Discovery of selective RNA-binding small molecules by affinity

Subscriber access provided by UNIVERSITY OF TOLEDO LIBRARIES

Article

Discovery of selective RNA-binding small molecules by affinity-selection mass spectrometry Noreen F. Rizvi, John A. Howe, Ali Nahvi, Daniel J. Klein, Thierry O. Fischmann, Haiyoung Kim, Mark A. McCoy, Scott S Walker, Alan Hruza, Matthew P. Richards, Chad Chamberlin, Peter Saradjian, Margaret T. Butko, Gabriel Mercado, Julja Burchard, Corey Strickland, Peter J. Dandliker, Graham F. Smith, and Elliott B. Nickbarg ACS Chem. Biol., Just Accepted Manuscript • DOI: 10.1021/acschembio.7b01013 • Publication Date (Web): 07 Feb 2018 Downloaded from http://pubs.acs.org on February 8, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

ACS Chemical Biology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Chemical Biology

ACS Paragon Plus Environment

ACS Chemical Biology 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 30

Discovery of selective RNA-binding small molecules by affinity-selection mass spectrometry

Noreen F. Rizvi†,∆, John A. Howe‡, Ali Nahvi§, Daniel J. Klein§, Thierry O. Fischmann‡, Hai-Young Kim†, Mark A. McCoy‡, Scott S. Walker‡, Alan Hruza‡, Matthew P. Richards†, Chad Chamberlin†, Peter Saradjian†, Margaret T. Butkoǁ, Gabriel Mercadoǁ, Julja Burchard†,┴, Corey Strickland‡, Peter J. Dandliker†, Graham F. Smith†,#, and Elliott B. Nickbarg†* †

Merck & Co., Inc., Boston, Massachusetts, 02115, USA Merck & Co., Inc., Kenilworth, New Jersey, 07033, USA § Merck & Co., Inc., West Point, Pennsylvania, 19486, USA ǁ Biodesy, Inc., South San Francisco, California, 94080, USA ┴ Current Address: Sera Prognostics, Inc., Salt Lake City, Utah, 84109, USA # Current Address: AstraZeneca, Cambridge Science Park, Cambridge, CB4 0WG, UK ∆ Current Address: Siemens Healthcare Diagnostics, Tarrytown, New York, 10591, USA *[email protected] ‡

TOC Graphic

1 ACS Paragon Plus Environment

Page 3 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Abstract Recent advances in understanding the relevance of non-coding RNA (ncRNA) to disease have increased interest in drugging ncRNA with small molecules. The recent discovery of ribocil, a structurally distinct synthetic mimic of the natural ligand of the flavin mononucleotide (FMN) riboswitch, has revealed the potential chemical diversity of small molecules that target ncRNA. Affinity-selection mass spectrometry (AS-MS) is theoretically applicable to high-throughput screening (HTS) of small molecules binding to ncRNA. Here we report the first application of the Automated Ligand Detection System (ALIS), an indirect AS-MS technique, for the selective detection of small molecule-ncRNA interactions, high-throughput screening against large unbiased small-molecule libraries, and identification and characterization of novel compounds (structurally distinct from both FMN and ribocil) that target the FMN riboswitch. Crystal structures reveal different compounds induce various conformations of the FMN riboswitch, leading to different activity profiles. Our findings validate the ALIS platform for HTS screening for RNA-binding small molecules, and further demonstrate that ncRNA can be broadly targeted by chemically diverse yet selective small molecules as therapeutics.



Page 4 of 30

Introduction As functional roles and biological relevance of non-coding RNA (ncRNA) become increasingly apparent, particularly in human disease, it is clear that ncRNA structural elements may represent a new and interesting class of targets for novel therapeutics and drug discovery1. To date, various classes of ncRNA have been targeted by small molecules, such as the bacterial ribosome, the HIV leader sequence, HCV IRES in the 5’ untranslated region (UTR), bacterial 5’UTR riboswitches, r(CUG) intronic repeats that cause myotonic dystrophy (DM1), the repeat expansion r(GGGGCC) responsible for frontal temporal dementia/ALS, and a 5’ splice site (exon-intron junction) responsible for spinal muscular atrophy (SMA) 2-4. Methods used to identify small molecules that bind to RNA have included high throughput screening methods including mass spectrometry (MS)-based, fluorescence-based, and FRET-based approaches5-8. Other biophysical methods have included coupling computational molecular dynamics with NMR9, and fragment-based approaches10. Small-molecule microarrays have identified small molecules that bind to several forms of RNA11. Informatics-based approaches have also aided in RNA drug discovery, including a selection-based 2D combinatorial screening (2DCS) in tandem with a statistical method to analyze structure-activity relationships through sequencing (StARTS) and additional structural information about the RNA (Inforna)12-15. RNA-ligand docking analyses have been used for structure-based drug design and virtual screening16, 17. In addition to these biophysical and informatics approaches, phenotypic assays have also been used18, 19. Despite these important efforts, technologies capable of screening large, unbiased small-molecule libraries using biophysical techniques remain limited. There is a need to develop screening approaches to identify small molecules that can target ncRNA. Potentially, one such technology is the Automated Ligand Identification System (ALIS) a label-free AS-MS platform for high-throughput screening of small-molecules as large combinatorial mixtures tested for binding to target macromolecules20. By coupling fast (< 20 s) size exclusion chromatography (SEC) to separate free ligand from target-ligand complexes to integrated LC-MS ligand identification, a single ALIS instrument can screen 500,000 compounds a day with minimal, label-free target consumption. Unlike direct AS-MS methods that rely on MS to detect the intact target-ligand complex, ALIS is an “indirect” AS-MS technique that uses size-exclusion chromatography to resolve the target-ligand complex from the unbound species, then dissociates the ligand from the complex using denaturing conditions and 3 ACS Paragon Plus Environment

