journal menu
research papers
 | STRUCTURAL BIOLOGY |
accessHigh-confidence placement of low-occupancy fragments into electron density using the anomalous signal of sulfur and halogen atoms
aSchool of Pharmacy, University College London, 29-39 Brunswick Square, London WC1N 1AX, United Kingdom, bPharmacognosy and Pharmaceutical Chemistry Department, College of Pharmacy, Taibah University, Al-Madinah Al-Mounawarah 30078, Saudi Arabia, cEuropean Molecular Biology Laboratory, 38000 Grenoble, France, and dDiamond Light Source, Harwell Science and Innovation Campus, Chilton, Didcot OX11 0DE, United Kingdom
*Correspondence e-mail: shumeng.ma.20@ucl.ac.uk, vitaliy.mykhaylyk@diamond.ac.uk, f.kozielski@ucl.ac.uk
Fragment-based drug design using X-ray crystallography is a powerful technique to enable the development of new lead compounds, or probe molecules, against biological targets. This study addresses the need to determine fragment binding orientations for low-occupancy fragments with incomplete electron density, an essential step before further development of the molecule. Halogen atoms play multiple roles in drug discovery due to their unique combination of steric effects and hydrophobic properties. Fragments incorporating halogen atoms serve as promising starting points in hit-to-lead development as they often establish halogen bonds with target proteins, potentially enhancing binding affinity and selectivity, as well as counteracting drug resistance. Here, the aim was to unambiguously identify the binding orientations of fragment hits for SARS-CoV-2 nonstructural protein 1 (nsp1) which contain a combination of sulfur and/or chlorine, bromine and iodine substituents. The binding orientations of carefully selected nsp1 analogue hits were focused on by employing their combined with Pan-Dataset Density Analysis (PanDDA). Anomalous difference Fourier maps derived from the diffraction data collected at both standard and long-wavelength X-rays were compared. The discrepancies observed in the maps of iodine-containing fragments collected at different energies were attributed to site-specific radiation-damage stemming from the strong X-ray absorption of I atoms, which is likely to cause cleavage of the C—I bond. A reliable and effective data-collection strategy to unambiguously determine the binding orientations of low-occupancy fragments containing sulfur and/or halogen atoms while mitigating radiation damage is presented.
Keywords: SARS-CoV-2 nsp1; fragment-based drug discovery; fragment orientations; anomalous scattering; PanDDA; halogens; COVID-19; low occupancy; radiation damage.
1. Introduction
In fragment-based drug discovery (FBDD), the use of small molecular fragments, typically with a molecular weight of less than 300 Da, offers distinct advantages. Due to their small size and low chemical complexity, fragments tend to yield a higher hit rate, efficiently cover a wider range of target binding sites and provide flexibility for subsequent structural modifications (Zimmermann et al., 2014![[Zimmermann, M. O., Lange, A., Wilcken, R., Cieslik, M. B., Exner, T. E., Joerger, A. C., Koch, P. & Boeckler, F. M. (2014). Future Med. Chem. 6, 617-639.]](../../../../../../logos/arrows/d_arr.gif "Zimmermann, M. O., Lange, A., Wilcken, R., Cieslik, M. B., Exner, T. E., Joerger, A. C., Koch, P. & Boeckler, F. M. (2014). Future Med. Chem. 6, 617-639."){kind=small}
; Baker, 2013). These attributes make fragment hits excellent starting points for drug development. High-throughput screening (HTS) methods are employed in FBDD to efficiently identify fragment hits with the potential to be developed into lead compounds. Among these HTS methods, X-ray crystallography is considered to be an indispensable technique due to its exceptional ability to elucidate the three-dimensional structures of fragment–protein complexes in addition to hit identification, which in turn empowers the rational design and optimization of lead compounds (Hartshorn et al., 2005). However, due to the features of low binding affinity, the co-existence of alternative binding orientations and low occupancy in binding sites, fragments tend to display incomplete electron density in the electron-density maps. In addition, due to their low molecular weight, highlighted within the `rule of three' (RO3; Congreve et al., 2003), and the incorporation of at least one ring (Giordanetto et al., 2019), a large proportion of fragments are planar or quasi-symmetric. These factors render fragment fitting into incomplete density and determination of their binding orientation(s) challenging. However, this problem can be overcome by using fragments incorporating halogen atoms. The distinctive combination of steric effects and hydrophobic properties allows halogen atoms to intricately modulate crucial aspects of drug binding, such as potency, metabolic stability, and permeability (Hernandes et al., 2010; Wang et al., 2022). Over the past decade, halogenated ligands have received increased attention because of their exceptional attributes, including enhanced binding affinity and selectivity for targets, and the potential to counteract drug resistance imparted by the formation of a noncovalent interaction with target molecules, called the `halogen bond' (Baumli et al., 2010; Hardegger et al., 2011). Various halogenated fragment libraries, such as the Halo Library, FragLites and HEFLib (Chopra et al., 2023; Heidrich et al., 2019; Wood et al., 2019), have been designed to probe the halogen-bond interaction in macromolecules of interest to speed up the drug-discovery process.
In addition to these attributes that are beneficial in interactions with proteins, halogens also have a significant anomalous signal at wavelengths commonly used at synchrotron beamlines. It is common to collect anomalous diffraction data for the identification and orientation of halogenated small molecules in the binding sites of macromolecules (Coleman et al., 2020; Pflug et al., 2012). However, for halogenated fragments this approach has only been applied in a limited number of studies. For example, in hot-spot identification studies the anomalous signal from bound halogenated fragments has been used to locate the binding pockets of HIV reverse transcriptase (Bauman et al., 2016; Chopra et al., 2023), HIV protease (Tiefenbrunn et al., 2014) and Thermus thermophilus EF-Tu (Grøftehauge et al., 2013). Other examples include cyclin-dependent kinase 2 (Wood et al., 2019) and glycerol-3-phosphate dehydrogenase (Choe et al., 2002). In these studies, the anomalous signal from halogen atoms was used to confirm the binding and the binding orientations of fragments with reasonable electron density. This raises an important question about the feasibility of this approach when the fragment density is incomplete, as is commonly observed for low-occupancy fragments. Additionally, most of these anomalous data were collected at wavelengths between 0.9 and 1.8 Å (Bauman et al., 2016; Blaney et al., 2006; Choe et al., 2002; Tiefenbrunn et al., 2014; Wood et al., 2019; Grøftehauge et al., 2013), where the protein crystals may experience site-specific radiation damage due to enhanced X-ray absorption above the L edge of iodine (5.2 keV) or the K edge of bromine (13.5 keV). If no account is taken of this phenomenon, it is possible that site-specific radiation damage could lead to a distortion of the peaks associated with iodine or bromine in the anomalous difference maps (Ravelli et al., 2003). This distortion has the potential to complicate the process of fragment fitting. Therefore, it is necessary to consider the influence of radiation damage on the anomalous difference Fourier maps of iodo or bromo fragments (Zwart et al., 2004) and to use a data-collection strategy that minimizes the effects of exposure to radiation.
