Effects of DNA Topology on Transcription from rRNA Promoters in Bacillus subtilis

The expression of rRNA is one of the most energetically demanding cellular processes and, as such, it must be stringently controlled. Here, we report that DNA topology, i.e., the level of DNA supercoiling, plays a role in the regulation of Bacillus subtilis σA-dependent rRNA promoters in a growth phase-dependent manner. The more negative DNA supercoiling in exponential phase stimulates transcription from rRNA promoters, and DNA relaxation in stationary phase contributes to cessation of their activity. Novobiocin treatment of B. subtilis cells relaxes DNA and decreases rRNA promoter activity despite an increase in the GTP level, a known positive regulator of B. subtilis rRNA promoters. Comparative analyses of steps during transcription initiation then reveal differences between rRNA promoters and a control promoter, Pveg, whose activity is less affected by changes in supercoiling. Additional data then show that DNA relaxation decreases transcription also from promoters dependent on alternative sigma factors σB, σD, σE, σF, and σH with the exception of σN where the trend is the opposite. To summarize, this study identifies DNA topology as a factor important (i) for the expression of rRNA in B. subtilis in response to nutrient availability in the environment, and (ii) for transcription activities of B. subtilis RNAP holoenzymes containing alternative sigma factors.


Introduction
Bacterial cells need to adapt to environmental changes. In nutrient-rich environments, cells grow and divide rapidly, and this requires a large number of ribosomes to satisfy the need for new proteins. In nutritionally poor environments, the synthesis of new ribosomes stops. As the production of new ribosomes is energetically costly for the cell, it must be tightly regulated. The number of ribosomes in the cell is regulated mainly on the level of transcription initiation of ribosomal RNA (rRNA) [1].
Transcription initiation can be divided into several steps. First, when the RNA polymerase (RNAP) holoenzyme (the core RNAP subunits [α2ββ ω] in complex with a σ factor) binds to specific DNA sequences, promoters, it forms the closed complex where DNA is still in the double-helical form [2]. The specificity of RNAP for promoter sequences is provided by the σ factor [3][4][5][6]. Subsequently, this complex isomerizes and forms the open complex where the two DNA strands are unwound, and the transcription bubble is formed. At this stage, initiating nucleoside triphosphates NTPs (iNTPs) can enter the active site and transcription can begin. RNAP then leaves the promoter and enters the elongation phase of transcription [7].
In bacteria, the concentrations of iNTPs act as key regulators of transcription and directly affect RNAP at some promoters. These promoters form relatively unstable open complexes where the time window available to iNTPs to penetrate into the active site and initiate transcription is relatively short. The higher the concentration of the respective iNTP, the higher the chance that it penetrates into the active site while the transcription bubble is Table 1. Bacterial strains used in a study.

Name
Original code Construct Description Reference
To determine the relative ATP/GTP concentrations after novobiocin treatment, LK134 was grown to OD 600~0 .3 (time 0) in medium supplemented with [ 32 P] H 3 PO 4 (100 µCi/mL), and at time 5 min treated with novobiocin (5 µg/mL). Samples were taken at time points 0, 5, 10, 20, and 30 min and processed in the same way as above.

Promoter Activity Monitored by Quantitative Primer Extension (qPE)
Promoter constructs were fused to lacZ and activities were assayed by primer extension of the short-lived lacZ mRNA that allows to observe rapid decreases in promoter activity in time. The experiments were conducted as described in [15]. Typically, 1 mL of cells was pipetted directly into 2 mL phenol/chloroform (1:1) and 0.25 mL lysis buffer (50 mM Tris-HCl pH 8.0, 500 mM LiCl, 50 mM EDTA pH 8.0, 5% SDS). After brief vortexing, the recovery marker (RM) was added. The RM RNA was made from B. subtilis strain LK41 as described in [15]. This was followed by immediate sonication. Water was then added to increase the aqueous volume to 6 mL to prevent precipitation of salts, followed by two extractions with phenol/chloroform, two precipitations with ethanol, and suspension of the pellet in 20-50 µL 10 mM Tris-HCl, pH 8.0.
Primer extension was performed with M-MLV reverse transcriptase as recommended by the manufacturer (Promega) with 1-10 µL purified RNA. The 32 P 5'-labeled primer (#2973) hybridized 89 nt downstream from the junction of the promoter fragment used for the creation of the lacZ fusion. Samples were electrophoresed on 7 M urea 5.5% or 9% polyacrylamide gels. The gels were exposed to Fuji Imaging Screens. The screens were scanned with Molecular Imager FX (Bio-Rad, Berkeley, CA, USA) and were visualized and analysed using the Quantity One software (Bio-Rad), and normalized to cell number (OD 600 ) and RM.
2.5. Promoter Activity Monitored by RT-qPCR rrnB P1 and Pveg promoters were fused to the marker lacZ gene (LK134 and LK135), yielding identical transcripts. The strains were grown to exponential phase (OD 600 0.5)-time point 0. Each culture was then divided into two flasks. Cells in one flask were treated with novobiocin (5 µg/mL) and cells in the other flask were left non-treated.  [15]. The final data were normalized to RM and the amount of cells (OD 600 ).