Page 5 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


employs MS to identify the previously bound ligand21, 22. ALIS has been used to efficiently screen a variety of classes of protein targets including polymerases, kinases, integral membrane proteins, G protein-coupled receptors, and enzymes21, and has also been used as an analytical tool for characterizing protein-ligand interactions in ligand competition and affinity ranking assays 23. While ALIS has been successful in identifying and characterizing important small molecules that bind to protein targets, ALIS analysis on other macromolecules such as RNA has not been explored previously. We chose to use bacterial riboswitches, regulatory ncRNA located in the 5’ UTR of bacterial genes, to explore the use of ALIS with ncRNA. Riboswitch aptamers fold to form sensitive, specific sensors that bind a natural small molecule ligand in a structure-dependent manner with high affinity. Ligand binding induces changes in RNA aptamer structure and thereby modulates gene expression through translation initiation, transcription elongation of mRNA, and/or the splicing of mRNA transcripts 24, 25. The discovery of ribocil showed that druglike synthetic small molecule ligands can also selectively bind to riboswitches 18. Thus riboswitches are a class of ncRNA known to bind both natural and synthetic ligands, and are useful for exploration of ncRNA-small molecule interactions on the ALIS platform. Here, we report the use of the ALIS platform to detect the selective binding of five natural ligands to their respective RNA riboswitches. Natural (i.e. cognate) and synthetic ligand binding to the FMN riboswitch was further characterized in ALIS. In an HTS screen of the FMN riboswitch using 53,000 antibacterial compounds, ALIS successfully detected binding of ribocil, which was originally identified through a phenotypic screen18. Additionally, ribocil analogs originally missed in the phenotypic screen were also identified in ALIS, along with other structurally diverse compounds that bind to the FMN riboswitch. ALIS competition studies and affinity ranking assays identified the highest-affinity compounds that competitively bind the FMN riboswitch. Further structural and conformational characterization, including crystal structure determination, 1D 1H-NMR, and second-harmonic generation (SHG) of these novel compounds demonstrates that the FMN riboswitch is able to bind structurally diverse small molecules in multiple, distinct conformations, leading to varied bioactivity. High-throughput ALIS screening aids in the discovery of novel chemical classes that bind ncRNA that may be missed by non-biophysical screening approaches. Our studies show that ALIS is suitable for the screening and analysis of RNA-binding compounds and can be used in the future to identify 4 ACS Paragon Plus Environment


Page 6 of 30

other synthetic and natural small molecules that interact with important ncRNA.

Results & Discussion Detection of small molecule-ncRNA interactions using ALIS Five well-characterized bacterial riboswitches were tested for binding to their natural ligand using ALIS: the flavin mononucleotide (FMN) riboswitch 26, 27, S-adenosylhomocysteine (SAH) riboswitch 28, thiamine pyrophosphate (TPP) riboswitch 29, S-adenosylmethionine (SAM) riboswitch 30, and the B12 riboswitch 31. Each riboswitch construct was combined with its natural ligand or a mixture of ligands (FMN, SAH, TPP, SAM, and coenzyme B12), annealed and analyzed using ALIS. Binding of each of the natural ligands to their respective riboswitch is successfully detected through ALIS (Figure 1). Importantly, each of the riboswitches only binds its natural ligand and can discriminate against all of the other ligands, demonstrated in the ligand mixture condition (Figure 1). Notably, the SAM ligand differs from SAH only by a single methyl group and the associated positive charge at the sulfur. The SAH-riboswitch, however, is able to selectively recognize SAH with an affinity that is 1000-fold better than for SAM 28. Similarly, the SAM-riboswitch exhibits 100-fold discrimination against SAH 30. This discrimination is also observed through ALIS binding experiments, since each of these riboswitches only bound their natural ligands. Characterization of ligand binding to the FMN riboswitch using ALIS Ligand binding to the FMN riboswitch was further characterized in ALIS using two natural ligands of the FMN riboswitch -- FMN and roseoflavin, a chemical analog of FMN that also binds to the FMN riboswitch to regulate gene expression 27, 32 -- and two active synthetic mimics, ribocil and its close analog ribocil-C (Figure 2A)18. FMN has the strongest affinity to the FMN riboswitch (Kd = 1.75 ± 1.1 nM), followed by ribocil-C (Kd = 20.5 ± 10.5 nM), ribocil (Kd = 25.6 ± 14.3 nM) and finally roseoflavin (Kd = 297.2 ± 79.4 nM) (Figure 2B). Similar values have previously been reported 18, 27, 32, 33 for FMN, roseoflavin, ribocil, and ribocil-C binding to the FMN riboswitch as determined through in-line probing assays or fluorescence quenching assays (Figure 2B), suggesting that features of ALIS, such as the SEC separation, do not perturb known RNA-ligand interactions. The slight discrepancies in binding affinity values compared to 5 ACS Paragon Plus Environment

