Exploration of Inhibitors of the Bacterial LexA Repressor-Protease

Abstract

Resistant and tolerant bacterial infections lead to billions in healthcare costs and cause hundreds of thousands of deaths each year. The bulk of current antibiotic research efforts focus on molecules which, although novel, are not immune from acquired resistance and seldomly affect tolerant populations. The bacterial SOS response has been implicated in several resistance and tolerance mechanisms, making it an attractive antibiotic target. Using small molecule inhibitors targeting a key step in the deployment of the SOS response, our approach focused on preventing the deployment of mechanisms such as biofilm formation, horizontal gene transfer, and error-prone DNA repair. Herein we report the synthesis and testing of analogs of a triazole-containing tricyclic inhibitor of LexA proteolysis, the key event in the SOS response. Our results hint that our inhibitor’s may function by adopting a b-hairpin conformation, reminiscent of the native cleavage loop of LexA.

Graphical Abstract

Roughly 36,000 Americans die each year due to antibiotic resistant infections, incurring billions in healthcare costs.¹ Previously treatable infections are no longer responsive to typical antibiotics, forcing clinicians to resort to more extreme “last-defense” drugs. Bacterial tolerance poses an additional challenge—these quiescent populations of cells can withstand most antibiotics and are implicated in chronic and recurring infections. This trajectory of resistant and tolerant infections threatens to undo most of the achievements of modern medicine, rendering even simple procedures life-threatening.

Several resistance and tolerance pathways have been connected to the SOS response (a bacterial stress response, Figure 1).^2,3 In unstressed bacteria, a repressor called LexA binds to the promoters of DNA damage response genes, maintaining the SOS response in the “off” state. The SOS response is turned “on” in the presence of DNA damage, which can be caused by UV exposure, reactive oxygen species, or antimicrobials. Damaged DNA stalls replication, exposing single-stranded DNA (ssDNA), which then recruits the DNA damage sensor RecA. ssDNA templates the ATP-dependent oligomerization of RecA forming an activated nucleoprotein filament known as RecA*. RecA*, in turn, binds to LexA to induce its self-cleavage. This reaction is presumed to occur due to the RecA*-dependent stabilization of a conformational change in LexA. This conformational change is defined by the formation of a β-turn that positions an internal scissile peptide bond between Ala84-Gly85 adjacent to its internal serine protease active site (Figure 2). LexA autoproteolysis severs the DNA binding domain from the catalytic and dimerization interface-containing protease domain.⁴ Ultimately, depletion of intact LexA leads to its dissociation from promoters, allowing unrepressed transcription of the SOS response.

Figure 1:

The SOS response

Figure 2:

Crystal structures highlighting the transition from a non-cleavable conformation (red) to a cleavable conformation (green) with the target scissile bond highlighted in yellow. The triazole scaffold may mimic the β-turn adopted by the cleavable conformation, occluding the target residues from the active site.

This SOS response in E. coli includes roughly 40 genes, including low-fidelity DNA polymerases that promote mutagenesis and genes that induce growth arrest, dissemination of mobile gene elements, and resistance determinants.^5–12 Given the central role of the response in pathways leading to antibiotic evasion, the SOS response makes for a compelling antimicrobial target. Indeed, interruption of the SOS response with a non-cleavable LexA mutant mitigates acquired resistance to fluoroquinolones.¹³ Furthermore, attenuation of the SOS response via recA inactivation not only reduces acquired resistance of Escherichia coli to several antibiotics, but also sensitizes the bacteria to some antibiotics, even resensitizing resistant bacteria to fluoroquinolones.^14–16

The SOS response has also been connected to biofilm formation.^17,18 Biofilms are matrices of cells enveloped in a thick extracellular polymeric substance (EPS) composed of polysaccharides, proteins, nucleic acids, and lipids. The EPS provides bacterial colonies structural stability, enables adhesion to surfaces, facilitates intercellular coordination, and protects resident cells from UV damage, metal toxicity, acid, and dehydration (among other conditions).^19–24 LexA represses transcription of fnbB, a fibronectin-binding protein important for adherence to extracellular matrix components and, as a result, biofilm formation.^25,26 Indeed, ΔrecA mutants display diminished stress-induced biofilm formation (4x) whereas recA-complemented strains regain biofilm formation ability.¹⁷ These studies make the SOS response an attractive target for the abrogation of biofilm formation, among other adaptive responses.

The key step of the SOS response is the RecA-induced autoproteolysis of LexA. Thus far few inhibitors of RecA* activity or filamentation have been studied.^27–32 However, RecA is homologous to the fundamental eukaryotic Rad51 recombinase family, rendering it a problematic target. In comparison, LexA inhibitors have been understudied until recently.^32–34 We discovered one group of inhibitors—triazole-containing tricycles—in an academic-industry partnership in 2018.³³ Previous efforts to increase potency of this scaffold yielded an analog with an improved IC₅₀ (9 ± 1 μM) with activity against both E. coli and Pseudomonas aeruginosa LexA (1). The inhibitors also showed biological activity in a ciprofloxacin-induced mutagenesis assay, supporting the principle that SOS inhibition can suppress acquired antibiotic resistance. Although the exact binding site of the inhibitors is not known, triazole-containing compounds have been used before as β-turn mimetics, which led us to hypothesize that these compounds may be adopting β-turn-like conformations to compete with the cleavage loop within the active site of LexA (Figure 2).^35,36 In our prior SAR study, we found isolated changes to either position A or position B in the scaffold of the lead compound could improve potency. We speculated that if the central triazole moiety were indeed positioning A and B more proximate to each other, simultaneous changes in these two regions might be either synergistic or antagonistic.

