Diaryl ethers with carboxymethoxyphenacyl motif as potent HIV-1 reverse transcriptase inhibitors with improved solubility

Abstract In search of new non-nucleoside reverse transcriptase inhibitors (NNRTIs) with improved solubility, two series of novel diaryl ethers with phenacyl moiety were designed and evaluated for their HIV-1 reverse transcriptase inhibition potentials. All compounds exhibited good to excellent results with IC50 at low micromolar to submicromolar concentrations. Two most active compounds (7e and 7 g) exhibit inhibitory potency comparable or even better than that of nevirapine and rilpivirine. Furthermore, SupT1 and CD4+ cell infectivity assays for the most promising (7e) have confirmed its strong antiviral potential while docking studies indicate a novel binding interactions responsible for high activity.


Introduction
Non-nucleoside reverse transcriptase inhibitors (NNRTIs) have proven their effectiveness as components of highly active antiretroviral therapy [1][2][3] . Their relatively low toxicity, as compared to other antiretroviral drugs, makes them a very attractive class of compounds used in treating HIV-1 infections [4][5][6][7] . Currently, there are five registered NNRTIs, first generation: nevirapine (NVP), efavirenz (EFV), delavirdine, and second generation: etravirine (ETV) and rilpivirine (RPV). Because HIV-1 reverse transcriptase (RT) has a low fidelityits error rate was reported to be in the range of 10 À3 -10 À5 per nucleotide addition 8-10there is a very high mutation rate of the virus, and strains resistant to antiretroviral drugs emerge. Consequently, the pharmacotherapy may become ineffective, moreover, cross-resistance between NNRTIs is possible [11][12][13][14] . Another problem is that the NNRTIs binding site of RT favours non-polar compounds, which are usually poorly soluble in water. This is especially the case in second-generation NNRTIs, as both ETV and RPV are practically insoluble in water and require special formulations 15,16 . For these reasons there is a need to develop new NNRTIs with improved potency against resistant HIV mutants and better pharmacokinetics [17][18][19] . First generation NNRTIs like NVP and EFV are rigid molecules that bind well to the wildtype RT, but a single amino acid mutation in the binding site can significantly decrease their affinity to the enzyme. Second generation NNRTIs have flexible structures which allows them to adapt to a modified binding site of mutant RT 20 . Usually, second generation NNRTIs have 2-3 aromatic rings with an ether, thioether, short alkyl or amino group located between the rings that acts as a hinge that allows the inhibitors to bind in different conformations and overcome resistance mutations 20,21 . An excellent review on the chemical diversity of NNRTIs was written by Zhan et al. 18 . Diaryl ethers are one of the classes of second generation NNRTIs. There are several interesting inhibitors belonging to this class, including 1the most potent NNRTI reported to date (against wild type RT) and doravirine (2), which is in phase III clinical trials ( Figure 1) [22][23][24] .
As mentioned above, poor solubility in water results in reduced bioavailability, and there is an increasing awareness of the need to design NNRTIs with improved pharmacokinetics. Several approaches were used by different authors to achieve better solubility of NNRTIs: salt formation 25,26 , prodrug formation 27,28 , addition of polar substituents [29][30][31] , modification of crystal structure 23 or reduced halogenations 32 .

Synthesis
Compounds 7a-g (resorcinol type) and 8a-f (catechol type) were synthesised in several steps from commercially available starting materials. Diaryl ether parts (9a-f) of the new NNRTIs were synthesised from phenols and aryl fluorides in N-methylpyrrolidone ( Figure 3) as described earlier 34,35 . In case of 9b Chan-Lam coupling was used 37 . Hydroxyacetophenones were O-alkylated with ethyl chloroacetate. Subsequent exchange of ethyl to methyl afforded pure and solid methyl esters, which were selectively brominated with N-bromosuccinimide and p-toluenesulfonic acid in chloroform (10a-d) (Figure 3) 38 . Final deesterification was performed using potassium carbonate in a mixture of methylene chloride, methanol and water (room temperature, 1-2 days). Structures of obtained compounds are given in Table 1. Detailed synthetic procedures and characterisation data of reported compounds can be found in the supplemental material.
atoms with partial charges less than 0.2 were scaled down by a factor of 0.9 41 . Several docking protocols were used: Standard and extra precision (XP) Glide docking with default settings [42][43][44][45] . Three best poses were stored. For XP docking the threshold to reject a minimised pose was increased to 0.9 kcal/ mol. Custom protocol 1 (CP1): The following positional constraint was added during receptor grid preparationa 1 Å sphere centred at the oxygen atom of diphenyl ether motif of a reference ligand (from the crystallographic structure) had to be occupied by a neutral H-bond acceptor atom of a docked ligand. All ligand poses which did not comply with the constraint were rejected. Custom protocol 2: ligand poses obtained with the standard or custom (CP1) protocol were examined, and one best pose was manually selected (basing on the similarity to the reference ligand pose from crystallographic structure). All ligands were then aligned to the selected pose using the Flexible Ligand Alignment tool from Maestro, using a maximal common substructure. Subsequently, ligands were docked using Glide XP with the sampling mode set to "None (refine only)" Only the best scoring pose obtained with either protocol was kept for each docked ligand. The complete results can be found in the supplemental file.

