Comparison of techniques for monitoring antibody fragment production in E. coli fermentation cultures

Development of an automated platform for the optimal production of glycoconjugate vaccines expressed in Escherichia coli

Protein Glycan Coupling Technology (PGCT) uses purposely modified bacterial cells to produce recombinant glycoconjugate vaccines. This vaccine platform holds great potential in this context, namely due to its modular nature, the simplified production process in comparison to traditional chemical conjugation methods, and its amenability to scaled-up operations. As a result, a considerable reduction in production time and cost is expected, making PGCT-made vaccines a suitable vaccine technology for low-middle income countries, where vaccine coverage remains predominantly low and inconsistent. This work aims to develop an integrated whole-process automated platform for the screening of PGCT-made glycoconjugate vaccine candidates. The successful translation of a bench scale process for glycoconjugate production to a microscale automated setting was achieved. This was integrated with a numerical computational software that allowed hands-free operation and a platform adaptable to biological variation over the course of a production process. Platform robustness was proven with both technical and biological replicates and subsequently the platform was used to screen for the most favourable conditions for production of a pneumococcal serotype 4 vaccine candidate. This work establishes an effective automated platform that enabled the identification of the most suitable E. coli strain and genetic constructs to be used in ongoing early phase research and be further brought into preclinical trials.

Three-dimensional chromatography for purification and characterization of antibody fragments and related impurities from Escherichia coli crude extracts

Antibody fragments (Fab) are often produced by recombinant methods in Escherichia coli as no glycosylation is needed. Besides the correctly expressed Fab molecule, a multitude of host cell impurities and product related impurities are present in the crude sample. The identification and characterization of the product-related impurities, such as modified Fab-molecules or free light chain, are of utmost importance. The objective of this work was to design a purification strategy to isolate and characterize Fab and related impurities. A three-dimensional chromatography method was established, consisting of two affinity steps (Protein G and Protein L) and subsequent cation exchange chromatography, followed by mass spectrometry analysis of the purified samples. The procedure was automated by collecting the eluted target species in loops and directly loading the samples onto the high-resolution cation exchange chromatography column. As an example, four different Fab molecules are characterized. All four samples contained mainly the correct Fab, while only one showed extensive N-terminal pyroglutamate formation of the Fab. In another case, we found a light chain variant with uncleaved amino acids from the lead molecule, which was not used for the formation of whole Fab as only correct Fab was found in that sample. Impurities with lower molecular weights, which were bound on the Protein L column, were observed in all samples, and identified as fragments of the light chain. In conclusion, we have devised a platform for characterizing Fab and Fab-related impurities, which significantly facilitated strain selection and optimization of cultivation conditions.

Extraction of recombinant periplasmic proteins under industrially relevant process conditions: Selectivity and yield strongly depend on protein titer and methodology

In this work, we attempted to identify a method for the selective extraction of periplasmic endogenously expressed proteins, which is applicable at an industrial scale. For this purpose, we used an expression model that allows coexpression of two fluorescent proteins, each of which is specifically targeted to either the cytoplasm or periplasm. We assessed a number of scalable lysis methods (high‐pressure homogenization, osmotic shock procedures, extraction with ethylenediaminetetraacetic acid, and extraction with deoxycholate) for the ability to selectively extract periplasmic proteins rather than cytoplasmic proteins. Our main conclusion was that although we identified industrially scalable lysis conditions that significantly increased the starting purity for further purification, none of the tested conditions were selective for periplasmic protein over cytoplasmic protein. Furthermore, we demonstrated that efficient extraction of the expressed recombinant proteins was largely dependent on the overall protein concentration in the cell.

Improving Fab' fragment retention in an autonucleolytic Escherichia coli strain by swapping periplasmic nuclease translocation signal from OmpA to DsbA

Objectives

To reduce unwanted Fab’ leakage from an autonucleolytic Escherichia coli strain, which co-expresses OmpA-signalled Staphylococcal nuclease and Fab’ fragment in the periplasm, by substituting in Serratial nuclease and the DsbA periplasm translocation signal as alternatives.

Results

We attempted to genetically fuse a nuclease from Serratia marcescens to the OmpA signal peptide but plasmid construction failed, possibly due to toxicity of the resultant nuclease. Combining Serratial nuclease to the DsbA signal peptide was successful. The strain co-expressing this nuclease and periplasmic Fab’ grew in complex media and exhibited nuclease activity detectable by DNAse agar plate but its growth in defined medium was retarded. Fab’ coexpression with Staphylococcal nuclease fused to the DsbA signal peptide resulted in cells exhibiting nuclease activity and growth in defined medium. In cultivation to high cell density in a 5 l bioreactor, DsbA-fused Staphylococcal nuclease co-expression coincided with reduced Fab’ leakage relative to the original autonucleolytic Fab’ strain with OmpA-fused staphylococcal nuclease.

