RapA opens the RNA polymerase clamp to disrupt post-termination complexes and prevent cytotoxic R-loop formation

Abstract

Following transcript release during intrinsic termination, Escherichia coli RNA polymerase (RNAP) often remains associated with DNA in a post-termination complex (PTC). RNAPs in PTCs are removed from the DNA by the Swi2/Snf2 ATPase RapA. Here, we determined PTC structures on negatively-supercoiled DNA as well as of RapA engaged to dislodge the PTC. We found that core RNAP in the PTC can unwind DNA and initiate RNA synthesis but is prone to producing R-loops. Nucleotide binding to RapA triggers a conformational change that opens the RNAP clamp, allowing DNA in the RNAP cleft to reanneal and dissociate. We show that RapA helps control cytotoxic R-loop formation in vivo, likely by disrupting PTCs. We suggest that analagous ATPases acting on PTCs to suppress transcriptional noise and R-loop formation may be widespread. These results hold significance for the bacterial transcription cycle and highlight a role for RapA in maintaining genome stability.

RNA-polymerase (RNAP) is the central enzyme of transcription across all domains of life. In the canonical bacterial transcription cycle, the catalytic core RNAP (E; subunit composition α₂ββ'ω) combines with a promoter-specificity σ factor to form the holoenzyme (Eσ⁷⁰) capable of initiating transcription from specific promoter sequences ¹. Once transcription initiates and the RNAP becomes committed to elongating the nascent RNA chain, the σ subunit generally releases (although not always) ² and core RNAP completes the elongation and termination phases of the transcription cycle. Although the elongation phase is highly processive, during intrinsic termination the completed RNA transcript is rapidly released from the complex ³, followed in the canonical scheme by release of RNAP from the DNA template.

Recent single-molecule investigations have shown that the canonical bacterial transcription cycle must be reevaluated ^4,5. In these studies, core RNAP was observed to remain associated with the template DNA after RNA transcript release at intrinsic terminators in a post-termination complex (PTC) in vitro. The RNAP in PTCs can diffuse in both directions on the DNA by sliding or hopping, and can flip 180 degrees on the DNA ^4-6. The diffusing PTCs could sometimes initiate transcription in a promoter-dependent manner after reassociating with σ from solution, consistent with in vivo results ⁴. Available data indicate that complexes with the same properties as PTCs can be reconstituted by incubating core RNAP directly with DNA (reconstituted-PTCs, or rPTCs) ^4,7,8.

The Sw2/Snf2 ATPase RapA is widespread throughout bacteria and is expressed to equal abundance as σ⁷⁰ in Eco ^9,10. RapA binds to core RNAP (K_d ~5-10 nM) but not Eσ^{70 9,11}. Binding of RapA to RNAP strongly stimulates RapA ATPase activity, but nucleic acids neither bind strongly to RapA nor stimulate its ATPase activity ^7,9-11. While rapA null mutants do not display a clear in vivo phenotype in rich medium, stresses such as osmotic shock, exposure to sodium deoxycholate, or salt stress produce a protracted recovery phase in rapA knockouts ^12,13. Additionally, rapA deletion diminishes antibiotic resistance in biofilms ¹⁴. Lastly, in Vibrio cholerae, deletion of the rapA homolog produces a >1000-fold decrease in colonization proficiency when cells attempt to engage the acid-tolerance response, which is important for virulence ¹⁵.

Many cellular functions have been proposed for RapA ^16-21. In in vitro transcription reactions, RapA strongly stimulates multi-round transcript production, suggesting that it acts following termination ^12,16,22-24. A recent single-molecule analysis defined the molecular target for RapA in the transcription cycle as the PTC; RapA forms a transient RapA-PTC complex that rapidly (within seconds) resolves upon ATP hydrolysis, releasing RapA and the RNAP from the DNA ⁷. Although several structural studies have provided insights into the interactions between RapA and core RNAP or the RNAP elongation complex (EC) ^18,20-22, the structural mechanism for RapA function remains elusive in part because a structure of RapA with its true molecular target, the PTC, has not been determined.

Here, we structurally and biochemically characterize Eco rPTCs. We find that core RNAP can associate with negatively-supercoiled DNA, is capable of generating a transcription-competent transcription bubble de novo, and can initiate RNA synthesis at physiological nucleotide concentrations, all in the absence of a σ factor. We also used a non-hydrolyzable ATP analog, ADP-AlF₃ ²⁵, to trap RapA associated with its functional target, the rPTC. A cryo-EM structure of RapA(ADP-AlF₃)-rPTC reveals that RapA acts to impart a large conformational change in the RNAP that promotes release from the DNA. We find that rPTC-mediated σ-independent transcription generates R-loops, potentially contributing to mutagenesis and posing a threat to genome stability in vivo ²⁶, but that RapA function in vivo plays a role in controlling R-loop formation. These findings have important implications for the canonical bacterial transcription cycle and the role played by PTCs and RapA in anti-sense and/or pervasive transcription and in genome stability ²⁷.

Core RNAP forms a transcription bubble

For structural analysis of rPTCs, we sought to eliminate one-dimensional diffusion of the RNAP off the end of the DNA ⁴ as well as binding of RNAP to duplex DNA ends (end-binding) ²⁸. We therefore assembled rPTCs by combining Eco core RNAP with a negatively-supercoiled circular DNA template, incubated briefly at 37°C, then prepared grids for cryo-EM analysis (Fig. 1a). Visual inspection of micrographs revealed RNAP molecules associated with DNA, including segments of extended DNA decorated with chains of RNAP molecules like beads on a string (Fig. 1b). In our processing of the micrographs, we were unable to identify particles without DNA bound in the RNAP cleft (Extended Data Fig. 1). Steps of maximum likelihood classification ²⁹ revealed three distinct conformational classes (Figs. 1c-f, Table 1, Extended Data Figs. 1 and 2). The cryo-EM density for the DNA in each of the classes was poorly resolved compared with the RNAP density, consistent with the expected local heterogeneity in the DNA due to non-specific association of rPTCs and lateral diffusion. The DNA was modeled with the aid of previously determined DNA-RNAP complexes ³⁰, and features of the DNA were conservatively interpreted.

Fig. 1 ∣. Reconstituted-PTCs unwind negatively-supercoiled DNA.

a. Schematic illustrating the experimental setup. Eco core RNAP (green) was incubated with circular, negatively-supercoiled plasmid DNA in the absence of RNA, NTPs, and σ⁷⁰. Created in BioRender. Campbell, E. (2023) BioRender.com/s21n287.

b. Representative micrograph illustrating core RNAP molecules associated with DNA (5,043 micrographs were collected; Extended Data Figure 1).

c.-e. Structures of rPTCs, c. rPTCc, d. rPTCi, e. rPTCo, determined by cryo-EM. The core RNAP is shown as a transparent molecular surface, revealing the DNA and active-site Mg²⁺ in the RNAP active-site cleft. The DNA is shown in cartoon format along with transparent cryo-EM difference density.

f. RNAP clamp conformations for the rPTC structures. The rPTCo structure was used as a reference to superimpose rPTCc and rPTCi via α-carbon atoms of the RNAP structural core, revealing a common RNAP structure (shown as a grey molecular surface) but with clamp conformational changes characterized as rigid body rotations about a rotation axis perpendicular to the page (denoted by the black dot). The clamp modules are shown as backbone cartoons with cylindrical helices. The angles of clamp opening are shown relative to rPTCo (0°).

g. Superposed DNAs from rPTCc (magenta), rPTCi (cyan), rPTCo (green), and an active EC (6ALH ⁷¹; orange). The RNAP active site Mg²⁺ is shown as a brown sphere.

Table 1 ∣.

Cryo-EM data collection, refinement, and validation statistics for RapA^†-PTC and rPTC classes.

Dataset		rPTC		RapA†-PTC
Data collection and processing
Microscope		FEI Titan Krios		FEI Titan Krios
Voltage (kV)		300		300
Detector		Gatan K3		Gatan K3
Electron exposure (e–/Å2)		55.9		52.1
Defocus range (μm)		−0.25 to −4.2		−0.8 to −2.5
Data collection mode		Counting Mode		Counting Mode
Pixel size (Å)		1.076		1.076
Symmetry imposed		C1		C1
Initial particle images (no.)		1,261,036		3,560,985
Refinement
Structure	rPTCc	rPTCi	rPTCo	RapA†-PTC
EMDB	EMD-40930	EMD-40931	EMD-40922	EMD-40943
PDB	8T00	8T02	8SZW	8T0L
Final particle images (no.)	13,101	49,701	86,865	100,010
Map resolution (Å) - FSC threshold 0.143	4.7	3.8	3.6	3.6
Map resolution range (Å)	3.4 - 7.8	2.4 - 6.9	2.8 - 6.9	2.8 - 7.2
Initial model used (PDB code)	6ALH	6ALH	6ALH	Ab-initio via screening dataset
Map sharpening B factor (Å2)	−126.0	−108.9	−114.5	−117.5
Model composition
Non-hydrogen atoms	50,247	51,102	51,717	65,326
Protein residues	3,093	3,146	3,183	4,152
Nucleic acid residues (DNA)	52	51	52	-
Ligands	2 Zn2+, 1 Mg2+	2 Zn2+, 1 Mg2+	2 Zn2+, 1 Mg2+	2 Zn2+, 1 Mg2+, 1 ADP-AlF3
B factors (Å2)
Protein	289.6	109.2	69.94	166.2
Nucleic acid	515.0	247.9	134.2	-
Ligands	382.0	139.8	83.67	233.2
R.m.s. deviations
Bond lengths (Å)	0.003	0.004	0.003	0.005
Bond angles (°)	0.662	0.643	0.63	0.668
Validation
MolProbity score	1.82	1.99	1.83	2.62
Clashscore	7.07	9.67	7.68	9.74
Poor rotamers (%)	0	0.22	0.22	5.59
Ramachandran plot
Favored (%)	93.49	92.06	93.93	90.2
Allowed (%)	6.48	7.91	5.97	9.68
Disallowed (%)	0.03	0.03	0.09	0.12

In the closed-DNA rPTC (rPTCc, ~9% of the particle population; 4.7 Å nominal resolution; Fig. 1c, Extended Data Figs. 1 and 2a-c), the RNAP clamp ^31,32 was notably wide open, 24° open compared to a 0° reference (Fig. 1f). The RNAP cleft was occupied with severely kinked (~90°) but mostly duplex (closed) DNA (Fig. 1c). The vicinity of the DNA kink was poorly resolved and potentially a site of bubble nucleation, with the conserved Switch 2 (Sw2) and Fork Loop 2 (FL2) RNAP structural elements ³² proximal to the kink (Extended Data Fig. 3a). The RNAP β’rudder was well-resolved, situated in the major groove of the upstream double-stranded DNA (dsDNA) (Extended Data Fig. 3a).

