Scaled up fed-batch production of recombinant alpha-1-antitrypsin by CHO cells in single-use surface aerated orbital shaken bioreactor
Wen Qin Tang, Chloe Qiu Zhen Jiang, Zi Ying Zheng, Ally Lau, Xuezhi Bi, Wei Zhang, Say Kong Ng
TL;DR
Researchers scaled up the production of alpha-1-antitrypsin using Chinese Hamster Ovary cells in a bioreactor to ensure a consistent supply for medical treatments.
Contribution
The study demonstrates scalable fed-batch production of A1AT in a single-use bioreactor with productivity comparable to traditional methods.
Findings
Cell-specific productivity of A1AT reached 9.6 and 12 pg/cell/day in bioreactor runs.
Production stability was evaluated over 12 weeks across ten CHO clones.
The bioreactor system offers potential for consistent and scalable A1AT production.
Abstract
Augmentation therapy is a treatment option available in the market that has been approved by the U.S. Food and Drug Administration (FDA) for alpha-1-antitrypsin (A1AT) deficient patients. The treatment requires weekly injections of purified A1AT for the patients and relies on plasma donor. The demand for A1AT is also high due to its functional role in various diseases. However, scaling up production of purified human plasma A1AT remained costly and challenging. It is therefore of great interest to generate A1AT at larger scale in ensuring a consistent supply to the market. In this paper, we evaluated the stability and productivity of ten Chinese Hamster Ovary (CHO) single cell clones over 12 weeks. This was followed by scaling up the fed-batch production of A1AT with the selected cell clone in a 10L single-use surface aerated orbital shaken bioreactor SB10-X. The cell specific…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7- —Agency for Science, Technology and Research (A*STAR), Singapore
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsViral Infectious Diseases and Gene Expression in Insects · Protease and Inhibitor Mechanisms · Proteoglycans and glycosaminoglycans research
Introduction
A1AT, which is also commonly known as SERPINA1, is a secreted protein with a molecular weight of 52 kDa^1,2^. The protease inhibitor is known to have a high affinity for neutrophil elastase which plays a critical role in defending against infectious agents, but also inevitably degrades elastin, collagen and proteoglycans. One important role of A1AT is to inactivate neutrophil elastase in the human lungs to protect the delicate alveolar tissues from proteolytic degradation during inflammatory responses^3^.
An imbalance in neutrophil elastase and A1AT would have a devastating impact on the lungs. In patients with severe A1AT deficiency, A1AT protein is present at low levels due to genetic defects, resulting in the development of emphysema. The current treatment for A1AT deficient patients involves augmentation therapy, where human plasma-derived A1AT is administered intravenously to increase the A1AT levels in the patients’ blood and lungs. This is the only available treatment that is approved by the FDA in the market currently. Apart from intravenous administration, initial studies have shown that aerosol application of A1AT could possibly help to protect the lungs too^2,4^.
Under augmentation therapy, dosage required to raise the A1AT levels above the required threshold of 11 µM in the serum is determined to be 60 mg/kg/week^5–8^. The limited supply of human plasma from donors and high frequency dosage amounts makes the treatment an expensive option for the patients^4,5,9,10^. Apart from the high demand^10^ and cost involved to generate purified human plasma A1AT, patients also run the risk of viral transmission^9,11^. Furthermore, upscale production of A1AT proves to be another limiting factor for the therapy^12^.
Other than utilising A1AT in augmentation therapy for A1AT deficient patients, there has been a growing number of evidence indicating that A1AT derived from human serum reduces pro-inflammatory cytokines, triggers anti-inflammatory cytokines, improves insulin release and inhibits the maturation of dendritic cells^13–15^. Given these properties, A1AT can also be potentially used in treating graft versus host disease (GVHD) in organ transplant as well as diabetes^8^. Other therapeutic indications included bacterial and viral infections^16^, arthritis^14,17–19^, acute myocardial infarction^14,20^, islet allograft survival and transplantation^14,21,22^ stroke and acute liver failure^14,23,24^. It is therefore of great interest to generate A1AT in larger scale due to its functional role in various diseases.
There have been reports on the production of recombinant A1AT at a small scale using various platforms^5,12,25^ including yeast, insect, bacteria and plant expression systems. Issues with these include the lack of post translational modification, low expression levels and slow growth of the expression system chosen^12,26^. Recombinant A1AT was also purified from the milk of transgenic sheep, though this faced the difficulty in removing trace amounts of sheep native A1AT and alpha-1-antichymotrypsin that could undesirably elicit an immunogenic response in humans^27^.
