Purification of mAb using HiScreen Fibro™ PrismA and HiPrep™ 26/10 desalting columns

In process development, many samples are produced due to running multiple purifications. Automation of sample preparation and analysis reduce workload and time, minimize errors, and increase productivity.

This study demonstrates an automated two-step tandem purification protocol to prepare samples from mAb feed for analysis. It was performed using ÄKTA avant™ 150 system connected to a Teledyne ASX-560 autosampler.

Samples from a fed-batch culture expressing mAb (days 0-14) were purified on a HiScreen Fibro™ PrismA unit and the eluted mAb peak was automatically transferred to two HiPrep™ 26/10 Desalting columns containing Sephadex™ G-25 resin. The two-step purification, the initial capture, and the following buffer exchange showed high recovery and reproducibility.

The analyses of selected CQA’s i.e., mAb charge variants and size distribution (mAb monomer, aggregates, and fragments) clearly showed consistency with the feed during the cultivation process. Additional analyses with LC-MS revealed detailed information on intact mass, subunits level, and glycan profile.

Introduction

During bioprocess development, depending on study purpose and use, the initial sample volume and obtained target fraction size varies. There is a need to adapt system configurations, column sizes, and formats. This study was done to test rapid purification cycles that enable high throughput of samples for volumes ranging between 10 mL to 250 mL.

In the present study, automation for higher throughput and a more efficient workflow is shown using an automated two-step purification process on an ÄKTA avant 150 chromatography system with an attached autosampler.

The first purification step included capture of mAb feed samples on a HiScreen Fibro PrismA unit binding the target molecule on an alkali-stable protein A ligand. In the second step, rapid buffer exchange performed in tandem, the eluate from the capture step was automatically loaded onto two HiPrep 26/10 desalting columns, coupled in series.

The effective two step protocol allowed for production of target material sufficient for the subsequent analytical characterization of selected critical quality attributes (CQA) (Fig 1).

Fig 1. Schematic illustration of the automated two-step mAb purification.

Results

An ÄKTA avant 150 system was connected to a Teledyne ASX-560 autosampler to purify 13 mAb samples from days 2-14 of a fed-batch culture. The study demonstrates an automated two-step tandem purification protocol (a) to purify mAb feed samples, where the captured peak from the HiScreen Fibro PrismA unit was automatically transferred to two HiPrep 26/10 desalting columns containing Sephadex G-25 resin. In the process, high recovery between the initial capture and the following buffer exchange step was obtained. Purification of three replicates (b) of a different mAb sample indicated a reproducible setup of the automated two-step purification protocol. The analysis of some of selected CQA’s, mAb charge variants (c) and the distribution of mAb monomer, aggregates, and fragments (d) clearly shows productivity changes and consistency in the target material as a result of the cultivation process.

    a. Automated two-step tandem purification of mAb

In the automated two-step purification, the target mAb was eluted from the HiScreen Fibro PrismA unit and directly loaded onto the two HiPrep 26/10 desalting columns (two columns coupled in series). The system setup and mode of operation using “flush out” enabled a more extensive cleaning of the sample probe and tubing using multiple buffers and solutions. This mode of operation also removes potential air trapped in the probe and autosampler tubing.

The maximum recommended sample volume to be loaded on desalting columns is 30% of the total bed volume. The loaded sample volume i.e., the peak volume from the Fibro elution (amount of target from the cultivation) increased for the samples purified from day 2 to 14.

In the developed purification protocol care was taken to minimize the risk of sample loss due to overloading the desalting column (Fig 2). In practice this is achieved by a watch function which initially bypasses the desalting columns. This is indicated by a pH level and conductivity level drop in the chromatogram. Elution and the final peak fractionation from the desalting columns were based on the second UV in the system setup. At the end of the chromatogram, as shown in Figure 2, dips and peaks in conductivity and pH appear due to the buffer exchange.

Fig 2. Automated two-step tandem purification of mAb (day 7) using a HiScreen Fibro PrismA unit (blue curve, UV280) and two HiPrep 26/10 desalting columns (green curve, second UV280). The conductivity curve is depicted in orange and the pH curve in black. Section 1: Fibro elution phase, desalting columns in tandem. Section 2: End of Fibro elution phase, desalting columns in bypass. Section 3: Elution from desalting columns. Section 4: Shift in conductivity and pH due to the buffer exchange.

