87208811 1 A COVID Moonshot: assessment of ligand binding to the SARS-CoV-2 main protease by saturation 1 transfer difference NMR spectroscopy 2 3 Anastassia L. Kantsadi1, Emma Cattermole1, Minos-Timotheos Matsoukas2, Georgios A. Spyroulias2 4 and Ioannis Vakonakis1* 5 1Department of Biochemistry, University of Oxford, South Parks Road, Oxford OX1 3QU, United 6 Kingdom 7 2Department of Pharmacy, University of Patras, Panepistimioupoli Campus, GR-26504, Greece 8 *To whom correspondence should be addressed, e-mail: ioannis.vakonakis@bioch.ox.ac.uk, Tel.: 9 +44 1865 275725, Fax: +44 1865 613201 10 11 Short title: Assessment of ligand binding to SARS-CoV-2 Mpro by STD-NMR 12 Keywords: SARS-CoV-2, COVID-19, Moonshot, Mpro, NMR, STD, screening, fragments, molecular 13 dynamics, MD, competition 14 15 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 2 Abstract 16 Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is the etiological cause of the 17 coronavirus disease 2019, for which no effective therapeutics are available. The SARS-CoV-2 main 18 protease (Mpro) is essential for viral replication and constitutes a promising therapeutic target. Many 19 efforts aimed at deriving effective Mpro inhibitors are currently underway, including an international 20 open-science discovery project, codenamed COVID Moonshot. As part of COVID Moonshot, we used 21 saturation transfer difference nuclear magnetic resonance (STD-NMR) spectroscopy to assess the 22 binding of putative Mpro ligands to the viral protease, including molecules identified by 23 crystallographic fragment screening and novel compounds designed as Mpro inhibitors. In this 24 manner, we aimed to complement enzymatic activity assays of Mpro performed by other groups with 25 information on ligand affinity. We have made the Mpro STD-NMR data publicly available. Here, we 26 provide detailed information on the NMR protocols used and challenges faced, thereby placing these 27 data into context. Our goal is to assist the interpretation of Mpro STD-NMR data, thereby accelerating 28 ongoing drug design efforts. 29 30 31 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 3 Introduction 32 Infections by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) resulted in 33 approximately 1.8 million deaths in 2020 (1) and led to the coronavirus 2019 (COVID-19) pandemic 34 (2-4). SARS-CoV-2 is a zoonotic betacoronavirus highly similar to SARS-CoV and MERS-CoV, which 35 caused outbreaks in 2002 and 2012, respectively (5-7). SARS-CoV-2 encodes its proteome in a single, 36 positive-sense, linear RNA molecule of ~30 kb length, the majority of which (~21.5 kb) is translated 37 into two polypeptides, pp1a and pp1ab, via ribosomal frame-shifting (8, 9). Key viral enzymes and 38 factors, including most proteins of the reverse-transcriptase machinery, inhibitors of host translation 39 and molecules signalling for host cell survival, are released from pp1a and pp1ab via post-40 translational cleavage by two viral cysteine proteases (10). These proteases, a papain-like enzyme 41 cleaving pp1ab at three sites, and a 3C-like protease cleaving the polypeptide at 11 sites, are primary 42 targets for the development of antiviral drugs. 43 The 3C-like protease of SARS-CoV-2, also known as the viral main protease (Mpro), has been the 44 target of intense study owing to its centrality in viral replication. Mpro studies have benefited from 45 previous structural analyses of the SARC-CoV 3C-like protease and the earlier development of 46 putative inhibitors (11-14). The active sites of these proteases are highly conserved, and 47 peptidomimetic inhibitors active against Mpro are also potent against the SARS-CoV 3C-like protease 48 (15, 16). However, to date no Mpro-targeting inhibitors have been validated in clinical trials. In order 49 to accelerate Mpro inhibitor development, an international, crowd-funded, open-science project was 50 formed under the banner of COVID Moonshot (17), combining high-throughput crystallographic 51 screening (18), computational chemistry, enzymatic activity assays and mass spectroscopy (19) 52 among the many methodologies contributed by collaborating groups. 53 As part of COVID Moonshot, we utilised saturation transfer difference nuclear magnetic 54 resonance (STD-NMR) spectroscopy (20-22) to investigate the Mpro binding of ligands initially 55 identified by crystallographic screening, as well as molecules designed specifically as non-covalent 56 inhibitors of this protease. Our goal was to provide orthogonal information on ligand binding to that 57 which could be gained by enzymatic activity assays conducted in parallel by other groups. STD-NMR 58 is a proven method for characterising the binding of small molecules to biological macromolecules, 59 able to provide both quantitative affinity information and structural data on the proximity of ligand 60 chemical groups to the protein. Here, we provide detailed documentation on the NMR protocols 61 used to record these data and highlight the advantages, limitations and assumptions underpinning 62 our approach. Our aim is to assist the comparison of Mpro STD-NMR data with other quantitative 63 measurements, and facilitate the consideration of these data when designing future Mpro inhibitors. 