IC 50 measurements
HIV RT inhibitory activity of the new compounds was measured using the colorimetric Reverse Transcriptase Assay kit (Roche, Basel, Switzerland). The assay was performed according to the manufacturer's instruction with the only modification that 2 ng (instead of suggested 4-6 ng which caused the reaction to run too fast) of enzyme supplied was used for the each reaction. Stock solutions of examined and reference compounds (NVP (TCI, Tokyo, Japan) and RPV) containing 5% dimethyl sulfoxide (Merck, Darmstadt, Germany) were used to prepare dilutions ranging from 0.1 to 50 mM. Averaged results from at least two measurements were used to obtain an inhibition curve. IC 50 values were obtained by interpolating the curve using non-linear regression.

Solubility
Substances were dissolved in ultrapure deionised water at room temperature up to concentrations of 50 g/L. Solutions were centrifuged and their aliquots were transferred into weighed glass vials (d ¼ 0.01 mg). The samples were dried under reduced pressure and the vials were weighed again. The solubility was calculated from mass difference divided by sample volume.
Lyophilised compound was dissolved in dimethyl sulfoxide to a concentration of 5 mM, and further diluted in cell culture medium (SupT1: IMDM and CD4 þ T cells: RPMI). Cells were incubated 2 h prior infection in appropriate medium as described above, now also supplemented with diluted compound. Subsequently, cells (SupT1: 50,000 per 96 well, CD4 þ T cells: 250,000 per 96 well) were infected with HIV (HIV NL4-3-GFP-I, an infectious virus expressing green fluorescent protein (GFP) from gfp-IRES-nef mRNA expressed from the nef locus, as described earlier 48 . Medium was as described above, now also supplemented with diluted compound (For CD4 þ T cells, PHA was left out). Medium was refreshed after 24 h, keeping compound and supplement concentrations constant. After 72 h, cells were harvested, measured by flow cytometry for GFP expression and counted, using a MACSQuant flow cytometer (Miltenyi Biotec, Bergisch Gladbach, Germany). Infection rate measured by GFP expression was between 6 and 20%. Cell numbers were used to measure cytotoxicity (cell numbers in non-infected cultures supplemented with compound were compared to cell numbers obtained in parallel cultures without compound added: a reduction with more than 10% was considered to indicate cytotoxicity).

Results and discussion
RT inhibitory activity 7a and 8a were tested for the inhibitory activity against HIV-1 RT. Both compounds were found to be NNRTIs with IC 50 of 1.23 ± 0.05 lM and 23.4 ± 1.6 lM, respectively. Using 7a and 8a as lead compounds several of their analogues were prepared and tested in vitro. Replacement of nitrile with chlorine in 7b was detrimental to the inhibitory activity (Table 2). Changing the position or removing the chlorine atom in the central ring of 7a also resulted in increased IC 50 values (7c and 7d, Table 2), but in case of catechol ethers moving the position of chlorine from 4 in 8a to 5 in 8b was beneficial for activity (about seven-fold decrease the IC 50 value, Table 2). Modifications introduced to the phenacyl moiety of the new inhibitors resulted in several interesting findings. 7e was found to be the most active NNRTI in this study with IC 50 0.36 ± 0.01 mM, more potent than the drug NVP (IC 50 ¼ 0.75 ± 0.02 mM) and nearly as potent as RPV (IC 50 ¼ 0.32 ± 0.04 mM). Analogous modification in catechol series resulted in 8c, more potent that its parent compound 8a (IC 50 1.9 ± 0.11 mM). Methylation of 7a in the phenacyl ring yielded 7f (less potent) and 7 g, which was found to be another potent inhibitor of HIV-1 RT, slightly more active than NVP (IC 50 0.65 ± 0.03 mM). Catechol analogues 8d-f were found to possess comparable, low micromolar activity (Table 2).

