Scholarly article on topic 'Automated harvesting and 2-step purification of unclarified mammalian cell-culture broths containing antibodies'

Automated harvesting and 2-step purification of unclarified mammalian cell-culture broths containing antibodies Academic research paper on "Chemical sciences"

CC BY
0
0
Share paper
Academic journal
Journal of Chromatography A
OECD Field of science
Keywords
{Automation / Antibody / "Mammalian expression" / "Tandem chromatography" / "Bench-top instrument"}

Abstract of research paper on Chemical sciences, author of scientific article — Fabian Holenstein, Christer Eriksson, Ioana Erlandsson, Nils Norrman, Jill Simon, et al.

Abstract Therapeutic monoclonal antibodies represent one of the fastest growing segments in the pharmaceutical market. The growth of the segment has necessitated development of new efficient and cost saving platforms for the preparation and analysis of early candidates for faster and better antibody selection and characterization. We report on a new integrated platform for automated harvesting of whole unclarified cell-culture broths, followed by in-line tandem affinity-capture, pH neutralization and size-exclusion chromatography of recombinant antibodies expressed transiently in mammalian human embryonic kidney 293T-cells at the 1-L scale. The system consists of two bench-top chromatography instruments connected to a central unit with eight disposable filtration devices used for loading and filtering the cell cultures. The staggered parallel multi-step configuration of the system allows unattended processing of eight samples in less than 24h. The system was validated with a random panel of 45 whole-cell culture broths containing recombinant antibodies in the early profiling phase. The results showed that the overall performances of the preparative automated system were higher compared to the conventional downstream process including manual harvesting and purification. The mean recovery of purified material from the culture-broth was 66.7%, representing a 20% increase compared to that of the manual process. Moreover, the automated process reduced by 3-fold the amount of residual aggregates in the purified antibody fractions, indicating that the automated system allows the cost-efficient and timely preparation of antibodies in the 20–200mg range, and covers the requirements for early in vitro and in vivo profiling and formulation of these drug candidates.

Academic research paper on topic "Automated harvesting and 2-step purification of unclarified mammalian cell-culture broths containing antibodies"

ELSEVIER

Contents lists available at ScienceDirect

Journal of Chromatography A

journal homepage www.elsevier.com/locate/chroma

Automated harvesting and 2-step purification of unclarified mammalian cell-culture broths containing antibodies

Fabian Holensteina, Christer Erikssonb, Ioana Erlandssonb, Nils Norrmanb, Jill Simonb, Ake Danielssonb, Adriana Milicova, Patrick Schindlera, Jean-Marc Schlaeppia*

a Novartis Institutes for Biomedical Research, Biologies Center, Novartis Campus, CH-4056 Basel, Switzerland b GE Healthcare Life Sciences, Bjorkgatan 30, SE-751 84 Uppsala, Sweden

ARTICLE INFO ABSTRACT

Therapeutic monoclonal antibodies represent one of the fastest growing segments in the pharmaceutical market. The growth of the segment has necessitated development of new efficient and cost saving platforms for the preparation and analysis of early candidates for faster and better antibody selection and characterization. We report on a new integrated platform for automated harvesting of whole unclarified cell-culture broths, followed by in-line tandem affinity-capture, pH neutralization and size-exclusion chromatography of recombinant antibodies expressed transiently in mammalian human embryonic kidney 293T-cells at the 1-L scale. The system consists of two bench-top chromatography instruments connected to a central unit with eight disposable filtration devices used for loading and filtering the cell cultures. The staggered parallel multi-step configuration of the system allows unattended processing of eight samples in less than 24 h. The system was validated with a random panel of 45 whole-cell culture broths containing recombinant antibodies in the early profiling phase. The results showed that the overall performances of the preparative automated system were higher compared to the conventional downstream process including manual harvesting and purification. The mean recovery of purified material from the culture-broth was 66.7%, representing a 20% increase compared to that of the manual process. Moreover, the automated process reduced by 3-fold the amount of residual aggregates in the purified antibody fractions, indicating that the automated system allows the cost-efficient and timely preparation of antibodies in the 20-200 mg range, and covers the requirements for early in vitro and in vivo profiling and formulation of these drug candidates.

© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license

(http://creativecommons.org/licenses/by/4.0/).

