Molecular Stripping Underpins Derepression of a Toxin-Antitoxin System

Abstract

Transcription factors control gene expression; amongst these transcription repressors must liberate the promoter for derepression to occur. Toxin-antitoxin (TA) modules are bacterial elements that autoregulate their transcription by binding the promoter in a T:A ratio dependent manner, known as conditional cooperativity. The molecular basis of how excess toxin triggers derepression remained elusive, largely because monitoring the rearrangement of promoter-repressor complexes, which underpin derepression, is challenging. Here, we dissect the autoregulation of the Salmonella enterica tacAT3 module. Using a combination of DNA binding and promoter activity assays, and structural characterization, we determine the essential TA and DNA elements required to control transcription and reconstitute a repression-to-derepression path. We demonstrate that excess toxin triggers molecular stripping of the repressor complex off the DNA through multiple allosteric changes causing DNA distortion and ultimately leading to derepression. Thus, our work provides important insight in the elucidation of the mechanisms behind conditional cooperativity.

Introduction

Transcription factors (TFs) can affect all steps of the transcription process. Bacterial repressors typically bind DNA sequences (known as operators) that overlap the target promoter, thereby preventing RNA polymerase from gaining access to the promoter (Rojo, 1999). Productive transcription in such systems relies on a derepression step where the repressor dissociates from the operator, often relying on allostery and structural rearrangements upon binding or releasing of inducer molecules (Zhang et al, 1987; Lewis et al, 1996). Monitoring the rearrangement of operator-repressor complexes during derepression is challenging owing to the substantial change in the affinity of the repressor towards the operator during this process (Matthews & Nichols, 1998; Reedstrom et al, 1997).

Type II toxin-antitoxin (TA) modules are small bicistronic operons encoding a toxin protein that usually inhibits an essential cellular process and an antitoxin protein that neutralizes the toxin through direct binding (Jurėnas et al, 2022). The antitoxin also binds DNA operators in the promoter region of the operon, allowing for tight autoregulation (Page & Peti, 2016). Typically, at a low toxin:antitoxin (T:A) ratio the toxin acts as a corepressor bridging two antitoxin dimers that bind two adjacent operators, creating avidity. At a high T:A ratio, the toxin acts as a derepressor in a phenomenon named conditional cooperativity (Afif et al, 2001; Magnuson & Yarmolinsky, 1998; Overgaard et al, 2008). The derepression of TA modules is thought to depend on the formation of a high T:A ratio complex with reduced or incompatible binding for the operator DNA (Boggild et al, 2012; Garcia-Pino et al, 2010; Overgaard et al, 2008). However, to date, there is no detailed understanding of the structural changes that underpin promoter derepression in TA modules.

Salmonella enterica serovar Typhimurium contains three paralogous tacAT TA modules encoding TacT1–3 toxins, which are members of the GCN5-related N-acetyltransferase (GNAT) protein family. When not neutralized by their cognate TacA1–3 antitoxins, these toxins acetylate the primary amine group of the glycine residue of glycyl-tRNA molecules (Rycroft et al, 2018; Cheverton et al, 2016; Bikmetov et al, 2022). The TacA1–3 antitoxins contain an N-terminal ribbon-helix-helix (RHH) DNA binding motif and a flexible C-terminal region responsible for specific toxin neutralization (Grabe et al, 2021). Homologous TA systems found in Escherichia coli (ataRT) or Klebsiella pneumoniae (kacAT) were shown to control their expression by binding to operator sequences in their promoters as heterohexameric complexes with a 2:4 T:A ratio (Jurėnas et al, 2019; Qian et al, 2019). In both cases increased T:A ratio leads to derepression of the operon in a typical conditional cooperativity scenario but besides formation of non-repressive complexes, no detailed molecular insights into how the excess toxin leads to derepression is available.

Here, we set out to investigate the mechanisms of derepression of the S. Typhimurium tacAT3 system. Using a combination of DNA binding experiments, promoter activity assays, and structural analysis we identify the critical elements required to control transcription, and synthesize a model for the entire repression-to-derepression cycle of the tacAT3 promoter. We find that at a high T:A ratio, excess toxin leads to molecular stripping of the repressor through the formation of an octameric TacA3-TacT3-DNA transition complex where allostery at several levels leads to DNA bending and kinking, and ultimately derepression of the operon. Our work provides in-depth mechanistic understanding on an operonic derepression event that is reliant on dynamic stoichiometry-responsive multi-protein repressor complexes, creating a paradigm for regulatory phenomena that are likely applicable to other promoters.

Results

Identification of the tacAT3 operator

Type II TAs are self-regulated elements where the antitoxin, sometimes facilitated by the toxin, binds its promoter region on palindromic operator sites. The tacAT3 promoter region contains two palindromic sites, OP1 and OP2, upstream of the transcriptional start site that may serve as operators (Fig. 1a). To explore if and how the tacAT3 TA operon is self-regulated we created transcriptional fusions between the tacAT3 promoter region encompassing the two potential operators (OP1 and OP2) and a gfp reporter gene. As a proxy for promoter activity, we measured population GFP intensity by flow cytometry in wild-type and tacAT3 deletion mutant Salmonella strains grown overnight in a rich laboratory medium. The tacAT3 promoter was substantially more active in the tacAT3 deletion mutant than in the wild-type bacteria, indicating that, like other type II TAs, TacAT3 auto-represses its expression (Fig. 1b). To determine which palindrome within the promoter sequence constitutes an operator site for TacAT3, we altered the inverted sequence within OP1 or OP2 of the transcriptional fusions and tested their role in the control of the tacAT3 operon expression. Alteration of the OP1, but not OP2, sequence abrogated the tacAT3-dependent repression of the promoter activity, suggesting that it is an operator bound by TacAT3 (Fig. 1a–b). We also recapitulated these observations when the tacA3 antitoxin and tacT3 toxin genes were encoded on the pBAD33 plasmid in E. coli BL21(DE3) that does not encode any tacAT genes (Extended Data Fig. 1a). We then performed an electrophoretic mobility shift assay (EMSA) and found that the purified TacA3-TacT3 complex binds the OP1 sequence, but not the OP1* sequence with disrupted inverted repeats (Fig. 1c). We also observed a shift revealing the binding of the TacA3-TacT3 complex to similar palindromes present in the promoter of homologs of tacAT3 in the related ataRT and kacAT (Extended Data Fig. 1b–c). The results of these experiments show that TacAT3 represses the expression of tacAT3 by binding the OP1 inverted repeat.

Figure 1. S. Typhimurium tacAT3 is transcriptionally controlled at the OP1 site.

a, tacAT3 promoter region with two putative operator sites (OP1 and OP2) containing inverted repeat sequences in bold. Promoter variants with altered operator regions, OP1* and OP2*, are shown, with nucleotide changes depicted in bold red and green, respectively. b, Promoter activity measured by quantification of GFP intensity where wild-type and mutated promoters are fused to gfp-lva in wild-type and ΔtacAT3 S. Typhimurium strains. A negative control “empty” stands for bacterial strains without reporter plasmid. Representative flow cytometry histograms (left) and quantification of GFP intensity (right) are shown. The GFP intensity was normalized to that obtained for the wild-type promoter in the ΔtacAT3 S. Typhimurium mutant. Statistical analysis (unpaired two-sided t tests; n.s. - not significant) with p values and average activity are shown for N= 3 biologically independent experiments. c, A representative EMSA (left) and quantification of DNA binding (right) by the purified TacA3-TacT3^Y143F complex are shown for wild-type and mutated operators (OP1 and OP1*; two technical repeats per combination). Average binding and ±SEM are shown for N= 3 biologically independent experiments.