In the intermediate-bubble rPTC (rPTCi; 33% of the particle population; 3.8 Å nominal resolution; Fig. 1d, Extended Data Figs. 1 and 2d-f), the RNAP clamp was relatively closed (0.6°; Fig. 1f). The RNAP cleft was occupied by DNA with a clear melted bubble of ~5 nucleotides (nts) enclosed completely within the cleft (Fig. 1d). The single-stranded non-template strand (nt-strand) DNA was relatively well-resolved, but the template-strand (t-strand) was dynamic, poorly-resolved, and not near active site (Fig. 1g, Extended Data Fig. 3b).

The downstream fork junction of the rPTCi bubble appears to be stabilized by the insertion of Sw2 and FL2 elements between the melted strands at the downstream edge of the bubble (Extended Data Fig. 3b). The stabilized downstream fork junction observed together with the Sw2-FL2 insertion is consistent with prior biochemical findings implicating Sw2 in DNA melting, initiating-nucleotide binding, and promoter escape, along with EC stability ^33,34. The β’rudder, previously implicated in EC and RPo stability ³⁵, was disordered, owing to clash between the rudder and the duplex DNA immediately upstream of the growing bubble (Extended Data Fig. 3b).

In the open-bubble rPTC (rPTCo, 58% of the particle population; 3.6 Å nominal resolution; Fig. 1e, Extended Data Figs. 1 and 2g-i), the RNAP clamp was closed on the DNA (Fig. 1f). The RNAP cleft was occupied by DNA with a more extensive bubble of ~7-8 nt (the upstream edge of the bubble was poorly-resolved), which propagated in the upstream direction compared with the smaller bubble in rPTCi (Figs. 1d and 1e). As in rPTCi, the single-stranded nt-strand DNA was relatively well-resolved while the t-strand was dynamic and poorly-resolved but occupied positions near the RNAP active-site Mg²⁺, similar to a bona fide elongation complex (EC; Fig. 1g).

As in rPTCi, the Sw2 and FL2 elements of rPTCo were inserted between the melted stands at the downstream fork of the transcription bubble (Extended Data Fig. 3c). The β’rudder was resolved and fully inserted between the two melted strands, likely stabilizing the upstream edge of the bubble (Extended Data Fig. 3c).

Rotation of a swivel module is associated with paused elongation complexes and appears to inhibit RNAP motions required to complete NTP binding, catalysis, and translocation ^30,36-38. Compared to an unswiveled reference structure (8EG8) ³⁰, both rPTCi and rPTCo were significantly swiveled (4.4° and 4.0°, respectively; Extended Data Fig. 4).

Cryo-EM structure of RapA(ADP-AlF₃)-rPTC

Recent work identified PTCs as the target of the Swi2/Snf2 ATPase RNAP-recycling factor RapA ⁷. Inlow et al. ⁷ used single-molecule analyses to identify two kinetically distinct RapA-rPTC assemblies formed during ATP-dependent RNAP recycling (denoted RapA-PTC and RapA^†-PTC) and proposed a kinetic scheme in which these assemblies are intermediates in disruption of PTCs by RapA (Extended Data Fig. 5a). Initial RapA association with PTCs was independent of nucleotide in solution, consistent with the first intermediate, RapA-PTC, containing apo-RapA (Extended Data Fig. 5a). Cryo-EM structures of apo-RapA engaged with core RNAP (7MKQ ¹⁸) and with ECs (7MKN ¹⁸, 7M8E ²¹) all yielded closed-clamp RNAP structures bound to apo-RapA in a nearly identical pose with each other, suggesting that the configuration of the complex of apo-RapA with RNAP is independent of the nucleic acid binding status of the RNAP. We therefore propose that the apo-RapA-RNAP and apo-RapA-EC structures represent good models for the structure of the first RapA-PTC intermediate.

The second kinetically distinct intermediate, RapA^†-PTC, leads to PTC disruption at a rate that is strongly dependent on nucleotide in solution ⁷. In the presence of ATP, PTC disruption by RapA was at least 150-fold faster than when ATP was absent or when a non-hydrolyzable ATP analog was present. In the presence of the analog, RapA^†-PTC was >20-fold more stable than the first intermediate, and more than half of the assemblies exhibited a characteristic lifetime close to 5 min ⁷, a time scale compatible with cryo-EM grid preparation. Therefore, we generated rPTCs on negatively-supercoiled circular DNA (Fig. 1a) and subsequently introduced RapA complexed with the non-hydrolyzable ATP analog ADP-AlF₃ ²⁵ and analyzed the resulting complexes by single-particle cryo-EM (Extended Data Fig. 5b). The results revealed a RapA(ADP-AlF₃)-PTC structure (3.6 Å nominal resolution; Fig. 2a, Extended Data Fig. 5b and 5c, Table 1) with novel characteristics compared to previously described RapA-RNAP or RapA-EC complexes ^18,21 and that we equate with RapA^†-PTC (Extended Data Fig. 5a). These novel properties include: i) the conformation of RapA, ii) the overall disposition of RapA with respect to the RNAP, and iii) the conformation of RNAP and the associated DNA. Cryo-EM density in the RNAP cleft of RapA^†-PTC was consistent with kinked duplex DNA (Fig. 2a), very similar to the DNA occupying the rPTCc RNAP cleft (Fig. 1c). The cryo-EM density for RapA was relatively poorly-resolved compared with the RNAP (Extended Data Fig. 5c), presumably due to conformational heterogeneity that could not be resolved by classification. The molecular model for RapA was initially constructed by rigid-body fitting of previously determined RapA domains ^21,22. Modeling of the nucleotide bound in the RapA active site (Fig. 2b) was guided by the structure of a RecA-ATPase homolog bound to a non-hydrolyzable ATP analog (chromodomain-ATPase portion of yeast Chd1 chromatin remodeler; 3MWY) ³⁹.

Fig. 2 ∣. RapA opens the RNAP clamp to promote DNA dissociation.

a. Cryo-EM map (local-resolution filtered) ⁷² of RapA(ADP-AlF₃)-rPTC (RapA^†-PTC).

b. Conformational changes in RapA. (lower left) Overall view of RapA^†-PTC. The relationship between this view and the view in (a) is shown. RapA, partially obscured by RNAP, is outlined in green. (center) RapA from apo-RapA-PTC (7M8E model) ²¹ and RapA^†-PTC were superimposed via the 1A domain (Supplementary Data Table 1). RapA^† is in cartoon format with transparent molecular surfaces; apo-RapA domains are colored outlines. Rotations for the conformational changes of each domain (apo-RapA -> RapA^†) are shown. Note the large 51° rotation of the RapA-NTD (green). (lower right) Zoomed-in view of the nucleotide-binding site (between the 1A and 2 domains). Modeled ADP-AlF₃ and cryo-EM difference density (transparent surface) corresponding to the bound nucleotide are shown.

c. Overall change in architecture of apo-RapA-PTC and RapA^†-PTC.

(left) apo-RapA-PTC shown as a molecular surface. RNAP surfaces are transparent, showing nucleic acids inside the RNAP cleft (t-strand DNA, dark gray; nt-strand DNA, light gray; RNA, red). RapA and the RNAP clamp are outlined in green and red, respectively.

(right) RapA^†-PTC shown similarly with transparent RNA surfaces revealing DNA inside the cleft.

(middle) Superimposed outlines of RNAP clamps from apo-RapA-PTC (lighter shades) and RapA^†-PTC. The entire RapA molecule rotates 65° on the RNAP surface, and the clamp opens 24°.

d. Boxed regions of (c) are magnified, highlighting the RapA-NTD:β'ZBD interface.

(left) apo-RapA-PTC

(right) RapA^†-PTC. The green and red outlines denote positions of the RapA-NTD (green) and RNAP clamp (red) from apo-RapA-PTC. A 51° rotation of RapA^†-NTD results in a 24° clamp opening.

e. Results from single-molecule fluorescence microscopy measuring effective dissociation rates (reciprocal of the average RNAP dwell time on DNA, <τ>⁻¹) of surface-tethered rPTCs formed with RNAP or a ΔZBD-RNAP mutant, by RapA plus ATP or in controls lacking RapA or lacking ATP. Number of complexes from left to right: N=308, 184, 306, 122, 272, 130. Data are presented as mean values ± SEM. Also see Extended Data Fig. 6c.