CrossMark

Article history:

Received 25 May 2015

Received in revised form 8 September 2015

Accepted 13 September 2015

Available online 18 September 2015

Keywords:

Automation

Antibody

Mammalian expression Tandem chromatography Bench-top instrument

1. Introduction

Therapeutic monoclonal antibodies (mAbs) and antibody-related molecules such as immunoglobulin Fc-fusion proteins represent a fast growing class of therapeutics, with nearly 50 products approved or pending registration in the US and the EU [1]. The development of these new biologics requires effective discovery platforms including high throughput antibody production and characterization. Powerful display methods such as phage display, used in combination with large combinatorial antibody libraries allow the rapid generation of a large panel of molecules [2]. In this early selection phase, when dealing with hundreds to thousands of candidates, only a few |ig up to a couple of mg of antibodies prepared on robotic platforms are required to carry out the

* Corresponding author. Tel.: +41 61 324 9570/+41 61 696 3781. E-mail address: jm.schlaeppi@vtxmail.ch (J.-M. Schlaeppi).

necessary biological and biophysical analysis [3]. Major developments in high throughput (HT) technology for producing a large panel of recombinant proteins have been driven initially by structural genomics and proteomics initiatives [4-6]. Back to back fully automated robotic installations for mammalian expression and purification of thousands of recombinant secreted proteins at the sub-mg scale have been reported [7]. In parallel, important progress has been made in the development of novel micro-bioreactors for HT small-scale parallel expression to support process development [8], and has been validated with small-scale disposable reactors up to culture volumes of 250 mL [9]. The downstream processing has relied mainly on modified liquid-handling systems able to carry out affinity-capture purification of recombinant antibodies or tagged-proteins [10-12]. Later on during the generation of therapeutic antibodies, the selection focuses on the 20-50 remaining candidates having shown the required biological and developability properties. Their further profiling, including biophysical screens and formulation studies, yet also confirming their

http://dx.doi.org/10.1016/j.chroma.2015.09.040

0021-9673/© 2015 The Authors. Published by Elsevier B.V. This is an open access article underthe CC BY license (http://creativecommons.org/licenses/by/4.0/).

biological efficacy in rodents or other animals, will require larger amounts, usually in the 50-200 mg range. Whereas many of the analytical platforms used for profiling these candidates have been optimized towards automation and higher throughput [13], the upstream and downstream preparative steps to produce sufficient amounts of materials rely largely on low throughput manual or semi-automated processes only. Upstream, a cost-effective and fast way to produce these quantities is to express them as recombinant antibodies transiently in mammalian cells such as human embryonic kidney (HEK)-293 cells. This is done usually in 1-L roller-bottles or in 10-100 L Wave bioreactors for gram amounts [14]. Downstream, the purification is done mainly by affinity capture (AC) chromatography such as protein-A, followed by one or several polishing steps. To reduce bottlenecks in purification, several systems based on modified bench-top chromatography instruments have increased the throughput by integrating automated multiple chromatographic steps, using intermediate pool collection between each chromatography step [15-18]. Recent improvements have been achieved by using a direct flow-through system with on-line neutralization [19] or by combining up to four purification columns [20]. Another way to increase throughput, yet keeping the advantages offered by benchtop chromatography instruments such as UV monitoring, is to include a dedicated auto-sampler for unattended tandem chromatography on the AKTA™ platform. The system is optimally designed to accommodate clarified sample-volumes up to 50 mL [21]. In this report, we have dealt with our downstream bottlenecks often encountered in preparing enough material for extended profiling of therapeutic antibody candidates, by building a bench-scale platform, which integrates the harvesting step and the purification in the automated process. Up to eight 1-L unclarified HEK293 cell culture-broths (CB) containing recombinant antibodies are clarified in-line by disposable filters, before being purified by tandem purification, avoiding time-consuming manual harvesting steps. The whole process runs in 19 h including the complete cleaning in place (CIP) and column re-equilibration to avoid sample cross-contamination, allowing the purification of 40 cell-cultures per week with a minimum of attendance. The automated system results in an approximatively 4-fold increase in throughput compared to the conventional downstream process relying on manual steps, yet provides material of comparable quality. The system includes additional features such as air bubbles and constant pressure monitoring, in-line pH neutralization after the capture step and efficient aggregate removal by adding an in-line filtration step before the size-exclusion chromatography (SEC).

