key: cord-0782704-th34d21s authors: Capon, Daniel J.; Troitskaya, Larisa; Chan, Nelson Lap Shun; Fomin, Marina; Edman, Ursula; Frank, Brendon; Jin, Jing; Martinelli, Rachel; Capon, Benjamin Z.; Ferguson, Ginger A.; Gefter, Malcolm L.; Simmons, Graham title: Topologically engineered antibodies and Fc-fusion proteins: a new class of multifunctional therapeutic candidates for SARS-CoV-2, cancer, and other disease date: 2022-03-23 journal: bioRxiv DOI: 10.1101/2022.03.23.485397 sha: c886361544a77087194238b0643d7ad35e38ed5a doc_id: 782704 cord_uid: th34d21s The ability of antibodies and Fc-fusion proteins to bind multiple targets cooperatively is limited by their topology. Here we describe our discovery that ACE2 Fc-fusion proteins spontaneously cross-dimerize, forming topologically distinct “superdimers” that demonstrate extraordinary SARS-CoV-2 intra-spike cooperative binding and potently neutralize Omicron B.1.1.529 at least 100-fold better than eight clinically authorized antibodies. We also exploited cross- dimerization to topologically engineer novel superdimeric antibodies and Fc-fusion proteins with antibody-like plasma half-lives to address cancer and infectious disease therapy. These include bispecific ACE2-antibody superdimers that potently neutralize all major SARS-CoV-2 variants, and bispecific anti-cancer and anti-viral antibody superdimers that are more potent than two-antibody cocktails. Superdimers are efficiently produced from single cells, providing a new therapeutic approach to many disease indications. illustrates the structures of the molecules we describe in these studies. Structures were confirmed by specific cleavage by IdeZ protease and multi-angle light scattering coupled with size-exclusion chromatography (SEC-MALS) (figs. S1-S3, table S1). Figs. 1A-1C illustrate three topologically distinct types of ACE2 dimers that are formed when dimerization is driven solely by the Fc domain (Fig. 1A) , solely by the ACE2 collectrin-like domain (Fig. 1B) , or by both the Fc and collectrin-like domains (Fig. 1C ). The first construct (ACE2-615 homodimer) and second construct (ACE2-740 heterodimer) were each produced as a single species. We discovered that the third construct was a mixture of two topologically distinct forms ( Fig. 2A) . The smaller form is the predicted ACE2-740 homodimer (Fig. 1C) . The larger form is an ACE2-740 superhomodimer (Fig. 1D) , featuring a novel topology reflecting cross-dimerization of its two pairs of CLD dimerizing polypeptides and two pairs of Fc dimerizing polypeptides. The structure of the superdimer suggests a tetrahedral-like configuration formed by the outward projection of its two ACE2 peptidase dimers (8) (9) (10) and two Fc dimers from a central "nucleus". Stoichiometric competition binding studies revealed extraordinary activity associated with the superdimer. When impure mixtures of ACE2-740 superhomodimer and ACE2-740 homodimer were titrated with limiting amounts of aggregate-free, individual spike trimers, we observed a preferred "order-of-binding" in which superdimer-spike complexes formed at the expense of dimer-spike complexes (Fig. 2D ). Similar results were obtained for equimolar mixtures of purified ACE2-740 superhomodimer with each of the three topologically distinct ACE2 dimers shown in Figs. 1A-1C ( fig. S4 ). Figs. 1E-1H illustrate "superheterodimers" that can be produced as a single topological form rather than mixtures by limiting the constructs to a single pair of CLD-dimerizing polypeptides. We refer to these constructs as GEM-DIMERs™ (Gemini dimers) given their twin-like structures. Superheterodimers may comprise any combination of antibodies and Fc-fusion proteins. Figs. 2B-2C show that ACE2-740/615 superheterodimer (Fig. 1E) , with four ACE2 peptidase domains, and ACE2-740/B13A superheterodimer (Fig. 1F) , with two ACE2 peptidase domains and two Fab domains from the anti-spike B13A antibody (11) , are each produced as a single species. ACE2-740/615 