CROI 2018 Abstract eBook

Abstract eBook

Poster Abstracts

301 DEFINING CORRELATES OF HUMORAL IMMUNE PROTECTION AGAINST EBV Christina B. Karsten 1 , Sally A. Shin 1 , Howard M. Heller 2 , Kumaran Kolandaivelu 3 , Jaap M. Middeldorp 4 , Bruce D. Walker 1 , Boris Julg 1 , Galit Alter 1 1 Ragon Institute of MGH, MIT and Harvard, Cambridge, MA, USA, 2 MIT, Cambridge, MA, USA, 3 MIT Institute for Medical Engineering & Science, Cambridge, MA, USA, 4 VU University Medical Center, Amsterdam, Netherlands Background: Epstein-Barr virus (EBV)-associated cancers including Burkitt, Hodgkin and non-Hodgkin lymphoma pose a deadly health threat, especially among HIV-1 infected populations. However, a protective vaccine to prevent EBV infection and or limit disease has yet to be developed. While previous vaccine efforts have focused on the induction of either neutralizing antibodies or CD8+ T-cell responses to prevent or kill infected cells, mounting evidence points to a potentially critical role for cytotoxic antibodies, able to deploy the antiviral activity of the innate immune system, in the control and clearance of EBV. Thus, here we aimed to define whether non-neutralizing antibody functions, able to recruit a broad array of innate immune effector functions, might selectively evolve following EBV infection. Methods: Systems serology profiling was performed in a cohort of 12 EBV- infected patients that were enrolled during the acute stage of infection and sampled for a year. Both functional and biophysical assays were performed to define cross-sectional and longitudinal differences in the humoral immune response against acute- and latency- associated EBV antigens. Results: As anticipated, IgG1 and IgG3 antibody subclasses, known to drive enhanced antiviral function, predominated the IgG immune response to the EBV viral envelope protein (gp350/220), capsid antigen (p18), early antigen (p47/54) and latent protein (EBNA 1). However, variability was observed among infected subjects suggesting the existence of different responder groups or patterns. While all EBV protein-specific antibodies were unable to recruit monocyte-dependent phagocytosis, p18-specific antibodies induced low levels of phagocytosis by neutrophils (ADNP) early in infection (0-2 weeks post enrollment; average median fold over background response (SD)= 2.33±0.16). Moreover, ADNP was associated with p18-specific IgM antibodies, indicating that an IgM, rather than an IgG response to this target, may be largely responsible for this antiviral function (r=0.77, p=0.044). Further, EBNA 1-specific IgMs were found to correlate more strongly (r=0.99, p= 0.000013) with the severity of EBV symptoms than the previously described p18-specific IgA suggesting a potential role of IgMs in the pathogenesis of EBV infection. Conclusion: Our systems serology approach identified ADNP as a first antibody mechanism of action against EBV with the potential to be exploited for future treatment and vaccine strategies. Nickita Mehta , Magdalena Sips, Jishnu Das, Marina Krykbaeva, Anne-Sophie Dugast, Fernanda Cerqueira, Galit Alter Ragon Institute of MGH, MIT and Harvard, Cambridge, MA, USA Background: Neutrophils are the most abundant white blood cell and are rapid responders to infections. Neutrophils are also the most potent mediators of Fc-effector functions and in addition to phagocytosis, neutrophils degranulate and can kill through the formation of extracellular traps. However, little is known about the role of neutrophils in antiviral control of HIV. Thus, here we aimed to define the role of neutrophil-recruiting antibodies in antiviral control of HIV. Methods: Because neutrophils express both an IgA and IgG Fc-receptor, both IgA and IgG were purified from HIV infected patient serum and compared to healthy controls. The functional activity of these antibodies were compared for phagocytic, degranulation, NET formation, and release of elastase activity. Results: Elevated levels of all neutrophil recruiting antibody subclasses and isotypes were observed in spontaneous controllers of HIV including higher HIV- specific IgG3, IgA1 and IgA2 titers. Moreover, while both IgA and IgG induced HIV-specific neutrophil phagocytosis, IgA1 drove enhanced phagocytosis compared to IgG or IgA2. Conversely, IgA2 induced enhanced degranulation and NET formation, which were not induced by IgA1. Additionally, IgA2 synergized with IgG1 and IgG3 to recruit neutrophil functions. However the depletion of IgA significantly reduced neutrophil activation, despite the presence of IgG1 and IgG3. Conclusion: These data suggest that HIV-specific IgA responses are not only enriched among spontaneous controllers of HIV, but may be selectively enriched to drive robust and rapid recruitment of neutrophils to clear and destroy HIV 302 LEVERAGING NEUTROPHILS TO CONTROL HIV VIA IGA

