Sage Journals: Discover world-class research

Abstract

Background:

HIV-1 integrase is a clinically validated therapeutic target for the treatment of HIV-1 infection, with one approved therapeutic currently on the market. This enzyme represents an attractive target for the development of new inhibitors to HIV-1 that are effective against the current resistance mutations.

Methods:

A fragment-based screening method employing surface plasmon resonance and NMR was initially used to detect interactions between integrase and fragments. The binding sites of the fragments were elucidated by crystallography and the structural information used to design and synthesize improved ligands.

Results:

The location of binding of fragments to the catalytic core of integrase was found to be in a previously undescribed binding site, adjacent to the mobile loop. Enzyme assays confirmed that formation of enzyme–fragment complexes inhibits the catalytic activity of integrase and the structural data was utilized to further develop these fragments into more potent novel enzyme inhibitors.

Conclusions:

We have defined a new site in integrase as a valid region for the structure-based design of allosteric integrase inhibitors. Using a structure-based design process we have improved the activity of the initial fragments 45-fold.

Introduction

Integration of viral DNA into the host cell genome is important for viral replication and is achieved by the two steps of 3′-processing (3′P) that involve excision of 3′-GT bases and strand transfer (ST), both carried out by HIV-1 integrase. These reactions occur in a complex that has been postulated to resemble a Holliday junction complex [1] containing integrase and components of the pre-integration complex, including the lens epithelium-derived growth factor (LEDGF) protein (which acts to anchor the complex to chromatin), the viral DNA termini and the host DNA. Enzymatic activity of integrase requires two magnesium ions to be coordinated by a DDE motif (D64, D116, E152) in the active site of integrase [2], with the position of E152 modulated by the ‘mobile loop’ (residues G140 to G149), suggesting that the loop plays a role in positioning the metals. Furthermore, mutagenesis studies have demonstrated that the mobility of this loop is important for catalytic activity [3] and different conformations of the loop and active site have been observed in model systems with and without integrase inhibitors [4].

Raltegravir, launched in 2007, was the first inhibitor to validate integrase as a therapeutic target [5]. The metal binding pyrimidinone scaffold of raltegravir was initially shown to bind to the magnesium ions present in hepatitis C NS5b polymerase, and subsequently optimized for binding to the magnesium ions bound to the catalytic triad DDE of integrase [6]. Both in vitro studies and clinical use of raltegravir have given rise to mutations in the integrase protein [7], in particular the residues Q148R/H, G140S/N/H/E/Q and Y143R, all of which are within the mobile loop. Recent structures of the integrase protein from the foamy virus (PFV) [4], including one with the protein complexed with two magnesium ions, viral DNA and raltegravir (PDB code 3L2T) show an interaction between Y143 (Y212 in PFV) and the inhibitor, which is significant because Y143R is a primary resistance mutation seen in response to raltegravir [8,9]. The Pommier group carried out a convincing study [10] of variants of the 3′-CA overhang which implicated a direct interaction of the Q148 residue with the cytosine base during the catalytic process. Detailed kinetic studies by Langley et al. [11] led to a model in which the adenine opens or closes the active site for DNA processing. Mutation of Q148 to R or H would interfere with these actions and has recently been shown to interfere with both 3′P and ST reactions [12]. NMR studies of the loop have indicated stable clusters of conformations even in the absence of the DNA ligand, suggesting functionally favoured orientations [13]. Modelling studies [14] of the mobile loop suggest that the G140S/Q148H double mutant blocks raltegravir binding to the active site directly by reducing the number of possible docking orientations. Taken together, these observations indicate a role for structural re-arrangement of the integrase complex and more specifically the requirement for mobility of mobile loop residues, including Q148. The requirement for mobility of this loop and the numerous reported crystal structures represents a potential target for the design of allosteric inhibitors. Structure-based design of integrase inhibitors has been complicated by the lack of crystal structures of small ligands bound to the active site of the isolated protein or the complex. The only known active site complex of HIV-1 integrase with a ligand is of 1-(5-chloroindol-3-yl)-3-hydroxy-3-(2H tetrazol-5-yl)-propenone [15] and this is of limited descriptive power because the ligand's orientation is influenced by a symmetry-related protomer in the crystal. For the avian sarcoma virus (ASV), allosteric inhibitors of its integrase protein, compounds 1 and 2 (Figure 1), have been reported [16].

Figure 1.

Structures of compounds 1, 2 and 3

A fragment-based screen of HIV-1 integrase by surface plasmon resonance (SPR) and saturation transfer difference NMR identified a number of small molecule ligands that bind to one or more sites on the integrase catalytic core domain (CCD; DIR, unpublished observations). By crystallography we determined some of these to be at a site adjacent to the mobile loop. We used these as a basis for structure-based design of novel allosteric inhibitors of HIV-1 integrase with the potential to overcome resistance to raltegravir binding.

Methods

Protein preparation

N-terminally hexa-His tagged CCD integrase (residues 50 to 210) containing the mutations C56S, F139D and F185H (core3H) was cloned into the E. coli expression vector pET28b(+) (Novagen, Madison, WI, USA) and expressed and purified essentially as described for core4H in Wielens et al. [17], however the current studies retained the His tag.

3′-processing strand transfer combined assay An assay procedure similar to that already published [18,19] was used to determine the inhibitory effect of the compounds. The assay was adapted to a 96-well plate format. Briefly, 400 ng integrase was incubated with 30 nM substrate DNA, consisting of annealed U5 LTR sequence oligonucleotides containing a digoxigenin (DIG) tag at the 3′-end (5′-ACTGCTAGAG ATTTTCCACACTGACTAAAAGGGTC-DIG-3′) or biotin (Bio; 5′-Bio-GACCCTTTTAGTCAGTGTGGA AAATCTCTAGCAGT-3′) so that each substrate has either a DIG or Bio tag on opposite strands. Reactions were carried out for 2 h at 37°C and products generated as a result of 3′P and ST activity were bound to streptavidin plates and, following denaturation, were detected using an anti-DIG-alkaline phosphatase conjugate (Roche Diagnostics, Mannheim, Germany) and p-nitro phenyl phosphate substrate (Sigma–Aldrich, St Louis, MO, USA). In this assay system raltegravir (Isentress^TM) returns a 50% inhibitory concentration (IC₅₀) of 0.13 μM (n=8) which compares to the Merck ST only assay system giving a published value of IC₅₀ 15 nM [5]. The second control compound, the diketo acid (L731,988) [20] gives an IC₅₀ of 0.18 μM.

