Structural Basis of Mos1 Transposase Inhibition by the Anti-retroviral Drug Raltegravir

DNA transposases catalyze the movement of transposons around genomes by a cut-and-paste mechanism related to retroviral integration. Transposases and retroviral integrases share a common RNaseH-like domain with a catalytic DDE/D triad that coordinates the divalent cations required for DNA cleavage and integration. The anti-retroviral drugs Raltegravir and Elvitegravir inhibit integrases by displacing viral DNA ends from the catalytic metal ions. We demonstrate that Raltegravir, but not Elvitegravir, binds to Mos1 transposase in the presence of Mg2+ or Mn2+, without the requirement for transposon DNA, and inhibits transposon cleavage and DNA integration in biochemical assays. Crystal structures at 1.7 Å resolution show Raltegravir, in common with integrases, coordinating two Mg2+ or Mn2+ ions in the Mos1 active site. However, in the absence of transposon ends, the drug adopts an unusual, compact binding mode distinct from that observed in the active site of the prototype foamy virus integrase.

T ransposons and viruses are mobile genetic elements that survive and propagate by integrating into their hosts' genomes. DNA transposons are cut out from one genomic location and pasted into another by a DNA transposase, often encoded within the transposon sequence. This genetic rearrangement provides a driving force for genomic variation and evolution but can also generate genomic instability. Some transposons have become domesticated within their host's genome and provide useful new functions: for example the V(D)J recombination system, which generates antibody diversity, and the methyltransferase-DNA transposase fusion protein SETMAR involved in DNA repair. 1,2 The mechanism of DNA transposition is closely related to the integration of retroviruses, such as human immunodeficiency virus 1 (HIV-1). DNA transposases specifically recognize short inverted repeat (IR) sequences that mark the transposon ends. Excision of the transposon and its integration at a new genomic site is coordinated within a nucleoprotein complex, the transpososome, in which the two transposon ends are paired. Likewise, viral DNA ends contain long terminal repeat (LTR) sequences that are recognized specifically by a retroviral integrase and are brought together in a nucleoprotein complex, the intasome. The integrase cleaves two nucleotides from the reactive DNA strand before joining the processed viral ends irreversibly to the host's genome.
The mechanistic similarities of DNA transposases and retroviral integrases are reflected in common active site architectures and similar structural features. 3,4 The catalytic core domains of these enzymes adopt a RNase-H like fold 5 bringing together a triad of catalytic acidic amino acids: the DDE/D motif. The carboxylate oxygens coordinate the Mg 2+ or Mn 2+ ions required for DNA cleavage and integration. 6 A number of crystal structures of isolated catalytic core domains of DNA transposases and integrases have been determined: these include the active mariner family transposase Mos1 (from Drosophila mauritiana 7 ), the closely related mariner transposase catalytic domain of SETMAR (human 8 ), and the retroviral integrase HIV-1 9,10 (shown in Figure 1). Furthermore there are crystal structures of the Mos1, 11 Tn5, 12 and Mu 13 transpososomes and the Spumavirus Prototype Foamy Virus (PFV) intasome, 14 each of which contains the full length enzyme in a synaptic complex with two cognate DNA ends.
