Biological Characterization, Mechanistic Investigation and Structure‐Activity Relationships of Chemically Stable TLR2 Antagonists

Abstract Toll‐like receptors (TLRs) build the first barrier in the innate immune response and therefore represent promising targets for the modulation of inflammatory processes. Recently, the pyrogallol‐containing TLR2 antagonists CU‐CPT22 and MMG‐11 were reported; however, their 1,2,3‐triphenol motif renders them highly susceptible to oxidation and excludes them from use in extended experiments under aerobic conditions. Therefore, we have developed a set of novel TLR2 antagonists (1–9) based on the systematic variation of substructures, linker elements, and the hydrogen‐bonding pattern of the pyrogallol precursors by using chemically robust building blocks. The novel series of chemically stable and synthetically accessible TLR2 antagonists (1–9) was pharmacologically characterized, and the potential binding modes of the active compounds were evaluated structurally. Our results provide new insights into structure‐activity relationships and allow rationalization of structural binding characteristics. Moreover, they support the hypothesis that this class of TLR ligands bind solely to TLR2 and do not directly interact with TLR1 or TLR6 of the functional heterodimer. The most active compound from this series (6), is chemically stable, nontoxic, TLR2‐selective, and shows a similar activity with regard to the pyrogallol starting points, thus indicating the variability of the hydrogen bonding pattern.


Introduction
The first barrier in the innate immune response is formed by the family of structurally conserved Toll-like receptors (TLRs). [1] In humans ten functional subtypes (TLR1-TLR10) have been identified. TLRs recognize intruding pathogens or endogenous danger signals released after cell damage or cell death and activate the innate immune response against them. This happens through the specific binding of pathogen-associated molecular patterns (PAMPs) and danger-associated molecular patterns (DAMPs), respectively. [2] TLR2 forms heterodimers with TLR1 and TLR6 and specifically recognizes several components of the cell wall of gram positive bacteria like di-and tri-acylated lipoproteins, lipoteichoic acids or lipomannans. After ligand binding, the preformed dimer undergoes conformational changes that trigger an intracellular signaling cascade that leads to the activation of NF-kB and the secretion of proinflammatory cytokines like tumor necrosis factor (TNF) and interleukin (IL)-8. [3] Under certain circumstances this response is excessive and leads to severe conditions like sepsis, rheumatoid arthritis, autoimmune diabetes, asthma and certain types of allergies. [1,4] The modulation of TLR2 function by small molecules has been postulated as a promising strategy to treat these conditions. To date only few compounds that modulate TLR2 activity have been identified and pharmacologically characterized. In 2010, four small organic molecules with agonistic activity on the receptor were discovered by high-throughput screening by Guan et al. [5] One of them was later chemically optimized. [6] In 2012 the first competitive antagonist CU-CPT22 was discovered by Yin et al. (Figure 1, left). [7] Virtual screening has successfully been applied to discover agonists and antagonists for TLR2, [8] but also for other TLR subtypes. [9] In a previous study, we identified a potent, competitive and selective TLR2 antagonist MMG-11; [10] however, it still contained the pyrogallol fragment (Figure 1, right). As the pyrogallol scaffold is notorious for its drawbacks including low chemical stability and poor synthetic accessibility, the modification of this scaffold to one that is less prone to oxidation is essential for further optimization steps.
In this work, we explored the chemical space around the pyrogallol-containing antagonists, MMG-11 and CU-CPT22, to enhance synthetic accessibility and chemical stability, and get insights into the structure-activity-relationships (SARs) of TLR2 antagonists. We performed synthetic modifications and analogue searches. The synthesized small molecules and selected analogues were biologically tested for their ability to inhibit TLR2 signaling. This leads to several novel TLR2 antagonists, a better understanding of their SAR and provides a way to rationalize binding modes of TLR2 antagonists.