2. Materials and methods

2.1. Automated instrumentation

Automated harvesting, purification and CIP were performed on two bench-top chromatography instruments AKTA™ pure 25 M, equipped with two system pumps capable of a flow rate of 25 mL/min and with both the U9-M triple variable wavelength and the U9-L fixed wavelength UV monitors (GE Healthcare Life Sciences, Uppsala, Sweden). Each instrument held the AC and SEC columns for the two-step purification, and performed identical operations. The two instruments were connected to a central unit containing eight disposable filtration devices, the two sample pumps S9 and sample inlet valves. The whole system was controlled by the UNICORN™ software version 6.4 (GE Healthcare Life Sciences). The instruments operated at room temperature (RT), only the two fraction collectors F9-C were kept at 7 °C in a custom-made three-door cooling cabinet (Koch-Kaelte AG, Appenzell, Switzerland).

Unclarified feed 1-L bottle

Sample application 5 pm filter

Affinity capture MabSelect Sure

Neutralization

Aggregate removal 0.45 pm filter

SEC polishing Superdex 200 PG

Fractionation Peak collection

Fig. 1. Overview of the main steps of the automated process.

2.2. Automated unclarified sample application, tandem purification and cleaning-in-place - a process overview

A brief overview of the automated process is shown in Fig. 1. Operations were phased and staggered to process eight 1-L unclarified CB in 19 h. Up to four bottles were connected to the sample-inlet valve of the corresponding chromatography instrument. Two samples were loaded in parallel at a flow rate of 6 mL/min through individual single-use filtration devices ULTATM Disc GF 47 mm, 5.0 |im pore size (GE Healthcare Life Sciences), to remove cells and cell debris before being purified in-line using tandem chromatography. MAbs were captured by protein-A affinity chromatography on two interconnected 5-mL HiTrapTM MabSelect™ SuRe™ columns equilibrated with Dulbecco phosphate buffered saline pH 7.3 (dPBS). This affinity medium was chosen for its high mAb binding and specificity properties and its alkali tolerance for efficient CIP. We used two interconnected 5mL columns to ensure enough binding capacity. After loading, the HiTrap columns were washed with 15 column volumes (CV) ofdPBS followed by a one-step elution at 5 mL/min with 50 mM Na-OAc buffer (Merck, Darmstadt, Germany). When the A2g0nm exceeded a threshold value of 400 mAU, the in-line pH neutralization of the eluted protein fraction was triggered by applying a gradient of 320 mM Tris base (Sigma-Aldrich, Steinheim, Germany). In parallel, the A600nm was monitored for indication of light scattering due to the presence of aggregates. The samples were then passed through a filter Filtopur S, 0.45 | m at 5 mL/min (Sarstedt, Numbrecht, Germany) for potential aggregate removal and were

then applied to the SEC column HiLoad™ 26/600 Superdex™ 200 prep grade (320 mL column) equilibrated with dPBS.The Superdex medium was chosen for its fractionation range and low non-specific interactions. Collection of the mAb monomeric peak was triggered, when the UV signal exceeded a given threshold value. The automated process also included a CIP of both the AC and SEC columns to avoid any cross-contamination. The AC columns were cleaned with 0.5 M NaOH at 5 mL/min, for 15 min (using the sample pump), and then re-equilibrated with dPBS. The SEC columns were cleaned with a 60 mL pulse of 0.5 M NaOH at 4 mL/min (15 min contact time) using the buffer pump, followed by equilibration with dPBS until the pH monitored at the SEC column outlet was stable for 10 min.

2.3. Generation of recombinant mAbs by transient expression in HEK293 cells

The human and mouse mAbs were cloned into CMV-promoter-driven expression plasmids for mammalian cell expression. The mAbs (mostly lgG1) were subsequently produced by transient transfection in HEK293-T at a 1-L scale in roller bottles, essentially as described previously [14]. The cells in proprietary M11V3 medium were transfected with a DNA:PEl ratio of 1:3 (Polyscience, Warrington, USA). The cells were fed with the same volume of Ex-Cell VPro medium (Sigma-Aldrich, St.Louis, USA) and cultivated at 37 °C, 7.5 rpm, 5% CO2 for 7 days. Our standard 2-step purification was done by transferring the unclarified feeds into 500 mL centrifugation tubes (Corning lncorporated, Oneonta, USA) and centrifuged for 15 min at 4500 rpm followed by clarification through a 0.22 |im sterile filter (Millipore, Zug, Switzerland). The supernatants were purified on ÄKTA explorer at 7 °C, on HiTrap columns as described in Section 2.2. The mAbs were eluted with eight CV of 50 mM Na-citrate, 90 mM NaCl, pH 3.2. Fractions were collected, neutralized to pH 7.0 and concentrated to a volume of 5 mL before being further purified on a 120-mL HiLoad 16/600 Superdex™ 200 pg column equilibrated with dPBS. Protein concentration of the purified antibody samples was measured by recording the absorbance at 280 nm using an UV/vis spectral-photometer Nanodrop 1000 (Thermo Scientific, Switzerland) and using the mAb specific extinction coefficient. The final recoveries were calculated as the percentage of mAb measured in the purified fractions by Nanodrop to that measured in the unclarified CB by Protein-A-HPLC (see Section 2.4).

