Discovery of Selective Bioactive Small Molecules by Targeting an RNA Dynamic Ensemble

Transcription

1 Discovery of Selective Bioactive Small Molecules by Targeting an RNA Dynamic Ensemble Andrew C. Stelzer 1,4,5, Aaron T. Frank 2,5, Jeremy D. Kratz 1, Michael D. Swanson 3, Marta J. Gonzalez-Hernandez 3, Janghyun Lee 1, Ioan Andricioaei 2, David M. Markovitz 3 and Hashim M. Al-Hashimi 1 * 1. Department of Chemistry & Biophysics, University of Michigan, 930 North University Avenue, Ann Arbor, Michigan 48109, USA 2. Department of Chemistry, University of California Irvine, 1102 Natural Sciences 2, Irvine, California 92697, USA 3. Department of Internal Medicine, Division of Infectious Diseases, University of Michigan Medical Center, Ann Arbor, Michigan 48109, USA 4. Current address Nymirum 3510 West Liberty Road, Ann Arbor, MI These authors contributed equally to this work Correspondence and requests for materials should be addressed to H. M. A. ( hashimi@umich.edu)

2 SUPPLEMENTARY METHODS Docking: All docking simulations were performed using the ICM (Molsoft LLC) docking program 1. ICM employs internal coordinate mechanics based on bond angles, bond lengths, torsion angles, and phase angles in grid-based docking simulations to efficiently determine interaction energies and optimal small molecule binding conformations in receptor-ligand complexes. ICM docking simulations were conducted using rigid RNA receptors and fully flexible small molecule ligands. In all cases, RNA structures were initially converted into ICM objects. Binding pockets were predefined for known RNA-small molecule structures and identified using ICM for unknown binding sites (see below). For structures with multiple binding sites, or in the event that multiple binding pockets were identified in one structure, each binding pocket was docked independently and the minimum score for each small molecule recorded. The small molecules were either hand drawn in MOL format using ChemDraw (CambridgeSoft Corp.) and converted into SDF format for ICM (validation sets of Fig. 1b and c and in-house library of ~2,000 small molecules used in virtual screen) or obtained in SDF format from CCG (University of Michigan). Small molecules were converted to SDF files using the openbabel software. Their protonation states were computed over the ph range using the major microspecies module of ChemAxon ( When multiple protonation states were found, each small molecule protonation state was docked independently and the minimum score recorded. For the validation simulations with known RNA-small molecule structure (Fig. 1b), small molecule protonation states were assigned as reported. Benchmark docking simulations against known RNA structures: Structures of all RNA-small molecule complexes (excluding ribosome) as of January, 2009 were downloaded from the Protein Data Bank ( A total of 96 structures that include riboswitches, aptamers, and various groove-binding RNA-small molecule complexes with N flex < 20 were used in the ICM validation simulations. A total of 48 RNA-small molecule structures (22 with metalmediated interactions and 26 without a reported K d ) were excluded from K d but not success rate analysis. The 22 structures with metal-mediated contacts were excluded because the ICM forcefield does not take into account such interaction energies. In all simulations, the RNA was separated from the small molecule, converted into an ICM object, and a receptor binding pocket defined as all atoms within 5 Å of the bound small molecule. The maximum number of Monte- Carlo iterations (thoroughness of 10) was used in these simulations. Docking simulations involving known TAR binders: The 20 conformers in the TAR NMR-MD ensemble were generated as previously described 2 using one round of SAS selection. X-ray (TAR X-ray ) (397D) 3 and NMR (TAR NMR ) (1ANR) 4 structures of apo-tar were downloaded from the PDB. The chemical structure of 38 TAR-binding compounds was obtained from the literature The TAR MD ensemble was constructed by randomly selecting twenty snap-shots from an 80 ns simulation of apo-tar with a backbone RMSDs ranging 3-80 Å. Binding pockets were predicted using the ICM PocketFinder module, which calculates the surface area and volume of cavities on the receptor surface. All binding pockets that fall within the druggable range (volume= Å 3 and surface area= Å 2 ), as determined by ICM, were set as receptor centroids and the final receptor defined as all atoms within 5 Å of the predicted binding site.

