Development of Covalent Inhibitors for Bacterial Histidine Kinases

Abstract

Antivirulence therapies offer an alternative strategy against antimicrobial resistance. We synthesized 2-aminobenzothiazole sulfonyl fluoride analogs for covalent histidine kinase inhibition. In vitro and in cellulo assays also identified lead inhibitors for affinity-based probe development. However, alkyne incorporation reduced bioactivity, emphasizing the challenges of designing functional probes while preserving inhibitor potency.

Antibiotic-resistant infections have emerged as one of the most urgent threats to global health. Existing treatments target essential bacterial processes, exerting strong selective pressure that accelerates the evolution of antibiotic resistance (ABR) and diminishes drug efficacy. Emerging alternative treatment strategies involve targeting non-essential bacterial virulence and resistance mechanisms to reduce this pressure. Attenuation of virulence can render pathogens susceptible to clearance by the host immune system or by existing antibiotics.

Bacteria have multiple mechanisms to enhance their ability to cause infections and persist in a host. For example, Salmonella enterica serovar Typhimurium is a Gram-negative foodborne pathogen that expresses a myriad of virulence factors—including motility, nutrient acquisition, and envelope modifications—enabling adaptation to hostile host environments and the successful establishment of infection (Figure 1A). Expression of many key virulence factors is triggered when environmental signals activate signal transduction networks known as the two-component systems (TCSs). TCSs typically consist of a sensor histidine kinase (HK) and a cognate response regulator (RR) that often function as a transcription factor facilitating the regulation of phenotypic responses (Figure 1B).^1–6 These signaling proteins are ubiquitous in bacteria with Salmonella species encoding 20–30 TCSs. A study by Murret-Labarthe and co-workers found that deletion of 24 of 30 TCSs in S. enterica serovar Typhi had little effect on its viability but altered infectivity.⁷ Similar results in other pathogens highlight TCSs as promising candidates for antivirulence therapies.^8–11

Figure 1.

Virulence and resistance mechanisms expressed in Salmonella enterica, many of which are associated with TCSs. A. Examples of common virulence and resistance mechanisms. B. TCS proteins and signaling cascade. HKs (yellow and blue) are multimeric structures containing a periplasmic signaling domain (SD), a transmembrane (TM) spanning region connecting to the cytosolic kinase domain comprised of the dimerization and histidine phosphorylation domain (DHp) and a catalytic-ATP (CA) binding domain. HKs have a cognate response regulator (RR; mustard), which typically act as a transcription factor when activated by phosphoryl group transfer to a conserved aspartate residue from the HK. Light blue star represents and inhibitor.

Conservation of the catalytic and ATP-binding (CA) domain across HKs and bacterial species presents an opportunity for pan-inhibition that could yield broad-spectrum antivirulence strategies.¹² We have previously identified a series of CA-domain binding inhibitors featuring a 2-aminobenzothiazole scaffold (Figure 2A).¹³ One of these molecules is riluzole (herein named R-1), an FDA approved drug for amyotrophic lateral sclerosis.¹⁴ We profiled the activity of a small panel of related structures in S. enterica and found that R-1 re-sensitizes a polymyxin-resistant strain to polymyxins and results in the downregulation of genes linked to the PhoPQ TCS.¹⁵ Taken together, these data suggest that the compounds are putative HK inhibitors and may act on TCSs such as PhoPQ. Encouraged by these results, we sought to identify the cellular targets of our inhibitors in S. enterica through the generation of affinity-based probes for application in pull-down experiments.^{16, 17} Determining the molecular targets of these inhibitors will assist future inhibitor design to develop more potent and selective analogs.

Figure 2.

Investigation of the covalent inhibitor library. A. A small library of 2-aminobenzothiazole-based covalent inhibitors were generated and investigated for activity inhibition in HK853. B. Specific growth rates and summation of the area under the curve were derived from S. enterica WT and polymyxin-resistant growth experiments grown in the presence of inhibitors (250 μM) or DMSO (2.5% v/v) for 24 h. Growth rates are represented with standard deviation from biological and technical triplicates. AUC data is represented in a histogram format with standard deviations. Statistical significance was determined through a Brown-Forsythe one-way ANOVA comparing each inhibitor-treatment to the DMSO control. p-value

Our affinity-based probe design first required the installation of an electrophilic group to covalently tag the bound protein(s). The active site of HKs does not contain a conserved cysteine residue preventing the use of a haloacetamide or micheal accepter. Instead, we selected the aryl sulfonyl fluoride (SO2F) moiety (SI Figure 1A). Sulfonyl fluoride-bearing compounds offer favorable characteristics, including stability under physiological conditions, high modification yields and broad proteome coverage.^18–21 Importantly, SO2Fs can react with a diverse set of nucleophilic amino acid side chains, such as tyrosine, cystine, lysine, and histidine.²⁰

Guided by previous structure-activity relationship studies, we focused on the addition of this moiety at the six-position of the benzothiazole core, the only site where modification yielded potent inhibitors (Figure 2A).^{22, 23} We connected the aryl sulfonyl fluoride to the main scaffold through a secondary amine, amide or ester (1b – 2b, Figure 2A). Previous findings indicate that the orientation of the SO₂F group on the aryl ring influences protein labeling and reactivity.²⁰ To interrogate optimal placement, we sought to generate analogs with the SO₂F group in the ortho-, meta-, and para-positions relative to the linker (Figure 2A).²⁰ Interestingly, the ortho-analog (1a) could not be isolated as it underwent intramolecular cyclization, resulting in fluorine displacement by the proximal amine and formation of a sultam (see SI 1a characterization data).