Production of recombinant A1AT in mammalian cell line has also been reported in a few studies, as summarised in Table 1. The production cell lines were generally cultured in small volumes only. To our knowledge, there has also not been any reports on the generation of rhA1AT by CHO cells in orbitally shaken bioreactor (OSB).Table 1. Summary of reported cell lines producing recombinant A1AT.Cell lineConcentration of A1AT at harvestCulture modeVessel typeSpecific productivity (pcd)SourceAGE1.HN ~ 0.2–0.4 g/LFed batch50 ml bioreactor filter tubesNot reported^28^AGE1.HN0.4 g/LBatch125 ml shake flasksNot reported^29^HEK293Not reportedBatch400 ml bioreactorNot reported^4^CHO1.15 g/LBatch125 ml shake flasks65^30^CHO-S0.117–0.123 g/LBatch250 ml shake flasks4–5.8 (early phase)2–3 (late phase)^9^CHO1.2 g/LFed batch, with temperature shiftPlate, scale not reportedNot reported^31^
The use of OSB can be a less complex and cost-effective alternative to stirred tank reactor (STR) in producing rhA1AT at larger scales. Reportedly, operation of larger scale single use OSB could be easier than STR as the mounting of the reactor with the integrated stirrer can be eliminated. Similar working principles among the shake flasks, tube spin bioreactors and OSB for gassing and mixing simplified the scaling up process too. This would be beneficial as the shake flasks and tube spin bioreactors are low-cost and straightforward options for small scale and high throughput production^32^.
Apart from being a cost effective alternative to STR, OSB could also be employed in a multi-product facility with faster turn-around time as it utilized single-use technology^33^. At a matching kLa, the OSB consumed less energy and had a faster mixing time. A lower specific power consumption rate in OSB would be ideal when CHO cells exhibited poor cell growth and viability in STR due to the high shearing forces^32^. Currently, there have been reported use of OSB in the production of influenza A virus, adeno-associated virus and biosimilar of a monoclonal antibody, up to a reactor scale of 12L^33–35^.
CHO cell line proves to be a popular option in the manufacture of recombinant protein therapeutics due to several advantages: (1) Regulations with regards to therapeutic application are clearly defined; (2) Usage and adaptation of cell line in serum free media are well established, thereby excluding any animal components in the production process. This reduces risk of viral contamination from animal sources. In addition, there are less lot-to-lot variations, achieving a more controlled process with consistent protein yields^36^; (3) Similar glycosylation profile is observed between CHO and human cell lines, thereby achieving the desired therapeutic efficacy and half-life in human body^37^, as well as being less likely to trigger an immune response in humans; (4) High yield could be generated from fed-batch cultures as CHO cells are able to achieve high density in suspension culture in large scale bioreactor^38^. As summarised in Table 1, CHO cell line was able to produce A1AT at approximately three-fold higher in concentration relative to human cell lines. (5) For mammalian cell lines, cell lysis and protein refolding is not required as most of the products are secreted, unlike bacteria and prokaryotes expression system^39^. Therefore, in consideration for all the points, it is of clinical and commercial interest to produce rhA1AT using CHO platform.
In our previous work, a few novel combinations of engineered motifs for further selection marker attenuation to enhance rhA1AT generation were evaluated^30^. In this work, the single cell clones were obtained from the rhA1AT producing cell pools. The stability of the clones was studied. A high producing and stable cell clone was subsequently selected and expanded for fed batch production in the 10L single use bioreactor. The product was then purified and tested for its activity.
Materials and methods
Cell line stability
Recombinant CHO cell line producing rhA1AT was previously generated from suspension CHO-DG44 cells (Gibco™ Catalog number 12609–012, Invitrogen, Carlsbad, CA)^30^. High producing cell pools were single cell cloned in HyQ PF-CHO (Hyclone, USA) media and amplified to 300 nM MTX. This was followed by the batch characterization of the cell clones in HyQ PF-CHO media with 300 nM MTX. Concurrently, cells were routinely passaged every 3 to 4 days in 125 mL disposable Erlenmeyer flasks (Corning, USA) in the commercial media over a span of 12 weeks. Sample supernatant was collected every week on the fourth day of passage for titre check via enzyme linked immunosorbent assay (ELISA). After which, the cell line was adapted to in house protein free chemically defined media.
