Stable high-level expression of factor VIII in Chinese hamster ovary cells in improved elongation factor-1 alpha-based system

Cloning of p.1.1-F8BDD

The DNA fragment encoding the FVIII-BDD ORF with Kozak consensus sequence was obtained by PCR using primers AD-8-AbsF (5’-ACC TCG AGG CCG CCA CCA TGG AAA TAG AGC TCT CC-3’) and AD-ONheI-R (5’-AAG CTA GCT CAG TAG AGG TCC TGT GCC-3’), the plasmid pOptivec/F8BDD [18] as a template, and Tersus polymerase mix (Evrogen, Moscow, Russia). The PCR product was subcloned into T-vector, sequenced using the primers to FVIII ORF listed in [18] and then inserted into the expression vector p1.1 (described in [17]) using AbsI (Sibenzyme, Novosibirsk, Russia) and NheI (Fermentas, Vilnius, Lithuania) restriction enzymes. Resulting expression plasmid p1.1-F8BDD for the transfection was purified by EndoFree Plasmid MaxiKit (Qiagen, Valencia, CA, USA) and sequence-verified.

Cell culture

CHO DG-44 (Cat. No. A1100001, Invitrogen, Carlsbad, CA, USA) cell line, adapted to suspension culture in a chemically defined medium was used. Cells were cultured in Erlenmeyer flasks (VWR International, Radnor, PA, USA) in 30 ml of CD DG-44 medium (Invitrogen) supplemented with 8 mM L-glutamine (Invitrogen) and 0.18% of surfactant Pluronic F-68 (BASF Inc., Florham Park, NJ, USA) at 37 °C, 8% CO2, with stirring on an orbital shaker at 130 rpm. Cells were passaged every 2–3 days at the culture density 1.2 x 10⁶ live cells per ml, and diluted with fresh medium in a ratio of 1:4.

Transfection was performed using Fugene HD (Origen Biomedical, Austin, TX, USA), in a ratio of 18 μg of plasmid DNA per 1.5 × 10⁷ cells in 30 ml of culture medium. Plasmid for the transfection was linearized by the PvuI or used in the supercoiled form. Transfection efficiency was determined by fluorescence microscopy 48 h post-transfection for control plasmid mixture (p1.1-F8BDD plasmid mixed 95:5 by weight with green fluorescent protein encoding plasmid pEGFP-N2). After transfection cells were cultured for 48 h without medium exchange, then transferred into selective medium ProCHO 5 (Lonza, Basel, Switzerland) and grown until cell viability was restored to 90% (15–20 days). During the cultivation in selective medium, cells were passaged every 3–5 days to the cell density 5 × 10⁵ cells/ml or less until positive growth was observed. After this, cells were passaged every 3 days at the seeding density 3 × 10⁵ cells/ml. Stably transfected cell population from the suspension culture was transferred to the 96-well plates with the culture medium Ex-Cell CHO Cloning Medium (Sigma-Aldrich, St. Louis, MO, USA), supplemented by the 8 mM of L-glutamine and 50–400 nM MTX. The growth of colonies was monitored at the 10th and 14th days of cultivation. Growth-positive wells were screened by ELISA, culture medium in antigen-positive wells was changed and colonies were grown close to the confluence. Alternatively, at 48 h post-transfection cells were seeded into 96-well plates at a density of 10,000 cells per well, 200 μl per well of AOF Ex-Cell CHO Cloning Media (Sigma-Aldrich) medium with 50 nM MTX and 8 mM of L-glutamine at 37 °C, 5% CO2. Cells were maintained as described above.

Cells from most productive colonies obtained by both techniques were transferred to 48-well plates, cultivated for 5 d and screened again by ELISA. Best secreting oligoclonal lines were further expanded and readapted to suspension growth in ProCHO 5 medium with 8 mM L-glutamine during three sequential passages in 24-, 12- and 6-well plates.

