Computationally restoring the potency of a clinical antibody against Omicron

Abstract

The COVID-19 pandemic underscored the promise of monoclonal antibody-based prophylactic and therapeutic drugs^1-3 and revealed how quickly viral escape can curtail effective options^4,5. When the SARS-CoV-2 Omicron variant emerged in 2021, many antibody drug products lost potency, including Evusheld and its constituent, cilgavimab^4-6. Cilgavimab, like its progenitor COV2-2130, is a class 3 antibody that is compatible with other antibodies in combination⁴ and is challenging to replace with existing approaches. Rapidly modifying such high-value antibodies to restore efficacy against emerging variants is a compelling mitigation strategy. We sought to redesign and renew the efficacy of COV2-2130 against Omicron BA.1 and BA.1.1 strains while maintaining efficacy against the dominant Delta variant. Here we show that our computationally redesigned antibody, 2130-1-0114-112, achieves this objective, simultaneously increases neutralization potency against Delta and subsequent variants of concern, and provides protection in vivo against the strains tested: WA1/2020, BA.1.1 and BA.5. Deep mutational scanning of tens of thousands of pseudovirus variants reveals that 2130-1-0114-112 improves broad potency without increasing escape liabilities. Our results suggest that computational approaches can optimize an antibody to target multiple escape variants, while simultaneously enriching potency. Our computational approach does not require experimental iterations or pre-existing binding data, thus enabling rapid response strategies to address escape variants or lessen escape vulnerabilities.

Fig. 1. Application of the GUIDE computationally driven drug engineering platform to Omicron.

Given a parental antibody and target antigens, a design space was defined and a collection of co-structures were estimated (left). Within the computational design phase (centre), a sequence generator used predictions of multiple properties to propose multi-point mutant antibody candidates, and a Bayesian optimization agent selected proposed sequences that were then simulated. On the basis of Pareto optimality, mutational distance and sequence diversity, 376 computationally evaluated sequences were selected and experimentally evaluated for binding in immunoassays (centre right). The best sequences were then evaluated for neutralization of SARS-CoV-2 variants, and the single best sequence was identified (right). See Supplementary Methods for details. FEP, free energy perturbation; MD, molecular dynamics; SFE, structural fluctuation estimation.

Fig. 2. Computationally designed IgG antibodies improve Omicron binding and maintain parental thermostability and binding to historical strains.

a, The parental COV2-2130 (orange circles) and computationally designed antibodies (2130-1-0114-112 in purple triangles, 2130-1-0104-024 in blue diamonds and remainder in black) were assayed for thermal shift (n = 3 technical replicates). Melting temperature (T_m ) calculated based on the Boltzmann method. Data are mean and s.d. b–e, The parental COV2-2130 antibody and computationally designed antibodies (see symbols in a) and cross-reactive positive control antibody S309 (magenta squares) were analysed for relative binding to four SARS-CoV-2 spike RBD variants in the Gyrolab immunoassay: WA1/2020 (b), Delta B.1.617.2 (c), Omicron BA.1 (d) and Omicron BA.1.1 (e). Lines represent a four-parameter logistic regression model fit using GraphPad Prism to each titration, executed without technical replicates. mAb, monoclonal antibody.

Fig. 3. Designed antibodies improve pseudoviral neutralization over COV2-2130.

a–e, The parental COV2-2130 antibody (orange circles), the cross-reactive positive control antibody S2K146 (magenta squares), the negative control antibody DENV-2D22 (grey x) and down-selected computationally designed antibodies (symbols as indicated in the key) were assayed by neutralization with lentiviruses pseudotyped with spike variants of WA1/2020 D614G (a), Omicron BA.1 (b), Omicron BA.1.1 (c), Omicron BA.2 (d) and Omicron BA.4 (e). Curves are four-parameter logistic regression models fit to two (a–d) or four (e) replicate serial dilutions using GraphPad Prism.

Fig. 4. 2130-1-0114-112 provides in vivo prophylactic protection against SARS-CoV-2 variants.

a–i, Eight-week-old female K18-hACE2 mice were administered 100 μg (approximately 5 mg kg⁻¹) of the indicated monoclonal antibody treatment by intraperitoneal injection 1 day before intranasal inoculation with 10⁴ focus-forming units (FFU) of WA1/2020 D614G (a,d,g), Omicron BA.1.1 (b,e,h) or Omicron BA.5 (c,f,i). Tissues were collected 4 days after inoculation. Viral RNA levels in the lungs (a–c), nasal turbinates (d–f) and nasal washes (g–i) were determined by RT–qPCR (lines indicate median of log₁₀ values); n  =  9 (WA1/2020 D614G and BA.1.1 isotype control groups) or 10 (all others) mice per group, from two experiments. The limit of assay detection is shown as a horizontal dotted line. Statistical comparisons between groups were by Kruskal–Wallis ANOVA with Dunn’s multiple comparisons post-test; P values are as listed or not significant (NS) if P > 0.05. All analyses were conducted in GraphPad Prism. Source data

