Antibody Therapeutics

Monoclonal antibodies (mAbs) and fusion proteins for immune applications (FPIA) play a crucial role in treating autoimmune diseases and cancers by targeting cell-surface proteins and triggering multiple immune mechanisms. These functions are mediated by the fragment crystallizable (Fc) region of mAbs and fusion proteins, whose interaction with Fc gamma receptors (Fc{gamma}Rs) can be modulated through Fc amino acid (AA) engineering. To address this, we developed the IMGT/FcVariantsExplorer tool (https://www.imgt.org/fcvariantsexplorer/) to identify AA changes within the Fc region in mAb and fusion proteins sequences from IMGT/2Dstructure-DB, the AA sequence database of IMGT(R), the international ImMunoGeneTics information system(R). We used the IMGT(R) nomenclature of engineered Fc variants involved in antibody effector properties and formats, applying a standardized classification in five categories: Effector, Half-life, Physicochemical properties, Structure, and Hybrid. We analyzed sequences of 1,107 mAbs and fusion proteins, identifying 483 entries with Fc AA changes, resulting in 211 unique Fc variants in the dataset. We also used web scraping to retrieve associated biological data from literature. All data have been integrated into IMGT/mAb-DB, with links to sequences in IMGT/2Dstructure-DB, enabling users to query Fc variants by their Category or Effect. This curated dataset reveals key trends in antibody engineering.

Multispecific antibodies are prominent therapeutic agents, but many molecular formats and drug candidates that show promise during molecular discovery stages cannot be scaled up and developed into drugs due to inadequate developability. During the discovery stages, the selection of molecule format(s), molecule design, purity, and initial physiochemical stability testing criteria largely rely on scientists experience. Machine learning, however, can identify hidden trends in large datasets, aiding in the selection of drug candidates with improved developability. In this study, we present a machine learning approach to predict antibody purity, measured by the percentage of monomer after protein A purification. Using the amino acid sequences of variable regions, molecular formats, germlines and germline pairings, and calculated physiochemical properties as inputs, machine learning models were trained to predict the percentage of monomer for a given multispecific antibody (Figure 1). The dataset employed in this study consists of [~]500 multi-specific antibodies generated during BIs internal drug discovery programs. Our results indicate that machine learning, when applied to sequence, germline, and format data, can effectively predict antibody percentage of monomer. Incorporating this approach into high-throughput multispecific antibody screening processes can save time and resources by reducing the need to test a large subset of potentially unstable antibodies. While this study focused on percentage of monomer as a test case, similar approaches can be employed to predict other antibody properties, such as melting temperature (Tm), hydrophobicity (aHIC), and solution stability properties (AC-SINS). O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=47 SRC="FIGDIR/small/570217v1_fig1.gif" ALT="Figure 1"> View larger version (11K): org.highwire.dtl.DTLVardef@f23246org.highwire.dtl.DTLVardef@c2b81aorg.highwire.dtl.DTLVardef@1c4e59dorg.highwire.dtl.DTLVardef@1bed683_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOFigure 1C_FLOATNO Overview of ML model for predicting multispecific antibody purity from sequence, germline and format information. C_FIG

Antibody based therapeutics targeting mesothelin (MSLN) have shown limited anticancer activity in clinical trials. Novel antibodies with high affinity and better therapeutic properties are needed. In the current study, we have isolated and characterized a novel VH domain 3C9 from a large size human immunoglobulin heavy chain variable (VH) domain library. 3C9 exhibited high affinity [KD (dissociation constant) < 3 nM] and binding specificity in a membrane proteome array (MPA). In a mouse xenograft model, 3C9 fused to human Fc became visible at tumor sites as early as 8 hours post infusion and persisted at the tumor site for more than 10 days. Both CAR-T cells and antibody domain drug conjugations (DDCs) generated with 3C9 were highly effective at killing MSLN positive cells in vitro without off-target effects. The X-ray crystal structure of full-length MSLN in complex with 3C9 reveals interaction of the 3C9 domains with two distinctive residues patches on the MSLN surface. 3C9 fused to human Fc domain drug conjugate was efficacious to inhibit tumor growth in a mouse xenograft model. This newly discovered VH antibody domain holds promise as a therapeutic candidate for MSLN-expressing cancers.