TacT3 facilitates the operon repression

Type II toxins can facilitate operator binding by the antitoxin, often by avidity through bridging several molecules. We tested whether TacT3 facilitated repression of tacAT3 and enhanced the associated OP1 binding by TacA3 through transcriptional reporter assays and EMSA, respectively. We detected complete promoter repression and direct operator binding only when both TacA3 and TacT3 were present (Fig. 2a–b). To understand how TacT3 facilitates repression of tacAT3 expression by TacA3, we generated a model of the TacA3-TacT3-OP1 complex using SWISS-MODEL and energy minimization performed with Gromacs (Fig. 2c). The model showed that TacA3 antitoxin dimers occupy the major grooves of the operator. Each antitoxin contributes a single N-terminal β1 strand with three polar amino acids (R11, S13, and R15) facing the DNA side and two hydrophobic residues (L12, and L14) facing the dimer interior, an architecture typical for RHH DNA binding proteins (Fig. 2c–d). To test the validity of this model, we assessed the contribution of these amino acids to the repression of the promoter where TacA3 is expressed alone or with TacT3. While in the tacA3-only condition every substitution affected the promoter repression activity, when both tacA3 and tacT3 were expressed only the R11A and R15A substitutions reduced promoter repression (Fig. 2e). This is likely due to the fact that toxins stabilize antitoxins (LeRoux et al, 2020). In agreement with this, the purified TacA3^R15A-TacT3 complex did not bind the operator in the EMSA assay (Fig. 2F). These results support our model where positively charged R11 and R15 are the key amino acids of the TacA3-DNA interface and suggest that the toxin molecule can compensate for small perturbations in the antitoxin-mediated interaction.

Figure 2. Critical repression determinants of the TacA3-TacT3 module.

a, Promoter activity measured by quantification of GFP intensity where wild-type tacAT3 promoter is fused to gfp in ΔtacAT3 S. Typhimurium deletion cells expressing wild-type or mutants of tacA3, and tacT3^Y143A (asterisk) or additional mutants from pBAD33 plasmid. The GFP intensity was normalized to that obtained for cells expressing both wild-type tacA3 and tacT3. Statistical analysis (unpaired two-sided t tests), and average activity are shown for N= 3 biologically independent experiments. b, A representative EMSA (left) and quantification of OP1 binding (right) by different combinations of wild-type and point mutant, purified TacA3 and TacT3^Y143F. Average binding, ±SEM, and statistical analysis (unpaired two-sided t tests) with p values are shown for N= 3 biologically independent experiments. For remaining EMSA see Figure S2A. c, Two orientations of a model of the structure of a hexameric TacA3-TacT3-DNA complex with a cartoon representation of TacA3 (yellow and orange) and TacT3 (blue), and DNA operator (light violet). d, Cartoon representation of the structure model of the TacA3-DNA interface in the major (left) and minor (right) grooves regions (light violet) with the main residues contributing to the interface shown as sticks. e, Promoter activity measured by quantification of GFP intensity where wild-type tacAT3 promoter is fused to gfp in E. coli BL21(DE3) with tacA3 antitoxin variants expressed from a pCOLA-Duet1 vector alone (top) or in combination with tacT3^Y143A toxin (bottom). The GFP intensity was normalized to that obtained for cells expressing wild-type tacA3 (top) or wild-type tacA3 and tacT3^Y143A (bottom). Statistical analysis (unpaired two-sided t tests) with p values and average activity are shown for N= 4 biologically independent experiments. f, A representative EMSA (top) and quantification of OP1 binding (bottom) by purified TacA3-TacT3^Y143F and TacA3^R15A-TacT3^Y143F complexes. Average binding and error bars (±SEM) are shown for N= 3 biologically independent experiments. A, B, and E, the asterisk indicates the use of a TacT3^Y143A catalytic mutant. g, Schematic illustration of the hexamer TacA3-TacT3-DNA model with TacA3 antitoxins (A3 and A3’) colored in yellow and orange, TacT3 toxins colored in blue (T3 and T3’), and the operator DNA colored in light violet. Primary (P), secondary (S), and minor groove (M) sites are shown. The ‘n.s.’ stands for not significant.

In this model, the operator is bound by two TacA3 antitoxin dimers connected via two peripherally localized TacT3 monomers. Each toxin is bound by two antitoxin molecules of a different dimer, forming either a primary (P) or a secondary (S) interface with respective surfaces of 1,590 and 599 Ų (Fig. 2c, g). As reported for many other TAs, TacT3 toxin molecules seem to create avidity of the antitoxin for the operator binding sites by bridging two TacA3 dimers. However, upon closer inspection of the TacA3-TacT3-OP1 model, we noticed a hydrogen bond between the TacA3 N34 and TacT3 R32 residues in the vicinity of the minor groove that forms an additional small TacA3-TacT3-DNA interface (Fig. 2c–d). To determine the significance of this interaction for tacAT3 repression and DNA binding we tested TacA3 N34A and TacT3 R32A/E substitutions individually, or in combination. While single substitutions had little effect, their combination resulted in almost complete loss of promoter repression and DNA binding (Fig. 2a–b and Extended Data Fig. 2a). A similar interaction site placing several positively charged or polar amino acids of the toxin in the vicinity of the operator minor groove is present in the KacAT-DNA structure (Extended Data Fig. 2b–c). This suggests that in addition to providing avidity to the antitoxin for the operator, the toxin itself engages with the operator, which likely further enhances repression.

Critical determinants of repression specificity

Of the three paralogous tacAT modules in S. Typhimurium, tacAT1 and tacAT3 have most similar operator sequences (Fig. 3a). We determined the crystal structure of the TacA1 and TacA3 antitoxin dimers, which adopt an identical RHH fold (Fig. 3b; Main Table 1). Despite the high similarity between their β1 strand sequences (Fig. 3c), we did not detect any cross-binding of the OP1 sites, nor could we see major cross-repression of the tacAT3 promoter by TacA1-TacT1 (Fig. 3d–e). Since TacA3 R11 and R15 are the main contributors to promoter repression and TacA1 displays a corresponding Q6/R10 pair (Fig. 3c), we hypothesized that TacA1 Q6 is the main determinant of operator binding specificity. We purified a TacA1^Q6R variant and tested its DNA binding activity against tacAT1 and tacAT3 OP1 operators. Purified TacA1^Q6R-TacT1 retained the ability to bind the tacAT1 operator while gaining the ability to bind the tacAT3 OP1 sequence (Fig. 3d). In agreement, overexpressed tacA1^Q6R and tacT1 repressed the tacAT3 promoter (Fig. 3e). We then tested a chimeric tacAT1–3 sequence with tacAT3-like substitutions introduced within the tacAT1 operator (Fig. 3a). We observed that both TacA1-TacT1 and TacA3-TacT3 purified complexes bound moderately, whereas TacA1^Q6R-TacT1 bound strongly, again revealing a loss of specificity (Fig. 3d). These results identify critical determinants of binding specificity within both TacAT3 and the operator sequence, demonstrating that although the two paralogous TacAT1 and TacAT3 systems do not seem to cross-regulate, a single mutation in the antitoxin or minimal changes in the operator sequence are sufficient for crosstalk.

