RESEARCH ARTICLE – Pharmaceutical Biotechnology

Bioreactor Process Parameter Screening Utilizing a Plackett–Burman Design for a Model Monoclonal Antibody CYRUS D. AGARABI,1 JOHN E. SCHIEL,2 SCOTT C. LUTE,3 BRITTANY K. CHAVEZ,3 MICHAEL T. BOYNE II,4 KURT A. BRORSON,3 MANSOOR A. KHAN,1 ERIK K. READ3 1

Division of Product Quality Research, Office of Testing and Research, OPS, CDER, FDA, Silver Spring, Maryland Biomolecular Measurement Division, Bioanalytical Science Group, National Institute of Standards and Technology, Gaithersburg, Maryland 3 Division of Monoclonal Antibodies, Office of Biotechnology Products, OPS, CDER, FDA, Silver Spring, Maryland 4 Division of Pharmaceutical Analysis, Office of Testing and Research, OPS, CDER, FDA, St. Louis, Missouri 2

Received 4 December 2014; revised 13 January 2015; accepted 10 February 2015 Published online 11 March 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.24420 ABSTRACT: Consistent high-quality antibody yield is a key goal for cell culture bioprocessing. This endpoint is typically achieved in commercial settings through product and process engineering of bioreactor parameters during development. When the process is complex and not optimized, small changes in composition and control may yield a finished product of less desirable quality. Therefore, changes proposed to currently validated processes usually require justification and are reported to the US FDA for approval. Recently, design-ofexperiments-based approaches have been explored to rapidly and efficiently achieve this goal of optimized yield with a better understanding of product and process variables that affect a product’s critical quality attributes. Here, we present a laboratory-scale model culture where we apply a Plackett–Burman screening design to parallel cultures to study the main effects of 11 process variables. This exercise allowed us to determine the relative importance of these variables and identify the most important factors to be further optimized in order to control both desirable and undesirable glycan profiles. We found engineering changes relating to culture temperature and nonessential amino acid supplementation significantly impacted glycan profiles associated with fucosylation, ␤-galactosylation, and sialylation. All of these C 2015 Wiley Periodicals, Inc. and the American Pharmacists Association J Pharm are important for monoclonal antibody product quality.  Sci 104:1919–1928, 2015 Keywords: biotechnology; glycoprotein; mass spectrometry; glycosylation; monoclonal antibody; quality by design (QbD); cell culture; design of experiments (DoE); Plackett–Burman; glycan profiling

INTRODUCTION The goal of commercial cell culture is to economically manufacture a protein suitable for downstream processing/purification that will yield a safe and effective final drug product. A balance must be struck between product yield and consistency of critical biochemical quality attributes. Factors that influence cell growth, physiology, and subsequently affect protein yield include: bioreactor process parameters (e.g., temperature,1 dissolved oxygen (DO),2 pH,3 agitation4 ), nutritional supplementation (e.g., glucose,5 amino acids,6 cortisone7 ), and the addition of chemical induction factors (e.g., butyrate8 ). Yet changes in these same factors can adversely affect product quality, for example, by altering the distribution of charge variants9 or glycoforms6,10 and can even lead to product crystallization.11 In particular, glycosylation has been identified as a critical quality attribute (CQA) for many antibody-based products, particularly oncology products where tumor cell lysis is a critical component of the mechanism of action and changes to the Fc glycan structure directly impact Fc effector function(s).12 For example, antibodydependent cellular cytotoxicity (ADCC), mediated by Natural Killer cells through Fc(RIIIa bridging to antibody bound on the target cell, is a likely effector function responsible for Correspondence to: Mansoor A. Khan (Telephone: +301-796-0016; Fax: +301796-9816; E-mail: [email protected]) This article contains supplementary material available from the authors upon request or via the Internet at http://onlinelibrary.wiley.com/. Journal of Pharmaceutical Sciences, Vol. 104, 1919–1928 (2015)  C 2015 Wiley Periodicals, Inc. and the American Pharmacists Association