Page 7 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


previously reported data may be attributed to different aptamer sequences from different bacterial species and different monovalent (NaCl, KCl) and divalent (MgCl2) ion concentrations in binding conditions 27, 34. ALIS competition experiments have been used to classify protein target-ligand interactions as competitive or non-competitive/allosteric, and also to rank the affinity of multiple ligands from a mixture 21, 23, 35. These experiments were similarly implemented to rank ligand affinities by competing the various ligands of the FMN riboswitch with the natural ligand, FMN in the presence of excess FMN riboswitch. The ability of the competitor ligand, FMN, to displace each of the RNA-bound ligands (roseoflavin, ribocil and ribocil-C) indicates each of these ligands competes directly with FMN for binding to the riboswitch (Figure 2C). The relative binding affinities can be expressed using the affinity competition experiment 50% inhibitory concentration (ACE50), which is defined as the total titrant concentration at which each RNAligand complex concentration is reduced to one-half its value, and is dependent upon the Kd of the ligand 35, 36. By analyzing the ACE50 value of each of the competitive ligands, the ligands can be ranked by affinity: ribocil-C > ribocil ≥ roseoflavin (Figure 2C). These values correspond to those determined by affinity (Kd) titration (Figure 2B) and previously through fluorescence quenching assays 18. The ALIS competition assay enables multiple competing ligands to be characterized simultaneously, an advantage over fluorescence methods. ALIS binding analysis can also be used to explore bacterial resistance to antibiotics. In previous studies, ribocil-resistant (ribocilR) mutants of the FMN riboswitch were identified, including six individual point-mutations (G37U, G93U, C111U, C100U, U218C, and C219U) and one deletion (∆94-102) 18. The binding of ribocil (as well as FMN) to each of these mutants was examined in ALIS. The FMN riboswitch wild-type (WT) and each of the seven mutants was annealed in the presence of either ribocil or FMN and run in ALIS. The MS signal was normalized relative to the FMN riboswitch WT signal for each ligand to compare the changes in binding of each mutant. Binding of ribocil is significantly lower or completely abolished in many of the mutants (G37U, G93U, C111U, C100U and ∆94-102) compared to the FMN riboswitch WT (Figure 2D). Therefore resistance of these mutants to ribocil can, in part, be explained by this reduced binding of ribocil. Previous studies show similarly reduced or abolished ribocil binding to the G37U, G93U, and ∆94-102 mutants18, 37. The U218C and C219U mutants map to the expression platform rather than the FMN aptamer region where these ligands bind18. Instead 6 ACS Paragon Plus Environment


Page 8 of 30

of a decrease in binding of ribocil, these mutations alter the sequestration loop to expose the Shine-Delgarno sequence for deregulation of ribB expression18, 37. Therefore, as expected, ribocil binding to these mutants is not significantly reduced compared to the WT (Figure 2D). In contrast, binding of FMN is not significantly different in the majority of the mutants (G37U, C111U, C100U, U218C, and C219U) compared to the WT (Figure 2D). However, binding of FMN to the G93U and ∆94-102 mutants in ALIS is significantly lower or completely abolished (Figure 2D), in agreement with other binding assays 18, 37. The FMN riboswitch controls expression of ribB via a negative feedback loop; when its effector ligand, (i.e. FMN) is bound, a conformational change abolishes ribB expression inhibiting riboflavin biosynthesis 26, 27

. The G93U and ∆94-102 mutations significantly alter binding of FMN to the FMN riboswitch

and therefore produce high levels of riboflavin18 but the resulting elevated riboflavin levels are not lethal to the bacterial cell, indicating that the regulatory FMN riboswitch is not essential to the riboflavin pathway. Previous studies support this notion, as deletion of the entire FMN riboswitch in E. coli did not affect cell growth nor negatively interfere with other physiological processes 38. To confirm and extend our findings with ALIS, binding of FMN and ribocil to the WT FMN riboswitch and to two mutants (G93U and C111U) was also analyzed using surface plasmon resonance (SPR; Biacore T200). For each RNA aptamer, results from serial injections at 10 different compound concentrations (10 - 0.0005 µM, 3-fold dilution) were fit to a 1:1 binding model (Figure S1). The binding kinetics of FMN to the WT riboswitch are characterized by a slower dissociation rate (t1/2 ~ 10 min) with a binding affinity (Kd) of ~ 1 nM, consistent with ALIS generated results. FMN binding to the C111U mutant is comparable to that to WT, but binding to the G93U mutant is significantly weaker, expressed by both a slower kon and faster koff, resulting in a ~1,300-fold decrease in affinity. Binding kinetics of ribocil to both the WT and C111U mutant riboswitches were essentially identical (Kd ~25 nM), but no appreciable binding of ribocil to the G93U mutant was observed. The SPR results are generally consistent with ALIS affinity measurements, and the kinetic analysis further supports the overall trends observed in ALIS. Together, these data show that ALIS is a robust platform to characterize compound binding to RNA targets and variants, and can provide insights into the connections between phenotypic changes such as ribocil-resistance and differences in biophysical binding. ALIS can also allow rank-ordering for compound affinity comparison, alongside methods such as 7 ACS Paragon Plus Environment

Page 9 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


fluorescence quenching and SPR.