FVIII concentration in the culture medium was measured 48 h post-transfection and at the end of the selection for stable transfectants. Oligoclonal cell line with the highest level of secreted FVIII, obtained in the presence of 50 nM MTX, was used for further amplification of bicistronic FVIII-dhfr cassette. It was grown in the ProCHO 5 medium with 8 mM of L-glutamine in the presence of increasing concentrations of methotrexate (MTX). Each amplification step lasted for 14–25 days until cell viability restored to 90%. Concentration of secreted FVIII was measured by ELISA and/or clotting test at the end of each step. Transgene-amplified oligoclonal cell line, obtained by the highest MTX selection pressure possible (4 μM) was cloned by limiting dilution in AOF Ex-Cell CHO Cloning Media (Sigma-Aldrich) with 8 mM alanyl-glutamine (Invitrogen) and HT (hypoxanthine/thymidine) supplement (Invitrogen), without MTX. Cells were seeded into 96-well plates at a density of 0.5 cells per well, 200 μl of medium per well at 37 °C, 5% CO2. The growth of single colonies was monitored at the 10th and 14th days of cultivation. The colonies were screened by ELISA, culture medium in positive wells was changed and colonies were grown close to the confluence. According to the results of the subsequent ELISA of actively growing colonies the most productive of them were transferred to 48-well plates. Clonal lines with the best levels of secretion were further expanded and readapted to suspension growth in ProCHO 5 medium with 8 mM L-glutamine during three sequential passages in 24-, 12- and 6-well plates.

Conditioned medium from 6-well plates was analyzed by ELISA. The top three cell lines were moved to Erlenmeyer flasks and evaluated by growth rate and specific productivity. Preparative batch cultivation was conducted in 500 ml Erlenmeyer flasks, 125 ml of culture medium per flask. Cells were seeded at a concentration of 2.5 × 10⁵ cells/ml and cultured without medium change to a density of 3 × 10⁶ cells/ml (4–5 days).

ELISA

Concentration of the factor VIII antigen (FVIII:Ag) was measured by ELISA using polyclonal antibodies to FVIII (LifeSpan BioSciences, Seattle, WA, USA) at 50 ng per well and specific MAb A2 to FVIII heavy chain described in [18]. Serial dilutions of normal calibrated human plasma (NPO Renam, Moscow, Russia) in PBS with 1% bovine serum albumin (BSA) were used as a standard. Samples were also diluted in PBS with 1% BSA.

Clotting assays

One stage clotting assay with chromogenic substrate was performed by the TECHNOCHROM® F VIII:C kit (Technoclone GmbH, Vienna, Austria); two stage clotting assay was performed by the optical coagulation analyzer ThromboScreen 400c (ThermoFisher Scientific, USA) and the Factor VIII- test reagent kit (NPO Renam), using normal calibrated human plasma (NPO Renam) as the initial standard and one stage assay-characterized sample of the FVIII-BDD preparation as the working standard. Specific activity of several FVIII-BDD samples was 1.7 times lower in the two-stage assay than in the one-stage assay if human plasma was used as the activity standard. Samples were diluted with the imidazole buffer with 1% BSA (NPO Renam).

Medium modifications

Effect of alkanoic acids in the culture medium on FVIII secretion was assessed in batch cultures, seeded as 3.75 × 10⁵ cells/ml and cultured for 3 days unless stated otherwise. Sodium butyrate and sodium propionate were dissolved in water at 1 M and adjusted to the pH 7.2 by HCl prior to addition to the culture medium. Antioxidant butylated hydroxyanisole was dissolved in DMSO at 2.5 M prior to addition to the culture medium, final concentration 0.1 mM; addition of dichlorofluorescein (DCF, Invitrogen) and subsequent flow cytometry analysis for oxidative stress level was performed as described in [19].

Oxidative stress induction and flow cytometry

Cells of the 11A4H line were seeded as 3.5 × 10⁵ cells/ml in the protein-free ProCHO-5 medium and cultivated for 4 days in the shake flask. On 3^rd and 4^th day 1 × 10⁶ cells were sampled. Then samples were stained with 5 μM of CM-H2-DCF-DA (Invitrogen) for 20 min on ice and counterstained with 1.5 μM propidium iodide just before flow cytometry analysis on Cytomics FC500 instrument (Beckman Coulter). Direct reads were gated on FS/SS scatter against intact control cells sample and then subsequently gated for propidium iodide exclusion (channel FL-3 620 nm). Generation of ROS were assessed as green fluorescence of activated DCF by measuring in FL-1 (525 nm) channel for 10⁵ gated cell reads, every obtained FL-1 scatter histogram were unimodular, mean values were calculated for core 5 ml of intact cell culture with H₂0₂ in final concentrations 10 μM, 100 μM and 1000 μM directly prior to live staining.

