44 45 D I A M O N D L I G H T S O U R C E A N N U A L R E V I E W 2 0 2 0 / 2 1 D I A M O N D L I G H T S O U R C E A N N U A L R E V I E W 2 0 2 0 / 2 1 Biological Cryo-Imaging Group eBIC (and beamline I03 from theMacromolecular Crystallography Group) Mapping antibody recognition of SARS-CoV-2 spike protein Related publication: DejnirattisaiW., Zhou D., Ginn H. M., Duyvesteyn H. M., Supasa P., Case J. B., ZhaoY.,WalterT. S., Mentzer A. J., Liu C.,Wang B., Paesen G. C., Slon-Campos J., Lopez-Camacho C., Kafai N. M., Bailey A. L., Chen R. E.,Ying B.,Thompson C. P., Fyfe A., Gupta S.,TanT., Gilbert-Jaramillo J., JamesW., Buttigieg K., Coombes N., Carroll M.W., Skelly D., Dold C., PengY., Levin R., DongT., Pollard A., Knight J., Klenerman P.,Temperton N., Hall D. R.,Williams M. A., Paterson N. G., BertramF. K. R., Seibert C. A., Clare D. K., Howe A., Raedecke J., SongY.,Townsend A., Huang K.-Y. A., Fry E. E., Mongkolspaya J., DiamondM. S., Ren J., Stuart D. I. & Screaton G. R.The Antigenic Anatomy of SARS-CoV-2 Receptor Binding Domain. Cell 184 , (2021). DOI: 10.1016/j.cell.2021.02.032 Publication keywords: SARS-CoV-2; Antibody; Germline;V-gene; Receptor-binding-domain; Spike; Neutralisation; Protection; Glycosylation T he COVID-19 vaccines currently in use in Europe and the USA aim to generate antibodies against the virus spike. We need to know where these antibodies bind and how they neutralise the virus. Viruses mutate to escape antibody binding, and changes to the spike structure can stop antibody attachment. An international team of researchers identified 377 human monoclonal antibodies (mAbs) from recovered SARS-CoV-2 patients that recognised the virus spike. They used cryo-electron microscopy (cryo-EM) at the Electron Bio-Imaging Centre (eBIC) and Macromolecular Crystallography (MX) on beamline I03 at Diamond Light Source to investigate the complex structures of the SARS-CoV-2 spike or the receptor binding domain (RBD) with the antibody Fabs (a region on an antibody that binds to antigens). Theteamdeterminedthestructuresof11spike/Fabcomplexesbycryo-EMand18RBD/fabcomplexesbycrystallography.Usingacombination of structural methods, and a novel computational algorithm utilising competition bio-layer interferometry data, they localised binding epitopes of 80on the surfaceof theRBD.Threeof thepotentneutralisingmAbs areglycosylated, and theglycans contribute toneutralisation. Their results identify the precise binding sites on the spike and their detailed interactions. This information can guide combinations for antibody cocktail therapy. More potent neutralising monoclonal antibodies (mABs) can be designed through structural analysis. Understanding how the binding of these antibodies is affected in variant viruses is helpful in understanding how we might design next- generation vaccines. The project involvedmany groups working closely together, and success in such a difficult timewas only possible due to tremendous support and collaboration with eBIC, the Oxford Particle Imaging Centre (OPIC) and I03. The current COVID-19 pandemic has had an unprecedented impact on both human health and economics globally. The causative agent, SARS-CoV-2, is related to SARS-CoV-1 and MERS-CoV, and all are beta coronaviruses causing severe respiratory syndromes. The spike, a glycoprotein, is the only molecule protruding outside the virus membrane, and is composed of two subunits, S1 which mediates receptor binding and S2 responsible for viral and host cell membrane fusion. The prefusion spike has an elongated trimeric structure that transitions to the post-fusion state by cleavage between S1 and S2 following receptor binding or trypsin treatment. This conformational change allows the virus to infect the host cell. The S1 fragment occupies the membrane distal tip of S and can be subdivided into an N-terminal domain (NTD) and the receptor binding domain (RBD). SARS-CoV-2 infected or vaccinated individuals produce a large number of antibodies which are crucial for immune protection against the virus. We obtained 377 monoclonal antibodies (mAbs) targeting the viral spike glycoprotein from B-cells in the blood of convalescent patients. 