Comparative high-throughput ALIS screening of the FMN riboswitch against a large collection of bioactive compounds recapitulates phenotypic data The same set of ~53,000 synthetic antibacterial small molecules used in a riboflavin pathway phenotypic screen in the discovery of ribocil18 was screened for molecules that bind the FMN riboswitch in ALIS. In addition to the FMN riboswitch WT, a ribocilR mutant (the G93U mutant was selected since this point mutation showed dramatic decreases in ribocil binding), and a scrambled sequence of the FMN riboswitch (termed “Scramble FMN”) were screened as a comparison and selectivity controls, respectively. ALIS screening was done using pooled mixtures of 500 compounds (0.5 µM/compound), allowing for ~53,000 compounds to be screened against each of the three RNA constructs in duplicate on six parallel instruments in one day. A total of 61 compounds bound to the FMN riboswitch WT, of which 22 compounds are selective and do not bind the G93U mutant or the Scramble FMN (Figure 3A, “W” compounds, named for WT). Importantly, ribocil and roseoflavin are in this subset of compounds that bind exclusively to the FMN riboswitch WT, consistent with previously reported data 18, 27, 32. These findings indicate that known hits are detectable in mixture-based highthroughput ALIS screening. We also found a ribocil-analog that was missed in the original phenotypic assay along with other not previously observed compounds. 73 total compounds bind to the G93U mutant, of which 20 compounds overlap with the FMN riboswitch WT (“WG” compounds, named for WT and G93U). The Scramble FMN binds 52 compounds, 15 of which overlap the FMN riboswitch WT and G93U mutant (“WGS” compounds, named for WT, G93U, and Scramble). The compounds that bind the Scramble FMN may be general RNA-binders, intercalators, or there may be a secondary/tertiary structure introduced into the Scramble FMN that is contributing to the binding of these compounds 39. Non-selective hits can be distinguished from selective binders using this method of comparative screening. The compounds that bind exclusively to the FMN riboswitch WT were ranked using the recently established protein titration ALIS (PT-ALIS) method40. While comparable to the ACE50 competition assay, PT-ALIS triages a larger number of compounds at once and does not require a strong titrant. The FMN riboswitch WT was titrated into a pool containing the 16 selective 8 ACS Paragon Plus Environment


Page 10 of 30

compounds (Figure 3A, W compounds). By decreasing the RNA concentration, the competition among the pooled compounds increases and at the lowest concentration, only the highest affinity compounds bind to the limiting RNA. As a standard for binding affinity the natural ligand FMN (Kd = 1.75 nM) was spiked in to the pool of 16 compounds. The data are plotted by quantifying the lowest RNA concentration at which ligand binding is observed to the FMN riboswitch WT. After FMN, the strongest affinity ligand for the FMN riboswitch WT is ribocil (W-1, Figure 3B), the lead compound from the phenotypic screen18. Compound W-2 is a potent analog of ribocil that was originally missed in the phenotypic screen (false negative), but later identified through analog searches of ribocil. Compound W-3 is roseoflavin, an antimicrobial natural product known to target the FMN riboswitch 32. Compounds W-4 to W-16 have weaker affinities than roseoflavin (Kd > 300 nM). The affinities of compounds that are not detected in this experiment (i.e. W-9 to W-16) are likely very low since these compounds do not bind to the FMN riboswitch when in competition with avid ligands such as FMN and ribocil. Notably, none of the compounds besides the ribocil-analog is structurally similar to ribocil according to the Tanimoto similarity score (Table S1). In fact, of the 124 total binders to all three RNAs screened (Figure 3A), only 29 compounds contain one or more binding compound with a similar structure. The remaining 95 compounds that bind to one of the screened targets are structurally unique. All of the non-ribocil species show relatively low similarity to ribocil, as judged by their Tanimoto similarity score. This indicates that the RNA is capable of binding a rather diverse array of chemical types. These results indicate the successful adaptation of PT-ALIS to accurately rank high-affinity RNA-binding compounds. Since ribocil is the lead selective compound after a combination of comparative screening and affinity ranking, we have shown ALIS to be a complementary, target-based approach to phenotypic screening that originally identified ribocil18. Characterization of compounds that bind to both FMN riboswitch WT and G93U mutant In an effort to circumvent the riboswitch-mediated mechanism of ribocil resistance, the compounds that bind to both the FMN riboswitch WT and the G93U mutant were analyzed by PT-ALIS. The FMN riboswitch WT and the G93U mutant were each titrated (0.31-20 µM) into a pool of 14 compounds (1 µM/compound) that bound to both aptamers (Figure 3A, WG compounds). FMN was again used as a standard, and the data were analyzed as above. The 9 ACS Paragon Plus Environment