Pan-Dataset Density Analysis (PanDDA) is a powerful tool for the identification of low-occupancy ligands that leverages ensemble and multi-crystal averaging to enhance the accuracy of electron-density maps, revealing subtle differences that pinpoint the binding sites of fragment hits (Pearce, Bradley et al., 2017; Pearce, Krojer et al., 2017). It has been reported that the effectiveness of fragment binding detection by the inspection of anomalous difference Fourier density maps is superior (Wood et al., 2019) or equal (Davison et al., 2022) to that of PanDDA. Therefore, it is interesting to compare their effectiveness in the determination of binding orientations of low-occupancy fragments.
In this study, we describe our approach to the high-confidence placement of fragments binding to SARS-CoV-2 nsp1 that contain S and/or halogen atoms (Cl, Br and I) into electron density using anomalous diffraction. We chose SARS-CoV-2 nsp1 as a model system because its crystals routinely diffract to high resolution, fragments can easily be soaked in without compromising crystal quality and several hundred data sets are at our disposal. For our investigation, we selected fragments that had already displayed low occupancy and incomplete electron density. The effectiveness of anomalous difference Fourier maps and PanDDA in the placement of these low-occupancy fragments into incomplete electron density was investigated. Furthermore, a study was conducted to determine how site-specific radiation damage develops with X-ray absorption during the collection of diffraction data, exemplified by an iodine-containing fragment. Finally, we demonstrate that the integration of anomalous difference Fourier maps and PanDDA offers a reliable and effective strategy for fitting fragments into challenging electron density with a high degree of confidence.
2. Experimental
2.1. Expression, purification and crystallization of SARS-CoV-2 nsp110–126
The N-terminal domain of SARS-CoV-2 nsp110–126, containing residues 10 to 126 and named nsp1 throughout this manuscript, was expressed, purified and crystallized as described previously (Ma et al., 2022). The crystallization condition used was Index screen (catalogue No. HR2-944-71; Hampton Research, Aliso Viejo, California, USA) condition 71 consisting of 0.1 M bis-Tris pH 6.5, 0.2 M NaCl, 25%(w/v) polyethylene glycol 3350.
Fragment hits 6A6, 1E7, 7G3, 9D4 and 11A7 were obtained from the Maybridge Ro3 library (Thermo Fisher Scientific, United Kingdom), while fragment hits 7H2_AL1, 7H2_AL2, 11A7_AL5 and 11A7_AL6 were purchased from Molport (Beacon, New York, USA), with a purity nominally exceeding 95%. The compounds were dissolved in DMSO-d6 (Sigma, St Louis, Missouri, USA; CAS No. 2206-27-1) at a concentration of 200 mM. Each stock (2 µl) was mixed with 8 µl crystallization condition, resulting in a fragment concentration of 40 mM in the crystallization condition, which also contained 20% DMSO-d6. Fragments or DMSO-d6 solutions (1.5 µl) were added to the approximately 1 µl volume of the crystallization drop, resulting in a final fragment concentration of 24 mM containing approximately 12% DMSO-d6. These drops were incubated at room temperature for 4–5 h. No additional cryoprotection was necessary as 25%(w/v) polyethylene glycol 3350 was present in the crystallization condition. The crystals were harvested using loops, cryocooled in liquid nitrogen and stored in pucks for sample shipment.
2.2. Data collection, and refinement
High-resolution data for nine nsp1–fragment complexes were collected on MASSIF-1 (Bowler et al., 2015; Svensson et al., 2015) at the European Synchrotron Radiation Facility (ESRF) at an incident beam energy Ex of 12.8 keV with 360° of rotation and fine-slicing. The data were analysed in both the single-crystal and multi-crystal systems using PanDDA.
The long-wavelength diffraction experiments were carried out on beamline I23 at Diamond Light Source (DLS). The measurements were performed in a vacuum environment using the semi-cylindrical PILATUS 12M detector and multi-axis goniometer (Wagner et al., 2016). During the measurements, the temperature of protein crystals mounted on copper sample holders was estimated to be ∼80 K. Each sample was exposed to X-rays at an incident beam energy of 4.5 keV. At this energy, the X-ray attenuation lengths of protein soaked with fragments containing sulfur, chlorine, bromine and iodine are practically the same (114, 114, 111 and 111 µm, respectively) and are very similar to that of protein alone (117 µm; values calculated at https://henke.lbl.gov/optical_constants/atten2.html). Therefore, when exposed to the same at 4.5 keV, samples with the same geometry (100 ×100 × 80 µm) absorb a very similar average dose per whole crystal as they are fully exposed, being smaller than the 200 × 200 µm X-ray beam. The dose was equal to 5.4 MGy per data set in this experiment, as calculated by RADDOSE-3D (Bury et al., 2018). Consequently, the radiation damage induced in all fragments is expected to have the same dynamics and magnitude. This was essential in the interpretation of the results and their comparison, as exemplified in our studies of iodine-containing samples (7H2_AL2 and 11A7_AL5) which were also exposed at 5.3 and 9.0 keV (corresponding to wavelengths of 2.8, 2.3 and 1.4 Å, respectively).
For each sample, a 360° rotation data set with fine-slicing (0.10°) was collected to obtain high multiplicity, with the resolution limited only by the detector dimensions (1.8, 1.5 and 1.0 Å at wavelengths of 2.8, 2.3 and 1.4 Å, respectively). Data-collection statistics for data sets obtained on beamlines I23 and MASSIF-1 at wavelengths of 2.8 Å (4.5 keV) and 1.0 Å (12.8 keV), respectively, are summarized in Table 1.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
The data-processing pipelines fast_dp (version 1.6.2), xia2 (version 3.12.0), xia2_3dii and xia2_dials (Winter et al., 2022) were automatically employed. PDB entry 8a55 and the protein sequence were provided in ISPyB (Delagenière et al., 2011) to trigger the Dimple processing pipeline (version 2.6.2; Wojdyr et al., 2013) to generate anomalous difference Fourier maps, and MrBUMP (version 2.2.6; Keegan & Winn, 2008) was employed to find a molecular-replacement solution. Within the Dimple run, ten cycles of rigid-body were performed, followed by four cycles of jelly-body and eight cycles of before starting to identify anomalous difference peaks (Wojdyr et al., 2013).