Chemistry

In an SPR, NMR and crystallography screening programme the N-benzyl indolinone 3 (Figure 2) was discovered to form a complex with integrase. Analogues 4–23 were synthesized to probe the structure–activity relationship. Three series of structures were investigated, N-benzyl isatins, N-benzyl indolin-2-ones and 3-benzyl indolin-2-ones.

Figure 2.

Compounds 3–23 assayed in the 3′-processing strand transfer HIV-1 integrase assay and calculated LE

General procedure for N-benzyl isatin

Indoline-2,3-dione (isatin), and its ring-substituted analogues, could be conveniently alkylated with the appropriate benzyl bromide, using piperidine as a base, in 60–70% yield. The resulting ester 25 underwent quantitative saponification to the corresponding acid 26 by treatment with lithium hydroxide (Figure 3). Compounds 14, 18, 20 and 21 (Figure 2) were prepared in this way.

Figure 3.

Alkylation of isatin core

General procedure for N-benzyl indolin-2-ones

An attempt of direct N-alkylation of indolin-2-one via deprotonation with lithium diisopropyl amide or sodium hydride failed, giving instead dialkylation of the 3-methylene group of the indoline. We therefore used an alternative route via the isatin derivatives previously prepared. Wolff–Kishner reduction with hydrazine produced the required N-benzyl indolin-2-ones, usually in good yield. In some cases, most notably compound 3, this method gave poor yields. An alternative procedure involved chlorination of the isatin (in this case the benzodioxole ester rather than the carboxylic acid) with phosphorus pentachloride in toluene to give the dichloro intermediate, which was subsequently reduced with zinc metal in acetic acid, with concomitant hydrolysis of the ester, to give the target N-benzyl indolin-2-ones (Figure 4). Compounds 3–9, 15 and 19 (Figure 2) were prepared in this way.

Figure 4.

General routes to N-benzyl indolin-2-ones

General procedure for 3-benzyl indolin-2-ones

Compounds 10–13, 16 and 17 (Figure 2) were prepared by initial Knoevenagel reaction of the appropriate benzaldehyde with indolin-2-one, followed by hydrolysis of the ester with sodium hydroxide, and subsequent reduction of the exocyclic double bond by hydrogenation on Raney nickel or Pd/C to give the 3-benzyl indolin-2-ones (Figure 5).

Figure 5.

Synthesis of C-alkylated 3-benzyl indolin-2-ones

Ring-substituted isatins

To introduce acidic and basic functionality into the isatin ring we investigated metal-mediated coupling with 5-halosubstituted analogues. A palladium catalyzed Heck reaction of 20 with methyl acrylate was carried out using the method of Larhed and Hallberg [21] under microwave conditions. The resulting diester was hydrolyzed by potassium hydroxide in methanol to give 22 (Figure 6). The analogous iodo compound 21 underwent Suzuki coupling with the pinacol ester of 3-pyridine boronic acid to give a 5-pyrid-3-yl derivative, which underwent ester hydrolysis as previously described to yield compound 23.

Figure 6.

Synthesis of compounds 22 and 23

Analytical HPLC was performed on a Waters Alliance 2690 system (Waters, Milford, MA, USA) with 996 PDA and Phenomenex PFP Luna PFP 2×150 mm 5 μ, 100 Å column (Phenomenex, Torrance, CA, USA), LCMS (compounds 19–23) was performed on a Shimadzu LCMS-2010EV system (Shimadzu, Kyoto, Japan) with the same Luna PFP column. Semi-prep HPLC (compounds 22 and 23) was performed on a Waters prep 600 system with 2489 detector (Waters) at 254 nM, using 30 ml/min flow rate and 1–60% MeCN/0.1% formic acid. All HRMS were carried out on a ThermoQuest MAT95XP (ThermoQuest, Waltham, MA, USA) employing electron impact (EI) at 70 eV, and employing perfluorokerosene as a reference sample. TLC were carried out on Merck precoated silica gel 60 F₂₅₄ (Merck, Darmstadt, Germany). Representative and key synthetic procedures are described below.

5-((2-Oxindolin-1-yl)methyl)benzo[d] [1,3] dioxole-4-carboxylic acid 3

To a stirred mixture of isatin (52 mg, 0.35 mmol), K₂CO₃ (73 mg, 0.53 mmol) and KI (10 mg) in DMF (4 ml) was added ethyl 5-(bromomethyl)benzo[d] [1,3] dioxole-4-carboxylate [22] (110 mg, 0.34 mmol). The mixture was heated to 80°C for 1 h, after which H₂ O (40 ml) was added and the mixture extracted with EtOAc (2×50 ml). The combined extract was washed with saturated brine and dried (Na₂SO₄), filtered then concentrated under reduced pressure to give the crude product, ethyl 5-((2,3-dioxoindolin-1-yl)methyl) benzo[d] [1,3] dioxole-4-carboxylate, which was purified by flash chromatography (eluant: EtOAc/Pet. ether:1/2, v/v, silica gel, Rf 0.40) to give the intermediate in 82% yield. ¹H-NMR: 1.45 (t, 3H, J=7.2 Hz), 4.60 (q, 2H, J=7.2 Hz), 5.22 (s, 2H), 6.10 (s, 2H), 6.72 (d, 1H, J=8.0 Hz), 6.80–6.84 (m, 2H), 7.13 (t, 1H, J=8.0 H), 7.50 (td, 1H, J=8.0, 1.2 Hz), 7.65 (dd, 1H, J=8.0, 1.2 Hz).