2.4. Titer determination by Protein A-HPLC - in process control (IPC)

Concentration determinations of antibody-containing super-natants were performed on a 1260 Infinity HPLC system (Agilent Technologies, Basel, Switzerland) holding an HPLC cartridge packed with 62.5 |L of MabSelect Sure material (GE Healthcare Life Sciences). Running buffers were 50 mM H3BO3, 200 mM Na2SO4 pH 7.5 (buffer A) and pH 2.5 (buffer B). The flow rate was 0.75 mL/min and 0.1 mL of supernatant was injected into the protein A column and rinsed with buffer A for 5.5 min. The bound antibody was eluted with buffer B for 2.5 min. The 215 nm UV trace was monitored and the elution peak area was used for quantita-tion.

2.5. Analytics of the purified materials

2.5.1. Aggregation determination by SEC

The aggregation level of purified mAbs was measured by SEC on an Agilent 1290 Infinity system equipped with a Superdex 200 10/300 GL column (GE Healthcare Life Sciences). The sample (50 | L) was loaded on the column equilibrated with dPBS at pH 7.3. The flow rate was 0.5 mL/min and the protein absorbance

was measured at 230 nm. The percentage of aggregates was calculated from the peak area at different retention times. Peaks which did not exhibit baseline separation were resolved by constructing lines at the point of peak intersection orthogonal to the absorbance baseline. Fractional concentrations were calculated by dividing individual peak areas by the sum of peak areas. The column was calibrated using an internal IgG standard previously validated to be 98% monomeric on a comparable SEC column coupled to a multi-angle light scattering detector [13].

2.5.2. Sodium dodecyl sulfate polyacrylamide gel-electrophoresis (SDS-PAGE)

SDS-PAGE was performed with 4-20% Mini-Protean® TGX™ Gels and Precision Plus Dual Color standards (Bio-Rad, Hercules, USA). Two |ig of sample was loaded on the gel and run in Tris/Glycine/SDS buffer according to the manufacturer's instructions.

2.5.3. Mass determination by mass spectrometry

Mass determination by mass spectrometry was done by LC-MS using a Waters Acquity UPLC system coupled to a Waters Q-TOF Premier Mass Spectrometer. Around 5 |g of antibody was injected at a flow rate of 0.1 mL/min onto a MassPrep micro-desalting cartridge (Waters, 186002785) heated at 80 °C, and eluted with a 12-min gradient (2-90%) of water/acetonitrile containing 0.1% formic acid.

2.5.4. Endotoxin detection

Endotoxin levels were measured in purified mAb samples by the Limulus Amebocyte Lysate (LAL) assay, using Endosafe disposable test cartridges and the PTS™ Portable Test System (Charles River Laboratories, France). The cartridges were adjusted to RT and samples were diluted to a minimum concentration of 0.01 mg/mL and were added into the cartridges followed by spectrophotometric measurement.

3. Results and discussion

3.1. instrumental platform

A picture of the complete system and the flowchart depicting the overall automated process are shown in the supporting information Figs. S1 and S2, respectively. The process consists of a number of generic protein preparation steps, which were integrated and optimized with the goal of obtaining endotoxin-free monomeric IgG (>95%) with recoveries at least comparable to those of our standard downstream procedures, and allowing us to prepare unattended, eight different 1-L cell-culture bottles in less than 24 h. The two, connected bench-top chromatography systems were identically equipped and configured, to perform identical yet in staggered phased operations. Optimization of the expression conditions, size and type of the filtration and chromatographic media, flow-rate, elution and pH neutralization buffers were performed as part of the system set-up. Below, the critical aspects of the system configuration are highlighted (flow paths and connected parts are illustrated for only one of the two instruments).