64 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 4 Materials and Methods 65 Protein production and purification 66 We created a SARS-CoV-2 Mpro genetic construct in pFLOAT vector (23), encoding for the viral 67 protease and an N-terminal His6-tag separated by a modified human rhinovirus (HRV) 3C protease 68 recognition site, designed to reconstitute a native Mpro N-terminus upon HRV 3C cleavage. The Mpro 69 construct was transformed into Escherichia coli strain Rosetta(DE3) (Novagen) and transformed 70 clones were pre-cultured at 37 °C for 5 h in lysogeny broth supplemented with appropriate 71 antibiotics. Starter cultures were used to inoculate 1 L of Terrific Broth Autoinduction Media 72 (Formedium) supplemented with 10% v/v glycerol and appropriate antibiotics. Cell cultures were 73 grown at 37 °C for 5 h and then cooled to 18 °C for 12 h. Bacterial cells were harvested by 74 centrifugation at 5,000 x g for 15 min. 75 Cell pellets were resuspended in 50 mM trisaminomethane (Tris)-Cl pH 8, 300 mM NaCl, 10 mM 76 imidazole buffer, incubated with 0.05 mg/ml benzonase nuclease (Sigma Aldrich) and lysed by 77 sonication on ice. Lysates were clarified by centrifugation at 50,000 x g at 4 °C for 1 h. Lysate 78 supernatants were loaded onto a HiTrap Talon metal affinity column (GE Healthcare) pre-79 equilibrated with lysis buffer. Column wash was performed with 50 mM Tris-Cl pH 8, 300 mM NaCl 80 and 25 mM imidazole, followed by protein elution using the same buffer and an imidazole gradient 81 from 25 to 500 mM concentration. The His6-tag was cleaved using home-made HRV 3C protease. The 82 HRV 3C protease, His6-tag and further impurities were removed by a reverse HiTrap Talon column. 83 Flow-through fractions were concentrated and applied to a Superdex75 26/600 size exclusion 84 column (GE Healthcare) equilibrated in NMR buffer (150 mM NaCl, 20 mM Na2HPO4 pH 7.4). 85 86 Nuclear magnetic resonance (NMR) spectroscopy 87 All NMR experiments were performed using a 950 MHz solution-state instrument comprising an 88 Oxford Instruments superconducting magnet, Bruker Avance III console and TCI probehead. A Bruker 89 SampleJet sample changer was used for sample manipulation. Experiments were performed and 90 data processed using TopSpin (Bruker). For direct STD-NMR measurements, samples comprised 10 91 μM Mpro and variable concentrations (20 μM – 4 mM) of ligand compounds formulated in NMR 92 buffer supplemented with 10% v/v D2O and deuterated dimethyl sulfoxide (D6-DMSO, 99.96% D, 93 Sigma Aldrich) to 5% v/v final D6-DMSO concentration. In competition experiments, samples 94 comprised 2 μM Mpro, 0.8 mM of ligand x0434 and variable concentrations (0 – 20 μM) of competing 95 compound in NMR buffer supplemented with D2O and D6-DMSO as above. Sample volume was 140 96 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 5 μL and samples were loaded in 3 mm outer diameter SampleJet NMR tubes (Bruker) placed in 96-97 tube racks. NMR tubes were sealed with POM balls. 98 STD-NMR experiments were performed at 10 oC using a pulse sequence described previously (20) 99 and an excitation sculpting water-suppression scheme (24). Protein signals were suppressed in STD-100 NMR by the application of a 30 msec spin-lock pulse. We collected time-domain data of 16,384 101 complex points and 41.6 μsec dwell time (12.02 kHz sweepwidth). Data were collected in an 102 interleaved pattern, with on- and off-resonance irradiation data separated into 16 blocks of 16 103 transients each (256 total transients per irradiation frequency). Transient recycle delay was 4 sec and 104 on- or off-resonance irradiation was performed using 0.1 mW of power for 3.5 sec at 0.5 ppm or 26 105 ppm, respectively, for a total experiment time of approximately 50 minutes. Reconstructed time-106 domain data from the difference of on- and off-resonance irradiation (STD spectra) or only the off-107 resonance irradiation (reference spectra) were processed by applying a 2 Hz exponential line 108 broadening function and 2-fold zero-filling prior to Fourier transformation. Phasing parameters were 109 derived for each sample from the reference spectra and copied to the STD spectra. 1H peak 110 intensities were integrated in TopSpin using a local-baseline adjustment function. Data fitting to 111 extract Kd values were performed in OriginPro (OriginLab). The folded state of M pro in the presence 112 of each ligand was verified by collecting 1H NMR spectra similar to Fig. 1A from all samples ahead of 113 STD-NMR experiments. 114 115 Ligand handling 116 Compounds for the initial STD-NMR assessment of crystallographic fragment binding to Mpro were 117 provided by the XChem group at Diamond Light Source in the form of a 384-well plated library (DSI-118 poised, Enamine), with compounds dissolved in D6-DMSO at 500 mM nominal concentration. 1 μL of 119 dissolved compounds was aspirated from this library and immediately mixed with 9 μL of D6-DMSO 120 for a final fragment concentration of 50 mM, from which NMR samples were formulated. For 121 titrations of the same crystallographic fragments compounds were procured directly from Enamine 122 in the form of lyophilized powder, which was dissolved in D6-DMSO to derive compound stocks at 10 123 mM and 100 mM concentrations for NMR sample formulation. 124 STD-NMR assays of bespoke Mpro ligands used compounds commercially synthesised for COVID 125 Moonshot. These ligands were provided to us by the XChem group in 96-well plates, containing 0.7 126 μL of 20 mM D6-DMSO-disolved compound per well. Plates were created using an Echo liquid 127 handling robot (Labcyte) and immediately sealed and frozen at -20 oC. For use, ligand plates were 128 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 6 thoroughly defrosted at room temperature and spun at 3,500 g for 5 minutes. In single-129 concentration STD-NMR experiments, 140 μL of a pre-formulated mixture of Mpro and NMR buffer 130 with D2O and D6-DMSO were added to each well to create the final NMR sample. For STD-NMR 131 competition experiments, 0.5 μL of ligands were aspirated from the plates and immediately mixed 132 with 19.5 μL of D6-DMSO for final ligand concentration of 0.5 mM from which NMR samples were 133 formulated. 