Figure 3. The specificity of repression is determined by the antitoxin β1 strand and sequence of the DNA major groove.

a, Operator sequences of tacAT3 (orange), tacAT1 (yellow), and chimeric tacAT1–3. Bold letters indicate palindromic sequences, asterisks denote nucleotide identity. Red dash frames point to the swap region generated in the tacAT1–3 operator sequence. b, Cartoon representation of the structures of the Ribbon-helix-helix (RHH) domains of the two TacA1 (left) and TacA3 (right) antitoxin dimers. TacA1^Q6 and TacA3^R11 are colored violet, whereas TacA1^R10 and TacA3^R15 in aquamarine. c, Sequence of the β1 strand of TacA3, TacA1, and a chimeric TacA1^Q6R antitoxins. d, A representative EMSA (top) and quantification of operator binding (bottom) of the tacAT3, tacAT1, and tacAT1–3 operator sequences listed in A by purified TacA3-TacT3^Y143F, TacA1-TacT1^Y140F, and TacA1^Q6R-TacT1^Y140F complexes. Average binding and ±SEM are shown for N= 3 biologically independent experiments. e, Activity of the tacAT3 promoter measured by quantification of GFP intensity where wild-type promoter is fused to gfp in E. coli BL21(DE3) with tacA3 and tacT3 variants expressed from pCOLA-Duet1. The GFP intensity was normalized to that obtained for cells expressing wild-type tacA3 and tacT3. Asterisks indicate the use of a TacT1^Y140A and TacT3^Y143A catalytic mutants. Statistical analysis (unpaired two-sided t test) with p values and average activity are shown for N= 4 biologically independent experiments.

Main Table 1.

Data collection and refinement statistics

PDB ID	7ZG5	7ZG6
Name	TacAT3-DNA	TacAl5–57
Data Collection
Space group	C 2 2 21	H 3
Cell dimension; a, b, c (Å)	79.6, 120.1, 172.0	69.0, 69.0, 69.5
Cell angles α, β, γ (°)	90.0, 90.0, 90.0	90.0, 90.0, 120.0
Rpim	0.048 (0.773)	0.044 (0.328)
l / σl	8.7 (1.2)	7.0 (0.8)
CC1/2	0.988 (0.407)	0.997 (0.897)
Completeness (%)	100 (100)	99.9 (98.6)
Redundancy	12.8 (12.8)	4.9 (5.0)
Unique reflections	57175	9077
Refinement
Resolution (Å)	43.39 (2.0)	34.50 (1.94)
No. reflections	55893	9077
Rwork/ Rfree	0.21/0.25	0.20/0.24
No. atoms	4689	886
Macromolecules	4426	839
Ligand/ion	134	1
Water	129	46
B-factors	65.7	57.0
Macromolecules	55.4	56.5
Ligand/ion	62.3	145.5
Water	59.2	64.5
RMS Deviations:
Bond lengths (Å)	0.009	0.009
Bond angles (°)	1.10	1.15
Assymetric unit contents (Chain IDs)	2x Toxin (A, B), 2x Antitoxin (C, D), 2x DNA (E, F)	2x Antitoxin (A, B)

An octameric complex is formed in derepression conditions

For many TAs, at a low T:A ratio the toxin facilitates the repression of the operon by the antitoxin, but excess toxin leads to derepression, a phenomenon called conditional cooperativity. In our transcriptional fusion experiments, we observed that low or moderate levels of expression of the tacAT3 operon repressed the tacAT3 promoter (Fig. 2a and 4a), whereas high levels led to derepression only in some bacteria of the population (Fig. 4a). As expected, expression of the DNA binding mutant tacA^R15AT3 did not lead to promoter repression. Because of the inherent higher stability of the toxin compared to that of the antitoxin, we reasoned the derepression observed at high levels of induction of tacAT3 expression may be due to an excess T:A ratio. Whereas we observed that TacAT3 cooperate in DNA binding (Fig. 2b), increasing TacT3:TacA3 ratios above 1:1 disrupted operator binding (Fig. 4b). The reduced DNA binding was accompanied by the formation of a higher molecular weight (HMW) TacA3-TacT3-DNA complex (Fig. 4b). We reasoned that the formation of the HMW species may facilitate the dissociation of the TacAT3 from DNA. To understand better the dissociation between the TacAT3 complex and the operator, we purified TacA3-TacT3 in complex with the operator and crystallized the complex in excess toxin. We obtained the structure of a heterooctameric TacA3-TacT3-DNA complex with a 4:4 T:A ratio (Fig. 4c; Main Table 1). In this complex, two TacT3 dimers bridge two TacA3 dimers (Fig. 4c–d). The two primary TacA3-TacT3 interfaces, P and P’ are almost identical (interface surfaces of 1,856 and 1,886 Ų, respectively), with differences contained in their β2 strand and α2 helix regions of the antitoxin (Extended Data Figure 3a–b).

Figure 4. Structure of an octameric TacA3-TacT3 complex on DNA.

a, Activity of the tacAT3 promoter by quantification of GFP intensity where wild-type promoter is fused to gfp in E. coli BL21(DE3) with tacAT3 or tacA^R15AT3 are expressed from pBAD33 under the control of an arabinose inducible promoter. (top) Columns are ordered by increasing inducer concentration (0, 0.001%, 0.01%, and 0.1% arabinose). Statistical analysis (unpaired two-sided t test), average activity, and ±SEM are shown for N= 3 biologically independent experiments. (bottom) Representative flow cytometry histograms are shown for 0.01 and 0.1% arabinose concentrations. b, A representative EMSA (left) and quantification of OP1 binding (right) by purified TacA3 and TacT3^Y143F proteins mixed at various molar ratios. DNA binding is quantified in the form of low (LMW; gray) or high (HMW; black) molecular weight band shifts. White, gray, and black triangles point to free DNA, LMW and HMW bound forms, respectively. Average binding, ±SEM and statistical analysis (unpaired two-sided t test for free DNA) with p values are shown for N= 3 biologically independent experiments. c, Cartoon representation of the structure of the TacA3-TacT3-DNA octamer complex in two orientations with toxins in blue and green, antitoxins in yellow and orange, and DNA in light violet. Arrows indicate toxin (top) and antitoxin (bottom) dimer regions. Coenzyme A molecules are shown as black sticks. d, Schematic illustration of the octameric TacA3-TacT3-DNA structure with TacA3 antitoxins (A3 and A3’) colored in yellow and orange, TacT3 toxins colored in blue and green (T3 and T3’). Primary (P and P’) toxin-antitoxin interfaces are shown.

The octamer formation is required for promoter derepression

Toxin dimerization is the main difference between the hexameric and octameric TacA3-TacT3 complexes on DNA (Fig. 4d). The toxin dimer there is substantially more elongated than toxin-only dimer that carries out the aminoacyl-tRNA acetylation (Rycroft et al, 2018) (Extended Data Fig. 4a). This elongated toxin dimer primarily engages the loop 4 element, which contributes 43% (~509 Ų) to the total interface surface, contrasting with only 8% (~88 Ų) in the condensed toxin-only dimer (Figure 5a–b and Extended Data Fig. 4a–b). Superimposition of the two dimeric states revealed a drastic repositioning of monomer 2 with a ~72° rotation and an average Cα distance shift of 22 Å (Extended Data Fig. 4c–d), with dynamic loop 4 in both monomers (Extended Data Fig. 4d). SEC-MALS analysis revealed that purified TacT3 elutes as two well-resolved peaks, each with a calculated molecular mass corresponding to a dimer (Extended Data Fig. 4e). We then simulated the transition from a condensed to an elongated dimer state using molecular dynamics. The presence of energetic minima for both dimer forms suggests that the TacT3 toxin is a dynamic protein capable of changing its conformation even in a toxin only setting (Movie S1). We found the two dimer states to be separated by a kinetic barrier of about 5 kcal/mol and the elongated dimer state to be less stable by approximately 3 kcal/mol than the condensed state (Extended Data Fig. 4g and Movie S1).