ADP-AlF₃ triggers global conformational changes in RapA

The cryo-EM density of RapA^†-PTC clearly showed ADP-AlF₃ occupying the RapA nucleotide binding site between the 1A and 2 (RecA) domains (Fig. 2b; the RapA domains discussed herein are defined in Supplementary Data Table 1). As expected for a RecA-type ATPase ⁴⁰, occupancy of the nucleotide binding site gave rise to a large change in the orientation of the RecA ATPase domains with respect to each other (RapA domains 1A and 2, Fig. 2b) compared to apo-RapA structures ^18,21, corresponding to a 33° rotation of domain 2 towards domain 1A (Fig. 2b). This large conformational change induced by ADP-AlF₃ binding triggered a complex series of allosteric conformational changes in the other RapA domains (with respect to the reference domain 1A; Fig. 2b, Supplementary Video 1), most notably resulting in a motion of the RapA-NTD corresponding to a 51° rotation and a 27 Å translation of the domain center of mass (Fig. 2b).

RapA conformational changes mechanically open the RNAP clamp

Comparing apo-RapA-EC (7M8E ²¹) with RapA^†-PTC, the binding of ADP-AlF₃ induced significant conformational changes in RapA (Fig. 2b, Supplementary Video 1). These changes were accompanied by a large rearrangement of RapA as a whole, corresponding to a 65° rotation of RapA relative to the RNAP (Fig. 2c, Supplementary Video 2).

Major RapA-RNAP interfaces include the RecA domains (RapA domains 1A and 2; Fig. 2b) with the RNAP βflap-tip and the RapA-NTD with the RNAP β'ZBD (Figs. 2c and 2d). The main anchor point for the 65° RapA rotation (with respect to the RNAP) is the RapA-1A:βflap-tip interaction. The rotation of RapA about this anchor point, which is accommodated by flexibility of the βflap-tip (Extended Data Fig. 6a, Supplementary Video 2), results in the large motion (51° rotation, 27 Å translation) of the RapA-NTD described previously (Figs. 2b and 2c). The RapA-NTD forms a significant interface with the β'ZBD [583 Ų interface area ⁴¹]; consequently, the large motion of the RapA-NTD pulls the β'ZBD and associated RNAP clamp with it, resulting in a 24° opening of the RNAP clamp (Figs. 2c and 2d, Supplementary Video 2). The RapA-Spacer domain also appears to wedge itself into the now open RNAP cleft, possibly stabilizing the open-clamp conformation (Extended Data Fig. 6b).

The relatively closed clamps observed in the rPTCo and rPTCi structures were accompanied by DNA melting observed in those structures (Figs. 1d and 1e), whereas rPTCc, with its open clamp, contained apparently closed, duplex DNA in its RNAP cleft (Fig. 1c). Similary, the cryo-EM density in the RNAP cleft of the open-clamp RapA^†-PTC is consistent with duplex DNA (Fig. 2a). Clamp opening in RapA^†-PTC would not only allow the DNA to anneal into a duplex but would also be expected to destabilize DNA binding and promote dissociation of the DNA.

The RapA-NTD pulls on the β'ZBD to open the RNAP clamp

Our model for RapA function predicts that the RapA-NTD:β'ZBD interface is crucial, allowing the motion of the RapA-NTD to pull on the β'ZBD and open the RNAP clamp. We tested this model by assessing the ability of RapA to disrupt PTCs generated from an RNAP derivative lacking the β'ZBD (ΔZBD-RNAP). Non-supercoiled fluorescently-labeled DNA circles tethered to the surface of a glass flow chamber were preincubated with either 1.5 nM RNAP or 4 nM ΔZBD-RNAP, each dye-labeled via a SNAP-tag fusion to the β′ subunit C-terminus (RNAP⁵⁴⁹ and ΔZBD-RNAP⁵⁴⁹, respectively). Single-molecule total internal reflection microscopy (smTIRF) revealed individual molecules of RNAP at the locations of single DNA molecules, indicating rPTC formation. Note that a higher concentration of ΔZBD-RNAP⁵⁴⁹ was required to achieve quantities of rPTCs optimal for experiments; at 1.5 nM RNAP⁵⁴⁹, we attained 40-50% DNA occupancy by RNAP, compared to ~25% occupancy with 4 nM ΔZBD-RNAP⁵⁴⁹.

As previously reported ⁷, replacement of the flow chamber solution with a buffer containing 1 mM ATP and a fluorescent RapA derivative (5 nM RapA⁶⁵⁰) accelerated the loss of RNAP from the surface compared to controls without RapA or without ATP (Fig. 2e, Extended Data Fig. 6c), indicating rPTC disruption. By contrast, no disruption above control was detected with ΔZBD-RNAP, highlighting the essential role of the interaction between the RapA-NTD and the RNAP-β'ZBD in stimulating RNAP clamp opening and subsequent dissociation of RNAP from DNA.

Core RNAP can initiate transcription

Given the DNA disposition in rPTCo and the possibility that the t-strand DNA can occupy a position near the RNAP active-site similar to an active EC (Fig. 1g), we compared the transcription activity of the rPTCs with that of Eσ⁷⁰ on a circular DNA plasmid template containing the strong T7A1 promoter and T7 intrinsic terminator sequence, expected to yield a specific 161 nt transcript with Eσ⁷⁰ (Fig. 3a). We used relatively high NTP concentrations (500 μM each) to approximate physiological concentrations ⁴², but note that the relatively high E concentration (0.5 μM) was chosen to drive rPTC formation and not to mimic physiological conditions, where the concentration of free E is very low ⁴³. The resulting RNA transcripts were extracted, quantifed, and the size distributions analyzed.

Fig. 3 ∣. Transcription initiation by core RNAP.

a. Schematic illustrating the experimental setup. Eco core RNAP (E) or Eσ⁷⁰ was incubated with circular, negatively-supercoiled plasmid DNA (paAR1707)⁷³ in the presence of NTPs (500 μM each) for 15 min at 37°C. The total amount and size distribution of the resulting RNA transcripts were then analyzed.

b. Histogram plot showing the total amount of RNA produced from each transcription reaction. The bars denote the average of three to six independent measurements (individual data points shown). Data are presented as mean values ± SEM. Statistical significance of differences between samples was determined using unpaired, two-tailed t-test.

c. Size distribution of transcripts resulting from two independent Eσ⁷⁰ transcription reactions (red shades) and two independent core RNAP reactions (green shades). The area under the curve for each profile was normalized according to the total RNA produced.

d. Size distribution of transcripts resulting from the core RNAP reactions (green shades) along with similar reactions treated with RNase HI (blue shades). The area under the curve for each profile was normalized according to the total RNA produced.

The overall amount and size distribution of the RNA produced by core RNAP vs. Eσ⁷⁰ were clearly distinct (Figs. 3b and 3c), confirming the absence of significant σ⁷⁰ contamination in our purified core RNAP. As expected, the transcription reactions with core RNAP produced less total RNA than Eσ⁷⁰ (about 3-fold less; Fig. 3b). Reactions using plasmid DNA pre-incubated with topoisomerase I to relax supercoils showed a marked (more than 2-fold) decrease in core RNAP transcription whereas Eσ⁷⁰ transcription was unaffected (core RNAP transcription on the topoisomerase I-treated template was ~6-fold less than Eσ⁷⁰; Fig. 3b). Transcription from the T7A1 promoter has been shown to be relatively insensitive to DNA supercoiling ⁴⁴. The decreased core RNAP transcription on relaxed DNA could be explained by a decrease in the overall number of PTCs, or a decrease in the transcription activity of the PTCs. Since PTCs on relaxed DNA are extremely stable (K_d < 15 nM) ⁸ we favor the hypothesis that negative supercoiling shifts the equilibrium away from transcriptionally inactive PTCc and PTCi and towards transcriptionally active PTCo (Figs. 1c-e).

The size profile of the RNA products for the Eσ⁷⁰ reactions was dominated by the expected specific transcript of 161 nt (Fig. 3c). Additional smaller peaks were superimposed onto a broad distribution of RNAs spanning approximately 1 kb to > 6 kb in length (Fig. 3c). By contrast, the core RNAP transcription products were characterized by a broad, relatively featureless distribution of RNAs over similar RNA lengths (1 kb to > 6 kb; Fig. 3c), consistent with non-specific transcription initiation by the rPTCs. We repeated the transcription reactions, comparing Eσ⁷⁰ with E on a derivative of pAR1707, pJB2, that lacked the strong T7 A1 promoter (as well as almost all other annotated promoters; see Extended Data Fig. 7). As expected, the amount and size profile of Eσ⁷⁰ transcripts from pJB2 were substantially altered compared to pAR1707 (Extended Data Figs. 7c and 7d). By contrast, the amount and size profile of E transcripts from pJB2 and pAR1707 were nearly identical (Extended Data Figs. 7c and 7e), indicating that the presence of the strong T7 A1 promoter on pAR1707 had little to no effect on transcription initiated by E. These results clearly indicate that rPTCs can non-specifically initiate transcription, consistent with the cryo-EM structure of rPTCo that shows a complex containing a bubble with the t-strand loaded into the active site (Fig. 1e).

PTC-initiated transcription is prone to R-loop formation

We noted that the broad distribution of large RNAs (> ~1 kb) for both Eσ⁷⁰ and core transcription reactions contained significant amounts of RNA chains longer than the plasmid itself (4.836 kb). We hypothesize that non-specific rPTC initiation generates elongating complexes with R-loops (persistent RNA/DNA hybrid) in their wake. The upstream RNA/DNA hybrid would prevent the formation of RNA secondary structure in the upstream transcript (such as terminator hairpins), causing the RNAP to ignore terminators and continuously transcribe around the circular DNA template in a rolling-circle transcription mechanism (Eco RNAP can produce > 7 kb transcripts from circular DNA templates in this manner) ⁴⁵.

Treatment of the PTC transcription reactions with Eco RNase HI, which specifically hydrolyzes the RNA phosphate backbone when the RNA is hybridized to DNA, significantly reduced the size distribution of the RNAs (Fig. 3d), confirming extensive production of R-loops. The RNase HI treatment reduced the peak of the RNA size distributions by about 4.8 ± 0.1 kb, corresponding to the size of the circular DNA template (Fig. 1a). This result supports the hypothesis that the core RNAPs transcribed all the way around the plasmid, generating an RNA/DNA hybrid the length of the DNA template. Further transcription would displace the RNA downstream of the RNAP from the DNA template, allowing the production of RNA transcripts longer than the DNA template but always leaving an RNA/DNA hybrid the length of the DNA template.