134 135 Molecular dynamics (MD) simulations 136 The monomeric complexes of Mpro bound to chemical fragments were obtained from the RCSB 137 Protein Data Bank entries 5R81 (ligand x0195), 5REB (x0387), 5RGI (x0397), 5RGK (x0426), 5R83 138 (x0434) and 5REH (x0540) for MD simulations with GROMACS version 2018 (25) and the 139 AMBER99SB-ILDN force field (26). All complexes were inserted in a pre-equilibrated box containing 140 water implemented using the TIP3P water model (26). Force field parameters for the six ligands 141 were generated using the general Amber force field and HF/6 – 31G*– derived RESP atomic charges 142 (27). The reference system consisted of the protein, the ligand, ~31,400 water molecules, 95 Na and 143 95 Cl ions in a 100 x 100 x 100 Å simulation box, resulting in a total number of ~98,000 atoms. Each 144 system was energy-minimized and subsequently subjected to a 20 ns MD equilibration, with an 145 isothermal-isobaric ensemble using isotropic pressure control (28), and positional restraints on 146 protein and ligand coordinates. The resulting equilibrated systems were replicated 4 times and 147 independent 200 ns MD trajectories were produced with a time step of 2 fs, in constant temperature 148 of 300 K, using separate v-rescale thermostats (28) for the protein, ligand and solvent molecules. 149 Lennard-Jones interactions were computed using a cut-off of 10 Å and electrosta�c interac�ons were 150 treated using particle mesh Ewald (29) with the same real-space cut-off. Analysis on the resulting 151 trajectories was performed using MDAnalysis (30, 31). Structures were visualised using PyMOL (32). 152 153 Notes 154 The enzymatic inhibition potential of Mpro ligands, measured by RapidFire mass spectroscopy 155 (17), was retrieved from the Collaborative Drug Discovery database (33). 156 157 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 7 Results 158 STD-NMR assays of M pro ligand binding 159 Mpro forms dimers in crystals via an extensive interaction interface involving two domains (15). 160 Mpro dimers likely have a sub-μM solution dissociation constant (Kd) by analogy to previously studied 161 3C-like coronavirus proteases (34). At the 10 μM protein concentration of our NMR assays Mpro is, 162 thus, expected to be dimeric with an estimated molecular weight of nearly 70 kDa. Despite the 163 relatively large size of Mpro for solution NMR, 1H spectra of the protease readily showed the presence 164 of multiple up-field shifted (<0.5 ppm) peaks corresponding to protein methyl groups (Fig. 1A). In 165 addition to demonstrating that Mpro is folded under the conditions tested, these spectra allowed us 166 to identify the chemical shifts of Mpro methyl groups that may be suitable for on-resonance 167 irradiation in STD-NMR experiments. Trials with on-resonance irradiation applied to different methyl 168 group peaks showed that irradiating at 0.5 ppm (Fig. 1A) produced the strongest STD signal from 169 ligands in the presence of Mpro, while simultaneously avoiding ligand excitation that would yield 170 false-positive signals in the absence of Mpro (Fig. 1B). Further, we noted that small molecules 171 abundant in the samples but not binding specifically to Mpro, such as DMSO, produced pseudo-172 dispersive residual signal lineshapes in STD spectra, while true Mpro ligands produced peaks in STD 173 with absorptive 1H lineshapes. We surmised that STD-NMR is suitable for screening ligand binding to 174 Mpro, requiring relatively small amounts (10-50 μgr) of protein and time (under 1 hour) per sample 175 studied. 176 The strength of STD signal is quantified by calculating the ratio of integrated signal intensity of 177 peaks in the STD spectrum over that of the reference spectrum (STDratio). The STDratio factor is 178 inversely proportional to ligand Kd, as �������� � � ��� � where [L] is ligand concentration. 179 Measuring STDratio values over a range of ligand concentrations allows fitting of the proportionality 180 constant and calculation of ligand Kd. However, time and sample-amount considerations, including 181 the limited availability of bespoke compounds synthesized for the COVID Moonshot project, made 182 recording full STD-NMR titrations impractical for screening hundreds of ligands. Thus, we evaluated 183 whether measuring the STDratio value at a single ligand concentration may be an informative 184 alternative to Kd, provided restraints could be placed, for example, on the proportionality constant. 