The Mos1 and human SETMAR mariner transposases show a higher degree of structural similarity compared with integrases ( Figure 1 and Supplementary Figure 1). The active sites of HIV-1 and PFV integrase contain DD-35-E motifs, whereas the mariner family DNA transposase Mos1 active site has a DD-34-D triad. The SETMAR mariner transposase catalytic domain has a DD-34-N motif, which supports DNA cleavage and integration, 15,16 and shares 38.7% sequence identity and 48.4% sequence similarity to Mos1. In all four enzymes the loop preceding the third catalytic residue contains conserved Tyr and Pro residues; these are Try276 and Pro278 in Mos1 Transposase (Tnp). In the Mos1 Tnp and SETMAR catalytic domain crystal structures, this loop is ordered due to its stabilizing interactions with the N-and C-terminal capping helices, α1 and α7 respectively ( Figure 1). As a result the active sites are fully structured without DNA. By contrast, in the crystal structure of the isolated HIV-1 integrase catalytic core domain, 10 the loop was disordered. NMR relaxation measurements indicated that loop residues are dynamic, moving between several distinct conformational clusters. 17 This is consistent with the proposal that the integrase active site does not adopt a well-defined conformation, capable of binding divalent metal ions and inhibitor, until the integrase has assembled on viral ends. 18 The DNA integration step of the retroviral life cycle has been targeted for the development of anti-retroviral therapies. Currently, several integrase strand transfer inhibitors (INSTIs) are available or in development for the treatment of HIV-1 infections, including Raltegravir 19 and Elvitegravir. 20 Attempts to crystallize the HIV-1 integrase with viral DNA ends have so far been unsuccessful, and surrogate models have been sought in order to better understand how INSTIs act. The Tn5 transposase was considered as a model system, because of the wealth of structural information on its interactions with inverted repeat DNA. Several diketoacid HIV-1 integrase inhibitors were identified using Tn5 transposase as the target in a chemical library screen; 21 these affected cleavage, synapsis, or integration steps of Tn5 transposition. 22 More recently, the PFV integrase, a closer relative of the HIV-1 enzyme, has provided an amenable model system. 23−25 Co-crystal structures of the PFV intasome with Raltegravir or Elvitegravir revealed that the drugs inhibit viral DNA integration by coordinating two divalent metal ions bound to the DDE motif carboxylates and displacing the reactive viral DNA end from the active site. 14,26 Recently it was shown that Raltegravir and Elvitegravir can also inhibit the nuclease activity of the transposase domain within the human fusion protein SETMAR. 27 We asked if Raltegravir and Elvitegravir could bind to and inhibit the DNA cleavage and integration activities of Mos1, 28       Crystals diffracted X-rays to 1.7 Å resolution (Table 1), and structures were determined by molecular replacement using the structure of the Mos1 catalytic domain (PDBID: 2F7T) as the model. The initial 2F o − F c map contained clear additional electron density near the active site into which Raltegravir was built. The structures were refined to a final R free of 24.1% and 24.2% for the structures containing Mg 2+ and Mn 2+ , respectively (Figure 3a−c, Table 1). We also collected diffraction data from crystals grown in MnCl 2 at the Mn Kedge (λ = 1.896 Å, Table 1), and the peaks in an anomalous difference map confirmed the positions of the two Mn 2+ ions in the active site (Figure 3b).
Raltegravir adopts a compact, curved conformation in the Mos1 active site (Figure 3a). The fluorobenzyl ring is oriented on the same side as the methyl-oxadiazole group and fills a hydrophobic pocket on Mos1 Tnp lined by Ala 251 and Pro 252. The six-membered fluorobenzyl ring tops a four-tiered aromatic ring stack, which also includes the five-membered oxadiazole moiety of Raltegravir and the side chains of Tyr 276 and His 122 of Mos1 Tnp (Figure 3a). In addition, the Raltegravir isopropyl group makes hydrophobic contacts with the conserved residue Pro 278. A similar "folded-over" binding mode was predicted for Raltegravir docked in the HIV-1 integrase active site in the absence of viral DNA, using computational methods. 30 Our crystal structures show that the carboxylate side chains of the active site aspartic acid triad coordinate two divalent metal ions (Figure 3a Rotation about the Raltegravir CBC−CBF bond also swaps the relative positions of the three metal-chelating oxygen atoms. However, their spatial arrangement is intact, enabling coordination of the two divalent metal ions in the active site of Mos1 transposase in a manner similar to that observed in the PFV intasome co-crystal structures, despite the drug adopting a distinctly different conformation. Thus, the rotational freedom within Raltegravir allows it to adopt a previously unappreciated bound conformation, highlighting that a drug can have distinct binding modes in subtly different molecular environments. The chemical structure of Elvitegravir (Figure 2c) is inherently less flexible than that of Raltegravir (Figure 2b), and there is a different separation of the three, rigidly coplanar chelating oxygen atoms. It also lacks the oxadiazole group that forms key stacking interactions in the Raltegravir-bound Mos1 structure. Taken together, these factors may explain our observation that Elvitegravir did not bind to Mos1 transposase.