Quantitative PCR, PCR, RT-PCR

Copy number of the expression cassettes in the genome was determined by the quantitative real-time-PCR (qPCR). A calibration curve was prepared using serial dilutions of highly purified p1.1-F8BDD plasmid. Genomic DNA was isolated by Wizard SV Genomic DNA Purification System (Promega, USA) and quantified by the Qubit Fluorometer (Invitrogen), using dsDNA HS kit (Invitrogen) and external DNA concentration standard, prepared in-house from the highly purified plasmid DNA, quantified by UV spectrophotometry. Concentration of the genomic DNA was validated by control qPCR with primers to PPIB gene region, presumably unique to the CHO cells genome according to the BLAST search results. Samples of genomic DNA with the determined copy numbers of PPIB less than 0.3 per haploid genome or more than 2.0 were discarded. Weight of one CHO haploid genome was established as 3 pg according to [20].

The mRNA levels were assayed by reverse transcription and quantitative PCR (RT-qPCR). Total RNA was isolated by the RNeasy Mini Kit (Qiagen), its concentration was determined by UV spectroscopy, purification quality was confirmed by monitoring the A260/A280 ratio (1.8–2.0); integrity was verified by electrophoresis in 1% agarose gel. The cDNA was synthesized with the reagent kit Mint (Evrogen) using 1 μg of total RNA per sample. The ΔΔCq method [21] was used to calculate mRNA expression relative to β-actin.

Primers (listed in Additional file 1: Table S1) were designed by the Beacon Designer v7.51 program (PREMIER Biosoft International, Palo Alto, CA). The unique primers allowing selective amplification of the expression cassette were chosen for the genome copy number analysis, and primers located in the different exons or at the exon junctions were used for mRNA expression level analysis.

QPCR was performed using qPCRmix-HS SYBR reaction mixture (Evrogen) and iCycler iQ thermocycler (Bio-Rad, USA). Calculations of threshold cycles, calibration curves, PCR efficiency and copy numbers were made by the iCycler Iq4 program. All determinations were repeated 3 times, in 3–5 replicates, sample volume 25 μl.

PCR analysis of the whole FVIII ORF area was performed for genomic DNA and cDNA in the essentially same conditions. Templates were taken as 10 ng per tube, reaction volume was 10 μl. Primers AD-8-AbsF and AD-ONheI-R, described in the Cloning subsection, were used; Encyclo PCR kit (Evrogen) was employed. Temperature gradient was from 53 °C to 68 °C. Amplification program was 3’ at 95 °C; 25 cycles as 15” at 95 °C, 53–68 °C for 15”, 72 °C for 3’. Final elongation was 72 °C for 5’. Amplification products were electrophoresed on 0.8% agarose gels and stained with ethidium bromide.

Southern blot hybridization

Biotinylated probes for Southern blotting were prepared by the Biotin DecaLabel DNA Labeling Kit (Fermentas). Template for the probes was pAL-ID plasmid (containing regions present in expression plasmids p1.1 including origin of replication, a beta-lactamase gene, EMCV IRES, DHFR ORF [17]) or a PCR product corresponding to the fragment of FVIII-BDD ORF. Genomic DNA was digested with EcoRI (Fermentas) for 16 h, precipitated by ethanol, separated by 0.8% agarose gel. Gel transfer to an Amersham Hybond-N+ membrane (GE Healthcare, USA) was performed according to the manufacturer’s protocol in 20x SSC buffer (3 M NaCl, 0.3 M Na₃C₆H₅O₇) for 16 h. DNA was fixed by heating the dried membranes at 80 °C for 2 h. Prehybridization and hybridization were conducted according to [22] in the buffer containing 7% SDS, 0.5 M Na₂PO₄, pH 7.2, 1% BSA for 16 h at 65 °C. Membrane was washed according to the manufacturer’s protocol and stained by Biotin Chromogenic Detection Kit (Fermentas).

SDS-PAGE and immunoblotting

Samples of conditioned medium were clarified by centrifugation and concentrated 30 times by precipitation with trichloroacetic acid. Cell lysates were prepared using a modified RIPA buffer (50 mM Tris -HCl, pH 7.4, 1% NP- 40, 0.25% sodium deoxycholate, 150 mM NaCl, 1 mM Na-EDTA) with protease inhibitor cocktail (Sigma-Aldrich) and normalized to total protein concentration, determined by QuantiPro ™ BCA Assay Kit (Sigma-Aldrich). Samples were separated by electrophoresis in 7.5% denaturing polyacrylamide gel, sample load was 10 μg per well for cell lysates and the equivalent of 0.5 ml of conditioned medium. Gel slabs were stained by colloidal Coomassie stain (Fermentas) according to manufacturer's procedure or used for blotting. Transfer to the membrane, blocking, hybridization and staining were performed as described previously [18].

Purification of FVIII

Purification of FVIII from the cell culture supernatant was carried out as described in [23] with changes.