To a stirred mixture of ethyl 5-((2,3-dioxindolin-1-yl)methyl)benzo[d] [1,3] dioxole-4-carboxylate (50 mg, 0.14 mmol) in benzene (1 ml) was added PCl₅ (67 mg, 0.32 mmol). The mixture was stirred for 24 h at room temperature, then water (10 ml) was added and the mixture was extracted with EtOAc (2×10 ml). The combined organic extract was washed with saturated brine, dried (Na₂SO₄), filtered and concentrated under reduced pressure to give a solid, which was purified by preparative TLC to give ethyl 5-((3,3-dichloro-2-oxindolin-1-yl)methyl)benzo[d] [1,3] dioxole-4-carboxylate (26 mg, yield 45%) as a light yellow solid (petroleum ether/EtOAc:4/1, v/v, Rf=0.40). To a stirred mixture of ethyl 5-((3,3-dichloro-2-oxindolin-1-yl)methyl)benzo[d] [1,3] dioxole-4-carboxylate (20 mg, 0.05 mmol) in AcOH (1 ml) was added zinc powder (32 mg, 0.5 mmol) The mixture was stirred for a further 5 min at room temperature after which TLC analysis indicated the starting material had been consumed. The zinc powder was filtered off and the filter cake was washed with EtOAc. Water (10 ml) was added into the filtrate and the mixture was extracted with EtOAc (2×10 ml). The organic layers were combined, washed with saturated NaHCO₃, dried (Na₂SO₄) and concentrated under reduced pressure to give ethyl 5-((2-oxindolin-1-yl)methyl)benzo[d] [1,3] dioxole-4-carboxylate (17 mg, yield 100%) as a light yellow solid (petroleum ether/EtOAc: 6/1, v/v, Rf=0.20). The ester (4.0 mg, 0.01 mmol) was taken up in AcOH (0.5 ml) and HCl (6 M, 0.5 ml) and the mixture was heated to reflux. After 5 h the reaction mixture was neutralized with 1 M NaOH and extracted with EtOAc (2×10 ml). The combined organic extract was washed with saturated brine, dried (Na₂SO₄) and concentrated under reduced pressure to give 5-((2-oxoindolin-1-yl) methyl)benzo[d] [1,3] dioxole-4-carboxylic acid 3. δ_H (400 MHz, CDCl₃ + CD₃OD) 3.50 (s, 2H), 5.11 (s, 2H), 5.96 (s, 2H), 6.61 (d, 1H, J=8.0 Hz), 6.68 (d, 1H, J=8.0 Hz), 6.78 (d, 1H, J=8.0 Hz), 7.02 (t, 1H, J=7.2 Hz), 7.41 (t, 1H, J=7.2 Hz), 7.51 (d, 1H, J=7.2 Hz). δ_C (100 Mhz, CDCL₃/DMSO-D6) 175.3, 166.7, 147.6, 144.2, 128.4, 127.8, 1248, 124.6, 122.4, 119.3, 115.0, 110.0, 109.1, 102.0, 41.1, 40.1, 39.9, 39.7, 39.5, 39.3, 39.1, 38.9, 35.6. m/z (ESI) C₁₇H₁₃NO₅ requires 311.0788, found 311.0792.

3-Hydroxy-2-methoxy-6-((2-oxindolin-3-yl)methyl) benzoic acid 12

To a mixture of ethyl 6-formyl-2-methoxy-3-(methoxymethoxy)benzoate (450 mg, 1.68 mmol) and indolin-2-one (436 mg, 3.28 mmol) in EtOH (10 ml) was added piperidine (2 drops) under an atmosphere of N₂. The mixture was heated to reflux overnight and then allowed to cool forming a precipitate. The condensation product, ethyl 2-methoxy-3-(methoxymethoxy)-6-((2-oxindolin-3-ylidene) methyl)benzoate, was obtained (517 mg, 45.3%) as a yellow solid by recrystallization from EtOH. A sample of this material (130 mg, 0.34 mmol) in MeOH (3 ml) was treated with 2 M NaOH (0.42 ml, 0.84 mmol) at 40°C for 2 days and then the mixture was concentrated. The residue was dissolved in EtOAc (5 ml) and washed by water (5 ml). The aqueous phase was separated and adjusted to pH 4–5, and then extracted with EtOAc (2×5 ml). The combined organic extracts were washed with brine, dried (Na₂SO₄), filtered and concentrated to give 3-hydroxy-2-methoxy-6-((2-oxindolin-3-ylidene)methyl)benzoic acid (76 mg, 63.3%, crude) as a yellow solid (TLC: DCM/MeOH:10:1, v/v, Rf=0.2). To the acid (70 mg, 0.22 mmol) was added Pd/C (10 mg) in MeOH and the reaction mixture was stirred at room temperature under H₂ overnight. Pd/C was removed by filtration (celite) and the filtrate was concentrated to give the crude product. It was purified by preparative TLC to obtain title compound 3-hydroxy-2-methoxy-6-((2-oxindolin3-yl)methyl)benzoic acid 12 (23 mg, 32.8%) as a yellow solid. (DCM/MeOH:10:1, v/v, Rf=0.2). δ_H (400 MHz, CDCl₃): 3.25–3.26 (t, 1H), 3.27–3.28 (d, 1H), 3.64 (s, 3H), 6.61 (s, 1H), 6.71–6.72 (d, 1H), 6.85–6.86 (d, 2H), 7.07–7.08 (d, 1H), 7.26–7.29 (m, 2H). LC-MS:312 (M-1). m/z (ESI) C₁₇H₁₅NO₅ requires 313.0915, found 313.0919.

(E)-6-((5-(2-Carboxyvinyl)-2,3-dioxoindolin-1-yl)methyl)-2,3-dihydrobenzo[b] [1,4]dioxin-5-carboxylic acid 22