To unambiguously identify the binding orientations of the fragment hits, the nsp1–fragment coordinates and anomalous difference Fourier maps were overlaid with 2mFo − DFc and mFo − DFc maps calculated from the high-resolution data (12.8 keV) for inspection in Coot (version 8.0; Emsley et al., 2010). For 1E7, 6A6, 7G3, 9D4, 11A7, 11A7_AL5 and 11A7_AL6, nsp1 binding site A is located in proximity to Lys125, whereas for 7H2_AL1 and 7H2_AL2 the shallower binding site B is located adjacent to Pro109, as recently described (Ma et al., 2022). Fragment hits were manually fitted into the fragment density, with S and/or halogen atoms positioned at the peak centres of the associated anomalous difference Fourier maps. Occupancy was then conducted, which involved initially assigning a reasonable occupancy value to the fragments in Coot and then initiating occupancy in Phenix (version 1.20.1-4487; Afonine et al., 2012). The mFo − DFc maps around the fragments were then visually inspected, guiding the adjustment of the initial occupancy values. This was followed by iterative rounds of occupancy in Phenix and further inspection in Coot, which continued until the electron density in the mFo − DFc maps around the fragments reached a minimum. are shown in Table 2. Fig. 1 was prepared using ChemDraw (version 20.1; Cousins, 2005), whereas Supplementary Fig. S2, Fig. 3(a) and Figs. 4–6 were generated with PyMOL (version 2.4.1; Schrödinger).
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
2.3. PanDDA
Given the incomplete mFo − DFc maps that were observed for more than half of the fragment hits collected at 12.8 keV, PanDDA (version 0.2.14; Pearce, Bradley et al., 2017) within CCP4 (version 7.1; Agirre et al., 2023) was employed to determine whether more complete fragment density could be observed in event maps. To prepare for PanDDA, the coordinate and map files for each protein–fragment complex, together with the chemical structure files of the soaked fragments in PDB and formats, were grouped into a single folder. The coordinate and map files for each complex were generated following MR by Dimple (Wojdyr et al., 2013) in CCP4 (version 7.0.072; Agirre et al., 2023) using PDB entry 8a55 as a search model (Ma et al., 2022), while the chemical structure files for each fragment were generated by eLBOW in Phenix (version 1.19.1; Liebschner et al., 2019). In addition, 40 high-resolution data sets of native nsp1 from the protein soaked in 12% DMSO were used in PanDDA to construct a `ground-state' model of nsp1. To identify hits, pandda.analyse was run following the instructions at https://pandda.bitbucket.io/pandda/tutorials.html. Each interesting event was inspected with pandda.inspect through the Coot interface to confirm clear electron density for bound fragments. PanDDA maps of fragment density were captured and used for the preparation of Figs. 4, 5 and 6.
3. Results
Nine new nsp1-binding fragment analogues containing sulfur and/or halogen substituents were selected for this study. The chemical structures of the analogues and their parental fragment hits are shown in Fig. 1. The SMILES string computer-readable identifiers are provided in Supplementary Table S1. Except for 7H2, 7H2_AL1 and 7H2_AL2, which bind to binding site B, all other fragments were located in binding site A of nsp1 (Ma et al., 2022).
![[Figure 1]](gm5104fig1thm.gif) | Figure 1 Chemical structures of fragment analogues containing S atoms and/or a chloro, bromo and iodo substituent that bind to nsp1. 2E10 and 7H2 (boxed) are two parental fragments that were reported in our previous publications (Borsatto et al., 2022 ; Ma et al., 2022). |
3.1. Comparison of the quality of data collected on MASSIF-1 and a dedicated long-wavelength beamline
To validate the binding of the fragment analogues, we prepared analogue-soaked nsp1 crystals and collected diffraction data at 12.8 keV on MASSIF-1 at ESRF. However, when fitting the fragment analogues into the mFo − DFc maps, half of the maps showed incomplete fragment density, which is not uncommon for fragments with low binding occupancy. Therefore, we took advantage of the from the heavy atoms contained in these fragment analogues by calculating anomalous difference Fourier maps from the 12.8 keV data sets using Dimple (version 2.6.2; Wojdyr et al., 2013). The quality of the maps obtained was validated by inspecting the anomalous signal from the S atoms of Met9, Cys51 and Met85 of nsp1 and the sulfur or halogen signals from the fragment analogues in Coot (Emsley et al., 2010; Fig. 2).
![[Figure 2]](gm5104fig2thm.gif) | Figure 2 A 3D column chart comparing anomalous peak heights, σ, for S atoms in Met9, Cys51 and Met85 of nsp1 and for S and/or halogen atoms in fragment analogues calculated from the data collected at 12.8 keV compared with those collected at 4.5 keV. If two anomalous signals from fragment analogues appeared (one from S and the other from halogen atoms), only the halogen anomalous peak value was plotted. This chart was prepared in Origin 2018 (Moberly et al., 2018 ). |
Overall, the anomalous sulfur signal from the cysteine, methionine and sulfur-containing fragment analogues (1E7, 7G3, 6A6, 11A7, 11A7_AL5 and 11A7_AL6) cannot be observed consistently, probably due to the low anomalous contribution to the f′′ (0.2 e) of sulfur at 12.8 keV, which is far from the sulfur (Supplementary Fig. S1). Therefore, the anomalous difference Fourier maps calculated from the data sets collected at a standard wavelength (1.0 Å, 12.8 keV) are not sufficient to facilitate the fitting of sulfur-containing fragments. Likewise, the anomalous signal of chlorine could not be observed in the fragment density of 9D4 and only appeared weakly in the density for 7H2 as the f′′ of chlorine at 12.8 keV is also low (0.3 e; Supplementary Fig. S1). However, this scenario changed for the bromine or iodine-containing analogues (11A7_AL5, 11A7_AL6, 7H2_AL1 and 7H2_AL2), for which peaks can be observed in the anomalous difference Fourier maps at 12.8 keV because the f′′ of iodine and bromine are as high as 3.0 and 0.5 e, respectively, at this beam energy. Site-specific radiation damage was observed in the anomalous difference Fourier map calculated from the 7H2_AL2 data set collected at 12.8 keV, where two adjacent anomalous peaks of 9.2σ and 6.5σ appeared (only the higher peak was plotted in Fig. 2). This suggests that although anomalous signal from iodine can be observed in the data collected at a standard wavelength, it would still be beneficial to collect data using low doses or at an incident beam energy far from the iodine absorption peak at 5.2 keV to avoid site-specific radiation damage, which will be discussed in more detail below. Additionally, for the quasi-symmetric and planar fragment analogue 7H2_AL1, in which the two substituents are in para positions, a single anomalous signal from bromine is adequate to fit the fragment into the density. In contrast, for the asymmetric analogues 11A7_AL5 and 11A7_AL6, a single anomalous signal from iodine or bromine is not sufficient. Therefore, for those analogues that contain both S and Br/I atoms, it is suggested that anomalous data should be collected at lower energy, close to and above the sulfur to obtain the anomalous signals from both heavy atoms in order to unambiguously fit them into the electron density.
