SARS-CoV-2 mechanistic correlates of protection in non-human primates: insight from modelling response to vaccines.
Marie Alexandre (1), Romain Marlin (2), Mélanie Prague (1), Séverin Coleon (3,4), Nidhal Kahlaoui (2), Sylvain Cardinaud (3,4), Christine Lacabaratz (3,4), Aurelie Wiedemann (3,4), Sandra Zurawski (5), Gerard Zurawski (5), Olivier Schwartz (3,6,7), Rogier W Sanders (8), Roger Le Grand (2), Yves Levy (3,4,9), Rodolphe Thiébaut (1,3,10)
(1) Bordeaux University, Department of Public Health, Inserm Bordeaux Population Health Research Centre, Inria SISTM, UMR 1219; Bordeaux, France. (2) Center for Immunology of Viral, Auto-immune, Hematological and Bacterial diseases (IMVA-HB/IDMIT), Université Paris-Saclay, Inserm, CEA; Fontenay-aux-Roses, France. (3) Vaccine Research Institute; Creteil, France. (4) Inserm U955, Equipe 16; Créteil, France. (5) Baylor Scott and White Research Institute and INSERM U955; Dallas, Texas, United States of America. (6) Virus & Immunity Unit, Department of Virology, Institut Pasteur; Paris, France. (7) CNRS UMR 3569; Paris, France. (8) Department of Medical Microbiology, Amsterdam UMC, University of Amsterdam, Amsterdam Infection & Immunity Institute; 1105 AZ Amsterdam, the Netherlands. (9) AP-HP, Hôpital Henri-Mondor Albert-Chenevier, Service d'Immunologie Clinique et Maladies Infectieuses; Créteil, France. (10) CHU Bordeaux, Department of Medical information; Bordeaux, France
Objectives: Despite the development of efficient SARS-CoV-2 vaccines, the identification of correlates of protection (CoP) is essential for further vaccine development. Today, binding and neutralizing antibodies appear to be robust surrogate markers of vaccine-induced protection [1]. Nevertheless, evidence of a causal relationship with virus control in vivo remains unclear. Accordingly, our goal is to propose a modelling approach for identifying the biological effect of immunological markers on the virus-host interaction defining mechanistic CoP (mCoP) [2].
Methods: We developed a mathematical building strategy using mechanistic models based on ordinary differential equations (ODEs) coupled with a model building approach in order to identify immunological mCoP against SARS-CoV-2. As a first step, a within-host mechanistic model was used to describe the interactions between the immune system and the virus and to identify the mechanisms of action of vaccine-induced and natural immunity. This multi-layered model combined: i) adapted target-cell ODE-based models as structural models, ii) statistical models describing model parameters as mixed-effect models to account for covariates and inter-individual variability, and iii) a model of observation linking observations to ODE compartments. In a second step, a selection algorithm for time-varying covariates was applied to identify mCoP among numerous immune markers (binding and neutralizing antibodies, cytokines, and T cell responses). Three criteria were proposed based on Prentice's definition [3] of mCoP: the best fits, the capture of the effect of the group of intervention and the best explanation of inter-individual variability. The method was independently applied to data from 3 non-human primates (NHPs) vaccine studies testing: (a) a protein-based vaccine targeting the binding of the RBD domain of the SARS-CoV-2 spike protein to CD40, with 6 naive NHPs and 12 SARS-CoV-2 convalescents, 6 of whom were vaccinated, (b) a two-component spike nanoparticle vaccine on 4 naive NHPs and 10 vaccinated NHPs, and (c) the mRNA-1273 vaccine with 8 PBS and 16 vaccinated NHPs receiving 10 or 100 μg of vaccine.
Results: We identified viral control by the blockage of new infection of target cells as the main mechanism of natural SARS-CoV-2 and vaccine-elicited immunity in the three studies with a mean infectivity reduced by more than 82% (95%CI: [12%; 96%]) and 97% (95%CI: [59%; 100%]), respectively. With the exception of the mRNA-1273 vaccine, the induced immune response also increased the rate of loss of infected cells, whose estimated clearance was increased by 80% in convalescent or vaccinated NHPs (p<0.007). In studies a) and b), neutralization quantified by the RBD-ACE2 binding inhibition assay appeared as a consistent mCoP for the decrease in viral infectivity, while neutralization measured on live cells with the luciferase marker played this role for the mRNA vaccine. Binding antibodies were associated with the increase in loss of infected cells, but the three criteria were not always fulfilled.
Conclusions: Our original statistical approach allowed us to defined mCoP in NHP studies. Blocking infection of new target cells was identified as the main mechanism of action of both studied vaccines and natural infection. We showed that this mechanism of protection is captured by (pseudo-) neutralization assays. In addition, we pointed out that other mechanisms could also be identified, particularly those that promote elimination of target cells. However, more specific markers should be used and deeper analyses should be performed for this purpose.
References:
[1] J. Perry, S. Osman, J. Wright, M. Richard-Greenblatt, S. A. Buchan, M. Sadarangani, S. Bolotin, Does a humoral correlate of protection exist for SARS-CoV-2? A systematic review.medRxiv , 2022.01.21.22269667 (2022).
[2] M. Alexandre, R. Marlin, M. Prague, S. Coleon, N. Kahlaoui, S. Cardinaud, T. Naninck, B. Delache, M. Surenaud, M. Galhaut, N. Dereuddre-Bosquet, M. Cavarelli, P. Maisonnasse, M. Centlivre, C. Lacabaratz, A. Wiedemann, S. Zurawski, G. Zurawski, O. Schwartz, R. W. Sanders, R. L. Grand, Y. Levy, R. Thiébaut, SARS-CoV-2 mechanistic correlates of protection: insight from modelling response to vaccines.bioRxiv , 2021.10.29.466418 (2021).
[3] R. L. Prentice, Surrogate endpoints in clinical trials: Definition and operational criteria. Statistics in Medicine 8, 431–440 (1989).