The emergence of highly immune invasive and transmissible variants of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has decreased the effectiveness of existing vaccines. It is, therefore, critical to develop effective and safe therapeutics for SARS-CoV-2 infections, especially for the most vulnerable and immunocompromised patients. Neutralizing antibodies have been shown to be successful at preventing severe disease from early SARS-CoV-2 strains, although their efficacy has diminished with the emergence of new variants. Here, we aim to develop fully human and broadly neutralizing monoclonal (mAb) and bispecific (BsAb) antibodies against SARS-CoV-2 and its variants. Specifically, we first identified two antibodies from human transgenic mice that bind to the receptor binding domain (RBD) of the SARS-CoV-2 spike protein and are capable of neutralizing SARS-CoV-2 and variants of concern with high to moderate affinity. Two non-competing clones with the highest affinity and functional blocking of ACE2 binding were then selected to be engineered into two BsAbs, which were then demonstrated to have relatively improved affinity, ACE2 blocking ability, and pseudovirus inhibition against several variants, including Omicron (B.1.1.529). Our findings provide one mAb candidate and two bsAb candidates for consideration of further clinical development and suggest that the bispecific format may be more effective than mAbs for SARS-CoV-2 treatment.

Monoclonal antibodies (mAbs) with long systemic persistence are widely used as therapeutics. However, antibodies with atypically fast clearance require more dosing, limiting their clinical usefulness. Deep learning can facilitate using sequence-based modeling to predict potential pharmacokinetic (PK) liabilities before antibody generation. Assembling a dataset of 103 mAbs with measured nonspecific clearance in cynomolgus monkeys (cyno), and using transfer learning from large protein language models, we developed multiple machine learning models to predict mAb clearance as fast/slow clearing. Focusing on minimizing misclassification of potentially promising molecules as fast clearing, our results show that using physicochemical properties yielded up to 73.1+/-1.1% classification accuracy on hold-out test data (precision 65.2+/-2.3%). Using only sequence-based features from deep learning protein language models yielded a comparable performance of 71+/-1.4% (precision 65.5+/-2.5%). Combining structural and deep learning derived features yielded a similar accuracy of 73.9+/-1.1%, and slightly improved precision (68.3+/-2.4%). Features important for classifying fast/slow clearance point to charge, moment, and surface area properties at pH 7.4 as well as deep learning derived features. These results suggest that the protein language models provide comparable information and predictive performance of clearance as physicochemical features. This work provides a foundation for in silico prediction of protein pharmacokinetics to inform antibody candidate generation and early deprioritization of designs with high risk of fast clearance. More generally, it illustrates the value of transfer learning-based application of protein language models to address characteristics of importance for protein therapeutics.

The epidermal growth factor receptor (EGFR) is a member of a family of transmembrane tyrosine kinase receptors that plays a pivotal role in regulating diverse cellular processes such as cell proliferation, survival, migration, and differentiation. Aberrant activation of EGFR signaling has been implicated in various pathological conditions, particularly cancer, making it an attractive target for therapeutic intervention. While several anti-EGFR monoclonal antibodies have been developed and demonstrated their clinical value for the treatment of various solid tumors, smaller antibody fragments such as nanobodies (Nb) offer distinct advantages over conventional antibodies, including reduced immunogenicity and enhanced tumor penetration. In this paper, we report the isolation and characterization of two novel high-affinity Nb targeting EGFR. These Nb were identified and characterized using ELISA, flow cytometry, microscopy, and SPR. Furthermore, these Nb and bivalent Nb engineered from them were tested for their effects on cancer cell proliferation. We demonstrate that the novel Nb exhibit high affinity and potent anti-tumor activity in vitro in their bivalent form, positioning them as promising candidates for cancer treatment.