3 Fluorescence-based binding assay: A fluorescence-based assay employing a TAR construct labeled with 2-aminopurine at bulge residue U25 (TAR-ap25). A second TAR construct labeled with fluorescein attached to bulge residue U25 (TAR-fl25) was used to measure binding of small molecules due to spectral overlap with 2-aminopurine 7. The TAR-fl25 binding assay yielded within experimental error an identical K d for neomycin ( M) as obtained with the TAR-ap25 assay ( M). Fluorescence intensity measurements were recorded in triplicate following incremental addition of small molecules and values normalized to unbound- TAR. K d s were determined by non-linear regression by fitting the measured fluorescence intensities (F) to the following equation 7 using the Origin software (OriginLab Corporation), F A [RNA] T B ([RNA] T [L] T K d ) ([RNA] T [L] T K d ) 2 4[RNA] T [L] T 2 in which [RNA] T is the total TAR concentration (100 nm), [L] T is the total ligand concentration, A and B are weighting factors that were allowed to float during curve fitting. The assay was validated by measuring K d s for the known TAR binders neomycin B 16, paromomycin 5, kanamycin 5, ribostamycin 5, tobramycin 5, and argininamide 16 (Fig. S4a). Similar fluorescence-binding assay was used to measure binding of the small molecules to HIV- 2 TAR 7,17, prokaryotic A-site 18, RRE 19, and HIV-2* TAR (Fig 2a and Supplementary Fig 8a). Competition binding experiments employed the fluorescence binding assay described above in presence of the unlabelled trna, A-site, RRE, and HIV-2 RNA constructs (see Supplementary Fig 8b). Fluorescence-based TAR-Tat displacement assay: We used a fluorescence polarization (FP) based displacement assay to measure inhibition of TAR-Tat binding by the small molecules. This assay employed an N-terminus fluorescein labeled Tat peptide (F-Tat) N-AAARKKRRQRRR-C, (Fig. S3) (Genscript Corp), which contains the arginine rich motif (RKKRRQRRR) that is critical for interaction with TAR 9. An elongated HIV-1 TAR (E-TAR) construct (Fig. S3) was prepared by in vitro transcription and purified by PAGE, was used in these experiments to increase the dynamic range of FP measurements. Binding of F-Tat to E-TAR was initially assayed by measuring FP for F-Tat (10 nm) as a function of increasing E-TAR concentration. An apparent K d was obtained by fitting the observed FP (FP obs ) to the following equation 20 using the Origin software (OriginLab Corporation), FP obs FP free (FP bound FP free ) ([L] T [RNA] T K d ([L] T [RNA] T K d ) 2 (4[L] T [RNA] T ) 2[L] T in which FP free and FP bound are the measured FP values for free and bound F-Tat, respectively, [L] T is the total concentration of F-Tat, and [RNA] T is the total E-TAR concentration. P bound, P free, and K d were allowed to float during the fit. The E-TAR-F-Tat K d (~40 nm) obtained in this assay was is in very good agreement with values reported in the literature (25 nm) 9. FP measurements in the presence of small molecules were performed by incubating E-TAR (60 nm) with increasing amounts of small molecules for 10 mins followed by addition of F-Tat (10 nm) and a second round of incubation for 10 mins. For each small molecule increment, data was

4 collected for 30 minutes at 5 minute intervals to ensure full equilibration has been reached. The FP data was collected in triplicate and IC 50 values calculated by fitting the observed FP (FP obs ) to the following equation from the Prizm software (GraphPad Software Inc.), FP obs FP free (FP bound FP free ) 1 ( [I] IC 50 ) m, in which FP free and FP bound are the FP values measured for free and E-TAR bound F-Tat peptide, respectively, [I] is log the total concentration of small molecule inhibitor, IC 50 is the 50% small molecule inhibitory concentration, m is the slope of the linear portion of the sigmoidal curve. IC 50 and m were allowed to float during the fit. Inhibition constants (K i ) were calculated from the Prizm software (GraphPad Software Inc.) using the following equation, log(ic 50 ) log(10 log(k i ) (1 [FTat] T K d )), where [F-Tat] T is the total concentration of F-Tat and K d is the F-Tat/E-TAR dissociation constant. All K i s reported in this manuscript are exact K i s calculated from the IC 50 and a range of K i s based on the 95% confidence interval. The displacement assay was benchmarked by measuring K i s for known TAR binders neomycin B, paromomycin, kanamycin, ribostamycin, tobramycin, and argininamide (Fig. S4b). NMR spectroscopy: Uniformly 13 C/ 15 N labeled TAR was prepared by in-vitro transcription and purified by PAGE as described previously 21. Chemical shift mapping experiments were performed by recording 2D HSQC spectra following incremental addition of small molecules from a stock solution (100 mm) to 13 C/ 15 N labeled TAR (~ mm). The magnitudes of the chemical shift perturbations ( av ) shown in Fig. 4 were computed using av ( H ) 2 ( C ( C )) 2, in which H and C is the change in 1 H and 13 C chemical shift H (in ppm), respectively, and C and H are the gyromagnetic ratios for carbon and hydrogen, respectively. NMR chemical shift perturbations were also used to measure the dissociation constant (K d ) for butirosin A (Table 1 and Fig. S7), by fitting the observed chemical shift at a given butirosin A concentration to the following equation using the Origin software (OriginLab Corporation), obs free ( t ) ([But] t [RNA] t K 2 d ) ([But] [RNA] K 2 t t d ) (4[But] t [RNA] t ) 2[RNA] t, where obs is the weighted average chemical shift measured at each ligand concentration, free is the chemical shift in the free state (in ppm), [But] T is the total butirosin A concentration, [TAR] T is the total TAR concentration based on UV absorbance at 260 nm, T is the difference in