Fig. 5. COV-2130 and 2130-1-0114-112 escape mapping using DMS.

a,b, Comparison between IC₅₀ values measured using DMS for COV-2130 and 2130-1-0114-112 antibodies in BA.1 (a) and BA.2 (b) backgrounds, with key mutations highlighted. Arbitrary units in both plots are on the same scale. Interactive plots that display each mutation can be found at https://dms-vep.org/SARS-CoV-2_Omicron_BA.1_spike_DMS_COV2-2130/compare_IC50s.html for the BA.1 background and at https://dms-vep.org/SARS-CoV-2_Omicron_BA.2_spike_DMS_COV2-2130/compare_IC50s.html for the BA.2 background. c,d, Heatmaps of mutation escape scores at key sites for each antibody in BA.1 (c) and BA.2 (d) backgrounds. Escape scores were calculated relative to the wild-type amino acid in the same virus background. X marks wild-type amino acid in the relevant background. Amino acids not present in the DMS libraries lack squares; grey squares are mutations that strongly impair spike-mediated infection. Mutations identified in a,b are shown with a heavy line surrounding the corresponding box in c,d. Interactive heatmaps for full spike can be found for the BA.1 background at https://dms-vep.org/SARS-CoV-2_Omicron_BA.1_spike_DMS_COV2-2130/COV2-2130_vs_2130-1-0114-112_escape.html and https://dms-vep.org/SARS-CoV-2_Omicron_BA.2_spike_DMS_COV2-2130/COV2-2130_vs_2130-1-0114-112_escape.html for the BA.2 background.

Fig. 6. Cryo-EM structure of neutralizing antibodies 2130-1-0114-112 in complex with Omicron BA.2 RBD.

a, Atomic model of the RBD–Fab complex, coloured by chain: BA.2 RBD in red, 2130-1-0114-112 HC in yellow and 2130-1-0114-112 LC in green. BA.2 RBD mutations are in orange, and 2130-1-0114-112 mutations are in cyan and blue (HC and LC) (left). A close-up view of the RBD–Fab interface, showing WA1 RBD (Protein Data Bank 7L7E, light brown shading) aligned with the BA.2 RBD (right). b–d, Details showing the 2130-1-0114-112 modified residues and their interaction with BA.2 RBD, coloured as in a. Residue labels are shown in black for the BA.2 complex and brown for the overlaid WA1-2130 complex. The orange and green dashed lines indicate hydrogen bond and hydrophobic interactions, respectively; the yellow dashed lines are labelled with distances. CDRH3 residue Glu112 (left) and with the surface coloured by electrostatic potential (right), showing the positive and negative charges of RBD Lys444 and CDRH3 Glu112 (b). CDRL1 Ala32 and Ala33 hydrophobic network (left) and with the nearby RBD surface coloured by hydrophobicity (right; orange to cyan indicates hydrophobic to hydrophilic) (c). CDRL2 Glu59 salt bridge with RBD residue Arg498 (d). e, 2D diagram of Fab 2130-1-0114-112 paratope and epitope residues involved in hydrogen bonds and salt bridges (yellow and red dashed lines, respectively; distances in Å) and hydrophobic interactions (curved lines with rays). Atoms are shown as circles, with oxygen, carbon and nitrogen in red, black and blue, respectively. Interacting residues that belong to CDR loops are coloured in corresponding shades. The asterisks indicate mutated residues. Image created with Ligplot+.

Extended Data Fig. 1. Optimized, single-concentration screening data in Gyrolab immunoassays allow selection of candidates from Set 1 (n = 230) for down-stream characterization.

a-d, Parental mAb COV2-2130 (orange circles) and positive control mAb S309 24 (magenta squares) serve as references for computationally designed mAbs in single-concentration immunoassays. Computationally designed antibodies are shown as gray diamonds; selected computationally designed antibodies are highlighted with other colors and symbols as shown in the legend. Target antigens are (a) wild type WA1/2020, (b) Delta, (c) Omicron BA.1 and (d) Omicron BA.1.1. Each screened antibody and antigen combination was evaluated in two replicate assays, shown as individual points in the plot. Each of the controls was replicated in two replicate assays for each of three groups of antibodies, resulting in six replicate assays for each point on the control curves. Control curve points and error bars indicate mean and standard deviation. Curves represent a four-parameter logistic curve fit to the control data. All analysis performed in GraphPad Prism.