superheterodimer and ACE2-740/B13A superheterodimer demonstrate highly potent binding to individual spike trimers, comparable to ACE2-740 superhomodimer (Figs. 2E-2F). The qualitative results from our competition binding experiments were confirmed by quantitative solution binding analysis (table S2). The three topologically distinct ACE2 dimers, ACE2-615 homodimer, ACE2-740 heterodimer, and ACE2-740 homodimer, bound spike protein with a KD of 1.56 nM, 1.92 nM, and 1.39 nM, respectively. ACE2-740 superhomodimer, ACE2-740/615 superheterodimer, and ACE2-740/B13A superheterodimer bound spike protein with a KD of 27.2 pM, 26.7 pM and 15.8 pM, respectively. These results demonstrate a cooperative binding advantage of ACE2 superdimers over ACE2 dimers of 51-to 126-fold. To assess our ACE2 superdimers as candidates for treatment of COVID-19 and other ACE2 utilizing viruses, we compared their ability to neutralize live SARS-CoV-2 betacoronavirus and NL63 alphacoronavirus (Figs. 2G-2I, fig. S5 , table S3). The results recapitulate the relative potencies of superdimers and dimers in our solution binding studies (table S2) . ACE2-740 superhomodimer and ACE2-740/615 superheterodimer demonstrate 403-fold and 143-fold greater potency than ACE2-615 homodimer against SARS-CoV-2, and 87-fold and >78-fold greater potency against NL63. ACE2-740/B13A superheterodimer demonstrates 646-fold and 29-fold greater potency against SARS-CoV-2 compared with ACE2-740 heterodimer and B13A antibody, respectively. We also compared the ability of ACE2 superdimers and dimers to neutralize spike protein binding to cell surface ACE2 receptors ( Fig. 2J-2K, table S4 ). ACE2-740 superhomodimer and ACE2-740/615 superheterodimer neutralized spike binding with a cooperative binding advantage over ACE2-740 homodimer of 8.9-fold and 10.6-fold. ACE2-740/B13A superheterodimer neutralized spike binding with a cooperative binding advantage over ACE2-740 homodimer and B13A antibody of 11.2-fold and 3.8-fold. We further evaluated ACE2-740/B13A superheterodimer using an in vivo prophylactic model of SARS-CoV-2 infection in golden Syrian hamsters (12) . Fig. 2L demonstrates that ACE2-740/B13A superheterodimer (25 mg/kg) exhibits comparable efficacy in preventing weight-loss compared to a two-antibody cocktail of REGN10987 (25 mg/kg) and REGN10933 (25 mg/kg), and superior efficacy to REGN10933 alone (25 mg/kg). To determine the utility of superdimerization to increase the potency of therapeutic antibodies, we created ACE2-740/Fab superheterodimers (Fig. 1F) , Fab/Fab superheterodimers (Fig. 1G) , and bispecific Fab1/Fab2 superheterodimers (Fig. 1H ) using eight clinically authorized antibodies (13) (14) (15) (16) (17) (18) . In stoichiometric competition binding experiments with individual spike trimers, we observed a preferred order-of-binding in which antibody superheterodimer-spike complexes generally formed at the expense of parent antibody-spike complexes (figs. S6-S7). Pseudovirus experiments similarly demonstrate a dramatic increase in the neutralization potency of antibody superheterodimers compared to their parent antibodies ( fig. S8 ). In particular, RG2-RG2 and CT1-CT1 neutralized the B.1.531 variant approximately 100-fold better than the REGN10933 and CT-P59 parent antibodies (table S5) . It is noteworthy that ACE2 dimers, whether formed by CLD dimerization, Fc dimerization, or both, appear to bind with relatively low avidity to individual spike trimers, in spite of the trivalency of the spike protein, suggesting that the configuration of the three receptor binding domains (RBDs) is incompatible with bivalent binding by ACE2 dimers. The fact that ACE2 superdimers as well as antibody superdimers bind more strongly to individual spike trimers suggests that the topology of the superdimer allows it to assume a configuration that is capable of an intra-spike interaction in which at least two binding sites on the spike trimer are simultaneously