infected cells. Surprising differences in functionality were observed among the IgA subclasses, each poised to leverage distinct innate immune effector functions. Overall, these data point to a potentially critical immunoprotective role for IgA via the recruitment of the tremendous anti-viral potential of neutrophils. 303 THE DEVELOPMENT OF A NOVEL MACHINE-LEARNING-GUIDED HIV IMMUNOGEN DESIGN Wen-Han Yu 1 , Galit Alter 1 , Christina B. Karsten 1 , Douglas Lauffenburger 2 , Daniel Lingwood 1 , Dan Barouch 3 1 Ragon Institute of MGH, MIT and Harvard, Cambridge, MA, USA, 2 MIT, Cambridge, MA, USA, 3 Beth Israel Deaconess Medical Center, Boston, MA, USA Background: HIV envelope glycoprotein (Env) is the major target for broadly neutralizing antibodies (bNAbs). Both the underlying protein and glycan shield represent antigenic determinants for many potent bNAbs. Several state-of- the-art protein-engineering approaches have been developed to generate HIV immunogens, among which the native-like trimer (SOSIP.664) has shown promise in inducing NAb responses against autologus tier-2 viruses. However, the roadmap for HIV vaccine development remains hampered by the inability to drive broad NAb responses to heterologous tier2 and 3 viruses. While several stragies have focused on improving SOSIP.664 or developing scaffolds to selectively induce mature/gemline bNAb, via immunofocusing approach, these immunogens do not take sequence and glycan diversity into account. Methods: To address this challenge, we developed a machine-learning model, aimed at deeply learning the complexity of the HIV sequence/glycan diversity linked to bNAb neutralization sensitivities, to guide an in silico rational immunogen design aimed at selectively inducing bNAbs while simultaneously blocking non-bNAb binding. Specifically, this Bayesian Support Vector Regression (SVR) model were based on 6,886 Env sequences, glycan occupancy and bNAb neutralization sensitivities (IC50). The predictive model was next used to guide a de novo sequence evolutionary algorithm. The algorithm imitated the process of natural selection to iteratively evolve a given trimeric gp140 Env sequence. The final convergent immunogen was designed to enhance/impair specific bNAb/non-bNAb binding without compromising structural stability and sequence conservation. Results: This machine-learning models were able to predict total 130 bNAb and non-bNAb IC50 profiles, and 80% of the predictive models showed a robust prediction performance (average correlation R-square between predicted and actual IC50 profile larger than 0.9), and the feature signature that enhances bNAbs, while simultaneously occluding non-bNAb binding was observed. This design approach has produced novel proof-of-principle immunogens that successfully skew binding of bNAbs PGT121 or PGT128 in vitro, with design of an enhanced PGT121 and/or PGDM1400 immunogen in process. Conclusion: This approach provides an innovative design strategy, rooted in deep learning of glycobiology and large sequence data, to predictably modulate NAb binding aimed at focusing the humoral immune response to sites of vulnerability. 304 DEFINING THE COMMON CORRELATES OF HIV VACCINE TRIALS CROSS- PREDICTS VACCINE EFFICACY Wen-Han Yu 1 , Dan Barouch 2 , Douglas Lauffenburger 3 , Margaret Ackerman 4 , Caitlyn Linde 1 , Thomas Broge 1 , Galit Alter 1 1 Ragon Institute of MGH, MIT and Harvard, Cambridge, MA, USA, 2 Beth Israel Deaconess Medical Center, Boston, MA, USA, 3 MIT, Cambridge, MA, USA, 4 Dartmouth College, Hanover, NH, USA Background: Over the last decade, several different HIV vaccine candidates have shown robust protection against SIV challenges in non-human primate (NHP) models, and immunological correlates of protection have often been identified. However, as these correlates are frequently different for each of the vaccine regimens, a major hurdle has been the ability to translate these data sets into a computational model in order to accurately cross-predict protective efficacy in future human vaccine trial. Methods: To address this, we used Systems Serology to comprehensively profile the biophysical and functional properties of the humoral immune response in two independent NHP trials. In the first trial, NHPs were primed with adenovirus (Ad) vectors expressing SIV Env/Gag/Pol, and then boosted with Ad or SIV Env gp140 proteins. The second trial was performed under the similar vaccine regimens, but the SIV antigens were substituted by HIV proteins. The

Poster Abstracts

CROI 2018 106