Raltegravir Binds to the Mos1 Paired-End Complex. Mos1 transposase recognizes specific IR sequences at each transposon end and brings them together in a paired-end complex (PEC) for DNA cleavage, before inserting the cleaved ends into TA dinucleotide sites on target DNA. To test if Raltegravir could also interact with and inhibit Mos1 Tnp bound to IR DNA, we first performed thermal denaturation assays using the Mos1 PEC. This complex was prepared using 'precleaved' IR DNA substrates as before. 32 Similarly to Mos1 Tnp, the Mos1 PEC was stabilized by divalent metal ions (Figure 5a Figure  5e).
Next, we tested the effect of Raltegravir on Mos1 strand transfer by monitoring the integration of fluorescently labeled, precleaved transposon IR DNA substrates into a 50-mer target DNA duplex containing one TA dinucleotide (Figure 6a). Integration into the top or bottom strand of the target duplex produces two major strand transfer products, of 68 and 40 nt, respectively, which are then detected by denaturing PAGE (Figure 6b, lane 3). Other minor products result from integration into the IR DNA substrate, the sequence of which contains two TA dinucleotides. When Raltegravir is added (in the concentration range 1 to 100 μM), integration is inhibited (Figure 6b, lanes 4−10) with an IC 50 of ∼2 μM (Figure 6c). This result implies that Raltegravir binds more tightly to the Mos1 PEC than to the transposase in the absence of DNA. The tighter binding could reflect additional interactions between precleaved IR DNA and the drug and/or a different Raltegravir binding mode; a co-crystal structure of Raltegravir and the Mos1 PEC would shed light on these possibilities.  34 (e.g., leukemia) and enhances the efficiency and accuracy of DNA repair by non-homologous end-joining. 2 It has been proposed that targeting the nuclease activity of SETMAR with small molecules could augment current chemotherapies 27 by inhibiting DNA repair. We have shown here that Raltegravir can also inhibit the in vitro DNA cleavage and integration of the mariner transposase Mos1. Given the wealth of structural and mechanistic understanding of the Mos1 DNA transposase and the close sequence and structural similarities between Mos1 and the transposase domain of SETMAR, the Mos1 transposase may provide an ideal model system to aid the development of drugs to target SETMAR.
■ METHODS Sequence Alignments. Structure-based sequence alignments of the transposase and integrase catalytic domains were performed using Expresso 35 and displayed using ESPript2.2. 36 Purification of Mos1 Transposase. Mos1 transposase containing the solubilizing mutation T216A (referred to as Mos1 throughout) was expressed and purified as previously described 37 with some changes. The protein was extracted from resuspended cells in a cell disruptor (Constant Systems, Ltd.) at 27 kPsi. After cation exchange chromatography of cleared lysate, the protein was further purified by hydrophobic interaction chromatography. Ammonium sulfate was added (to 1 M) before loading Transposase onto a HiTrap Phenyl HP column (GE Healthcare). Transposase was eluted using a decreasing ammonium sulfate gradient (1 to 0 M), then exchanged into 50 mM Tris pH7.5, 0.35 M KCl, 1 mM DTT, and concentrated using a 6 mL Vivaspin column (10 kDa cutoff). The purity of the protein was assessed by SDS-PAGE.
Annealing of IR DNA Duplex. Oligonucleotides were synthesized by Integrated DNA Technologies and dissolved in water prior to annealing. Duplex IR DNA containing the right inverted repeat sequence and with a 3 base overhang, corresponding to an excised transposon end, was prepared by annealing a 28-mer (5′-AAAC-GACATTTCATACTTGTACACCTGA-3′) and a complementary 25mer (5′-GGTGTACAAGTATGAAATGTCGTTT-3′). For the target integration assays the 28-mer had an IRDye 700 fluorescent tag at the 5′ end. Samples were heated from 20 to 80°C in increments of 0.5°C every 30 s. The fluorescence of Sypro Orange was excited at 485 nm and measured at 575 nm in relative fluorescence units (RFU). The sensitivity of Sypro Orange to the hydrophobicity of the chemical environment results in an increase in fluorescence intensity as the sample unfolds and hydrophobic residues are exposed. The transition unfolding temperature (T m ) of each sample was taken as the minimum value of the derivative −δRFU/δTemp. Each condition was measured in triplicate and an average T m for each condition calculated.