Fed-batch characterization in shake flasks and production in SB10-X
The inoculum train was initiated from vial thaw and expanded in shake flasks with in-house protein free chemically defined medium. Two shake flasks were inoculated at a target cell density of 5 × 10^5^ cells/mL with an initial working volume of 100 mL. The shake flasks were maintained in an incubator (Kühner AG, Switzerland) at 8% CO_2_ and 37ᵒC with a shaking speed of 110 rpm.
As for SB10-X (Kühner AG, Switzerland), the runs were inoculated at 5 × 10^5^ cells/mL with an initial working volume of 8L in the same in-house protein free chemically defined medium. The amount of 1 M glucose solution and protein-free feed required was determined based on the specific consumption rates of glucose and glutamine. The dissolved oxygen level was maintained above 40% with a gas flow of 400-1000 ml/min via the overlay. The O_2_ composition in the gas flow was kept in the range of 16–26%. The pH was regulated at 7.1 ± 0.05, with CO_2_ flowing in through the overlay to reduce the pH. 7.8% sodium bicarbonate was used for upward control of pH. The shaking speed and temperature were maintained at 80 rpm and 37ᵒC respectively throughout the culture.
Cell culture sample analysis
Daily sampling from the shake flask and bioreactor was performed and samples were analysed immediately. Viable cell density (VCD) and viability were determined via Vi-CELL™ Cell Viability Analyzer (Beckman Coulter, USA). Other key aspects of cell culture tracked daily included pH, glutamine, glutamate, glucose, lactate, ammonium, sodium and potassium. These components were analysed with BioProfile®100 Plus Analyzer (NOVA Biomedical, USA), where the cells were first removed via centrifugation at 6010xg for 10 min in a micro-centrifuge (Eppendorf, Germany). The osmolality of the clarified supernatant was also measured offline using VAPRO® Vapor Pressure Osmometer (Wescor Inc., USA).
Quantification of commercial human A1AT
Lyophilized hA1AT (Abcam, United Kingdom, Cat. No. ab91136) were constituted to a concentration of 5 mg/ml. Verification of hA1AT concentration was performed to confirm that the reconstituted hA1AT was suitable as a standard for ELISA quantification. This ensured the accuracy of the protein titre results to be measured. This was necessary as it has been reported in other study that the protein standard provided in commercial kit may potentially result in inaccurate titre reading^40^.
The concentration of the reconstituted human A1AT was quantified using both Bicinchoninic acid assay (BCA) assay (Thermofisher Scientific, USA) and Nanodrop Microvolume Spectrophotometer (Thermofisher Scientific, USA). For the BCA assay, the absorbances and concentrations of albumin (Thermofisher Scientific, USA) were correlated linearly (refer to Supplementary Figure S1) to generate a standard curve. The absorbance of the reconstituted A1AT at 10 and 20 fold dilutions were then used to determine the concentration of the reconstituted A1AT using the standard curve. For Nanodrop Microvolume Spectrophotometer, 1uL of the sample was loaded to the instrument and it was analyzed at a wavelength of 280 nm.
ELISA quantification of rhA1AT concentration
Clarified supernatant from daily sampling were aliquoted and stored in -20ᵒC freezer. The titre was quantitated using Human alpha-1-antitrypsin ELISA Quantitation kit (GenWay Biotech, Inc., USA). Lyophilized hA1AT (Abcam, United Kingdom, Cat. No. ab91136) were reconstituted in deionised (DI) water and diluted as standards in the assay.
Upon the completion of the culture run, the samples from daily sampling were thawed out and diluted to a range of 3.1-25 ng/mL, before being loaded in duplicates onto the coated 96 well plate. Wells were then blocked with 1% bovine serum albumin (BSA) in TRIS buffer, followed by the incubation with the horseradish peroxidase conjugate. The absorbances of the wells were finally measured at 450 nm wavelength using colorimetric analysis. Readings obtained from the standards were subjected to a four degree polynomial curve fit.
pH stability
To characterize the A1AT stability at different pH, freshly harvested culture supernatant (with pH measured to be 7.0) was clarified and aliquoted. The pH of these clarified supernatant were then adjusted to 3, 3.5, 4, 4.5 and 5 and equilibrated at room temperature for 15 min. Samples were subsequently centrifuged at 3220xg for 10 min to remove any precipitates present and the supernatant were stored frozen till analysis. The A1AT concentrations of the samples for respective pH were then measured using ELISA.
Harvest & purification of A1AT
The harvest was centrifuged at 2095xg for 20 min. The pooled supernatant was then filtered through 0.22 µm filter before storing it in -80ᵒC freezer prior to purification.