engaged. The molar masses of superdimer-spike binding complexes formed in the presence of excess superdimer (table S6) provide evidence for two distinct types of intra-spike interaction. The first type, exemplified by ACE2-740/615 superheterodimer and most of the Fab/Fab superheterodimers, is characterized by 1:1 superdimer-spike trimer complex, indicating an intra-spike, inter-subunit interaction in which a single superdimer simultaneously engages a single type of epitope on at least two of the three S protein subunits constituting the spike trimer. The second type, exemplified by ACE2-740/B13A superheterodimer and bispecific Fab1/Fab2 superheterodimers, is characterized by a 2:1 or greater complex, indicating an intra-spike interaction in which each superdimer engages two distinct epitopes in an inter-subunit or intra-subunit fashion, but which in either event allows two or more superdimers to access an individual spike trimer simultaneously. We suggest that the superdimer configuration may provide a general solution to the inability of antibodies and Fc-fusion proteins to engage multiple targets that are refractory to simultaneous binding, whether of a monospecific or multispecific nature. We next confirmed that antibody superheterodimers have antibody-like plasma half-lives using the human FcRn Tg32 homozygous transgenic mouse model. This model is used to predict the clearance (CL) of human antibodies in humans with greater accuracy than non-human primate studies (19). Each superheterodimer was administered as a single i.v. bolus (10 mg/kg) and ACE2 plasma levels were determined by spike binding activity ( fig. S9 -S10, table S7). Two ACE2-740/B13A superheterodimers, one possessing and one lacking angiotensin-converting enzyme activity (fig. S11, table S8), demonstrate comparable terminal half-lives (T1/2) of 10.6 and 11.2 days, respectively, representing a 33.5-to 35.4-fold improvement compared with the T1/2 of 7.6 hours observed for soluble recombinant human ACE2 (APN01) in humans (single i.v. bolus at 1.2 mg/kg) (20). Two Fab/Fab superheterodimers (Fig. 1G ), RG2-RG2 and VR1-VR1, demonstrate a T1/2 of 17.6 and 16.3 days, respectively, which compared favorably with the T1/2 of 12.9 and 14.7 days observed for the parent REGN10933 and VIR-7831 antibodies. The Omicron B.1.1.529 variant, which eliminates or substantially reduces the effectiveness of most clinically authorized antibodies, highlights a potential therapeutic advantage for ACE2 superdimers in addressing SARS-CoV-2 resistance. fig. S12A -S12K, table S9). ACE2-740/B13A superheterodimer generally demonstrates equal or better neutralization activity than the same eight antibodies against all twelve variants. In keeping with our prediction of the particular therapeutic value of ACE2 molecules without affinity-enhancing mutations, SARS-CoV-2 has not, to date, successfully evaded neutralization by our ACE2 superdimers. We suggest that Omicron B.1.1.529, known for its high transmissibility, have been selected in a manner resulting in its particular susceptibility to ACE2-740/615 superheterodimer with its four ACE2 domains. We created bispecific Fab1/Fab2 superheterodimers (Fig. 1H ) to determine whether they can neutralize certain resistant variants more effectively than two-antibody cocktails containing the parent antibodies. A bispecific Fab1/Fab2 superheterodimer (HB1701) consisting of the REGN10987 and REGN10933 Fabs neutralized the N439K/B.1.351 variant 25-fold better than the REGN10987/REGN10933 two-antibody cocktail (Figs. 4A-4C, fig. S13 , table S10). The neutralizing activity of HB1701 was comparable to the highly potent ACE2-740/615 superheterodimer and ACE2-740/B13A superheterodimer, demonstrating that bispecific antibody superheterodimers can potently neutralize SARS-CoV-2 variants that are resistant to cocktails of their parent antibodies. We next examined whether bispecific antibody superheterodimers can effectively target cell surface