Crystallization and Crystal Soaks. Crystals of the Mos1 catalytic domain were grown by hanging drop vapor diffusion as previously described. 37 Purified Mos1 transposase at 8 mg mL −1 was mixed in the ratio 1:1 with well solution containing MgCl 2 (5 mM) or MnCl 2 (5 mM), 22% (w/v) PEG 4000, 100 mM Tris pH 6.8. Crystals appeared after incubation at 290 K for 10 days. For soaking experiments, crystals were transferred to a solution containing 10 mM MgCl 2 or MnCl 2 , 22% (w/v) PEG 4000, 100 mM Tris pH 6.8, 20% (v/v) glycerol, and 1 mM Raltegravir and incubated over the well solution for 18 h. Crystals were then flash frozen in liquid nitrogen prior to data collection.
Structure Determination and Refinement. X-ray diffraction data were collected at beamline I02 at the Diamond Light Source. Images were collected on a ADSC Q315r detector, integrated with iMosflm, and scaled using SCALA within the CCP4 suite. 38 Phases were calculated by molecular replacement in PHASER using the coordinates of the Mos1 catalytic domain (PDB ID: 2F7T) as the search model. Structure refinement was performed with REFMAC and the final data collection and refinement statistics are shown in Table 1 Reactions were incubated at 30°C for 20 min, stopped by the addition of EDTA to 10 mM, and separated on a 1% (w/v) agarose gel in 1x TAE buffer at 70 V for 2 h. DNA was stained using SafeView (NBS Biologicals) and visualized and quantified on a GelDoc EZImager (BioRad).
Target Integration Assays. A 50-mer target DNA substrate containing one TpA dinucleotide was prepared by annealing the 50 nt sequence 5′ AGCAGTCCACTAGTGCACGACCGTTCAAAGCT-TCGGAACGGGACACTGTT with its complementary strand. Annealed target and IR DNA oligonucleotides were purified by HPLC. Assays were performed in 20 μL reactions containing 15 nM 50-mer target DNA, 1.5 nM IR DNA, and 15 nM Mos1 Transposase in buffer containing 25 mM Hepes pH 7.5, 50 mM potassium acetate, 10% (v/v) glycerol, 0.25 mM EDTA, 1 mM DTT, 10 mM MgCl 2 , 50 μg mL −1 BSA, and 20% (v/v) DMSO. Raltegravir was added to reactions as indicated. Reactions were incubated for 2 h at 30°C, and the products were separated on an 8% denaturing polyacrylamide gel as described previously. 11 To visualize the products, the IRDye 700 was excited at 680 nm and detected on a LI-COR Odyssey system. The fluorescence intensities of the product bands were quantified using Image Studio software.

* S Supporting Information
Supplementary figures. This material is available free of charge via the Internet at http://pubs.acs.org.

Accession Codes
The molecular coordinates of the Mg 2+ and Mn 2+ containing crystal structures of Raltegravir bound to the Mos1 transposase catalytic domain have been deposited in the protein data bank (PDB) with accession codes 4MDB and 4MDA, respectively.

Notes
The authors declare no competing financial interest.
■ ACKNOWLEDGMENTS X-ray data were collected at the Diamond Light Source (beam line I02), and we thank the beam line staff for their help and expertise. J.M.R. is funded by a Wellcome Trust University Award (085176/Z/08/Z), M.T. was supported by a Darwin Trust PhD studentship, and E.R.M. is funded by the BBSRC (BB/J000884/1). We thank A. Cook for critical comments on the manuscript.