80 of these react with the RBD and include nearly all the potent neutralisers (which we define as giving 50% virus neutralisation at < 0.1 µg/ml). To acquire greater insight to the mAb binding sites on the RBD, we measured pairwise competition between antibodies using biolayer interferometry (BLI) in a 96- well plate format. 79 of the current set of mAbs (one has known structure), and three ‘external’ antibodies of known binding positions, were used. Combining the BLI data and the four known binding positions, we were able to map the positions of the mAbs on the surface of the RBD. The predicted locations, covering most of the RBD surface, clustered into five groups and were named by analogy to a human torso (Fig. 1). There is generally good correlation between overlap with ACE2 footprint and neutralisation. However, there were notable examples of non-neutralising antibodies that were good ACE2 blockers. Interestingly, antibodies co-locating with known neutralising/protecting antibodies EY6A and S309, in the left and right flank clusters respectively, did not show appreciable neutralisation in our assays 1,2 . To confirm the mapped positions of the mAbs and visualise detailed interactions with the antigen, we determined 11 spike/Fab cryo-EM and 18 RBD/Fab crystal structures with excellent support from eBIC, Oxford Particle Imaging Centre (OPIC) and the Macromolecular Crystallography (MX) beamline I03 (a Fab is a fragment of the antibody that binds the antigen). The complex structures include 10 natural and two hybrid mAbs, one mAb (159) that binds at the NTD, some had two Fabs bound at the same time, and some were with mutant RBD of the virus variants of concern 3,4 . The binding positions of the mAbs agreed with the predicted locations, with an average error of 7.6 Å . In the isolated prefusion trimeric spike, the RBD is mainly found in two orientations, ‘up’ and ‘down’. The cell receptor ACE2 can only bind the RBD in the up conformation. The most common configuration of the spike observed in the present study is one RBD-up and two RBD-down, with a Fab bound only to the RBD in the up conformation (Fabs 40, 150, 158, and the chimeras 253H55L and 253H165L). Fabs 316 and 384 bind the spike in all RBD-down configuration with Fab:spike stoichiometry of 3:1 and 1:1 respectively. The chimera 253H55L can also bind the spike in all RBD-down configuration with 1:1 stoichiometry. In contrast, 3 Fabs of mAb 88 bind the spike in all RBD-up configuration, which we observed previously for EY6A, despite the fact that these two mAbs recognise quite different epitopes on the RBD 2 . In the case of mAb 159, the most potent neutraliser (IC 50 of 5ng/mL) that interacts with the NTD, three Fabs bind the spike in either all RBD-down or one RBD-up configuration (Fig. 2). Antibodies that recognise glycan epitopes are well documented. In contrast, the role of sugars on the variable part of the antibody, although found in 15%-25% of immunoglobulin Gs (IgGs), is poorly studied at the molecular level. Fourteen of 80 RBD-binding antibodies in the present study contain glycosylation sequons arising from somatic mutations in their variable region. Three of the Fabs we determined structures for, 88, 253 and 316 have glycosylation sites, in CDRs H1, H3 and H2 respectively. The glycans are presented close to the top of the left shoulder of the RBD. In two out of three cases the sugars interact directly but rather weakly with the antigen. mAbs 88 and 316 could be de-glycosylated without denaturation, and BLI analysis showed that this had negligible effect on RBD/Fab binding affinities. However, mutations that eliminate glycosylation had a deleterious effect on neutralisation for these two and the 253H165L chimera. The most potent RBD-binding mAbs bind in the regions of the neck and left shoulder, achieving neutralisation by blocking the ACE2 binding (Fig. 3). There is a close association between potent neutralisers and public V-genes (antibody genes shared