Molecular docking analysis
Using one of the three docking protocols, a good pose for every inhibitor could be found for all four receptors. Averaged docking scores are given in Table 2 (individual scores are in Table S4 of supplemental material). Interestingly, the obtained mean docking scores show quite good qualitative correlation with IC 50 values ( Table 2). For example, for compounds 7a-g the docking scores almost correctly rank the inhibitors by their activity: 7e, 7 g, 7f, 7a, 7c, 7 b, 7d, with only 7f being swapped with 7a, as well as 7c with 7b. For 8a-f the docking scores also were able to identify the three most active inhibitors. The results show that the molecular docking may be a valuable tool in further development of NNRTIs from this chemical class.
The docking analysis was also useful in gaining some insight into the binding interactions of the new inhibitors with RT. The diaryl ether part of examined compounds binds in the hydrophobic cavity formed by non-polar aromatic amino acid residues of Tyr181, Tyr188, Phe227, Trp229 and Tyr318 (Figure 4(a)). The 3chloro-5-cyanophenyl ring forms a p-p stacking interactions with Trp229 and Tyr188. The chlorine atom in the central aromatic ring points towards the carbonyl oxygen of Tyr188, forming a halogen bond. The carbonyl group of phenacyl moiety forms a hydrogen bond with the backbone oxygen of Lys103. Finally, the carboxyl group of three-ringed inhibitors locates itself at the solvent exposed entrance to the binding site, forming a hydrogen bond with Val106. In case of compounds 7e, 8c and 8f, the additional phenyl ring between phenacyl and glycolic acid synthons is predicted to bind in an unexplored pocket adjacent to the entrance to the NNRTIs binding site (Figure 4(b)). This pocket may be an attractive target for the further optimisation of the new inhibitors. The predicted binding modes of 7e, 8c and 8f suggest the existence of a strong ionic interaction between the inhibitors carboxyl and guanidine of Arg199, which explains the unusually high-docking scores for these three compounds (e.g. -19.62 for 7e in 3C6T receptor, see Table S3), and may also be the cause of 7e potency.
In order to assess expected activity of our best compound against mutated forms of the enzyme we have performed docking studies of 7e and RPVclinically used drug of the second generation that is active against these mutations. For wild-type, K103N, and Y181C forms 3DRP, 3DRS and 3DRR structures were chosen,   Table 3. Threshold toxic concentrations and antiviral activity at 10-fold lower concentration of compounds examined.

Compound
Threshold toxicity [mM] Antiviral activity  Figure 5). Interestingly, 7e shows practically the same scores for WT, Y181C and K103N/Y181C forms, and an increased score for K103N ( Figure 5). This suggests that 7e is not sensitive to these mutations.

Solubility
The aqueous solubility (at pH ¼7) of synthesised compounds is presented in Table 2. Many of the compounds show a good solubility, exceeding 50 g/L. Exact measurements of the maximum solubility proved to be infeasible due to the formation of micelles, aggregates and ultimately gels for increasing concentrations of the compounds. This behaviour results from the amphiphilic character of the examined compounds and elongated shapes of their molecules, which gives them surfactant-like properties. The general trend observed for all inhibitors is that catechol-based ethers (8a-f) are better soluble than their resorcinol counterparts (7a-g). Interestingly, removing or replacing the chlorine atom in two-position of resorcinolbased ethers like in case of 7c and 7d significantly increased their solubility. However, as discussed above, the two-position of the chlorine atom is optimal for the inhibitory activity. Methyl substitution in the phenacyl ring was also detrimental for the solubility. The most active compound 7e shows relatively low solubility, but it is still ca. 100,000 times greater than that of RPV. Calculated octanolwater partition coefficients (logD) for examined compounds range from 0.97 to 3.05, which is close to that of NVP (2.49), and significantly lower than that of RPV (5.47, Table S5).

Infection assays
First, toxicity of the compounds was determined in SupT1 cells.
Since even minor toxicity affects the support of viral replication by the cell, compounds should be active well below threshold toxic concentration (i.e. concentration at which toxicity is observed in 10-fold titration series). The threshold toxic concentration, and whether antiviral activity was observed at 10-fold lower concentration than this threshold toxic concentration are given in Table 3. Several compounds tested in infectivity assay did not show antiviral activity at less than 10% of toxic concentration (Table 3). Nonetheless, of those who did, 7e inhibited infection most clearly below concentrations which affected cellular viability. Therefore, this compound was tested more extensively in SupT1 cells, as well as in peripheral blood CD4 þ cells. As shown in Figure 6, both in SupT1 cells as in primary T cells, IC 50 was around 0.25 mM (toxicity was only apparent above 20 mM). The measurements were run in parallel with NVP as a reference, and it showed IC 50 of 0.04 mM, in line with literature 50 .

Conclusions
Two novel scaffolds of diaryl ether NNRTIs with phenacyl moiety were designed in this study. With the aid of molecular modelling several modifications of the core structures were prepared. Using molecular docking to several crystallographic structures and thorough conformational search of ligands, it was possible to obtain quite accurate predictions of structure-activity relationship. All synthesised compounds showed inhibitory activity against wildtype HIV-1 RT. In general, resorcinol-based compounds possessed better activities than catechol-based compounds, and are more promising candidates for further development. One of the compounds, 7e, was found to be a very potent NNRTI in enzymatic assay, and is predicted to display novel interactions with the RT. The presented compounds were designed to possess a good aqueous solubility, which was achieved in all cases. Given those encouraging results, the inhibitors were subjected to biological evaluation of their efficacy against HIV infection in vitro. 7e proved to be a potent anti-viral in SupT1 and CD4 þ T cell infectivity assays. These results show our design could deliver highly watersoluble NNRTIs, at least one compound displays potent antiviral activity in infection assays in vitro.

Disclosure statement
The authors report no conflicts of interest.

Funding
The reported studies were supported by the grant 2011/02/A/ST4/ 00246 (2012-2017) from the Polish National Science Centre (NCN) to PP, and by Ghent University grant BOF17/GOA/013, HIVERA