3.1.1. Sample application

Each of four 1-L cell culture bottles, containing different mAbs expressed transiently in HEK293-T cells, were connected to an inlet valve containing an integrated air sensor to control a safe and complete sample loading. Feedback from the air sensor ended the sample application, when completed. The sample pump, located downstream of the inlet valve, moved the sample through a second valve with an integrated pressure sensor (pre-filter) and through one of four 5 | m nominal pore size single-use filters and back

Fig. 2. Dual-flow valves for parallel loading and eluting the affinity-capture (AC)-units. (A) Parallel mAb adsorption on the AC-unit #1 and mAb elution (from another sample) on AC-unit #2. (B) By-passing AC-unit #1 and mAb adsorption on AC-unit #2. The different active flow paths are designated with broken and solid lines and arrows, respectively. The grey flowpath is inactive. The filled circles represent valves. Only the active valve ports are indicated.

through the same valve and a post-filter pressure sensor. Delta pressure was registered to track filter status and detects potential clogging. The sample was then applied to the protein-A column. The sample application part of the system's flow path, as described above, was driven by the sample pump.

3.1.2. Dual-flow for parallel affinity capture and elution

The two chromatography purification steps (AC and SEC) were performed in tandem with no intermediate peak storage. Both chromatography steps were thus operated at the same flow rate. A flow cell for A280nm monitoring was placed after each column outlet. Each chromatography system had two AC-units, consisting each of two interconnected 5-mL HiTrap columns to provide enough mAb binding capacity. Indeed, we have observed that loading the column below its maximum binding capacity prevents potential aggregation of the antibodies (Holenstein unpublished observations). A key to high productivity was that different operations

were carried out simultaneously on the two AC-units. The liquid flow through one of the AC-units was driven by the sample pump while the flow through of the second AC-unit was driven by the buffer pump (Fig. S2). This enabled parallel mAb adsorption on one AC-unit and elution of another sample on the other (Fig. 2A). This configuration also enabled one AC-unit to be by-passed (e.g., during CIP of the SEC column) while the sample was loaded on the other AC-unit (Fig. 2B). Also, it enabled sample loading on one AC-unit while the other was being cleaned (not shown).

3.1.3. Neutralization and aggregate removal

In order to minimize mAb exposure to acidic pH, an in-line neutralization step was included after elution from the AC-unit. Both AC-units were connected to an outlet valve, and an alkaline neutralization buffer was introduced in the flow path immediately after the valve. To secure a smooth pH transition, the alkaline buffer was introduced by a separate pump as a gradient, and was added

Fig. 3. Overall timelines for processing eight unclarified cell-culture broths in 19 h.

Fig. 4. Result overview of one representative sample purification showing: (A) chromatogram of the automated in-line tandem affinity-capture (AC) and size-exclusion chromatography (SEC) as described in Section 2.2. Both absorbance at 280 nm (mAU, blue) and 600 nm (mAu, red) are shown. The recovery was 75.4%, and the purity was 99.9%; (B) AP monitoring (mPa) during the in-line filtration ofthe unclarified cell-culture both; (C)SDS-PAGEofthe collected antibody SEC fractions; lanes 1 and3, unreduced and reduced samples; lane 2, molecular weight markers. The mAb titer in the cell-culture broth was 107.2 mg/Land the cell density at time of harvest was 2.8 x 106 cells/mL.

into the system flow in an active mixer (Fig. S3). The neutralized mAb fraction then passed through a flow cell for simultaneous monitoring of A280nm (for protein quantitation) and A600nm (for indication of light scattering due to aggregation). An in-line filtration step (0.45 |im nominal pore size) was used to remove larger aggregates, if any, before the sample was applied to the SEC column. A separate filter was used for each mAb sample. The flow path through the 0.45 | m filter was opened when elution from the AC-column was started, and it was closed, thus by-passing the filter, (Fig. S3) when the UV peak signal was below a set threshold.

3.1.4. Phasing and programming

A single, optimized and automated purification cycle (including sample loading, filtration, AC, SEC, CIP and column equilibration) took 7.5 h. In order to be able to achieve 40 such cycles within a working week, using two chromatography instruments, it was not only essential to perform different operations in parallel on each chromatography instrument as described above, but also to phase the purifications. The staggered phasing scheme is shown in Fig. 3. To allow for flexibility regarding the number of samples, and to avoid unnecessary complexity with respect to programming, five different UNICORN methods were developed for performing

0 50 100 150 200 250

Antibody concentration (mg/L)

Fig. 5. Total capacity of the automated system. Antibody spiked at various concentrations was subjected to the automated tandem purification. The final recovery corresponds to the percentage of mAb measured in the purified fractions by Nano-drop (A280 nm) to that initially loaded on the system.

the purification of four samples. The methods were arranged in a method queue and were executed sequentially.