185 Theoretical and practical considerations suggested that three parameters influence our 186 evaluation of single-concentration STDratio values towards an affinity context. Firstly, the STDratio 187 factor is affected by the efficiency of NOE magnetisation transfer between protein and ligand, which 188 in turn depends on the proximity of ligand and protein groups, and the chemical nature of these 189 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 8 groups (20-22). To minimize the influence of these factors across diverse ligands, we sought to 190 quantify the STDratio of only aromatic ligand groups, and only consider those showing the strongest 191 STD signal; thus, that are in closest proximity to the protein. Second, STD-NMR assays require ligand 192 exchange between protein-bound and -free states in the timeframe of the experiment; strongly 193 bound compounds that dissociate very slowly from the protein would yield reduced STDratio values 194 compared to weaker ligands that dissociate more readily. Structures of Mpro with many different 195 ligands show that the protein conformation does not change upon complex formation and that the 196 active site is fully solvent-exposed (18), which suggests that ligand association can proceed with high 197 rate (107 – 108 M-1s-1). Under this assumption, the ligand dissociation rate is the primary determinant 198 of interaction strength. Given the duration of the STD-NMR experiment in our assays, and the ratios 199 of ligand:protein used, we estimated that significant protein – ligand exchange will take place even 200 for interactions as strong as low-μM Kd. Finally, uncertainties or errors in nominal ligand 201 concentration skew the correlation of STDratio to compound affinities; as shown in Fig. S1, STDratio 202 values increase strongly when very small amounts of ligands are assessed. Thus, overly large STDratio 203 values may be measured if ligand concentrations are significantly lower than anticipated. 204 205 Quantitating M pro binding of ligands identified by crystallographic screening 206 Mindful of the limitations inherent to measuring single-concentration STDratio values, and prior to 207 using STD-NMR to evaluate bespoke Mpro ligands, we used this method to assess binding to the 208 protease of small chemical fragments identified in crystallographic screening experiments (18). In 209 crystallographic screening campaigns of other target proteins such fragments were seen to have 210 very weak affinities (> 1 mM Kd, e.g. (35)), thereby satisfying the exchange criterion set out above. 39 211 non-covalent Mpro interactors are part of the DSI-poised fragment library to which we were given 212 access, comprising 17 active site binders, two compounds targeting the Mpro dimerisation interface 213 and 20 molecules binding elsewhere on the protein surface (18). We initially recorded STD-NMR 214 spectra from these compounds in the absence of Mpro to confirm that we obtained no or minimal 215 STD signal when protease is omitted, and to verify ligand identity from reference 1H spectra. Five 216 ligands gave no solution NMR signal or produced reference 1H spectra inconsistent with the 217 compound chemical structure; these ligands were not evaluated further. Samples of 10 μM Mpro and 218 0.8 mM nominal ligand concentration were then formulated from the remaining 34 compounds 219 (Table S1), and STD-NMR spectra were recorded, from which only aromatic ligand STD signals were 220 considered for further analysis. 221 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 9 We observed large variations in STD signal intensity and STDratio values in the presence of M pro 222 across compounds (Fig. 2A,B; Table S1), with many ligands producing little or no STD signal, 223 suggesting substantial differences in compound affinity for the protease. However, we also noted 224 that ligand reference spectra different substantially in intensity (Fig. 2C), despite compounds being 225 at the same nominal concentration. Integrating ligand peaks in these reference spectra revealed 226 differences in per-1H intensity of up to ~15-fold, indicating significant variation of ligand 227 concentrations in solution (Table S1). Such concentration differences could arise from errors in 228 sample formulation or from concentration inconsistencies in the compound library. To evaluate the 229 former we also integrated the residual 1H signal of D6-DMSO in our reference spectra, and found it to 230 vary by less than 35% across any pair of samples (11% average deviation). As DMSO was added 231 alongside ligands in our samples, we concluded that sample formulation may have contributed 232 errors in compound concentration of up to ~1/3, but did not account for the ~15-fold differences in 233 concentration observed. 234 Given that differences in compound concentration can skew the relative STDratio values of ligands 235 (Fig. S1), and that such concentration differences were also observed among newly designed Mpro 236 inhibitors (see below), we questioned whether recording STDratio values under these conditions can 237 provide useful information. To address this question we attempted to quantify the affinity of 238 crystallographic fragments to Mpro, selecting ligands that showed clear differences in STDratio values 239 in the assays above and focusing on compounds binding at the Mpro active site; hence, that are of 240 potential interest to inhibitor development. We performed Mpro binding titrations monitored by STD-241 NMR of compounds x0195, x0354, x0426 and x0434 in 50 μM – 4 mM concentrations (Fig. S2), and 242 noted that only compounds x0434 and x0195, which show the highest STDratio (Fig. 2A), bound 243 strongly enough for an affinity constant to be estimated (Kd of 1.6 ± 0.2 mM and 1.7 ± 0.2 mM, 244 respectively). In contrast, the titrations of x0354 and x0426, which yielded lower STDratio values, 245 could not be fit to extract a Kd indicating weaker binding to M pro. 246 To further this analysis, we assessed the binding of fragments x0195, x0387, x0397, x0426, x0434 247 and x0540 to the Mpro active site using quadruplicate atomistic molecular dynamics (MD) simulations 248 of 200 nsec duration. As shown in Fig. S3A,B, and Movies S1 and S2, fragments with high STDradio 249 values (x0434 and x0195) always located in the Mpro active site despite exchanging between 250 different binding conformations (Fig. S4), with average