Extended Data Fig. 2. ELISA screening allows set for down selection of candidates from Set 2 (n = 204), for down-stream characterization.

a-d, Parental mAb COV2-2130 (orange circles; 3 technical replicates) and positive control mAb S309 24 (magenta squares; 3 technical replicates) serve as references for computationally designed mAbs (black curves). Purple triangles are 2130-1-114-112; blue diamonds are 2130-1-0104-024. Each designed antibody had a single measurement (n = 1) at each concentration. All curves are 4-parameter logistic fits, produced in GraphPad Prism. Target antigens are RBD from (a) wild type WA1/2020, (b) Omicron BA.1 (Acro), which is biotinylated, (c) Omicron BA.1, and (d) Omicron BA.1.1. For the biotinylated antigen in (b) (Acro Biosystems, cat. SPD-C82E4), an additional coat and wash cycle was required to prepare the ELISA plate with streptavidin.

Extended Data Fig. 3. 2130-1-0114-112 is potent in focus reduction neutralization tests with authentic virus in Vero-TMPRSS2 cells.

a-i, 2130-1-0114-112 potently neutralizes (a) WA1/2020 D614G (b) Delta B.1.617.2, (c) Omicron BA.1, (d) Omicron BA.1.1, (e) Omicron BA.2, (f) Omicron BA.2.12.1, (g) Omicron BA.4, (h) Omicron BA.5, and (i) Omicron BA.5.5 authentic viruses in focus reduction neutralization assays in Vero-TMPRSS2 cells. Symbols indicate the values of two technical replicates; curves are 4-parameter logistic regression models fit of normalized data. j, IC50 values and 95% confidence intervals corresponding to a-i. “>” indicates IC50 values > 10,000; “NC” indicates fits that were unconverged, unstable, or with positive hill slope. Analyses were performed in GraphPad Prism.

Extended Data Fig. 4. Authentic virus neutralization in plaque assays using Vero E6-TMPRSS2-T2A-ACE2 (VAT) cells.

a-c, Plaque assay neutralization of (a) Delta, (b) Omicron BA.1 and (c) Omicron BA.1.1 viruses. Data are represented as the normalized infection of mAb-treated virus to virus treated with control human IgG (Invitrogen). For S309, each point shows one of four technical replicates; all other points show one of two technical replicates at each concentration. Curves are two-parameter (IC50, hill-slope) logistic fits to normalized response. d, IC50 values (ng/ml, “>” indicates IC50 greater than 10,000 ng/mL) show that 2130-1-0104-024, while having only two mutations from COV2-2130, remains potent against BA.1 and suffers a 20-times loss in potency against BA.1.1. 2130-1-0114-112 is strongly potent against all three tested variants. All analysis performed in GraphPad Prism.

Extended Data Fig. 5. Normalized pseudoviral neutralization of SARS-CoV-2 VOCs by 2130-1-0114-112.

a, f, COV2-2130 and 2130-1-0114-112 potently neutralize WA1/2020 D614G. b, COV2-2130 does not potently neutralize BA.1, whereas 2130-1-0114-112 does, consistent with other pseudoviral neutralization assays. c, 2130-1-0114-112 potently neutralizes BA.2.75, outperforming COV2-2130 by 90-fold. d, e, 2130-1-0114-112 loses substantial potency in the context of BA.4.6 and artificially-produced BA.2.75 + R346T but retains measurable neutralization, demonstrating mitigation of this critical weakness of COV2-2130. g-j, 2130-1-0114-112 maintains weak neutralization of BA.2.75 R346T F486S (g), but loses detectable neutralization activity against BQ.1.1 (h) and XBB (i) and exhibits a near-complete loss of neutralization activity against BN.1 (j). Symbols indicate two independent technical replicates at each concentration. Curves represent a four-parameter logistic curve fit to the data with fixed minimum and maximum values (0 and 100, respectively). k, Neutralization IC50 values in ng/ml. “>” indicates IC50 greater than 10,000 ng/mL. “NC” indicates failure to converge. All analysis conducted in GraphPad Prism.