Conclusions

We successfully rescued Fab’ leakage back to acceptable levels and established a basis for future investigation of the linkage between periplasmic nuclease expression and leakage of co-expressed periplasmic Fab’ fragment to the surrounding growth media.

Matrix effects during monitoring of antibody and host cell proteins using attenuated total reflection spectroscopy

Production of recombinant proteins, e.g. antibodies, requires constant real-time monitoring to optimize yield and quality attributes and to respond to changing production conditions, such as host cell protein (HCP) titers. To date, this monitoring of mammalian cell culture-based processes is done using laborious and time consuming enzyme-linked immunosorbent assays (ELISA), two-dimensional sodium dodecylsulphate polyacrylamide gel electrophoresis, and chromatography-based systems. Measurements are usually performed off-line, requiring regular sample withdrawal associated with increased contamination risk. As information is obtained retrospectively, the reaction time to adapt to process changes is too long, leading to lower yield and higher costs. To address the resulting demand for continuous online-monitoring systems, we present a feasibility study using attenuated total reflection spectroscopy (ATR) to monitor mAb and HCP levels of NS0 cell culture in situ, taking matrix effects into account. Fifty-six NS0 cell culture samples were treated with polyelectrolytes for semi-selective protein precipitation. Additionally, part of the samples was subjected to filtration prior to analysis, to change the background matrix and evaluate effects on chemometric quantification models. General models to quantify HCP and mAb in both filtered and unfiltered matrix showed lower prediction accuracy compared to models designed for a specific matrix. HCP quantification in the range of 2,000-55,000 ng mL(-1) using specific models was accurate for most samples, with results within the accepted limit of an ELISA assay. In contrast, mAb prediction was less accurate, predicting mAb in the range of 0.2-1.7 g L(-1) . As some samples deviated substantially from reference values, further investigations elucidating the suitability of ATR for monitoring are required.

Growth and productivity impacts of periplasmic nuclease expression in an Escherichia coli Fab' fragment production strain

Host cell engineering is becoming a realistic option in whole bioprocess strategies to maximize product manufacturability. High molecular weight (MW) genomic DNA currently hinders bioprocessing of Escherichia coli by causing viscosity in homogenate feedstocks. We previously showed that co-expressing Staphylococcal nuclease and human Fab' fragment in the periplasm of E. coli enables auto-hydrolysis of genomic DNA upon cell disruption, with a consequent reduction in feedstock viscosity and improvement in clarification performance. Here we report the impact of periplasmic nuclease expression on stability of DNA and Fab' fragment in homogenates, host-strain growth kinetics, cell integrity at harvest and Fab' fragment productivity. Nuclease and Fab' plasmids were shown to exert comparable levels of growth burden on the host W3110 E. coli strain. Nuclease co-expression did not compromise either the growth performance or volumetric yield of the production strain. 0.5 g/L Fab' fragment (75 L scale) and 0.7 g/L (20 L scale) was achieved for both unmodified and cell-engineered production strains. Unexpectedly, nuclease-modified cells achieved maximum Fab' levels 8-10 h earlier than the original, unmodified production strain. Scale-down studies of homogenates showed that nuclease-mediated hydrolysis of high MW DNA progressed to completion within minutes of homogenization, even when homogenates were chilled on ice, with no loss of Fab' product and no need for additional co-factors or buffering.

Evaluation of cell disruption effects on primary recovery of antibody fragments using microscale bioprocessing techniques

Intracellular antibody Fab' fragments periplasmically expressed in Escherichia coli require the release of Fab' from the cells before initial product recovery. This work demonstrates the utility of microscale bioprocessing techniques to evaluate the influence of different cell disruption operations on subsequent solid-liquid separation and product recovery. Initially, the industrial method of Fab' release by thermochemical extraction was established experimentally at the microwell scale and was observed to yield Fab' release consistent with the larger scale process. The influence of two further cell disruption operations, homogenization and sonication, on subsequent Fab' recovery by microfiltration was also examined. The results showed that the heat-extracted cells give better dead-end microfiltration performance in terms of permeate flux and specific cake resistance. In contrast, the cell suspensions prepared by homogenization and sonication showed more efficient product release but with lower product purity and poorer microfiltration performance. Having established the various microscale methods the linked sequence was automated on the deck of a laboratory robotic platform and used to show how different conditions during thermochemical extraction impacted on the optimal performance of the linked unit operations. The results illustrate the power of microscale techniques to evaluate crucial unit operation interactions in a bioprocess sequence using only microliter volumes of feed.