Ethyl 6-((5-bromo-2,3-dioxoindolin-1-yl)methyl)-2,3-dihydrobenzo[b] [1,4] dioxin-5-carboxylate was prepared using the method described in step one for compound 3. To a sample of this material (46 mg, 0.11 mmol) in DMF (0.5 ml) was added tributylamine (20 mg, 0.11 mmol), palladium(II)acetate (10 mol%) and methyl acrylate (12 mg, 0.14 mmol). The reaction mixture was heated in a CEM-Discover^TM microwave reactor (CEM Corporation, Matthews, CA, USA), with stirring, for 4 min at 60 Watts constant power. The reaction mixture was then cooled and diluted with EtOAc (5 ml), filtered (celite), washed with 0.1 M HCl (5 ml), dried (MgSO₄), filtered and purified by column chromatography to give ethyl (E)-6-((5-(3-methoxy-3-oxoprop-1-en-1-yl)-2,3-dioxindolin-1-yl)methyl)-2,3-dihydrobenzo[b] [1,4] dioxin-5-carboxylate. δ_H (400 MHz, CDCl3): 7.78 (1H, m), 7.62 (2H, M), 7.56 (1H, s), 6.93 (1H, d, J=8.2 Hz), 6.87 (1H, d, J=8.7 Hz), 6.73 (1H, d, J=8.6 Hz), 6.35 (1H, d, J=16.0 Hz), 4.90 (2H, s), 4.42 (2H, q, J=7.15, 7.33, 6.91 Hz) 4.29 (4H, m), 3.81 (3H, s), 1.26 (3H, t, J=7.21, 7.33 Hz). The diester was then dissolved in MeOH (5 ml) with 2 M KOH (2 ml) and the mixture was heated at 80°C for 5 h. The solution was neutralized by careful addition of 1 M HCl and extracted by EtOAc (2×20 ml). The organic extracts were dried (Na₂SO₄), filtered, and concentrated under reduced pressure. Purification of the residue by semi-preparative HPLC gave the title product 22 eluting at 32–34% MeCN/H₂O, δ_H (400 MHz, CDCl3): 12 (1H, br), 8.17 (1H, s), 7.70 (1H, M), 7.35 (1H, m), 6.90 (3H, m), 6.39 (1H, d, J=16.03 Hz), 4.3 (4H, m). HPLC Rt 2.0 min. LCMS 410 (M+H⁺).

6-((2,3-Dioxo-5-(pyridin-3-yl)indolin-1-yl)methyl)-2,3-dihydrobenzo[b] [1,4] dioxine-5-carboxylic acid 23

Ethyl 6-((5-iodo-2,3-dioxoindolin-1-yl)methyl)-2,3-dihydrobenzo[b] [1,4] dioxine-5-carboxylate was prepared using the method described in step one for compound 3. To a sample of this material (25.3 mg, 0.05 mmol) in DMF and H₂O (3:1, 1 ml) was added 3-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)pyridine (10.5 mg, 0.05 mmol) and potassium carbonate (7 mg) and the solution degassed under reduced pressure for 5 min. Tetrakis(triphenylphosphine)palladium(0) was added and the reaction heated for 15 min in the microwave at the constant temperature of 100°C. The reaction mixture was diluted with EtOAc (5 ml), filtered (celite), washed with saturated NaHCO₃ (5 ml), dried (MgSO₄), filtered and purified on semi-preparative HPLC eluting at 38–40% MeCN. The combined fractions were freeze dried to give ethyl 6-((2,3-dioxo-5-(pyridin-3-yl)indolin-1-yl)methyl)-2,3-dihydrobenzo[b] [1,4] dioxine-5-carboxylate (19 mg, 84%). LCMS 445 (M+H⁺). A sample of this material (15 mg) was dissolved in MeOH (5 ml) with 2 M KOH (2 ml) and the mixture was heated at 80°C for 5 h. The mixture was evaporated to dryness and the residue purified by semi-preparative reverse phase chromatography, eluting at 23% MeCN and the combined fractions were freeze dried to give the title compound 23 (7 mg, 50%). HPLC Rt 1.5 min. LCMS 417 (M+H⁺) δ_C (100 MHz, DMSO-D6) 178.1, 172.2, 150.7, 147.4, 142.7, 129.0, 127.6, 126.6, 124.5, 121.9, 120.9, 117.3, 114.4, 109.1, 55.6, 46.5, 40.1, 39.9, 39.7, 39.5, 39.3, 39.1, 38.9, 34.9. m/z (ESI) 417 (M+H⁺), C₂₃H₁₆N₂O₆ requires 416.1003, found 416.0987.

Crystallography

For crystallization, the protein sample was concentrated to 5.5 mg/ml and exchanged into 40 mM Tris-Cl pH 8.0, 250 mM sodium chloride, 30 mM magnesium chloride and 5 mM DTT. Crystals were generated at the Bio21 Collaborative Crystallization Centre (C3; Parkville, Australia), with initial screening focused around previously published conditions. All drops were set up in SD-2 (IDEX Corp, Lake Forest, CA, USA) sitting drop plates using a Phoenix robot (Art Robbins Industries, Sunnyvale, CA, USA) with 50 μl of crystallant in the reservoir and droplets consisting of 200 nl of the reservoir and 200 nl of the protein sample. For large-scale production, the final conditions were: 100 mM sodium acetate at pH 5.0–5.5 and 1.6–2.0 M ammonium sulfate. A cryosolution was made up of 100 mM sodium acetate pH 5.5, 1.75 M ammonium sulfate, 25% ethylene glycol and 5% compound in neat DMSO. Twenty-four h prior to data collection 1.5 μl of the cryosolution was added to the crystal containing droplet and the crystallization plate was resealed. When compounds as dry powders were used for soaking, the cryosolution was the same as above but without compound, and the compound was ‘sprinkled’ on top of the drop after the addition of the cryosolution. For data collection, the crystal was gently looped out using MicroLoops (MiTeGen, Ithica, NY, USA) and cryo-cooled in the nitrogen stream. All data collection was done at 100 K. Typically 181 frames of data were collected with an oscillation angle of 1 degree. All data were indexed with Mosflm [23] scaled with SCALA (CCP4) [24] and molecular replacement was done using Phaser (CCP4) [24]. Manual rebuilding was done with the molecular graphics programme Coot [25] and the compounds were initially fitted into the density using the programme Afitt [26]. Subsequent refinement was done using Refmac [27]. All figures were prepared using Afitt [26], except Figure 7, which was made using the programme PyMol (Schrödinger, New York, NY, USA). Refined and validated structure data are reported in Table 1.

Table 1.