To obtain higher quality anomalous difference Fourier maps, nsp1 crystals soaked with the distinct fragment analogues prepared under the same conditions were measured on beamline I23 at DLS at an incident X-ray energy of 4.5 keV. The anomalous signal from sulfur in methionine and cysteine side chains present in nsp1, and from heavy atoms in the fragments, were again visually inspected and the peak heights of these anomalous signals were compared with those extracted from the 12.8 keV data (Fig. 2). The anomalous signals originating from the S atoms in these residues are visible in all anomalous difference Fourier maps except for that of nsp1–7G3, for which that of the sulfur in Met9 was not observed. This is possibly because of the low I/σ(I) of this specific data set and the flexibility of Met9 as the first residue at the N-terminus of nsp1. It is clear that the anomalous signals of S atoms from the protein and fragment analogues are significantly stronger and appear consistently in the anomalous difference Fourier maps calculated from the data collected at I23, where the measurements were performed at Ex = 4.5 keV in a vacuum environment to maximize the signal-to-noise ratio (El Omari et al., 2023). The anomalous signals of halogen atoms in the fragment analogues are also strong. No anomalous peak splitting, which might be caused by site-specific radiation damage, was observed.
3.2. Strategy for data collection and the effects of radiation damage
The wavelength chosen for anomalous data collection on I23 was based on the following considerations. For sulfur- and chloride-containing fragments, 4.5 keV is above their absorption edges, allowing strong anomalous signals to be obtained without compromising resolution (the maximum achievable resolution at 4.5 keV is 1.8 Å due to the I23 beamline detector geometry and the fixed sample-to-detector distance). While the K of bromine is 13.5 keV, which is beyond the tuneable range, the anomalous contribution to the from the L edge of Br at 4.5 keV (f′′ = 3.4 e) is sufficiently high to allow the signal to be confidently observed in the anomalous difference Fourier maps. For iodine-containing fragments, although the L (5.2 keV) is within the tuneable range, data collection close to and above its should be avoided due to the corresponding strong X-ray absorption As site-specific radiation damage was observed at 12.8 keV, where the f′′ of iodine is 3.0 e, we collected data at three incident energies, just below the absorption peak (4.5 keV), above the peak (5.3 keV) and significantly above the peak (9 keV), for both analogues to establish the best approach for data collection for the iodine-containing analogues 7H2_AL2 and 11A7_AL5. By comparing the fragment-binding sites in the three data sets, we observed that a single anomalous peak appeared around the I atom in both the 4.5 and 9 keV maps, while two adjacent anomalous peaks appeared in the 5.3 keV data set for both analogues (Supplementary Fig. S2). This supports our conjecture that site-specific radiation damage is likely to occur due to the strong absorption of iodine at the L edge (5.18 keV). Radiation-induced structural changes in proteins are not uncommon, but are a major concern when the measurements are carried out at energies that correspond to the absorption edges of ions, although these changes can be utilized for experimental phasing (Fütterer et al., 2008; Schiltz et al., 2004). To test this, we collected 22 data sets at 9 keV, where the absorption of iodine is reduced by an order of magnitude in comparison with the peak value (calculated at https://henke.lbl.gov/optical_constants/atten2.html), from a single crystal soaked with 11A7_AL5. This was conducted to capture the moment of initiation of the site-specific radiation-induced changes. The average dose absorbed by the whole crystal during the collection of one data set at 9 keV was 0.34 MGy, as calculated by RADDOSE-3D (Bury et al., 2018). By inspecting the 22 data sets in the order in which they were collected, we observed that the radiation-induced structural changes occurred when the anomalous signal of iodine gradually shifted to the second anomalous peak (starting from the ninth data collection) and that the anomalous signal is redistributed between the two sites until the peak-height ratio reduces to 1. Representative transitions of the anomalous difference Fourier maps are displayed in Fig. 3(a), while the peak-height ratio between the two anomalous signals is plotted in Fig. 3(b) from data set 9. We believe that the absorbed dose could trigger cleavage of the C—I bond, leading to a shift of the I atom away from the C atom to the nearest available space as the absorbed dose increases. The displacement of metal ions induced by radiation has recently been shown in X-ray crystallographic studies of metalloproteins (Lennartz et al., 2022), while instances of radiation-induced bond cleavage and the subsequent shift of a Br atom have also been previously documented (Ravelli et al., 2003). In the case of shifted anomalous signals in the maps from the diffraction data of 11A7_AL5 in complex with nsp1, the peaks are derived from an I atom in two distinct locations. As data collection proceeds, the fraction of cleaved C—I bonds increases, which is manifested as a gradually stronger second anomalous peak in the superposed electron-density maps. A rigorous validation of this hypothesis would require dedicated experiments and theoretical calculations which are outside the scope of this study. Nonetheless, this investigation allowed us to determine a strategy to prevent C—I bond cleavage when using iodine-containing fragments. For anomalous data collection we chose 4.5 keV where, despite being below the L of iodine, the f′′ value (3.4 e) is large enough for iodine to be observed in the anomalous difference Fourier maps. At the same time, the absorption at this energy is relatively low and thus is unlikely to induce radiation damage at typical doses for data collection.
![[Figure 3]](gm5104fig3thm.gif) | Figure 3 (a) Representative transitions of the anomalous difference Fourier maps of the iodine-containing fragment analogue 11A7_AL5 showing site-specific radiation damage that occurs during data collection at Ex = 9 keV. Panels (i)–(vi) show the gradual and continuous development of the second anomalous peak from iodine present in the analogue. For simplicity, only the maps from data sets 8, 9, 13, 17, 20 and 22 are shown. I, N, S and C atoms are coloured purple, blue, yellow and cyan, respectively. The anomalous difference Fourier maps are shaded as an orange mesh (4σ). (b) Line graph showing the sigma ratio of iodine anomalous peak heights between the first (initial) and the second (gradually appearing) anomalous peaks. The second peak did not appear in the maps for the first eight data sets, and therefore the graph starts at data set 9. |
3.3. Low-occupancy and planar fragment fitting using anomalous signals and PanDDA maps
Fragment fitting was guided by overlaying anomalous difference Fourier maps generated from data collected at 4.5 keV onto mFo − DFc maps calculated from the MASSIF-1 data (Figs. 4, 5 and 6). spectra of halogen-containing fragments were also collected to identify chemical ions of interest that might be in the crystal or potentially bind to the protein. The measured at 9.0 keV, as exemplified by the nsp1–11A7_AL5 complex, shows clear peaks assigned to the Kα lines of sulfur at 2.3 keV and chlorine at 2.6 keV, as well as a peak due to the Lβ line of iodine at 4.3 keV (Supplementary Fig. S3).