5 chemical shifts between the free and ligand-bound states (in ppm). t and K d are allowed to float during the fit. NMR-based validation of predicted TAR-small molecule structures: Populations of TAR conformers shown in Fig. 3a were determined using the partition function e G i / RT P. Apparent K d values were calculated for each docking score using the best-fit line equation determined in Fig. 1c. All six small molecules were re-docked against the 20 TAR NMR-MD conformers 5 times populations for each conformer calculated, and the final populations (shown in Fig. 3a) determined as the average over the 5 runs. Sisomicin and amikacin showed the largest standard deviation in scores between runs (8.7 kcal/mol and 6.0 kcal/mol respectively) while the other four small molecules had standard deviations <2.6 kcal/mol. Correspondingly, sisomicin and amikacin also showed the widest distributed populations across the 20 TAR NMR-MD conformers. Fig. 3c shows the TAR conformer selected with highest frequency over the five runs for mitoxantrone (conformer 15), sisomicin (conformer 15), netilmicin (conformer 18), butirosin A (conformer 18), and DMA (conformer 14) in the lowest scoring binding pose. Amikacin bound two distinct conformers with equal frequency, shown is the lowest scoring conformer (18). Subtle aspects of the predicted binding modes are supported by the chemical shift perturbations (Fig. 3c and Fig. S9); unique stacking of mitoxantrone on G26 is supported by observation of large and directionally unique perturbations at G26(C8), the distinct binding modes of netilmicin and sisomicin, which differ only by a single ethylene group, is supported by observation of significant perturbations at G28 and A35 with netilmicin but not sisomicin; binding of butirosin A and amikacin to similar pockets is supported by observation of significant perturbations at many common sites; and finally binding of DMA to a unique pocket within the apical loop which is supported by observation of extensive directionally unique perturbations at many apical loop residues (G32(C8/C1 ), G33(C1 ), G34(C1 ) and A35(C1 )). Thus, not withstanding the limitations in quantitatively interpreting chemical shift perturbations, the data support global aspects of the docking predictions and particularly the predicted differences in the binding modes of the six small molecules. i i 1 e G / RT

6 SUPPLEMENTARY RESULTS SUPPLEMENTARY FIGURES Figure S1 The 20 conformers from the TAR NMR-MD dynamic ensemble.

7 Figure S2 a, Percentage of ICM predicted ligand poses that fall within a heavy-atom RMSD relative to the real structure using a benchmark of known RNA small molecule bound structures. b, Correlation plots between experimental binding energies ( G = RTln(K d ) and ICM docking scores for 38 TAR binders when docking small molecules onto X-ray (blue), NMR (red) structures of apo-tar, and an MD ensemble consisting of 20 random snapshots from an 80 ns apo-tar MD simulation (green).

8 Figure S3 Binding of F-Tat peptide to E-TAR by FP. Measured FP values upon incremental addition of E-TAR to 10 nm F-Tat. Average values and standard deviations over triplicate runs are shown.