Structure refinement and model validation

	Compound ID (PDB structure code)
Factor	3 (3FN6)	4 (3FN7)	5 (3FN8)	8 (3FN9)	20 (3FNA)
Resolution range, Å	31.5–1.90	61.8–1.8	62.0–1.9	45.3–1.95	18.9–1.95
Number of reflections used in refinement	28,583	33,527	28,736	26,262	26,422
Final overall R factor (Rfree)	17.9 (20.8)	16.2 (20.7)	16.9 (19.8)	16.4 (20.8)	16.2 (20.4)
Number of reflections for R_free	1,510	1,768	1,532	1,401	1,407
Overall average B factor, Å²	27.5	26.2	31.2	35.6	35.4
Total number of atoms	2,673	2,800	2,737	2,669	2,724
Number of solvent atoms	112	169	102	89	107
Number of ligand atoms	92	48	150	48	104
Ramachandran: most favoured regions, %	98.9	99.6	97.3	99.2	98.8
Ramachandran: additionally allowed regions, %	1.1	0.4	2.7	0.8	1.2

PDB, Protein Data Bank.

Figure 7.

Modelling of compunds 22 and 23 based on the crystal complex of ligand 4

Structures of the compounds in complex with integrase have been deposited at the Protein Data Bank (PDB) with codes 3NF6, 3NF7, 3NF8, 3NF9 and 3NFA for compounds 3, 4, 5, 8 and 20 respectively.

Modelling studies

Compounds 22 and 23 were detected in the assay, yet were not visualized in the crystal structures. In order to estimate their potential bound conformations, these compounds were docked into the allosteric site using the crystal structure of the complex containing compound 4 and core3H as a guide. The cis- and trans- isomers of compound 22, as well as compound 23 were prepared by normalizing protonation states for pH 7.4, assigning partial atomic charges using the AM1 BCC Hamiltonian [28], and enumerating conformers using the Omega conformer generation programme [29]. Docking was performed using an in-house software package (DL, unpublished data). This software uses the attached-point statistical description of a receptor pocket [30] and aligns input ligand conformations using the OEShape toolkit [31,32]. Conformations that did not clash with the receptor atoms were scored using a number of standard scoring functions including DOCK [33], PMF [34], SMoG [35], ChemScore [36], SCORE [37] and AutoDock [38]. A consensus score was generated for each docked pose from a product of individual ranks. All docked poses were minimized in the MMFF94s force field [39]. The top ranking conformation with greatest overlap to the known crystallographic ligand 4 is shown in Figure 7A for cis- and trans-isomers of compound 22 as well as compound 23 Figure 7B.

Following the initial docking, 60 ns of molecular dynamics was performed for both stereo isomers of compound 22, compound 23 and compound 4 (as a control because it was observed in crystallographic experiments). Molecular dynamics simulations were performed in explicit TIP3P solvent [40], using the GROMACS software package [41] with a time step of 2 fs. The Amber99sb force field [42] was used to model the receptor and GAFF [43] was used to model the ligands. Following an initial 2 ns equilibration at constant temperature and pressure, the simulations were performed in the NVT (number of particles, volume, temperature) ensemble using the method of Bussi et al. [44] for temperature coupling. Electrostatic interactions were treated with the particle-mesh Ewald method. All backbone atoms were restrained to their crystallographic positions, as were all non-backbone heavy atoms farther than 6 Å from the binding site.

Results

Several constructs of the integrase CCD were made and tested for their ability to yield diffraction quality crystals. These constructs were made with two goals in mind, to solubilize the protein and to keep its activity intact. Two constructs, core3H and core4H (core3H containing C56S, F139D and F185H mutations [15] and core4H containing a W131D mutation in addition to these three), both containing a hexa-his tag at the N-terminus, were found to crystallize. During optimization of the crystallization process for integrase CCD, we observed that the core4H crystallized in a space group (P2₁2₁2₁) where the active site is occluded by symmetry-related molecules. However, the core3H construct crystallized in a different space group (P3₁), with an accessible active site and relatively ordered mobile loop.

A complex of compound 3 and the core3H protein showed good density for residues 140–149 and the DDE motif (Figure 8A). The 3′ST assay (a composite of two biochemical processes, the 3′P and ST steps) of compound 3 gave an IC₅₀ of 259 μM, equating to a ligand efficiency (LE) [45] of 0.21. The crystallographic data indicated that although compound 3 was observed in a single position in the complex with the protein in the crystal, as seen by the close fit of the Fourier density map to the compound (Figure 8B), no hydrogen bonding interactions were made, suggesting that significant optimization was possible. Analogues of compound 3 also showed partial density in the LEDGF site (TSP, unpublished observations) and complete density in the newly identified site. In the absence of the LEDGF co-factor in a 3′ST assay with full length protein, inhibition was still observed, indicating inhibition of LEDGF binding is not the primary mechanism of inhibition by these compounds. Furthermore, the compounds did not inhibit the enhancement of the 3′ST activity when LEDGF is added [46] and additional testing of these compounds in an interaction assay [47] between LEDGF and integrase confirmed they did not inhibit LEDGF binding (NV, data not shown). Taken together, these assays results are consistent with the allosteric inhibition of integrase activity by these compounds through interactions with the newly identified pocket.

Figure 8.

HIV-1 integrase in complex with compound 3

Structural information for full length HIV-1 integrase is not available and we have not found the full length protein to be amenable to the SPR and NMR studies conducted here with the CCD constructs. Therefore, in order to determine whether this newly identified pocket is likely to be present in full length integrases we overlayed the structures of the protein and ligands generated in this study with the only full length structure available, that of the PFV integrases [4]. As anticipated, a high degree of structural homology was observed together with the existence of the pocket in the full length protein with space to accommodate the binding of ligands (TSP, data not shown). The observation of high structural homology and the data from the 3′ST assays showing ligands that bind to this pocket in the CCD inhibit the activity of full length enzyme is consistent with the pocket being present in the full length HIV-1 integrase.