Figure 5. Formation of the TacA3-TacT3 octamer is required for derepression.

a, Cartoon representation of the structure of the elongated toxin dimer engaged in the TacA3-TacT3 octameric complex on DNA. Black sticks represent coenzyme A molecules. The dashed red box highlights the main region of the toxin dimerization (D) interface. Arrows point to TacT3 loop 4 element. b, Cartoon and stick representation of the structure of the toxin dimerization (D) interface with main contributing residues. Black dashed lines depict hydrogen bonds formed by R84. c, A representative EMSA (top) and quantification (bottom) of tacAT3 operator binding by purified TacA3 in combination with TacT3^Y143F (TacT3) or TacT3^{Y143F, I78A, L82A, R84E} mutant (TacT3^Loop) mixed at various molar ratios. DNA binding is quantified in the form of low (LMW; grey) or high (HMW; black) molecular weight band shifts. White, gray, and black triangles point to free DNA, LMW and HMW bound forms, respectively. Average binding, ±SEM, and statistical analysis (unpaired two-sided t test) with p values are shown for N= 3 biologically independent experiments. d, Activity of the tacAT3 promoter by quantification of GFP intensity (top) where wild-type promoter is fused to gfp in E. coli BL21(DE3) with tacAT3 or tacAT3^Loop variant are expressed from pBAD33 under the control of an arabinose inducible promoter. Representative flow cytometry histograms (middle) are shown for 0.01 and 0.1% arabinose concentrations. Dashed black line represents the GFP intensity threshold above which cells were classified as having high promoter activity. Quantification of subpopulations of cells displaying low (black) and high (grey) promoter activity (bottom). (top and bottom) Columns with individual data points are ordered by increasing inducer concentration (0, 0.001%, 0.01%, and 0.1% arabinose). Average geometric mean of GFP intensity (top), proportions of cells displaying low and high promoter activities (bottom), ±SEM, and statistical analysis (unpaired two-sided t test between samples with equal inducer concentration) with p values are shown for N= 3 biologically independent experiments.

To test if the transition from a hexameric to octameric complex on the DNA is responsible for the derepression of the operon, we sought to disrupt the toxin dimerization interface (D interface) that supports the elongated dimer form. The three main amino acids constituting the D interface are I78, L82, and R84 of loop 4 (Fig. 5b). We created a TacT3^Loop mutant carrying I78A, L82A, and R84E substitutions. SEC-MALS analysis confirmed that this toxin variant forms only one of the two dimer peaks (Extended Data Fig. 4e). Its dependence on loop 4 and lower calculated stability supported that the earlier eluting peak represents the elongated form that can exist outside of the octameric TacA3-TacT3-DNA complex. We then confirmed that the TacA3-TacT3^Loop purified complex forms a hexameric state, both alone, and with the operator (Extended Data Fig. 4f). We then tested OP1 binding by purified TacT3^Loop in combination with TacA3 in EMSA experiments. Whereas TacT3^Loop displayed similar cooperativity to TacT3 at low T:A ratio, there was no disruption of DNA binding nor formation of HMW at high T:A ratios (Fig. 5c). This strengthened our hypothesis that the octameric complex on DNA is a key species required for the dissociation of the Toxin-Antitoxin-DNA complex. We then proceeded to test how the TacA3-TacT3^Loop mutant influenced the activity of the tacAT3 promoter. Low and moderate levels of expression of the tacAT3^Loop variant led to repression of the promoter comparably to wild-type tacAT3. However, the repression activity of the tacAT3^Loop variant was significantly more potent than that of the wild-type complex at higher inducer concentrations and no derepression was observed (Fig. 5d). Overall, these results demonstrate that a hexamer-to-octamer switch at the operator site mediated by loop 4 of TacT3 is required for promoter derepression at high T:A ratios.

The octamer enables molecular stripping of the repressor

To understand how the switch from hexamer to octamer enables promoter derepression, we compared the conformations adopted by the TacA3 antitoxin in the two states. The main difference is a switch of the antitoxin C-terminal regions from a secondary S interface formed with the pre-existing toxin to a primary P’ interface formed with the incoming toxin molecule (Fig. 2g and 4d). Superimposition in the N-terminal region of the two antitoxin states revealed a break in the α2 helix (S) that transitions into a β2 strand and α3 helix (P’) (Extended Data Fig. 5a). Moreover, the previously disordered 17 amino acids long C-terminal region in the S form is structured and forms an α4 helix in the P’ interface (Extended Data Fig. 5a). While TacA3 molecules engaging in the P interface show no major change in their RHH region, there is a considerable rotation of the C-terminal β2-α3-α4 region with an average Cα shift of 15 Å (Extended Data Fig. 5b). Consequently, the interacting TacT3 molecule is also substantially repositioned (average Cα shift of 16 Å) (Fig. 6a and Extended Data Fig. 5b, Movie S2). This repositioning results in a contact loss between the R32 residue of TacT3 and N34 of TacA3 and pulls the positive region of the toxin away from the minor groove (Fig. 6a and 2c, Extended Data Fig. 5b, Movie S2).

Figure 6. Conformational rearrangements in TacA3-TacT3-DNA octamer.

a, Cartoon representation of the structure of two states of the single P interface in hexameric (top) and octameric (bottom) TacA3-TacT3-DNA complexes. TacA3 molecules are shown in yellow and orange, TacT3 in blue, and operator in light violet. The red dashed line shows the change in position of the TacT3 R32 residue. Black dashed lines depict hydrogen bonds. b, Top view of the structure of two TacA3 dimers extracted from hexameric (top) and octameric (bottom) complexes superimposed on the left dimer RHH domain (amino acids 10–52; TacA3 dimer 1). DNA is shown in light violet. c, Side view of the hexameric (top) and octameric (bottom) TacA3-TacT3-DNA complexes illustrating DNA bending with angle values shown in green. d, Two (bottom and top) views of cartoon putty representation of the KacAT-DNA and TacAT3-DNA complexes with the B-factor indicated in a red-white-blue gradient (values 30 to 180) obtained by PyMOL. e, Schematic illustration of the transition from hexameric to octameric TacAT3 complex on DNA, with TacA3 antitoxins (A3 and A3’) colored in yellow and orange, TacT3 toxins colored in blue and green (T3 and T3’). Primary (P and P’), secondary (S) toxin-antitoxin, toxin dimerization (D), and minor groove (M) sites are shown.

Antitoxin superimposition on dimer 1 revealed a drastic shift of TacA3 dimer 2 with an average Cα distance of 12 Å in the RHH region (Fig. 6b and Extended Data Fig. 5c, Movie S3). This rearrangement is accompanied by a change in the operator curvature from an initial 33° to 55° when transitioning from hexameric to octameric complex (Fig. 6c). Moreover, the operator-antitoxin region in the octameric complex is characterized by high B-factor thermal motion values, a feature contrasting with relatively low operator disorder in the structure of the related hexameric KacAT-DNA complex (Fig. 6d and Extended Data Fig. S5d). This suggested that the strength of the interaction between the octameric TacAT3 complex and the operator DNA is weaker than that of the hexameric complex and that this could be a feature promoting dissociation from the DNA (Fig. 6e).