Extended Data Fig. 6. Details and VOC comparisons for Cryo-EM structure of neutralizing antibody 2130-1-0114-112 in complex with Omicron BA.2 RBD.

a, RBD residues within 7 Å of 2130-1-0114-112. RBD shown in red, with BA.2 mutated residues in orange, and 2130-1-0114-112 in yellow/green. b, CDRH3 Glu112 and RBD Lys440, shown with the EM map and distance between the side chains. c, d, 3D representation of the interaction plot between RBD and 2130-1-0114-112 HC (c, yellow) and LC (d, green). 2130-1-0114-112 is shown as stick and RBD as gray spheres with the contact residues in red. Contact residues are labelled and numbered. e, Fab COV-2130 paratope and epitope residues involved in hydrogen bonding (dashed lines; distances in Å) and hydrophobic interactions with WA1/2020 RBD; compare with Fig. 6d showing BA.2/2130-1-0114-112 interactions. Residues forming hydrophobic interactions are shown as curved lines with rays. Atoms are shown as circles, with oxygen, carbon, and nitrogen in red, black, and blue respectively. Image created with Ligplot + . f, Atomic model of the RBD-Fab complex superimposed with WA1-RBD (light brown PDB: 7L7E), XBB1.1-RBD (pink PDB: 8IOS), and BQ1.1 (gray PDB 8IF2). BA.2 RBD is shown in red, with BA.2 mutations in orange. 2130-1-0114-112 HC and LC are yellow and green, with mutations in cyan and blue. Hydrogen bonds are shown as dashed lines.

Extended Data Fig. 7. Sequence logos of candidate antibody designs.

a, The set of 376 designed IgG includes mutation at 16 positions in the heavy chain (blue; mutations in green) and 9 positions in the light chain (magenta; mutations in pink). Height of each letter is proportional to the frequency of the amino acid in the group. b, c, This set of 376 sequences is divided into two overlapping sets, Set 1 (b; n = 230) and Set 2 (c; n = 204). d, From these two sets, a set of eight sequences was selected for production at larger scale and further evaluation including assessment of their thermostability and binding performance (Fig. 2). Selected sequences show reduction in mutations throughout the CDRH3 residues (103-118) mutated in (a), especially in residues 103-108.

Extended Data Fig. 8. Strains with mutated epitopes are differentially neutralized by COV2-2130 and 2130-1-0114-112.

a, b, Starting from the 7L7E structure (a), shown after being separated and rotated to show the contact surfaces (b), complexes of COV2-2130 and 2130-1-0114-112 were naively composed by rigid body superposition of the highest resolution structures available of the VOC RBDs onto the WA1/2020 RBD of 7L7E. Epitope and paratope are outlined (solid black) and colored according to charge (blue for positive, red for negative). Residues are correspondingly outlined according to charge or, if uncharged, hydrophilicity (cyan) or hydrophobicity (yellow). The models were used to infer possible intermolecular clashes and loss of key interactions that could account for loss of affinity and, conversely, relief of clashes or new favorable interactions that could account for gain of affinity. For putative interactions that differ among the antibody and RBD combinations displayed, lines show salt bridges (purple) and hydrogen bonds (cyan). c, The interface between COV2-2130 and WA1/2020 RBD shows a number of favorable electrostatic interactions, including between RBD R346 and HC D56 and RBD K444 and HC D107. d-f, In the modeled interaction between 2130-1-0114-112 and BA.1 (d, RBD from PDB 7X66), separated and rotated (e), and with interactions identified (f) the RBD G446S substitution changes the center of the epitope, resulting in a loss of binding with COV2-2130, perhaps by means of clashes. 2130-1-0114-112 may rescue lost affinity by introduction of LC 59E that may form a favorable interaction with R498 and S446. Also 2130-1-0114-112 may introduce a favorable interaction with all of the Omicron variants by forming a salt bridge between HC 112E and K440. Further, the designed mutations LC S32A and S33A in 2130-1-0114-112 could enhance favorable hydrophobic interaction across the interface in the A484 region. g, BA.1.1 RBD (RBD from PDB 7XAZ) has all of the mutations of BA.1, but additionally contains the mutation R346K, which may disrupt the favorable interaction with HC D56. h, BA.2 RBD (RBD from PDB 8GB8) lacks the particularly unfavorable G446S mutation present in BA.1 and BA.1.1, and the R346K mutation of BA.1.1. BA.2 also has the K440 and A484 residues that are the interaction partners of 2130-1-0114-112’s designed mutations HC E112, LC A32 and LC A33. This combination of preservation of WA1/2020 interactions and the addition of newly exploited interaction partners results in strong neutralization by 2130-1-0114-112. i, j, BQ.1.1 (i, RBD from 8IF2), and XBB.1 (j, RBD from 8IOS) which are not neutralized by either COV2-2130 or 2130-1-0114-112, contain substitutions in the epitope at positions 346 and either 444 or 445 as compared to WA1/2020. All three of R346, K444, and V445 are among the most highly buried residues in the epitope, and the mutation of R346 and K444 removes the two salt bridges formed with these residues. Consequently, substitutions at these positions heavily impact the shape and charge complementarity of both COV2-2130 and 2130-1-0114-112 to the RBD, likely explaining much of the impact to their affinities.