minimal concentration at which 90% of bacterial growth is inhibited) was found by serially diluting antibiotics and spectrophotometrically testing the ability of different dilutions to inhibit bacterial growth. On a 96-well plate, all wells (excluding column 12) were filled with 60 μl of sterile LB broth. One milliliter of solutions of antibiotics to be tested was prepared at the highest desired concentration in LB. Three hundred microliters of the prepared antibiotics were added to the wells of column 12 and homogenized by pipetting up and down five times with a multichannel pipette. Two hundred and forty microliters were transferred to the next wells in column 11, homogenization was repeated and serial 1:1.25 dilutions were sequentially continued until column 3.
The final 240 μl from column 3 were discarded. All the wells were inoculated with bacterial cultures adjusted to an OD 600 of 0.003 except column 1. In this way, column 1 contained negative controls (no bacteria and no antibiotic), column 2 contained positive controls (no antibiotics), and columns 3-12 contained serially diluted antibiotics inoculated with the identical number of bacteria. The plate was covered, sealed between plastic sheets, and incubated for 18 hr at the standard growth conditions. The plate was read by measuring OD 600 values on a plate reader (Multiskan EX, Thermo Electronics).

| SPI-1 and SPI-2 reporter assays
Salmonella pathogenicity island (SPI) activity was accessed by measuring the expression of gfp + reporter gene fusions to promoters of prgH and ssaG to look at SPI-1 and SPI-2 expression, respectively.
The gfp + reporter fusions were transduced into each strain by P22 generalized transduction and selected with chloramphenicol.

| Global supercoiling determination
Global DNA supercoiling was assayed in bacterial strains transformed with a reporter plasmid pUC18 (Table 3). An overnight culture of pUC18-containing strain was adjusted to an OD 600 of 0.003 and grown to the late stationary growth stage (24 hr) in 25 ml of LB broth or in 25 ml of minimal medium N of the required MgCl 2 concentration pre-conditioned as above. Fourteen OD 600 units (6 OD 600 units for minimal medium) were harvested and pUC18 was isolated with the aid the of QIAprep Spin Miniprep kit (QIAGEN, Hilden, Germany) according to manufacturer's guidelines.
To observe the range of DNA supercoiling states characteristic of a strain at a given growth stage, extracted pUC18 samples were resolved on 0.8% agarose gel supplemented with the DNA intercalating agent chloroquine. Two liters of 1× TBE buffer (89 mM Tris base, 89 mM boric acid, 2 mM EDTA pH 8.0) and 1 ml of 25 mg/ml of chloroquine were made. 0.8% agarose solution was made from 300 μl of TBE and melted in a Tyndall steamer. When the gel cooled down, it was supplemented with 2.5 μg/ml of chloroquine. The 27 cm long gel was poured, left to polymerise for 2 hr, and covered with 1.7 liters of the running buffer containing 1x TBE and chloroquine at 2.5 μg/ml. One microgram or 500 ng of the plasmid samples in 15 μl volumes was mixed with 5x loading dye (80% glycerol, 0.01% bromophenol blue) and loaded on a gel. The gel was electrophoresed for 16 hr at 100 V. The gel was washed in distilled water for 24 hr changing water a few times, stained in 1 μg/ml of ethidium bromide for 1 hr rocking in the dark. The stain was poured off and the gel was washed in distilled water for further 1 hr. The plasmid topoisomers were visualized under UV on the ImageQuant LAS 4,000 imager.
ImageJ software was used to outline plasmid topoisomer distribution profiles.

| Determining the patterns of gyrA and gyrB locations in bacterial chromosomes
The location of oriC in each organism examined was determined using the DoriC 10.0 database (tubic.org/doric) (Luo & Gao, 2019) and the coordinates of their gyrA and gyrB genes were obtained using the Ensembl bacteria browser (bacteria.ensembl.org). Distance in base pairs between the oriC and the gene was calculated and converted into the percentage of the total chromosome size. An attempt was made to cover bacterial taxonomy as broadly as possible, encompassing members of the major bacterial phyla, well studied, and clinically important organisms in the analysis (Table 1). The table is neither complete nor does it claim to include all the existing possibilities of gyrA and gyrB arrangements in bacterial chromosomes, but instead, exemplifies the arrangement possibilities mentioned in this work. Closely related species and those belonging to the less diverse phyla were found to share the chromosomal positions of gyrA and gyrB frequently. Thus, one representative of a taxonomic rank was often deemed sufficient for the purpose of inclusion in the table.
Lower classification ranks were analyzed within more diverse and studied phyla (Table 1).