ligand root-mean-square-deviation (RMSD) of 251 3.2 Å and 5.1 Å respectively after the first 100 nsec of simulation. Medium STDratio value fragments 252 (x0426 and x0540, Fig. S3C,D, and Movies S3 and S4) show average RMSDs of approximately 9 Å in 253 the same simulation timeframe, frequently exchanging to alternative binding poses and with x0540 254 occasionally exiting the Mpro active site. In contrast, fragments showing very little STD NMR signal 255 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 10 (x0397 and x0387, Fig. S3E,F, and Movies S5 and S6) regularly exit the Mpro active site and show 256 average RMSDs in excess of 15 Å with very limited stability. Combining the quantitative Kd and MD 257 information above, we surmised that, despite limitations inherent in this type of analysis and 258 uncertainties in ligand amounts, STDratio values recorded at single compound concentration can act 259 as proxy measurements of Mpro affinity for ligands. 260 261 Assessment of M pro binding by COVID Moonshot ligands 262 We proceeded to characterise by STD-NMR the Mpro binding of bespoke ligands created as part of 263 the COVID Moonshot project and designed to act as non-covalent inhibitors of the protease (17). 264 Similar to the assays of crystallographic fragments above, we focused our analysis of STD signals to 265 aromatic moieties of ligands binding to the Mpro active side and extracted STDratio values only from 266 the strongest STD peaks. Once again, we noted substantial differences in apparent compound 267 concentrations, judging from reference 1H spectral intensities (Fig. 3A), which could not be 268 attributed to errors in sample preparation as the standard deviation of residual 1H intensity in the 269 D6-DMSO peak did not exceed 5% in any of the ligand batches tested. Crucially, out of 650 different 270 molecules tested, samples of 35 compounds (7.6%) contained no ligand and 86 (13.2%) very little 271 ligand (Fig. 3A). In these cases, NMR assays were repeated using a separate batch of compound; 272 however, 96.2% of repeat experiments yielded the same outcome of no or very little ligand in the 273 NMR samples. 274 We measured STDratio values from samples were ligands produced sufficiently strong reference 1H 275 NMR spectra to be readily visible, and deposited these values and associated raw NMR data to the 276 Collaborative Drug Discovery database (33). Some of these ligands were assessed independently for 277 enzymatic inhibition of Mpro using a mass spectroscopy method as part of the COVID Moonshot 278 collaboration (17). Where both parameters are available, we compared the STDratio values and 50% 279 inhibition concentrations (IC50) of these ligands. As shown in Fig. 3B, STDratio and IC50 values show 280 weak correlation (R2=30%) for most ligands tested; however, a subset of ligands displayed 281 conspicuously low or even no STD signals considering their effect on Mpro activity, and presented 282 themselves as outliers in the correlation graph. As these outlier ligands had IC50 values below 10 μM, 283 suggesting that their affinities to the protease may be in the μM Kd region, we considered whether 284 our approach gives rise to false-negative STD results, for example through slow ligand dissociation 285 from Mpro. 286 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 11 To address this question, we derived an assay whereby the bespoke, high-affinity Mpro inhibitor 287 would outcompete a lower-affinity ligand known to provide strong STD signal from the protease 288 active site. In these experiments the lower-affinity ligand would act as ‘spy’ molecule whose STD 289 signal reduces as function of inhibitor concentration. We used fragment x0434, which yields 290 substantial STD signal with Mpro (Fig. 1B and 2A), as ‘spy’, and tested protease inhibitors EDJ-MED-291 a364e151-1, LON-WEI-ff7b210a-5, CHO-MSK-6e55470f-14 and LOR-NOR-30067bb9-11 as x0434 292 competitors. Of these inhibitors, EDJ-MED-a364e151-1 gave rise to substantial STD signal in earlier 293 assays, whereas the remaining produced little or no STD signal; yet, all four inhibitors were reported 294 to have low-μM or sub-μM IC50 values based on M pro enzymatic assays. In these competition 295 experiments, both EDJ-MED-a364e151-1 and LON-WEI-ff7b210a-5 yielded Kd parameters 296 comparable to the reported IC50 values (Fig. S5A,B), showing that at least in the case of LON-WEI-297 ff7b210a-5 the absence of STD signal in the single-concentration NMR assays above represented a 298 false-negative result. In contrast, CHO-MSK-6e55470f-14 and LOR-NOR-30067bb9-11 were unable to 299 compete x0434 from the protease active site (Fig. S5C,D), suggesting that in these two cases the 300 reported IC50 values do not reflect inhibitor binding to the protease, and that the weak STD signal of 301 the initial assays was a better proxy of affinity. We surmised that although some low STDratio values 302 of Mpro inhibitors may not accurately reflect compound affinity to the protease, such values cannot 303 be discounted as a whole as they may correspond to non-binding ligands. 304 305 306 307 308 309 310 311 312 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 12 Discussion 313 Fragment-based screening is a tried and tested method for reducing the number of compounds 314 that need to be assessed for binding against a specific target in order to sample chemical space (36). 