A molecule, compound 1 (Figure 1), also known as Y3, has previously been identified [16] that binds adjacent to the mobile loop in the ASV integrase crystal and in an ST assay this compound had an IC₅₀ of 10.9 μM against HIV-1 integrase. The compound 1 (Y3) analogue, compound 2 (7-amino-3,8-dibromo-4-hydroxynapthalene-2-sulphonic acid, 2BrNSA; Figure 1), was subsequently shown to have an improved activity, IC₅₀ of 2.4 μM in a 3′P HIV-1 integrase assay, however, although biological assays suggested that 2BrNSa did not bind at the catalytic site of integrase, no binding mode was identified [48]. The complex of compound 3 with the HIV-1 integrase CCD identified in our study was compared (Figure 8C) with the complex of compound 1 (Y3) with ASV integrase (PDB-1A5X) [16]. When the two integrase constructs (HIV-1 to ASV) are compared, an alignment of 124 aligned residues with 5 gaps and a sequence identity of 23% is seen. A structural alignment of the two cores gave a root mean square deviation of 2.1 Å indicating a high degree of structural similarity despite the low sequence conservation and makes comparison of binding locations possible. In HIV-1 integrase, compound 3 is seen to bind in a pocket formed by protein segments V150→I161, V77→I84, V180 and H183. V150 (HIV-1) is within 2.7 Å of compound 3; the analogous residue of ASV (M155) forms a 3.1 Å hydrogen bond from the M155 backbone amide proton to the Y3 sulfate. The segment V150-I161 contains the E152 in the DDE motif and interactions with this loop could conceivably alter the positioning of E152 and effect enzyme activity. The π-nitrogen of H183 (HIV-1) is 2.9 Å from the O1 dioxolane oxygen of compound 3, and forms a favourable hydrogen bonding interaction. The side chains of M154 and I151 are within 3.8 Å and 3.5 Å of the six-membered ring of the indolinone of compound 3, respectively, forming a hydrophobic wall in the binding pocket (Figure 8D). These areas represent additional areas that can be targeted to improve the activity of compounds binding in this region. Q62 is conserved between the proteins and is also close to both compounds (Figure 8C). In the PDB-1A5X structure with Y3, the mobile loop residues from 148 to 153 are disordered, whereas in the compound 3 complex the loop shows sufficient order to be traced, with B-factors of 46–60 Å². To develop a structure-activity relationship around the lead compound 3, a series of analogues (4 to 23) were prepared (Figure 2).

As expected, crystallographic studies showed that analogues 4, 5, 8 and 15 bound in the same pocket as compound 3 (see Figure 8E for compound 5). Substitution of the indolin-2-one ring with a halogen (Figure 2; entries 4, 5 and 8) had little effect on activity, possibly because of decreased solubility. Incorporating a nitrogen at the 5-position gave improved inhibitory activity (Figure 2; entry 6) with an improved LE for compound 6 of 0.23. Inversion of the indolin-2-one ring to give attachment via the C-3 carbon (Figure 2; entries 10–12) did not improve binding, although compound 12, with the dioxolane ring opened, has better affinity than compound 10. Based on the result for compound 12, analogues 14–19 were made without the dioxolane group, either as N-linked compounds 18 and 19 or carbon-linked compounds 16 and 17. Compound 18 exhibits comparable LE to the limited compound 3. LE, as reviewed by Congreve et al. [45], is a value determined for each compound and is proportional to the logarithm of the IC₅₀ divided by the number of non-hydrogen atoms in the molecule. Comparing LE in a series helps distinguish where adding a functional group has improved IC₅₀ simply because of increased hydrophobic interactions with the receptor, generally decreasing LE, compared to a specific favourable interaction, which leads to improved LE. For example, compound 8 with 6-Cl has lower LE than compound 3 with 6-H (LE 0.20 compared to 0.21; Figure 2), suggesting that the change has not improved interactions of the compound with integrase despite the similar IC₅₀s of the two compounds (IC₅₀ 290 and 295 μM, for compounds 8 and 3, respectively; Figure 2).

Examination of the complexes of compounds 3, 4, 5, 8 and 15 indicated that the indolin-2-one ring was oriented with its 5-position toward the carboxamide of residue Q62. This might explain the improved binding of compounds 6 and 7 with a basic nitrogen oriented toward Q62, however the separation estimated from the position of 3 was not optimal (>5 Å), and Q62 was seen to occur in two conformations of equal occupancy.

In an attempt to target the Q62 residue for H-bonding interactions, and an anticipated improvement in compound IC₅₀, a further series of analogues were prepared with either acidic (22; Figure 2) or basic (23; Figure 2) groups substituted at position 5. It can be seen from Figure 2, that compound 22, with an extended acidic side chain, had 45-fold better affinity (determined by IC₅₀) than compound 3, suggesting that interactions with Q62 are indeed possible. These interactions are attractive in a molecule as Q62 is highly conserved in HIV-1 integrase sequences [8] and being close to the catalytic residue D64, it might have an effect on activity. Compounds 22 and 23 were detected in the assay, yet could not be observed in the structures of soaked crystals. To provide a reasonable estimate of the bound conformations of these compounds, they were modelled into the allosteric site (V150–I161, V77–I84 and V180 and H183) using the bound conformation of 4 as a guide. The cis- and trans-isomers of compound 22, as well as compound 23 were prepared [28] and the conformers were enumerated [29]. After docking (DL, unpublished data) of the conformers, those that did not clash with the integrase atoms were scored using a number of standard scoring functions [33 ,–38] to generate a consensus score for each docked pose from a product of individual ranks. All docked poses were minimized in the MMFF94s force field [39]. The top ranking conformation with greatest overlap to the known crystallographic ligand 4 is shown in Figure 7A for cis- and trans-isomers of compound 22 as well as compound 23 (Figure 7B). To validate the top ranking docked conformation of compounds 22 and 23, 60 ns of molecular dynamics simulations were performed on each as well as the crystallographic conformation of compound 4 (as a control). The molecular dynamics simulations of compound 4 agreed well with the crystallographic data, indicating that the chosen simulation parameters are reasonable for a qualitative assessment of the docked conformations of compounds 22 and 23 (Additional file 1).

Discussion

We initiated this study by firstly successfully identifying a construct, hexa-his tagged core3H, and crystallization conditions that enabled the generation of crystals with a defined loop and accessible active site suitable for use in identifying fragments binding to pockets in the HIV-1 integrase CCD. These crystals were then utilized with fragment hits from an SPR and NMR screening campaign to identify a new site in the CCD of HIV-1 integrase, where ligands act as allosteric inhibitors of the catalytic activity of the enzyme.

