rAAV Manufacturing: The Challenges of Soft Sensing during Upstream Processing

Drug product formulation and fill/finish manufacturing process considerations for AAV-based genomic medicines

Adeno-associated viruses (AAVs) have become the delivery medium of choice for a variety of genomic medicine applications i.e., gene therapy, gene editing/regulation, and ex-vivo cell therapy. AAVs are protein-DNA complexes which have unique stability characteristics that are susceptible to various stress exposure conditions commonly seen in the drug product (DP) life cycle. This review takes a comprehensive look at AAV DP formulation and process development considerations that could impact critical quality attributes (CQAs) during manufacturing, packaging, shipping, and clinical use. Additional aspects related to AAV development reviewed herein are: (1) Different AAV serotypes with unique protein sequences and charge characteristics potentially leading to discrete stability profiles; (2) Manufacturing process challenges and optimization efforts to improve yield, recovery and purity especially during early development activities; and (3) Defining and identifying CQAs with analytical methods which are constantly evolving and present unique characterization challenges for AAV-based products.

High-yield recombinant adeno-associated viral vector production by multivariate optimization of bioprocess and transfection conditions

Recombinant adeno-associated viral vectors (rAAVs) are one of the most used vehicles for gene therapy, with five rAAV therapeutics commercially approved by the FDA. To improve product yield, we optimized the suspension production process of rAAV8 vectors carrying a proprietary transgene using a commercially available transfection reagent, FectoVIR-AAV. Using a miniaturized automated 250 mL scale bioreactor system, we generated models of vector genome (vg) titer, capsid (cp) titer, and Vg:Cp percentage from two multivariate design of experiment studies, one centered around bioreactor operating parameters, and another based on the transfection conditions. Using the optimized process returned from these models, the vector genome titer from the bioreactor was improved to beyond 1 × 1012  vg/mL. Five critical parameters were identified that had large effects on the pre-purification vector quantity-the transfection pH, production pH, complexation time, viable cell density at transfection, and transfection reagent to DNA ratio. The optimized process was further assessed for its performance extending to six AAV serotypes, namely AAV1, AAV2, AAV5, AAV6, AAV8, and AAV9 carrying a transgene encoding for green fluorescent protein (GFP). Five of the six serotypes returned higher vector genome titers than the control condition. These data suggest that the choice of transfection reagent is a major factor in improving vector yield. The multivariate design of experiment approach is a powerful way to optimize production processes, and the optimized process from one AAV vector can to some extent be generalized to other serotypes and transgenes to accelerate development timelines of new programs.

How Not to Make the Joint Extended Kalman Filter Fail with Unstructured Mechanistic Models

The unstructured mechanistic model (UMM) allows for modeling the macro-scale of a phenomenon without known mechanisms. This is extremely useful in biomanufacturing because using the UMM for the joint estimation of states and parameters with an extended Kalman filter (JEKF) can enable the real-time monitoring of bioprocesses with unknown mechanisms. However, the UMM commonly used in biomanufacturing contains ordinary differential equations (ODEs) with unshared parameters, weak variables, and weak terms. When such a UMM is coupled with an initial state error covariance matrix P(t=0) and a process error covariance matrix Q with uncorrelated elements, along with just one measured state variable, the joint extended Kalman filter (JEKF) fails to estimate the unshared parameters and state simultaneously. This is because the Kalman gain corresponding to the unshared parameter remains constant and equal to zero. In this work, we formally describe this failure case, present the proof of JEKF failure, and propose an approach called SANTO to side-step this failure case. The SANTO approach consists of adding a quantity to the state error covariance between the measured state variable and unshared parameter in the initial P(t = 0) of the matrix Ricatti differential equation to compute the predicted error covariance matrix of the state and prevent the Kalman gain from being zero. Our empirical evaluations using synthetic and real datasets reveal significant improvements: SANTO achieved a reduction in root-mean-square percentage error (RMSPE) of up to approximately 17% compared to the classical JEKF, indicating a substantial enhancement in estimation accuracy.