Fig. 6. Scatter plot representation and mean of the percentage of recovery of 45 feeds run by the automated process compared to 102 feeds run by the conventional manual process. (T-test analysis, p < 0.0001). Recovery is calculated as the percentage of mAb measured in the purified fractions by Nanodrop (A280nm)tothat measured in the unclarified cell-culture broths by in-process control (IPC) as described in Section 2.4.

2-fold, namely to a single 5-mL HiTrap column and a 120-mL SEC column, in addition to an optimization of the separation conditions, would be an alternative for samples with low titers.

3.2. Validation of the automated system

3.2.1. Performances and reliability

The reliability of the automated instrumentation was tested by repeated loading of unclarified CBs or spiked samples at different antibody concentrations. The intra- and inter-run variations were 1.6% (n = 23) and 2.1% (n = 19) respectively, with an average mAb recovery of 80%, indicative of good reproducibility of the automated tandem chromatography. A representative chromatogram of the automated harvest and purification of one unclarified sample is shown in Fig. 4A. The one-step elution is started after washing the HiTrap column and when the absorbance at 280 nm exceeds a threshold value of 400 mAU, the peak collection and in-line pH neutralization of the eluted protein fraction is triggered by applying a gradient of 320 mM Tris base. The sample is passed through the Filtopur filter for potential aggregate removal and is applied to the SEC column. In parallel, the A600nm is monitored for indication of light scattering from potential aggregates, which in the example is negligible. In parallel, the A600nm monitored the presence of potential aggregates, which is in our example is negligible. The AC peak collection ends when the A280 nm returns to 600 mAU. Collection of the purified mAb monomeric peak after SEC chromatography is triggered likewise, when the UV signal exceeded a given threshold value. Fig. 4B shows the pressure monitoring during the pre-capture filtration, in-line AC loading and filter washing triggered by the air sensors. The Delta pressure (AP) is monitored to track filter status, an increase in pressure indicating filter clogging. Fig. 4C shows the purity of the final purified mAb analyzed by SDS-PAGE. Transient expression in mammalian cells such as HEK293-T results after 7 days of cultivation in an average cell density of 3.0 x 106 cells/mL, with around 86% cell-viability. Under these conditions, little build-up of pressure during the cell-culture filtration was observed and did not interfere with the completion of the clarification, even with those antibodies having aggregation issues. Antibody titers obtained by transient expression ranged between 15 and 200 mg/L of cell-culture. Therefore, we selected the sizes of the AC and SEC columns according to these specifications. As shown in Fig. 5, the best recoveries were indeed, obtained with antibody concentrations between 25 and 300 mg/L. Samples with titers lower than 25 mg/L gave poor recoveries and were less suitable for the automated set-up. Reducing the column sizes by

3.2.2. Comparison between automated and manual downstream process

The performances of the automated harvesting and tandem purification were compared to those of our manual process including manual harvesting and 2-step purification. A random panel of 45 unclarified CBs was run on the system and the recoveries were compared to those of 102 antibodies prepared by the manual process. As shown in Fig. 6, the results indicated that the recoveries were significantly higher with the automated process compared to the manual one, the mean recoveries being 66.7% and 46.6% respectively (p < 0.0001). The quality of the purified antibodies was measured by various analytical methods, including SEC for aggregation level, SDS-PAGE for overall purity (Fig. 4C), LC-MS for integrity and identity, and LAL assay for endotoxin level. Overall the automated process resulted in samples with the same quality as measured by SDS-PAGE and LC-MS. The comparison by ESI-TOF mass spectrometry of a representative antibody purified by conventional and automated methods is shown in the supporting Figs. S4-S9. Fully comparable mass spectra were measured. Glycoforms represented the main source of microheterogeneity. The average residual content of aggregates was reduced in the automated process (0.7%) compared to the manual one (2.3%). During our manual process, some problematic antibodies tend to precipitate upon pH neutralization of the protein-A fractions resulting in low recoveries. The standardized, gentle and fast in-line pH neutralization seems to reduce these losses. Indeed, by comparing the recoveries of eight culture-broths split between both the manual and the automated processes, we found the same trend towards higher recoveries with the automated process (Table 1). The comprehensive end-product quality control analysis showed that the quality of the material prepared by the 2-step automated purification was as comparable to that of the material purified by conventional manual harvesting and purification (Table 1). Overall, by increasing the throughput without reducing the quality of the final products, the automated process is a valuable approach to improve the productivity of the antibody-selection process. Noteworthy was the low endotoxin level measured in the final purified samples, despite running the automated process at RT, in contrast to the manual process run entirely at cold temperature. Indeed, all values remained well below the threshold of 5 EU/mg of mAb, required for most preclin-ical animal applications [22].