Compound selection, design and synthesis
A selection of the compounds that resulted from both chemical synthesis and similarity-based analogue search is shown in Scheme 1. The design was guided by binding mode investigations of MMG-11 in complex with TLR2 regarding spatial requirements of the binding site and potential receptor-ligand interactions. Especially, we intended to modify the polyphenolic core structure, with the aim of avoiding the most easily oxidized 1,2-diphenols and 1,2,3-triphenols or the phenoxy ethers derived from them. Since the three hydroxy groups of the pyrogallol scaffold are involved in hydrogen bonding with the receptor (Figure 2A), [10a] we had to systematically evaluate these interactions. Therefore, we reduced the number of hydroxy groups capable to function as both hydrogen bond donors and acceptors (1, 3, 5 and 6) and varied the substitution pattern. For a systematic control, two compounds still comprising the 1,2,3trihydroxy motif, 7 and 8, were included in the study. Furthermore, we introduced methoxy groups, which can only serve as hydrogen bond acceptors (2 and 4). Considering the flexibility of the lead structure, we introduced an amide moiety to rigidify the molecules (1-4 and 6). In order to enhance synthetic accessibility and the chemical stability, we exchanged the furan moiety by a phenyl ring in all synthesized compounds. This resulted in a set of eight synthesis-derived compounds (1-8, Schemes 1 and 2 and the Supporting Information).
In a complementary approach we searched for structural analogues in the databases which were used for the discovery of MMG-11 by virtual screening. [10a] MMG-11 was used as the query structure and the databases were searched for similar commercially available molecules with a Tanimoto coefficient higher than 0.8. We found three closely related compounds in the Enamine database (Enamine Ltd, Kiev, Ukraine) which were ordered for biological evaluation (9-11). Molecular weight and purity (> 95 %) were confirmed by LC-MS.
The resulting set of eleven derivatives obtained by synthesis and analogue search, provides the possibility to conceive the SAR of TLR2 antagonists, in particular for the rationalization of the hydrogen bond pattern of polyphenolic ligands.
Compounds 1, 3, and 6 were synthesized starting from the corresponding 2,4-, 3,4-, or 3,5-dihydroxy-benzoic acids as exemplified for compound 6 in Scheme 2. First, both the phenolic hydroxy groups and the carboxylic acid residues were protected in one step as O-benzyl-ethers and esters, respectively, using benzyl bromide with iodide addition and furnishing the tri-O-benzyl-protected intermediates 12-14.
The tri-O-methyl-protected 3,4-dihydroxy benzoic acid 15 needed for the synthesis of compound 2 was prepared by an analogous protocol using methyl iodide for alkylation. Saponifi- Figure 1. Previously discovered competitive TLR2 antagonists CU-CPT22 and MMG-11 both containing the pyrogallol scaffold. [7,10a] Scheme 1. Synthesized and selected compounds. The starting point MMG-11 is depicted on the top with the different variations highlighted in color. Compounds 1 to 8 were rationally designed and synthesized and are shown on the left side. The compounds selected by analogue search (9 to 11) are shown on the right side.
Aromatic ketone 5, in which the amide linker between two benzene rings is replaced by a ketomethylene unit, was obtained via the direct C-acylation of resorcinol (1,3-diphenol) with 3,4-dimethoxy-phenyl acetic acid using boron trifluoride diethyl etherate as the activating Lewis acid in 19 % yield. Compounds 1-6 were isolated with > 95 % purity by column chromatography. Synthesis of the trihydroxy-derivatives, ethyl 3-(2,3,4-trihydroxy-benzamido)benzoate 7 and ethyl 3-(3,4,5trihydroxy-benzamido)benzoate 8 was attempted following the same strategy as in Scheme 2. While the preparation of the tri-O-benzyl-protected precursors of 7 and 8 proceeded smoothly, debenzylation of the protected intermediates led to the instantaneous decomposition of these products due to oxidation. 7 and 8 therefore could not be isolated and tested biologically.

