Mutations of six amino acid residues in a B domain-deleted blood coagulation factor VIII have a cumulative effect on increasing its secretion

Abstract

Background

High levels of factor VIII (FVIII) expression are needed for various applications to treat Hemophilia A. Besides deletion of the B-domain (BDD) and codon-optimization, F309S mutation, and other 5 mutations (X5) in FVIII were previously described to increase its expression.

Objectives

To investigate whether combining the 6 aforementioned mutations (X6) results in a further increase of FVIII expression.

Methods

The 5 (X5) and 6 (X6) mutations were introduced into a BDD-FVIII (wild-type [WT]), and proteins were expressed in cell cultures with different transgene copy numbers, purified, and tested for specific activity, binding to von Willebrand factor and a low-density lipoprotein receptor-related protein fragment, tyrosine sulfation levels and immunogenicity in silico and in human T-cell culture. The 6 mutations were also reproduced in full-length FVIII (FL-FVIII) and tested for secretion levels.

Results

From the single-copy transgene cell lines, secretion levels of X5 and X6 increased 1.6-fold and 2.3-fold, respectively, compared with WT. These levels increased proportionally with increasing transgene copy number, approaching saturation. The specific activity, binding to von Willebrand factor and lipoprotein receptor-related protein fragment, and assessments of immunogenicity of X6 in model systems were similar to WT, while tyrosine sulfation levels, were moderately lower at the highest gene dose. However, the 6 mutations reproduced in FL-FVIII did not result in increased secretion.

Conclusion

Combining the 6 mutations in BDD-FVIII improved its expression and did not affect general protein properties, making it promising for future product development. The data also indicate that BDD-FVIII and FL-FVIII have different expression mechanisms.

Essentials

• •Expression of blood coagulation factor VIII (FVIII) is low and requires improvement.
• •This improvement is needed for the use of recombinant FVIII for treatment of Hemophilia A.
• •Specific mutagenesis of FVIII resulted in increased protein secretion.
• •The protein’s general properties were preserved, making it a candidate for further development.

1. Introduction

Congenital deficiency in factor (F) VIII results in a bleeding disorder (Hemophilia A) that can be life-threatening. The disease is generally treated by infusions of therapeutic FVIII (1-4 times per week), although recently, FVIII gene therapy has also been approved.

FVIII is a large multidomain heterodimeric protein (∼300 kDa) that carries posttranslational modifications such as glycosylation and tyrosine sulfation [1] and is found in plasma at very low concentration (∼1 nM or ∼0.3 μg/mL) [2]. Due to the low gene expression level and high complexity of FVIII, researchers have sought to increase the expression level in cell culture to meet the needs of industry and research.

Different approaches to increase expression of recombinant FVIII in cell culture have been described. Historically, the first was genetic deletion of the large B-domain (B-domain deleted [BDD]), which is considered dispensable for FVIII functions. While deleting the B-domain results in only a moderate increase of FVIII expression (∼1.5-fold), the advantages include (i) achieving uniformity in the protein molecule with protein purification convenience, as full-length (FL) FVIII is heterogenous due to various C-terminal truncations of the B-domain, and (ii) decreasing the gene size to fit in a viral vector capsid for FVIII gene therapy. The second approach is gene codon optimization that increased expression efficacy of the BDD-FVIII by several-fold [[3], [4], [5]] and was used in a recently approved gene therapy for Hemophilia A [6]. It should be noted, however, that codon-optimized proteins should be evaluated to ensure that their properties (conformation, specific activity, ligand binding, etc.) are not adversely affected [5].

Several point mutations in FVIII have also been described that increase its expression in cell culture. A mutation, F309S in FL-FVIII increased both its secretion level and activity in similar amounts, indicating preservation of specific activity [7]. Notably, F309 is located within a binding site in FVIII for the chaperon BiP (GRP78) in the endoplasmic reticulum, and the mutation’s effect was attributed to decreased intracellular aggregation of FVIII during expression [8]. In another study, 5 replacements with homologous porcine amino acid residues (I86V, A108S, G132K, M147T, and L152P) in a BDD-FVIII resulted in increase of secretion level, while they did not affect protein specific activity [9]. Notably, testing specific activity provides a strong criterion for the integrity of the structure and function of FVIII, as revealing its activity requires interaction with multiple ligands within the tenase complex [1].