All solutions used for the purification, except the final composition, contained 0.02% Tween-80. To the conditioned culture medium or cell suspension, NaCl was added to 0.3 M and CaCl₂ to 27 mM. The resulting suspension was incubated for 30 min with stirring at 37 °C, clarified by centrifugation, supplemented with 10 mM of L-histidine pH 7.0 and 0.02% of Tween-80 and applied in 100 ml portions to the 1 ml HiTrap Capto MMC column (GE Healthcare), equilibrated with a solution 1-A (0.3 M NaCl, 10 mM CaCl2, 10 mM L-histidine, pH 7.0), flow rate 1 ml/min.

Column was washed by 10 volumes of solutions 1-B, 1-C and 1-D at 2 ml/min. Solution 1-B contained 1 M NaCl, 50 mM CaCl₂, 50 mM L- histidine, pH 6.5. Solution 1-C contained 0.1 M NaCl, 50 mM CaCl₂, 50 mM L-histidine, pH 6.5. Solution 1-D contained 0.3 M NaCl, 10 mM CaCl₂, 10 mM L-histidine, 0.25 M L-arginine-HCl, 10% ethylene glycol, pH 6.5. FVIII was eluted at 1 ml/min by the solution 1-E (0.3 M NaCl, 20 mM CaCl₂, 20 mM L- histidine, 0.8 M L- arginine-HCl, 50% ethylene glycol, pH 6.5).

Eluted FVIII solution was diluted 8-fold with a solution 2-A (10 mM NaCl, 10 mM L-histidine, 10 mM CaCl₂, pH 6.5) and applied to the 1 ml Tricorn 5/50 column (GE Healthcare) with the SP Sepharose FF (GE Healthcare), equilibrated with a solution 2-B (0.15 M NaCl, 10 mM L- histidine, 10 mM CaCl₂, pH 6.5), flow rate 2 ml/min. Column was washed with 20 ml of the solution 2-B and FVIII was eluted by the solution 2-C (0.34 M NaCl, 35 mM CaCl₂, 45 mM L-arginine-HCl, 0.2 M sorbitol, 10 mM L-histidine, pH 6.5). Resulting solution of the semi-purified FVIII was immediately frozen in liquid N₂ and stored in the −80 °C freezer.

Thawed FVIII solution was applied on the 1 ml Tricorn 5/50 column, filled with the immunoaffinity resin VIII Select (GE Healthcare) at 0.2 ml/min. Prior to the application of the sample, the column was equilibrated with the solution 3-A (0.3 M NaCl, 20 mM CaCl₂, 20 mM L- Histidine, pH 6.5). The column was washed sequentially at a rate of 2 ml/min by 10 ml of the solution 3-A and 3 ml of the solution 3-B (1 M NaCl, 20 mM CaCl₂, 20 mM L-Histidine, pH 6.5). FVIII was eluted with the solution 3-C (1.5 M NaCl, 20 mM CaCl₂, 20 mM L- Histidine, 50% ethylene glycol, pH 6.5).

Eluate was diluted 15-fold with the solution 4-A (10 mM NaCl, 10 mM L- histidine, 10 mM CaCl₂, pH 6.5) and applied at 2 ml/min to the 1 ml HiTrap Capto Q column (GE Healthcare), equilibrated with the solution 4-B (0.1 M NaCl, 0.02 M L-histidine, 20 mM CaCl₂, pH 6.5). Column was washed with 20 ml of the 4-C solution (0.3 M NaCl, 20 mM L-histidine, 20 mM CaCl₂, pH 6.5). Elution was performed by the 4-D solution (0.4 M NaCl, 0.02 M CaCl₂, 20 mM L-histidine, pH 6.5).

FVIII solution was applied in 2 ml portions at 0.4 ml/min flow rate to the size exclusion column Tricorn 10/300 Superdex 200 (GE Healthcare), equilibrated with the 5-A solution(9 g/L NaCl, 0.25 g/L CaCl₂, 1.5 g/L L-histidine, 0.01% Tween-80, 3 g/L sucrose, pH 7.0) at the flow rate 1 ml/min. Monomer of the FVIII in the long-term storage solution was collected in the final volume of 5–8 ml, divided into small aliquots, frozen in liquid N₂ and stored frozen for further analysis.