Upon thawing of the stored supernatant, it is passed through a depth filter (Pall Corporation, USA) for clarification. The filtered solution is then purified via an anion exchange column (Pall Corporation, USA), followed by hydrophobic interaction (Pall Corporation, USA) to achieve a single peak chromatogram. The final purified rhA1AT product is then subjected to buffer exchange to the buffer solution of 30 mM HEPES with 300 mM NaCl via tangential flow filtration using 30 kDa filter (Sartorius AG, Germany).
Samples were diluted below 1 g/L and 0.22 µm filtered prior to loading in the high performance liquid chromatography (HPLC) instrument for the purity check. The mobile phase, which consisted of 0.2 M arginine, 50 mM MES, 5 mM EDTA and 0.05% (w/w) sodium azide at pH 6.5, had a flow rate of 0.6 mL/min for 40 min. The monomeric peak had a retention time of 13 to 14 min. High molecular weight species were defined as peaks that appeared before the main peak, whereas low molecular weight species appeared after the main peak.
rhA1AT activity analysis
rhA1AT activity was determined using the elastase inhibition assay^41^: The concentration of the purified samples was determined via the ELISA quantitation kit as used for titre analysis. The same commercial A1AT standard that was used for titre analysis was diluted in in house protein free media to the range of 0.2–1.6 ng/uL. 20-100 ng of samples was loaded in duplicates to a non-adhesion 96 well plate followed by the addition of 0.84 ng/uL of elastase (Merck, Germany, Cat. No.324682). 125ug of N-succinyl-ala-ala-pro-phe p-nitroanilide (SAPNA) (Sigma-Aldrich Corporation, USA, Cat. No. S4760) dissolved in N,N-Dimethylformamide (DMF) (Sigma-Aldrich Corporation, USA, Cat. No. D4551) was then added to the wells and the plate was incubated at 37ᵒC for 10 min. The absorbance of the well was measured at 410 nm wavelength at an interval of 5 min over 30 min. The gradient of the readings against time were then subjected to a natural logarithm curve fit with the corresponding A1AT amounts.
Verification of purified rhA1AT production and signal peptide cleavage by LC–MS/MS peptide mapping
5 µg purified rhA1AT protein were denatured and reduced in Laemmli buffer (62.5 mM Tris–HCl, pH 6.8, 25% glycerol, 2% SDS, 0.01% Bromophenol Blue, 25 mM DTT), at 95 °C for 10 min and separated on 12% Mini-PROTEAN® TGX™ precast gels (Bio-Rad), stained with 0.1% Coomassie blue R250. After destaining, rhA1AT bands was excised and in-gel digested with sequencing grade trypsin, pepsin and chymotrypsin (Promega) respectively according to manufacturer’s instruction, peptide extraction, concentration as described previously^42^.
LC–MS/MS analysis was performed using a nanoACQUITY UPLC (Waters, Milford, USA) coupled to LTQ Orbitrap Elite ETD Mass Spectrometer (ThermoFisher Scientific, Waltham, MA, USA) according to previously described^42^. Spectra were obtained by data-dependant acquisition tandem mass spectrometry in which one full MS scans at 120,000 resolution from 350 to 1600 m/z, was followed by HCD Orbitrap tandem MS scans of the 12 most intense peptide ions with normalized collision energy of 30% at a resolution of 15,000 over a 90 min gradient from 5 to 35% Acetonitrile-0.1% Formic acid.
Chymotrypsin, pepsin and trypsin digest peptide tandem mass spectrometry data were combined for peptide mapping and N-terminal sequence analysis using PEAKS studio X plus software (Bioinformatics Solutions Inc.) de novo sequencing and database searching against human uniprot database (555,426 entries, downloaded in 201,709), semispecific digestion and two missed cleavages were allowed, carbamidomethylation of cysteine was included as a fixed modification, oxidation of methionine, N/Q deamidation were considered as variable modification. Peptide and fragment ion mass tolerances used were ± 10 ppm and ± 0.5 Da, respectively. The results were filtered with false detection rate (FDR) of 0.5%.
