High-resolution mapping of the neutralizing and binding specificities of polyclonal sera post-HIV Env trimer vaccination

Mapping polyclonal serum responses is critical to rational vaccine design. However, most high-resolution mapping approaches involve isolating and characterizing individual antibodies, which incompletely defines the polyclonal response. Here we use two complementary approaches to directly map the specificities of the neutralizing and binding antibodies of polyclonal anti-HIV-1 sera from rabbits immunized with BG505 Env SOSIP trimers. We used mutational antigenic profiling to determine how all mutations in Env affected viral neutralization and electron microscopy polyclonal epitope mapping (EMPEM) to directly visualize serum Fabs bound to Env trimers. The dominant neutralizing specificities were generally only a subset of the more diverse binding specificities. Additional differences between binding and neutralization reflected antigenicity differences between virus and soluble Env trimer. Furthermore, we refined residue-level epitope specificity directly from sera, revealing subtle differences across sera. Together, mutational antigenic profiling and EMPEM yield a holistic view of the binding and neutralizing specificity of polyclonal sera.

Abstract 1554: Uncoupling tumor-cell cytotoxicity from cytokine release with novel T-cell engaging bispecific antibodies

T-cell recruiting bispecific antibody (T-BsAb) treatment is an emerging cancer immunotherapy notable for its highly-specific tumor killing activity in humans. However, the success of T-BsAbs in the clinic has been limited due to cytokine-release related toxicities. In order for this technology to reach its potential, the next generation of T-BsAbs must avoid this unintended toxicity by uncoupling cytokine-release from tumor-cell killing. To this end, we used a sequence-based discovery method on CD3-immunized OmniFlic rats, which express human fixed light-chain antibodies, to identify several novel anti-CD3 antibodies. These antibodies bind to multiple CD3 epitopes, each with varying affinities and degrees of T-cell activation. Furthermore, when placed in the context of a T-BsAb, these novel anti-CD3 antibodies achieve similar levels of T-cell mediated tumor-specific lysis, but with a thousand-fold range in potency. Among these antibodies, we identified one lead with particularly favorable properties, in that it stimulates very low levels of cytokine production without sacrificing its potent antigen-specific tumor lysis in vitro, in mouse xenograft assays, as well as ex vivo patient sample studies. This unique behavior gives our novel anti-CD3 antibody the potential to widen the therapeutic window when treating patients and ultimately usher in the next generation of T-BsAbs.

Citation Format: Kyle Lorentsen, Preethi Sankaran, Nathan D. Trinklein, Duy Pham, Ute Schellenberger, Ben Buelow, Andrew Boudreau, Priya Choudhry, Starlynn C. Clarke, Kevin Dang, Katherine E. Harris, Suhasini Iyer, Brett Jorgensen, Payal Pratap, Udaya S. Rangaswamy, Harshad S. Ugamraj, Omid Vafa, Arun P. Wiita, Wim van Schooten, Roland Buelow, Shelley Force Aldred. Uncoupling tumor-cell cytotoxicity from cytokine release with novel T-cell engaging bispecific antibodies [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1554.

Multispecific antibodies targeting CD38 show potent tumor-specific cytotoxicity

57

Background: Multivalent antibodies targeting either CD38 alone or CD38 in conjunction with PD-L1 may yield therapeutics with superior biological activities and provide benefit for treating malignancies expressing low levels of CD38 (MCL, NHL, T cell lymphomas and Daratumumab refractory MM). Multivalent, multispecific antibodies kill CD38low cells through a variety of mechanisms including stronger and more specific engagement of CD38. Potent and directed immune checkpoint inhibition is realized by adding an anti-PD-L1 binding domain. Teneobio’s discovery platform utilizes VH domains (UniDabs) of fully human heavy chain antibodies (UniAbs) to develop bi-, tri-, and tetravalent antibodies. Methods: Individual UniDabs targeting CD38 and PDL1 were identified using our unique sequence-based discovery platform and high-throughput lead evaluation pipeline (TeneoSeek). This robust screening workflow enables evaluation of a large diversity of natural fully human antibodies, targeting multiple epitopes on a single antigen and uncovering important sequence activity relationships. UniDabs from transgenic rats are ideal building blocks for the generation of potent and highly manufacturable multivalent antibody therapeutics. Results: We have identified UniDabs that efficiently block PD-1/PD-L1 interaction as well as additional UniDabs that bind to five different functional epitopes on human CD38. Using different combinations and arrangements of UniDabs, a variety of multivalent antibodies were constructed and evaluated in in vitro models. Specific combinations of UniDabs show more potent cytotoxic effects than Daratumumab for multiple mechanisms including CDC and direct apoptosis. Conclusions: Data from a range of assay types show that multivalent UniAbs targeting CD38 can be engineered to display superior tumor cell cytotoxicity through multiple mechanisms of action.

