Synthesis and Molecular Modeling Provide Insight Into a Pseudomonas aeruginosa Quorum Sensing Conundrum

Abstract

The triphenyl amide/ester 12 was originally reported to be a potent mimic of the natural 3-oxo-dodecanoyl homoserine lactone quorum sensing molecule in Pseudomonas aeruginosa. However, explicit synthesis/chemical characterization was lacking, and a later report providing protein crystallographic data inferred 12 to be incorrect with 9 now being the surmised structure. Because of these inconsistencies, and our interest in quorum sensing molecules utilized by Gram-negative bacteria, we found it necessary to synthesize 9 and 12 to test for agonistic activity in a P. aeruginosa reporter assay. Despite distinct regiochemical differences, both 9 and 12 were found to have comparable EC₅₀ values. To reconcile these unanticipated findings, modeling studies were conducted and both compounds were revealed to have comparable binding properties to the LasR receptor.

Quorum sensing (QS) is the process in which bacteria communicate intercellularly through chemical signals termed autoinducers.¹ QS, a population-density dependent phenomenon, allows bacterial populations to regulate a variety of physiological processes such as bioluminescence, antibiotic biosynthesis, biofilm differentiation, and production of virulence factors. Pseudomonas aeruginosa, an opportunistic pathogen, is found in many immunocompromised patients suffering from diseases such as cystic fibrosis, AIDS, burns, or neutropenic cancer.² P. aeruginosa, like many other Gram-negative bacteria, regulates QS through the exchange of acyl homoserine lactone-based autoinducers. One of the major autoinducers of P. aeruginosa, 3-oxo-dodecanoyl homoserine lactone (3OC12-HSL), along with its cognate receptor LasR, has been the subject of intense investigation.³ Accordingly, and due to the dire effects of P. aeruginosa to human health, a myriad of approaches to the modulation of QS have been examined,⁴ including the construction of structural analogs of 3OC12-HSL to act as QS agonists and antagonists.⁵

Recently, a library of 200,000 compounds was screened by Greenberg and coworkers in the hopes of discovering potent inhibitors or activators of the LasR-dependent QS pathway.⁶ Both activators and inhibitors of QS were uncovered and, excitingly, one compound exhibited more potent QS activation than the natural 3OC12-HSL signal. This compound, triphenyl 12 (termed TP-1P), was structurally unrelated to 3OC12-HSL with an activity greater than the natural ligand (EC₅₀ of 14 nM v. 140 nM).⁷ However, the chemical identity of this novel activator was never explicitly confirmed and was subsequently proven incorrect in an X-ray crystallography study of LasR by Zou and Nair. According to the electron density maps of TP-1 bound to LasR, it was proposed by Zou and Nair that the chlorine atom on the first ring and the nitro group on the third ring were transposed in the actual structure of TP-1, 9, (herein referred to as TP-1R).⁸ Thus, the proposed potency of TP-1P, coupled with the quandary of its chemical structure, provided an impetus to synthesize both previous and revised compounds to be fully characterized and tested in a P. aeruginosa QS reporter assay system. Herein, we report these findings and the unexpected activity of both compounds.

Synthesis of both TP-1R and TP-1P began from commercially available 3,5-dibromosalicylaldehyde 1, which was protected as MOM ether 2 (Scheme 1). Reduction of the aldehyde using NaBH₄, followed by conversion of alcohol 3 to the mesylate afforded compound 4. At this point, the Gabriel synthesis was invoked to provide primary amine 6, which serves as the common intermediate for both TP-1R and TP-1P. Towards TP-1R, amide 7 was formed in the presence of 2-nitrobenzoic acid, EDC, and HOBt, followed by cleavage of the MOM ether to achieve phenol 8. Finally, esterification of 8 with 2-chlorobenzoic acid yielded TP-1R (9).

Scheme 1.

Synthesis of TP-1R^a and TP-1P^b

^aConditions: (a) MOMCl, K₂CO₃, DMF, rt, 16 h, 93%; (b) NaBH₄, MeOH, THF, rt, 14 h, 96%; (c) MsCl, Et₃N, CH₂Cl₂, rt, 3 h, 65%; (d) Potassium phthalimide, DMF, rt, 19 h, 80%; (e) H₂NNH₂-H₂O, EtOH, reflux, 17 h, 85%; (f) (2-NO₂)PhCO₂H, EDC-HCl, HOBt, DMF, rt, 19 h, quant.; (g) TFA, CH₂Cl₂, 0°C-rt, 3h, 61%; (h) (2-Cl)PhCO₂H, EDC-HCl, DMAP, THF, rt, 13 h, 50%.

^bConditions: (a) (2-Cl)PhCO₂H, EDC-HCl, HOBt, rt, 19 h, 66%; (b) TFA, CH₂Cl₂, 0°C-rt, 3 h, quant.; (c) (2-NO₂)PhCO₂H, EDC-HCl, DMAP, THF, rt, 13 h, 52%.

The synthesis of TP-1P from 6 follows the same sequence with the following modifications: amide formation was performed in the presence of 2-chlorobenzoic acid to afford 10, followed by MOM deprotection and ester formation in the presence of 2-nitrobenzoic acid to provide TP-1P (12).