Results
Production stability of single cell clones
Ten single cell clones were obtained from A1AT producing pools^30^. The characterization of the clones was performed as a batch culture with 300 nM MTX, where the maximum integral viable cell density (IVCD) of the clones ranged from 5.42 × 10^6^ to 1.34 × 10^7^ cells.d/mL. As for the maximum titres achieved relative to Clone 1, it ranged from 0.61 to 1.26. Based on the characterization of the cell clones, Clone 2 had the highest specific growth rate of 0.017 h^-1^ while Clone 9 gave the highest relative titre at 1.26. (Fig. 1a-c).Fig. 1. Comparing the characteristic of the cells clones (a) IVCD, (b) specific growth rate and (c) maximum titre relative to clone 1.
To evaluate production stability, the growth and specific productivity of the ten cell clones in the presence and absence of MTX were studied (Fig. 2). Relative specific productivity was calculated for alternate weeks from Week 2 to 12 of the stability study. The values were normalized to the specific productivity of Week 0. An average of the relative specific productivity for all ten clones in the absence and presence of MTX, as well as the average specific growth rates during passage every 3 or 4 days were summarised in Table 2.Fig. 2. Titre and specific productivity relative to Clone 1 with (a) no MTX and (b) 300 nM MTX over a period of 12 weeks.Table 2. Average specific growth rate and relative specific productivities of the ten clones from Week 2 to 12.No MTX300 nM MTXClone noAverage µ (h^-1^)Average relative qpAverage µ (h^-1^)Average relative qp10.0148 ± 0.00330.51 ± 0.350.0104 ± 0.00291.04 ± 0.4420.0213 ± 0.00200.58 ± 0.300.0196 ± 0.00220.77 ± 0.2330.0141 ± 0.00360.39 ± 0.260.0075 ± 0.00191.16 ± 0.3840.0146 ± 0.00210.37 ± 0.350.0104 ± 0.00110.65 ± 0.3850.0188 ± 0.00210.29 ± 0.050.0147 ± 0.00150.45 ± 0.1660.0213 ± 0.00230.28 ± 0.100.0152 ± 0.00140.57 ± 0.2870.0138 ± 0.00140.34 ± 0.160.0107 ± 0.00160.52 ± 0.2380.0225 ± 0.00220.45 ± 0.150.0206 ± 0.00200.65 ± 0.2690.0222 ± 0.00270.36 ± 0.210.0209 ± 0.00240.52 ± 0.12100.0166 ± 0.00380.41 ± 0.380.0119 ± 0.00230.90 ± 0.30
In the presence of MTX, most of the clones have a relative specific productivity ranging from 0.45–0.90, except for Clone 1 and 3 that have values exceeding 1. However, the average specific growth rates for the two clones were the lowest in the group (Fig. 1b).
It was also observed in the absence of MTX, the titre for the first two passages would generally be higher. This could possibly be explained by the better growth of the cells, regardless of its productivity due to lower selection pressure.
Taking into consideration the average relative specific productivity and growth rate of the cell clones, Clone 2 is one of the highest performing clones. It was therefore chosen for further analysis and larger scale culture production of rhA1AT.
Fed-batch production of rhA1AT in shake flasks and 10L shaken bioreactor
With an average specific growth rate of 0.0196 h^-1^, and relative specific productivity averaging at 0.77 from week 2 to 12, Clone 2 was evaluated to be the best performing among the ten clones. To characterize the fed-batch production process, Clone 2 was first cultured in the shake flasks. Subsequently, two fed-batch production runs were performed in the 10L scale shaken bioreactor (SB10-X, Kühner AG) to demonstrate the potential scalability of the shake flask process.
The culture was also grown with a starting working volume of 100 mL in duplicates in 500 mL shake flasks, which achieved maximum VCD of 1.17 × 10^7^ cells/mL on day 7. The harvest titre on day 13 was determined to be 437 mg/L, which is approximately 30% lower than the maximum achievable titre in the shake flasks on day 10 (Fig. 3b).Fig. 3. Comparison of rhA1AT fed batch production among the two bioreactor runs, SB10-X R1 (∆), SB10-X R2 (♦) and shake flask duplicates (■) (a) VCD and viability, (b) titre, (c) glucose, (d) glutamine, (e) glutamate, (f) lactate, (g) ammonium and (h) osmolality were characterized up to 14 days.
The production process was successfully scaled up to a 10L SB10-X reactor in fed batch mode. Similar to a shake flask, bubble free surface gassing was involved and the mode of agitation is via gentle orbital shaking. However, unlike the shake flasks in an incubator, dissolved oxygen level and pH could be actively controlled and maintained in the shaking bioreactor.