SAR of novel antagonists
In order to rationalize the bioactivity and elucidate the potential binding modes of the tested compounds docking studies were performed. As all characterized compounds both inhibited TLR2/1 and TLR2/6 signaling, without showing effects at other TLRs, we hypothesized that they would bind to the TLR2 monomer and share the same ligand binding pocket like its pyrogallol-containing precursor, MMG-11. [10a] This compound has been previously shown to be a competitive antagonist for the binding of Pam 3 CSK 4 and Pam 2 CSK 4 , which further supports our proposed binding mode. However, an experimental confirmation of the binding mode is beyond the scope of this study. Extensive docking analyses were performed and plausible binding poses selected based on their similarity to the binding mode of MMG-11 and the amount of performed interactions between the ligand and the TLR2 binding site. The resulting proposed binding mode for the most active compound 6 is shown in Figure 2B.  Figure S2) and the dimethoxybenzamidobenzoates (2 and 4, Figure S3). The geometry of the 2,4-dihydroxy-benzamide 1 causes it to form H-Bonds to Ser346 and Lys347 but not Leu350 and Phe349 leading to a weak activity. The 3,4dihydroxy-benzamide 3 is more active than 1 as the mandatory H-Bonds to Leu350 and Phe349 are formed, but less active than 6 as the stabilizing H-Bond to Ser326 is missing. In the case of the dimethoxybenzamidobenzoates the methoxy groups are worse and bulkier acceptors than the hydroxy groups thus making the formation of the key H-Bonds less favorable. For these compounds we hypothesize a flipped binding mode which allows the carbonyl oxygen of the ester to interact with the backbone of Phe349 and Leu350, without the formation of further stabilizing H-Bonds towards Ser346 the resulting activity still is low. The dihydroxyphenyl moiety of compound 5 is surmised to form two H-Bonds towards Phe349 and Leu350, however its scaffold puts the methoxy groups into proximity of hydrophobic residues, which is unfavorable for binding and leads to a reduced activity ( Figure S4).
In addition to the synthesized compounds we tested three dihydroxy-phenyl-oxoethyl-furan-carboxylates (9, 10 and 11) of which one 11 was cytotoxic and therefore omitted in further analysis. The other two compounds, 9 and 10 show a binding mode nearly identical to that of MMG-11 ( Figure S5). Surprisingly, their activity is much lower, which shows the importance of the third hydroxy group for the stabilization of ligands containing the furan carboxylate scaffold. Interestingly, this is not the case for the newly developed and rigidified dihydroxybenzamidobenzoates like lead structure 6, which virtually retains the binding affinity despite of having only two remaining hydroxy groups.

Conclusion
Taken together, this study reports a set of TLR2 antagonists derived by chemical synthesis and analogue search based on the previously reported pyrogallol-containing TLR2 antagonists CU-CPT22 and MMG-11. The series of compounds (1-11) designed and tested here provides novel insights into the structure-activity relationships of TLR2 antagonists and allows for rationalization of structural binding characteristics. The findings prove the possibility of varying the pyrogallol fragment, the hydrogen bonding pattern, the second aryl aryl fragment and the linker moiety connecting both fragments. These tolerated variations enabled the development of chemically stable drug candidates without the sensitive pyrogallol scaffold. The nontoxic and most active compound from this series, inhibitor 6, shows a comparable activity toward TLR2 as its pyrogallol precursors without being sensitive toward oxidative degradation like its pyrogallol derivatives. Furthermore, the biological activity of this inhibitor supports our hypothesis that this group of TLR antagonists binds solely to TLR2 and does not directly interact with the TLR1 or TLR6 part of the functional heterodimer.

Experimental Section
Chemical synthesis: Chemicals were purchased from Acros, Alfa Aesar, Fluka, Sigma-Aldrich, and VWR and were used without further purification or distillation unless otherwise stated. All solvents that were used during the reactions were obtained from a column-based solvent system (MBraun, MB-SPS-800). Air or moisture-sensitive reactions were carried out under a positive pressure of dry nitrogen or argon in oven-dried glassware. Solution phase reactions were monitored either by LC-MS techniques or by thinlayer chromatography (TLC) using Analtech silica gel plates (60 F254) and the spots were examined under UV light at 254 nm or stained with developing reagents. LC-MS data were recorded on an Agilent 1100 series chromatography workstation (Agilent Technologies) equipped with a single quadruple mass spectrometer and electrospray ionization (ESI). Eluents A (0.1 % formic acid in Millipore water) and B (0.1 % formic acid in acetonitrile) were used in a linear gradient (5-99 % B in 5 min). Solvents were removed in rotary evaporators from Heidolph. The Biotage®Isolera TM Spektra One was used for regular flash purifications with pre-pack flash chromatog-ChemMedChem Full Papers doi.org/10.1002/cmdc.202000060 raphy cartridges from Biotage® employing individual gradients derived from analytical runs or thin layer chromatography (TLC). Non-HRMS measurements were recorded with an ESI single quad spectrometer (Agilent) coupled with an analytical HPLC system (Agilent 1100). HRMS measurements were conducted with an Agilent 6210 ESI-TOF mass spectrometer. NMR spectra ( 1 H and 13 C NMR) were recorded on a Bruker AVANCE 300 MHz, 400 MHz and 500 MHz instrument and chemical shifts (δ) were measured in parts per million (ppm). The elemental analyzer VARIO EL was used for C, H, and N analysis. Details for compound synthesis and analytic data are given in the Supporting Information.
General procedure for the preparation of perbenzylated benzoic acid esters 12-14: Benzyl bromide (6 equiv), sodium iodide (6 equiv) and potassium carbonate (8 equiv) were added to a solution of the respective benzoic acid in acetone (40 mL mmol À 1 ) at room temperature. The mixture was stirred under reflux for 16 h. Then it was cooled to room temperature and the solvent removed under reduced pressure. Water was added and the resulting mixture was extracted with EtOAc. The combined organic layers were dried over anhydrous Na 2 SO 4 and filtered. EtOAc was evaporated under reduced pressure and the resulting crude product was purified by column chromatography over SiO 2 (hexane/EtOAc as eluent) to afford the desired perbenzylated compounds.