9 Figure S4 Benchmark experiments for fluorescence assays. a, Validating the TAR-fl25 binding assay using five known TAR binders neomycin 16, kanamycin 5, paromomycin 5, tobramycin 5, ribostamycin 5, and argininamide 16. b, Validation of the displacement assay using six inhibitors of the TAR-Tat interaction. The K i values obtained for neomycin ( M) and argininamide ( M) are in very good agreement with values reported previously (1 M 16 and 1 mm 22 respectively).

10 Figure S5 TAR chemical shift perturbations induced by spermine. 2D HSQC spectra of free 200 M TAR (black) and upon addition of 200 M (red), 400 M (orange), 800 M (green), and 1600 M(blue) spermine.

11 Figure S6 Docking-based predictions. a, Examples of small molecules that are correctly predicted to bind TAR weakly as verified using fluorescence binding assays. b, Representative examples of false positives. The docking score is indicated next to each small molecule.

12 Figure S7 Discovery of TAR binding small molecules by virtually screening its dynamic structure ensemble. a, Binding of small molecules to HIV-1 TAR using fluorescence intensity measurements (mitoxantrone, sisomicin, netilmicin, amikacin, and DMA) in the absence (filled circles) and presence (open circles) of 100-fold excess (relative to TAR) S. Cerevisiae trna and NMR chemical shift perturbations (butirosin A, due to commercial unavailability at the onset of these experiments). Computed dissociation constants (K d ) are shown in each case. b, Displacement of N-terminal-labeled-fluorescein Tat peptide from TAR as measured by fluorescence polarization as a function of small molecule concentration. Computed inhibition constants (K i ) are shown in each case.

13 Figure S8 Measuring binding specificity of small molecules. Binding of small molecules to HIV-2 TAR, RRE, prokaryotic A-site, and HIV-2* (TAR construct lacking bulge residue C24) using fluorescence intensity measurements in the a, absence and b, presence of unlabelled competitor RNA. HIV-0 has the same sequence as HIV-1 TAR but lacks the UCU bulge. Computed dissociation constants (K d ) are shown in each case. Mitoxantrone binding was not detectable in background of competitor RNA due to non-specific binding. Note that DMA binds the HIV-1 TAR apical loop sequence and no binding is observed with the distinct RRE and A- site tetraloops.

14 a

15 b

16 c

17 d

18 e

19 f Figure S9 TAR chemical shift perturbations a, Mitoxantrone. Shown are 2D HSQC spectra of free 100 M TAR (black) and upon addition of 100 M (red), 200 M (orange), and 400 M (green) mitoxantrone. b, Sisomicin. Shown are 2D HSQC spectra of free 200 M TAR (black) and upon addition of 200 M (red), 400 M (orange), 800 M (green), and 1600 M (blue) sisomicin. c, Netilmicin. Shown are 2D HSQC spectra of free 200 M TAR (black) and upon addition of 200 M (red) and 400 M (orange) netilmicin. d, Amikacin. Shown are 2D HSQC spectra of free 200 M TAR (black) and upon addition of 200 M (red), 400 M (orange), 800 M (green), and 1600 M (blue) amikacin. e, Butirosin A. Shown are 2D HSQC spectra of free 100 M TAR (black) and upon addition of 100 M (red), 200 M (orange), 400 M (green), and 650 M (blue) butirosin A. f, DMA. Shown are 2D HSQC spectra of free 200 M TAR (black) and upon addition of 200 M (red), 400 M (orange), 800 M (green), and 1600 M (blue) DMA.

20 Figure S10 Netilmicin specifically inhibits Tat-mediated transactivation and HIV-1 replication. Jurkat T cells were treated with 100 μm netilmicin, or vehicle, for 24 hours and then transfected with the either HIV-2 LTR with HIV-1 Tat, HIV-1 LTR with HIV-2 Tat constructs, or empty vector (pcdna3.1) with each corresponding LTR. Following transfection, netilmicin or vehicle was added to the media and luciferase activity measured 24 hours later.