Page 11 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


highest affinity compound, WG-1, binds to the FMN riboswitch WT and G93U mutant with a similar affinity as FMN (Figure 3C), while other compounds show lower binding affinities to both the FMN riboswitch WT and G93U. An ALIS competition assay was then run to classify compound binding for WG-1 and WG-3 (Figure 4A; WG-2 was not evaluated due to its weak MS signal in ALIS). Increasing concentrations of FMN (0-50 µM) displace the binding of WG-3 and ribocil (Figure 4B). This indicates WG-3 competes directly with FMN for binding to the FMN riboswitch WT. While WG-1 also competes with FMN, its displacement is incomplete suggesting that WG-1 may be binding non-specifically at an additional binding site or sites, or with a different stoichiometry. In the presence of the RNA denaturant urea41, 42, binding of FMN, ribocil and WG-3 to the FMN riboswitch is fully abolished, but not the binding of WG-1 (Figure S2). The ability of WG-1 to bind in the presence of high urea concentration suggests that compound binding is not completely specific to the tertiary RNA structure required for FMN and ribocil binding18. These findings are further supported by SPR analysis of the binding kinetics of WG-1 to the FMN riboswitch (Supplementary Results). In comparison, compounds that bind to all three riboswitch constructs including the Scramble FMN (Figure 3A, 15 WGS compounds) do not bind competitively with FMN (Figure S3). These results corroborate the lack of specificity of these compounds and are consistent with their ability to also bind to the Scramble FMN, suggesting binding is not fully dependent upon the nature of the tertiary folded structure of the RNA. Additionally, some WGS compounds appear to be non-specific intercalators when tested in an intercalation assay (Figure S4). The activity of FMN-competitive compounds WG-1 and WG-3 was measured in a bacterial growth inhibition assay. Since the riboflavin biosynthetic pathway is conditionally essential in the absence of exogenous riboflavin18, 26, 38, inhibition by WG-1 and WG-3 was measured in the absence and presence of exogenous riboflavin to study bioactivity specific to the riboflavin biosynthetic pathway (regulated by the FMN riboswitch). Surprisingly, neither WG-1 nor WG-3 exhibit riboflavin-dependent antibacterial activity in cells, as evidenced by their potent antibacterial activity (MIC = 0.1 – 3.1 µM) in the presence of sufficient exogenous riboflavin to render the riboflavin pathway non-essential (Figure 4C). The activities of the other WG (and W) compounds were also reexamined, and none proved to have FMN-dependent antibacterial activity, as predicted from a previous phenotypic screen18 (data not shown). In 10 ACS Paragon Plus Environment


Page 12 of 30

contrast, ribocil is active only in the absence of exogenous riboflavin, implicating that its target is in the riboflavin pathway, consistent with previous studies18. Although WG-1 and WG-3 bind to the FMN riboswitch competitively and with high affinity, their inactivity against the FMN riboswitch may be due to their binding failing to mimic the bioactivity of the natural ligand FMN in downregulating riboflavin production, or their activity against the FMN riboswitch may be masked by effects against other target(s) not related to riboflavin biosynthesis. Although we did not observe bioactivity of WG-1 and WG-3 against targets in the riboflavin pathway, both compounds display potent activity against the bacterial cells. WG-1 is a close analog of ellipticine 43, a DNA-intercalating antitumor agent that inhibits DNA topoisomerase II44, phosphorylation of p53 in Lewis lung carcinoma cells 45, and P-450-1A146. This is supported by data which show that WG-1 is in fact an intercalator (Figure S4), which is in agreement with the incomplete dissociation of WG-1 in ALIS competition studies (Figure 4B), sustained binding in the presence of a denaturant (Figure S2), and SPR kinetic studies (Supplementary Results). WG-3 is in the quinolone class of antibacterials 47 that inhibit DNA gyrase, topoisomerase II or IV 48. WG-3 activity was therefore tested in an E. coli strain with a mutation in the hotspot region of Gyrase A (D87N) conferring resistance to quinolones49. A 4fold decrease in WG-3 potency was seen in the GyrA mutant when compared to the WT strain (Figure S5), suggesting WG-3 acts at least in part via gyrase inhibition. These data suggest that WG-3 may be binding to both protein and structured RNA targets. Similar to the antibiotic roseoflavin which is known to bind and be active against the FMN riboswitch 32 as well as other mechanisms 50, 51, WG-1 and WG-3 may be exhibiting polypharmacology against multiple targets (supported by the fact that both WG-1 and WG-3 are both planar, strongly stacking molecules that may suggest a variety of non-specific interactions), which may be masking their riboflavin-dependent activity against the FMN riboswitch. NMR and SHG studies reveal distinct conformations of the FMN riboswitch WT when bound to WG-1 and WG-3 To explore the hypothesis that WG-1 and WG-3 lack activity against the FMN riboswitch due to differences in ligand-induced conformations of the FMN riboswitch compared to those induced by actives such as FMN or ribocil, 1D 1H-NMR studies were conducted on the FMN riboswitch bound to WG-1, WG-3, ribocil and FMN. Analysis of the aromatic proton region of the FMN 11 ACS Paragon Plus Environment