RapA contributes to the control of cytotoxic R-loops in vivo

Excessive PTCs on the DNA can lead to non-specific transcription initiation and generate extensive R-loops (Fig. 3d). Excessive R-loop formation in vivo is a threat to genomic stability and can be lethal ^26,46-48. We hypothesize that RapA may contribute to the control of R-loops in vivo by removing PTCs, a potential source of R-loops. We first tested this hypothesis by comparing the growth of an Eco rapA null mutant (ΔrapA) with the wild-type parent strain (wt) under conditions of R-loop stress (Extended Data Fig. 8), induced by growing the cells in the presence of bicyclomycin (BCM), a selective inhibitor of the Rho termination factor ⁴⁹. Rho plays an essential role in Eco by suppressing R-loop formation ^47,50. The cells carried pBAD18rnhA (a plasmid expressing rnhA, encoding Eco RNase HI, under control of the arabinose-inducible P_BAD promoter) ⁵¹ or the empty pBAD18 plasmid as a negative control. We reasoned that a growth defect of the ΔrapA strain under R-loop stress would be compensated by overexpression of RNase HI via induction of pBAD18rnhA, directly confirming the role of R-loops. We analyzed cell growth (monitored by OD_{600 nm}) using two parameters, the doubling time during log-phase growth and t_1/2, the time for the cells to reach half their OD_{600 nm} plateau (Extended Data Figs. 8a-c).

In the absence of BCM, the ΔrapA strain had a small but reproducible growth defect (less than 1.2-fold) compared with wt under all conditions tested (Extended Data Figs. 8d and 8e). The growth rate defect of ΔrapA (relative to wt) increased significantly (doubling time, 1.3-fold; t_1/2, 1.7-fold) when the cells were grown in the presence of BCM at 0.5X MIC (Figs. 4a and 4b, Extended Data Figs. 8f and 8g). The increased growth defect of ΔrapA in the presence of BCM was corrected by the expression of RNase HI (Figs. 4a and 4b, Extended Data Figs. 8f and 8g).

Fig. 4 ∣. RapA suppresses cytotoxic R-loops in vivo.

a.-b. Growth parameters (doubling times and t_1/2) for wt and ΔrapA Eco cells carrying pBAD18 (empty vector) ⁷⁴ or pBAD18rnhA (RNase HI) ⁵¹ without (BCM=0) or with 0.5X MIC BCM (BCM=0.5) were determined (Extended Data Fig. 8). Histograms show the ratios (ΔrapA/wt). Error bars denote standard error (N=3 measurements). Statistical significance of differences between samples was determined using an unpaired, two-tailed t-test.

a. Ratios of doubling times at matching conditions (ΔrapA/wt). Data are presented as best fit values ± SEM (calculated from 95% confidence limits of best fit).

b. Ratios of t_1/2 at matching conditions (ΔrapA/wt). Data are presented as best fit values ± SEM (calculated from 95% confidence limits of best fit).

c. The ΔrapA mutant cannot grow on LB plates + 1 M NaCl (upper right quadrant), but overexpression of RNase HI enables growth (upper left quadrant).

To test the hypothesis that RapA contributes to the control of R-loops in vivo at a distinct growth condition and without the use of BCM, we took advantage of a previous finding that deletion of rapA renders Eco unable to grow on LB plates with 1 M NaCl (Fig. 4c) ¹². The apparent lethality of 1 M NaCl to ΔrapA was rescued by expression of RNase HI (Fig. 4c). This points to a role for R-loop toxicity in the inability of ΔrapA to grow on 1 M NaCl.

Eco responds to osmotic stress (such as 1 M NaCl in the surrounding medium) by accumulating high concentrations of osmolytes in the cytoplasm; major osmolytes include K⁺-ion, glutamate, and trehalose ^52,53. These high concentrations of osmolytes can have significant effects on protein-DNA interactions ⁵⁴. We suggest that the altered cytoplasmic conditions under osmotic stress may stabilize PTCs but that PTC accumulation is prevented by RapA activity in wt Eco. In the absence of RapA, PTCs accumulate, leading to increased levels of PTC-mediated initiation and production of R-loops to a lethal level.

Discussion

Mechanisms underlying pervasive transcription initiation are of growing interest as genome-wide sequencing techniques have allowed deeper coverage of in vivo transcription to be achieved ⁵⁵. The role played by PTCs in the production of this promoter-independent transcription has been difficult to structurally probe due to challenging biochemical reconstitution constraints. PTCs exhibit one-dimensional diffusion on the DNA ^4-6 (meaning they could slide off the ends of a linear DNA fragment) and core RNAP binds tightly to the ends of linear DNA fragments ²⁸. These features preclude the use of linear DNA fragments typically used in cryo-EM analyses of protein-DNA complexes. By reconstituting PTCs on a negatively-supercoiled circular duplex DNA substrate, our results provide insight into the structural nature of PTCs and promoter-independent transcription initiation more broadly, and the role of RapA in preventing the excessive build-up of PTCs, leading to a revised model for the bacterial transcription cycle (Fig. 5) ⁴.

Fig. 5 ∣. Model for the Role of RapA in the Bacterial Transcription Cycle.

Results from this work (highlighted in red-shaded boxes) yield a model for the role of RapA in the bacterial transcription cycle: i) In promoter-dependent transcription (light blue background), free core RNAP (green but with dark red clamp) combines with σ⁷⁰ (blue) to form Eσ⁷⁰, which locates promoter DNA sequences and initiates promoter-specific transcription; ii) Following intrinsic termination, core RNAP can remain on the DNA in a PTC, equilibrating between PTCo PTCi PTCc (Post-termination complexes, light green background); iii) PTCo can initiate transcription independent of σ but is prone to cytotoxic R-loop formation (promoter-independent transcription, light orange background); iv) The toxic formation of R-loops is suppressed by RapA (orange), which specifically binds PTCs and pulls open the RNAP clamp in an ATP-dependent manner. This allows the transcription-bubble to reanneal and remove the core RNAP from DNA, facilitating RNAP recycling and averting the harmful build-up of R-loops (RapA-mediated RNAP recycling, light yellow background). Created in BioRender. Campbell, E. (2024) BioRender.com/f68z555.

σ-dependent transcription begins with base-flipping and capture of conserved nt-strand bases (normally at position −11 and −7 with respect to the transcription start site at +1) ⁵⁶ within the −10 core promoter element, triggering transcription bubble nucleation and bubble propagation downstream to the transcription start site ^57,58. Our rPTC structures show that core RNAP can bind and bend the DNA to nucleate a transcription bubble on negatively supercoiled DNA independent of σ factor or conserved promoter elements (Figs. 1c-f). Core RNAP appears to nucleate its transcription bubble from within the RNAP active-site cleft and to propagate in the upstream direction while engaging SW2, FL2, and the β'rudder in the reverse order from σ-dependent transcription initiation (Extended Data Fig. 3).

Evolved strategies to suppress promoter-independent transcription appear manifold. Termination factor Rho targets ECs producing aberrant transcripts ⁵⁹. Additional mechanisms for preventing non-specific association of RNAP with DNA outside of promoter sequences include auto-inhibitory σ⁷⁰_1.1 ⁶⁰, DNA packaging proteins (such as H-NS) ⁶¹, and RapA ⁷. Our approach for generating rPTCs suitable for cryo-EM analysis allowed us to visualize RapA engaged with its true target substrate (Fig. 2a). Our RapA^†-PTC structure, along with comparisons to previously determined apo-RapA-RNAP and apo-RapA-EC structures, illustrate how nucleotide binding drives a complex allosteric conformational rearrangement of RapA that pulls open the RNAP clamp, allowing melted regions of the DNA to rewind and promoting dissociation of the DNA. Further work will be required to determine the role of ATP hydrolysis (as distinct from nucleotide binding) in the RapA functional cycle.

Dey et al. ⁶² found that a population of Eco RNAP ECs paused at the U-rich pause site following the HK022 putL element released their RNA transcript but remained associated with the DNA template - essentially PTCs. The RNAP conformation in these PTCs (8AC2) ⁶² closely matches the rPTCc RNAP conformation, and both of these are very similar to the PTC portion of our RapA^†-PTC structure (Supplementary Data Table 2). Thus, RapA uses ATP binding energy to stabilize a pre-existing PTC state. Dey et al. ⁶² did not observe closed-clamp/open-bubble structures corresponding to rPTCi or rPTCo, presumably because their complexes were formed on linear DNA fragments that lacked supercoiling.

Our in vitro and in vivo results point to the formation of cytotoxic R-loops via σ⁷⁰-independent transcription mediated by PTCs (Figs. 3d and 4). Our in vivo experiments show that overexpression of RNase HI rescues growth defects of a rapA null strain (ΔrapA) at two distinct growth conditions, R-loop stress induced by sub-MIC BCM (Figs. 4a and 4b, Extended Data Fig. 8) and osmotic stress induced with 1 M NaCl (Fig. 4c). As observed previously ¹², the ΔrapA strain appeared to be completely unable to grow on LB agar + 1 M NaCl (Fig. 4c), suggesting that rapA may be essential under this condition (conditional essentiality). Strikingly, overexpressing RNase HI allowed growth at this otherwise lethal condition for the ΔrapA strain. This result strongly suggests that under osmotic stress, the loss of RapA function leads to a lethal accumulation of R-loops, presumably generated by uncontrolled PTC-mediated initiation.