To further investigate the underlying mechanism leading to dissociation of the octameric TacAT3 complex from DNA, we ran multiple molecular dynamics simulations. In each replica, the initially contiguous base pair stacking present in the octamer-operator complex became rapidly disrupted to form a kink, even when we enforced Watson-Crick base pairing. In all these restrained simulations, the kink formed in the central region of the operator at the 5’-TG-3’ sequence (Extended Data Fig. 6a). To test if this distortion of the operator facilitates derepression of the operon, we sought to decrease tensions by manipulating the spacing between the inverted repeats within the operator. We added one nucleotide (Fig. 7a) and performed EMSA experiments with the purified wild-type TacA3-TacT3 and octamer-incompetent TacA3-TacT3^Loop complexes. Size alteration of the spacer still allowed DNA binding by both TacA3-TacT3 and TacA3-TacT3^Loop complexes (Extended Data Fig. 6b). Notably, purified TacA3-TacT3 formed a HMW on the OP1+1 operator, a pattern not observed with a TacA3-TacT3^Loop mutant complex (Extended Data Fig. 6b). This suggests that a TacA3-TacT3 octameric state binds preferentially the OP1+1 DNA sequence. We then tested binding of the OP1+1 operator by mixing the separately purified TacA3 and TacT3 at increasing T:A ratios. Even though the DNA binding of OP1+1 was less efficient than that of OP1 at low T:A ratios, it did not decrease at high T:A ratios (Fig. 7b). To further investigate the relative inability of the hexameric complex to adjust to the increased spacer length, we ran multiple equilibrium simulations of the protein components alone, observing the internal dynamics of both heterooligomers. Then, by aligning one binding interface to its target DNA site and observing the conformational ensemble of the other, we could infer the role of protein elasticity in adapting to changes in spacer length. As shown in Extended Data Figure 6c, the dominant distortion mode of the hexamer resembles a hinge movement in which the β-sheet (formed by the TacA3 dimer) moves away from the DNA in a direction perpendicular to DNA axis, one that essentially fixes the relative distance between the DNA-binding interfaces. While no clear pattern emerges for the octamer, the combined increased flexibility of the protein along the DNA axis and the elasticity of the kinked DNA counterpart give more room to adjust to non-native spacer lengths. This indicates that the spacing between the inverted repeats of OP1 is optimized for conditional cooperativity, with the hexameric TacAT3 complex having an advantage over the octamer for binding and the octameric complex suboptimal binding facilitates complex dissociation off the DNA.

Figure 7. Operator size is optimal for conditional cooperativity.

a, Sequence of wild-type and altered OP1 tacAT3 tested. Bold letters indicate inverted repeat while the insertion is shown in red. b, A representative EMSA (left) and quantification of operator binding (right) of the wild-type and altered tacAT3 operators represented in A by purified TacA3 and TacT3^Y143F proteins mixed at various molar ratios. DNA binding is quantified in the form of low (LMW; grey) or high (HMW; black) molecular weight band shifts. Average binding and ±SEM are shown for N= 4 biologically independent experiments. Unpaired two-sided t test p values for free DNA and HMW shifts in corresponding T:A ratios are shown. c, Model of derepression of tacAT3. The hexameric TacA3-TacT3 complex establishes a strong repressed state on the operator DNA by binding major (TacA3 dimers) and minor (M sites) grooves. Upon stress-induced release of additional cytosolic toxin, TacT3 monomers or dimers (green and grey triangles) invade the hexameric complex through interaction with the accessible C-terminal parts of the antitoxin (orange) molecules engaged in an S interface with the toxin. This initiates a cascade of allosteric rearrangements: i) TacT3 dimerization D interface replaces the bridging S interface; ii) the toxin-antitoxin S interface transforms into a P’ interface; iii) the loss of the minor groove interactions (M site); iv) the antitoxin dimer repositioning; and v) DNA distortion. All these changes result in the eventual molecular stripping of the transcriptional repressor from the DNA operator site, leading to operon transcription.

Altogether, these findings lead us to propose a model where the transcriptional derepression of the tacAT3 TA system occurs via an increased ratio of T:A, leading to a reflected stoichiometric (hexameric to octameric) switch taking place on the DNA. Available disordered C-terminal regions of the TacA3 molecules engaged in an S interface with the toxin would initiate contact with incoming toxin molecules in monomeric or dimeric form. This switch results in multiple allosteric effects leading to i) the disruption of contacts between the toxin-antitoxin and the operator minor grooves and ii) the rearrangement of the antitoxin dimers causing DNA bending that ultimately drives molecular stripping of the repressor complex from the DNA (Fig. 7c; Movie S4–5).

Discussion

Transcriptional control is essential for the adequate response of living organisms to dynamically changing environmental conditions. In bacteria, tight control of the expression of operons modulating cell growth, such as TA modules, is key for survival (Gollan et al, 2019; Leroux & Laub, 2022; Page & Peti, 2016). Here, we demonstrate how the S. Typhimurium TacAT3 module has evolved a stoichiometry-driven molecular switch of its repressor to accommodate both transcription derepression and buffering of fluctuations in cytosolic T:A ratios.

The RHH fold adopted by TacA3 antitoxin is a common DNA binding frequently found in transcription repressors such as MetJ, NikR, and CopG (Schreiter & Drennan, 2007). When in dimeric form, the N-terminal β-strand forms an antiparallel β-sheet that resides in the major groove of the DNA and specifically recognizes the operator sequence. We found that the closely related S. Typhimurium paralogue, TacAT1, does not bind the tacAT3 OP1 operator but a single Q6R substitution in TacA1 suffices to initiate binding and repression crosstalk with the tacAT3 element (Fig. 3d–e). Interestingly, RHH proteins with the same β-strand polar residues contacting DNA often recognize various DNA sequences (e.g., NikR, CcdA and CopG) suggesting existence of additional sequence specificity determinants (Schreiter & Drennan, 2007). We show that transcriptional repression of tacAT3 occurs by binding of TacA3-TacT3 at the OP1 operator site; however, it is plausible that cross-regulation in other promoter regions (such as OP2 site) also takes place.Notably, GNAT TA operator sites are positioned differently at different promoters, and while the TacAT3 operator flanks the −35 element on either side, KacAT and AtaRT bind an inverted repeat that overlaps the transcription start site (Extended Data Fig. 1c). Bending of the −35 region is required for the efficient docking of RNA polymerase holoenzyme (Shin et al, 2021), and it might be that the TacAT3 system exploits the spatial flexibility of this region for the promotion of derepression. DNA bending induced by RHH repressors has been observed before in the case of CopG where a tetramer (two CopG dimers) bends the operator DNA ~60° by compressing the minor and major grooves of the DNA (Xavier Gomis-Rüth et al, 1998). Interestingly, in a study with copG operator sites that were artificially introduced, CopG-mediated DNA bending could either activate or repress the expression of noncognate promoters depending on the position of the bending center (Chen et al, 2006). It is also noteworthy that the use of bending and kinking to induce derepression allows for a degree of energetic fine-tuning as different DNA sequences are characterized by varying degrees of flexibility (Lavery et al, 2010). Free energy profiling for the bending of the operator sequence (Extended Data Fig. 6d) reveals that the 5’-TG-3’ central sequence favors kinking, thereby lowering the energetic cost of the large structural operator distortion induced by the hexamer-to-octamer transition. Indeed, the identified high propensity of the 5’-TG-3’ subsequence to accommodate the kink is consistent with its previously reported tendency to assume bent, high-roll or kinked geometries (Lankaš et al, 2006; Chen et al, 2001; McNamara et al, 2012; Nagaich et al, 1994), suggesting that this sequence feature (or, more generally, a 5’-YR-3’ pattern in the central site) enables the formation of a strongly distorted binding site for the octameric complex. Intriguingly, the related tacAT1 and kacAT operator sequences do not contain the centrally positioned YR motif, which suggests that derepression of these modules could rely on other mechanisms, such as the loss of auxiliary repression sites at the minor groove regions, for example.