![[Figure 4]](gm5104fig4thm.gif) | Figure 4 Comparison of various maps in the binding site of analogues 6A6, 11A7, 11A7_AL5 and 11A7_AL6 (row 1 to row 4, respectively). (a, d, g, j) PanDDA event maps [blue, 1.0σ, background density correction factor (BDC) = 0.37, 0.39, 0.18 and 0.25, respectively] and Z-maps (green/red, ±4.0σ). (b, e, h, k) Sulfur, iodine and bromine anomalous difference Fourier maps in the fragment region calculated from data collected at 4.5 keV (orange, 4σ) overlaid with mFo − DFc maps (green, 3.0σ) calculated from the 12.8 keV data. (c, f, i, l) Refined 2mFo − DFc maps (blue, 1.0σ) calculated from the 12.8 keV data with the S, I or Br atoms placed in the centres of their anomalous peaks. The 2mFo − DFc map is almost complete for 6A6 (c), 11A7 (f) and 11A7_AL6 (l) but is partial for 11A7_AL5 (i). In Figs. 4 , 5 and 6 and Supplementary Fig. S2, the C, N, O, S, F, Cl, Br and I atoms in the fragments are coloured cyan, blue, red, yellow, light blue, green, dark red and purple, respectively. |
![[Figure 5]](gm5104fig5thm.gif) | Figure 5 Comparison of various maps in the binding site of analogues 1E7, 7G3 and 9D4 (row 1 to row 3, respectively). (a, d, g) PanDDA event maps (BDC = 0.44, 0.27 and 0.21, respectively) and Z-maps (green/red, ±4.0σ). (b, e, h) Sulfur and chlorine anomalous difference Fourier maps calculated from data collected at 4.5 keV (orange, 4σ) overlaid with mFo − DFc maps (green, 3.0σ) calculated from the 12.8 keV data in the fragment region. (c, f, i) Refined 2mFo − DFc maps (blue, 1.0σ) calculated from the 12.8 keV data with the S or Cl atoms placed in the centres of the anomalous peaks. The electron density completely accounts for 1E7, while only partial density is visible for 7G3 and 9D4, possibly due to their low occupancy. While one binding orientation was identified for 1E7, two binding orientations were evident for both 7G3 and 9D4. |
![[Figure 6]](gm5104fig6thm.gif) | Figure 6 Comparison of various maps in the binding site of analogues 7H2, 7H2_AL1 and 7H2_AL2 (row 1 to row 3, respectively). (a, d, g) PanDDA event maps (BDC = 0.34, 0.33 and 0.48, respectively) and Z-maps (green/red, ±4.0σ). (b, e, h) Chlorine, bromine and iodine anomalous difference Fourier maps calculated from data collected at 4.5 keV (orange, 4σ) overlaid with mFo − DFc maps (green, 3.0σ) calculated from the 12.8 keV data in the fragment region. (c, f, i) Refined 2mFo − DFc maps (blue, 1.0σ) of 7H2, 7H2_AL1 and 7H2_AL2 calculated from the 12.8 keV data with the Cl, Br or I atom placed in the centre of the anomalous peaks. The electron density mostly accounts for 7H2, while for the two analogues the density systematically degrades, possibly due to their low occupancy. |
A multi-crystal method for extracting weak binding states from conventionally uninterpretable electron density, PanDDA, was run on the data collected at 12.8 keV to evaluate its effectiveness in identifying partial occupancy features in the crystallographic data. As expected, PanDDA maps show more complete fragment density compared with the standard maps for all fragment analogues. However, the binding orientations and potential alternative orientations of the analogues are still challenging to interpret based solely on PanDDA maps (Figs. 4a, 4d, 4g, 4j, 5a, 5d, 5g, 6a, 6d and 6g).
Among the seven analogues of fragment hit 2E10 (Fig. 1), 6A6, 11A7, 11A7_AL5 and 11A7_AL6 contain a benzothiazole ring system and an amine substituent at the 2′ position. The only difference between these structures is the substituent at the 6′ position, which is either hydrogen, fluorine, bromine or iodine.
The PanDDA map for 6A6 could indicate the location of its single amine substituent in the fragment density, but provides less information on the direction of its ring system due to its quasi-symmetry (Fig. 4a). The binding orientation is clear when the anomalous signal of sulfur is present (Fig. 4b). Although the PanDDA map for 11A7_AL5 (Fig. 4g) is interpretable, the mFo − DFc map (Fig. 4h) of 11A7_AL5 is at best partial, and both provide little information on binding orientation. Therefore, fitting the fragment analogue into the maps (Fig. 4g or 4h) would be challenging. Similarly, through the location of the anomalous peaks from sulfur and iodine and the assignment of the two peaks by comparing the difference in anomalous peak heights between sulfur and iodine (with iodine having a higher anomalous peak height due to its larger f′′ value; Fig. 4h), the binding orientations can be unambiguously determined (Fig. 4i). The PanDDA maps (Figs. 4d and 4j) are complete for 11A7 and 11A7_AL6 and are better than the mFo − DFc maps (Figs. 4e and 4k). However, both types of map are sufficient for an experienced crystallographer to manually fit the analogues in the correct orientations (Figs. 4f and 4l). Nonetheless, the anomalous difference Fourier map provides further confidence in fitting. Overall, the binding orientations of the four analogues are the same, as expected from their high structural and chemical similarity (Fig. 4).
The other three analogues of 2E10, namely 1E7, 7G3 and 9D4 (Fig. 1), demonstrate more diversity in the five-membered rings fused to the benzene ring. 1E7 and 7G3 share the same ring scaffold, benzothiophene, with a single substituent at distinct positions. For 1E7, an amine is positioned para to the sulfur, while a more flexible acetic acid substituent is located in the meta position to the sulfur in 7G3, which may explain the missing density for this substituent (Figs. 5d, 5e and 5f). The mFo − DFc map obtained from a single crystal is as good as the PanDDA map of 1E7 (Fig. 5a), indicating nearly full occupancy and a clear binding orientation (Fig. 5c). In the anomalous difference Fourier map of 1E7 combined with the mFo − DFc map, only one sulfur anomalous peak was observed (Fig. 5b).
In contrast to 1E7, 7G3 represents a fragment analogue that binds with low occupancy, resulting in difficult-to-interpret electron-density maps (Figs. 5d, 5e and 5f). Whereas the PanDDA map still covers the core ring structure, it provides no indication of its substituent and binding orientations (Fig. 5d). Interestingly, even at a resolution of 1.2 Å the mFo − DFc map of 7G3 is only partially visible (Fig. 5e), and it is difficult to fit it confidently into the density. To complicate matters, two peaks were observed for the S atom in 7G3 (Fig. 5e). In such a challenging case, the anomalous difference Fourier map allowed confident fragment fitting, exemplifying how useful anomalous signals can be when working with low-occupancy fragments that bind in two distinct orientations.