3 H-Incorporation in Total RNA
This experiment was conducted as described previously [30]. Briefly, strain LK134 was grown in LB medium to OD 600~0 .3 (early exponential phase). Newly synthesized RNA in the cells was labeled with 3 H-uridine (1 µCi/mL) (cold [non-radioactive] uridine was added to a final concentration of 100 µM); time point 0. The bacterial culture was divided into three flasks-non-treated, treated with novobiocin (5 µg/mL), and treated with rifampicin (2 µg/mL), respectively (time point 5). At 0, 5, 10, 20 and 30 min, 100 µL and 250 µL of cells were withdrawn to measure cell density and determine 3 H-incorporation, respectively. The 250 µL cell sample was mixed with 1 mL of 10% trichloroacetic acid (TCA) and kept on ice for at least 1 h. Thereafter, each sample was vacuum filtered, using Glass Microfiber Filters (Whatman, Little Chalfont, UK), washed twice with 1 ml of 10% TCA and three times with 1 mL of ethanol. The filters were dried, scintillation liquid was added, and the radioactivity was measured. The signal was normalized to cell density (OD 600 ).

RNAP Levels in Time
Cells (strain LK134) were grown in LB rich medium to OD 600 0.3 (time point 0). Subsequently, every 30 min 10 mL of cells were pelleted and OD 600 was measured. Pellets were washed with Lysis Buffer (20 mM Tris-HCl, pH 8, 150 mM KCl, 1 mM MgCl 2 ) and frozen. Next day, pellets were resuspended in Lysis Buffer (100-500 µL, according to the size of pellet) and disrupted by sonication 2 × 1 min, with 1 min pause on ice between the pulses. After centrifugation (5 min, 4 • C) to remove cell debris, the amounts of proteins were measured with the Bradford protein assay and 5 µg was resolved by SDS-PAGE and analyzed by Western blotting, using mouse monoclonal antibodies against the β subunit of RNAP (clone name 8RB13, dilution 1:1000, Genetex, Irvine, CA, USA) and anti-mouse secondary antibody conjugated with HRP (dilution 1:800,000, Sigma, Munich, Germany). Subsequently, the blot was incubated for 5 min with SuperSignal TM West Femto PLUS Chemiluminiscent substrate (Thermo scientific, Waltham, MA, USA), exposed on film and developed. Genes encoding σ B , σ E , σ F and σ H were amplified from genomic wt DNA by PCR with Expand High Fidelity PCR System (Roche) with respective primers ( Table 1, Material and Methods section) and cloned into pET-22b(+) via NdeI/XhoI restriction sites and verified by sequencing. Primers for cloning of σ E were designed for the active form of protein, as its first 27 AA are in the cell posttranslationally removed [31,32]. The resulting plasmids were transformed into expression strain BL21(DE3), yielding strains LK1207 (σ B ), LK2580 (σ E ), LK1425 (σ F ), and LK1208 (σ H ). His-SUMO-σ N fusion protein in an expression plasmid pBM05 [25] was transformed to BL21(DE3), resulting in strain LK2531.
The SigA subunit of RNAP (LK22) was overproduced a purified as described [27]. σ B , σ E , σ F , σ H expression strains were grown to OD 600~0 .5 when IPTG was added to a final concentration of 0.8 mM. Cells were allowed to grow for 3 h at room temperature, cells were harvested, washed and resuspended in P buffer (300 mM NaCl, 50 mM Na 2 HPO 4 , 3 mM β-mercaptoethanol, 5% glycerol). Cells were then disrupted by sonication and the supernatant was mixed with 1 mL Ni-NTA agarose (QIAGEN, Hilden, Germany) and incubated for 1 h at 4 • C with gentle shaking. Ni-NTA agarose with the bound protein was loaded on a Poly-Prep ® Chromatography Column (Bio-Rad, Berkeley, CA, USA), washed with P buffer and subsequently with the P buffer with the 30 mM imidazole. The protein was eluted with P buffer containing 400 mM imidazole and fractions containing σ factor were pooled together and dialyzed against storage buffer (50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 50% glycerol and 3 mM β-ME). The proteins were stored at −20 • C.
σ D was purified from inclusion bodies as described in [28]. Cells containing the plasmid for overproduction of σ N were grown to OD 600~0 .5 and IPTG was added to final concentration 0.3 mM. Cells were then allowed to grow for 3 h at room temperature; afterwards the cells were harvested, washed, and resuspended in P buffer. All purification steps were done in P2 buffer (the same composition as P buffer, but pH 9.5). Cells were then disrupted by sonication and the supernatant was mixed with 1 mL Ni-NTA agarose (QIAGEN) and incubated for 1 h at 4 • C with gentle shaking. Ni-NTA agarose with the bound His-SUMO-σ N was loaded on a Poly-Prep ® Chromatography Column (Bio-Rad), washed with P2 buffer and subsequently with the P2 buffer with the 30 mM imidazole. The protein was eluted with P2 buffer containing 400 mM imidazole and fractions containing His-SUMO-σ N were pooled together and dialyzed against P2 buffer.
The SUMO tag was subsequently removed by using SUMO protease (Invitrogen). The cleavage reaction mixture was again mixed with the 1 mL Ni-NTA agarose and allowed to bind for 1 h at 4 • C and centrifuged to pellet the resin. Supernatant was removed, the resin was washed once more with P2 buffer with 3 mM β-ME. The supernatants (containing σ N ) were pooled together and dialysed against storage P2 buffer (P2 buffer and 50% glycerol). The protein was stored at −20 • C.
The purity of all proteins was checked by SDS-PAGE.

