Design, Synthesis and Structure-activity Studies of Rhodanine Derivatives as HIV-1 Integrase Inhibitors

Raltegravir was the first HIV-1 integrase inhibitor that gained FDA approval for use in the treatment of HIV-1 infection. Because of the emergence of IN inhibitor-resistant viral strains, there is a need to identify innovative second-generation IN inhibitors. Previously, we identified 2-thioxo-4-thiazolidinone (rhodanine)-containing compounds as IN inhibitors. Herein, we report the design, synthesis and docking studies of a series of novel rhodanine derivatives as IN inhibitors. All these compounds were further tested against human apurinic/apyrimidinic endonuclease 1 (APE1) to determine their selectivity. Two compounds showed significant cytotoxicity in a panel of human cancer cell lines. Taken together, our results show that rhodanines are a promising class of compounds for developing drugs with antiviral and anticancer properties.


Introduction
Raltegravir (MK-0518) is the first HIV-1 integrase (IN) inhibitor approved for the treatment of HIV-1 infection in treatment-experienced patients, and heralds a new era in HIV/AIDS treatment [1,2]. IN inhibitors differ from other classes of antiretroviral agents in that they target the integration stage in the viral life cycle. The integration step, catalyzed by IN, results in the insertion of proviral DNA into the host genome. This occurs as a two-stage reaction. First, the proviral DNA is primed for integration by an endonucleolytic cleavage at its 3' end. This step is called 3'-processing and takes place in the cytoplasm of an infected cell. IN remains bound to the processed viral DNA and translocates into nucleus aided by several interacting cellular cofactors. Inside the nucleus, IN catalyzes a staggered cleavage in the host DNA. The nucleophilic hydroxyl group at the 3' end of the viral DNA is then inserted into the host DNA by a trans-esterification reaction. This process is called the strand transfer reaction. These two steps catalyzed by IN are critical for HIV-1 replication and infection [3][4][5].
Nearly two decades of intensive efforts towards developing clinically useful HIV-1 IN inhibitors has resulted in a rich database of structural information [6][7][8]. This is particularly useful in designing second generation inhibitors in order to overcome the challenges faced with emerging resistant strains against current IN inhibitors, including raltegravir. Among the various structural classes, β-diketoacid based pharmacophores have shown high potency and selectivity towards inhibition of the IN-catalyzed strand transfer [9,10]. S-1360, a β-diketoacid bioisostere, was the one of first IN targeted antiretroviral agent to undergo human clinical trials. Although it failed to show clinical efficacy, S-1360 has served as a good template for the design of several second-generation IN inhibitors [11][12][13].
Previously, we identified rhodanine-containing compounds as IN inhibitors through a common feature pharmacophore search based on β−diketoacid bioisosteres ( Figure 1) [14]. We reported that compounds containing both a rhodanine moiety and a salicylic acid substitution exhibited significant IN inhibition. The most potent compound 1 inhibited IN catalytic activities with IC 50 values of 15 μM and 11 μM for 3'-processing and strand transfer, respectively. This prompted us to explore other modifications in the substructures of compound 1 ( Figure 1) and study their effects on IN inhibition. Another earlier study also similarly applied a caffeic acid phenylester (CAPE)-based pharmacophore approach and identified rhodanine-containing compounds bearing a phenylsulfonamide moiety as IN inhibitors [15].
Several rhodanine-based small molecules have also known to inhibit various enzymes such as dual-specificity phosphatases, HCV NS3 protease, phosphodiesterases, etc. (Figure 2) [16][17][18][19]. The central rhodanine scaffold is commonly found in several antivirals, antimicrobials and antitumor agents [20,21]. In all these structures, modifications in the alkylidene chain and in the rhodanine moiety appear to influence biological activity. While there is diverse literature on rhodanine-containing compounds as various enzyme inhibitors [22], there is limited information regarding their use as HIV-1 IN inhibitors. Herein, we report the synthesis and structure-activity relationship (SAR) for a series of rhodanine derivatives that inhibit IN in the low micromolar range. We further characterized the selectivity of these compounds for IN by counter-screening against APE1, another DNA binding enzyme, and analyzed their binding modes in the IN and APE1 active sites. Finally, we evaluated the antiproliferative effects of these rhodanines against a panel of human cancer cell lines.   [17][18][19][20]; (2) JSP-1 inhibitor; (3) PDE4 inhibitor; (4) HCV NS3 Protease inhibitor; (5) Antitumor agent (rhodanine moiety is highlighted in red).