Figure 3 shows the increase in titer of the harvest from day 2 to 14. As expected, the increase in titer was linear in the beginning and levelled out at the end of the cultivation. The volume load on the Fibro unit was the same for each sample. Table 1 shows more detailed data from the integration performed on the chromatograms and similarity in the obtained peak areas for the Fibro and desalting results indicating a minimal sample loss.

Fig 3. Amount of purified mAb from each sample taken daily from the fed-batch culture.

Table 1. Peak table for the purified mAb samples from a fed-batch culture. The data shows the material balance between the two consequent steps in the purification protocol. Samples from days 2-14 of a fed-batch cultivation process were purified on an ÄKTA avant 150 system on a HiScreen Fibro PrismA unit connected in tandem with two HiPrep 26/10 desalting columns. UV signals are within the linear range of the system.

    b. Reproducibility

To verify reproducibility of the developed purification protocol and system setup, three identical mAb samples were purified (Fig 4). The analyzed data show good reproducibility between the performed experiments as indicated in Table 2.

Fig 4. Replication of the automated two-step tandem purification of mAb using HiScreen Fibro PrismA (blue curves, UV280) and HiPrep 26/10 desalting (green curves, second UV280). The conductivity curves are displayed in orange. Triplicate runs are displayed (curves overlapping).

Table 2. Recovery of two-step tandem purification for mAb samples. An average of three runs are reported.

    c. Charge variants profile

In Figure 5 the relative abundance distribution of mAb charge variants over the cell culture process is displayed. The data shows a shift in profile from main peak to acidic variants around day 9 in the cultivation process.

Fig 5. Charge variants profile over the fed-batch culture process for a mAb after capture on a HiScreen Fibro PrismA unit and buffer exchange on two HiPrep 26/10 Desalting columns. Cation exchange chromatography (CIEX) was used for analysis of acidic variants, main peak, and basic variants.

    d. Aggregate and fragment profile

In Figure 6, the relative distribution of mAb monomer, aggregates and fragments over the fed-batch culture process is displayed. There is a slight increase of aggregates and fragments over the culture period, being minor (up to ~ 1.4%), but the small change indicates a stable production process.

Fig 6. Relative abundance of monomer (main), aggregates (high molecular weight, HMW), and fragments (low molecular weight, LMW) over the fed-batch culture process for a mAb after capture with a HiScreen Fibro PrismA unit and buffer exchange on two HiPrep 26/10 desalting columns. Analysis was conducted using size exclusion chromatography (SEC).

nanoDSF analysis

The protein stability was determined by comparing unfolding onset and melting temperatures. No differences in protein stability were found, which indicate a stable production process (Fig 7).

Fig 7. Schematic picture of a melting curve and determined temperatures for unfolding onset and melting for each purified mAb8 sample from day 2- day 14.

LC-MS of intact mAb samples

As presented in Figure 8, expected MW of full-length mAb8 was determined (148186 Da, G0F/G0F) along with small amounts of free light-chain (23439 Da, 23558 Da [+Cys], and 23744 Da [+GSH]), light-chain dimer (46876 Da), and half-ab (74093 Da). The UV signal at 210 nm shows the small quantitative increase in light-chain species across the cultivation process. Glycoforms were accurately determined for the full-ab with G0F/G0F and G0F/G1F being the two predominant species (Figure 9).

Fig 8. Analysis of intact MW using LC-MS with UV signal at 210 nm (left) and total ion current (TIC) signal (right). The data shows samples from D3, D5, D7, D10, and D14.

Fig 9. Deconvoluted mass spectra for the main peak (full-length mAb8) in Figure 8. The data shows D2 and D14.

LC-MS analysis of subunits

The mAb samples were digested with IdeS and reduced with DTT, resulting in FC/2 (25200 Da, G1F), light-chain (LC, 23439 Da), and Fd’ fragments (25476 Da, Figure 10). Several modifications were detected, i.e. deamidations (+1/+2 Da) and something of +115/116 Da occurring on all three subunits. The modified variants increased only slightly throughout the cultivation, which indicates a stable production.