In our study, we aimed to combine the 6 amino acid replacements in BDD-FVIII to investigate whether this results in further enhancement of secretion and also test that for FL-FVIII, as recent studies indicated that both FVIII variants have different expression pathways [10]. If secretion is increased, we aim to evaluate the major functional properties, primarily specific activity and binding to selected FVIII ligands.

The 6 substitutions were introduced in a codon-optimized BDD-FVIII (Wild-Type [WT]), previously shown to have increase in secretion compared with nonmutated protein while preserving functional properties [5]. We generated (i) a variant with the 5 mutations (X5) and (ii) a variant with the additional mutation F309S (X6). In parallel, the 6 mutations were also reproduced in FL-FVIII. Consequently, we evaluated the effects of the mutations on protein secretion efficacy and selected properties.

2. Materials and Methods

Moroctocog alfa, BDD-FVIII drug product (Pfizer), was used as a reference standard (S1) for protein quantitation and control in FVIII activity assays, and turoctocog alfa (Novo Nordisk) was used as an additional control (S2) in selected assays. Low-density lipoprotein receptor-related protein (LRP1) cluster II was produced as described [11]. FIX, von Willebrand Factor (VWF), and Human α-thrombin were purchased from Prolytix. Anti-Human FVIII:C Affinity Purified (Polyclonal) (sheep immunoglobulin [Ig]G) antibodies (CL20035AP) were from Cedarlane Laboratories and mAb GMA-012 (Mab HFVIII R8B12) targeting an epitope on FVIII, formed by residues 497-510 and 584-593, was from Green Mountain Antibodies. Rabbit polyclonal anti-bovine VWF antibody ab6994 was purchased from Abcam. For FVIII activity assays, the following reagents were used: Chromogenix Coatest SP4 Factor VIII kit (Chromogenix), SynthASil, human hereditary FVIII deficient plasma (HRF Inc), and WHO 8th International Standard for FVIII Concentrate (NIBSC code 07/350) was used as a standard in FVIII activity assays.

2.1. Plasmid construction, mutagenesis, and generation of recombinant lentiviruses

A codon-optimized BDD-FVIII gene (wt) (GenBank: MT568995.1), inserted in a lentivirus vector [5], was introduced with 5 mutations, I86V, A108S, G132K, M147T, and L152P [9] (x5), and a construct with an additional sixth mutation, F309S [7] (x6, GenBank: PQ450447) was generated using the service of GenScript. Recombinant lentiviruses were produced as described [5] and their titers (genome copies per mL) were measured using a real-time quantitative polymerase chain reactions (qPCR) lentivirus titer kit, LV900 (Applied Biological Materials) and QuantStudio 3 Real-Time PCR system (Thermo Fisher Scientific) according to vendor instructions.

2.2. Transduction of cultured cells with recombinant lentiviruses

CHO cells (Sigma-Aldrich) were cultured in F12 media supplemented with 10% fetal bovine serum, L-glutamine/glutamax (1X), and antibiotic-antimycotic (1X). Each lentivirus (x5 and x6) was transduced into cells, and secretion of respective FVIII variants (X5 and X6) was confirmed by polyacrylamide gel electrophoresis (PAGE) and western blot (WB). In second experiment, CHO cells were serially transduced with lentiviruses wt and x6 4 times, and clonal cell lines stably expressing WT and X6 were generated using standard single-cell cloning by limiting dilution and testing protein secretion by an electrochemiluminescence assay (ECLA, see below) as described [5]. In the third experiment, CHO cells were transduced with lentiviruses wt, x5, and x6 at multiplicity of infection (MOI) of 2, 4, or 8, and genomic DNA was quantified for the integrated transgene copy number with quantifying secreted FVIII variants levels by ECLA (see below).

2.3. Transgene copy number quantitation in clonal cell lines

Genomic DNA was extracted by MagMAX DNA multi-sample kit (Fisher Scientific). Analysis via qPCR was performed using QuantStudio 3 real-time PCR system and TaqMan Fast Advanced Master Mix (Thermo Fisher Scientific). TaqMan Assays, 18S rRNA (Hs99999901_s1 (Thermo Fisher Scientific) was used for endogenous control. To amplify transgenic FVIII cDNA, we used a forward primer GCAGCCTGATCAGCTACGA, reverse primer GTTGGGCTTCACGAAGTTCTTC, and FAM-conjugated primer CTGCCGCTGGTCCTC. The conditions were as follows: UNG incubation for 2 minutes at 50 °C; initial denaturation for 2 minutes at 95 °C; 40 cycles including denaturation for 10 seconds at 95 °C and annealing/extension for 30 seconds at 60 °C.