Promoter DNA Construction
Promoter regions of alternative σ-dependent genes were amplified from genomic wt DNA of B. subtilis with primers listed in Table 2 (Material and Methods section) by PCR. All fragments were then cloned into p770 (pRLG770 [33]) using EcoRI/HindIII restriction sites and transformed into DH5α. All constructs were verified by sequencing. Transcription experiments were performed with the B. subtilis RNAP core reconstituted with a saturating concentration of σ A (ratio 1:5) in storage buffer (50 mM Tris-HCl, pH 8.0, 0.1 M NaCl, 50% glycerol) for 15 min at 30 • C. The 1:5 ratio was used also for σ B , σ D , σ E , σ F , and σ H . For σ N , the ratio was 1:8. Multiple round transcription reactions were carried out in 10 µL reaction volumes with 30 nM RNAP holoenzyme. The transcription buffer contained 40 mM Tris-HCl pH 8.0, 10 mM MgCl 2 , 1 mM dithiothreitol (DTT), 0.1 mg/mL BSA and 150 mM KCl, and all four NTPs and 2 µM radiolabeled [α-32 P] UTP.
In K GTP determination experiments, the amount of DNA (SC or LIN form) was 100 ng, ATP, CTP were 200 µM; UTP was 10 µM and GTP was titrated from 0 to 2000 µM.
To determine the affinity of RNAP to DNA, ATP, CTP were at 200 µM; UTP was 10 µM, GTP was 1000 µM and DNA (SC/LIN) was titrated from 0 to 900 ng per reaction. In reactions with alternative σ, DNA (SC or LIN form) was 100 ng, CTP were at 200 µM; UTP was 10 µM and GTP/ATP was 1000 µM, depending on the identity of the base in the +1 position of the transcript.
All transcription reactions were allowed to proceed for 15 min at 30 • C and then stopped with equal volumes of formamide stop solution (95% formamide, 20 mM EDTA, pH 8.0). Samples were loaded onto 7 M urea-7% polyacrylamide gels and electrophoresed. The dried gels were scanned with Molecular Imager FX (Bio-Rad) and were visualized and analysed using the Quantity One software (Bio-Rad).