| Mammalian cell culture conditions
RAW264.7 murine macrophages were maintained in Dulbecco's Modified Eagle's Medium (DMEM), (Sigma, catalog number D6429) supplemented with 10% fetal bovine serum (FBS) in a humidified 37°C, 5% CO 2 tissue-culture incubator grown in 75 cm 3 tissueculture flasks. When approximately 80% confluent growth was achieved, cells were split into a fresh flask. Cells within the 9-16 passage number range were used for infections. All media and PBS used for cell culture were pre-warmed to 37°C. To split cells, old DMEM was removed and the monolayer was rinsed with 10 ml of sterile PBS. Ten milliliters of fresh DMEM were pipetted into the flask and the monolayer was scraped gently with a cell scraper to dislodge the cells. Scraped cells were centrifuged at 450 x g for 5 min in an Eppendorf 5810R centrifuge and the cell pellet was resuspended in 5 ml of DMEM + FBS. One milliliter of the cell suspension was added to 14 ml of fresh DMEM + FBS in a 75 cm 3 flask, gently rocked to mix, and incubated at 37°C, 5% CO 2 . To seed cells for infection, cells were treated as for splitting. After resuspension in 5 ml of DMEM + FBS, viable cells were counted on a haemocytometer using trypan blue exclusion dye. A 24-well tissue culture plate was filled with 500 μl of DMEM + FBS. 1.5 × 10 5 cells were added to each well, gently rocked to mix, and incubated at 37°C, 5% CO 2 for 24 hr.

| Macrophage viability assay in SPI-1 inducing conditions
Overnight bacterial cultures were subcultured 1:33 in 10 ml of fresh LB broth in 125 ml conical flask and grown for 3.5 hr to maximize SPI-1 expression (Steele-Mortimer et al., 1999). Five hundred microliters of the culture were centrifuged at 16,000 ×g for 1 min and resuspended in 500 μl of HBSS−/−. Monolayers were washed twice with 500 μl of HBSS+/+ and infected with bacteria at MOI of five in three technical replicates for each timepoint and strain. The plate was centrifuged at 200 ×g for 10 min to synchronize infections and incubated for 30 min at 37°C, 5% CO 2 . In the meantime, the infection medium was plated for enumeration on LB agar plates-T = 0 hr.
Gentamicin protection assay was used to determine bacterial counts inside macrophages. To kill all extracellular bacteria, the monolayers were washed once with HBSS+/+ and high gentamicin (100 μg/ml) treatment diluted in DMEM + FBS was added to the wells. The plate was incubated at 37°C, 5% CO 2 for 1 hr. At 1 hr post-infection, the monolayers were washed three times with HBSS+/+, macrophages were lysed by adding 1 ml of ice-cold water, pipetting up and down ten times with scraping and intracellular bacteria were plated for enumeration. The monolayers which were intended for other timepoints were washed once with HBSS+/+, low gentamicin (10 μg/ml) treatment in DMEM + FBS was added and the plate was incubated at 37°C, 5% CO 2 . The low gentamicin concentration is to ensure that any extracellular bacteria are killed, but at the same time to avoid gentamicin permeabilizing the plasma membrane of a macrophage (Kaneko et al., 2016). At later timepoints monolayers were washed three times with HBSS+/+, macrophages were lysed by adding 1 ml of ice-cold water, pipetting up and down ten times with scraping and intracellular bacteria were plated for enumeration.

Research in the CJD laboratory is supported by Science Foundation
Ireland Investigator Award 13/IA/1875.

CO N FLI C T S O F I NTE R E S T
The authors declare that they have no conflicts of interest.