In TA modules regulated through conditional cooperativity, the repression occurs at a low T:A ratio where the toxin increases the affinity of antitoxin to the operator through avidity by bridging it to two or more DNA binding sites (De Bruyn et al, 2021). We find that TacT3 cooperates with TacA3 in DNA binding and promotes repression beyond simple avidity with TacA3 N34 and TacT3 R32 engaging the operator minor grooves (M site) (Fig. 2a–d and Extended Data Fig. 2). At the M site the positively charged region of the α1 helix of the toxin likely strengthens the hexamer-DNA interaction. Thus, the positive surface of the toxin, which is involved in binding of the target amino acyl-tRNA during toxin activity, is remarkably repurposed for DNA binding during repression.

It has been observed previously that the derepression of a TA promoter relies on formation of an intermediate complex of higher T:A ratio that weakens the interaction between the repressor complex and DNA, which is part of the mechanism of conditional cooperativity (Overgaard et al, 2008). The increased T:A ratio in tacAT3 shifts the complex from a hexameric to a toxin saturated octameric state. Octamer formation causes a cascade of conformational rearrangements from the operator-bound hexamer state, leading to operator dissociation and promoter derepression. In our model and simulation of the hexamer-to-octamer shift, incoming toxins anchor to the hexamer via the disordered C-terminal tail of antitoxins (Fig. 7c, Movie S4–5). We envision two non-mutually exclusive scenarios regarding the state of the invading toxin. In one model, monomeric toxins are released either from cytosolic toxin-antitoxin complexes, or from spontaneous dissociation of TacT3 dimers. Indeed, in TacAT and homologous toxin-antitoxin complexes, the antitoxin splits the toxin dimer (Yashiro et al, 2019, 2021; Grabe et al, 2021; Qian et al, 2019). Gradual degradation of the antitoxin could thus release monomeric toxin molecules. Moreover, dimers with a dissociation free energy of ca. 8 kcal/mol as is the case for TacT3 (based on our free energy simulations) display dissociation kinetics in the milliseconds-to-minutes timescale, which could also constitute a source of monomeric toxin. These monomers would then initiate invasion of the operator bound heterohexamer on both flanks resulting in heterooctameric-operator complex (Movie S4). Alternatively, toxin dimers would directly invade the heterohexameric-DNA complex at one flank (Movie S5) docking one monomer on the repressor complex. The residual monomer would then associate with the other side of the DNA-bound complex forming an octamer. This anchoring initiates a sequence of spatial restructuring in form of (i) replacement of the bridging S interface with a new bridging toxin dimeric D interface; (ii) transformation of the secondary S interface into a fully structured primary P’ interface formed with an incoming toxin; (iii) rotation of the pre-existing P interface (TacA3-TacT3) and loss of contact with the minor groove; and (iv) repositioning of TacA3 antitoxin dimers leading to DNA bending and kinking; that ultimately results in derepression (Fig. 7c; Movie S4). A similar molecular stripping mechanism has been proposed for the NF-κB-DNA complex, where incoming IκB protein forms a transition complex initiating DNA dissociation (Potoyan et al, 2016).

Interestingly, the octameric TacA3-TacT3-OP structure reported here is the first complex where full-length antitoxins are bound to dimeric acetyltransferase toxins. In the related K. pneumoniae KacAT system, a higher order oligomeric state of the TA complex accompanied by DNA release was observed at high T:A ratio (Qian et al, 2019). Whether this system relies on octamer and elongated toxin dimer formation, like TacAT3, or on a higher stoichiometry state, remains to be determined. Surprisingly, in another ortholog of TacAT3, the E. coli AtaRT, operator binding at increasing T:A ratios resulted in hexamer-to-tetramer switch, suggesting that even closely related TA systems evolve unique transition states initiating derepression (Jurėnas et al, 2019).

It is undoubtedly rare to be able to observe the transition state from full repression to derepression and indeed most structures of reduced or non-binding states for other repressors such as LacI or TrpR have been obtained on DNA-free forms (Lewis et al, 1996; Zhang et al, 1987). These transition repressor-DNA complexes occur for other TA systems; however, the dramatically lower binding DNA affinities and overall instability have so far precluded visualization and structural characterization (Bendtsen et al, 2017; Garcia-Pino et al, 2010; Boggild et al, 2012). Our study fills in the gap in the cycle of conditional cooperativity reported for TA modules and provides evidence for an on-operator transition state required for derepression.

Methods

Expression and purification of recombinant proteins

Individual Salmonella TacT1, TacT3, TacA1, and TacA3 protein variants were expressed as N-terminal (7xHis) TEV (tobacco etch virus) -cleavable fusions from the pQlinkH vector (Scheich et al, 2007). Point mutations were introduced using overlap extension PCR or standard PCR with one of the primers containing desired mutation. E. coli BL21(DE3) PC2 strain was grown in lysogeny broth (LB) at 37 °C until an optical density at 600 nm (OD600) of 0.8 was reached. Protein expression was induced overnight at 18 °C by addition of 0.5 mM isopropyl β-d-1-thiogalactopyranoside (IPTG). Cells were lysed by sonication in the lysis buffer (50 mM Tris, pH 8.0, 500 mM NaCl, 1 mM phenylmethanesulfonyl fluoride). Cleared lysate was incubated with Ni–NTA resin (Thermo Fisher Scientific) for 2 h at 4 °C with agitation. The lysate/resin mixture was applied to a column and washed with wash buffer (50 mM Tris, pH 8.0, 500 mM NaCl, 20 mM imidazole). The protein was eluted in 10 ml of elution buffer (50 mM Tris, pH 8.0, 500 mM NaCl, 500 mM imidazole) and subjected to TEV cleavage and dialysis (50 mM Tris, pH 8.0, 500 mM NaCl). Complexes of TacT and TacA were expressed from the multiple co-expression pQlink vectors composed of 7xHis-tagged TEV-cleavable TacA antitoxins and a non-tagged TacT toxin variant.

Size exclusion

Purified proteins were subjected to size exclusion chromatography using AKTA pure chromatography system (Cytiva) and a Superdex 200 Increase 10/300 GL column with a flow rate of 0.5 mL/min previously equilibrated with a 20 mM Tris pH 8.0 running buffer containing variable NaCl concentration ranging from 250 to 500 mM.

Crystallization and structure determination

Purified TacA3-TacT3^Y143F complex was mixed with the DNA duplex at a 1:1.5 molar ratio with 10 mg/ml protein and incubated for 1 h at 4 °C. Purified TacA1 (1), and TacA3-TacT3-DNA (2) complexes (20–40 mg × ml⁻¹) were crystallized at 4 °C by sitting drop vapor diffusion against the following reservoir solutions: 1% Tryptone, 1 mM sodium azide, 50 mM HEPES pH 7.0, 12% PEG 3350 for (1), and 0.08 M NaCl, 0.02 M BaCl, 0.04 M sodium cacodylate pH 7.0, 40% MPD, 0.012 M spermine tetrahydrochloride for (2), respectively. Crystals were harvested and either submerged in a crystallization buffer containing 20% glycerol for 10 s in case of (1), or not (2), followed by an immediate freezing step by immersion in liquid nitrogen. Diffraction data were acquired at Diamond Light Source (Harwell Science and Innovation Campus) on beamline I03. High-resolution native data sets were autointegrated using xia2 dials software (Clabbers et al, 2018). Phases were obtained with PHENIX Phaser software using a molecular replacement function and a model scaffold of either TacA2 (PDB 7AK7) or TacT3 (PDB 7AK9) for (1) and (2), respectively. Models were then manually improved in accordance with the amino acid sequence using Coot (Emsley et al, 2010) and refined using PHENIX and/or Refmac (Winn et al, 2011), against the high-resolution native data to give final structures (Main Table 1).