The Activity of rrnB P1 Decreases during Entry into Stationary Phase
As the main model rRNA promoter, we selected the rrnB P1 promoter as it is one of the best-characterized rRNA promoters in B. subtilis that is regulated by [iNTP], [11,[34][35][36]. Furthermore, the dynamic range of the activity of rrnB P1 is wide, which facilitated the design and interpretation of the experiments. As the main control promoter, we selected the strong Pveg promoter that forms stable open complexes and is saturated with a relatively low level of its iNTP. This promoter drives transcription of the veg gene that is involved in biofilm formation [37,38]. Promoter sequences are shown in Figure 1A.
To monitor promoter activities, we used core promoter-lacZ fusions. The endogenous copy of Pveg initiates transcription with ATP (+1A). Here, we used a +1G variant of Pveg so that both transcripts (from rrnB P1-lacZ and Pveg-lacZ) were identical, excluding any effects due to, e.g., potentially differential decay of the transcripts. The +1G Pveg promoter variant behaves identically with the +1A variant [11]. Throughout the study, promoter activity was determined by quantitative primer extension (qPE) or reverse transcription followed by quantitative PCR (RT-qPCR).
We used defined rich MOPS medium to grow the cells and measured (i) relative GTP level ([GTP]) and (ii) relative promoter activity (rrnB P1 and Pveg) from early exponential phase till approximately two hours into stationary phase by qPE (Figure 1).
We detected a moderate decrease in [GTP] already during exponential phase ( Figure 1B). This moderate decrease was followed by a precipitous decline during the transition between the two phases. This correlated with a sharp spike in the (p)ppGpp level (Supplementary Figure S1). However, early on in the stationary phase, [GTP] even slightly increased and then remained at the same level till the end of the experiment. The activities of both rrnB P1 and Pveg decreased during the time course of the experiment-the activity of the former more than of the latter, consistent with the behavior of these promoters as reported in previous studies [10,11].
Surprisingly and interestingly, the activity rrnB P1 decreased even after the relative GTP concentration had been stabilized at a constant level. This strongly suggested that another mechanism, besides rRNA promoter regulation by [GTP], exists in the cell. DNA supercoiling is known to change between growth phases, typically the negative supercoiling from exponential phase becomes more relaxed in stationary phase, as demonstrated for Escherichia coli [39] and also B. subtilis [40]. Also, we noticed that the activity of Pveg significantly decreased, although the decrease was not as pronounced as that of the ribosomal promoter. As DNA topology is an important factor for gene expression regulation, we decided to address the potential of B. subtilis rRNA promoters to be regulated by the level of supercoiling.

Chromosome Relaxation Inhibits Total RNA Synthesis In Vivo
To test whether DNA topology could affect rRNA expression in vivo, we used novobiocin. Novobiocin is an antimicrobial compound that binds to the β subunit of gyrase and blocks its function by inhibiting ATP hydrolysis [41][42][43]. Gyrase relieves tension in DNA caused by transcribing RNAPs or helicases by creating more negatively supercoiled DNA. Hence, the inhibition of gyrase causes DNA in the cell to be more relaxed [44].
In this experiment, we first used total RNA as a proxy for rRNA synthesis as in exponential phase most of RNA synthesis comes for rRNA operons (~80% of RNA in cell is rRNA and tRNA [29,45]). We treated early-exponentially growing cells (OD 600~0 .3) with novobiocin or mock-treated them, and measured the rates of total RNA synthesis by following incorporation of radiolabeled 3 H-uridine into RNA. As a positive control, where we expected cessation of RNA synthesis, we treated cells with rifampicin, a wellcharacterized inhibitor of bacterial RNAP. Figure 2 shows that in the presence of novobiocin the synthesis of total RNA decreased/stopped, similarly as in the presence of rifampicin, suggesting that relaxation of the chromosome affects total RNA synthesis in the cell (Figure 2A).

Novobiocin-Induced Relaxation of DNA Affects the Activity of rrnB P1 In Vivo
Next, by RT-qPCR we monitored the response of rrnB P1 and Pveg to novobiocin treatment, using the same conditions as in the previous experiment. We grew cells carrying the appropriate fusions (rrnB P1-lacZ (LK134) and Pveg-lacZ (LK135)) to early-exponential phase (OD 600~0 .3) and either treated them with novobiocin or mock-treated them. In the case of rrnB P1, the promoter activity decreased after novobiocin treatment (as opposed to mock treatment), but in the case of Pveg, the promoter activity displayed the same moderate decline regardless of the novobiocin treatment, suggesting that rrnB P1 is more sensitive to changes in DNA topology ( Figure 2B,C).
We also measured the GTP levels in novobiocin treated cells. We observed that novobiocin-induced relaxation resulted in a massive increase in the GTP level in cell ( Figure 2D). The levels of ATP increased only slightly (Supplementary Figure S2). Thus, the activity of rrnB P1 and the level of GTP became uncoupled. These experiments suggested that DNA topology might affect the activity rRNA promoters, but it was also possible that unknown, secondary effects of the novobiocin treatment could be the cause.