Chemical and biological characterisation of a sensor surface for bioprocess monitoring

This paper describes the step-wise fabrication and characterisation of a multi-layer dual polarization interferometry (DPI) based biosensor utilising Protein G (ProG) as the bio-recognition layer for the detection of a fragment antibody (Fab'). The biosensor is capable of monitoring the concentration of Fab' product within the extracellular medium of a fed-batch fermentation after leakage from Escherichia coli (E.coli). The activity, stability and functionality of each sensor layer were analysed in situ using DPI, whilst the chemical identity and homogeneity of the chemical layers were assessed ex situ using X-ray photoelectron spectroscopy (XPS) and secondary ion mass spectrometry (SIMS). Two different biotin linkers were found to produce hugely differing surfaces after the capture of NeutrAvidin™ (NA) and biotinylated Protein G (b-ProG). The hydrophilic (PEG)(4)-biotin linker resulted in a surface where the b-ProG layer was deposited and organised above the NA layer producing an active and stable surface, whilst the hydrophobic LC-biotin linker generated a surface where the b-ProG layer was buried within the NA layer leading to variable surfaces and poor binding of the Fab' target. The biosensor has a detection limit of 1.7 μg/ml with a dynamic range covering two orders of magnitude. The sensor can detect the onset of Fab' leakage as early as 2h following product induction, with high signal-to-noise ratios and little interference from extracellular components. Leakage of Fab' followed a biphasic profile, switching to a more rapid rate 20 h after induction, indicating accelerated product loss and the need for cultivation harvest.

Mimic of a large-scale diafiltration process by using ultra scale-down rotating disc filter

Ultra scale-down (USD) approach is a powerful tool to predict large-scale process performance by using very small amounts of material. In this article, we present a method to mimic flux and transmission performance in a labscale crossflow operation by an USD rotating disc filter (RDF). The Pellicon 2 labscale system used for evaluation of the mimic can readily be related to small pilot and industrial scale. Adopted from the pulsed sample injection technique by Ghosh and Cui (J Membr Sci. 2000;175:5-84), the RDF has been modified by building in inserts to allow the flexibility of the chamber volume, so that only 1.5 mL of processing material is required for each diafiltration experiment. The reported method enjoys the simplicity of dead-end mode operation with accurate control of operation conditions that can mimic well the crossflow operation in large scale. Wall shear rate correlations have been established for both the labscale cassette and the USD device, and a mimic has been developed by operating both scales under conditions with equivalent averaged shear rates. The studies using E. coli lysate show that the flux vs. transmembrane pressure profile follows a first-order model, and the transmission of antibody fragment (Fab') is independent of transmembrane pressure. Predicted flux and transmission data agreed well with the experimental results of a labscale diafiltration where the cassette resistance was considered.

Evaluation of the effects and interactions of mixing and oxygen transfer on the production of Fab’ antibody fragments in Escherichia coli fermentation with gas blending

Fermentations carried out at 450-L and 20-L scale to produce Fab' antibody fragments indicated a serious problem to control levels of dissolved oxygen in the broth due to the large oxygen demand at high cell densities. Dissolved oxygen tension (DOT) dropped to zero during the induction phase and it was hypothesised that this could limit product formation due to inadequate oxygen supply. A gas blending system at 20-L scale was employed to address this problem and a factorial 2(2) experimental design was executed to evaluate independently the effects and interaction of two main engineering factors: agitation rate and DOT level (both related to mixing and oxygen transfer in the broth) on Fab' yields. By comparison to the non-gas blending system, results in the gas blending system at same scale showed an increase in the production of Fab' by 77% independent of the DOT level when using an agitation rate of 500 rpm level and by 50% at an agitation rate of 1,000 rpm with 30% DOT. Product localisation in the cell periplasm of >90% was obtained in all fermentations. Results obtained encourage further studies at 450-L scale initially, to evaluate the potential of gas blending for the industrial production of Fab' antibody fragments.

Extreme scale-down of expanded bed adsorption: Purification of an antibody fragment directly from recombinantE. coli culture

Scale-down is a methodology that combines the use of very small volumes of process fluid in dedicated devices to predict accurately the behaviour of process-scale biotechnological unit operations and for the production of comparable material for use in further devices which, taken together, facilitate the mimic of a complete full-scale process. This article provides the rationale behind the development of a small-scale mimic and demonstrates the use of a highly scaled-down expanded bed to predict hydrodynamic, kinetic, and adsorptive performance using less than 5-mL sample volumes. Data acquired on a specially developed 1.9 mm ID column was compared with that obtained in a standard 25 mm ID column. A homogenised E. coli system expressing an antibody fragment (F(ab)) adsorbed onto an rProtein A matrix was used to characterise the full adsorptive performance. Breakthrough curve studies using BSA in buffer were used to characterise binding kinetics. Performance at the two scales was comparable both in terms of expansion, axial dispersion, binding isotherms, and elution behaviour of the antibody fragment. The eluted F(ab) material was further purified by ion exchange chromatography to demonstrate the similarity between the profile of the product material obtained at both scales. The high level of scale-down (approximately 200-fold) provides for rapid process evaluation early in development, where material is at a premium and where a fast appreciation of the likely merits of one process strategy will lead to greater confidence in process selection and more robust flowsheets.