The growth, production, nutrients and metabolites profiles were found to be similar to the shake flasks when culture was scaled up in the first SB10-X run. Furthermore, the specific productivity of the cells in both the shake flasks and SB10-X were similar too (Figure S2). This demonstrates the scalability of A1AT production from shake flask to the benchtop 10L shaking bioreactor. The maximum achievable titre in SB10-X was approximately 20% higher as compared to the shake flasks culture (Fig. 3b). This was likely due to a better aeration and pH control in the bioreactor relative to the shake flasks, which enabled the culture to maintain at a cell density beyond 1e7cells/ml for an extended period.
As it was established that the cell line could be scaled up from the shake flasks to the 10L shaking bioreactor, under similar conditions as the first reactor run, the second bioreactor culture run was conducted. The growths of the culture in the two SB10-X reactor runs were comparable. A maximum VCD of 1.0 × 10^7^ and 1.29 × 10^7^ cells/mL were achieved on day 8 for the two bioreactor runs, which occurred a day later as compared to the shake flasks. Similarly, the VCD was also kept above 1 × 10^7^ cells/mL for a longer period as compared to the shake flasks duplicates (Fig. 3a).
It was observed that the titre of the culture decreases on day 12 for the first SB10-X run as well as the shake flasks. This occurred when the viability of the culture has dropped below 50%. Hence, for the second SB10-X run, the culture was terminated on day 10 at a viability of 68% to ensure the titre remained high at the point of harvesting. In addition, less undesired impurities, such as host cell proteins and cell debris^43^, would be present when the product is harvested earlier. This would hence put on less stress on the purification process.
The titres in the bioreactor runs were also comparable. The maximum titre in the bioreactor runs were at 789 and 729 mg/L for run 1 and 2 respectively (Fig. 3b). It was also noted that the specific production rate improved by 20% in Run 2 (Figure S2).
Daily analysis of the nutrients indicated that there was a gradual increase in the consumption of both the glucose and glutamine (Fig. 3c-d) over time, which corresponds to the cell growth. The amount of glucose during the second run of SB10-X ran low on day 6 to 9 (Fig. 3c), which could potentially inhibit the cell growth due to limited availability of the nutrient.
Glutamate accumulated over time in the shake flasks as well as the first reactor run (Fig. 3e). The ammonium levels also increased in the shake flasks and first reactor run, which surpassed 10 mM by Day 4. Hence, to reduce the accumulated ammonium and glutamate, the starting concentration of glutamine was lessened to 4 mM for the second bioreactor run. In addition, a maximum post feeding concentration of glutamine was also implemented for the second run (Fig. 3d).
This resulted in the ammonium level at the point of harvest being halved in the second reactor run relative to the first (Fig. 3g). Glutamate level had also started to decline on Day 8 in the second reactor run (Fig. 3e). On the other hand, lactate accumulation was observed to be more significant in second SB10-X run (Fig. 3f). Despite the differences in the accumulation of the metabolites, the viability trends for all the fed batch runs appeared to be similar.
In general, the consumption rates for both the substrates were observed to be higher in the exponential growth phase, followed by a decrease till the end of the culture. This is consistent to what was reported in literature for recombinant CHO cell line by other authors^38^. It was also observed that the specific consumption rate of glucose was slightly higher in the second SB10-X run after day 5 relative to the first run as well as the shake flasks experiment. Correspondingly, the glutamine level was close to zero and lowest on day 5 in the second reactor run. As a lower initial concentration of glutamine and a maximum glutamine concentration limit was applied, the specific consumption of glutamine was generally lower relative to the shake flasks and first reactor run in the earlier stages of culture (Figure S2).
With the accumulation of metabolites, the osmolality of the fed-batch cultures increased over time. The second SB10-X run had higher osmolality towards the end of culture relative to the shake flasks and first reactor run. This corresponded with the higher lactate concentration measured in the second SB10-X run (Fig. 3f), which triggered more base addition to maintain at the desired pH set point. Consequently, this also attributed to the increase in osmolality (Fig. 3h).
pH stability of rhA1AT
Purification of rhA1AT expressed in CHO cultures typically depended on anion exchange chromatography, affinity chromatography or size exclusion chromatography^9,31,44^. However, a dedicated virus inactivation step was not commonly reported^45^. For CHO cultures producing monoclonal antibody, the removal of potential viral contamination was typically via low pH inactivation^46^. The stability of rhA1AT was thus investigated to verify whether low pH viral inactivation is possible. The culture supernatant from a production run with a titre of 777 mg/L, was subjected to an acidic pH range of 3.0–7.0. Thereafter, the amount of rhA1AT that remained quantifiable by ELISA was analysed and stability was defined by the relative concentration of rhA1AT when normalized to the harvest sample supernatant. The results were presented in Fig. 4.Fig. 4. Percentage of A1AT that remained quantifiable by ELISA between pH 3.0 to 7.0.
rhA1AT appeared to be most stable at pH 7.0, where the highest proportion of A1AT remained quantifiable. A decrease in the amount of A1AT was observed at lower pH conditions (Fig. 4). This suggested that viral inactivation of the harvest via lowered pH, which typically ranges from 3.3 to 3.8^46^, may not be a feasible approach.