Changes in DNA Topology Affect the Affinity of RNAP for iNTP In Vitro
To test directly whether DNA topology affects the activity of rRNA promoters, we performed in vitro transcription experiments. We had speculated that the in vivo decrease in the activity of rrnB P1 during stationary phase and in response to novobiocin treatment could be due to altered affinity of RNAP for iGTP at this promoter (induced by changes in supercoiling levels): the GTP level does not change but the open promoter becomes less stable, requiring more iGTP for maximal transcription. To address this hypothesis experimentally, we performed in vitro transcriptions with defined components. We used promoter core variants of rrnB P1 and Pveg cloned in the pRLG770 plasmid [11] (for details see Table 1 in Material and Methods section). The DNA templates were used in two different topological forms-in the negatively supercoiled plasmid form (SC), and in the relaxed form (LIN), using the same DNA construct but linearized with the PstI restriction enzyme (Supplementary Figure S3).
We performed multiple round transcriptions in vitro with increasing [GTP] (Figure 3). The GTP concentration required for half-maximal transcription (K GTP ) was used as a measure of the affinity of RNAP for iGTP at the promoter. A characteristic of rRNA promoters is their requirement for relatively high levels of iGTP for maximal transcription (due to unstable open complexes), reflected in high values of K GTP in vitro. Pveg, to the contrary, has a low value of K GTP . Experiments with SC templates confirmed previously published results [46], the K GTP for rrnB P1 was 277 ± 24 µM, and for Pveg 36 ± 9 µM. Experiments with the LIN templates then revealed that K GTP values for both promoters increased (rrnB P1 = 440 ± 25 µM, Pveg = 511 ± 78 µM). In the case of rrnB P1 the K GTP increased from SC to LIN~1.5x, and in the case of Pveg K GTP~1 4x. Surprisingly, the K GTP value of LIN Pveg was even higher than the value for rrnB P1 ( Figure 3B). Importantly, the experiments showed that the strength (the maximal level of transcription) of the rrnB P1 promoter dramatically decreased on the LIN template whereas in the case of Pveg the maximal level of transcription was comparable for both types of the template ( Figure 3A, primary data), confirming the hypothesis that DNA relaxation decreases the activity of rrnB P1 more than the activity of Pveg.
As the preceding experiments were done with the core version of the rrnB P1 promoter, we also decided to use an extended version of the promoter region to assess whether the surrounding sequence has significant effects. Therefore, we used a DNA fragment containing both rrnB P1 and rrnB P2 promoters in their native tandem arrangement. Each of them contained their respective native -60 to -40 regions encompassing the UP elements. UP elements are A/T-rich sequences that enhance promoter activity by binding the Cterminal domains of α-subunits of RNAP [47][48][49]. Although their stimulatory effect on rRNA promoters in B. subtilis is less pronounced than, e.g., in E. coli (~30x), it is still significant [11]. Experiments with these promoter versions yielded virtually the same results as with the core version ( Figure 3C). The K GTP for rrnB P1 (from the tandem promoter fragment) was 242 ± 31 µM for SC and 361 ± 46 µM for LIN. K GTP for rrnB P2 was 62 ± 13 µM for SC and 427 ± 61 µM for LIN (see Supplementary Table S1 and Supplementary Figure S4A,B). Similar results were obtained also with rrnO P1 and rrnO P2 promoters (Supplementary Figure S4C,D).
Hence, we concluded that for transcription from LIN templates higher concentrations of GTP are needed, regardless of the promoter. The increased K GTP of Pveg suggested that this change in RNAP affinity for the substrate iNTP might be responsible, at least in part, for the decrease in its activity during the transition from exponential to stationary phase. However, the moderate increase in K GTP of rrnB P1 suggested that other factor(s) must be involved in the decrease of this promoter's activity in vivo. A likely candidate factor was the affinity of RNAP for promoter DNA, i.e., formation of the closed complex or/and the intracellular level of RNAP.