proteins, in particular targets for cancer therapeutics. The clinical application of anti-CD20 and anti-CD19 antibodies is associated with treatment-resistant and/or recurrent disease (21, 22) suggesting the potential therapeutic value of a bispecific superheterodimer that targets CD20 and CD19 simultaneously. We created bispecific Fab1/Fab2 superheterodimers using Rituxan, an anti-CD20 antibody widely used as a frontline treatment in B cell lymphoma (23), and FMC63, an anti-CD19 antibody used as a component of chimeric antigen receptors for CAR-T cell therapy of relapsed/refractory B-cell lymphoma (24). We evaluated the ability of two anti-CD20/anti-CD19 bispecific superheterodimers (HB1905, HB1906) to bind cell surface CD20 and CD19 by pre-incubating each with the CD20/CD19-positive human Toledo cell line followed by addition of fluorescently labeled Rituxan and FMC63. Figs. 4D-4I demonstrate that HB1905 and HB1906 completely extinguish binding of both labeled antibodies, whereas unlabeled Rituxan and unlabeled FMC63 controls each only compete against its labeled counterpart. Surface plasmon resonance binding studies demonstrate that HB1905 and HB1906 bind monovalent CD20 and CD19, as well as Fc gamma receptors and complement component C1q, with an affinity comparable to Rituxan and FMC63 (fig. S14). We evaluated the functional bispecificity of HB1905 and HB1906 by antibody-dependent cell-mediated cytotoxicity (ADCC) of Toledo cells. As shown in Figs. 4J-4K, the ADCC activity observed for HB1905 and HB1906 was 39-to 44-fold better than a two-antibody cocktail containing Rituxan and FMC63 (table S11). These results indicate that the bispecific superheterodimer configuration may be generally valuable in anti-cancer therapeutics to combat resistance. In summary, we have used topological engineering to enhance the cooperative binding of spike protein by ACE2 in the form of ACE2 superdimers that potently neutralize all major SARS-CoV-2 variants. We have also shown that anti-cancer and anti-viral antibodies in the form of bispecific superdimers are more potent than two-antibody cocktails. We anticipate that superdimerization may be applied to any combination of antibodies and fusion proteins or their derivatives. Having the ability that we describe here to expand the universe of allowable binding targets of antibodies and Fc-fusion proteins should enhance their utility as effective therapeutics. We believe these findings have considerable implications for basic research and clinical investigation. Nine or more concentrations were tested for each protein. Pseudovirus in culture media without protein was used as a negative control to determine 100% infectivity. The mixtures were then incubated with 293T-hsACE2 cells at 2.5x10 5 cells/mL in the 96-well plates. Infection took place over approximately 72 hrs at 37 o C with 5% CO2. The luciferase signal was measured using the Renilla-Glo Luciferase Assay System, Cat#E2710 (Promega, Madison, WI) with the luminometer set at 1 ms integration time. The relative luminescence signals (RLU) obtained from the negative control wells were normalized and used to calculate the neutralization percentage for each concentration. All samples were run in duplicate. The data were processed by GraphPad Prism v9.3.1 to fit a 4-parameter logistic curve and calculate the log IC50. The study was conducted at BIOQUAL, Inc. (Rockville, MD). A total of 24 male golden hamsters 6-8 weeks old were assigned to four groups (n=6). Animals were treated with the appropriate material for their group on study day minus one (-1) via intraperitoneal (IP) injection. Noncompartmental pharmacokinetic data analysis was performed using Phoenix WinNonlin v8.3 (Certara, Princeton, NJ). Parameters were estimated for individual mice using a module for IV bolus administration. The area under the plasma concentration versus time curve was estimated using the linear trapezoid method. Terminal slopes for all animals were