315 Combined with X-ray crystallography, which provides information on the target site and binding 316 pose of ligands, initial fragments can quickly be iterated into potent and specifically-interacting 317 compounds. The COVID Moonshot collaboration (17) took advantage of crystallographic fragment-318 based screening (18) to initiate the design of novel inhibitors targeting the essential main protease 319 of the SARS-CoV-2 coronavirus; however crystallographic structures do not report on ligand affinity 320 and inhibitory potency in enzymatic assays does not always correlate with ligand binding. Thus, 321 supplementing these methods with solution NMR tools highly sensitive to ligand binding can provide 322 a powerful combination of orthogonal information and assurance against false starts. 323 We showed that STD-NMR is a suitable method for characterising ligand binding to Mpro, allowing 324 us to assess ligand interactions using relatively small amounts of protein and in under one hour of 325 experiment time per ligand (Fig. 1B). However, screening compounds in a high-throughput manner is 326 not compatible with the time- and ligand-amount requirements of full STD-NMR titrations. Thus, we 327 resorted to using an unconventional metric, the single-concentration STDratio value, as proxy for 328 ligand affinity. Although this metric has limitations due to its dependency on magnetisation transfer 329 between protein and ligand, and on relatively rapid exchange between the ligand-free and -bound 330 states, we demonstrated that it can nevertheless be informative. Specifically, the relative STDratio 331 values of chemical fragments bound to the Mpro active site provided insight on fragment affinity (Fig. 332 2A), as crosschecked by quantitative titrations (Fig. S2) and MD simulations (Fig. S3). Furthermore, 333 STDratio values of COVID Moonshot compounds held a weak correlation to enzymatic IC50 parameters 334 (Fig. 3B), although false-negative and -positive results from both methods contribute to multiple 335 outliers. Thus, in our view the biggest limitation of using the single-concentration STDratio value as 336 metric relates to its supra-linear sensitivity to ligand concentration (Fig. S1), which as demonstrated 337 here can vary substantially across ligands in a large project (Fig. 3A). 338 How then should the STD data recorded as part of COVID Moonshot be used? Firstly, we showed 339 that at least for some bespoke Mpro ligands the STDratio value obtained is a better proxy for 340 compound affinity compared to IC50 parameters from enzymatic assays (Fig. S5). This, inherently, is 341 the value of employing orthogonal methods thereby minimizing the number of potential false 342 results. Thus, when one is considering existing Mpro ligands to base the design of future inhibitors, a 343 high STDratio value as well as low IC50 parameters are both desirable. Second, due to the 344 aforementioned limitations of single-concentration STDratio value as proxy of affinity, and the 345 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 13 influence of uncertainties in ligand concentrations, we believe that comparisons of compounds and 346 derivatives differing by less than ~50% in STDratio is not meaningful. Rather, we propose that the 347 STDratio values of M pro ligands measured and available at the CDD database should be treated as a 348 qualitative metrics of compound affinity. 349 In conclusion, we presented here protocols for the assessment of SARS-CoV-2 Mpro ligands using 350 STD-NMR spectroscopy, and evaluated the relative qualitative affinities of chemical fragments and 351 compounds designed as part of COVID Moonshot. Although development of novel antivirals to 352 combat COVID-19 is still at an early stage, we hope that this information will prove valuable to 353 groups working towards such treatments. 354 355 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 14 References 356 1. WHO. Coronavirus disease 2019 [Available from: 357 https://www.who.int/emergencies/diseases/novel-coronavirus-2019. 358 2. Kucharski AJ, Russell TW, Diamond C, Liu Y, Edmunds J, Funk S, et al. Early dynamics of 359 transmission and control of COVID-19: a mathematical modelling study. Lancet Infect Dis. 360 2020;20(5):553-8. 361 3. Wu F, Zhao S, Yu B, Chen YM, Wang W, Song ZG, et al. A new coronavirus associated with 362 human respiratory disease in China. Nature. 2020;579(7798):265-9. 363 4. Zhu N, Zhang D, Wang W, Li X, Yang B, Song J, et al. A Novel Coronavirus from Patients with 364 Pneumonia in China, 2019. N Engl J Med. 2020;382(8):727-33. 365 5. Bermingham A, Chand MA, Brown CS, Aarons E, Tong C, Langrish C, et al. Severe respiratory 366 illness caused by a novel coronavirus, in a patient transferred to the United Kingdom from the 367 Middle East, September 2012. Euro Surveill. 2012;17(40):20290. 368 6. Kuiken T, Fouchier RA, Schutten M, Rimmelzwaan GF, van Amerongen G, van Riel D, et al. 369 Newly discovered coronavirus as the primary cause of severe acute respiratory syndrome. Lancet. 370 2003;362(9380):263-70. 371 7. Zaki AM, van Boheemen S, Bestebroer TM, Osterhaus AD, Fouchier RA. Isolation of a novel 372 coronavirus from a man with pneumonia in Saudi Arabia. N Engl J Med. 2012;367(19):1814-20. 373 8. Thiel V, Ivanov KA, Putics A, Hertzig T, Schelle B, Bayer S, et al. Mechanisms and enzymes 374 involved in SARS coronavirus genome expression. J Gen Virol. 2003;84(Pt 9):2305-15. 375 9. Bredenbeek PJ, Pachuk CJ, Noten AF, Charite J, Luytjes W, Weiss SR, et al. The primary 376 structure and expression of the second open reading frame of the polymerase gene of the 377 coronavirus MHV-A59; a highly conserved polymerase is expressed by an efficient ribosomal 378 frameshifting mechanism. Nucleic Acids Res. 1990;18(7):1825-32. 