Mass spectrometry

Coomassie-stained bands of FVIII chains were cut from the polyacrylamide gel, washed twice (to remove the dye) with 100 μl of 40% acetonitrile in 0.1 M NH₄HCO₃ for 20 min at 37 °C, dehydrated by 100 μl of acetonitrile and completely dried in vacuo. Four μl of 15 μg/ml modified trypsin solution (Promega) in 50 mM NH₄HCO₃ added to each 3–4 mm³ gel piece. Hydrolysis was carried out for 5 h at 37 °C, then 7 μl of 0.5% trifluoroacetic acid (TFA) in 10% aqueous acetonitrile solution was added and mixed thoroughly. The upper layer of the solution above the gel was used for MALDI-TOF analysis.

2 μl aliquots of the peptide mixtures were mixed with 0.5 μl of 2.5-dihydroxybenzoic acid solution (Aldrich, 20 mg/ml in 20% aqueous acetonitrile, 0.5% TFA), applied to target spots and air dried.

Mass spectra were obtained on a MALDI-TOF instrument UltrafleXtreme (BrukerDaltonics, Germany) in the positive ion mode with the reflectron. For each sample analyzed the accuracy of the monoisotopic mass measurements was maintained as at least 0.003% (30 rrm) by the calibration on trypsin autolysis peaks. Spectra were obtained in the mass range 700–4500 m/z. Peptide mapping was conducted using the GPMaw program (Lighthouse data, Denmark).

Establishment and characterization of BDD-FVIII producing cell line

The expression construct p1.1-F8-BDD [GenBank: KY682701] was obtained by ligating the FVIII-BDD SQ ORF preceded by the Kozak consensus sequence into expression vector p1.1. Natural FVIII UTRs were completely removed; therefore in the resulting expression construct, FVIII ORF was entirely relocated into the context of EEF1A1 gene (Fig. 1a).

The FVIII:Ag level for the CHO DG-44 cells transiently transfected with this plasmid in the supercoiled form was 9 mIU/ml. The observed low level of FVIII:Ag in the transiently transfected cells corresponds to the low transfection efficiency, not exceeding 3% for the 16 kbp plasmid used. Meanwhile, CHO DG-44 cells stably transfected with the CMV promoter-based plasmid pOptivec/F8BDD displayed 2-fold lower FVIII:Ag level under the same growth conditions [18] despite higher transfection efficiency. Transfection of cells by the linearized expression plasmid p1.1-F8-BDD gave no detectable product 48 h post-transfection and no stably transfected cells after 3 weeks of selection.

Two selection strategies were used for transfected cells (cell lines generation scheme is depicted in the Additional file 2: Figure S1). Generation of a stably transfected population in suspension culture in the absence of both MTX and HT with subsequent selection of oligoclonal lines in adherent culture or direct selection of stably transfected oligoclonal lines in adherent culture immediately after transfection.

Stably transfected population (with FVIII:Ag level 134 mIU/ml) was plated into 96-well plates in the presence of 50–400 nM MTX. Multiple FVIII:Ag-positive colonies were observed in all cases (data not shown). It should be noted that most of wells in the plates with 50–200 nM MTX in the culture medium gave multiple colonies, but in the case of 400 nM MTX supplementation the proportion of wells with growing colonies dropped to less than 10% from total. Subsequently, the ten most productive wells in the plates with 50, 100 and 200 nM MTX were chosen from 200–300 wells with actively growing colonies and the 10 most productive wells for plates with 400 nM MTX supplementation were chosen from less than 40 wells with colonies. In these conditions, 200 nM MTX supplementation resulted in the sharp increase of the maximal titer in first 10 most productive wells (Fig. 1b).

For direct selection transiently transfected cells were seeded 48 h post-transfection into 96-well plates in the presence of 50 nM MTX and grown in the same manner.

Cells from the most productive wells, obtained by selection and subsequent amplification in the presence of 200 nM MTX or by direct selection with 50 nM MTX, were expanded and re-adapted to suspension culture. Levels of FVIII:Ag in the culture medium (Fig. 1c) ranged from 5 to 10 IU/ml which is comparable to typical published values for clonal cell lines producing FVIII-BDD [24]. Thus one-step transgene amplification under adherent conditions and direct selection of transiently transfected cells in the presence of 50 nM MTX resulted in oligoclonal cell lines, secreting FVIII-BDD in the range of 2–8 IU/ml.

Oligoclonal line 3-B8, obtained by the direct selection in the presence of 50 nM MTX and secreting mostly the intact two-chain form of FVIII-BDD and virtually no degraded FVIII chains, according to Western blot data (data not shown), was selected for further amplification in the suspension culture.