Vascular cell adhesion molecule-1 (VCAM-1) plays an important role in inflammation, where it facilitates the recruitment of leukocytes to the inflamed area via leukocytes VLA-4 and endothelial cells VCAM-1 interaction. VCAM-1 expression is also upregulated in certain cancers. VCAM-1 has 7 Ig-like domains, with domains 1 and 4 shown to be critical for VLA-4 binding. However, the specific functions of individual VCAM-1 Ig-like domains remain poorly understood. In this study, we identified single-chain variable fragment (scFvs) antibodies targeting domains 2, 3, and 5 of VCAM-1, and investigated the ability of these antibodies to block VCAM-1-mediated cell adhesion to macrophages. We show that scFv antibodies against Ig-like domains 2 and 3 significantly interfere with the ability of macrophages to bind endothelial cells, suggesting that these domains also play a role in facilitating this interaction. These results emphasize the need to more carefully study the role of each domain on VCAM-1 function and highlight the potential of targeting these VCAM-1 domains for more tailored therapeutic interventions in inflammatory diseases and cancer.

Antibodies with high affinity against the receptor binding domain (RBD) of the SARS-CoV-2 S1 ectodomain were identified from screens using the Retained Display (ReD) platform employing a 1 x 1011 clone single-chain antibody (scFv) library. Numerous unique scFv clones capable of inhibiting binding of the viral S1 ectodomain to the ACE2 receptor in vitro were characterized. To maximize avidity, selected clones were reformatted as bivalent diabodies and monoclonal antibodies (mAb). The highest affinity mAb completely neutralized live SARS-CoV-2 virus in cell culture for four days at a concentration of 6.7 nM, suggesting potential therapeutic and/or prophylactic use. Furthermore, scFvs were identified that greatly increased the interaction of the viral S1 trimer with the ACE2 receptor, with potential implications for vaccine development.

We report a functional pipeline for facile conversion of variable Fv domains, typically discovered in antibody discovery programs, into chimeric monoclonal antibodies (mAbs). Often, in initial screenings, a set of candidate mAbs is produced in small volumes and purified from supernatant for testing. Our pipeline also simplifies purification of mAbs by using an extended histidine tag (His-10) fused to the C-terminus of the light chain. Both the length of the His-10 and its location have been shown to affect the efficacy of mAb purification using an inexpensive nickel-based resin at neutral pH. Our antibody cloning and purification pipeline, when followed together with detection and affinity measurements, can be smoothly incorporated into an antibody discovery workflow.

Accurate spectral unmixing is a critical step for flow cytometry data analysis and requires a single stain control for every fluorescent parameter used in an experiment. Currently, compensation particles are often used for making single stain controls when a target protein is of low abundance or a cell type is of low frequency. However, compensation particles introduce incongruencies in emission spectra compared to cells resulting in spectral unmixing or compensation errors. To enable the use of cells regardless of the abundance of target proteins or immune cell type, we generated a bispecific antibody that links a human anti-CD45 and mouse anti-IgG variable region. We refer to this new bispecific tool as CaptureBody (CB) and highlight the benefits of its final nanobody-based design. We provide all sequences and methods necessary for the in-house expression of a CaptureBody to disseminate their use for spectral flow cytometry experiments.

Dose selection and confirmation are critical tasks in the development of therapeutic antibodies. These tasks could become particularly challenging in the absence of robust pharmacodynamics biomarkers or at very flat dose-response curves. Although much knowledge has been acquired in the past decade, it remains uncertain which factors are relevant and how to select doses more rationally. In this study, we developed a quantitative metric, Therapeutic Exposure Affinity Ratio (TEAR), to retrospectively evaluate up to 60 approved antibodies and their therapeutic doses (TDs), and systematically assessed the factors that are relevant to antibody TDs and dose selection patterns. This metric supported us to analyze many factors that are beyond antibody pharmacokinetics and target binding affinity. Our results challenged the traditional perceptions about the importance of target turnovers and target anatomical locations in the selection of TDs, highlighted the relevance of an overlooked factor, antibody mechanisms of action. Overall, this study provided insights into antibody dose selection and confirmation in the development of therapeutic antibodies.