A novel T-cell bispecific antibody platform for efficient T-cell mediated killing of tumor cells with minimal cytokine release

209

Background: Bispecific antibodies that recruit cytotoxic T cells to kill tumor cells are popular due to their targeted mechanism of action. Despite their attractiveness, there are limitations in the clinic due to undesirable toxicities associated with cytokine release. We describe here a platform for generation of a large collection of human anti-CD3 antibodies obtained from custom transgenic rats. Combining these unique anti-CD3 arms with different tumor targeting arms enables creation of bispecific antibodies of varying tumor cell killing capability. These anti-CD3 arms were combined with tumor antigen binding arms, namely, an anti-BCMA arm for multiple myeloma therapy, or an anti-CD22 arm for B-cell acute lymphoblastic leukemia therapy. Methods: Our platform utilizes a discovery approach involving antibody repertoire deep sequencing, high-throughput gene assembly, and recombinant expression, generating a highly diverse panel of antibodies with varying affinities. The CD3 antibodies were tested in in vitro T cell assays using human PBMCs to measure activation and cytokine release. Bispecific antibodies were evaluated for their ability to kill target cell lines upon co-culture with primary human PBMCs. The in vivo efficacy of bispecific antibodies was evaluated in a xenograft mouse model. Results: The in vitro T cell activity of these antibodies as measured by interleukin-2, interferon gamma levels and upregulation of the activation marker CD69 covered a broad spectrum of EC50 values. In co-culture systems with human PBMCs, anti-BCMA or anti-CD22 bispecific antibodies potently killed their respective target expressing cells with varying strengths. Additionally, the cytokine release from T cell activation correlated with the affinity of the anti-CD3 arms. The in vivo efficacy of the bispecific antibodies in a xenograft model with human PBMCs transferred into NSG mice showed striking tumor clearance at a wide range of doses. Conclusions: Our platform is highly suitable for creation of an extensive collection of bispecific antibodies for a variety of disease models by selecting the ideal anti-CD3 arm for each unique tumor antigen binding arm.

Development of a fully human t-cell engaging bispecific antibody for the treatment of multiple myeloma

60

Background: T-cell engaging bispecific antibody (T-BsAb) therapies are highly efficacious and well suited for targets with low expression on tumor cells. Recently, T-BsAbs with high activation of CD3 have been shown to overstimulate T cells, leading to toxicity and decreased efficacy. Teneobio has developed a fully human BCMA-specific T-BsAb using a low-activating αCD3 that is highly effective in vitro and in vivo against MM but stimulates minimal cytokine release. Methods: UniRats were immunized with either CD3 or BCMA antigens and antigen-specific UniAbs were identified by Ab repertoire sequencing and high-throughput gene assembly, expression, and screening. Antigen-specific VH sequences with the desired target affinity were selected using recombinant proteins and cells. In vitro efficacy studies included T-cell activation by cytokine- and tumor cell kill by calcein-release assays. In vivo efficacy of the molecules was evaluated in NSG mice harboring myeloma cells and human PBMCs. Results: BCMA-specific UniAbs bound plasma cells with sub-nM affinity. Strong and weak T cell agonists were identified that bound human T cells with high and low affinities respectively. T-BsAbs with a strong and a weak αCD3 demonstrated T-cell activation and tumor-cell cytotoxicity in vitro; T-BsAbs with a weak αCD3 showed markedly reduced cytokine production even at doses that showed maximum tumor cell lysis. In vivo, BCMAxCD3 T-BsAbs reduced tumor load and increased survival when co-administered with human PBMCs as compared to controls. Conclusions: Our results suggest that T-BsAbs with low-activating αCD3 arms may have a favorable toxicity profile while maintaining efficacy in the treatment of MM.