Page 13 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


riboswitch RNA spectra (6.5-7.5 ppm) reveals differences in peak shape and chemical shifts when FMN, ribocil, or WG-1 is bound compared to the unbound (apo) FMN riboswitch (Figure S6). These results imply the FMN riboswitch adopts different conformations when bound to each of these ligands. There is also a slight shift in the WG-3-bound state, but this change is not as substantial compared to other bound ligands. The observed peaks are from the RNA only; they are not well-resolved likely as a result of the large RNA size and conformational flexibility. Second-harmonic generation (SHG) was also used to compare FMN riboswitch conformational changes upon binding of WG-1, WG-3, ribocil and FMN. The non-specific binder WGS-1 was used as a negative control. SHG has been used to study ligand-induced conformational changes in proteins 52, 53 and RNA54 by measuring the SHG intensity produced by an SHG dye attached to the biomolecule tethered to a surface. Because the intensity of the SHG signal is highly sensitive to the angular orientation of the dye relative to the surface plane, this technique can resolve conformational changes that alter the net, average dye orientation. In agreement with the NMR results, SHG signals show significantly distinct conformational states of the FMN riboswitch when bound to each WG-1, WG-3, ribocil, and FMN compared to the apo FMN riboswitch, plotted as the percent (%) change in SHG intensity from apo RNA (Figure S6). FMN and ribocil induce a similar increase in SHG signal upon binding. In contrast, WG-3 induces a small decrease in SHG signal, suggesting that this compound induces a different conformation than FMN or ribocil. Finally, WG-1 induces a larger magnitude decrease in SHG signal, indicating a third distinct conformational state. As expected, non-specific binder WGS-1 does not induce a RNA-conformational change upon binding to the FMN riboswitch. Importantly, compared to the RNA conformation of the FMN-bound state, the WG-1-bound and WG-3-bound states are also significantly different, whereas the conformation of the ribocilbound state is not (Table S2). In agreement with SHG analysis, previously reported structural analysis of the ribocil-bound FMN riboswitch revealed it binds in the same binding pocket as FMN, and ribocil-binding has key similarities to FMN-binding in participating RNA residues 18. Here, we show that compounds WG-1 and WG-3 induce distinct conformational changes compared to the FMN- bound (and ribocil-bound) state, which may explain the lack of riboflavin-dependent activity of these compounds.



Page 14 of 30

Structure of WG-3–bound FMN riboswitch A 2.8 Å co-crystal structure of WG-3 bound to the Fusobacterium nucleatum impX FMN riboswitch was solved (attempts to co-crystalize the E. coli FMN riboswitch with WG-3 were unsuccessful, as previously reported with other ligands 18, 27, 37). Like both FMN27 and ribocil18, WG-3 is positioned inside the junctional region of the six RNA stems (Figure 5A). Similar conformations in portions of the global structures of FMNbound and the apo form55 are consistent with the global structure of the WG-3-bound FMN riboswitch. Whereas ribocil and FMN both stack face-to-face between the A-48 and A-85 bases, A-48 is flipped when bound to WG-3 and instead, WG-3 stacks face-to-face between A-48 and G-62 (Figure 5B; Figure S7; Figure S8). In fact, this conformational rearrangement of A-48 resembles that which is observed in the apo form55. Interestingly, the fact that WG-3 does not display riboflavin-dependent activity is likely related to the observation that the WG-3 costructure is more alike the apo conformation. A similar off-pathway conformation was observed upon fragment binding to the of the TPP riboswitch56, suggesting that despite ligand binding in the natural ligand binding site, non-natural ligands can stabilize alternative structures in RNA. The only regions of the FMN riboswitch that show conformational changes upon FMN binding are the junctional regions that immediately contact FMN. FMN-induced stabilization of these junctional regions prevents pairing of the switching sequence (J1/2) with the complementary sequence in the downstream expression platform to form the anti-terminator sequence55. Like ribocil or FMN, binding of WG-3 complements many stacking interactions with polar hydrogen bonds to several nucleotides in junctional regions of the FMN riboswitch, including G-10, G11, G-47, G-62, and A-84 (Figure 5C). However, both ribocil and FMN interact edge-to-face with key base A-49, whereas WG-3 largely interacts face-to-face. Distinct interactions specific to WG-3 binding may affect the stability of the bound-state riboswitch, or may be incompatible with the necessary rearrangement of the expression platform resulting in constitutive expression of the downstream ribB ORF and increased biosynthesis of riboflavin38, which is non-toxic to bacteria18. This is the first example of a FMN riboswitch structure in which the ligand induces an RNA conformation structurally distinct from that found in complex with ribocil and FMN. As seen with FMN, some of the interactions between WG-3 and the FMN riboswitch are mediated by a metal ion. This metal ion is assigned as Mg2+, which directly coordinates with the 13 ACS Paragon Plus Environment

Page 15 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


carboxylate group and the ring carbonyl oxygen of WG-3 (Figure 5C). Four additional water molecules are presumed to complete the coordination sphere for the Mg2+ ion but were not built in to the crystallographic model due to their low resolution. The specific importance of Mg2+ for coordinating WG-3 and RNA was confirmed by additional crystal soaking experiments in which Mg2+ was substituted by Ba2+ and Mn2+. The removal of Mg2+ leads to the diffusion of WG-3 out of the crystal and no ligand density is observed. These interactions are not surprising, given that quinolones such as WG-3 are known to form metal complexes via the carboxyl group and Mg2+ facilitates RNA-ligand interactions including FMN binding to the FMN riboswitch, TPP binding to the TPP riboswitch, and cyclic diguanylate (c-di-GMP) binding to the c-di-GMP riboswitch 57, 58

. This work presents the structure and binding of an FMN riboswitch ligand which is

structurally distinct from both ribocil and the natural ligand, FMN. From the comparison, a common pattern in ligand-RNA interactions emerges, where specificity arises through π-stacking interactions and H-bonds.