Although maintaining a fused two-ring core, 9D4 has an indazole ring system and two substituents, one on each ring. An amine substituent is present at the 3′ position, while a chloro substituent is at the 4′ position (Fig. 1). For this fragment analogue, the mFo − DFc map is uninterpretable (Fig. 5h). The PanDDA map nearly covers the core ring system of 9D4 and again highlights the strength of this pan-data-set approach, but it provides limited information about the position of its substituent and potential binding orientations (Fig. 5g). Facilitated by the anomalous difference Fourier map, two binding orientations were clearly suggested by chlorine anomalous peaks in the density (Figs. 5h and 5i).
7H2_AL1 and 7H2_AL2 are two analogues of the previously reported fragment hit 7H2 (Ma et al., 2022) in which the chlorine substituent is replaced with bromine and iodine (Fig. 1), respectively. 7H2 displays reasonable PanDDA and mFo − DFc maps; however, the direction of its two substituents was ambiguous. The combination of the anomalous difference Fourier map and the mFo − DFc map allowed unambiguous fitting of the fragment (Ma et al., 2023; Figs. 6b and 6c). Similarly, the nearly complete PanDDA (Fig. 6d) and partial mFo − DFc maps (Fig. 6e) of 7H2_AL1 present two possible binding orientations. By locating the anomalous signal from bromine, an unambiguous binding orientation of 7H2_AL1 can be determined (Figs. 6e and 6f). However, for 7H2_AL2 the density is clearly compromised and distorted by site-specific radiation damage in the mFo − DFc map (Fig. 6h). By locating the anomalous signal of the iodine substituent, the fragment can still be confidently placed into the remaining density, in particular in the PanDDA map. The three analogues share the same binding orientation, with the halogen atoms pointing towards the protein to anchor the fragments in the binding pocket (Figs. 6c, 6f and 6i).
4. Discussion
FBDD has emerged as a powerful strategy for developing novel lead compounds and advancing drug development. However, FBDD also presents challenges that differentiate it from traditional small-molecule drug-discovery approaches. A fundamental aspect is the markedly weak binding affinity of fragments for their targets, which is typically in the low-millimolar range. Additionally, the fragments are small, simple and typically incorporate at least one aromatic ring, therefore having fewer rotatable bonds, which renders them planar and quasi-symmetric. These intrinsic features introduce additional complexities in determining their binding orientations.
This challenge can be addressed by implementing the PanDDA approach, which identifies binding events by comparing data sets from crystals soaked with fragments with native data sets, allowing the identification of fragments with a statistically reliable degree of confidence (Pearce et al., 2015). However, the fragment density in PanDDA maps for low-occupancy binding events may still be insufficient to ascertain the binding orientation(s) of hits due to their inherent features, such as small size and planarity. To overcome this issue, the combination of PanDDA maps with anomalous signals generated from atoms in fragment analogues not only confirms the binding orientation but can also suggest the presence of multiple binding orientations. The occupancy of each orientation for the same fragment analogue can also be estimated from the ratio of anomalous peak heights. Consequently, these two methods are complementary, and when employed in conjunction they can offer unequivocal information about fragment binding conformations.
This study has the potential to inform good practice for the problem of correctly placing low-occupancy fragments into incomplete electron density in FBDD. The method is well suited to determine the binding orientations of fragments. A schematic summary of the general workflow is provided in Fig. 7. X-ray diffraction data are first collected at a standard wavelength from crystals soaked with fragments. Fourier maps are generated in the single-crystal system, where well defined ligand density can unambiguously guide ligand fitting. Challenging-to-fit low-occupancy fragments, showing partial or uninterpretable fragment density, are then selected for long-wavelength experiments with careful design of the data-collection parameters, considering potential radiation-damage effects caused by the absorbed dose. Anomalous difference Fourier maps are then calculated from the long-wavelength data, and event maps are computed by PanDDA in the multi-crystal system. Determining the binding orientation(s) of low-occupancy fragments is achieved by considering the complementary information from both PanDDA event maps and anomalous difference Fourier maps.
![[Figure 7]](gm5104fig7thm.gif) | Figure 7 Schematic workflow for correctly placing low-occupancy fragments into incomplete electron density during FBDD. |
Radiation-damage effects in data collection for fragments containing bromine and iodine have rarely been considered. Anomalous data have also rarely been applied to simultaneously determine the binding modes of fragments containing S and halogen atoms. This study of data-collection strategies for fragments containing S and/or halogenated atoms or substituents suggest that a carefully designed strategy for data collection is necessary depending on the purpose of the study. For the identification of fragment hits containing iodine, a low dose during data collection is recommended to avoid site-specific radiation damage, in particular when the X-ray energy is close to the iodine For the determination of binding orientations of fragments containing both S/Cl and halogen atoms, the incident X-ray energy should be above and close to the sulfur/chlorine to ensure that the anomalous signals of both can be observed for unambiguous manual fragment placement.
5. Conclusion
In conclusion, this study has shown how the challenge of determining the binding orientation(s) for low-occupancy and difficult-to-fit fragments can successfully be addressed for sulfur- and/or halogen-containing fragments. This method is poised to significantly improve the efficacy and success rate of FBDD during rational drug design.
Supporting information
Supplementary Table and Figures. DOI: https://doi.org/10.1107/S2059798324004480/gm5104sup1.pdf
Acknowledgements
This publication comprises part of the doctoral thesis of Shumeng Ma. We acknowledge Diamond Light Source and the European Synchrotron Radiation Facility for time on the I23 and MASSIF-1 beamlines under proposals MX30393 and MX2322, respectively. This research received no external funding.