estimated over the time period from approximately 5 days through the end of sampling at approximately 28 days and used to calculate half-life and AUC extrapolated to infinity. The carboxypeptidase reaction was initiated when 0.025 μg (wild-type) or 15 μg (ACE2 mutant) of protein was added to 0. Toledo cells were used as target cells and maintained in RPMI with GlutaMax, Cat#61870036, Thermo Fisher Scientific, Waltham, MA) and 10% fetal bovine serum, Cat#SH30070.03IH30-45, (Cytiva). To differentiate target cells from natural killer (NK) cells, the Toledo cells were labeled with 5 µM CellTrace Yellow, Cat#C34567 (Thermo Fisher) for 20 min at 37℃. After quenching and washing according to the manufacturer's recommendation, 1 x 10 3 cells, in duplicate, were incubated with CD20-and/or CD19-binding proteins and 2 µM CellEvent Caspase-3/7 Green Detection Reagent, Cat#C10423 (Thermo Fisher) for 10 min at RT in 96-well ultra-low attachment 96-well microplates, Cat#7007 (Corning, Corning, NY). Human peripheral blood NK cells, Cat#70036, (Stemcell Technologies, Vancouver, Canada) were added at an effector:target ratio of 10:1. Caspase-3/7 cleavage was measured every hr for 8 hrs using an IncuCyte SX5 (Sartorius, Göttingen, Germany) at 37℃ with 5% CO2. Caspase-3/7 positive target cells were discriminated from unlabeled apoptotic NK cells using multi-channel fluorescence. Kinetic data were obtained using IncuCyte 2021A software (Sartorius) before area-under-the-curve (AUC) (mean ± SEM) and non-linear regression analysis were performed using GraphPad Prism v9.3.1. Recombinant CD20 (167 nM-0.68 nM), Cat#CD0-H52H3 (Acrobiosystems) was diluted in a threefold dilution step with HEPES buffer containing 0.05% DDM and 0.01% CHS. Proteins were injected onto the CMD-P chip for 5 min during the association phase followed by HEPES buffer injection for another 5 min during the dissociation phase. The CMD-P chip was regenerated by flowing Pierce IgG Elution Buffer, Cat#21004 (Thermo Fisher) with 1 M NaCl, 3 times with 1 min intervals after each round of analyte injection. Data were analyzed using the Carterra Kinetics Software (Carterra). fig. S1 . Strategy for the confirmation of the structures of the topologically distinct molecules described in these studies using specific cleavage with IdeZ and TCEP reduction. Molecules were digested with IdeZ protease followed by incubation with protein A beads to remove Fc fragments and incompletely digested molecules. The untreated, IdeZ-treated, and IdeZ-treated/TCEPreduced molecules were analyzed by SEC/MALS ( fig. S2, table S1 ). Predicted fragments are depicted for (A) ACE2-615 homodimer (Fig. 1A) , (B) ACE2-740 heterodimer (Fig. 1B) and its Fabsubstituted counterpart, (C) ACE2-740 homodimer (Fig. 1C) , (D) ACE2-740 superhomodimer (Fig. 1D) , (E) ACE2-740/615 superheterodimer (Fig. 1E) , (F) ACE2-740/B13A superheterodimer (Fig. 1F) , (G) Fab/Fab superheterodimer (Fig. 1G) , and (H) Fab1/Fab2 superheterodimer (Fig. 1H) . . SE-HPLC analysis of the topologically distinct molecules described in these studies following IdeZ cleavage and TCEP reduction. SE-HPLC elution profiles of untreated, IdeZ-treated, and IdeZ-treated/TCEP-reduced molecules are shown by black, red, and blue curves, respectively. Arrow 1 points to the intact molecule, arrow 2 points to the IdeZ cleavage product, and arrows 3 and 4 point to the TCEP-reduced IdeZ cleavage product(s). As expected, IdeZ cleavage products that are dimerized solely by the Fc domain were dissociated upon TCEP reduction of the hinge region interchain disulfides, while IdeZ cleavage products dimerized by the collectrin-like domain did not dissociate upon TCEP reduction. To further confirm each structure, the molar mass of each of the observed peaks was determined by SEC-MALS as described in table S1. The following molecules were analyzed: (A) ACE2-615 homodimer (Fig. 1A) , (B) Fab heterodimer ( fig. S1B ), (C) ACE2-740 homodimer (Fig. 1C) , (D) ACE2-740 