SARS-CoV-2 (Covid-19) has caused currently ongoing global plague and imposed great challenges to health managing systems all over the world, with millions of infections and hundreds of thousands of deaths. In addition to racing to develop vaccines, neutralizing antibodies (nAbs) to this virus have been extensively sought and are expected to provide another prevention and therapy tool against this frantic pandemic. To offer fast isolation and shortened early development, a large human naive phage display antibody library, was built and used to screen specific nAbs to the receptor-binding domain, RBD, the key for Covid-19 virus entry through a human receptor, ACE2. The obtained RBD-specific antibodies were characterized by epitope mapping, FACS and neutralization assay. Some of the antibodies demonstrated spike-neutralizing property and ACE2-competitiveness. Our work proved that RBD-specific neutralizing binders from human naive antibody phage display library are promising candidates to for further Covid-19 therapeutics development.

Plants are an efficient production platform for manufacturing glycoengineered monoclonal antibodies and antibody-like molecules. Avaren-Fc (AvFc) is a lectin-Fc fusion protein or lectibody produced in Nicotiana benthamiana, which selectively recognizes cancer-associated high-mannose glycans. In this study, we report the generation of a glycovariant of AvFc that is devoid of plant glycans, including the core 1,3-fucose and {beta}1,2-xylose residues. The successful removal of these glycans was confirmed by glycan analysis using HPLC. This variant, AvFc{Delta}XF, has significantly higher affinity for Fc gamma receptors and induces higher levels of luciferase expression in an antibody-dependent cell-mediated cytotoxicity (ADCC) reporter assay against B16F10 murine melanoma cells without inducing apoptosis or inhibiting proliferation. In the B16F10 flank tumor mouse model, we found that systemic administration of AvFc{Delta}XF, but not an aglycosylated AvFc variant lacking affinity for Fc receptors, significantly delayed the growth of tumors, suggesting that Fc-mediated effector functions were integral. AvFc{Delta}XF treatment also significantly reduced lung metastasis of B16F10 upon intravenous challenge whereas a sugar-binding-deficient mutant failed to show efficacy. Lastly, we determined the impact of anti-drug antibodies (ADAs) on drug activity in vivo by pretreating animals with AvFc{Delta}XF before implanting tumors. Despite a significant ADA response induced by the pretreatment, we found that the activity of AvFc{Delta}XF was unaffected by the presence of these antibodies. These results demonstrate that glycoengineering is a powerful strategy to enhance AvFcs antitumor activity.

Defining the binding epitopes of antibodies is essential for understanding how they bind to their antigens and perform their molecular functions. However, while determining linear epitopes of monoclonal antibodies can be accomplished utilizing well-established empirical procedures, these approaches are generally labor- and time-intensive and costly. To take advantage of the recent advances in protein structure prediction algorithms available to the scientific community, we developed a calculation pipeline based on the localColabFold implementation of AlphaFold2 that can predict linear antibody epitopes by predicting the structure of the complex between antibody heavy and light chains and target peptide sequences derived from antigens. We found that this AlphaFold2 pipeline, which we call PAbFold, was able to accurately flag known epitope sequences for several well-known antibody targets (HA / Myc) when the target sequence was broken into small overlapping linear peptides and antibody complementarity determining regions (CDRs) were grafted onto several different antibody framework regions in the single-chain antibody fragment (scFv) format. To determine if this pipeline was able to identify the epitope of a novel antibody with no structural information publicly available, we determined the epitope of a novel anti-SARS-CoV-2 nucleocapsid targeted antibody using our method and then experimentally validated our computational results using peptide competition ELISA assays. These results indicate that the AlphaFold2-based PAbFold pipeline we developed is capable of accurately identifying linear antibody epitopes in a short time using just antibody and target protein sequences. This emergent capability of the method is sensitive to methodological details such as peptide length, AlphaFold2 neural network versions, and multiple-sequence alignment database. PAbFold is available at https://github.com/jbderoo/PAbFold.