References
Afonine, P. V., Grosse-Kunstleve, R. W., Echols, N., Headd, J. J., Moriarty, N. W., Mustyakimov, M., Terwilliger, T. C., Urzhumtsev, A., Zwart, P. H. & Adams, P. D. (2012). Acta Cryst. D68, 352–367. Web of Science CrossRef CAS IUCr Journals Google Scholar
Agirre, J., Atanasova, M., Bagdonas, H., Ballard, C. B., Baslé, A., Beilsten-Edmands, J., Borges, R. J., Brown, D. G., Burgos-Mármol, J. J., Berrisford, J. M., Bond, P. S., Caballero, I., Catapano, L., Chojnowski, G., Cook, A. G., Cowtan, K. D., Croll, T. I., Debreczeni, J. É., Devenish, N. E., Dodson, E. J., Drevon, T. R., Emsley, P., Evans, G., Evans, P. R., Fando, M., Foadi, J., Fuentes-Montero, L., Garman, E. F., Gerstel, M., Gildea, R. J., Hatti, K., Hekkelman, M. L., Heuser, P., Hoh, S. W., Hough, M. A., Jenkins, H. T., Jiménez, E., Joosten, R. P., Keegan, R. M., Keep, N., Krissinel, E. B., Kolenko, P., Kovalevskiy, O., Lamzin, V. S., Lawson, D. M., Lebedev, A. A., Leslie, A. G. W., Lohkamp, B., Long, F., Malý, M., McCoy, A. J., McNicholas, S. J., Medina, A., Millán, C., Murray, J. W., Murshudov, G. N., Nicholls, R. A., Noble, M. E. M., Oeffner, R., Pannu, N. S., Parkhurst, J. M., Pearce, N., Pereira, J., Perrakis, A., Powell, H. R., Read, R. J., Rigden, D. J., Rochira, W., Sammito, M., Sánchez Rodríguez, F., Sheldrick, G. M., Shelley, K. L., Simkovic, F., Simpkin, A. J., Skubak, P., Sobolev, E., Steiner, R. A., Stevenson, K., Tews, I., Thomas, J. M. H., Thorn, A., Valls, J. T., Uski, V., Usón, I., Vagin, A., Velankar, S., Vollmar, M., Walden, H., Waterman, D., Wilson, K. S., Winn, M. D., Winter, G., Wojdyr, M. & Yamashita, K. (2023). Acta Cryst. D79, 449–461. Web of Science CrossRef IUCr Journals Google Scholar
Baker, M. (2013). Nat. Rev. Drug Discov. 12, 5–7. CrossRef CAS PubMed Google Scholar
Bauman, J. D., Harrison, J. J. E. K. & Arnold, E. (2016). IUCrJ, 3, 51–60. Web of Science CrossRef CAS PubMed IUCr Journals Google Scholar
Baumli, S., Endicott, J. A. & Johnson, L. N. (2010). Chem. Biol. 17, 931–936. Web of Science CrossRef CAS PubMed Google Scholar
Blaney, J., Nienaber, V. & Burley, S. K. (2006). Fragment-based Approaches in Drug Discovery, edited by W. Jahnke & D. A. Erlanson, pp. 215–248. Weinheim: Wiley-VCH. Google Scholar
Borsatto, A., Akkad, O., Galdadas, I., Ma, S., Damfo, S., Haider, S., Kozielski, F., Estarellas, C. & Gervasio, F. L. (2022). eLife, 11, e81167. CrossRef PubMed Google Scholar
Bowler, M. W., Nurizzo, D., Barrett, R., Beteva, A., Bodin, M., Caserotto, H., Delagenière, S., Dobias, F., Flot, D., Giraud, T., Guichard, N., Guijarro, M., Lentini, M., Leonard, G. A., McSweeney, S., Oskarsson, M., Schmidt, W., Snigirev, A., von Stetten, D., Surr, J., Svensson, O., Theveneau, P. & Mueller-Dieckmann, C. (2015). J. Synchrotron Rad. 22, 1540–1547. Web of Science CrossRef CAS IUCr Journals Google Scholar
Bury, C. S., Brooks-Bartlett, J. C., Walsh, S. P. & Garman, E. F. (2018). Protein Sci. 27, 217–228. Web of Science CrossRef CAS PubMed Google Scholar
Choe, J., Suresh, S., Wisedchaisri, G., Kennedy, K. J., Gelb, M. H. & Hol, W. G. J. (2002). Chem. Biol. 9, 1189–1197. CrossRef PubMed CAS Google Scholar
Chopra, A., Bauman, J. D., Ruiz, F. X. & Arnold, E. (2023). J. Med. Chem. 66, 6013–6024. CrossRef CAS PubMed Google Scholar
Coleman, J. A., Navratna, V., Antermite, D., Yang, D., Bull, J. A. & Gouaux, E. (2020). eLife, 9, e56427. CrossRef PubMed Google Scholar
Congreve, M., Carr, R., Murray, C. & Jhoti, H. (2003). Drug Discov. Today, 8, 876–877. Web of Science CrossRef PubMed Google Scholar
Cousins, K. R. (2005). J. Am. Chem. Soc. 127, 4115–4116. CrossRef CAS Google Scholar
Davison, G., Martin, M. P., Turberville, S., Dormen, S., Heath, R., Heptinstall, A. B., Lawson, M., Miller, D. C., Ng, Y. M., Sanderson, J. N., Hope, I., Wood, D. J., Cano, C., Endicott, J. A., Hardcastle, I. R., Noble, M. E. M. & Waring, M. J. (2022). J. Med. Chem. 65, 15416–15432. CrossRef CAS PubMed Google Scholar
Delagenière, S., Brenchereau, P., Launer, L., Ashton, A. W., Leal, R., Veyrier, S., Gabadinho, J., Gordon, E. J., Jones, S. D., Levik, K. E., McSweeney, S. M., Monaco, S., Nanao, M., Spruce, D., Svensson, O., Walsh, M. A. & Leonard, G. A. (2011). Bioinformatics, 27, 3186–3192. Web of Science PubMed Google Scholar
El Omari, K., Duman, R., Mykhaylyk, V., Orr, C. M., Latimer-Smith, M., Winter, G., Grama, V., Qu, F., Bountra, K., Kwong, H. S., Romano, M., Reis, R. I., Vogeley, L., Vecchia, L., Owen, C. D., Wittmann, S., Renner, M., Senda, M., Matsugaki, N., Kawano, Y., Bowden, T. A., Moraes, I., Grimes, J. M., Mancini, E. J., Walsh, M. A., Guzzo, C. R., Owens, R. J., Jones, E. Y., Brown, D. G., Stuart, D. I., Beis, K. & Wagner, A. (2023). Commun. Chem. 6, 219. Web of Science CrossRef PubMed Google Scholar
Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. (2010). Acta Cryst. D66, 486–501. Web of Science CrossRef CAS IUCr Journals Google Scholar
Fütterer, K., Ravelli, R. B. G., White, S. A., Nicoll, A. J. & Allemann, R. K. (2008). Acta Cryst. D64, 264–272. Web of Science CSD CrossRef IUCr Journals Google Scholar