SEC–MALS

Size-exclusion chromatography-multiangle light scattering (SEC–MALS) was performed using Sepax SRT SEC-300 column (Sepax Technologies, Inc.) previously equilibrated (overnight; room temperature; flow rate 0.5 ml × min⁻¹) with running buffer (20 mM HEPES, pH 7.5, 250 mM NaCl). Purified TacT3 and TacA3-TacT3 complexes previously dialyzed in running buffer were applied (0.1 ml) on to the column and analyzed using a liquid chromatography system (1260 Infinity II LC System, Agilent) and a Wyatt Dawn Heleos II Multi-Angle Light Scattering Detector, with in-line DLS detector and Optilab TrEX refractive index detector and ASTRA software (Wyatt Technologies, Inc.), assuming dn/dc (refractive index increment) of 0.185.

Electrophoretic mobility shift assay

Oligonucleotides used are listed in Supplementary Table S3. Electrophoretic mobility shift assays (EMSA) were performed on 6% TBE gels. Typically, separately purified antitoxin (20 pM) was mixed with toxin variants at different ratios (0, 20, 40, 80, or 160 pM), and a 2.5 pM HEX-labelled DNA duplex in a binding buffer (20 mM Tris, 150 mM NaCl, and 20% glycerol) followed by 1 h incubation step at 37 ˚C. When testing pre-formed TacA3-TacT3 complex, 10 or 20 pM protein was mixed with the DNA duplex. Following the protein-DNA incubation, 5x loading buffer (20 mM Tris pH 8.0, 150 mM NaCl, 60% glycerol) was added and samples were loaded onto a 6% TBE gel that was previously pre-run for 1 h at 100 V with TBE buffer (50 mM Tris, 50 mM boric acid, and 10 mM EDTA) at 4 °C. Samples were electrophoresed at 100 V for 2 h in the dark at 4 °C. Gels were imaged using ChemiDoc MP Imaging System (BIO-RAD). Observed DNA duplex shifts intensities were quantified using ImageJ software (Schneider et al, 2012).

Promoter activity assays in vitro

Salmonella enterica serovar Typhimurium or Escherichia coli BL21(DE3) strains containing pFPV25 plasmid borne tacAT3 reporter (PtacAT3_gfp [Figs. 1, S1, and 2A] or PtacAT3_gfpLVA [Figs. 2E, 3, 4, 5]) were grown overnight in LB, appropriate antibiotics, and various inducer concentrations in the case of vector combinations including arabinose promoter. Each sample was collected and stored in PBS prior analysis by flow cytometry on a BD LSR II. GFP reporter activity data were analyzed using FlowJo software.

Software

Amino acids contributing to interfaces and surface area were obtained using the PDBePISA software (Krissinel & Henrick, 2007). Structural figures were generated using PyMOL, VMD, or Chimera software. Alpha carbon distances were obtained in PyMOL. Morph movies were generated in PyMOL2 with Morphing plugin. Promoter elements −35, −10, and TSS were predicted using BPROM and Salcom software (Solovyev & Salamov, 2013; Kröger et al, 2013). Analysis of simulation data was performed with in-house Python scripts, and MDTraj (McGibbon et al, 2015) was used to read and process trajectories. Movies S1, S4, and S5 were produced with Molywood (Wieczór et al, 2020). Hexameric form of DNA bound TacA3-TacT3 was obtained in SWISS-MODEL (Waterhouse et al, 2018) using KacAT-DNA as reference (PDB 5ZGN) followed by energy minimization performed with Gromacs (Abraham et al, 2015) (1,000 step steepest descent with Amber99SB force field and parmbsc1 correction).

System preparation for molecular dynamics and energy calculation

In the octameric form, the missing central base pair (25 bp system) or two base pairs (26 bp system) were introduced through geometric manipulation in VMD (Humphrey et al, 1996), followed by energy minimization and short equilibration with a restraint on the DNA molecule. All sequence manipulations were handled using the mutate_bases utility of X3DNA (Lu & Olson, 2003).

Molecular dynamics

All molecular dynamics simulations were performed with the Gromacs (Abraham et al, 2015) suite patched with the Plumed (Tribello et al, 2014) plugin (versions 2020.5 and 2.8.0, respectively). The Amber99SB-ILDN (Lindorff-Larsen et al, 2010) force field with the parmbsc1 (Ivani et al, 2015) correction for DNA was used along with their corresponding standard water (TIP3P) and ion parameters. The systems were embedded in dodecahedron boxes with a 1.5 nm padding, and the concentration of potassium/chloride ions was adjusted to correspond to a physiological value of 0.15M while neutralizing the electric charge of the system. Standard simulation conditions were applied: a CSVR thermostat with a reference temperature of 300 K, a Berendsen barostat maintaining the pressure at 1 bar, PME with periodic boundary conditions, and a 2 fs timestep with SHAKE/LINCS constraints. In the case of equilibrium simulations, 4 replicas were launched for each setup, with 400 ns per replica for the oligomer-DNA systems and 1000 ns per replica for the protein-only systems.

Free energy calculations

Free energy calculations were performed using multiple-walker well-tempered metadynamics (Barducci et al, 2008) as implemented in Plumed. For DNA bending, the reaction coordinate was defined as the angle between the centers of mass of the (1) four leftmost, (2) five [25 bp] or four [26 bp] central, and (3) four rightmost base pairs. The definitions of “leftmost” and “rightmost” exclude the terminal base pair that is prone to fraying. 8 walkers were used, 4 of which were started from the straight and 4 from the bent conformation; at least 250 ns of trajectory length was produced per walker. For the transition between two dimer geometries, the reaction coordinate was defined separately for each geometry (native and elongated), and consisted of 1/3 native contacts, 1/3 non-specific contacts, and 1/3 distance between monomer CoMs, so that values of 0 and 1 unequivocally corresponded to well-separated monomers and the precise geometry of the bound state; 12 walkers were used per geometry, all seeded initially from different frames along the WESTPA trajectory (see below); at least 900 ns of trajectory length was produced per walker.

Enforcing conformational transitions

The transition between two dimer geometries shown in Movie S1 was generated with Gromacs using the WESTPA suite (Zwier et al, 2015) through “information biasing” in the weighted ensemble framework, i.e. through adaptive cloning or discontinuation of short simulation segments in order to populate the entire range of a desired collective variable without applying external forces. The collective variable used to drive the transition was the difference between number of residue contacts characteristic for each geometry, where a residue contact is considered characteristic if within top N inter-chain residue pairs with largest differences in distances between the two geometries. An identical variable was already used in our previous work (Jurasz et al, 2021); here N was taken to be 56 and 36 for the native and elongated dimer.

A separate set of simulations was prepared to produce Movie S4 and S5. Here, external forces were added with Plumed to generate six 10-ns and three 5-ns trajectories for Movie S4 and S5, respectively, recapitulating the major events in the derepression pathway, as inferred from experimental and simulation data. The GOdMD model (Sfriso et al, 2013) was used to explore complex conformational transitions in a physically meaningful way, and coarse-grained pathways guided fully atomistic systems through the use of collective path variables (Branduardi et al, 2007) as implemented in Plumed.