Our observation that σ⁷⁰-independent transcription initiation is prone to R-loop formation points to a crucial role for σ⁷⁰ in initiating proper RNA transcript strand separation from the RNA-DNA hybrid during initial transcription at promoters. The σ⁷⁰-family σ factors ¹ contain a conserved structural element, the σ-finger (also called the σ⁷⁰_3.2-loop) ^63,64, that loops into the RNAP cleft and helps pre-organize the t-strand DNA near the active site but also blocks the path of the elongating nascent RNA ^63,64. As the RNA chain extends during initiation, a steric clash with the σ-finger either promotes abortive initiation or the σ-finger is displaced, facilitating promoter escape ^63-66. We hypothesize that the absence of the σ⁷⁰-finger in the RNAP cleft results in unsuccessful strand separation of the RNA-DNA hybrid during PTC transcription initiation. We stress, however, that the detailed molecular mechanism for R-loop generation by PTC-mediated non-specific transcription initiation is immaterial to our main conclusions that: i) PTCs can initiate promoter-independent transcription and are prone to R-loop formation, and ii) PTC-mediated promoter-independent initiation and R-loop formation are both suppressed by disruption of PTCs by RapA.

Microbes are ubiquitous across a vast assortment of environments and consequently have developed survival strategies for frequent osmotic shock ⁵². Eco has effectively solved the biophysical challenge of surviving in aqueous environments ranging from highly dilute solutions to those containing molar concentrations of salts ⁵³. Here, our results highlight the lethal threat posed by cytotoxic R-loops and how the activity of RapA in evicting PTCs permits cell survival at the high osmotic strength of 1 M NaCl (Fig. 4c).

All-in-all, our findings implicate transcription initiation by PTCs, either before or after the association of σ, in promoter-independent transcription and the generation of cytotoxic R-loops in bacteria. Additionally, our structure of the RapA^†-rPTC delineates the mechanisms underlying the suppression of promoter-independent transcription by RapA. We reveal a previously unappreciated role for RapA in vivo; contributing to the suppression of deleterious R-loops. RapA is spread widely among bacterial lineages, but not universally so. We hypothesize that bacterial lineages that lack RapA may harbor analagous ATPases that fulfill similar roles as Eco RapA ^67-70. The behavior of bacterial RNAPs other than Eco RNAP after intrinsic termination is unknown; further studies will be required to understand how the potential for promoter-independent transcription and R-loop accumulation is suppressed in bacteria that lack RapA.

Methods

Structural biology software was accessed through the SBGrid consortium ⁷⁵. No statistical methods were used to predetermine sample size. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment.

Protein expression, purification, reconstitution, and and labeling.

Eco core RNAP and σ⁷⁰ were separately overexpressed and purified as previously described ⁵⁷. Briefly, a pET-based plasmid overexpressing each subunit of Eco RNAP (full-length α, β, ω) as well as β'-PPX-His10 (PPX; PreScission protease site, LEVLFQGP, Cytiva) was co-transformed with a pACYCDuet-1 plasmid containing Eco rpoZ (encoding ω) into Eco BL21(DE3) (Novagen). Protein expression was induced with 1 mM isopropyl ß-D-thiogalactopyranoside (IPTG) for 4 hr at 30°C. Cells were harvested and lysed with a French Press (Avestin) at 4°C. Lysate was precipitated using polyethyleneimine [PEI, 10% (w/v), pH 8.0, Acros Organics]. Pellets were washed and RNAP was eluted. The PEI elutions were precipitated with ammonium sulfate. Pellets were harvested, resuspended and loaded on to HiTrap IMAC HP columns (Cytiva) for purification by nickel affinity chromatography. Bound RNAP was washed on column, eluted and dialyzed. Dialyzed RNAP was loaded onto a Biorex-70 column (Bio-Rad) for purification by ion exchange chromatography. Eluted RNAP was concentrated by centrifugal filtration, then loaded onto a HiLoad 26/600 Superdex 200 column (Cytiva) for purification by size exclusion chromatography. Purified RNAP was supplemented with glycerol to 20% (v/v), flash frozen in liquid N₂, and stored at −80°C.

Plasmid encoding Eco His₁₀-SUMO-σ⁷⁰ was transformed into Eco BL21(DE3) (Novagen). Protein expression was induced with 1 mM IPTG for 1 hr at 30°C. Cells were harvested and lysed with a French Press (Avestin) at 4°C. Lysate was loaded onto a HiTrap IMAC HP column (Cytiva) for purification by nickel affinity chromatography. Eluted σ⁷⁰ was cleaved with ULPI SUMO protease (Thermo Fisher Scientific) to remove the His₁₀-SUMO-tag from σ⁷⁰, followed by dialysis. Cleaved sample was further purified on a HiTrap IMAC HP column (Cytiva). Tagless σ⁷⁰ was collected in the flowthrough and concentrated by centrifugal filtration. The sample was then loaded onto a HiLoad 16/60 Superdex 200 for purification by size exclusion chromatography. Purified σ⁷⁰ was supplemented with glycerol to a final concentration of 20% (v/v), flash-frozen in liquid N₂, and stored at −80°

Eco core RNAP containing a SNAP-tag on the C-terminus of β' ⁷⁶ was fluorescently labeled with SNAP-Surface DY-549 dye (New England Biolabs), yielding WT-RNAP⁵⁴⁹ as described ⁴. Briefly, RNAP-SNAP was dialyzed into labeling buffer (10 mM Tris-HCl, pH 8.0, 40 mM KCl, 5 mM MgCl₂, 20 μM ZnCl₂, 1 mM DTT) at 4 °C for 4 h. The resulting product was then mixed with equimolar SNAP-Surface DY-549 (1 mM in DMSO) at room temperature for 30 min. The product was mixed with an equal volume of labelling buffer supplemented with 60% glycerol (v/v) to yield RNAP⁵⁴⁹ in reconstitution buffer [10 mM Tris-HCl, pH 8.0, 30% glycerol (v/v), 0.1 mM EDTA, 100 mM NaCl, 20 mM KCl, 20 μM ZnCl₂, 3 mM MgCl₂, 0.6 mM DTT) then flash frozen and stored at −80 °C. A SNAP-tagged β' zinc binding domain deletion mutant (ΔZBD-RNAP⁵⁴⁹; β' residues 64-94 replaced with a -GS- linker) was similarly purified and labeled. Native mass spectrometry (nMS) analysis of unlabeled SNAP-RNAP-SNAP samples showed that the WT-RNAP-SNAP was mostly assembled core (80%); by contrast, only 9% of ΔZBD-RNAP-SNAP was fully assembled with 85% of the mutant core RNAP lacking the ω subunit. To avoid potential issues due to the low abundance of ω in ΔZBD-RNAP-SNAP, WT-RNAP-SNAP and ΔZBD-RNAP-SNAP were labeled with DY-549 dye as above and then incubated with three-fold and four-fold excess ω subunit (respectively) as follows: For WT-RNAP⁵⁴⁹, 12 μM WT-RNAP⁵⁴⁹ was mixed with 36 μM ω subunit and incubated on ice for 30 min. For ΔZBD-RNAP⁵⁴⁹, 1.83 μM ΔZBD-RNAP⁵⁴⁹ was mixed with 7.68 μM ω subunit and incubated on ice for 30 min. The samples were then buffer-exchanged into 10 mM Tris-HCl, pH 8, 100 mM NaCl, 2.5 mM MgCl₂, 1 mM DTT, 25% glycerol (v/v) using Centri-Spin10 (5 kDa) spin columns (Princeton Separations) and stored at −80°C. Subsequent nMS analysis revealed that both WT-RNAP⁵⁴⁹ and ΔZBD-RNAP⁵⁴⁹ core complexes were fully assembled and completely labelled with DY-549.

Eco RapA was overexpressed and purified as previously described ⁷. Briefly, Eco RapA protein was overexpressed in Eco BL21(DE3) cells transformed with pQE80L (Qiagen) expression vector (encoding N-terminally His₆-tagged full-length RapA) ²² and grown in LB medium with ampicillin (100 μg/mL) at 37 °C. Expression of RapA was induced with 1 mM IPTG and shaken at 37 °C and harvested after 3 h. RapA protein was purified by affinity and size-exclusion chromatography using prepacked 5 mL Ni-affinity (HisTrap HP), 5 mL heparin (HiTrap Heparin), and Superdex 200 columns (Cytiva) and stored in storage buffer (10 mM HEPES, pH 7.5, 50 mM NaCl, 0.1 mM EDTA, pH 8.0, 5 mM DTT) at −80 °C. The fluorescently-labelled SNAP-RapA construct, RapA⁶⁵⁰, was prepared as previously described ⁷. Plasmid pKI1 (Addgene #199118) encoding His₆-SNAP-RapA was transformed into NEBExpressIQ cells (New England Biolabs), and the His₆-SNAP-RapA protein was expressed and purified as above. To make RapA⁶⁵⁰, His₆-SNAP-RapA and JFX-650 fluorophore SNAP substrate (kind gift of Luke Lavis, Janelia Farm Research Campus) were mixed at a 1:2 molar ratio with 1 mM DTT in SNAP reaction buffer (50 mM Tris–HCl, pH 7.5, 100 mM NaCl, 0.05% Tween 20). Excess dye was removed using an Ultra-0.5 mL 30K spin column (Amicon). Purified RapA⁶⁵⁰ was frozen in liquid N₂ and stored at −80 °C in 10 mM Tris–HCl, pH 7.9, 50% glycerol (v/v), 0.1 mM EDTA, 0.1 M NaCl, 1 mM DTT. The functionality of RapA⁶⁵⁰ compared to His₆-RapA was verified through single-molecule washout experiments ⁷.

Native MS analysis.