Table 1

Comparison between automated and manual process of the quality of purified mAbs.

Automated process Manual process

Recoverya [%] Aggregates [%] Monomeric [%] Endotoxin [EU/mg] Recoverya [%] Aggregates [%] Monomeric [%] Endotoxin [EU/mg]

mAb 1 55.4 0.3 99.7 1.4 53.3 1.1 98.9 0.6

mAb 2 66 0.3 99.3 <0.8 61.6 1.1 98.9 <0.5

mAb 3 80.6 0.4 99.6 <0.5 68.4 0.2 99.8 <0.3

mAb 4 59.3 0.1 99.9 nd 42.8 1.1 88.3 nd

mAb 5 37.0 0.1 99.9 nd 27.9 0.4 99.6 nd

mAb 6 71.1 0.3 99.7 <0.4 48.9 0.3 99.7 <0.3

mAb 7 67.6 0.1 91.8 <0.4 40.8 0.1 91.2 <0.3

mAb 8 85.3 0.9 99.1 <0.7 63.5 1.7 98.3 <0.5

nd: not determined.

a The final recovery represents the percentage of mAb measured in the purified fractions by Nanodrop (A28onm) to that measured in the unclarified CB by IPC as described in Section 2.4.

4. Conclusion

We have set-up an automated system for in-line harvesting and purification of 1-L antibody containing cell-culture broths to cover our needs for the efficient preparation of therapeutic antibody candidates in the early in vitro and in vivo profiling phase. By integrating the time-consuming harvesting step in the automated process, we could reduce our downstream bottlenecks that we experienced, when covering many requests for antibodies in the 50-200 mg range. Indeed, to remain cost effective, we have concentrated our efforts on the downstream part, leaving limited manual steps for the upstream transient transfection and cell cultivation. A back to back fully automated solution at this volume scale would imply a very complex and space-intensive robotic installation, probably necessitating specially trained operators. The flexibility of our system made of bench-top instruments equipped with standard components and software should be operational for all kinds of secreted recombinant proteins purified by any capture step, such as Fc-proteins or else. Furthermore, we conclude that the automated process eliminates subjective user decisions, minimizing waste and potential contamination due to intermediate pooling and fraction collection, and thereby increases process yields and reproducibility.

Acknowledgments

We thank the Novartis Biologics Center team in Basel, who helped to set-up the platform and provided various reagents, especially B. Kerins, A. Lavoisier, M. Coulot, S. Popp, N. Lageyre, J. Koelln, and S. Irigaray. We thank A. Tschupp for providing his expertise in automation, M. Glaettli, K. Stein, and L. Erni from GE Healthcare for helpful discussions. We thank T. Pietzonka and J. Hastewell for support throughout the project.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.chroma.2015.09. 040.

References

[1] J.M. Reichert, Antibodies to watch in 2015, MAbs 7 (2015) 1-8.

[2] D. Ponsel, J. Neugebauer, K. Ladetzki-Baehs, K. Tissot, High affinity, devel-opability and functional size: the holy grail of combinatorial antibody library generation, Molecules 16(2011)3675-3700.

[3] L. Turunen, K. Takkinen, H. Soderlund, T. Pulli, Automated panning and screening procedure on microplates for antibody generation from phage display libraries, J. Biomol. Screen. 14 (2009) 282-293.

[4] S.C. Almo, S.J. Garforth, B.S. Hillerich, J.D. Love, R.D. Seidel, S.K. Burley, Protein production from the structural genomics perspective: achievements and future needs, Curr. Opin. Struct. Biol. 23 (2013) 335-344.

[5] M.A. Elsliger, A.M. Deacon, A. Godzik, S.A. Lesley, J. Wooley, K. Wüthrich, I.A. Wilson, The JCSG high-throughput structural biology pipeline, Acta Crystallogr. 1 (2010) 1137-1142.

[6] S.A. Lesley, Parallel methods forexpression and purification, Methods Enzymol. 463 (2009) 767-785.