Conclusions This work demonstrates that the ALIS platform is a compatible and powerful tool for screening structured RNA and for characterizing RNA-small molecule interactions. We have shown that known binders such as natural riboswitch ligands are selectively detectable in ALIS. Using the FMN riboswitch, this work validates ALIS compound competition assays and affinity ranking analysis for accurate and reproducible characterization of small molecule RNA-binders. Furthermore, RNA targets can be successfully screened in ALIS against large compound collections as compound mixtures. By comparative screening of control RNA targets, nonselective compounds can be readily distinguished. In the case of the FMN riboswitch, we successfully identified ribocil as a selective and high-affinity binder, complementary to the phenotypic screen used to discover ribocil18 as an antibacterial compound acting against a structured RNA target (Figure 6). In addition to identifying ribocil as a binder to the FMN riboswitch, ALIS screening identified potent ribocil-analogs that were missed through phenotypic screening approaches. This work shows the importance of biophysical screening methods such as ALIS, which are not only complementary to functionally-based screens, but also provide an efficient, target-based approach to investigating RNA-small molecule 14 ACS Paragon Plus Environment


Page 16 of 30

interactions. In addition to recapitulating the discovery of ribocil in ALIS, we have identified novel compounds (WG-1 and WG-3) that bind competitively and with high-affinity to the FMN riboswitch. The structural diversity of these compounds indicates that the FMN riboswitch has the ability to bind a range of chemically diverse, drug-like small molecules, yet with varied functional activity. Crystal structure analysis of the FMN riboswitch bound to WG-3 reveals distinct ligand-bound RNA conformations and small changes in riboswitch conformation related to inactivity of WG-3 (Table 1). From this work we learn that RNA shows conformational flexibility that can be exploited for the discovery of RNA-targeting small molecules. Furthermore, by using a biophysical approach like ALIS to identify multiple structural diverse, high-affinity compounds binding to the FMN riboswitch, we have uncovered structure-function relationships important to targeting RNA in a phenotypically useful way. As the role of RNA in crucial biological function continues to be studied, ALIS screening and the comparison of active and inactive small molecule binders will aid in target-based drug design that can be applied to a broad range of disease-relevant RNA targets.


Page 17 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Materials and Methods Complete experimental methods may be found in the Supporting Information section.

Acknowledgements The authors would like to thank A. Buevich for compound NMR on WG-3, A. Brunskill for small molecule crystallography on WG-3, C. Andrews for advice on affinity ranking experiments, R. Boinay for compound management, and G. Schroeder for expertise on SPR kinetic analysis. Supporting Information Supporting information is available free of charge on the ACS Publications website at http://pubs.acs.org. Materials and methods section; Description of SPR results on WG-1 and WG-3; Supplementary Figure S1 (SPR results), Figure S2 (urea denaturation results), Figure S3 (ALIS competition studies on WGS compounds), Figure S4 (DNA intercalator assay results), Figure S5 (Gyrase and MIC assay results), Figure S6 (NMR results for WG-1 and WG-3, Figure S7 (illustration of base-stacking of WG-3 vs ribocil on FMN riboswitch), Figure S8 (electron density difference map of ligand WG-3); Supplementary Table S1 (structural comparison of hits from Figure 3A), Table S2 (Dunnett’s comparisons tests on SHG data), Table S3 (riboswitch RNA WT and mutant sequences), Table S4 (RNA sequences for FMN Scramble FMN, SAH, TPP, SAM, and B12 riboswitches), Table S5 (RNA sequences for FMN used in SHG experiments), and Table S6 (X-ray crystal data collection and refinement statistics).

Author contributions N.F.R. performed all ALIS experiments, led all follow-up experiments, and prepared the manuscript. J.A.H. generated the FMN riboswitch RNA constructs and conducted microbiological studies. A.N., D.J.K., and J.B. provided intellectual input to the experimental design and overall ncRNA drug discovery studies. N.F.R., A.H, and T.O.F generated the x-ray co-crystal structural data. H-Y.K. and M.M. performed the NMR experiments. N.F.R. and S.S.W. conducted microbiological studies. M. P. R. and C.C. aided in ALIS setup and screening. P.S. performed compound sample preparation for ALIS screening. M.T.B. and G.M. conducted the SHG conformational analysis. C.S. directed the x-ray co-crystal structure determination.



Page 18 of 30

P.J.D. advised on ALIS experiments. G.F.S and E.N. directed and provided intellectual input throughout the overall ncRNA drug discovery studies.


Page 19 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figures

Figure 1: ALIS determination of the specific binding of FMN-, SAH-, TPP-, SAM- and B12riboswitches to their native ligands. Each riboswitch was tested against its natural ligand, and a pool of all ligands. A structure of each native ligand is shown above its riboswitch. Data represents ALIS binding MS signals that are normalized to the signal from each native ligand. Each sample was run in duplicate and the experiment was repeated twice.



Page 20 of 30

Figure 2: Development and validation of ALIS methods with the FMN riboswitch. A) Structures of the natural and synthetic ligands that bind to and regulate the FMN riboswitch. B) Affinities of FMN, roseoflavin, ribocil and ribocil-C to the FMN riboswitch, as determined by ALIS. Data represent ALIS binding MS signals that are normalized to the signal from the highest concentration of each ligand. Error bars denote standard deviations of two injections from one ALIS experiment. Kd values are calculated and reported as an average and standard deviation of three replicate experiments. * Literature values determined by in-line probing or fluorescence quenching assays18, 27, 32, 33. C) FMN competition experiment with roseoflavin, ribocil, and ribocil-C for binding to the FMN riboswitch in ALIS. Data represent ALIS binding MS signals that are normalized to the signal from the highest concentration of each ligand (0.5 µM for roseoflavin/ribocil/ricobil-C; 50 µM for FMN). Error bars denote standard deviation of two ALIS injections from a single experiment. Experiment was replicated twice. D) FMN and ribocil binding to the FMN riboswitch and seven ribocil-resistant mutants. MS signal of each mutant is normalized to the MS signal of the FMN riboswitch WT for each ligand. Data represent ALIS binding MS signals that are normalized to the FMN riboswitch WT signal for each ligand. Error bars represent standard deviation of two ALIS injections. Experiment was replicated twice and significance is calculated using Student’s t-test. * p < 0.05