Giordanetto, F., Jin, C., Willmore, L., Feher, M. & Shaw, D. E. (2019). J. Med. Chem. 62, 3381–3394. Web of Science CrossRef CAS PubMed Google Scholar
Grøftehauge, M. K., Therkelsen, M. Ø., Taaning, R., Skrydstrup, T., Morth, J. P. & Nissen, P. (2013). Acta Cryst. F69, 1060–1065. Web of Science CrossRef IUCr Journals Google Scholar
Hardegger, L. A., Kuhn, B., Spinnler, B., Anselm, L., Ecabert, R., Stihle, M., Gsell, B., Thoma, R., Diez, J., Benz, J., Plancher, J. M., Hartmann, G., Banner, D. W., Haap, W. & Diederich, F. (2011). Angew. Chem. Int. Ed. 50, 314–318. Web of Science CrossRef CAS Google Scholar
Hartshorn, M. J., Murray, C. W., Cleasby, A., Frederickson, M., Tickle, I. J. & Jhoti, H. (2005). J. Med. Chem. 48, 403–413. Web of Science CrossRef PubMed CAS Google Scholar
Heidrich, J., Sperl, L. E. & Boeckler, F. M. (2019). Front. Chem. 7, 9. CrossRef PubMed Google Scholar
Hernandes, Z. M., Cavalcanti, S. M., Moreira, D. R., de Azevedo Junior, F. W. & Leite, A. C. (2010). Curr. Drug Targets, 11, 303–314. CrossRef CAS PubMed Google Scholar
Keegan, R. M. & Winn, M. D. (2008). Acta Cryst. D64, 119–124. Web of Science CrossRef CAS IUCr Journals Google Scholar
Lennartz, F., Jeoung, J.-H., Ruenger, S., Dobbek, H. & Weiss, M. S. (2022). Acta Cryst. D78, 238–247. Web of Science CrossRef IUCr Journals Google Scholar
Liebschner, D., Afonine, P. V., Baker, M. L., Bunkóczi, G., Chen, V. B., Croll, T. I., Hintze, B., Hung, L.-W., Jain, S., McCoy, A. J., Moriarty, N. W., Oeffner, R. D., Poon, B. K., Prisant, M. G., Read, R. J., Richardson, J. S., Richardson, D. C., Sammito, M. D., Sobolev, O. V., Stockwell, D. H., Terwilliger, T. C., Urzhumtsev, A. G., Videau, L. L., Williams, C. J. & Adams, P. D. (2019). Acta Cryst. D75, 861–877. Web of Science CrossRef IUCr Journals Google Scholar
Ma, S., Damfo, S., Lou, J., Pinotsis, N., Bowler, M. W., Haider, S. & Kozielski, F. (2022). Int. J. Mol. Sci. 23, 12448. CrossRef PubMed Google Scholar
Ma, S., Mykhaylyk, V., Bowler, M. W., Pinotsis, N. & Kozielski, F. (2023). Int. J. Mol. Sci. 24, 11197. CrossRef PubMed Google Scholar
Moberly, J. G., Bernards, M. T. & Waynant, K. V. (2018). J. Cheminform, 10, 5. CrossRef PubMed Google Scholar
Pearce, N. M., Bradley, A. R., Krojer, T., Marsden, B. D., Deane, C. M. & von Delft, F. (2017). Struct. Dyn. 4, 032104. Web of Science CrossRef PubMed Google Scholar
Pearce, N. M., Kelm, S., Shi, J., Deane, C. M. & von Delft, F. (2015). Acta Cryst. A71, s258. CrossRef IUCr Journals Google Scholar
Pearce, N. M., Krojer, T., Bradley, A. R., Collins, P., Nowak, R. P., Talon, R., Marsden, B. D., Kelm, S., Shi, J., Deane, C. M. & von Delft, F. (2017). Nat. Commun. 8, 15123. Web of Science CrossRef PubMed Google Scholar
Pflug, A., Johnson, K. A. & Engh, R. A. (2012). Acta Cryst. F68, 873–877. Web of Science CrossRef IUCr Journals Google Scholar
Ravelli, R. B. G., Leiros, H. S., Pan, B., Caffrey, M. & McSweeney, S. (2003). Structure, 11, 217–224. CrossRef PubMed CAS Google Scholar
Schiltz, M., Dumas, P., Ennifar, E., Flensburg, C., Paciorek, W., Vonrhein, C. & Bricogne, G. (2004). Acta Cryst. D60, 1024–1031. Web of Science CrossRef CAS IUCr Journals Google Scholar
Svensson, O., Malbet-Monaco, S., Popov, A., Nurizzo, D. & Bowler, M. W. (2015). Acta Cryst. D71, 1757–1767. Web of Science CrossRef IUCr Journals Google Scholar
Tiefenbrunn, T., Forli, S., Happer, M., Gonzalez, A., Tsai, Y., Soltis, M., Elder, J. H., Olson, A. J. & Stout, C. D. (2014). Chem. Biol. Drug Des. 83, 141–148. Web of Science CrossRef CAS PubMed Google Scholar
Wagner, A., Duman, R., Henderson, K. & Mykhaylyk, V. (2016). Acta Cryst. D72, 430–439. Web of Science CrossRef IUCr Journals Google Scholar
Wang, Z., Cherukupalli, S., Xie, M., Wang, W., Jiang, X., Jia, R., Pannecouque, C., De Clercq, E., Kang, D., Zhan, P. & Liu, X. (2022). J. Med. Chem. 65, 3729–3757. CrossRef CAS PubMed Google Scholar
Winter, G., Beilsten-Edmands, J., Devenish, N., Gerstel, M., Gildea, R. J., McDonagh, D., Pascal, E., Waterman, D. G., Williams, B. H. & Evans, G. (2022). Protein Sci. 31, 232–250. Web of Science CrossRef CAS PubMed Google Scholar
Wojdyr, M., Keegan, R., Winter, G. & Ashton, A. (2013). Acta Cryst. A69, s299. Web of Science CrossRef IUCr Journals Google Scholar
Wood, D. J., Lopez-Fernandez, J. D., Knight, L. E., Al-Khawaldeh, I., Gai, C., Lin, S., Martin, M. P., Miller, D. C., Cano, C., Endicott, J. A., Hardcastle, I. R., Noble, M. E. M. & Waring, M. J. (2019). J. Med. Chem. 62, 3741–3752. Web of Science CrossRef CAS PubMed Google Scholar
Zimmermann, M. O., Lange, A., Wilcken, R., Cieslik, M. B., Exner, T. E., Joerger, A. C., Koch, P. & Boeckler, F. M. (2014). Future Med. Chem. 6, 617–639. Web of Science CrossRef CAS PubMed Google Scholar
Zwart, P. H., Banumathi, S., Dauter, M. & Dauter, Z. (2004). Acta Cryst. D60, 1958–1963. Web of Science CrossRef CAS IUCr Journals Google Scholar
This is an open-access article distributed under the terms of the Creative Commons Attribution (CC-BY) Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original authors and source are cited.
| STRUCTURAL BIOLOGY |