The RNAP samples were buffer-exchanged into nMS solution (500 mM ammonium acetate, pH 7.5, 0.01% Tween-20) using Zeba microspin desalting columns (Thermo Scientific) with a 40-kDa MWCO ⁷⁷. For nMS analysis, 2–3 μL of the buffer-exchanged sample was loaded into a gold-coated quartz emitter that was prepared in-house and then electrosprayed into an Exactive Plus EMR instrument (Thermo Fisher Scientific) with a static nanospray source ⁷⁸. The nMS parameters used included: spray voltage, 1.2 kV; capillary temperature, 125 – 150 °C; in-source dissociation, 10 V; S-lens RF level, 200; resolving power, 8,750 or 17,500 at m/z of 200; AGC target, 1 x 10⁶; maximum injection time, 200 ms; number of microscans, 5; injection flatapole, 8 V; interflatapole, 4 V; bent flatapole, 4 V; high energy collision dissociation (HCD), 200 V; ultrahigh vacuum pressure, 5.5–6.5 × 10⁻¹⁰ mbar; total number of scans, at least 100. Mass calibration in positive EMR mode was performed using cesium iodide. The acquired MS spectra were visualized using Thermo Xcalibur Qual Browser (v. 4.2.47) and processed further using the deconvolution software UniDec version 4.2.0 ^79,80 to obtain the deconvolved masses. The resulting measured masses for the unlabeled SNAP-RNAP assemblies observed included WT core: 412,045 Da, WT core–ω: 402,035 Da, α₂β: 223,691 Da, β'-ΔZBD core: 408,544 Da and β'-ΔZBD core–ω: 398,427 Da. The measured masses for the DY549-labeled, SNAP-RNAP assemblies incubated with excess ω subunit were WT-RNAP⁵⁴⁹: 413,105 Da and β'-ΔZBD-RNAP⁵⁴⁹: 409,687 Da, which closely matched the mass of the corresponding RNAP core with one covalently attached DY-549 dye. The mass accuracies, calculated as the percent mass deviation between the measured and predicted masses, ranged from 0.03% - 0.07%.

DNA templates.

The initial discovery and characterization of PTCs by single-molecule analysis utilized linear DNA templates ^4,5, but subsequent experiments characterizing PTCs utilized promoterless circular DNA templates ⁸, including a small 586 base pair promoterless circular DNA template used in single-molecule studies of RapA function ⁷. The single-molecule studies herein (Fig. 2e) used this same 586 base pair circular DNA template. Direct comparison of the results from all of the experiments described herein (single-molecule, cryo-EM, in vitro transcription) using the same DNA template would be ideal, but the 586 base pair circular DNA template could not be generated in sufficient yield for the cryo-EM experiments, and we sought to use an in vitro transcription DNA template containing the well-characterized T7 A1 promoter and terminator to facilitate comparison of Eσ⁷⁰ and E transcription characteristics. The rationale for the various DNA templates and their construction is described below:

In vitro transcription reactions:

We used the 4.8 kb plasmid pAR1707 ⁷³ containing the strong T7 A1 promoter and terminator (Fig. 3a). This facilitated direct comparison of the in vitro transcription reactions using Eσ⁷⁰ or E; in contrast to E, Eσ⁷⁰ was expected to yield an easily identifiable 161 nucleotide transcript (Fig. 3c).

To test that the presence of the strong T7 A1 promoter (or other promoters) on pAR1707 did not influence our analysis of rPTCs, we generated a derivative of pAR1707 (pJB2) in which all of the annotated promoters (ref. 71 and https://www.ncbi.nlm.nih.gov/nuccore/J01749) were removed (with the exception of the weak pbla promoter required for ampicillin resistance; see Extended Data Figs. 7a and 7b).

Cryo-EM:

We reasoned that preparing cryo-EM grids for single-particle data collection would be facilitated by a smaller circular DNA template. Since we could not generate the 586 base pair circular DNA template (used in the single-molecule experiments) in sufficient yield for the cryo-EM experiments, we constructed pJB1 (1.9 kb plasmid, instead of 4.9 kb pAR1707; Fig. 1a) using Gibson assembly on amplicons generated from plasmids pAR1707 and pUC57. The pAR1707 insert amplicon contains a T7A1 promoter sequence and downstream 21-mer stall sequence derived from pAR1707 (for use outside of the scope of this work). The pUC57 backbone amplicon included an Ampicillin resistance gene and a high copy-number origin of replication.

Preparation of Eco rPTC complex for cryo-EM.

Eco core RNAP (0.5 mL of 5 mg/mL protein) was injected into a 10/300 Superose 6 Increase column (Cytiva) equilibrated with 10 mM Tris-HCl, pH 8.0, 100 mM KCl, 5 mM MgCl₂ and 2.5 mM DTT. The peak fractions of the eluted protein were concentrated by centrifugal filtration (EMD-Millipore - 30 kDa MWCO) to 25 μM protein concentration.

Plasmid pJB1 was grown overnight in DH5α cells in standard Luria broth with 100 μg/ml ampicillin and isolated using a Plasmid DNA Maxiprep kit (Qiagen). Plasmid DNA solution at 1750 ng/μL was added to core RNAP for a final concentration of 491 ng/μL of DNA (0.82 μM). The sample was incubated for 15 min at 37°C, then 3-([3-cholamidopropyl]dimethylammonio)-2-hydroxy-1-propanesulfonate (CHAPSO; Anatrace) was added to a final concentration of 8 mM ⁸¹ and the sample was kept at room temperature prior to grid preparation.

Preparation of RapA-rPTC complex for cryo-EM.

The rPTC complexes were prepared as described above but with the plasmid final concentration of 0.42 μM. RapA (20 μM) was pre-incubated with AlF₃ and ADP (2.5 mM each) (Sigma-Aldrich). The ADP-AlF₃-RapA solution was added to the rPTC sample to achieve a final concentration of 8 μM RapA, 1 mM AlF₃, and 1 mM ADP. CHAPSO was then added (8 mM final concentration) and the sample was kept at room temperature prior to grid preparation.

Cryo-EM grid preparation.

C-flat holey carbon grids (CF-1.2/1.3-4Au; Protochips) were glow-discharged for 20 s before the application of 3.5 μL of the sample (0.42 μM Plasmid DNA, 8 μM core RNAP, 8 μM RapA, 1 mM ADP, 1 mM AlF₃, 8 mM CHAPSO). After blotting for 3–4.5 s the grids were plunge-frozen in liquid ethane using an FEI Vitrobot Mark IV (FEI) with 100% chamber humidity at 37°C.

Cryo-EM data acquisition and processing.

Eco rPTCs.

Grids were imaged using a 300-keV Titan Krios (FEI) equipped with a K3 Summit direct electron detector (Gatan). Images were recorded with Leginon ⁸² in counting mode with a pixel size of 1.076 Å and a defocus range of −0.25 to −4.16 μm. Data were collected with a dose rate of 28 e⁻ per Ų per s. Images were recorded over a 2 s exposure with 0.05 s frames (40 total frames) to give a total dose of 55.9 e⁻/Ų. Dose-fractionated videos were gain-normalized, drift-corrected, summed, and dose-weighted using MotionCor2 ⁸³. The contrast transfer function (CTF) was estimated for each summed image using the Patch CTF module in cryoSPARC3 (CS3) ⁸⁴. Particles were picked and extracted from the dose-weighted images with a box size of 256 px using CS3 Blob Picker and Particle Extraction. Coordinates pointing to contaminating ice particles were extracted as faux particles and used to generate an initial decoy 3D model in CS3 (ab initio reconstruction) in order remove junk particles from initial particle stacks. Multiple rounds of CS3 Hetero Refinement of all blob-picked particles employing 3D templates from this 3D decoy along with an Eco core RNAP 3D template (PDB: 6ALH with all nucleic acids removed, low pass filtered to 20 Å resolution), were used to identify a 3D consensus reconstruction containing subclasses for rPTCo and rPTCi. Multiple rounds of CS3 Hetero Refinement of all blob-picked particles employing 3D templates of the 3d decoy along with Eco core RNAP 3D template (PDB: 6GH6 with all nucleic acids removed, low pass filtered to 20 Å resolution), were used to identify a 3D consensus reconstruction containing the subclass for rPTCc. Many classification schemes were tested that converged on the conclusion that three mid-to-high-resolution classes were present in the particle dataset. All three classes were subjected to two rounds of successive Bayesian Polishing in Relion3 ⁸⁵. CS3 CTF-refinement and non-uniform (NU) refinement were then performed for each resulting class, yielding three distinct structures: rPTC_c (13,101 p, 4.7 Å nominal resolution), rPTC_i (49,701 p, 3.8 Å nominal resolution), and rPTC_o (86,865 p, 3.6 Å nominal resolution) (Extended Data Figs. 1 and 2).

Eco rPTC + RapA.

Grids were initially screened using a 200 keV Talos Arctica (FEI) equipped with a K2 Summit direct electron detector. Datasets were recorded with a pixel size of 1.5 Å over a defocus range of −1.0 μm to −3.5 μm. Movies were recorded in counting mode at 8 electrons/physical pixel/second in dose-fractionation mode with subframes of 0.3 s over a 15 s exposure (50 frames) to give a total dose of 53.33 electrons/Ų. Dose-fractionated movies were gain-normalized, drift-corrected, summed, and dose-weighted using MotionCor2 ⁸³. The CTF was estimated for each summed image using the Patch CTF module in CS3 ⁸⁴. Particles were picked and extracted from the dose-weighted images with a box size of 256 px using CS3 Blob Picker and Particle Extraction. Particles were curated via CS3 2D classification and selection. CS3 ab initio reconstruction was used to produce a density map of rPTC + RapA. CS3 Non-uniform refinement was used to further refine this initial map to 6.16Å. Core RNAP subunits (no nucleic acids present) and all RapA domains, except for spacer domain, were rigid-body refined to fit the density of the map. The resulting incomplete molecular model was used to produce a simulated 20 Å resolution density map for downstream templating.