tumor death.13 Fucosylation of the biantennary N-linked glycans on antibodies can modulate the binding to Fc(RIII,14,15 and thus ADCC activity.16 Therefore, consistency and control of fucosylation on therapeutic antibodies on a batch-by-batch basis is considered to be a high priority of biopharmaceutical firms and regulatory authorities. The ICH Guidance for Industry Q8(R2) Pharmaceutical Development Quality by Design (QbD) annex17 calls upon biotechnology and pharmaceutical manufacturers to identify critical process parameter ranges where consistently high-quality product is produced.18,19 However, varying all of the potential cell culture factors would result in a colossal and costprohibitive study, even if performed in parallel cultures. Also, it is likely that not all possible bioreactor or cell culture parameters will impact CQAs, especially when maintained within typical commercial manufacturing ranges. Thus, alternative strategies for implementing QbD in cell culture are warranted.9 Unlike synthetic chemical reactors for small molecules, a cell culture process takes several days to months to complete. Therefore, in order to understand product and process variability, it may be necessary to run several bioreactors simultaneously. Laboratory-scale (0.5–15 L) parallel bioreactor systems have been developed by several vendors to run six or more simultaneous, independently controlled cultures. Such bioreactors rely on sophisticated process control software and feedback loop algorithms that maintain different critical parameter settings in each bioreactor, such as agitation, pH, DO, and temperature. Many small-scale systems have been designed to control

Agarabi et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:1919–1928, 2015

1919

A saturated 12 experiment Plackett–Burman design was chosen to screen 11 variables at two levels (Table 1). Meaningful operating ranges for all parameters were chosen based on Agarabi et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:1919–1928, 2015

Add Add Add Add Add Add

Add

Add

Add Add Add Add Add Add

Add Add

Add Add

Shift

Shift Shift

Shift Shift Shift

Shift Shift

Drip Drip Bolus Bolus Drip Drip Bolus Bolus Drip Bolus Bolus Drip Shift Shift Shift Shift

50,000 50,000 50,000 100,000 100,000 100,000 100,000 100,000 100,000 50,000 50,000 50,000 35.5 37 35.5 37 35.5 37 35.5 37 35.5 35.5 37 37 170 90 170 170 90 170 170 90 90 90 90 170 20 40 40 40 40 40 20 20 20 40 20 20

0.3 0.3 0.7 0.3 0.3 0.7 0.3 0.7 0.7 0.7 0.3 0.7

Fatty Acids Nonessential AA Feeding Strategy Temperature Shift pH Shift Inoculation Density (Cells/mL) Temperature (°C)

−−+−−+−+++− +−−+−+++−−− +++−−−+−−+− +−+++−−−+−− +−−−+−−+−++ +++++++++++ −−+−+++−−−+ −+−+++−−−+− −+−−+−+++−− ++−−−+−−+−+ −−−+−−+−+++ −+++−−−+−−+

Experimental Design Rationale

1 2 3 4 5 6 7 8 9 10 11 12

An in-house murine IgG3:6 antibody-producing hybridoma28 adapted to suspension culture was grown in CD–hybridoma media with 8 mmol/L glutamine29,30 in 2 L spinner flasks prior to transfer into 0.8 L of media in the bioreactors. Viable cell density (VCD, VC/mL) was monitored with manual hemocytometer counts.

Pattern

Murine Hybridoma Seed Culture

Agitation Rate (RPM)

Unless noted otherwise in the text, reagents were purchased from the same source and used at similar concentrations as described in Read et al.6 Chemically defined (CD) hybridoma AGT culture media (Gibco, Grand Island, NY) was prepared with supplemented amino acids (aspartate, cysteine, methionine, threonine, tryptophan, and tyrosine). Supplements used in the experimental design included: hydrocortisone (SigmaAldrich, St. Louis, MO), 1× MEM non-essential amino acids (NE-AA) (Corning Cell Gro, Manassas, VA), and CD lipid concentrate (Gibco).