Extended Data

Extended data Figure 1. Transcriptional control and DNA binding by TacAT3.

a, Promoter activity measured by quantification of GFP intensity in E. coli BL21(DE3) cells carrying pBAD33 (Empty) or pBAD33-tacAT3 (+tacAT3); pFPV25 containing gfp-lva fused to wild-type (black) or mutated (OP1*, red; OP2*, green) tacAT3 promoter. The GFP intensity was normalized to that obtained for the wild-type promoter in the E. coli carrying pBAD33 empty. Statistical analysis (unpaired two-sided t test) with p values and average activity are shown for N= 3 biologically independent experiments. b, A representative EMSA (top left) and quantification of DNA binding (top right) by the purified TacA3-TacT3^Y143F complex are shown for wild-type S. Typhimurium and related operator sequences in K. pneumoniae (kacAT) and E. coli (ataRT) (two technical repeats per combination). Average binding and ±SEM are shown for N= 3 biologically independent experiments. (bottom) Identity in the OP1 inverted repeat sequences (black dashed boxes) of E. coli ataRT, K. pneumoniae kacAT, and wild-type S. Typhimurium. Bold asterisks point to nucleotide identity. Mutated OP1* sequence with mutation sites (red bold letter) is also shown. c, Operators in closely related TA systems localize to different promoter regions. Start codon of the antitoxin is represented in blue. Bold black letters and half arrows depict operator inverted repeat sequences. Red triangles depict TSS.

Extended data Figure 2. Toxin plasticity displayed near the operator minor groove.

a, A representative EMSA of DNA operator binding by different combinations of wild-type and mutant purified TacA3 and TacT3 (for quantification see Figure 2B). The asterisk indicates the use of a TacT3^Y143F catalytic mutant. b, Cartoon representation of the structure of the KacA-KacT-DNA interface in the minor groove of the operator (light violet), where KacA antitoxin (orange) stabilizes the charged region of KacT toxin (blue) and orientates the positively charged residues of its α1 helix (blue sticks) towards the operator (PDB: 5ZGN). Black dashed lines indicate hydrogen bonds. c, Structures of KacT (left) and TacT3 (right) toxin-DNA interfaces. Polar and charged amino acids are shown as sticks. Toxins are each superimposed with their respective toxin-only states (grey). Superimpositions reveal rearrangement of the α1 helix in the toxin (from grey to blue) when in the toxin-antitoxin-operator complex. Representations are based on structures from PDB: 5XUN, 5ZGN (left) and PDB: 6G96 and TacAT3 hexamer-DNA model (right).

Extended data Figure 3. Comparison of P and P’ interfaces.

a, Cartoon representation of the structure of TacT3 molecules with their respective TacA3 antitoxins bound in a P (left) or P’ (right) interface. b, Cartoon representation of the structure of TacA3 α3 helix (top left), α4 helix (top right), and β2 strand (bottom) regions. Superimposed toxins forming the P and P’ interface are overlayed in top, and split in bottom respective panels. The main amino acids contributing to the interface are shown as sticks. Black sticks indicate coenzyme A. Black dashed lines depict hydrogen bonds.

Extended data Figure 4. Two types of TacT3 dimers.

a, Two side views (top and bottom) of TacT3 condensed (left; PDB 6G96) and elongated (right; PDB 7ZG5) dimers superimposed on blue monomer 1. Black dashed line represents the distance between the centers of mass between each TacT3 monomer within respective dimer. Red dashed circle (top left) depicts putative tRNA binding site of the condensed dimer. b, Surface contribution [Ų] of each secondary structure element of TacT3 to the dimerization interface for the condensed (orange) and elongated (pink) dimers. c, Side view representation of the spatial transition of monomer 2 from a condensed to elongated dimer state. d, Cα distance [Å] of each amino acid in monomer 1 (blue line) and monomer 2 (green line) measured between the two TacT3 dimer states superimposed on blue monomer 1 as shown in A. Secondary structure elements relative to amino acid position are depicted. e, SEC-MALS chromatograms of TacT3^Y143F (top), TacT3^{Y143F, Loop} (bottom), purified proteins. (E and F) Molecular mass, relative UV₂₈₀, and relative Rayleigh Ratio measurements are shown in green, violet, and red, respectively. f, SEC-MALS chromatograms of TacA3-TacT3^{Y143F, Loop} (top; Theoretical hexameric TacA3-TacT3 MW= 79 kDa, measured MW= 80 kDa) and TacA3-TacT3^{Y143F, Loop}-DNA complex (bottom; Theoretical hexameric TacA3-TacT3-DNA MW =110 kDa, measured MW =104 kDa). g, Free energy profile along the conformation transition between the extended (ξ= −1) and native (ξ= 1) geometries. The value of 0 refers to the dissociated state, in which the CoMs of the TacT3 monomers are separated by more than 5.5 nm. The free energy difference between the two states is ca. 3 kcal/mol.

Extended data Figure 5. Conformational changes in octamer vs. hexamer TacAT3 complex.

a, Architecture of secondary structure elements (top) in hexameric and octameric TacA3 antitoxins forming either P (dark yellow) or S/P’ (orange) interfaces. Cartoon representation of the TacA3 antitoxin molecules superimposed on their RHH regions (bottom) from the hexameric (shaded) and octameric (non-shaded) TacAT3 complex forming either S/P’ (bottom left) or P (bottom right) interfaces. Gray dashed lines indicate disordered regions. b, Cα distances [Å] of amino acids involved in the P interfaces measured between the octameric and hexameric complexes relative to antitoxin dimer region and the DNA. The TacA3 dimer was superimposed at the RHH region (amino acids 10–52). The dashed grey line indicates the disordered TacT3 loop 4 region in the hexameric complex. The red number specifies the change in distance between TacT3 R32 from the hexamer to octamer complexes (See Fig. 6a and Movie S2). c, Hexamer-to-octamer TacA3 dimer repositioning. Cα distances [Å] of each amino acid of the TacA3 RHH domain superimposed on antitoxin dimer 1 (amino acids 10–52; dashed orange and yellow lines). Distances of TacA3 molecules constituting dimer 2 are shown as solid yellow and orange lines. (See Fig. 6b and Movie S3). Secondary structure elements are shown (top) for orientation. d, Relative B factor distribution of antitoxins forming the P and P’ interface in related GNAT TA modules. Two (P and P’) TacA3 molecules from TacAT3-DNA, KacA (P) from KacAT-DNA (PDB 5ZGN), and AtaR (P) from AtaRT-DNA (PDB 6GTS) are shown.

Extended data Figure 6. Operator kink and flexibility of hexameric and octameric TacAT3 complex.

a, C5-C5 distances between consecutive bases in equilibrium simulations of octamer with the 25- and 26-bp DNA molecules in which Watson-Crick base-pairing was enforced. For each setup 4 replicates are shown, 400 ns each. In all the systems kink formation consistently occurs at the 5’-TpG-3’ step. b, A representative EMSAs (left) and quantification (right) of two tacAT3 operator binding by purified TacA3-TacT3^Y143F and TacA3-TacT3^{Y143F, Loop} protein complexes. Statistical analysis (unpaired two-sided t test) with p values for change in free DNA fraction are shown for N= 3 biologically independent experiments. c, Conditional probability density of finding the second DNA-binding interface at a given relative position with respect to X-ray DNA structure in pure protein systems, after aligning the first DNA-binding interface with the experimental binding site. The distance is measured from the center of the DNA major groove of each base pair to the center of the β-sheet interface. The position along the DNA is calculated as a continuous argmin function of the distance vector and is offset so that the experimental value (red dot) be equal to 0. The distance shown corresponds to the minimum value of the vector, or minimal distance between the β-sheet and the major groove. Probability densities are calculated using kernel density estimation. d, Free energy profile for the bending of a naked 25-bp double-stranded DNA, with the native operator sequence capable of forming the kink (blue) and a mutated sequence devoid of a central 5’-YR-3’ motif. The dashed line represents the angle found in the crystal structure; an angle of π would correspond to a perfectly straight DNA helix.

Data availability

The atomic coordinates for TacA1 and TacA3-TacT3-DNA have been deposited in the Protein Data Bank (PDB) under accession codes 7ZG6 and 7ZG5, respectively. Source data containing all experimental repeats of FACS histograms and EMSA gels are provided with this paper.