Crystallographic analysis and modelling of a series of compounds identified in this study has in more detail elucidated a pocket identified previously in ASV integrase and further extends this with a basis for inhibitor design, capitalizing on interactions with H183, Q62 and S147. We initially identified an isatin analogue bound in this pocket which is a member of a class of molecules with previously described biological activities such as antitumor and antiviral activities [49]. By using structure-based design we developed a compound, 22, showing significantly higher potency than the initial fragment with low micromolar affinity in a 3′ST enzyme activity assay. The docked model of the complex with compound 22 suggests the basis of the affinity of the compound arises from potential interactions with both H183 and Q62, as well as the side chain of S147. Unfortunately, this conformation was not favoured in the molecular dynamics simulations, indicating the possibility of more than one potential binding mode. Compound 23, however, did remain bound in the docked conformation throughout the duration of the simulations, indicating that it is a reasonable model for a potential binding mode of this compound. Whilst processing a bulky 3-pyridyl group, compound 23 does show improved affinity as compared to compound 7; however, this is at the loss of efficiency of binding (LE [45]) and might simply reflect increased hydrophobic contact. The confirmation of the binding modes as explored by this docking study validates the orientation observed in the crystal complex as a basis for design of higher affinity analogues. The pocket and approach identified would appear to represent an attractive target for the development of inhibitors of integrase and importantly also for integrase mutants that are resistant to the current generation of inhibitors. Finally, compound 22 has 45-fold better IC₅₀ than compound 3.

Footnotes

Acknowledgements

The work was supported by a commercial ready grant COMO4229 from the Commonwealth of Australia, Department of Innovation, Industry, Science and Research to Avexa Ltd. We thank the Australian Synchrotron and the beamline scientists at MX1 and MX2 for their help in data collection. We also thank the Bio21 Collaborative Crystallisation Centre (Parkville, Australia) for producing the protein crystals used in this study. In addition, we thank Roger Mulder and Jo Cosgriff for being more than accommodating in the timely collection and workup of the 13C NMR spectra and HRMS data. Compounds 3 to 17 were prepared by SYNthesis Pty Ltd.

The authors declare no competing interests.

Additional file 1: Simulation results testing docked conformations of compounds 22 and 23 using atomistic molecular dynamics can be found at

References

Hazuda

Hastings

Wolfe

Emni

. A novel assay for the DNA strand-transfer reaction of HIV-1 integrase. Nucleic Acids Res 1994;22:1121–1122.

Dyda

Hickman

Jenkins

Engelman

Craigie

Davies

. Crystal structure of the catalytic domain of HIV-1 integrase: Similarity to other polynucleotidyl transferases. Science 1994;266:1981–1986.

Greenwald

Butler

Bushman

Choe

. The mobility of an HIV-1 integrase active site loop is correlated with catalytic activity. Biochemistry 1999;38:8892–8898.

Hare

Gupta

Valkov

Engelman

Cherepanov

. Retroviral intasome assembly and inhibition of DNA strand transfer. Nature 2010;464:232–236.

Summa

Petrocchi

Bonelli

Discovery of raltegravir, a potent, selective orally bioavailable HIV-integrase inhibitor for the treatment of HIV–AIDS infection. J Med Chem 2008;51:5843–5855.

Summa

Petrochhi

Matassa

4,5-Dihydroxypyrimidine carboxamides and N-alkyl-5-hydroxy-pyrimidinone carboxamides are potent, selective HIV-integrase inhibitors with good pharmacokinetic profiles in preclinical species. J Med Chem 2006;49:6646–6649.

Reigadas

Anies

Masquelier

The HIV-1 integrase mutations Y143C/R are an alternative pathway for resistance to raltegravir and impact the enzyme functions. PLoS ONE 2010;5:e10311.

Ceccherini-Silberstein

Malet

D'Arrigo

Antinori

Marcelin

Perno

. Characterization and structural analysis of HIV-1 integrase conservation. AIDS Rev 2009;11:17–29.

Delelis

Thierry

Subra

Impact of Y143 HIV-1 integrase mutations on resistance to raltegravir in vitro and in vivo. Antimicrob Agents Chemother 2010;54:491–501.

10.

Johnson

Santos

Godwin Pais

GCG

Integration requires a specific interaction of the donor DNA terminal 5′-cytosine with glutamine 148 of the HIV-1 integrase flexible loop. J Biol Chem 2005;281:461–467.

11.

Langley

Samanta

Lin

Walker

Krystal

Dicker

. The terminal (catalytic) adenosine of the HIV LTR controls the kinetics of binding and dissociation of HIV integrase strand transfer inhibitors. Biochemistry 2008;47:13481–13488.

12.

Métifiot

Maddali

Naumova

Zhang

Marchand

Pommier

. Biochemical and pharmacological analyses of HIV-1 integrase flexible loop mutants resistant to raltegravir. Biochemistry 2010;49:3715–3722.

13.

Fitzkee

Masse

Shen

Davies

Bax

. Solution conformation and dynamics of the HIV-1 integrase core domain. J Biol Chem 2010;285:18072–18084.

14.

Perryman

Forli

Morris

A dynamic model of HIV integrase inhibition and drug resistance. J Mol Biol 2010;397:600–615.

15.

Goldgur

Dyda

Hickman

Jenkins

Craigie

Davies

. Three new structures of the core domain of HIV-1 integrase: An active site that binds magnesium. Proc Natl Acad Sci U S A 1998;95:9150–9154.

16.

Lubkowski

Yang

Alexandratos

Structure of the catalytic domain of avian sarcoma virus integrase with a bound HIV-1 integrase-targeted inhibitor. Proc Natl Acad Sci U S A 1998;95:4831–4836.

17.

Wielens

Headey

Jeevarajah

Crystal structure of the HIV-1 integrase core domain in complex with sucrose reveals details of an allosteric inhibitory binding site. FEBS Lett 2010;584:1455–1462.

18.

Debyser

Cherepanov

Pluymers

De Clercq

. Assays for the evaluation of HIV-1 integrase inhibitors. Methods Mol Biol 2001;160:139–155.

19.

Ovenden

Wan

Globoidnan A: A lignan from Eucalyptus globoidea inhibits HIV integrase. Phytochemistry 2004;65:3255–3259.

20.