For our full data collection, grids were imaged using a 300-keV Titan Krios (FEI) equipped with a K3 Summit direct electron detector (Gatan). Images were recorded with Leginon ⁸² in counting mode with a pixel size of 1.076 Å and a defocus range of −0.8 to −2.5 μm. Data were collected with a dose rate of 26 e⁻ per Ų per s. Images were recorded over a 2-s exposure with 0.05-s frames (40 total frames) to give a total dose of 56 electrons per Ų. Dose-fractionated videos were gain-normalized, drift-corrected, summed, and dose-weighted using MotionCor2 ⁸³. The CTF was estimated for each summed image using the Patch CTF module in CS4 ⁸⁴. Particles were picked and extracted from the dose-weighted images with a box size of 256 px using CS3 Blob Picker and Particle Extraction. An initial decoy 3D model was generated in CS4 (ab initio reconstruction) as described above. Multiple rounds of CS4 Hetero Refinement of all blob-picked particles employing 3D templates of this 3D decoy along with the incomplete 20 Å ab initio map mentioned above (all nucleic acids and RapA spacer domain missing), were used to identify a 3D consensus reconstruction containing rPTC+RapA. This class was subjected to focused CS4 3D-classification, masking around RapA to identify a clear class for RapA^†-PTC. Then, two rounds of successive Bayesian Polishing were performed in Relion3 ⁸⁵. Then, CS4 CTF-refinement and NU-refinement were performed, yielding RapA^†-PTC from 100,010 p (3.6 Å nominal resolution) (Extended Data Figs. 5b and 5c).

The heatmap distributions of particle orientations and half-map FSCs were calculated using CS3. 3D Fourier shell correlation calculations were performed using 3DFSC ⁸⁶. Local-resolution calculations were performed using blocres and maps were locally filtered using blocfilt (Bsoft package) ⁷².

Model building and refinement.

rPTCs.

The initial model for the rPTCs was derived from PDB 8EG7 ³⁰ with all of the nucleic acids removed. The model was manually fit into the cryo-EM density maps using ChimeraX ⁸⁷ and rigid-body and real-space refined using PHENIX real-space-refine ^88,89. For real-space refinement, rigid-body refinement was followed by all-atom and B factor refinement with Ramachandran and secondary structure restraints. Models were inspected and modified using COOT ⁹⁰.

RapA^†-PTC.

The initial model for RapA^†-PTC included the rPTCc model (determined herein) combined with RapA from PDB 7M8E ²¹. Steps of model building and refinement followed the same steps for the rPTCs described above.

Single-molecule experiments.

Construction of the 586 bp circular promoter-less DNA templates (npDNA^Cy5) used in single-molecule experiments was performed as described ⁷. In brief, a PCR product (template for Golden Gate Assembly, see below) was amplified from pDT4 (Addgene #199120) using primers 5′-GAA GGT CTC CAG CCG TAC CAA CCA GCG GCT TAT C-3′ and 5´-CCG GGT CTC ACC ATA CCC GCT GTC TGA GAT TAC G-3′. The npDNA^Cy5 template was then made via BsaI Golden Gate Assembly (New England Biolabs, Golden Gate Assembly Mix) in T4 DNA Ligase Buffer using equimolar PCR product and a synthetic duplex oligonucleotide containing internal biotin and Cy5 dye modifications made by annealing the complementary DNAs 5′-CGA TTA GGT CTC GGG CTA GTA CTG GTT TCT AGA G/iCy5/GT TCC AAG CC/iBio/ TCA CGG CGG CCG CCC ATC GAG ACC GGT TAA CC-3′ and 5′-GGT TAA CCG GTC TCG ATG GGC GGC CGC CGT GAG GCT TGG AAC CTC TAG AAA CCA GTA CTA GCC CGA GAC CTA ATC G-3′ (IDT). Reactions were incubated for alternating cycles of 5 min at 37 °C and 10 min at 16 °C, followed by 5 mins at 55 °C and then 10 min at 65 °C to inactivate T4 DNA ligase. The product was digested with 10 units of T5 exonuclease (New England Biolabs) for 30 min at 37 °C, followed by inactivation with 15 mM EDTA, and further purified with Qiaquick PCR Purification Kit (Qiagen).

Single-molecule fluorescence experiments were set up as described ⁷. Briefly, we use a micromirror total internal reflection fluorescence instrument with a ~65-μm-diameter circular field of view. We used excitation wavelengths 532 nm to monitor RNAP⁵⁴⁹ and 633 nm to locate DNA^Cy5. The temperature of the reaction chamber was maintained at 33.1 ± 0.5 °C using a custom temperature-control system. Single-molecule observations were performed in glass flow chambers (volume ~20 μL) passivated with succinimidyl (NHS) polyethylene glycol (PEG) and NHS–PEG–biotin (Laysan Bio Inc.). Streptavidin (#21125; Life Technologies) was introduced at 220 nM in wash buffer [50 mM Tris–acetate, pH 8.0, 100 mM potassium acetate, 8 mM magnesium acetate, 27 mM ammonium acetate, and 0.1 mg/mL bovine serum albumin (BSA) (#126615 EMB Chemicals)], incubated 1 min, and washed out (all wash-out steps used two flushes each of four chamber volumes of wash buffer). Streptavidin-coated fluorescent beads (T-10711, Molecular Probes), used as markers for stage drift correction, were loaded in the chamber at a dilution of ~1:400,000, and excess beads were washed out. The chamber was then incubated with 25 pM npDNA^Cy5 in wash buffer for 5 min, and excess npDNA^Cy5 was flushed. Locations of surface-tethered npDNA^Cy5 molecules were recorded by acquiring five 1 s exposure images with 633 nm excitation at a power of 400 μW (all laser powers measured incident to the objective lens). Next, we introduced either 1.5 nM RNAP⁵⁴⁹ or 4 nM ΔZBD-RNAP⁵⁴⁹ into the chamber in transcription buffer [wash buffer supplemented with 3.5% w/v PEG 8,000 (#81268; Sigma-Aldrich), 1 mg/mL BSA, and an O2-scavenging system (4.5 mg/mL glucose, 40 units/mL glucose oxidase, 1,500 units/mL catalase, 1 mM DTT), incubated for 10 min to allow binding to npDNA^Cy5 and rPTC formation, and washed out excess unbound RNAP⁵⁴⁹. Then, solutions of wash buffer, 5 nM RapA⁶⁵⁰, or 5 nM RapA⁶⁵⁰ + 1 mM ATP were mixed in transcription buffer and loaded into the glass chamber. Image acquisition began within 10 s after loading the reagents (at t = 0), with excitation alternating between 532 nm and 633 nm (400 μW each) at 1 frame/s for 40 min. Lifetimes (τ) of RNAP⁵⁴⁹ and ΔZBD-RNAP⁵⁴⁹ on npDNA^Cy5 were calculated by measuring the intervals where RNAP⁵⁴⁹ or ΔZBD-RNAP⁵⁴⁹ was present on DNA ⁷.

Transcription Assays.

Eσ⁷⁰ was reconstituted by incubating core RNAP (0.5 μM final) with σ⁷⁰ (2.5 μM final) at 37°C for 15 min. Either core RNAP or Eσ⁷⁰ (as indicated) were incubated with ATP, CTP, GTP, and UTP (500 μM each; TriLink Biotechnologies) and pAR1707 ⁷³ (35.8 nM final; Fig. 3a) or pJB2 (41.1 nM final; Extended Data Fig. 7b) plasmids at 37°C for 15 min in transcription buffer (100 mM Tris-HCl, pH 8.0, 500 mM KCl, 100 mM MgCl₂, 1 mM EDTA, 10 mM DTT, and 50 μg/mL BSA). Next, two units of Turbo DNase I (Invitrogen) were added to the sample along with Turbo DNase I reaction buffer (Invitrogen, final concentration 1X) and incubated at 37°C for 15 min in order to fully digest all DNA substrate present in the reaction, effectively halting transcription. DNase I digestion was stopped by adding EDTA to 15 mM (final) and RNA was extracted with RNA Clean and Concentrator Kit (Zymo Research). Eluted RNA was combined with Qubit HS RNA Reagent and Qubit HS RNA Buffer and measured using a Qubit fluorometer using the HS RNA protocol after RNA standardization (Life Technologies). The size profile of the RNA sample was analyzed on an Agilent 2200 TapeStation using a High sensitivity RNA Screentape with RNA sample buffer (Agilent). The results were visualized using the Agilent Tapestation Software (Agilent). For RNase H sensitivity experiments, transcription was halted prior to the introduction of Turbo DNase I by the addition and 5 min incubation of 5 μM Rifampicin at 37°C. Subsequently, 2.5 units of RNase H (New England Biolabs) was added to the reaction and incubated at 37°C for 5 minutes.

BCM growth assays.

RapA KO strain (F-, ΔhepA769::kan, Δ(araD-araB)567, ΔlacZ4787(::rrnB-3), λ⁻, rph-1, Δ(rhaD-rhaB)568, hsdR514 – Keio collection, National BioResource Project entry JW0058) and WT parental strain (E. Coli BW25113 - National BioResource Project entry ME9062) were both separately transformed with plasmids pBAD18 (empty vector) ⁷⁴ or pBAD18rnhA (RNase HI) ⁵¹ and grown overnight on LB-agar plates with 100 μg/mL ampicillin. Colonies were picked and grown overnight in LB with 100 μg/mL ampicillin, then back diluted to an OD_{600 nm} of 0.01 in LB. L-arabinose was added to a final concentration of 0.05% (w/v) for induction. Culture solutions (50 μL) were dispensed into dark well tissue culture plates (Greiner Bio-One CELLSTAR, 384-well) along with BCM (when used, from a 2.5 g/L stock in DMSO; Santa Cruz Biotechnology, CAS 38129-37-2) using an HP D300e Digital Dispenser (Tecan). The minimum inhibitory concentration (MIC) for BCM was determined to be 37.5 mg/L.