Three lead cell lines were selected and further expanded to shake flask cultures. The best clonal line 11A4H had specific productivity of 14.2 μIU/cell/d and acceptable doubling time of 28.5 h. Batch culture of 11A4H in shake flask, seeded at 3 × 10⁵ cells/ml in the protein-free ProCHO medium gave a maximal concentration of FVIII:C 39 ± 3 IU/ml at day 4, with final viable cell count (VCD) of 3.5 × 10⁶ cells/ml and viability over 85%. Further culturing without medium change resulted in sharp drop in FVIII:C level and subsequent drop in cell concentration and viability (data not shown).

Perfusion-like culturing was performed by seeding cells in fresh culture medium at 1.5 × 10⁶ cells/ml and changing medium daily by centrifugation and subsequent resuspension of the cell pellet in the fresh medium. In these conditions, FVIII:C level was in the range of 40–60 IU/ml for several days. We suggest that the observed increase of FVIII:C level in perfusion-like culture is due to the shorter time the secreted protein remained in culture medium exposed to proteases.

Since MTX was removed from the culture media at the stages of cell cloning and expansion, we expected that the resulting clonal cell line would maintain a high level of FVIII secretion for a sufficiently long time needed for the creation of cell banks and subsequent industrial manufacturing of the FVIII protein. For the 11A4H cell line the observed specific level of FVIII secretion (ratio of FVIII:C and VCD) showed no tendency to decrease during 120 days of cultivation (at least 70 generations) in the absence of MTX (Fig. 1d).

The transgene copy number in the genome of the cell lines expressing FVIII was determined by qPCR. Two transgene-specific areas of the integrated genetic cassette were used for analysis — SQ linker region of the FVIII ORF and the junction area between EMCV IRES and the DHFR ORF. Endogenous PPIB gene region, presumably unique in the CHO cells genome, was used as a single-copy control amplicon. For oligoclonal cell populations elevation of the transgene copy number during the first step of the MTX-driven transgene amplification and subsequent changes of this number in both directions are seen (Fig. 2a and c), but the observed FVIII:C levels remain mostly unchanged (Table 1). The transgene copy number for the lead clonal cell lines was not proportional to their specific productivity. Moreover, the transgene copy number for all three lines was strongly reduced compared to the parent oligoclonal line (Fig. 2b and d). Apparently, a significant portion of genomic inserts that occurred during the MTX-driven amplification was unstable or transcriptionally inactive. It should be noted that the copy number of the expression cassette in the genome of best-producing cell line 11A4H is moderate, only about 30 copies per haploid genome. While for CMV promoter-based producer cell lines, hundreds or even thousands of target gene copies per haploid genome were described elsewhere [25].

To ensure that genomic copies of inserted expression cassette in the candidate cell line 11A4H did not undergo large rearrangements (deletions) during the MTX-driven gene amplification process the genomic DNA isolated from A11H4 line was analyzed by PCR amplification of the whole FVIII ORF region using temperature gradient. The single PCR product corresponding to the full-length ORF (4818 bp) was observed, no truncated forms were detected. So we could conclude that all genome-integrated copies of the expression cassette contain the intact FVIII-BDD ORF without large deletions. (Fig. 3a). PCR product was also cloned and three plasmid clones were subjected to complete sequencing of the insert. All clones contained correct ORF of the FVIII without mutations, so we have assured that 11A4H cell lines codes only for correct FVIII protein (data not shown). Absence of incorrectly spliced or truncated FVIII mRNA was confirmed by PCR of total cDNA with temperature gradient, also showing single specific product of the right size (Fig. 3b).

Consistency of the genetic inserts in the genome of the candidate cell line 11A4H and the backup cell line 21A7B was investigated by Southern blotting using the FVIII-BDD ORF-specific probe (Fig. 3c) and another probe specific for the IRES, DHFR and the area of the bla gene (Fig. 3d). Single hybridization band of the correct size (app. 3700 bp) was observed for the FVIII-BDD ORF probe, indicating that both cell lines contain no truncated sequences of the target gene. In the case of IRES, DHFR, bla probe, three hybridization bands were visible. Two of them correspond to the predicted size of the restriction fragments, containing the IRES-DHFR area of the genetic cassette (app. 2200 bp) and the complete “bacterial maintenance” region of p1.1-F8BDD plasmid (2900 bp). Thus IRES-DHFR region apparently remains connected to the FVIII-BDD ORF sequence; this conclusion is confirmed by the qPCR data on copy numbers for FVIII and IRES-DHFR amplicons.