With this motivation in mind, we planned to further expand our structure-activity relationship (SAR) study by examining the substitutions on the outer rings of the scaffold. Guided by the findings of our first SAR study, we devised an initial panel of analogs (Figure 3). We synthesized 1, the previous lead, as a control. Analogs 2-5 combined the improvements seen in past analogs by melding ring substitutions in position A with the improvement of using an unsubstituted phenyl ring at position B. We designed and synthesized additional analogs decorated with various heteroatoms and electron-withdrawing groups to probe the effect on inhibitory activity. We also postulated that, if effective, analog 11 could serve as a starting point for diazirine photoaffinity probe studies, available in 4 steps.³⁷ These analogs could be easily synthesized using a convergent route that would also enable access to a number of outer ring combinations (Figure 4). First, we combined substituted anilines with 2-bromoacetyl bromide to generate various amides. Then, using either a cyanide or azide nucleophile, we assembled the two fragments before combining them via a base-catalyzed click to form the full scaffold.

Figure 3:

Tricyclic inhibitors of LexA autoproteolysis. A) Small molecule panel inspired by previous lead 1. B) Analogs 2–5 combine the best results from the initial SAR study. C) Proposed expanded library of A and B rings.

Figure 4:

Synthesis of small molecule panel. For full synthesis details, refer to the Supplemental Information.

Initial attempts to perform the base-catalyzed click using past conditions (either sodium ethoxide in ethanol or cesium carbonate in DMSO/water) were unsuccessful.³⁴ Curiously, the only product we were able to isolate was an unexpected dihydropyrrolone (analog 12), confirmed via NMR, HRMS, and X-ray crystallography. Reaction optimization, namely swapping the base and solvent for potassium carbonate and acetonitrile, respectively, provided triazole products in 18–83% yield (full optimization details available in the supporting information).³⁸ We suspect that the change for a less polar solvent favors the desired cycloaddition product, whereas more polar systems might bias the reaction toward ionic transition states.

We began by synthesizing analogs varied at the A ring, the activity of which we planned to use to guide a panel of B ring analogs. These analogs were tested using our FlAsH-LexA assay (Figure 5), which utilizes a truncated LexA furnished with a tetracysteine CCPGCC motif bound to the biarsenical fluorophore, FlAsH-EDT₂.³³ Addition of RecA* triggers LexA proteolysis, releasing a short fluorophore-labelled peptide measured by a corresponding decrease in fluorescence polarization. The addition of LexA inhibitors prevents release of this peptide and maintains fluorescence polarization. Regrettably, our inhibitor analogs displayed no improvement over the previous lead, 1. Only analogs 4 and 10 displayed any appreciable activity. Surprisingly, analogs 2, 3, and 5 were inactive, although their component A and B rings displaying potent activity separately.³⁴

Figure 5:

The FlAsH-LexA assay utilizes a truncated version of LexA in which the DNA-binding N-terminal domain is replaced with a FlAsH-binding tetracysteine tag. Upon RecA*-mediated autoproteolysis, the FlAsH-LexA construct (left) releases a short fluorophore-labeled peptide (right). The rate of proteolysis may then be measured as by a decrease in fluorescence polarization.

Despite no further improvement on our starting compound, our new analogs did offer some added insights. First, in our previous SAR study, we demonstrated that replacement of ortho,meta-dimethyl-phenyl group B with an unsubstituted phenyl ring resulted in ~3-fold inhibitor improvement while replacing the para-ethoxy group with an ortho-methyl on group A resulted in ~2-fold improvement.³⁴ Meanwhile, in our tested compounds here, combining these two improvements resulted in loss of inhibition in our assay (Figure 6, analog 3). The observation that these changes antagonize one another, rather than show additive or synergistic effects, supports the possibility that groups A and B could feasibly interact with another in a manner consistent with β-turn mimicry, attributed to the triazole scaffold. In order to explore this hypothesis, we decided to use GOLD, as part of Cambridge Structural Databases “Discovery” software package, for molecular docking and pose prediction of our best lead compound, compound 1 (Figure S1 and Table S1).^42,43 The triazole ring is centered near the target scissile bond with half of the molecule tracking with the backbone of the cleavage loop residues 81–85 in each the top five best performing poses (scored with either GoldScore or ChemScore). However, rather than completing the full β-turn the other half of the compound instead continues deeper into the pocket. Surprisingly, either functional moiety, A or B, can be accommodated on either side within the pocket. This may indicate that simultaneous substitutions at both sites may disrupt the central positioning of the triazole, leading to the observed loss of inhibitory activity. Based on these observations, further rational expansion of the SAR would be imperative in deciphering which of these interactions are indispensable for inhibitor function. As further opportunity for investigation, the observation that analog 10, with alternative para-position functionalization, retains inhibitory activity offers possibilities for generation of reactive probes to further explore binding and mechanism of action. Ongoing work to explore the binding mode and site are essential to further develop this promising lead series for combating the acquisition of antibiotic resistance and is currently under investigation.

Figure 6:

LexA Inhibition measured with our FlAsH-LexA assay. IC₅₀ fits are shown for analogs 1, 4, and 10. Percent inhibition was calculated by normalizing the change in anisotropy relative to the baseline of “uncleaved LexA” condition (mean, (−) ATPγS) and the “maximally cleaved LexA” condition (mean, (+) ATPγS), followed by scaling the data from 0 to 100. This data was then plotted and fit using [Inhibitor] vs. Normalized Response (variable slope) in GraphPad Prism 9.3.1 for Windows. Assay validation can be found in Supplemental Figure S2 and detailed fit statistics can be found in Table S2.

Materials and Methods

Full details available in the Supporting Information.