[7] R. Gonzalez, L.L. Jennings, M. Knuth, A.P. Orth, H.E. Klock, W. Ou, J. Feuerhelm, M.V. Hull, E. Koesema, Y. Wang, J. Zhang, C. Wu, C.Y. Cho, A.I. Su, S. Batalov, H. Chen, K. Johnson, B. Laffitte, D.G. Nguyen, E.Y. Snyder, P.G. Schultz, J.L. Harris, S.A. Lesley, Screening the mammalian extracellular proteome for regulators of embryonic human stem cell pluripotency, Proc. Natl. Acad. Sci. U.S.A. 107 (2010)3552-3557.

[8] Q. Long, X. Liu, Y. Yang, L. Li, L. Harvey, B. McNeil, Z. Bai, The development and application of high throughput cultivation technology in bioprocess development, J. Biotechnol. 192(2014)323-338.

[9] R. Bareither, N. Bargh, R. Oakeshott, K Watts, D. Pollard, Automated disposable small scale reactor for high throughput bioprocess development: a proof of concept study, Biotechnol. Bioeng. 110(2013)3126-3138.

[10] J. Steen, M. Uhlen, S. Hober, J. Ottosson, High-throughput protein purification using an automated set-up for high-yield affinity chromatography, Protein Expr. Purif. 46 (2006) 173-178.

[11] G.C. Barnard, M.D. Hougland, Y. Rajendra, High-throughput mAb expression and purification platform based on transient CHO, Biotechnol. Prog. 31 (2015) 239-247.

[12] S.C. Wiesler, R.O. Weinzierl, Robotic high-throughput purification of affinity-tagged recombinant proteins, Methods Mol. Biol. 1286 (2015) 97-106.

[13] A. Lavoisier, J.M. Schlaeppi, Early developability screen of therapeutic antibody candidates using Taylor dispersion analysis and UV area imaging detection, MAbs 7 (2015) 77-83.

[14] S. Geisse, C. Fux, Recombinant protein production by transient gene transfer into Mammalian cells, Methods Enzymol. 463 (2009) 223-238.

[15] J.A. Sigrell, P. Eklund, M. Galin, L. Hedkvist, P. Liljedahl, C.M.Johansson,T. Pless, K. Torstenson, Automated multi-dimensional purification of tagged proteins, J. Struct. Funct. Genom. 4 (2003) 109-114.

[16] Y. Kim, G. Babnigg, R.Jedrzejczak, W.H. Eschenfeldt, H. Li, N. Maltseva, C. Hatzos-Skintges, M. Gu, M. Makowska-Grzyska, R. Wu, H. An, G. Chhor, A. Joachimiak, High-throughput protein purification and quality assessment for crystallization, Methods 55 (2011) 12-28.

[17] R. Bhikhabhai, A. Sjöberg, L. Hedkvist, M. Galin, P. Liljedahl, T. Frigärd, N. Pet-tersson, M. Nilsson, J.A. Sigrell-Simon, C. Markeland-Johansson, Production of milligram quantities of affinity tagged-proteins using automated multistep chromatographic purification, J. Chromatogr. A 1080 (2005) 83-92.

[18] J.M. Schlaeppi, M. Henke, M. Mahnke, S. Hartmann, R. Schmitz, Y. Pouliquen, B. Kerins, E. Weber, F. Kolbinger H.P. Kocher, A semi-automated large-scale process for the production of recombinant tagged proteins in the Baculovirus expression system, Protein Expr. Purif. 50 (2006) 185-195.

[19] M. Shamashkin, R. Godavarti, T. Iskra, J. Coffman, A tandem laboratory scale protein purification process using Protein A affinity and anion exchange chro-matography operated in a weak partitioning mode, Biotechnol. Bioeng. 110 (2013)2655-2663.

[20] U. Riek, S. Ramirez, T. Wallimann, U. Schlattner, A versatile multidimensional protein purification system with full internet remote control based on a standard HPLC system, Biotechniques 46 (2009) ix-xii.

[21] D. Yoo, J. Provchy, C. Park, C. Schulz, K. Walker, Automated high-throughput protein purification using an ÄKTA purifier and a CETAC autosampler, J. Chromatogr. A 1344 (2014) 23-30.

[22] P. Malyala, M. Singh, Endotoxin limits in formulations for preclinical research, J. Pharm. Sci. 97 (2008) 2041-2044.