Chemistry
In our earlier studies, we identified rhodanine-containing compounds through a pharmacophorebased high-throughput screening. We found that the presence of both a rhodanine core and a salicylic acid group improved IN inhibitory effects [14]. Furthermore, substitutions in the aryl substructure and at the 4-position in the rhodanine nucleus (alkylidene substructure) also appeared to influence IN inhibition potency. Based on this, we synthesized several 2-thioxo-1,3-thiazolidin-4-ones bearing modifications in both the aryl and the alkylidene substructures to explore their effect on IN catalytic activities and understand their SAR. These rhodanines were synthesized according to previously a reported procedure (Scheme 1) [23]. All of these new analogues possessed a five-membered rhodanine moiety as a common structural unit and are linked to various substituted aromatic groups via alkylidene and four-carbon aliphatic thioxoamide linkers (Figure 1). Scheme 1. Synthesis of compounds 6-54. Ar

In vitro IN inhibitory activities and SAR of rhodanine derivatives
Inhibition of IN catalytic activities, 3'-processing and strand transfer, was determined using an in vitro enzymatic assay. Compounds with a weakly electron-withdrawing chlorophenyl group in substructure B (e.g., 6-8) showed no IN inhibition at the highest concentration tested (Table 1). Interestingly, compound 9, which has a methoxy substitution in ring A, inhibited both 3'-processing and strand transfer with an equal potency (IC 50 value of 33 μM). Substitution with the weakly basic and bulky indole group abolished any activity (compound 10). Compound 11, with a stronger electronwithdrawing substitution in substructure B, had a greater potency towards IN inhibition with an IC 50 value of 58 ± 4 μM and 20 ± 6 μM for 3'-processing and strand transfer, respectively. 58 ± 4 20 ± 6 >100 Compounds 12-21 have a relatively stronger electron-withdrawing nitrofuran ring in substructure B. Again, we observed that electron-withdrawing groups in ring A led to moderate inhibitory activity, while replacement by an electron-donating group such as a methoxyl improved activity further (Table 2).
The presence of a basic moiety such as tetrazole ring at position 4 in ring A led to a loss of activity. Interestingly, introduction of electron-donating group in substructure B did not seem favorable for IN enzyme binding, as all of the compounds 22-26 were inactive (Table 3).
Since the presence of an electron-donating methoxy group in ring A seemed important for IN inhibitory activity, we synthesized analogs with trimethoxy substitution on ring A (Table 4). Substructure B was also modified using various substituted phenols. Electron-withdrawing substituents like Cl, Br and NO 2 led to increased activity. Compound 35 with a salicylic acid had only a moderate effect on IN inhibition, while compound 34 with a 2-phenyldiazaphenol moiety inhibited 3'-processing and strand transfer, with IC 50 values of 33 ± 19 μM and 26 ± 14, respectively. Aliphatic and basic substitutions had a negative effect on activity.  In order to further understand the effect of electron-donating groups in ring A on activity, we synthesized analogs with hydroxyl groups at position 4. Here again, strong electron-withdrawing and hydrophobic substitutions in substructure B appears to influence IN inhibition ( Table 5). The most potent IN inhibition was seen with compound 53 having 3,5-diiodophenol substitution (IC 50 value of 7 ± 3 μM and 3 ± 2 for 3'-processing and strand transfer, respectively). Compared to compounds bearing trimethoxy substitution (Table 4), some of these compounds are more potent, suggesting that electron-donating substituents in ring A, together with strong electron-withdrawing and hydrophobic substitution in substructure B, leads to favorable IN inhibitory activity. Interestingly, substitution with a hydroxyl group in ring A improved selectivity towards strand transfer inhibition ( Figure 3A). This may be due to possible formation of stabilizing hydrogen bonds at the IN active site. Overall, these rhodanine derivatives with an aliphatic thioxoamide linker have an improved IN inhibitory activity over those previously reported with an aliphatic linker [14].