Methyl 3,4-dimethoxy-benzoate (15):
To a solution of 3,4dihydroxy-benzoic acid (0.80 g, 5.2 mmol, 1 equiv) in DMF (30 mL) was added methyl iodide (1.50 mL, 25 mmol, 5 equiv) and potassium carbonate (3.59 g, 25 mmol, 5 equiv) at room temperature. The mixture was stirred at 50°C for 16 h, cooled to room temperature and concentrated under reduced pressure. Water was added and the resulting mixture extracted with EtOAc (3 × 20 mL). The combined organic layers were dried over anhydrous Na 2 SO 4 and filtered. The EtOAc was evaporated under reduced pressure and the resulting crude product purified by column chromatography on SiO 2 (hexane/EtOAc as eluent) to afford 15 as a white solid Preparation of benzoic acids 16-18. General procedure: To a suspension of the appropriate benzyl benzoates (12)(13)(14) in EtOH at room temperature was added an excess of 5 M NaOH solution and the resulting mixture stirred under reflux for 8 h. The reaction mixture was cooled to room temperature and acidified to pH 3 by dropwise addition of 12 M HCl. The resulting precipitate was filtered off, washed with water and dried under reduced pressure to afford the corresponding acids 16-18.

Preparation of 3,4-dimethoxy-benzoic acid (19):
To a suspension of 15 (0.35 g, 1,79 mmol, 1 equiv) in H 2 O/MeOH (1 : 1 v/v, 30 mL) at room temperature was added LiOH monohydrate (0.23 g, 5.36 mmol, 3 equiv) and the resulting mixture stirred for 4 h. The reaction mixture was then brought to pH 3 using 1 M HCl and extracted with EtOAc. The combined organic layers were dried over anhydrous Na 2 SO 4  General procedure for the preparation of benzamides 2, 4, and 21-23: To a mixture of the appropriate acid (1 equiv) and 20 (1 equiv) in CH 2 Cl 2 (5 mL mmol À 1 ) was added HATU (1.2 equiv) and DIPEA (3 equiv) at 0°C and after which it was stirred at 30°C for 8 h. The solvent was removed under reduced pressure, water was added and the resulting mixture was extracted with EtOAc. The combined organic layers were dried over anhydrous Na 2 SO 4 and filtered. The EtOAc was evaporated under reduced pressure and the resulting crude product was purified by column chromatography over SiO 2 (hexane/EtOAc as eluent) to afford the desired amides 2, 4, 21-23.
Cell viability: Cell viability was assessed by the MTT assay in HEK Blue hTLR2 cells as previously described. [14] 10 % (v/v) DMSO (Carl Roth) served as control and the viability of untreated cells was defined as 100 %.
Statistical analysis: Data of the bar charts are shown as mean + SD. Potency (IC 50 ) data are presented as mean with the confidence interval (95 %). Statistical analysis was done by using GraphPad Prism 6.0 (GraphPad software, San Diego, USA). Nonlinear regression was used to plot and analyze concentration-response curves and to obtain IC 50 values.

Computational methods:
The crystal structure of the heterodimer of TLR2-TLR1 with bound Pam 3 CSK 4 (PDB ID: 2Z7X) [15] was retrieved from the Protein Data Bank [16] and used for docking studies with the tested compounds. Prior to docking the TLR1 monomer, all ligands and water molecules were removed using Molecular Operating Environment (MOE2019, Chemical Computing Group, Montreal, QC, Canada). The TLR2 monomer was protonated using the "Protonate 3D" application included in MOE2019. The GOLD Suite v.5.2 (Cambridge Crystallographic Data Centre, Cambridge, UK) [17] was used for docking with the GoldScore [18] as scoring function with "slow" parameters. Binding poses were minimized (MMFF94 force field) [19] and further analyzed in LigandScout 4.2 (Inte:ligand, Vienna, Austria). [20]