To elucidate how these modifications affected target-engagement and enzyme activity, we tested compounds 1b, 1c, 2a, and 2b against the CA domain of a model HK (HK853) using an in vitro end point assay developed by our group to derive IC₅₀ values (at a single protein concentration and time point).^{13, 24, 25} Found in Thermotoga maritima, HK853 is a well-characterized, thermally stable HK. It serves as a representative model that maintains the conserved CA-domain found across diverse bacterial species. This truncated construct lacking its sensory and transmembrane domains, is constitutively active which eliminates the need for external stimuli—features that make it a suitable model to assess HK-inhibitor activity. HK853 activity was detected with an activity-based probe, BODIPY-FL-ATPγS (B-ATPγS).²⁵ Upon binding to the CA-domain, B-ATPγS is used by the protein for autophosphorylation, resulting in covalent labeling of the histidine with the fluorophore, BODIPY-FL. Inhibitors occupying the CA-domain prevent B-ATPγS binding.

Fluorophore labeling, as determined by integrated band densities following SDS-PAGE, was used to calculate IC₅₀ values, providing relative potency information for our covalent analogs and approximate comparisons to our previously reported non-covalent inhibitor R-1.^{13, 22, 23} Indeed, a dose-dependent decrease in fluorescence was observed upon incubation with the covalent inhibitors, confirming their ability to bind the CA-domain of HK853 and inhibit autophosphorylation activity with similar potency to R-1 (Figure 2A). Native gel-based assays corroborated that HK853 inhibition by this small library of compounds is not due to protein aggregation, prompting further investigation of their activities (SI Figure 2). With the goal of developing affinity-based probes to capture target proteins, we selected 1b and 1c as suitable scaffolds for generating derivatives P-1 and P-2 (Figure 2A). The binding activity of these derivatives was substantially affected by the addition of an alkyne (Figure 2A).

We next employed intact protein mass spectrometry to determine which residue(s) were labeled in HK853. Upon whole protein analysis with MALDI-TOF mass spectrometry, we found that 2a and P-1 both modified the protein (SI Figure 3). To determine the binding site of the covalent analogs, we performed peptide mapping (site-of-labeling experiments). HK853 was incubated with P-1 or 1c, which are the most similar to our potent non-covalent inhibitors.^22,23 Labeled protein was enzymatically digested and the peptides analyzed by mass spectrometry. Comparison of the chymotrypsin-derived peptides following incubation with 1c, P-1 or DMSO revealed species with unique masses upon treatment (Figure 3, SI Figure 4). The corresponding modified peptide was lower in abundance following P-1 treatment, which may reflect ionization differences but is also consistent with the reduced potency of this molecule in comparison to 1c (IC50). This peptide mapped to the ATP-lid region of HK853, which contains several nucleophilic residues (Figure 3, SI Figure 4). To determine the exact site of labeling, we investigated trypsin-derived peptides and found an overlapping but distinct peptide in the ATP-lid region that enabled us to confirm that modification occurred at Tyr429 based on sulfonyl reactivity trends (Figure 3, SI Figure 5).^{20, 21} Together, these results suggest that P-1 and 1c target the active site as expected.

Figure 3.

Full length representation of combined peptides identified in chymotrypsin (pink) and trypsin (blue) with Tyr429 highlighted (yellow) as the only sulfonyl fluoride-reactive residue shared between both peptides.

A defining feature of antivirulence agents is that they target non-essential processes, limiting their general toxicity. To assess the bactericidal effects of our inhibitors, we performed minimum inhibitory concentration (MIC) assays. Compounds were used at a maximum concentration of 170 μg/mL (1b and 1c) and 176 μg/mL (2a and 2b). Higher concentrations could not be evaluated due to limited solubility. Following overnight incubation, both wild-type (WT) and polymyxin-resistant (EG9492) S. enterica displayed visible growth at all tested concentrations (SI Table 1) indicating no bactericidal effect at the tested concentrations. We next evaluated our leads in a growth curve assay to look for more subtle evidence of cell toxicity (OD₆₀₀). Treatment with 2a and 2b yielded lower final OD₆₀₀ readings compared to the DMSO control (SI Figure 6). Comparison of specific growth rates (μ_max) between inhibitor- and DMSO-treated samples revealed no statistically significant differences in WT cells (Figure 2B). However, polymyxin-resistant cells (EG9492) treated with compound 2b exhibited a statistically significant decrease in μ_max relative to the control (Figure 2B). In addition, area under the curve (AUC) analysis revealed a significant reduction upon treatment with compounds 2a and 2b in both strains, consistent with the observed lower final OD₆₀₀ values (Figure 2B and SI Figure 7). Together these results suggest that 2a and 2b may be impairing essential processes, leading to reduced viability. In contrast, 1b and 1c had no measurable impact on growth in either strain, indicating that these analogs are unlikely to target essential processes.