In vitro APE1 Inhibitory Activity
The specificity of these rhodanine derivatives against IN was assessed by counter-screening against human apurinic/apyrimidinic endonuclease 1 (APE1), another DNA-binding enzyme. APE1 plays an important role in DNA repair by the base excision pathway. Like IN, APE1 possesses endonucleolytic activity and cleaves the phosphodiester DNA backbone 5' to its recognition site. Like IN and other DNA-binding enzymes, optimal APE1 catalytic activity requires the presence of a divalent metal ion. APE1 is also an attractive target for designing cancer therapeutics [24][25][26].
Most compounds showed no inhibition against APE1 at the maximum tested concentration of 100 μM. Compound 53, with the most potent IN inhibitory activity, exhibited only a weak APE1 inhibition with an IC 50 value of 62 ± 3 μM. Its inhibitory activity is 20-fold more selective for IN than APE1. Similarly, compounds 48 and 51, with promising IN inhibitory activity, did not show any significant APE1 inhibition at 100 μM. On the other hand, two compounds, 7 and 8, showed a moderate to weak APE1 inhibition with IC 50 values of 45 ± 51 μM and 76 ± 5 μM, respectively, while they lacked any IN inhibitory activity. Similarly, compound 9 inhibited APE1 activity with an IC 50 value of 47 ± 2 μM ( Figure 3B). However, it also exhibited an equal inhibitory effect on IN (IC 50 value of 33 μM), suggesting possible similarities in its mechanism of action.

Docking Studies
To predict the possible binding modes and enzyme inhibition mechanism, compounds 6-54 were docked onto the active sites of both HIV-1 IN and APE1, using GOLD 4.0, the automated docking program, and Glide, the grid-based ligand docking with energetics software from Schrödinger. The docking scores for these compounds are reported in Table 6. A plot of the pIC 50 (ST activity) versus the predicted Glide docking score of these 49 compounds against IN is shown in Figure 4. Although a significant correlation between docking scores and inhibitory activities of the compounds was not routinely observed, there was a general trend where many of the active compounds scored high. Of the 32 active compounds with a ST IC 50 value less than 100 µM, 24 compounds were correctly predicted as true positives. Eight compounds, which were moderately active or least active, were predicted as false negatives. Similarly, out of the 18 inactive compounds with ST IC 50 values greater than 100 µM, 9 compounds were correctly predicted as true negatives, while 9 inactive compounds were over predicted as false positives. We found no correlation between the APE1 inhibitory activity and the docking scores.   Compounds 6-8, with a weak electron-withdrawing group on ring B, did not show any IN inhibitory activity at the highest tested concentration (Table 1). However, the introduction of a carbonyl group in substructure B (11) significantly improved strand transfer activity. Interestingly, docking simulations with Glide also estimated a high fitness score for this compound. The observed activity of the compound 11 supports the predicted binding interactions inside the IN active side. As shown in Figure 5, the oxygen atom of the rhodanine core and the oxygen atom of the substructure ring B are involved in interactions with both the Mg 2+ ion and the active site residues D64 and D116. Additionally, the donor nitrogen atom of 11 forms a hydrogen bond with E152. The importance of this additional carbonyl of 11 in substructure B can further be understood by comparing against the activity and binding mode of compound 9 ( Figure 5). A possible explanation for the lower activity of 9, as compared to 11, might be the lack of such interactions with key active residues D64 and D116 in the substructure B. Compounds 6-8 show a similar binding mode at the HIV-1 IN active site.
Compounds 12-21 with an electron-withdrawing nitrofuran ring in substructure B exhibit a moderate inhibition of strand IN transfer. Docking results predicted seven of these moderately active (12-17 and 20) compounds as false negatives, with their predicted activities lower than their actual activities. Compounds 27-40, with electron-donating trimethoxy group on ring A, have shown moderate IN inhibitory activity. Docking results predicted four inactive compounds (30-31, 33 and 36) as false positives, with their predicted activities being higher than their actual activities. The binding mode of the most active compound 53 is shown in Figure 6. The residues in close contact with the compound 53 were D64, C65, T66, H67, D116, I141, Q148, E152, and N155. As shown in Figure 6, the carbonyl oxygen atom of the rhodanine and the hydroxyl oxygen atom in ring B are involved in interactions with both the Mg 2+ ion and the amino acid residues D64 and C65. The two donor nitrogen atoms of 53 formed strong H-bonding interactions with the key active site residues D64 and E152. Additionally, the hydroxyl oxygen atom of the ring A was involved in a strong H-bonding interaction with the amino acid residue I141. The presence of these interactions with key active site residues explains the potency of 53. A close analogue to 53 is compound 39. The replacement of p-hydroxyl group (53) in ring A by a trimethoxy group (e.g., compound 39) resulted in a 10-fold decrease in potency. The binding mode of compound 39 is shown in Figure 6. A possible explanation for its lower activity, as compared to 53, might be the lack of H-bonding interactions with the amino acid residue I141. These docking results confirm the observations from the SAR studies. Indeed, the phydroxy group in ring A is able to form stabilizing H-bonds with I141. This accounts for the increased potency of compounds 41-54 as compared to their trimethoxy analogues 27-40.
Orientation of the rhodanine core and substructure B are also important for IN inhibitory activity, contributing to the formation of key interactions with the metal ion and active site residues. Accordingly, replacement of the iodine groups on ring B by t-butyl groups (54) led to a drastic change in the binding pattern of ring B and a loss of activity. Though Glide estimated similar scores for both 53 and 54, differences in activity can be clearly explained from the GOLD docking poses ( Figure 6). With the introduction of the bulky t-butyl groups, the rhodanine ring becomes inverted and no longer interacts with the Mg 2+ ion and the active site residues. This might be responsible for the loss of activity of compound 54.