Low stability of rhA1AT between pH 2.0 to 4.0 has also been reported previously, and additional incubation in basic media subsequently was necessary to recover partial protein activity^47^. Hence, alternative viral inactivation method such as the dry heat treatment was explored. As the dry heat treatment is effective against non-enveloped viruses only, additional virus inactivation step such as tri-n-butyl cholate treatment that targets enveloped viruses was reported as well^45,48^.
Purification of rhA1AT
After the two step column purification, it is crucial that the purified product is able to retain its activity to remain functional. This could be verified via the in house activity assay, where the activity was determined based on the rate of inhibition for 0.84 mg/L of elastase at various dilutions of the purified sample. The activity of the rhA1AT was observed to increase from approximately 37% to 65% after the first step column purification. This potentially suggested that some A1AT protein aggregates and fragments could be removed after the first step of purification. The activity then remained unchanged following the hydrophobic interaction and buffer exchange (Fig. 5).Fig. 5. Comparison of the percentage of A1AT that remained active after purification, as determined using elastase inhibition assay.
The raw and purified rhA1AT samples after each step of purification were also loaded to HPLC to check on the purity of the sample. A single peak chromatogram was observed at a retention time of 12.69 min (Fig. 6b) after two steps of purification. A similar peak, which is identified as rhA1AT, was also observed in the raw sample at 12.69 min (Fig. 6a). Peptide mapping was also performed to confirm the identity of the purified rhA1AT generated (Fig. 7).Fig. 6HPLC chromatogram of (a) harvest cell culture fluid and (b) purified sample.Fig. 7. Peptide mapping of purified rhA1AT from a bioreactor run.
Discussion
The use of CHO cell lines to generate recombinant proteins has been well established in the biopharmaceutical industry. It remained a key choice in production as the cells have been able to generate products with similar glycosylation profile as human and without any human virus. One of the approved treatment options for A1AT deficiency patients is human plasma derived A1AT, which is an expensive and volatile option. Furthermore, with the discovery of more A1AT functions, these materials would be even higher in demand. Hence, it is of considerable interest that CHO cell line is able to generate A1AT on a large scale at a lower cost^10^.
The specific productivity for the two reactor runs in SB10-X was determined to be 9.6pcd and 12pcd respectively. While the cell specific productivity was higher than some of the cell lines used to generate rhA1AT^9,49^, the harvest concentration appeared to be lower than previous reported work^30^. This could be attributed to the declining VCD after day 7 or 8 in the shake flasks and SB10-X, which led to a lower IVCD. To improve the harvest titre, the process could be optimized to prolong the culture at high VCD to achieve both high IVCD and specific productivity.
Fed batch production is commonly practiced in the industry to intensify the culture growth and maximise the titre of the product harvested. However, the accumulation of metabolites remained one of the key factors which led to the death of the culture. This was similarly observed in this study, where the lactate increased above 2 g/L for all the fed-batch culture. There are multiple ways to mitigate this issue, which includes incorporating the continuous mode of production, limiting the amount of nutrients present or minimizing the accumulation of carbon dioxide in the bioreactor^50–54^.
A higher starting glutamine concentration in the shake flask duplicates and first SB10-X run have resulted in ammonium accumulation over time. This was undesirable, as apart from being growth and production inhibitory, it would tend to alter the glycosylation pattern of the recombinant protein^53,55^. A lower starting glutamine concentration was therefore used in the second SB10-X bioreactor run to reduce the amount of ammonium accumulating in the culture. This resulted in the feeding of the in-house concentrated feed on day 3, a day earlier as compared to the shake flasks. Nevertheless, the ammonium level was kept well below 10 mM for the second bioreactor run (Fig. 3g). Furthermore, comparison of the cellular specific productivities during the exponential phases between the two runs further substantiated that the accumulation of ammonium has inhibited production.