379 10. Hilgenfeld R. From SARS to MERS: crystallographic studies on coronaviral proteases enable 380 antiviral drug design. FEBS J. 2014;281(18):4085-96. 381 11. Ghosh AK, Xi K, Grum-Tokars V, Xu X, Ratia K, Fu W, et al. Structure-based design, synthesis, 382 and biological evaluation of peptidomimetic SARS-CoV 3CLpro inhibitors. Bioorg Med Chem Lett. 383 2007;17(21):5876-80. 384 12. Verschueren KH, Pumpor K, Anemuller S, Chen S, Mesters JR, Hilgenfeld R. A structural view 385 of the inactivation of the SARS coronavirus main proteinase by benzotriazole esters. Chem Biol. 386 2008;15(6):597-606. 387 13. Yang H, Yang M, Ding Y, Liu Y, Lou Z, Zhou Z, et al. The crystal structures of severe acute 388 respiratory syndrome virus main protease and its complex with an inhibitor. Proc Natl Acad Sci U S A. 389 2003;100(23):13190-5. 390 14. Yang H, Xie W, Xue X, Yang K, Ma J, Liang W, et al. Design of wide-spectrum inhibitors 391 targeting coronavirus main proteases. PLoS Biol. 2005;3(10):e324. 392 15. Zhang L, Lin D, Sun X, Curth U, Drosten C, Sauerhering L, et al. Crystal structure of SARS-CoV-393 2 main protease provides a basis for design of improved alpha-ketoamide inhibitors. Science. 394 2020;368(6489):409-12. 395 16. Rut W, Groborz K, Zhang L, Sun X, Zmudzinski M, Pawlik B, et al. SARS-CoV-2 M(pro) 396 inhibitors and activity-based probes for patient-sample imaging. Nat Chem Biol. 2020. 397 17. , Achdout H, Aimon A, Bar-David E, Barr H, Ben-Shmuel A, et al. COVID Moonshot: Open 398 Science Discovery of SARS-CoV-2 Main Protease Inhibitors by Combining Crowdsourcing, High-399 Throughput Experiments, Computational Simulations, and Machine Learning. bioRxiv. 2020. 400 18. Douangamath A, Fearon D, Gehrtz P, Krojer T, Lukacik P, Owen CD, et al. Crystallographic 401 and electrophilic fragment screening of the SARS-CoV-2 main protease. Nat Commun. 402 2020;11(1):5047. 403 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 15 19. El-Baba TJ, Lutomski CA, Kantsadi AL, Malla TR, John T, Mikhailov V, et al. Allosteric Inhibition 404 of the SARS-CoV-2 Main Protease: Insights from Mass Spectrometry Based Assays. Angew Chem Int 405 Edit. 2020. 406 20. Mayer M, Meyer B. Characterization of Ligand Binding by Saturation Transfer Difference 407 NMR Spectroscopy. Angew Chem Int Ed Engl. 1999;38(12):1784-8. 408 21. Becker W, Bhattiprolu KC, Gubensak N, Zangger K. Investigating Protein-Ligand Interactions 409 by Solution Nuclear Magnetic Resonance Spectroscopy. Chemphyschem. 2018;19(8):895-906. 410 22. Walpole S, Monaco S, Nepravishta R, Angulo J. STD NMR as a technique for ligand screening 411 and structural studies. Methods in Enzymology. 615: Elsevier; 2019. p. 423-51. 412 23. Rogala KB, Dynes NJ, Hatzopoulos GN, Yan J, Pong SK, Robinson CV, et al. The Caenorhabditis 413 elegans protein SAS-5 forms large oligomeric assemblies critical for centriole formation. Elife. 414 2015;4:e07410. 415 24. Hwang TL, Shaka AJ. Water Suppression That Works - Excitation Sculpting Using Arbitrary 416 Wave-Forms and Pulsed-Field Gradients. Journal of Magnetic Resonance Series A. 1995;112(2):275-417 9. 418 25. Abraham MJ, Murtola T, Schulz R, Páll S, Smith JC, Hess B, et al. GROMACS: High 419 performance molecular simulations through multi-level parallelism from laptops to supercomputers. 420 SoftwareX. 2015;1:19-25. 421 26. Lindorff-Larsen K, Piana S, Palmo K, Maragakis P, Klepeis JL, Dror RO, et al. Improved side-422 chain torsion potentials for the Amber ff99SB protein force field. Proteins. 2010;78(8):1950-8. 423 27. Bayly CI, Cieplak P, Cornell W, Kollman PA. A well-behaved electrostatic potential based 424 method using charge restraints for deriving atomic charges: the RESP model. J Phys Chem. 425 1993;97(40):10269-80. 426 28. Bussi G, Zykova-Timan T, Parrinello M. Isothermal-isobaric molecular dynamics using 427 stochastic velocity rescaling. J Chem Phys. 2009;130(7):074101. 428 29. Darden T, York D, Pedersen L. Particle mesh Ewald: An N⋅ log (N) method for Ewald sums in 429 large systems. J Chem Phys. 1993;98(12):10089-92. 430 30. Michaud-Agrawal N, Denning EJ, Woolf TB, Beckstein O. MDAnalysis: a toolkit for the 431 analysis of molecular dynamics simulations. J Comput Chem. 2011;32(10):2319-27. 432 31. Gowers RJ, Linke M, Barnoud J, Reddy TJE, Melo MN, Seyler SL, et al., editors. MDAnalysis: A 433 Python package for the rapid analysis of molecular dynamics simulations. 15th Python in Science 434 Conference; 2016; Austin, TX. 435 32. DeLano WL. The PyMOL Molecular Graphics System. DeLano Scientific, San Carlos, CA, USA. 436 http://www.pymol.org.2002. 437 33. Collaborative Drug Discovery database public access 2020 [Available from: 438 https://www.collaborativedrug.com/public-access/. 439 34. Grum-Tokars V, Ratia K, Begaye A, Baker SC, Mesecar AD. Evaluating the 3C-like protease 440 activity of SARS-Coronavirus: recommendations for standardized assays for drug discovery. Virus 441 Res. 2008;133(1):63-73. 442 35. Davies TG, Wixted WE, Coyle JE, Griffiths-Jones C, Hearn K, McMenamin R, et al. Monoacidic 443 Inhibitors of the Kelch-like ECH-Associated Protein 1: Nuclear Factor Erythroid 2-Related Factor 2 444 (KEAP1:NRF2) Protein-Protein Interaction with High Cell Potency Identified by Fragment-Based 445 Discovery. J Med Chem. 2016;59(8):3991-4006. 446 36. Erlanson DA, Fesik SW, Hubbard RE, Jahnke W, Jhoti H. Twenty years on: the impact of 447 fragments on drug discovery. Nat Rev Drug Discov. 2016;15(9):605-19. 