BackgroundAs the most abundant immunoglobulin in blood and the most common human isotype used for therapeutic monoclonal antibodies, the engagement and subsequent activation of its Fc receptors by IgGs are crucial for antibody function. While generally assumed to be relatively constant within subtypes, recent studies have shown the antibody variable regions to exert distal effects of modulating antibody-receptor interactions on many antibody isotypes. Such effects are also expected for IgG and its subtypes with the in-depth understanding of these V-region effects highly relevant for engineering antibodies, antibody purifications, and understanding to how robust the microbial immune evasion proteins are. MethodsIn this study, we created a panel of IgG2/IgG3/IgG4 antibodies by changing the VH family (VH1-7) frameworks while retaining the complementarity determining regions of Pertuzumab and measured the interaction of the IgGs with Fc{gamma}RIa, Fc{gamma}RIIaH167, Fc{gamma}RIIaR167, Fc{gamma}RIIb/c, Fc{gamma}RIIIaF176, Fc{gamma}RIIIaV176, Fc{gamma}RIIIbNA1, and Fc{gamma}RIIIbNA2 receptors alongside antibody superantigens proteins L and G using biolayer interferometry. ResultsThe library of 21 IgGs demonstrated that the VH frameworks influenced receptor binding sites on the constant region of the subtypes significantly, providing non-canonical interactions and non-interactions. However, there was minimal influence on the binding of bacterial B-cell superantigens Proteins L and G on the IgGs, showing their robustness against V-region effects. ConclusionsThese results demonstrate the importance of the V-regions during humanization of therapeutic antibodies that can confer or diminish FcR-dependent immune responses, while remaining both suitable and susceptible to the binding by bacterial antibody superantigens in antibody purification and be present with normal flora. STATEMENT OF SIGNIFICANCEIgGs are the predominant isotype for clinical and research applications. Despite the vast amount of research to study it, particularly on IgG1, there remains a gap in understanding how the variable regions and the receptor binding sites can influence one another in the other IgG subtypes, across the IgG subtypes with different hinges and makeup. This study investigates the effect of these variable regions on the engagement of receptors and also how bacterial antibody superantigens present in microflora and used in antibody purification can exert distal effects.

As the primary response antibody with increasing interest as a therapeutic antibody format, IgM is also the largest antibody structure among the five major human isotypes. Spontaneously forming pentamer and hexamers, IgM has avidity effects that could compensate for weaker interactions, although steric hindrances can occur for certain epitopes. With recent evidence of the heavy chain constant region affecting antigen binding and the VH families of the V-regions affecting FcR engagement found on other isotypes, we investigated CDR-grafted Trastuzumab and Pertuzumab VH1-7 IgMs for biolayer interferometry. From our panel of the 14 IgM variants, the V-regions holistically affected FcR binding, and the IgM C-region modulated Her2 engagements with contributions from the V-regions and influences from protein L binding at the V{kappa}. These findings revealed the oligomerization effect of IgMs to play a significant role in both FcR and antigen binding that is distinct from the other isotypes that can guide the development and protein en-gineering of IgM therapeutics.