Fig 10. Analysis of subunits using LC-MS with UV signal at 210 nm (left) and total ion current (TIC) signal (right). The data shows samples from D3, D5, D7, D10, and D14.

LC-MS analysis of released N-glycans

The dominating N-glycosylation species for IgG1 type mAbs expressed in CHO cells are G0F (glycan F(6)A2) and G1F (glycan F(6)A2G(4)1) as depicted in Figure 11. The glycan profile gradually shifted with the cultivation process, where G0F was ca 50% and G1F was ca 40%. The remaining ca 10% was assigned to G2F (F(6)A2G(4)2), G0 (A2), and other glycans (e.g. sialylated glycans).

Fig 11. Released N-glycan profiles of mAb samples over the culture period.

II. Conclusions

In this study, two-step purification runs in high-throughput tandem mode was performed with the serial setup of an ÄKTA avant 150 system and an autosampler. The setup included the use of HiScreen Fibro PrismA unit and two HiPrep 26/10 desalting columns operating in tandem mode to automatically purify multiple mAb feed samples. The desalting step enabled for subsequent delivery of purified sample in an optimal buffer for use in a diverse range of analysis methods. Development of the protocol showed that adjustments to the sample volume loaded on the desalting columns may be needed to not exceed the recommended specifications. Essentially this could be done, as in this study, by including an instruction to turn the desalting columns out of flow at the end of the elution from the capture column. The current setup automated purification of large number of samples and sample volumes in the range of 10 mL to 250 mL, which is of great use in applications such as upstream process development to optimize culture conditions.

Offline analyses of critical quality attributes supported a stable cultivation process with successful daily sampling and purification. The analyses of mAb charge variants and size distribution (mAb monomer, aggregates, and fragments) showed consistency with the feed during the cultivation process. Additional analyses with LC-MS revealed detailed information on the presence of light-chain, light chain dimer, and half-ab and that deamidation species (= acidic variants) increasing with the progress of the cultivation process. The glycan profile was as expected from released N-glycan analysis and there were no differences in mAb protein stability as measured by nanoDSF.

Material and Methods

Fed-batch culture

A Chinese hamster ovary (CHO) cell line expressing the IgG monoclonal antibody (Cytiva, Uppsala) was cultured in an XDR-50 bioreactor in ActiPro™ medium. Starting at day 3, the reactor was fed with Cell Boost™ 7a Supplement and Cell Boost 7b Supplement once daily. The daily feed volumes, related to the initial working volume, were 2.5% and 0.25%, respectively. We added glucose to a final concentration of 4 to 6 g/L as levels decreased below 2 g/L at sampling. The culture was harvested at a viability of 92% on day 14. The culture was sampled daily from day 2 to 14, and cells were removed by centrifugation and filtration. Collected samples were stored at -20 °C during the culture time and transferred to -80 °C after completion of the culture run. IgG titer and metabolites were monitored by Cedex Bio.

Automated two-step purification

An automated two-step tandem purification of mAb feed samples was set up on an ÄKTA avant 150 system (Fig 12). The system was also equipped with three versatile valves (V9H-V), one external air sensor (L9-1.5) and an LED-UV (280 nm, U9-L). Two of the versatile valves and the external air sensor were setup as in the flush out configuration (wash of the sample probe in the autosampler) and used during sample application using the Teledyne ASX-560 autosampler (Teledyne CETAC Technologies Inc). The flush out configuration enabled the use of multiple cleaning solutions for a more extensive cleaning of the sample probe and to enter the appropriate buffer in the sample probe and tubing. In addition, this mode of operation also removes potential air trapped in the probe and tubing. The third versatile valve and the LED-UV were connected between the multiple wavelength UV (U9-M) and the conductivity monitor (U9-C) to direct the flow to or to bypass the desalting columns. The mixer (M9) was disconnected from the flowpath, as recommended for Fibro applications (1).