Page 21 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 3: A) ALIS screening of the FMN riboswitch WT, G93U mutant, and Scramble FMN against ~53,000 small molecules with antibacterial activity. Data are a result of two replicate primary screens in ALIS using pooled compound mixtures and a confirmation screen as single compounds run in duplicate injections. Affinity ranking of compounds that B) bind only to the FMN riboswitch WT (W compounds from Figure 3A) and C) bind to both the WT and G93U mutant (WG compounds from Figure 3A) by RNA titration using PT-ALIS. The target riboswitch concentration was varied from 0.3 to 20 µM, and the individual compound component concentrations in the ligands were set at 1 uM per compound. FMN was added in as a reference, which has a high affinity (Kd = 1.75 nM) to the FMN riboswitch WT. Data represent relative affinity determined by the lowest RNA concentration at which the compound was observed to bind. Experiment was run in two replicate ALIS injections, and the experiment was repeated twice.



Page 22 of 30

Figure 4: Characterization of WG-1 and WG-3. A) Structures of WG-1 and WG-3. B) Competition experiments with compounds that bind the FMN riboswitch WT competitively with FMN. Highest affinity compounds WG-1 and WG-3 were tested. Excess FMN riboswitch WT (2 µM) was annealed with a constant concentration (0.5 µM) of WG-1, WG-3, or ribocil (positive control), and FMN was added at increasing concentrations (0-59 uM) to displace the other ligands. Data represents ALIS binding MS signals that are normalized to the signal from the highest concentration of each ligand (1 µM for WG compounds; 50 µM for FMN). Error bars denote standard deviation of two ALIS injections from a single experiment. Experiment was replicated twice. C) Minimum inhibitory concentration (MIC) of antibacterial compound (WG-1, WG-3 or ribocil) needed to suppress growth of E. coli (strain MB5746) in the absence and presence of exogenous riboflavin (20 µM). Compounds that have activity specifically against the riboflavin pathway are rescued by the addition of exogenous riboflavin (i.e. ribocil).


Page 23 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 5: X-ray structure of the F. nucleatum FMN riboswitch with WG-3. A) Overall structure of the WG-3-bound FMN riboswitch aptamer. The aptamer is represented as a cartoon and colored by P-loop domain as labeled. WG-3 is represented as red spheres. B) Base stacking of FMN riboswitch when bound to WG-3, FMN, ribocil, and the apo form. RNA bases are colored blue, carbon atoms colored green for FMN, cyan for ribocil, and pink for WG-3. C) Interactions of WG-3 at the binding site. RNA bases are colored blue, carbon atoms colored in pink for WG3, and Mg2+ is represented as a yellow sphere. Interactions are shown with black dashes between WG-3 and bases and red between WG-3 and Mg2+.



Page 24 of 30

Figure 6. ALIS screening is a target-based approach to screen for small molecules that bind to RNA. This method (green steps) is a biophysical approach that is complimentary to phenotypic screening (blue steps).


Page 25 of 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Tables

Binding properties (ALIS) Compound Competition with FMN

FMN

-

Conformation

Affinity to WT riboswitch Specificity (nM)

1.2

Kinetics

Activity Phenotypic screen

Intercalator?

-

Very slow koff Shifted spectrum Significantly different and compared to conformation than apo is compound apo RNA is potent

-

Not tested

Shifted Significantly different spectrum & Faster koff rate conformation than additional peak apo, but not compared to compared to significantly different FMN apo RNA and than FMN-bound FMN-bound

Riboflavindependent activity

No

NMR

SHG

SPR studies

Ribocil

Competitive

50

WT only

Ribocil-analog

Competitive

~ 200

WT only

Not tested

Same conformation as ribocil-bound

Similar kon as ribocil but a faster koff

Riboflavindependent activity

Not tested

Roseoflavin

Competitive

200

WT only

Not tested

Significantly different Rapid k /k on off conformation than apo rates and FMN-bound

Not tested

Not tested

WG-1

Competitive, but incomplete dissociation

Binding does not saturate

WG-3

Competitive (confirmed with co-crystal structure)

130

WGS-1

Noncompetitive

High affinity

Inconclusive data (did not fit 1:1 binding model). Somewhat Not Dissociation is shifted Significantly different heterogeneous riboflavinWT & Yes conformation than apo spectrum dependent mutant (biphasic) compared to and FMN-bound MIC = 3.1 uM indicating apo RNA multiple binding sites/intercalation Shifted Partially Not spectrum & Inconclusive (showed that it Significantly different data (did not fit riboflavin- is partially acting WT & additional peak conformation than apo dependent mutant compared to 1:1 binding as an and FMN-bound apo RNA and model). MIC

Discovery of selective RNA-binding small molecules by affinity

Recommend Documents