Bispecific T cell engager (TCE) is an emerging anti-cancer modality which redirects cytotoxic T cells to tumor cells expressing tumor-associated antigen (TAA) thereby forming immune synapses to exerts anti-tumor effects. Considering the protein engineering challenges in designing and optimizing size and pharmacokinetically acceptable TCEs in the context of the complexity of intercellular bridging between T cells and tumor cells, a physiologically relevant and clinically verified computational modeling framework is of crucial importance to guide the process to understand the protein engineering trade offs. In this study, we developed a quantitative, physiologically based computational framework to predict immune synapse formation for a variety of molecular format of TCEs in tumor tissue. Our model incorporated the molecular size dependent biodistribution using the two pore theory, extra-vascularization of T cells and hematologic cancer cells, mechanistic bispecific intercellular binding of TCEs and competitive inhibitory interaction by shed targets. The biodistribution of TCE was verified by positron emission tomography imaging of [89Zr]AMG211 (a carcinoembryonic antigen-targeting TCE) in patients. Parameter sensitivity analyses indicated that immune synapse formation was highly sensitive to TAA expression, degree of target shedding and binding selectivity to tumor cell surface TAA over shed target. Interestingly, the model suggested a "sweet spot" for TCEs CD3 binding affinity which balanced the trapping of TCE in T cell rich organs. The final model simulations indicated that the number of immune synapses is similar ([~]50/tumor cell) between two distinct clinical stage B cell maturation antigen (BCMA)-targeting TCEs, PF-06863135 in IgG format and AMG420 in BiTE format, at their respective efficacious dose in multiple myeloma patients, demonstrating the applicability of the developed computational modeling framework to molecular design optimization and clinical benchmarking for TCEs. This framework can be employed to other targets to provide a quantitative means to facilitate the model-informed best in class TCE discovery and development. Author summaryCytotoxic T cells play a crucial role in eliminating tumor cells. However, tumor cells develop mechanisms to evade from T cell recognition. Bispecific T cell engager (TCE) is designed to overcome this issue with bringing T cells to close proximity of tumor cells through simultaneous bivalent binding to both tumor-associated antigen and T cells. After successful regulatory approval of blinatumomab (anti-CD19 TCE), more than 40 TCEs are currently in clinical development with a variety of molecular size and protein formats. In this study, we developed a quantitative computational modeling framework for molecular design optimization and clinical benchmarking of TCEs. The model accounts for molecular size dependent biodistribution of TCEs to tumor tissue and other organs as well as following bispecific intercellular bridging of T cells and tumor cells. The model simulation highlighted the importance of binding selectivity of TCEs to tumor cell surface target over shed target. The model also demonstrated a good agreement in predicted immune synapse number for two distinct molecular formats of TCEs at their respective clinically efficacious dose levels, highlighting the usefulness of developed computational modeling framework for best in class TCE discovery and development.

BackgroundIMV-M is a MUC16xDR5 bispecific antibody has demonstrated MUC16-selective anti-tumor activity. However, it remained unclear whether multiple binding of IMV-M on a single MUC16 molecule was required for IMV-M cytotoxicity, whether circulating CA125 could attenuate its efficacy or cause off-target toxicity, and whether anti-drug antibodies might induce IMV-M aggregation and related adverse effects. MethodsA comparative analysis of three bispecific antibodies, IMV-M (sofituzumabxDR5), 11D10xDR5, and fluorxDR5, sharing an identical IgG1-anti-DR5 scFv architecture, was performed. Sofituzumab binds to multiple epitopes on a single MUC16 molecule, whereas 11D10 binds a single MUC16 epitope, and fluor does not bind any human antigen. Antibody binding to shed and cell-surface MUC16 was evaluated by ELISA and flow cytometry. Cytotoxicity was assessed in a MUC16+/DR5+ tumor cell line and MUC16-/DR5+ hepatic cell lines. Additional studies examined the effects of soluble CA125 and Fc-directed polyclonal antibodies on IMV-M activity. ResultsIMV-M bound MUC16 to a markedly higher extent than the 11D10xDR5 comparator, consistent with its multivalent engagement, while binding of fluorxDR5 to MUC16 was negligent. Only IMV-M induced potent cytotoxicity in MUC16+ tumor cells, whereas 11D10xDR5 and fluorxDR5 control antibodies were inactive, demonstrating that multivalent clustering on MUC16 is required for apoptosis. IMV-M showed no significant cytotoxicity toward hepatic cell lines, even in the presence of Fc-directed polyclonal antibodies or clinically relevant concentrations of soluble CA125. ConclusionsThese findings indicate that IMV-M cytotoxic activity requires clustering on MUC16, that CA125 at clinically relevant concentrations does not mediate IMV-M neutralization, and that aggregate formation with secondary antibodies or soluble MUC16 does not induce off-target toxicity.

