Preclinical Efficacy of IMM-BCP-01, a Highly Active Patient-Derived Anti-SARS-CoV-2 Antibody Cocktail (preprint)

Using an unbiased interrogation of the anti-viral memory B cell repertoire of convalescent COVID-19 patients, we identified three human antibodies that when combined demonstrated both robust viral suppressive properties against all tested SARS-CoV-2 variants of concern in vitro and profound anti-viral efficacy in vivo. In this report, we describe the pre-clinical characterization of an antibody cocktail, IMM-BCP-01, that consists of three unique, patient-derived recombinant antibodies directed at non-overlapping surfaces on the Spike protein, each with particularly effective antiviral activity. One antibody has a composite epitope blocking ACE2 binding, one antibody bridges two Spike proteins, and one antibody neutralizes virus by binding to a conserved epitope outside of ACE2 binding site. These antibodies, when administered after viral infection, potently decreased viral load in lungs of infected Syrian golden hamsters in a dose-dependent manner, elicited broad anti-viral neutralizing activity against multiple SARS-CoV-2 variants, and induced a robust anti-viral effector function response, including phagocytosis, and activation of classical complement pathway. Our pre-clinical data demonstrate that the unique three antibody cocktail IMM-BCP-01 is a potent and dose-efficient approach to treat early viral infection and prevent SARS-CoV-2 in susceptible individuals.

Coronavirus associated molecular mimicry common to SARS-CoV-2 peptide (preprint)

Relationship of COVID-19 and immunity is complex and can involve autoimmune reactions through molecular mimicry. We investigated autoimmunity related pathological mechanisms involving molecular mimicry that are common to certain coronaviruses, including SARS-CoV-2, by means of a selected peptide sequence (CFLGYFCTCYFGLFC). Accordingly, coronavirus-associated sequences that are homologous to that 15mer sequence in the SARS-CoV-2 proteome are attained first. Then, homologous human and coronavirus sequences are obtained, wherein the coronavirus sequences are homologous to the 15mer SARS-CoV-2 peptide. All the identified query-subject sequences contained at least 7 residue matches in the aligned regions. Finally, parts of those coronavirus and host sequences, which are predicted to have high affinity to the same human leukocyte antigen (HLA) alleles as that of the SARS-CoV-2 sequence, are selected among the query and subject epitope-pairs that were both (predicted to be) strongly binding to the same HLA alleles. The proteins or the protein regions with those predicted epitopes include, but not limited to, immunoglobulin heavy chain junction regions, phospholipid phosphatase-related protein type 2, slit homolog 2 protein, and CRB1 isoform I precursor. These proteins are potentially associated with certain pathologies, but especially the possible CRB1 related coronavirus pathogenicity could be furthered by autoimmunity risk in HLA*A24:02 serotypes. Overall, results imply autoimmunity risk in COVID-19 patients with HLA*A02:01 and HLA*A24:02 serotypes in general, through molecular mimicry. This is also common to other coronaviruses than SARS-CoV-2. These results are indicative at the current stage, they need to be validated. Yet, they can pave the way to autoimmunity treatment options to be used in COVID-19 and its associated diseases.

Modeling mutational effects on biochemical phenotypes using convolutional neural networks: application to SARS-CoV-2 (preprint)

Biochemical phenotypes are major indexes for protein structure and function characterization. They are determined, at least in part, by the intrinsic physicochemical properties of amino acids and may be reflected in the protein three-dimensional structure. Modeling mutational effects on biochemical phenotypes is a critical step for understanding protein function and disease mechanism as well as enabling drug discovery. Deep Mutational Scanning (DMS) experiments have been performed on SARS-CoV-2's spike receptor binding domain and the human ACE2 zinc-binding peptidase domain - both central players in viral infection and evolution and antibody evasion - quantifying how mutations impact binding affinity and protein expression. Here, we modeled biochemical phenotypes from massively parallel assays, using convolutional neural networks trained on protein sequence mutations in the virus and human host. We found that neural networks are significantly predictive of binding affinity, protein expression, and antibody escape, learning complex interactions and higher-order features that are difficult to capture with conventional methods from structural biology. Integrating the intrinsic physicochemical properties of amino acids, including hydrophobicity, solvent-accessible surface area, and long-range non-bonded energy per atom, significantly improved prediction (empirical p<0.01) though there was such a strong dependence on the sequence data alone to yield reasonably good prediction. We observed concordance of the DMS data and our neural network predictions with an independent study on intermolecular interactions from molecular dynamics (multiple 500 ns or 1 s all-atom) simulations of the spike protein-ACE2 interface, with critical implications for the use of deep learning to dissect molecular mechanisms. The mutation- or genetically- determined component of a biochemical phenotype estimated from the neural networks has improved causal inference properties relative to the original phenotype and can facilitate crucial insights into disease pathophysiology and therapeutic design.

Early therapy with remdesivir and antibody combinations improves COVID-19 disease in mice (preprint)

Improving the standard of clinical care for coronavirus disease 2019 (COVID-19) is a global health priority. Small molecule antivirals like remdesivir (RDV) and biologics such as human monoclonal antibodies (mAb) have demonstrated therapeutic efficacy against SARS-CoV-2, the causative agent of COVID-19. However, the efficacy of single agent therapies has not been comprehensively defined over the time course of infection and it is not known if combination RDV/mAb will improve outcomes over single agent therapies. In kinetic studies in a mouse-adapted SARS-CoV-2 pathogenesis model, we show that single-agent therapies exert potent antiviral effects even when initiated relatively late after infection, but their efficacy is diminished as a function of time. RDV and a cocktail of two mAbs in combination provided improved outcomes compared to single agents alone extending the therapeutic window of intervention with less weight loss, decreased virus lung titers, reduced acute lung injury, and improved pulmonary function. Overall, we demonstrate that direct-acting antivirals combined with potent mAb can improve outcomes over single agents alone in animal models of COVID-19 thus providing a rationale for the coupling of therapies with disparate modalities to extend the therapeutic window of treatment.

Unbiased interrogation of memory B cells from convalescent COVID-19 patients reveals a broad antiviral humoral response targeting SARS-CoV-2 antigens beyond the spike protein (preprint)

Patients who recover from SARS-CoV-2 infections produce antibodies and antigen-specific T cells against multiple viral proteins. Here, an unbiased interrogation of the anti-viral memory B cell repertoire of convalescent patients has been performed by generating large, stable hybridoma libraries and screening thousands of monoclonal antibodies to identify specific, high-affinity immunoglobulins (Igs) directed at distinct viral components. As expected, a significant number of antibodies were directed at the Spike (S) protein, a majority of which recognized the full-length protein. These full-length Spike specific antibodies included a group of somatically hypermutated IgMs. Further, all but one of the six COVID-19 convalescent patients produced class-switched antibodies to a soluble form of the receptor-binding domain (RBD) of S protein. Functional properties of anti-Spike antibodies were confirmed in a pseudovirus neutralization assay. Importantly, more than half of all of the antibodies generated were directed at non-S viral proteins, including structural nucleocapsid (N) and membrane (M) proteins, as well as auxiliary open reading frame-encoded (ORF) proteins. The antibodies were generally characterized as having variable levels of somatic hypermutations (SHM) in all Ig classes and sub-types, and a diversity of VL and VH gene usage. These findings demonstrated that an unbiased, function-based approach towards interrogating the COVID-19 patient memory B cell response may have distinct advantages relative to genomics-based approaches when identifying highly effective anti-viral antibodies directed at SARS-CoV-2.