Hazuda

Felock

Witmer

Inhibitors of strand transfer that prevent integration and inhibit HIV-1 replication in cells. Science 2000;287:646–650.

21.

Larhed

Hallberg

. Microwave-promoted palladium-catalyzed coupling reactions. J Org Chem 1996;61:9582–9584.

22.

Ollero

Castedo

Dominguez

. Synthesis of 3,4-dihydroisocoumarins from α-lithiated o-toluates. Synlett 1997; 1047–1048.

23.

Leslie

AGW

. Joint CCP4 + ESF-EAMCB newsletter on protein crystallography 1992, No. 26.

24.

Collaborative Computational Project Number 4. The CCP4 suite: Programs for protein crystallography. Acta Crystallogr D Biol Crystallogr 1994;50:760–763.

25.

Emsley

Cowtan

. Coot: Model-building tools for molecular graphics. Acta Crystallogr D Biol Crystallogr 2004;60:2126–2132.

26.

Wlodek

Skillman

Nicholls

. Automated ligand placement and refinement with a combined force field and shape potential. Acta Crystallogr D Biol Crystallogr 2006;62:741–749.

27.

Murshudov

Vagin

Dodson

. Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr D Biol Crystallogr 1997;53:240–255.

28.

Jakalian

Jack

Bayly

. Fast, efficient generation of high-quality atomic charges. AM1-BCC model: II. Parameterization and validation. J Comput Chem 2002;23:1623–1641.

29.

OpenEye Scientific Software, Inc. Omega, version 2.3.2. (Accessed 30 December 2010.) Available from www.eyesopen.com/omega.

30.

Moreno

Leon

. Geometric and chemical patterns of interaction in protein–ligand complexes and their application in docking. Proteins 2002;47:1–13.

31.

OpenEye Scientific Software, Inc. OEShape, version 1.7.2. (Accessed 30 December 2010.) Available from www.eyesopen.com/shape-tk.

32.

Grant

Gallardo

Pickup

. A fast method of molecular shape comparison: A simple application of a Gaussian description of molecular shape. J Comput Chem 1996;17:1653–1666.

33.

Ewing

TJA

Makino

Skillman

Kuntz

. DOCK 4.0: Search strategies for automated molecular docking of flexible molecule databases. J Comput Aided Mol Des 2001;15:411–428.

34.

Muegge

Martin

. A general and fast scoring function for protein–ligand interactions: A simplified potential approach. J Med Chem 1999;42:791–804.

35.

DeWitte

Shakhnovich

. SMoG: de novo design method based on simple, fast, and accurate free energy estimates. 1. Methodology and supporting evidence. J Am Chem Soc 1996;118:11733–11744.

36.

Eldridge

Murray

Auton

Paolini

Mee

. Empirical scoring functions: I. The development of a fast empirical scoring function to estimate the binding affinity of ligands in receptor complexes. J Comput Aided Mol Des 1997;11:425–445.

37.

Wang

Liu

Lai

Tang

. SCORE: A new empirical method for estimating the binding affinity of a protein–ligand complex. J Mol Model 1998;4:379–394.

38.

Morris

Goodsell

Halliday

Automated docking using a Lamarckian genetic algorithm and an empirical binding free energy function. J Comput Chem 1998;19:1639–1662.

39.

Halgren

. Merck molecular force field. II. MMFF94 van der Waals and electrostatic parameters for intermolecular interactions. J Comput Chem 1996;17:520–552.

40.

Jorgensen

Chandrasekhar

Madura

Impey

Klein

. Comparison of simple potential functions for simulating liquid water. J Chem Phys 1983;79:926–935.

41.

Hess

Kutzner

van der Spoel

Lindahl

. GROMACS 4: Algorithms for highly efficient, load-balanced, and scalable molecular simulation. J Chem Theory Comput 2008;4:435–447.

42.

Hornak

Abel

Okur

Comparison of multiple amber force fields and development of improved protein backbone parameters. Proteins 2006;65:712–725.

43.

Wang

Wolf

Caldwell

Kollman

Case

. Development and testing of a general amber force field. J Comput Chem 2004;25:1157–1174.

44.

Bussi

Donadio

Parrinello

. Canonical sampling through velocity rescaling. J Chem Phys 2007;126:014101–014107.

45.

Congreve

Chessari

Tisi

Woodhead

. Recent developments in fragment-based drug discovery. J Med Chem 2008;51:3661–3680.

46.

Christ

Voet

Marchand

Rational design of small molecule inhibitors of the LEDGF/75-integrase interaction and HIV replication. Nat Chem Biol 2010;6:442–448.

47.

Hou

McGuinness

Prongay

Screening for antiviral inhibitors of the HIV integrase-LEDGF/p75 interaction using the AlphaScreen luminescent proximity assay. J Biomol Screen 2008;13:406–414.

48.

Daniel

Myers

Kulkosky

Characterization of a naphthalene derivative inhibitor of retroviral integrases. AIDS Res Hum Retroviruses 2004;20:135–144.

49.

Medvedev

Buneeva

Glover

. Biological targets for isatin and its analogues: Implications for therapy. Biologics 2007;1:151–162.

Structural Basis for a New Mechanism of Inhibition of H I V-1 Integrase Identified by Fragment Screening and Structure-Based Design

Abstract

Background:

Methods:

Results:

Conclusions:

Introduction

Methods

Protein preparation

Chemistry

General procedure for N-benzyl isatin

General procedure for N-benzyl indolin-2-ones

General procedure for 3-benzyl indolin-2-ones

Ring-substituted isatins

5-((2-Oxindolin-1-yl)methyl)benzo[d] [1,3] dioxole-4-carboxylic acid 3

3-Hydroxy-2-methoxy-6-((2-oxindolin-3-yl)methyl) benzoic acid 12

(E)-6-((5-(2-Carboxyvinyl)-2,3-dioxoindolin-1-yl)methyl)-2,3-dihydrobenzo[b] [1,4]dioxin-5-carboxylic acid 22

6-((2,3-Dioxo-5-(pyridin-3-yl)indolin-1-yl)methyl)-2,3-dihydrobenzo[b] [1,4] dioxine-5-carboxylic acid 23

Crystallography

Modelling studies

Results

Discussion

Footnotes

Acknowledgements

References