21 SUPPLEMENTARY Data Set (SEE ATTACHED FILE) SUPPLEMENTARY REFERENCES 1 Abagyan, R., Totrov, M. & Kuznetsov, D. ICM - a new method for protein modeling and design - applications to docking and structure prediction from the distorted native conformation. J. Comput. Chem. 15, (1994). 2 Frank, A. T., Stelzer, A. C., Al-Hashimi, H. M. & Andricioaei, I. Constructing RNA dynamical ensembles by combining MD and motionally decoupled NMR RDCs: new insights into RNA dynamics and adaptive ligand recognition. Nucleic Acids Res. 37, (2009). 3 Ippolito, J. A. & Steitz, T. A. A 1.3-angstrom resolution crystal structure of the HIV-1 trans-activation response region RNA stem reveals a metal ion-dependent bulge conformation. Proc. Natl. Acad. Sci. USA 95, (1998). 4 Aboul-ela, G., Karn, J. & Varani, G. Structure of HIV-1 TAR RNA in the absence of ligands reveals a novel conformation of the trinucleotide bulge (vol 24, pg 3974, 1996). Nucleic Acids Res. 24, (1996). 5 Blount, K. F. & Tor, Y. Using pyrene-labeled HIV-1 TAR to measure RNA-small molecule binding. Nucleic Acids Res. 31, (2003). 6 Blount, K. F., Zhao, F., Hermann, T. & Tor, Y. Conformational constraint as a means for understanding RNA-aminoglycoside specificity. J. Am. Chem. Soc. 127, (2005). 7 Bradrick, T. D. & Marino, J. P. Ligand-induced changes in 2-aminopurine fluorescence as a probe for small molecule binding to HIV-1 TAR RNA. RNA 10, (2004). 8 Ding, L., Zhang, X. X., Chang, W. B., Lin, W. & Yang, M. Studies of binding constants and interaction of drugs to trans-activation response RNA by capillary electrophoresis. Anal. Chim. Acta 543, (2005). 9 Hamasaki, K. & Ueno, A. Aminoglycoside antibiotics, neamine and its derivatives as potent inhibitors for the RNA-protein interactions derived from HIV-1 activators. Bioorg. Med. Chem. Lett. 11, (2001). 10 He, M. Z. et al. Synthesis and assay of isoquinoline derivatives as HIV-1 Tat-TAR interaction inhibitors. Bioorg. Med. Chem. Lett 15, (2005). 11 Ironmonger, A. et al. Scanning conformational space with a library of stereo- and regiochemically diverse aminoglycoside derivatives: the discovery of new ligands for RNA hairpin sequences. Org. Biomol. Chem. 5, (2007). 12 Mayer, M. et al. Synthesis and testing of a focused phenothiazine library for binding to HIV-1 TAR RNA. Chem. Biol. 13, (2006). 13 Mei, H. Y. et al. Inhibition of an HIV-1 Tat-derived peptide binding to TAR RNA by aminoglycoside antibiotics. Bioorg. Med. Chem. Lett. 5, (1995). 14 Xu, Y. L. et al. Synthesis and evaluation of novel neamine derivatives effectively targeting to RNA. Bioorg. Med. Chem. Lett. 19, (2009). 15 Zhao, P., Jin, H. W., Yang, Z. J., Zhang, L. R. & Zhang, L. H. Solid-phase synthesis and evaluation of TAR RNA targeted beta-carboline-nucleoside conjugates. Org. Biomol. Chem. 6, (2008).

22 16 Nandi, C. K., Parui, P. P., Brutschy, B., Scheffer, U. & Gobel, M. Fluorescence correlation spectroscopy at single molecule level on the Tat-TAR complex and its inhibitors. Biopolymers 89, (2008). 17 Zhao, F. et al. Molecular recognition of RNA by neomycin and a restricted neomycin derivative. Angew. Chem. Int. Edit. 44, (2005). 18 Kaul, M., Barbieri, C. M., & Pilch, D. S. Fluorescence-based approach for detecting and characterizing antibiotic-induced conformational changes in ribosomal RNA: comparing aminoglycoside binding to prokaryotic and eukaryotic ribosomal RNA sequences.. J. Am. Chem. Soc. 126, (2004). 19 DeJong, E. S., Chang, C. E., Gilson, M. K. & Marino, J. P. Proflavine acts as a Rev inhibitor by targeting the high-affinity Rev binding site of the Rev responsive element of HIV-1. Biochemistry 42, (2003). 20 Lundblad, J. R., Laurance, M. & Goodman, R. H. Fluorescence polarization analysis of protein-dna and protein-protein interactions. Mol. Endocrinol. 10, (1996). 21 Zhang, Q., Sun, X. Y., Watt, E. D. & Al-Hashimi, H. M. Resolving the motional modes that code for RNA adaptation. Science 311, (2006). 22 Tao, J. S. & Frankel, A. D. Specific binding of arginine to TAR RNA. Proc. Natl. Acad. Sci. USA 89, (1992).