An estimate (UV280, offline measurements, data not shown, CEDEX Bio Analyzer, F. Hoffmann-La Roche Ltd) was made based on the titer of the harvest from day 14 to determine the amount of sample to load on the Fibro™ unit. The same volume was loaded for each sample. Each mAb sample was treated with diatomaceous earth (60 g/L) and filtered (0.2 μm) before loaded onto the Fibro unit (2) . The captured peak from the HiScreen Fibro PrismA unit was directed automatically for a buffer exchange to appropriate buffer conditions using two HiPrep 26/10 desalting columns. In the UNICORN™ method a watch function was programmed on the UV level (20 mAu), which sends a command to turn the third versatile valve directing the eluate from the Fibro unit directly to the two HiPrep 26/10 desalting columns connected in series. For the samples purified from day 2 no watch instruction was used in the Fibro elution phase, in this case a fixed volume load was used for the desalting columns. A schematic view and details of the UNICORN two-step method are shown in Figure 13.

Fig 12. ÄKTA avant 150 system configured for automatic two-step purification scheme. Additional components are three versatile valves (VV1, VV2 and VV3), one external air sensor (L9-1.5) and a LED-UV (U9-L). The Teledyne autosampler uses the ÄKTA™ sample pump for injections. When the first UV detects the eluting peak from the HiScreen Fibro PrismA unit a watch function initiates, turning the versatile valve (VV3) and directs the peak to the two HiPrep 26/10 Desalting columns. The flow will then pass the second UV (U9-L), the conductivity and pH monitor, and via the outlet valve to the built-in fraction collector. The autosampler was connected via the I/O box using a custom-made cable for matching the input and output digital signals. All digital output I/O box signals were set to 0 in UNICORN before starting the method. The ASX-560 autosampler is connected to a host computer using a USB interface. The autosampler is controlled using the AScript software.

    a. UNICORN method

Table 3 shows the entire protocol used for the two-step purification including the automated and flexible use of autosampler probe wash. Figure 13 shows the UNICORN method and details for programming the different steps in the purification protocol. To ensure that the entire peak eluted from the Fibro column was loaded onto the desalting columns a watch instruction in UNICORN, based on UV level, was used to turn the third versatile valve (VV3), and place the desalting columns in flow during the Fibro elution step. In the current study this watch instruction was used for samples purified from day 3 to 14. For the sample purified from day 2, instead a fixed volume was set to turn the third versatile valve and place the desalting columns in flow during the Fibro elution step. An alternative way, when only small peaks and hence small volumes, could be to change the UV start and stop levels for the watch instructions to ensure the specific peak volume to be loaded onto the desalting columns.

Table 3. Protocols for protein A purification, buffer exchange and autosampler wash.

*The volume required in each autosampler wash step might have to be adjusted based on the capillary lengths. No sample pump wash instructions are used during the cleaning of the autosampler flowpath, instead a sample flow is set in these phases. Sample pump flow rate is set to a value preventing flooding in the autosampler wash station.

Fig 13. UNICORN method and details for programming the sample application phase, Fibro elution phase with watch function for direct the load onto the second chromatography step, the autosampler control phases and wash phases.

     b. AScript method

AScript software (v1.5) was used to control digital inputs I/O box signals. All signals were set to 0 in the AScript software before starting the method. A description of how-to setup the autosampler is available in the Cue card. A description and details of the AScript method is shown in Figure 14.

Fig 14. The AScript method used for the Flush out configuration. The 50 mL sample rack was used. When starting the AScript method the number of samples was specified in order to loop the script for each sample. An equal number of samples was also set in the scouting scheme for the UNICORN method.

mAb concentration

mAb concentration was measured using NanoDrop One system with quantification performed at A280 nm using an extinction coefficient of 1.4 mL g-1 cm-1 (blank D0).

Charge variants profile

Analytical cation exchange chromatography (CIEX) was performed using a ProPac WCX-10 2x250 mm column on an Agilent 1290 UPLC system with OpenLab software. Approximately 40 μg mAb was injected onto the column. Mobile phase A was 10 mM Tris and 10 mM Phosphate at pH 6.5. Mobile phase B was 10 mM Tris and 10 mM Phosphate at pH 9.5. Separation was performed in a stepwise gradient from 0 to 100 % B in 21 min at 0.3 mL/min. In the data analysis the full peak was integrated and split right before and after the main peak to calculate the relative abundance of acidic variants, main peak, and basic variants.Aggregate and fragment profile

Analytical size exclusion chromatography (SEC) was performed using a Superdex™ 200 Increase 10/300 GL column on an Agilent 1290 UPLC system with OpenLab. Approximately 40 μg mAb was injected onto the column. The mobile phase was 200 mM phosphate at pH 6.8. Separation was performed in 30 min at 0.75 mL/min. In the data analysis the full peak was integrated and split right before and after the main peak to calculate the relative abundance of high-MW (aggregates), monomer, and low-MW (fragments).