Pveg and rRNA Promoter Affinities for RNAP Change with DNA Relaxation In Vitro
We tested the relative affinity of RNAP for promoter DNA by performing in vitro transcriptions as a function of increasing promoter DNA concentration. We used the tandem rrnB P1+P2 DNA fragment and Pveg. The GTP concentration was set to 1 mM to ensure high efficiency of open complex formation for the tested promoters. Affinity for RNAP of both rRNA promoters was unchanged or slightly decreased on relaxed templates, but this effect was not statistically significant ( Figure 4). Therefore, it is possible that the observed decrease in bulk transcription from rrnB P1 (SC vs LIN) in vitro could be due to yet another factor (e.g., promoter escape). The opposite trend was observed with Pveg: a relatively low level of the relaxed promoter DNA was able to saturate RNAP compared to the supercoiled template. This behavior could then explain why the activity of Pveg decreased less than the activity of rrnB P1 both in vitro and in vivo. Importantly, it was previously reported that the levels of RNAP subunits decrease from exponential to stationary phase [50,51] and we also observed this trend ( Figure 5).

The Effect of Supercoiling on Transcription In Vitro with Alternative Sigma Factors
To extend the study, we tested the effect of supercoiling on transcription from promoters dependent on alternative sigma factors: σ B , σ D , σ E , σ F and σ H . σ B is a general stress response sigma factor [52,53], σ D transcribes genes linked with the cell motility and flagella formation [54]. σ E and σ F are sigma factors of early sporulation [55,56]. σ H is responsible for transcription of early stationary genes [57].
We tested also σ N (ZpdN) that is present only in the B. subtilis NCIB 3610 strain. This strain possesses a large, low-copy-number plasmid pBS32, which was lost during domestication of the commonly used laboratory strains [58]. pBS32 carries genes responsible for cell death after mitomycin C (MMC) treatment, and this effect is dependent on σ N [24,25]. MMC is an antitumor antibiotic that induces DNA strand scission by DNA alkylation leading to crosslinking [59][60][61]. This DNA damage could lead to the formation of linear DNA fragments.
Sequences of respective promoters are listed in Supplementary Table S2. We performed transcriptions in vitro on SC and LIN DNA templates with saturating concentration of iNTP. In all but one cases it was the SC DNA that was the better template for transcription, similarly to what we observed with σ A (Figure 6).
The exception was σ N , which displayed about the same or higher activity on LIN DNA than on SC DNA, depending on the promoter ( Figure 6B). To show that this effect was not due to some unknown properties of the plasmid DNA bearing these promoters, we also tested a longer sigN promoter construct (sigN P2+P3). This construct contains σ A -dependent sigN P2 and σ N -dependent sigN P3 promoters [25] and allowed us to test the effect of SC vs LIN topology for two sigmas with the same template. The results are shown in Figure 6C: σ A -dependent P2 is more active on SC DNA whereas σ N -dependent P3 prefers LIN DNA for efficient transcription. Representative primary data are shown (radioactively labelled transcripts resolved by polyacrylamide electrophoresis). SC stands for supercoiled promoter DNA, LIN for linear DNA. Letters above the gels indicate the sigma factor used-A for σ A , B for σ B etc. Each sigma factor is depicted with different color (σ A , black; σ B , dark blue; σ D , green; σ E , yellow; σ F , brown; σ H , red and σ N , purple). For each promoter three independent experiments were performed. Transcription from SC was set as 1 for each promoter. Quantitation of results is shown in graphs below the respective primary data. The graphs show averages ±SD. The reactions with σ A on all promoter fragments were used to show that the observed transcription was due to the addition of the specific σ factors and not due to (theoretical) contamination of the core with σ A . (A) Transcription in vitro from selected σ B , σ D , σ E , σ F and σ H -dependent promoters. (B) Transcription in vitro from σ N -dependent promoters. (C) Transcription in vitro using a longer construct, sigN P2+P3. P2 is σ A -dependent, P3 is σ N -dependent.