Third hybridization band had the size of app. 1500 bp and could correspond to the junction area between the “bacterial maintenance” region of the plasmid and the host cell DNA or to the junction between tandem copies of the genetic insert. We can assume that IRES-DHFR area of the p1.1-F8BDD plasmid remained intact in the genome of the A11A4H cell line and the non-functional “bacterial maintenance” region of the plasmid was present in both intact and truncated form. Our data cannot rule out the presence of unexpected genome integration points in the areas of the EEF1A1 UFR and EEF1A1 DFR parts of the plasmid. Due to the presence of the same DNA sequences in the genome of untransfected host cells, the genome insertion events of the EEF1A1 UFR and EEF1A1 DFR parts of the plasmid cannot be confirmed by Southern blotting and will be additionally studied by the new generation sequencing methods.

Relative levels of FVIII mRNA for three lead cell lines (Fig. 4a) were not proportional to their specific productivities, indicating that actual secretion level of FVIII is not simply limited by the transcribed mRNA levels. At the same time, relative level of FVIII mRNA in the parental line 3-B8 is at least 2.5 times lower, than in the clonal lines and the end-titer of FVIII for this line is approximately 6 times lower. Control clonal cell line DG-BDDFVIII-18 [18], expressing essentially the same FVIII molecule, coded by the CMV promoter-based plasmid, has even lower level of FVIII mRNA and lower secreted FVIII titer. Thus relatively high level of FVIII mRNA is required for high secretion rate, although quantitative levels of FVIII mRNA and the secreted protein are not proportional.

It is known, that high levels of translation of FVIII mRNA results in the retention of the immature FVIII pro-protein in the ER and drives the activation of the UPR (unfolded protein response) [26]. UPR is coupled to oxidative stress induction and increased cell death rate. These effects are much more pronounced in the case of full-length FVIII overexpression, although they are still detectable in the case of FVIII-BDD [26]. Common marker of the UPR activation is the up-regulation of the BiP chaperone, which directly interacts with the FVIII pro-protein in the ER. In two of the three lead cell lines obtained by us levels of BiP mRNA were found to be similar to the baseline (Fig. 4b). In the case of the most productive cell line 11A4H, level of BiP expression was increased approximately threefold. Similar increase of the BiP level was described for the BHK cells, secreting human FVIII at the 2–3 IU/10⁶ cells/d rate [26].

Modifications of the culture medium

The secretion level of several recombinant proteins can be increased by adding sodium butyrate to the culture medium [27]. Expression level of FVIII is also boosted by sodium butyrate [28], probably due to inactivation of transcription silencer, present in the mRNA of FVIII [29]. A similar effect was described for sodium propionate addition [24]. At the same time, addition of sodium butyrate to the culture medium induces endoplasmic reticulum stress in the FVIII-producing cells, as was shown in [19] and damages the cells. Major consequence of endoplasmic reticulum stress induction is the oxidative stress, which can be diminished by the membrane-permeable antioxidant butylated hydroxyanisole (BHA).

We applied sodium butyrate and sodium propionate supplements to batch cultures of 11A4H cells for 24 h. In the case of sodium propionate treatment both FVIII:C and VCD were decreased (Fig. 5a), but sodium butyrate, used at 0.25 mM, allowed the increase of FVIII:C and simultaneous decrease of VCD (Fig. 5b). Addition of up to 0.5 mM of sodium butyrate allowed cell proliferation, so 20% increase of volumetric productivity was recorded for batch culture in the presence of 62.5 μM of sodium butyrate (Fig. 5c) after 3 days of cultivation. Ratio of procoagulant activity of secreted FVIII and its concentration as ELISA antigen (FVIII:C/FVIII:Ag) was not affected significantly by the addition of sodium butyrate (Fig. 6a) up to 0.5 mM, indicating that sodium butyrate does not damage the specific activity of the FVIII.

We measured levels of reactive oxygen species (ROS) generation as a common marker of the oxidative stress and found that the ROS level is directly proportional to the concentration of sodium butyrate (Fig. 6c). The addition of 10 μM BHA or 10% of fetal bovine serum was sufficient to reduce the ROS generation level to baseline values (Fig. 6d). At the same time, both specific productivity and volumetric productivity in the 96 h batch culture of 11A4H cells were significantly increased (Fig. 6b). Thus the volumetric productivity of the 11A4H cell line in the batch culture can be almost doubled by using non-toxic and inexpensive medium supplements.