Protein stability

The protein stability was determined using a Prometheus NT.48 (NanoTemper) nanoDSF system. Each mAb sample D0-D14 was loaded into nanoDSF sample capillaries in triplicates according to the manufacturer’s instructions. In the ThermControl software a melting scan was performed using the following settings: excitation power: 20%, start temperature: 20°C, end temperature: 95°C and temperature slope: 1.0°C/min. The data analysis included assessing melting curves, onset melting temperatures, unfold temperature (TM) and aggregate formation (scattering).

LC-MS analysis of intact mAb samples

Samples were diluted in MQ prior to analysis. In the analysis of intact mAb samples a BioAccord (Waters), Acquity I-Class UPLC System coupled to Acquity RDa detector was used. The sample was prepared for analysis by loading 1 µg of mAb on a BioResolve RP mAb Polyphenyl column (Waters), 450Å, 2.7 µm, 2.1x50mm. The mobile phases used for separation was A: 0.1% formic acid in milli-Q water, B: 0.1% formic acid in MeCN at a flow rate: 0.5 mL/min. Gradient elution was performed 0-60% B in 1 to 17 min and the column temperature was set to 80 °C. UV detection was performed at 210 and 280 nm and the MS settings were set to full scan, positive mode, high mass range (m/z 400-7000), cone voltage 70V, scan rate 2Hz, capillary voltage 1.5 kV and a desolvation temperature 550 °C. In the data analysis the UV and TIC were inspected. MW was determined manually for selected samples by extraction of mass spectra at specific retention times followed by deconvolution into neutral masses using MaxEnt1 algorithm. MW for all peaks in all samples were determined automatically by Intact Mass App.

Subunit analysis

In the subunit analysis samples were digested with FabRICATOR® MagIC (IdeS) and reduced with DTT prior to analysis. 2µg mAb digested sample was loaded on a BioResolve RP mAb Polyphenyl column (Waters), 450Å, 2.7 µm, 2.1x50mm). The mobile phases used for separation was A: 0.1% formic acid in ultrapure 18.2 MΩ^·cm water, B: 0.1% formic acid in MeCN at a flow rate: 0.3 mL/min. Gradient elution was performed 20-40% B in 1 to 10 min and the column temperature was set to 60 °C. In the MS analysis using a BioAccord (Waters), Acquity I-Class UPLC System coupled to Acquity RDa detector UV detection was performed at 280 nm and the MS was set to full scan, positive mode, high mass range (m/z 400-7000), cone voltage 70V, scan rate 2Hz, capillary voltage 1.5 kV and a desolvation temperature 550 °C. In the data analysis the UV and TIC were inspected. MW was determined manually for selected samples by extraction of mass spectra at specific retention times followed by deconvolution into neutral masses using MaxEnt1 algorithm. MW for all peaks in all samples were determined automatically by Intact Mass App.

Released N-glycans

Analysis of released N-glycans was performed loading 10µL of final SPE eluate (labelled and purified glycans from 15µg mAb) on a Acquity Premier Glycan BEH Amide column (Waters, 130Å, 1.7 µm, 2.1x150mm) on a Waters BioAccord Acquity Premier UPLC system coupled to Acquity RDa detector. The mobile phases used for separation was A: 50 mM ammonium formate, pH 4.4 and B: 100% MeCN at a flow rate of 0.4 mL/min. Gradient elution was perfomed 25-46% A in 0 to 35 min at a column temperature of 60 °C. Fluorescence detection was used at a extinction and emission wavelength of 265 and 425 nm respectively. The MS was set to full scan, positive mode, low mass range (m/z 50-2000), cone voltage 30V, scan rate 1Hz, capillary voltage 1.5 kV, desolvation temperature 400 °C. In the data analysis FLD and TIC were inspected. Retention times were adjusted in the processing method for the dextran calibration ladder to have correct peak integration/assignment. Glycans were identified using glycose unit (from the dextran calibration) and accurate mass match provided in the Waters’ glycan library. Glycans were quantified using the fluorescence signal.