Development of a fully human T cell engaging bispecific antibody for the treatment of multiple myeloma

8017

Background: Although BCMA is a plasma cell specific surface molecule attractive as an antibody target in multiple myeloma, its scarcity on the cell surface may limit the efficacy of a conventional antibody. T-cell engaging bispecific antibody approaches are highly efficacious and are particularly well suited for a membrane target with limited expression, such as BCMA. Teneobio has developed a multivalent antibody platform based on modular human VH domains, which allowed us to build T cell engaging bispecific antibodies with low and high T cell agonistic activities. Methods: UniRats were immunized with either CD3 or BCMA antigens and antigen-specific UniAbs were identified by antibody repertoire sequencing and high-throughput gene assembly, expression, and screening. High affinity binding VH sequences were selected using recombinant proteins and cells. In vitro efficacy studies included T-cell activation by cytokine- and tumor cell kill by calcein-release assays. In vivo efficacy of the molecules was evaluated in NSG mice harboring myeloma cells and human PBMCs. Results: BCMA-specific UniAbs bound plasma cells with high affinities (100-700pM) and cross-reacted with cynomolgus plasma cells. Strong and weak T cell agonists were identified that bound human T cells with high and low affinities respectively and cross-reacted with cynomolgus T cells. T cell engaging bispecifics with a strong (H929 cytotoxicity:EC50=27pM) and a weak T cell activating arm (H929 cytotoxicity: EC50=1170pM) demonstrated T-cell activation and tumor-cell cytotoxicity in vitro; bispecifics with a weak CD3 engaging arm showed markedly reduced cytokine production even at doses saturating for cytotoxicity. In viv o, BCMAxCD3 bispecific antibodies reduced tumor load and increased survival when co-administered with human PBMCs as compared to controls. Conclusions: Our results suggest that T cell engaging bispecifics with low-affinity anti-CD3 arms could be preferred for the treatment of Multiple Myeloma.

A study of the physics and chemistry of TMC-1

We present a comprehensive study of the physical and chemical conditions along the TMC-1 ridge. Temperatures were estimated from observations of CH3CCH, NH3, and CO. Densities were obtained from a multitransition study of HC3N. The values of the density and temperature allow column densities for 13 molecular species to be estimated from statistical equilibrium calculations, using observations of rarer isotopomers where possible, to minimize opacity effects. The most striking abundance variations relative to HCO+ along the ridge were seen for HC3N, CH3CCH, and SO, while smaller variations were seen in CS, C2H, and HCN. On the other hand, the NH3, HNC, and N2H+ abundances relative to HCO+ were determined to be constant, indicating that the so-called NH3 peak in TMC-1 is probably a peak in the ammonia column density rather than a relative abundance peak. In contrast, the well-studied cyanopolyyne peak is most likely due to an enhancement in the abundance of long-chain carbon species. Comparisons of the derived abundances to the results of time-dependent chemical models show good overall agreement for chemical timescales around 10(5) yr. We find that the observed abundance gradients can be explained either by a small variation in the chemical timescale from 1.2 x 10(5) to 1.8 x 10(5) yr or by a factor of 2 change in the density along the ridge. Alternatively, a variation in the C/O ratio from 0.4 to 0.5 along the ridge produces an abundance gradient similar to that observed.

System analysis of Phycomyces light-growth response with Gaussian white noise and sum-of-sinusoids test stimuli