Before biological evaluation, the solubility and stability of TP-1P and TP-1R were measured to ensure optimal conditions for the cell-based QS assays. This is particularly important due to the presence of a seemingly labile o-nitro ester in TP-1P. Nevertheless, in agreement with previous reports,⁷ TP-1P and TP-1R were stable in buffer systems MES and TRIS with half-lives of 59.91 h and 607.5 h at pH 8 (Supplementary Information, Figure S1-S4 and Table S1). Furthermore, TP-1P and TP-1R were found to have a maximum solubility of 5 μM in 10% DMSO and a minimum tested solubility of 3.12 nM in 0.16% DMSO, thus ensuring the integrity of each compound under the assay conditions.

A reporter strain of P. aeruginosa based on the luxCDABE gene cassette was used to test agonistic activity of TP-1P and TP-1R compared to that of the natural autoinducer 3OC12-HSL.⁹ TP-1P, TP-1R, and 3OC12-HSL all demonstrated similar potent agonistic activity of LasR dependent signaling (Table 1). Interestingly, maximal luminescence was nearly equal in the case of TP-1R and 3OC12-HSL, but TP-1P only induced 50% luminescence relative to the former compounds (SI, Figure S5). Nevertheless, the finding of similar EC₅₀ values between the two triphenyl compounds was unforeseen in light of the perceived effect of the regiochemical change of the chloro and nitro positioning in the two agonists. This was even more puzzling considering that the LasR receptor protein has evolved to bind a vastly dissimilar ligand in 3OC12-HSL.

Table 1.

Biological activity and binding energies of TP-1P, TP-1R, and 3OC12-HSL.

Compound	EC50 (nM)	ΔG (kcal/mol)a
9, TP-1R	28.3 ± 4.3	−11.5
12, TP-1P	42.8 ± 4.0	−11.7
13	> 10000	n.d.b
3OC12-HSL	30.9 ± 1.3	n.d.b

^aRepresents relative free energy of binding between ligand and LasR.

^bn.d. = not determined.

Seeking to delineate the similar QS activity of TP-1P and TP-1R on a molecular level and elucidate the structural requirements of binding, computational studies were initiated to model the TP ligands into the LasR binding domain. Studies of TP-1P and TP-1R were simulated using Gromacs v4.0.7 to determine the necessity of the nitro and chloro substituents. Simulations of the 3IX4 complex⁸ did not show any highly occupied hydrogen-bonds between TP-1R and LasR. The most consistent interactions were between the amide proton of TP-1R and Asp73 and the amide carbonyl of TP-1R and Tyr56, but these interactions were transient as Asp73 also hydrogen-bonds with Thr75 and Tyr64 throughout the simulation and Tyr56 also hydrogen-bonds with Ser129 (Figure 1). The exchange of the nitro and chloro groups in TP-1R and TP-1P does not affect occupancy of either the hydrogen-bond with Asp73 or with Tyr56 in TP-1R. The only significant difference in the simulations was that with TP-1P, the amide carbonyl occasionally forms a hydrogen-bond with Ser129, which is unseen with TP-1R. Pleasingly, truncated Alchemical FEP simulations (Table 1) show only a small difference in the relative binding energy ΔΔG = 0.2 kcal/mol between TP-1R and TP-1P.

Figure 1.

Simulations from the 3IX4 structure show the amide in TP-1R forming hydrogen-bonds with Asp73 and Tyr56 (A). Whereas with TP-1P, Tyr56 has moved away and no longer interacts with TP-1P. In TP-1P the amide is rotated when compared to TP-1R, and makes hydrogen-bonds with Asp73 and Ser129 (B). In both simulations Asp73 is caged through interactions with Tyr64 and two interactions with Thr75 (A&B)

Due to the high degree of shape similarity of TP-1R and TP-1P, both compounds’ ability to make hydrogen-bonds through their amide moiety to nearby residues, and low ΔΔG of binding between the two molecules, it is not surprising from our simulations that they are both of similar potency. The positioning of the nitro and chloro groups within the binding pocket of LasR with no consistent hydrogen-bonding or electrostatic interactions with nearby residues sheds light on the similar activity observed with TP-1P and TP-1R. To investigate the effect of the scaffold alone, 2-(benzamidomethyl)phenyl benzoate (13) was synthesized in one step from 2-hydroxybenzyl amine and benzoic acid and tested in the P. aeruginosa reporter assay. This compound failed to give any QS activation at concentrations up to 10 μM (Table 1). Thus, while the regiochemical interchange of the nitro and chloro substituents does not seem paramount for activity, the presence of an electronegative substituent appears to be critical for LasR binding, dimerization, and ultimately gene expression.

In summary, through chemical synthesis, we have established the structural identity of TP-1, and through modeling, we have ascertained the minimal necessary chemical architecture for LasR activation. Lastly, we highlight the potential synthetic interchangeable pieces found within the 2-(benzamidomethyl)phenyl benzoate. The ester and amide units could readily serve as a viable grounding for the diversity and development of additional agonists and antagonists against LasR dependent QS in P. aeruginosa.