A notable feature of S. enterica pathogenicity is its ability to detect and resist cationic antimicrobial peptides (CAMPs), which are key components of the mammalian innate immune system due to their broad-spectrum antibacterial activity.²⁶ Secreted by epithelial cells at the mucosa and skin surfaces or within phagocytic vacuoles, CAMPs bind to the negatively-charged lipopolysaccharides on the outer membrane of Gram-negative bacteria. This interaction is followed by hydrophobic disruption of the inner membrane, ultimately resulting in cell lysis.^26–28 This mechanism is also used by the clinically-important drug, colistin (polymyxin E).²⁹ Gram-negative bacteria confer resistance to these molecules through the addition of cationic groups to the outer membrane resulting in electrostatic repulsion of CAMPS and colistin.^30–32 This modification is regulated by the PhoPQ and PmrAB TCSs in S. enterica and other pathogenic Gram-negative bacteria.^{28, 31–36}

We previously found that R-1 could potentiate polymyxin resistance upon co-treatment of a resistant S. enterica strain (EG9492s), resulting in a ~8-fold change in the colistin MIC.¹⁵ Using a checkerboard assay, we determined that R-1 and colistin act synergistically (fractional inhibitory concentration (FIC) value of 0.28, FIC < 0.5 indicates synergy; SI Table 2). While we found that co-treatment of colistin with R-2, 1b, 1c and 2a (250 μM) resulted in decreased polymyxin E MIC values (2–32-fold decrease; Table 1), we could not determine FIC values as these compounds did not kill S. enterica at their solubility limit.

Table 1.

2-aminobenzothiazoles enhance polymyxin E activity in resistant cells.

Combination	Polymyxin E MIC (μg/mL)
DMSO	64
R-1	8
R-2	2
1b	16
1c	8
2a	32
2b	64

[a] Inhibitors present at a final concentration of 250 μM in 2.5% (v/v) DMSO.

[b] Results are from 3-independent experiments in strain EG9492.

Given that CAMP/polymyxin resistance is regulated by PhoPQ and PmrAB, we postulated that these compounds are indirectly or directly modulating the activity of these TCSs or their downstream effectors. To investigate this possibility, we employed probe P-1. Despite its relatively weak activity (IC₅₀ = 37.2 μM), it enabled fluorescent tagging and pull-down of purified HK853 (SI Figure 8). Pull-down experiments with P-1 in an E. coli strain expressing HK853 showed extremely low/no labeling upon gel-based analysis of the eluted fraction, even under conditions to induce overexpression of this protein (SI Figure 9). These unsuccessful attempts in a controlled system with elevated HK expression levels indicate that P-1 is highly unlikely to enable the identification of native HKs, which are expressed at low levels relative to the total bacterial proteome.^{37, 38} These findings highlight how structural modifications to generate a pull-down probe from an inhibitor can significantly influence compound activity.³⁹

Finally, given that several of our non-covalent inhibitors exhibit activity across multiple pathogens, we sought to determine whether the covalent analogues exhibited efficacy in another organism. Prior studies with a related 2-aminobenzothiazole, R-2 (SI Figure 1B), in Pseudomonas aeruginosa (PA) showed significant repression of motility, specifically swarming. Given that the pathogenesis of P. aeruginosa is closely linked to its motility—facilitating environmental spread and colonization—we tested our compounds in a soft-agar assay designed to mimic the lung environment. Covalent inhibitors 1b and 1c significantly inhibited motility in a dose-dependent manner, similar to R-2, by 90 and 98% respectively, relative to the untreated control (SI Figure 10). Analogues 2a and 2b reduced motility by ~42–31%.

In summary, we developed a series of 2-aminobenzothiazole-based covalent inhibitors and found that SO₂F substitution at the meta- and para-positions was optimal placement for activity and that they target the ATP-binding domain of HK853, and likely other HKs. Notably, analogs 1b and 1c did not impact S. enterica viability, suggesting that they target non-essential processes. Moreover, they effectively potentiated resistance in a polymyxin-resistant S. enterica strain—a process mediated by TCSs. This study provides new insights into HK reactivity, as to the best of our knowledge, no reports of covalent inhibition of bacterial histidine kinases have been described to date. Unfortunately, structural modifications, such as the addition of SO₂F and alkyne groups, led to physicochemical and bioactivity perturbations including poor solubility and altered bioactivity, limiting their application in downstream experiments. This outcome highlights the need for further optimization to preserve biological activity while improving compound properties.

Overall, our results suggest that these covalent analogs may interfere with TCS-mediated processes and enhance the efficacy of existing antibiotics. This raises several important questions: can covalent inhibition of HKs be leveraged for therapeutic development? Can these compounds also synergize with other antibiotics or innate immune defenses, offering a broader strategy to combat S. enterica infections? Ongoing target identification studies aim to clarify the molecular basis of these effects and further define their relationship with TCS pathways.