Discussion
In this study we have identified the supercoiling level of DNA as a factor affecting the ability of Bacillus subtilis RNAP to transcribe from σ A -dependent rRNA promoters as well as from selected promoters depending on alternative σ factors.

rRNA Promoters and Pveg
In our experiments, the drop in rRNA promoter activity during transition to stationary phase was pronounced and concurrent with the onset of stationary phase. A decrease in the activity of B. subtilis rRNA promoters in stationary phase was also observed in [62]. However, they used promoter constructs fused with GFP and monitored promoter activity by measuring the intensity of fluorescent signal. GFP is relatively stable, so the decreases they reported were less pronounced than those observed in our experiments.
Here, we propose an updated model of regulation of B. subtilis rRNA promoters, revealing supercoiling as a factor involved in their control. The more negatively supercoiled DNA in exponential phase contributes to the high activity of B. subtilis rRNA promoters. As this negative supercoiling becomes more relaxed when the cell transitions into stationary phase, this likely contributes to the decrease in the activity of RNAP at rRNA promoters. This is in part due to the decreased affinity of RNAP at rRNA promoters for the initiating GTP but also to a so far unknown step during transcription initiation (e.g., isomerization, promoter escape). The activity of rRNA promoters in stationary phase is also likely affected by the decreased RNAP concentration. The decrease in the available RNAP pool is further exacerbated by the association of the RNAP:σ A holoenzyme with 6S-1 RNA that sequesters it in an inactive form in stationary phase [63]. The combined effect results in the shut-off of rRNA synthesis. Previously, for E. coli rRNA promoters, the decreased stability of the open complex was identified as the main kinetic intermediate affected by supercoiling [64]. We also note that in S. aureus in post-exponential growth phase the downregulation of rRNA is independent of ppGpp or NTP pools [17], and it is possible that DNA topology might be a factor contributing to this downregulation.
In B. subtilis, correlations between the supercoiling level and rRNA activity could be found also in the forespore. Within the developing spore, DNA becomes more negatively supercoiled then in stationary phase [40] and this correlates with an increase in rRNA activity in the forespore [62]. Interestingly, during novobiocin treatment the GTP level increases in B. subtilis and the changes in DNA topology override its stimulatory effect so that the net result is a decrease in the activity of rrnB P1. This is the first observation of a situation where the GTP level and rRNA promoter activity do not correlate in B. subtilis. We note that supercoiling was also reported to be involved in rRNA expression in yeast although the mechanistic aspects of this regulation are less understood [65].
The activity of the control Pveg promoter also decreases from exponential to stationary phase but the decrease is not as pronounced as in the case of rrnB P1. The decrease in the activity of Pveg can be attributed, at least in part, to its increased requirement for the concentration of the iNTP when DNA supercoiling relaxes. Nevertheless, the affinity of Pveg for RNAP seems to increase with DNA relaxation and this likely partially counteracts the negative effect on open complex formation.

Transcription with Selected Alternative σ Factors
Transcription experiments with promoters dependent on alternative σ factors revealed that linear templates are poorer substrates for the majority of them (σ B , σ D , σ E , σ F , and σ H ). This trend was previously reported also for RNAP:σ H transcribing from the spoIIA promoter [66]. For forespore-specific σ F , this is consistent with the DNA supercoiling increase in the forespore [40]. For σ H and σ E that are active in stationary phase, although activities of respective promoters strongly decreased with reduced supercoiling in vitro, this likely reflects the physiologically relevant requirements for their activities in the cell in stationary phase. Also, the decrease in the level of supercoiling in stationary phase is likely not as extreme as in our in vitro experiments where it was used to better visualize the effects.
The exception was σ N , where transcription (SC vs. LIN) is either relatively unaffected or even increased on linear templates. This is likely physiologically important as mitomycin, which induces σ N expression [24], causes also DNA relaxation and σ N may have evolved to be most active under such conditions. The proficiency of RNAP:σ N on linear templates then may stem from the relatively short spacers of σ N dependent promoters (15 bp compared to 17 bp for σ A , [67]), analogously to σ 70 and σ S of E. coli where the different σ activities were proposed to be due to preferences for differently DNA supercoiled templates [68][69][70].

Conclusions
To conclude, our findings extend the current model of rRNA promoter regulation in B. subtilis and reveal the effect of supercoiling on transcription with main and alternative σ factors.
Supplementary Materials: The following are available online at https://www.mdpi.com/2076-2 607/9/1/87/s1, Figure S1: Relative GTP and (p)ppGpp levels after entry into stationary phase. Figure S2: Effect of novobiocin-induced relaxation of chromosome on ATP levels. Figure S3: SC and LIN promoter DNA on agarose gel. Figure S4: The affinity of RNAP for iNTP in vitro changes on different DNA templates. Table S1: The K GTP values for the promoters tested in the transcriptions in vitro. Table S2: Alternative σ factor-dependent promoters used in the study. Funding: This work was supported by 20-12109S to LK from the Czech Science Foundation.