448 449 450 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 16 Acknowledgements 451 We are grateful to Nick Soffe for maintenance of the Oxford Biochemistry solution NMR facility, 452 to Claire Strain-Damerell, Petra Lukacik and Martin A. Walsh for advice on Mpro production, to 453 Anthony Aimon and Frank von Delft for providing the DSI-poised fragment library, to Adrián García, 454 Nil Casajuana and Clàudia Llinàs del Torrent for advice with MD analysis tools, and to Leonardo 455 Pardo for providing access to high-performance computing facilities. This work was supported by 456 philanthropic donations to the University of Oxford COVID-19 Research Response Fund and the 457 Oxford Glycobiology Institute Endowment. The Oxford Biochemistry NMR facility was supported by 458 the Wellcome Trust (094872/Z/10/Z), the Engineering and Physical Sciences Research Council 459 (EP/R029849/1), the Wellcome Institutional Strategic Support Fund, the EPA Cephalosporin Fund 460 and the John Fell OUP Research Fund. This work was also supported by the “Reinforcement of 461 Postdoctoral Researchers - 2nd Cycle” (MIS-5033021), implemented by the Greek State Scholarships 462 Foundation (ΙΚΥ). 463 464 465 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 17 Figure 1: 1D and STD-NMR spectra of SARS-CoV-2 M pro . A) Methyl regions from 1H NMR spectra of 466 recombinant SARS-CoV-2 Mpro. The spectrum on the left was recorded from a 10 μM protein 467 concentration sample in a 5 mm NMR tube at 25 oC using an excitation sculpting water-suppression 468 method (24). 512 acquisitions with recycle delay of 1.25 sec were averaged, for a total experiment 469 time of just over 10 min. The spectrum on the right was recorded from a 10 μM Mpro sample in a 3 470 mm NMR tube at 10 oC, using the same pulse sequence and acquisition parameters. For both 471 spectra, data were processed with a quadratic sine function prior to Fourier transformation. Protein 472 resonances are weaker in the 10 oC spectrum due to lower temperature and the reduced amount of 473 sample used for acquisition in the smaller NMR tube. The position where on-resonance irradiation 474 was applied for STD spectra is indicated. B) Vertically offset 1H STD-NMR spectra from ligand x0434 475 binding to Mpro. The reference spectrum is in black with the x0434, H2O and DMSO 1H resonances 476 indicated. The STD spectrum of x0434 in the presence of Mpro is shown in red while that in the 477 absence of Mpro is in green. STD spectra are scaled up 64x compared to the reference spectrum. 478 Bottom panels correspond to magnified views of the indicated spectral regions, with x0434 479 resonances assigned to chemical groups of that ligand as shown. 480 481 Figure 2: Assessment of fragment binding to M pro . A) STDratio values for chemical fragments identified 482 by crystallographic screening as binding to Mpro (18). Ligands binding to the Mpro active site are 483 coloured orange, at the Mpro dimer interface in red, and elsewhere on the protein surface in blue. B) 484 Overlay of STD-NMR spectra from fragments x0305, x0387 and x434, which bind the Mpro active site, 485 showing the ligand aromatic region in the presence of Mpro. Spectra are colour coded per ligand as 486 indicated. As seen, the three fragments yield significantly different STD signal intensities captured in 487 the STDratio values shown in (A). C) Overlay of reference spectra from fragments x305, x376 and x540, 488 showing the ligand aromatic region. Peak intensities vary substantially, suggesting significant 489 differences in ligand concentration. 490 491 Figure 3. STD-NMR of COVID Moonshot ligands binding to M pro . A) Overlay of reference spectra from 492 the indicated COVID Moonshot ligands, showing the ligand aromatic region in each case. in the 493 presence of Mpro. Spectra are colour coded per ligand as indicated. As seen, peak intensities vary 494 substantially, suggesting significant differences in ligand concentration. Peaks of ligand EDJ-MED-495 c8e7a002-1 (green) are indicated by arrows; ligand EDJ-MED-e4b030d8-12 (red) produced no peaks 496 in the NMR spectrum. B) Plot of STDratio values from COVID Moonshot ligands assessed by STD-NMR 497 against their IC50 value estimated by RapidFire mass spectroscopy enzymatic assays (17). Ligands in 498 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 18 blue show weak correlation between the two methods (red line, corresponding to an exponential 499 function along the IC50 dimension). Ligands in grey represent outliers of the STD-NMR or enzymatic 500 method as discussed in the text. 501 502 503 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 10 8 6 4 2 0 x0434 H2O DMSO Reference STD (+Mpro) STD (-Mpro) B 2 0 STD irradiation A 2 0 25 oC 10 oC δ 1H (ppm) N NH NH O x0434 1 δ 1H (ppm) δ 1H (ppm) 2 3 4 1 2 3 4 5 5 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ 0 10 20 30 40 50 60 x0434 x0195 x0540 x0426 x0305 x0072 x0161 x0107 x1249 x0395 x0354 x0387 x0397 x1187 x0390 x0194 X 1086 X 1237 X 0350 x1226 X 1235 X 0669 x0398 x0478 X 1119 X 0177 X 0376 X 1132 X 0499 X 1101 X 1163 x0464 X 0336 X 0165 x0425 S TD ra tio (x 1 0- 3 ) Ligand fragments x0305 x0387 x0434 [ppm] 8.5 8.0 7.5 7.0 6.5 δ 1H (ppm) B A [ppm] 8.0 7.5 7.0 6.5 6.0 5.5 x0305 x0376 x0540 C .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/ [ppm] 8.0 7.5 7.0 6.5 6.0 RAL-THA-6b94ceba-1 LOR-NOR-c954e7ad-2 EDJ-MED-c8e7a002-1 EDJ-MED-e4b030d8-12 A δ 1H (ppm) 0.01 0.1 1 10 100 0 100 200 300 400 500 R ap id Fi re IC 50 ( µ M ) STDratio (x 10-3)B R2=30% .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 6, 2021. ; https://doi.org/10.1101/2020.06.17.156679doi: bioRxiv preprint https://doi.org/10.1101/2020.06.17.156679 http://creativecommons.org/licenses/by/4.0/