superhomodimer (Fig. 1D) , (E) ACE2-740/615 superheterodimer (Fig. 1E) , (F) ACE2-740/B13A superheterodimer (Fig. 1F) , (G) Fab/Fab superheterodimer (Fig. 1G) , and (H) Fab1/Fab2 superheterodimer (Fig. 1H) . Purified preparations of ACE2-740 superhomodimer (Fig. 1D ) and ACE2-740 homodimer (Fig. 1C) were analyzed individually by SEC-MALS. The profiles are overlaid for comparison. Molar masses are summarized in table S1. Abbreviations: LS, light scattering; dRI, differential refractive index; MM, molar mass. fig. S4 . Order-of-binding analysis demonstrates that ACE2 superdimer-spike complexes form at the expense of ACE2 dimer-spike complexes. Titrations with aggregate-free spike trimer were carried out using mixtures of purified ACE2-740 superhomodimer (Fig. 1D ) with purified preparations of the following ACE2 dimers: (A) ACE2-615 homodimer (Fig. 1A) , (B) ACE2-740 heterodimer (Fig. 1B) , and (C) ACE2-740 homodimer (Fig. 1C) . table S6. Stoichiometry of spike trimer binding the topologically distinct molecules described in these studies. SEC-MALS was used to determine the molar mass of each spike-binding protein and the molar mass of each binding complex that formed when an excess of each spike-binding protein was incubated with aggregate-free, individual spike trimers (Mr=587.7kDa). *Each species was a well-formed symmetrical peak. Abbreviations: SBP, spike-binding protein The following equation was used to determine the molar ratio of spike binding protein to individual spike trimer for each complex, Designing CD4 immunoadhesins for AIDS therapy Engineering human ACE2 to optimize binding to the spike protein of SARS coronavirus 2 Engineered ACE2 receptor traps potently neutralize SARS-CoV-2 SARS-CoV-2 and Three Related Coronaviruses Utilize Multiple ACE2 Orthologs and Are Potently Blocked by an Improved ACE2-Ig De novo design of potent and resilient hACE2 decoys to neutralize SARS-CoV-2 3D complex: a structural classification of protein complexes Principles of assembly reveal a periodic table of protein complexes Recombinant Expression and Characterization of Human and Murine ACE2: Species-Specific Activation of the Alternative Renin-Angiotensin-System inventors; Apeiron Biologics AG, assignee. Recombinant ACE2 Polypeptide Dimer Structural basis for the recognition of SARS-CoV-2 by full-length human ACE2 Mouse antibodies with activity against the SARS-CoV-2 D614G and B.1.351 variants Simulation of the Clinical and Pathological Manifestations of Coronavirus Disease 2019 (COVID-19) in a Golden Syrian Hamster Model: Implications for Disease Pathogenesis and Transmissibility Antibody cocktail to SARS-CoV-2 spike protein prevents rapid mutational escape seen with individual antibodies Surface plasmon resonance (SPR) binding analysis of anti-CD20/anti-CD19 bispecific Fab1/Fab2 superheterodimers to CD20, CD19, Fc gamma receptors, and complement C1q Pseudovirus neutralizing activity of ACE2-740/Fab superheterodimers and antibody Fab/Fab superheterodimers compared with their parent antibodies. Neutralization activity (IC50) against Wuhan D614G and B.1.351 variants by ACE2-740/Fab superheterodimers, Fab/Fab superheterodimers, and their parent antibodies CT1-CT1. Abbreviations: ACE2 WT, ACE2 ACE2 with the H378A mutation which abrogates angiotensin-converting activity We thank Samuel Broder for invaluable discussions and critical reading of the manuscript; Carin HB1553) and ACE2-740/B13A superheterodimers (HB1515, HB1516) compared with eight clinically authorized antibodies. Neutralization activity (IC50) of HB1507 and HB1553 (Fig. 1F) , and HB1515 and HB1516 (Fig. 1G) table S11. ADCC activity of anti-CD20/anti-CD19 bispecific Fab1/Fab2 superheterodimers (HB1905, HB1906) compared with their parent antibodies. ADCC activity (EC50) against CD20/CD19-positive Toledo cells is shown for HB1905 and HB1906 and their parent antibodies Rituxan (anti-CD20) and FMC63 (anti-CD19) used as single agents or as a two-antibody cocktail.