Sparge Rate (SLPH)

Reagents

DO (%) Set Point

MATERIALS AND METHODS

Pattern Number

important physical settings and match important engineering coefficients [e.g., mixing rate, volumetric mass transfer coefficient (kLa)] to be representative of larger-scale bioreactors.20 The enhanced capabilities of small-scale parallel bioreactor cultures allows screening design of experiments (DoE) approaches such as fractional factorial designs to efficiently identify process parameters that can affect CQAs and have successfully been explored by Abu-Absi et al.21 and Banerjee et al.9 The Plackett–Burman design offers an alternative to the fractional factorial design and enables efficient screening of “main effects,” that is, the impact of particular parameters without directly evaluating their interactions.22 The DoE approach with a Plackett–Burman screening has extensively been utilized by one or more of the current authors for formulation development.23–25 The design can be fully saturated with 12 experiments capable of exploring 11 variables, versus a traditional full factorial design of 211 = 2048 experiments.26 The Plackett–Burman design has successfully been used in the development of serum-free cell culture media for CHO cells.27 This approach can rapidly identify pertinent variables for further optimization of operational ranges and interaction effects utilizing response surface methods and other advanced DoEs. Here, we prototype a DoE-based cell culture design strategy for rapid, laboratory-scale screening using an established model hybridoma cell culture system, which has been subjected to extensive process understanding activities in the past. The objective of the present screening experimental design is a hypothesis generating approach to evaluate 11 product and process variables. They were rank-ordered with respect to their relative importance in impacting the CQAs of the cell culture unit operation as a first step in an overall process optimization program. Narrowing down variables is critical for the design of subsequent follow up statistical studies for factor level optimization.

Hydrocortisone

RESEARCH ARTICLE – Pharmaceutical Biotechnology

Table 1. Culture Process Set Points, Events, and Supplements for the Plackett–Burman DoE

1920

DOI 10.1002/jps.24420

RESEARCH ARTICLE – Pharmaceutical Biotechnology

1921

Figure 1. Representative bioreactor cell growth (experimental pattern 4) demonstrates consistent and predicted growth from a healthy bioreactor culture, as measured by viable cells (VC), nonviable cells (nVC), total cells (TC), integral viable cell days (IVC), and viable fraction (Vf).

biologically relevant ranges, commercial practices, and previous empirical run data6 from the same cell line and bioreactor platform. Independent variable levels were set at either a low or high value for the five continuous variables and as action or no action for six categorical variables. Dependent variables were chosen based on standard lot release testing and advanced characterization techniques to understand product quality, summarized in Supplementary Table 1. Four control reactors were run at baseline historical conditions, representing intermediate process values for continuous factors, to account for inherent variability between cultures. JMP Version 10 (SAS Institute, Inc., Cary, North Caroline) was used to develop the experimental design, randomize parallel reactor order, and analyze the data from the statistical design. Bioreactor Processing Conditions A DasGip (DasGip Biotools, Shrewsbury, Massachusetts) parallel bioreactor system with eight 1.2 L vessels was run in fedbatch mode for 5 days (120 h). Our experimental plan consisted of 12 parallel runs varying typical bioreactor operating parameters in a Plackett–Burman statistical design, illustrated in Table 1. The initial pH set point of 7.2 was maintained automatically by the DasGip control software (version 4.5 rev 70) via CO2 gas sparging and 0.4 mol/L sodium hydroxide addition. For the pH shift, the controller set point was changed to pH 6.9 and allowed to reach the new value without further manipulations. The DO set point was maintained by the DasGip 4-gas algorithm (CO2 , N2 , O2 , and air) and vessels were sparged using a 10 :m pore size sintered tip microsparger. Temperature was maintained at 37°C or 35.5°C, and the temperature shift to 34°C on day 3 was achieved rapidly (

Bioreactor process parameter screening utilizing a Plackett-Burman design for a model monoclonal antibody.

Consistent high-quality antibody yield is a key goal for cell culture bioprocessing. This endpoint is typically achieved in commercial settings throug...
853KB Sizes 0 Downloads 11 Views