Drug-like Properties
Oral bioavailability is a desirable property of compounds under investigation in the drug discovery process. Lipinski's rule-of-five is a simple model to forecast the absorption and intestinal permeability of a compound [27,28]. In the rule-of-five model, compounds are considered likely to be well absorbed when they possess these attributes-molecular weight < 500, cLogP < 5, number of H-bond donors < 5, number H-bond acceptors < 10, and number of rotatable bonds < 10. Further, we calculated the polar surface area, which is also used to predict drug absorption [29,30]. Earlier studies by Palm et al. [31][32][33] and Kelder et al. [34] suggest that the compounds with a polar surface area >140 Å 2 would tend to show poor (<10%) absorption, whereas the compounds with polar surface area <60 Å 2 could be predicted to show complete (>90%) absorption. The calculated atom-based Log P (S + logP) values of most of the active compounds ranged from 2.0 to 5.0, and the H-bond donor and acceptor counts are ≤5 and <10, respectively. High lipophilicity is often an issue with many rhodanine-containing structures impeding their cell permeability and in vivo activity [22]. However, most of the compounds presented here have polar surface areas <140 Å 2 and log P value of <5. Therefore, most of these analogs have favorable drug-like properties and are potentially interesting for further optimization.

Cytotoxicity and Antiviral Activity
Several rhodanine structures such as the rhodacyanine dyes have been found to possess anticancer and antitumor properties. Hence, we evaluated the ability of these rhodanine-based compounds to inhibit the proliferation of various cancer cell lines. Compounds 13 and 17 were moderately active against a colon cancer cell line (HCT 116) and a pancreatic cell line (Panc-1) (GI 50 values provided in Table 8). The remaining compounds lacked any significant cytotoxicity at the maximum tested concentration of 10 μM. We also tested the antiviral activity of compounds 48, 51 and 53 against HIV-1 in MT-4 cell infectivity assay. Unfortunately, these compounds did not show any significant antiviral activity.
Optimization studies are underway to design rhodanines with more favorable physiochemical properties and improved antiviral activity.

General
The 1 H-NMR spectra were recorded on a Bruker AC spectrometer (200 MHz) using DMSO-d 6 as solvent. The data are given in parts per million (ppm) and are referenced to an internal standard of tetramethylsilane (TMS, δ 0.00 ppm). The spin-spin coupling constants (J) are given in Hz. Peak multiplicity is reported as s (singlet), d (doublet), dd (double doublet), t (triplet), m (multiplet), and br s (broad singlet). The mass spectra were obtained on a Kratos instrument using a direct inlet system; the ionization energy was 70 eV; the accelerating voltage was 1.75 kV. Melting points were measured on a Boetius hot-stage apparatus. 5-Benzylidene-4-oxo-2-thioxothiazolidines 6-54 were purified by silica gel flash chromatography column (eluent: 3:1 petroleum ether-ethyl acetate). 2-Thioxo-1,3-thiazolidin-4-ones were prepared according to a previously reported procedure (Scheme 1) [23].

Docking Studies
APE1. Three-dimensional coordinates of the crystallized structure of human APE1 endonuclease with bound abasic DNA and Mn 2+ ion were obtained from RCSB PDB (PDB ID: 1DE9). For the protein target, a 20 Å radius active site was defined using xyz coordinates of the 5'-phosphate fragment of the abasic DNA in the co-crystal structure. Prior to docking, the bound abasic DNA and chain B of the APE1 monomer were removed, and proper protonation states were assigned for the acidic and basic residues. Docking simulations with standard parameters and GOLD fitness score calculations were performed as described above. Based on the GOLD fitness scores for each molecule, a bound conformation with high fitness score was considered as the best bound-conformation.