NaCl growth assays.

RapA KO strain (F-, ΔhepA769::kan, Δ(araD-araB)567, ΔlacZ4787(::rrnB-3), λ⁻, rph-1, Δ(rhaD-rhaB)568, hsdR514 – Keio collection, National BioResource Project entry JW0058) and WT parental strain (E. Coli BW25113 - National BioResource Project entry ME9062) both separately transformed with plasmids pBAD18 (empty vector) and pBAD18rnhA and grown overnight on 100 ug/ml Ampicillin LB-agar plates. Colonies were picked and grown overnight in Luria broth with 100 ug/ml Ampicillin. Culture concentrations were standardized according to OD_{600 nm} measurements and were diluted in a 1/10 dilution series, using LB combined with a final concentration of 0.05% arabinose. 2 μl samples from the 1/10 dilution series of all four culture strains were plated onto separate quadrants of LB-agar plates composed of LB-agar, 0.9 M NaCl, and 0.05% arabinose. Plates were incubated at 37°C for 48 hours and then photographed.

Figures and statistical calculations.

Figures and statistical calculations were made using Graphpad Prism 10.3.1 (Figs. 3b, 4a, 4b and Extended Data Figs. 6c, 7c, 8), PyMOL 2.5.5 (Figs. 1f, 1g, 2b-d, and Extended Data Figs. 4, 5d, 6a, 6b) and ChimeraX 1.4 (Figs. 1c-e, 2a and Extended Data Figs. 2b, 2e, 2h, 3, 5c).

Extended Data

Extended Data Fig. 1 ∣. Cryo-EM processing pipeline for rPTC structures (rPTCo, rPTCi, rPTCc).

Cryo-EM processing pipeline for rPTC structures (rPTCo, rPTCi, rPTCc).

Extended Data Fig. 2 ∣. Cryo-EM of rPTCo, rPTCi, and rPTCc.

A, B, C. rPTCc class: Gold standard FSC calculations for cryo-EM density map (A), cryo-EM density map and cross section colored according to key (top of B), cryo-EM density map colored according to local resolution (bottom of B) ⁷¹, 3DFSC and sphericity of density map (C) ⁸⁴.

D, E, F. rPTCi class: Gold standard FSC calculations for cryo-EM density map (D), cryo-EM density map and cross section colored according to key (top of E), cryo-EM density map colored according to local resolution (bottom of E) ⁷¹, 3DFSC and sphericity of density map (F) ⁸⁴.

G, H, I. rPTCo class: Gold standard FSC calculations for cryo-EM density map (G), cryo-EM density map and cross section colored according to key (top of H), cryo-EM density map colored according to local resolution (bottom of H) ⁷¹, 3DFSC and sphericity of density map (I) ⁸⁴.

Extended Data Fig. 3 ∣. The role of RNAP structural elements in DNA melting.

A.-C. (left) View of the rPTC structures (same as Figs. 1c-e). The boxed regions are magnified on the right.

(right) Magnified views of boxed region; only the DNA, βFork-loop2 (FL2), β'rudder, and β'Switch2 (Sw2) are shown. DNA is shown as a backbone cartoon with a transparent molecular surface (t-strand, blue; nt-strand, orange). The protein elements are shown as backbone cartoons (β, cyan; β', pink).

A. rPTCc; the clamp is open 24°, resulting in a large separation between FL2 and Sw2.

B. rPTCi; the clamp closes, closing the gap between FL2 and Sw2, nucleating a ~5 nt bubble in the DNA. The β'rudder is completely disordered.

C. rPTCo; the bubble propagates in the upstream direction to ~7-8 nt, creating room for the β'rudder.

Extended Data Fig. 4 ∣. Swiveling in rPTCi and rPTCo.

RNAP structures are shown as a gray molecular surface but with the swivel modules ³⁸ shown as backbone cartoons with cylindrical helices [8EG7 ³⁸ (unswiveled reference EC), yellow; rPTCo, green; rPTCi, cyan]. Rotation (swiveling) of the rPTCo and rPTCi swivel modules with respect to 8EG7 is shown.

Extended Data Fig. 5 ∣. RapA^†-PTC.

a. Hypothesized mechanism for the disruption of the PTC by RapA - adapted from ⁷.

b. Cryo-EM processing pipeline for RapA^†-PTC class.

c. Gold standard FSC calculations for the cryo-EM density map (upper left), cryo-EM density map and cross section colored according to key (upper middle), cryo-EM density map colored according to local resolution (upper right) ⁷¹, 3DFSC and sphericity of density map (lower) ⁸⁴.

d. Range of RNAP clamp movement observed in the transition between RapA-PTC stand-in (PDB: 7M8E) and observed RapA^†-PTC. Clamp opens approximately 24° upon RapA-PTC conformation change driven by ATP binding.

Extended Data Fig. 6 ∣. Details of RapA^†-PTC structural rearrangements.

a. The overall rotation of RapA^† with respect to the RNAP (65°, see Fig. 2c) is accommodated by flexibility of the RNAP βflap-tip, which maintains its contacts with RapA but also rotates with respect to the rest of the RNAP.

b. The RNAP cleft opens as the RNAP clamp is pulled open by the RapA^† structural rearrangements (transition from left to middle panel; also see Fig. 2d). The RapA^† spacer domain (cyan to dark blue transition) wedges into the open RNAP cleft (transition from middle to right panel).

c. Results from single-molecule fluorescence microscopy. Plotted is the reciprocal of the effective dissociation rates [τ⁻¹ (s⁻¹), black scale on the left and black data points] and the reciprocal of the average RNAP dwell time on DNA [<τ>⁻¹ (s⁻¹), red scale on the right, red line representing mean values with error bars showing ±SEM) of surface-tethered rPTCs formed with RNAP or a ΔZBD-RNAP mutant, by RapA plus ATP or in controls lacking RapA or lacking ATP. Number of complexes from left to right: N=308, 184, 306, 122, 272, 130.

Extended Data Fig. 7 ∣. In vitro transcription reactions on pJB2.

a. Plasmid map showing relevant features of pAR1707 (including annotated promoters). The region deleted to generate pJB2 is highlighted in red.

b. Plasmid map showing relevant features of pJB2.

c. Histogram plot showing the total amount of RNA produced from transcription reactions. The bars denote the average of three to six independent measurements (individual data points shown). Data are presented as mean values ± SEM. Statistical significance of differences between samples was determined using an unpaired, two-tailed t-test.

d. Size distribution of transcripts resulting from two independent Eσ⁷⁰ transcription reactions using pJB2 as a template (light shades) compared with the original size distribution of transcripts resulting from Eσ⁷⁰ transcription from pAR1707 (hot pink, see Fig. 3c).

e. Size distribution of transcripts resulting from core RNAP (E) reactions using pJB2 as a template (light shades) compared with the original size distribution of transcripts resulting from E transcription from pAR1707 (see Fig. 3d).

Extended Data Fig. 8 ∣. Growth analysis of wt and ΔrapA Eco.

a-c. Growth curves (OD_{600 nm}). The solid lines plot the average of three replicates, the thin lines above and below show the 95% confidence limit.

(top panel) Semi-log plots of the growth curves during log-phase growth. The thick, transparent lines show the exponential fit used to calculate the doubling time.

(bottom panel) Linear scale showing the full growth curves. The thick arrows denote the t_1/2.

a. BCM=0 and in the absence of arabinose [uninduced pBAD18 (empty vector) or pBAD18rnhA (RNase H)].

b. BCM=0 but with arabinose induction (0.05% w/v) of pBAD18 (empty vector) or pBAD18rnhA (RNase H).

c. BCM=0.5X MIC (MIC = 37.5 mg/L) and with arabinose induction (0.05% w/v) of pBAD18 (empty vector) or pBAD18rnhA (RNase H).

d.-e. Histograms showing growth parameters [double times (d) and t_1/2 (e)) for wt and ΔrapA Eco cells carrying pBAD18 (empty vector) ⁷⁴ or pBAD18rnhA (RNase HI) ⁵¹, all without BCM (BCM=0). In d., combined is the average for all the measurements. Error bars denote standard error (N=3 independent growth curves for each condition). Statistical significance of differences between samples was determined using unpaired, two-tailed t-test.

d. Doubling times (N=3 independent growth curves for each condition). Data are presented as best fit values ± SEM (calculated from 95% confidence limits of best fit).

e. t_1/2 (N=3 independent growth curves for each condition). Data are presented as best fit values ± SEM (calculated from 95% confidence limits of best fit).

f.-g. Histograms showing growth parameters (doubling times and t_1/2;) for wt and ΔrapA Eco cells carrying pBAD18 (empty vector) ⁷⁴ or pBAD18rnhA (RNase HI) ⁵¹ without (BCM=0) or with 0.5X MIC BCM (BCM=0.5). Error bars denote standard error (N=3 independent growth curves for each condition). Statistical significance of differences between samples was determined using an unpaired, two-tailed t-tests.

f. Doubling times (N=3 independent growth curves for each condition). Data are presented as best fit values ± SEM (calculated from 95% confidence limits of best fit).

g. t_1/2 (N=3 independent growth curves for each condition). Data are presented as best fit values ± SEM (calculated from 95% confidence limits of best fit).

Data availability.

The cryo-EM density maps and atomic coordinates have been deposited in the EMDataBank (https://www.ebi.ac.uk/emdb) and Protein Data Bank (https://www.rcsb.org) as follows: rPTCc (EMD-40930, PDB 8T00), rPTCi (EMD-40931, PDB 8T02), rPTCo (EMD-40922, PDB 8SZW), RapA^†-PTC (EMD-40943, PDB 8T0L). The atomic models used for initial model building and analysis are available from the Protein Data Bank under the accession codes 6ALH, and 7M8E.