A critical aspect of a successful pandemic response is expedient antibody discovery, manufacturing and deployment of effective lifesaving treatments to patients around the world. However, typical drug discovery and development is a lengthy multi-step process that must align drug efficacy with multiple developability criteria and can take years to complete. In this context, artificial intelligence (AI), and especially machine learning (ML), have great potential to accelerate and improve the optimization of therapeutics, increasing their activity and safety as well as decreasing their development time and manufacturing costs. Here we present a novel, cost-effective and accelerated approach to therapeutic antibody discovery, that couples AI-designed human antibody libraries, biased for improved developability attributes with high throughput and sensitive screening technologies. The applicability of our platform for effective therapeutic antibody discovery is demonstrated here with the identification of a panel of human monoclonal antibodies that are novel, diverse and pharmacologically active. These first-generation antibodies, without the need for affinity maturation, bind to the SARS-CoV-2 spike protein with therapeutically-relevant specificity and affinity and display neutralization of SARS-CoV-2 viral infectivity across multiple strains. Altogether, this platform is well suited for rapid response to infectious threats, such as pandemic response. IMPORTANCEExpedient discovery and manufacturing of lifesaving therapeutics is critical for pandemic response. The recent COVID pandemic has highlighted the current inefficiencies and the need for improvements. To this end, we present our therapeutic antibody discovery platform that couples artificial intelligence (AI) and innovative high throughput technologies, and we demonstrate its applicability to rapid response. This platform enabled the isolation, characterization, and rapid identification of effective broadly neutralizing SARS-CoV-2 antibodies with good developability attributes, anticipated to fit our current process development and manufacturing platform. As such, this would benefit cost-of-goods and improve therapeutic access to patients. The AI-derived antibodies represent an advantageous therapeutic modality that can be developed and deployed fast, thus well suited for rapid response to infectious threats, such as pandemic response.

The emergence of coronavirus disease 2019 (COVID-19) pandemic led to an urgent need to develop therapeutic interventions. Among them, neutralizing antibodies play crucial roles for preventing viral infections and contribute to resolution of infection. Here, we describe the generation of antibody libraries from 17 different COVID-19 recovered patients and screening of neutralizing antibodies to SARS-CoV-2. After 3 rounds of panning, 456 positive phage clones were obtained with high affinity to RBD (receptor binding domain). Then the positive clones were sequenced and reconstituted into whole human IgG for epitope binning assays. After that, all 19 IgG were classified into 6 different epitope groups or Bins. Although all these antibodies were shown to have ability to bind RBD, the antibodies in Bin2 have more superiority to inhibit the interaction between spike protein and angiotensin converting enzyme 2 receptor (ACE2). Most importantly, the antibodies from Bin2 can also strongly bind with mutant RBDs (W463R, R408I, N354D, V367F and N354D/D364Y) derived from SARS-CoV-2 strain with increased infectivity, suggesting the great potential of these antibodies in preventing infection of SARS-CoV-2 and its mutations. Furthermore, these neutralizing antibodies strongly restrict the binding of RBD to hACE2 overexpressed 293T cells. Consistently, these antibodies effectively neutralized pseudovirus entry into hACE2 overexpressed 293T cells. In Vero-E6 cells, these antibodies can even block the entry of live SARS-CoV-2 into cells at only 12.5 nM. These results suggest that these neutralizing human antibodies from the patient-derived antibody libraries have the potential to become therapeutic agents against SARS-CoV-2 and its mutants in this global pandemic.