Membrane-bound fraction of the secreted FVIII

Significant fraction of the secreted FVIII binds to the cell membrane, preferably to the acidic phosphatidylserine-containing membranes [30]. In the case of full-length FVIII, membrane-bound fraction was estimated as ¾ of all extracellular FVIII [16] and in the case of FVIII-BDD it may increase up to 90% [31]. Membrane-bound FVIII-BDD may be liberated during the cultivation by the addition of vWF, annexin V or water-soluble phosphatidylserine analogue — o-phospo-L-serine [32]. Membrane-bound FVIII also may be dissociated from the cell membrane by the hypertonic solution with 20–30 mM CaCl₂ [32]. Both commercialize preparations of the full-length FVIII are manufactured in the presence of vWF in the culture medium, realized as co-expression of human vWF [10] or addition of human plasma proteins to the culture medium [9]. The presence of vWF in the culture medium diminishes the membrane-bound fraction of FVIII and protects the secreted FVIII from proteolytic degradation. It should be completely removed during the purification, thus increasing the complexity of downstream procedures and in-process controls. Commercialized process of obtaining the FVIII-BDD SQ does not involve the use of vWF in any form; and therefore the bio-analogues of FVIII-BDD SQ are expected to be completely free of vWF. Annexin V is the intracellular protein tightly binding to the acidic membranes, its use in the industrial process of FVIII production also seems non-practical. At the same time, the addition of o-phospho-L-serine to the culture medium is not expected to alter the downstream process or to increase the cost of cultivation significantly.

The proportion of the membrane-bound FVIII:C, liberated by NaCl + CaCl₂ treatment according to [32], for 11A4H cell line was below 15% in all cases, including the culture, induced by sodium butyrate (data not shown). Addition of fetal bovine serum, containing bovine vWF, enhanced FVIII:C level in the culture medium (growth time 72 h) by about 60% (Fig. 6a). We suggest, that this modest increase of FVIII:C level is the result of FVIII protection from proteolytic breakdown by the fWF. Similarly, the addition of o-phospho-L-serine to the culture medium caused no increase in the FVIII:C level, however, it caused the decrease of VCD (Fig. 5d). We believe that the lack of phospho-L-serine effect we observed was due to the low content of phosphatidylserine in the cell membrane of 11A4H cells.

BDD-FVIII purification and characterization of the recombinant protein

Purification of FVIII from the culture medium was performed using the five-stage chromatography process, including the solvent/detergent viral inactivation step. SDS-PAGE analysis and Western blotting of the protein fractions collected after the first three purification stages are shown on Fig. 7a and b, the purification balance sheet is shown in Table 2. Although the visually homogeneous FVIII was obtained after the first three chromatography steps, the specific procoagulant activity of FVIII was significantly increased by subsequent anion exchange chromatography and size exclusion chromatography steps. The specific procoagulant activity of the completely purified product is comparable to the specific activity of the pharmaceutical grade FVIII-BDD SQ (7’600–13’800 IU/mg protein, according to [33]). The total yield of the FVIII protein (as measured by the procoagulant activity) was above 20%, which proves that the purification process employed is industrially applicable.

#	Name	Resin	Volume, ml	Protein, mg/mla	FVIII:C, IU/ml	Specific activity, IU/mg	Total FVIII:C, kIU	Stage yield	Overall yield
0	Clarified medium	-	320	--	36.3	-	11.6	-	-
1	Multi-mode anion exchange	Capto MMC	17	0.233	496.9	2133	8.45	73%	73%
2	Cation exchange	SP Sepharose	10	0.271	538.1	1986	5.38	64%	46%
3	Affinity	VIII Select	11	0.046	382.0	8304	4.20	78%	36%
4 + 5	Anion exchange + size exclusion	Capto Q + Superdex 75	10	0.026	284.1	10925	2.84	68%	24%

^a1–3 protein by Bradford; stage 4 + 5 by OD₂₈₀ nm

Tryptic peptides of the purified BDD-FVIII were analyzed by MALDI-TOF mass spectrometry. Mass spectra of tryptic peptide mixtures of the two chains of BDD-FVIII or of the single-chain FVIII form did definitely show peaks corresponding to 44% of total BDD-FVIII protein length (Fig. 7c). None of the mass-spectra, obtained in the positive mode, contained peaks corresponding to the peptide with three sulfotyrosine residues and the peptides with three fully occupied N-glycosylation sites. At the same time, we identified unmodified peptides, containing partially occupied O- glycosylation sites and peptides with T1651 and T1652 normally sulfated tyrosine residues. Thus the purified BDD-FVIII sample contains a fraction of molecules with non-sulfated tyrosine residues T1651 and T1652.