Glide
Compounds 6-54 from Tables 1-5 were processed using the Schrödinger LigPrep utility (Schrödinger, LLC, USA). This program produces a number of low energy 3D structures, with various ionization states, tautomers, stereochemistries, and ring conformations, from each molecule input. For this study, a pH range of 6-8 was used. Both the neutral and the anionic states of carboxylic acid groups were generated. Three-dimensional coordinates of the crystal structures of HIV-1 IN catalytic domain and human APE1 endonuclease, with bound abasic DNA and Mn 2+ ion, were obtained from the RCSB PDB (PDB ID: 1BL3 and 1DE9). All the water molecules present in the protein were removed, and hydrogen atoms were added considering the appropriate ionization states for both the acidic and basic amino acid residues. All compounds (6-54) were docked onto the binding site in HIV-1 IN or human APE1 protein using version 5.0 of Glide (Grid-Based Ligand Docking With Energetics) software from Schrödinger [37][38][39]. A grid with a bounding box size of 20 Å was used. The coordinates of the enclosing box (1BL3: x = 22.23 Å; y = 1.33 Å; z = −19.16 Å and 1DE9: x = 47.866; y = 9.534; z = 40.274) were defined starting from the set of active site residues involved. The Glide algorithm is based on a systematic search of positions, orientations, and conformations of the ligand in the receptor binding site using funnel type approach. The search begins with a rough positioning and scoring phase. This significantly limits the search space, and reduces the number of poses to be selected for minimization on the precomputed OPLS-2005 Van der Waals and electrostatic grids for the protein. In order to obtain an accurate correlation between good poses and good scores, the Glide Extra-Precision (XP) Mode was used.

Drug-Like Property Prediction
The ligand structures were energy minimized using the Catalyst software package (Accelrys, Inc.) [35] and the lowest-energy conformation of each compound was exported to ADMET Predictor (Simulations Plus, Inc., Lancaster, CA, USA) [40] to calculate various ADME (absorption, distribution, metabolism and excretion) properties.

Cell Lines and Chemicals
Human melanoma cell line MDA-MB 435, pancreatic cancer cell line Panc-1, and colon cancer cell line HT 29 were purchased from the American Type Cell Culture (Manassas, VA, USA). HCT 116 p53 +/+ and HCT 116 p53 −/− were kindly provided by Dr. Bert Vogelstein, Johns Hopkins Medical Institutions (Baltimore, MD, USA). HCT 116 p53 +/+ , HCT116 p53 −/− , and HT 29 cells were maintained as monolayer cultures in RPMI 1640 supplemented with 10% fetal bovine serum (FBS; Gemini-Bioproducts, Woodland, CA, USA) and 2 mmol/L L-glutamine at 37 °C in a humidified atmosphere of 5% CO 2 . Similarly, MDA-MB 435 and Panc-1 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum as described above. To remove the adherent cells from the flask for passaging and counting, cells were washed with PBS without calcium or magnesium, incubated with a small volume of 0.25% trypsin-EDTA solution (Sigma-Aldrich, St. Louis, MO, USA) for 5 to 10 minutes and washed with culture medium and centrifuged. All experiments were done using cells in exponential cell growth.
Compounds showing at least 50% growth inhibition at 10 μM were selected for IC 50 determinations.
After 72 h exposure to the test compounds, MTT solution (5 mg/mL; 20 μL) was added to each well and cells were incubated for 4 h at 37 °C. After incubation, media from each well was removed and the dark blue formazan crystals formed by live cells were dissolved in DMSO (150 μL/well). The absorbance intensity was measured at 570 nm against appropriate blank controls using a microplate reader (Molecular Devices, Sunnyvale, CA, USA). All assays were done in triplicate. The growth inhibitory concentration GI 50 was determined by plotting the logarithm of compound concentration against percentage of cells killed to obtain the concentration that produced 50% cell kill.

Conclusions
In this study, we have investigated the SAR of rhodanine-based IN inhibitors by modifying the aryl and alkylidene sub-groups. Several of these compounds showed moderate to potent IN inhibition and appeared to be more selective for IN as compared to APE1. Results from this study confirm the importance of the rhodanine moiety as a useful structural scaffold with promising IN inhibitory activity and provides a framework for designing more potent and selective IN inhibitors. In addition, some of these compounds inhibited APE1 endonuclease activity and also showed antiproliferative effects against cancer cell lines; thus making rhodanine-based derivatives suitable leads for antiviral and anticancer drug development.