by many people) suggesting that vaccination responses should be strong. Three public heavy chain V-region genes are represented at least twice in our set, i) IGHV3-53: mAbs 150, 158, 175, 222 and 269, ii) IGHV3-66: 282 and 40 (this is very similar to IGHV3-53) and iii) IGHV1-58: 55, 165, 253 and 318. mAb 40 together with IGHV3-53 mAbs are all bound in the neck region with an almost identical pose determined by their highly conserved CDRs H1 and H2. Notably IGHV3-30 is found in 11 RBD binders, none of which are potent neutralisers. Structures of two representatives, 45 and 75, show binding on the left flank and right shoulder respectively. By switching light chains within the same IGHV set, we found a 10-fold increase in neutralisation potency when the heavy chain of mAb 253 was combined with the light chains of mAbs 55 and 165. mAb 384, an IGHV3-11 antibody, the most potent neutraliser with IC 50 of 2 ng/ml, adopts a unique pose with a footprint extending from the left shoulder epitope to the neck epitope, via its extended CDR H3. We tested neutralisation of the 20 most potent neutralisers against three variants of concern: B.1.1.7, B.1.351 and P.1. All three variants carry a N501Y mutation, B.1.351 and P.1 have additional E484K, and K417N in the former and K417T in the latter in the RBD. Neutralisation of the majority of the mAbs are affected by one or more variants, but 222 and 253 are unaffected 3-5 . 253 has no direct contact with any of the mutations, while 222 contacts residues 501 and 417. Complex structures of 222 with the RBD which carry the N501Y mutation show P30 of 222 makes stacking contact with Y501, strengthening the interaction with the mutant RBD. The structural information has prompted us to change the light chain of other IGHV3-53 mAbs to that of 222, which restored the activity of mAbs 150 and 158 against all three variants. In fact, the modified mAb 150 is 50-fold and 13-fold more potent against B.1.351 and P.1, than the original 150, respectively. References: 1. Huo J. et al. Neutralization of SARS-CoV-2 by Destruction of the Prefusion Spike. Cell Host Microbe 28 , 445-454.e6 (2020). DOI: 10.1016/j. chom.2020.06.010 2. Zhou D. et al. Structural basis for the neutralization of SARS-CoV-2 by an antibody from a convalescent patient. Nat. Struct. Mol. Biol. 27 , 950–958 (2020). DOI: 10.1038/s41594-020-0480-y 3. DejnirattisaiW. et al. Antibody evasion by the P.1 strain of SARS-CoV-2. Cell (2021). DOI: 10.1016/j.cell.2021.03.055 4. Supasa P. et al. Reduced neutralization of SARS-CoV-2 B.1.1.7 variant by convalescent and vaccine sera. Cell 184 , 2201-2211.e7 (2021). DOI: 10.1016/j.cell.2021.02.033 5. Zhou D. et al. Evidence of escape of SARS-CoV-2 variant B.1.351 from natural and vaccine-induced sera. Cell 184 , 2348-2361.e6 (2021). DOI: 10.1016/j. cell.2021.02.037 Funding acknowledgement: WellcomeTrust (grant 095541/A/11/Z). UK Medical Research Council (MR/ N00065X/1). Corresponding authors: Dr Jingshan Ren, Division of Structural Biology, University of Oxford,
[email protected]; Prof. David I. Stuart, Division of Structural Biology, University of Oxford, and Diamond Light Source,
[email protected] Figure 1: Epitope mapping of RBD binding antibodies. The RBD is shown as grey surface representation with region labelled in analogy to the torso. The footprint of the receptor ACE2 is coloured in green. Antibodies are shown as spheres positioned based on the BLI competition data and colour-coded according to their ability to neutralise SARS-CoV-2. Figure 2: Cryo-EM structures of SARS-CoV-2 spike and Fab complexes. Each structure is drawn as surface representation with chains A, B and C of the spike, and heavy chain and light chain of Fab coloured in green, cyan, magenta, yellow and salmon, respectively. Note, in panels (a), (d), (e), (j) and (k) only the VhVl domains of the Fabs are modelled. Figure 3: The binding positions and orientations of 12 Fabs whose structures with RBD or spike are determined by either cryo-EM or X-ray crystallography. The RBD is shown as grey surface, and Fabs are drawn as ribbons.