The sporangiophore (fruiting body) of the fungus Phycomyces modulates its elongation rate in response to changes in blue light intensity. This light-growth response of wild-type and behavioral mutant strains has been studied extensively by two methods of nonlinear system identification employing Gaussian white noise and sum-of-sinusoids test stimuli. Both methods are in the framework of the Wiener theory of nonlinear systems. The light-growth response is well described by the first-order Wiener G-functional; addition of the second-order functional improves the precision. The Wiener kernel of first-order resembles the light-growth response to a nonsaturating pulse stimulus. The second-order kernel indicates the nonlinear property of rectification. The kernels have been interpreted by system analysis methods in the frequency domain. A nonlinear dynamic model of the light-growth response has been developed from the kernels obtained by both methods. The model includes a nonlinear dynamic subsystem, including a squarer, followed by a linear dynamic subsystem (which, by itself, accounts for the first-order kernel). This parametric model has been used to evaluate light-growth response kernels under various conditions (viz. wavelength, temperature, and genetic strain). The kernels of single and double mutants have been analyzed jointly with a nonparametric model to reveal interactions among the products of eight genes that influence the light-growth response. The extent of interactions found suggests that these gene products function together in a molecular complex.

System analysis of Phycomyces light-growth response with sum-of-sinusoids test stimuli

The light-growth response of Phycomyces has been studied with the sum-of-sinusoids method of nonlinear system identification (Victor, J.D., and R.M. Shapley, 1980, Biophys. J., 29:459). This transient response of the sporangiophore has been treated as a black-box system with one input (logarithm of the light intensity, I) and one output (elongation rate). The light intensity was modulated so that log I, as a function of time, was a sum of sinusoids. The log-mean intensity was 10(-4) W m-2 and the wavelength was 477 nm. The first- and second-order frequency kernels, which represent the linear and nonlinear behavior of the system, were obtained from the Fourier transform of the response at the appropriate component and combination frequencies. Although the first-order kernel accounts for most of the response, there remains a significant nonlinearity beyond the logarithmic transducer presumed to occur at the input of the sensory transduction chain. From the analysis of the frequency kernels, we have derived a dynamic nonlinear model of the light-growth response system. The model consists of a nonlinear subsystem followed by a linear subsystem. The model parameters were estimated from a combined nonlinear least-squares fit to the first- and second-order frequency kernels.

System analysis of Phycomyces light-growth response. Photoreceptor and hypertropic mutants

The light-growth responses of Phycomyces behavioral mutants, defective in genes madB, madC, and madH, were studied with the sum-of-sinusoids method of system identification. Modified phototropic action spectra of these mutants have indicated that they have altered photoreceptors (P. Galland and E.D. Lipson, 1985, Photochem. Photobiol. 41:331). In the two preceding papers, a kinetic model of the light-growth response system was developed and applied to wild-type frequency kernels at several wavelengths and temperatures. The present mutant studies were conducted at wavelength 477 nm. The log-mean intensity was 6 X 10(-2)W m-2 for the madB and madC night-blind mutants, and 10(-4)W m-2 for the madH hypertropic mutant. The prolonged light-growth responses of the madB and madC mutants are reflected in the reduced dynamic order of their frequency kernels. The linear response of the hypertropic mutant is essentially normal, but its nonlinear behavior shows modified dynamics. The behavior of these mutants can be accounted for by suitable modifications of the parametric model of the system. These modifications together support the hypothesis that an integrated complex mediates sensory transduction in the light responses and other responses of the sporangiophore.

System analysis of Phycomyces light-growth response. Wavelength and temperature dependence

The light-growth response of the Phycomyces sporangiophore was studied further with the sum-of-sinusoids method of nonlinear system identification. The first- and second-order frequency kernels, which represent the input-output relation of the system, were determined at 12 wavelengths (383-529 nm) and 4 temperatures (17 degrees, 20 degrees, 23 degrees, and 26 degrees C). The parametric model of the light-growth response system, introduced in the preceding paper, consists of nonlinear and linear dynamic subsystems in cascade. The model parameters were analyzed as functions of wavelength and temperature. At longer wavelengths, the system becomes more nonlinear. The latency and the bandwidth (cutoff frequency) of the system also vary significantly with wavelength. In addition, the latency decreases progressively with temperature (Q10 = 1.6). At low temperature (17 degrees C), the bandwidth is reduced. The results indicate that about half of the latency is due to physical processes such as diffusion, and the other half to enzymatic reactions. The dynamics of the nonlinear subsystem also vary with wavelength. The dependence of various model components on wavelength supports the hypothesis that the light-growth response, as well as phototropism, are mediated by multiple interacting photoreceptors.