Interestingly, although lesser ammonium was accumulated, we have also observed a higher lactate concentration in the second reactor run. This suggested that in the absence of glutamine, more glucose would be utilised instead. This resulted in more lactate generated in the culture. Hence, it would be essential to balance the glucose and glutamine levels in the culture to reduce the accumulation of the metabolites^56^.
In this study, the production of rhA1AT was carried out in single-use reactors via headspace aeration. SB10-X would serve as an alternative option when gassing through a sparger is undesirable. This mode of aeration would be highly applicable to cell lines that are shear sensitive. Comparatively, aeration via headspace is less efficient as compared to sparging due to lower surface area to volume ratio available for gas transfer. Hence, to implement headspace aeration at higher cell density, it is critical to consider the liquid fill volume, diffusion coefficient of oxygen, kinetic viscosity of liquid, shaking speed and reactor diameter. These factors were known to have an impact on the k_L_a of orbital shaken bioreactors^57^.
The availability of cost-effective production capabilities such as single use reactor SB10-X has eased the upscale production of rhA1AT for research or early clinical studies. This could minimize the upfront capital costs and provide the manufacturer the flexibility to switch modality if necessary, since fabrication of stainless steel STR would not be necessary at this stage. In addition, SB10-X could be used as part of the seed train to inoculate large-scale STR^32^.
It was noted that the amount of quantifiable rhA1AT generally decreased when incubated at a lower pH. This could be due to the aggregation or denaturation of the protein at lower pH. This pose a challenge for downstream processing as low pH inactivation would not be feasible.
A two-step purification was performed to purify rhA1AT generated by the cell line. From the analysis of the purified samples, the activity of the product was found to be enhanced. Furthermore, approximately 70% of the products could be recovered after two steps of purification. The glycan analysis of the A1AT produced by the cell line utilised in this study has also been reported previously^58^.
This study has showcased that it was possible to generate rhA1AT at larger scale using the CHO platform and purify it. We foresee that it would be highly useful in ensuring a consistent and quality supply of rhA1AT to the market. There would hence be less reliance on human plasma derived A1AT.
The key focus of this study was on the scaling up capabilities in producing rhA1AT, and it was demonstrated that the productivity of the cell line in the shake flasks and OSB remained comparable. Areas to improve on the productivity and titre of the cell line were not fully explored. Culture conditions, such as temperature^59,60^, pH^61^, osmolality^50^, as well as feeding strategy^56,62^, could be modified to enhance on the cell growth and titre^63^. In the future studies, these could be explored to further increase on the rhA1AT production.
Calculations
Daily specific growth rate (µ) was determined by Eq. 1, where VCD_N_ is the viable cell density on Day N, VCD_N+1_ is the viable cell density the following day and t is the culture duration.
\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$VCD_{N+1}=VCD_N e^ {{\mu \Delta {\mathrm{t}}}}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mu = \frac{1}{\Delta t}1n(\frac{{VCD_{N + 1} }}{{VCD_{N} }})$$\end{document}The cumulative integrated viable cell density (IVCD) was calculated by trapezoidal rule according to Eq. 2.
\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$IVCD_{N} = \, IVCD_{N - 1} + \, 0.5 \, \times (VCD_{N} + VCD_{N - 1} ) \times \Delta t$$\end{document}Specific A1AT productivity (q_p_) between culture time t_0_ and t_1_ was determined by plotting A1AT titre (P) vs cumulative IVCD according to Eq. 3.
\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${q}_{p }= \frac{{P}_{1}-{P}_{0}}{{IVCD}_{1}-{IVCD}_{0}}$$\end{document}The specific consumption rate of nutrients between culture time t_1_ and t_2_ was determined by calculating according to Eq. 4, where C refers to the concentration of nutrient and V refers to the volume of culture.
Amount of nutrient fed = Concentration of nutrient in feed × Volume fed.
\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text{Specific consumption rate}\hspace{0.17em}=\frac{{C}_{N-1}{V}_{N-1}-{C}_{N}{V}_{N}+amount of nutrient fed}{0.5\times \Delta t\times ({VCD}_{N-1}\times {V}_{N-1}+{VCD}_{N}\times {V}_{N})}$$\end{document}Supplementary Information
Supplementary Information 1. Supplementary Information 2.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1David (Xiaojian) Zhao, B.N., Mark Stramaglia, Richard Fike. Improving Protein Production in CHO cells. 2008 (2008).
- 2Li, F., et al. Current Therapeutic Antibody Production and Process Optimization. 5. (2007).