2.4. FVIII variants protein quantitation by electrochemiluminescence assay

The assay (ECLA) was performed for quantitation of FVIII variants using MESO QuickPlex SQ 120MM instrument and Meso Scale Discovery (MSD) GOLD 96-well Small Spot Streptavidin SECTOR plates (MSD). Samples were serially diluted in cell culture media and processed using the FVIII reference standard, biotinylated Mab GMA-012 as a capture antibody, and the polyclonal anti-FVIII antibody conjugated with MSD GOLD SULFO-TAG reporter for detection according to manufacturer’s instructions. The results were expressed in relative units of mass (RUm) corresponding to the protein amount of FVIII standard with one International Unit (IU) of FVIII activity.

2.5. Generation of cell lines with a single-copy transgene of FVIII variants

Flp-In CHO cells (Invitrogen) stably expressing WT, X5, and X6 were generated following manufacturer instruction. Briefly, transfection was made using pcDNA5/FRT vectors encoding these FVIII variants and pOG44 vector. The respective cell lines were isolated using selection with hygromycin (Fisher Scientific) and single-cell sorting as described [5]. Cell lines were confirmed for gene integration by sequencing using service of LifeSct.

2.6. FVIII variant expression and purification

Proteins were expressed from the clonal lines as described [5] and purified using VIIISelect sorbent (Cytiva) (Olivares et al., manuscript under preparation). Protein identity and purity were tested using PAGE/WB, ECLA, and mass spectrometry; and endotoxin level was measured with Pierce Chromogenic Endotoxin Quant Kit (Thermo Fisher Scientific).

2.7. FVIII variant analysis by polyacrylamide gel electrophoresis and Western Blot

The electrophoresis (PAGE) used a 4% to 12% Bis-Tris gel (Fisher Scientific) to resolve protein bands. This was followed by gel staining with SimplyBlue Safe Stain (Thermo Fisher Scientific), and densitometry analysis was performed using Image Studio Lite software (LI-COR Biosciences). Alternatively, protein was analyzed by WB following a standard protocol of Li-Cor Biosciences with Pierce protein-free blocking buffer (Thermo Fisher Scientific), where protein bands were detected using anti-human factor VIII:C affinity purified (polyclonal) (sheep IgG) (Cedarlane Laboratories) and 680RD Donkey anti-Goat IgG (LI-COR Biosciences) or mouse monoclonal GMA-012 (Green Mountain Antibodies) and IRDye 800CW Donkey anti-Mouse (LI-COR Biosciences). Alternatively, protein was analyzed by WB using chemiluminescent detection with secondary Peroxidase AffiniPure Donkey anti-sheep antibody (Jackson ImmunoResearch Laboratories Inc) visualized with SuperSignal West Pico PLUS (Thermo Scientific) on a G:Box Mini 6 (Syngene).

2.8. FVIII variants activity and specific activity measurements

FVIII activity in samples was tested using a chromogenic substrate (CS) and one-stage clotting (OC) assays against the 8th International Standard for FVIII. The CS assay used Coatest SP4 kits, samples were diluted to ∼1 RUm/mL in FVIII-deficient plasma, and further serial dilutions were performed in a kit supplied buffer. The assay volume was scaled down to 120 μL for use in 96 well microplates and otherwise follows the manufacturer’s instructions. Pipetting was performed using 96-channel automated pipettor. Absorbance was recorded with Biotek Neo2 (BioTek Instruments Inc) at 410 nm. The maximal rate of substrate conversion (0 to 2.5 min) was used as the assay readout.

The OC assay was performed using SynthASil reagent and the ACL TOP 550 analyzer (Werfen), following the standard protocol with the sole modification of replacing lyophilized Werfen’s FVIII-deficient plasma with FVIII-deficient plasma (fresh/frozen).

The data from CS and OC assays were analyzed using OriginPro 2023 software (OriginLab). FVIII activity in samples was calculated using an in-house script. Calibration curves of the serially diluted standard were approximated with linear fitting in linear-log or log-log scale. FVIII activity in each sample (IU/mL) was estimated as an average for 2 to 7 dilutions. Relative specific activity of FVIII variants (ratio of FVIII activity by either CS or OC to protein mass by ECLA) was calculated as described in Supporting Information (page 1).

2.9. Surface plasmon resonance assay

The surface plasmon resonance assay was performed using Biacore T200 instrument (Cytiva). Various concentrations (0.5-120 nM) of WT and X6 were tested for binding to immobilized VWF and LRP1 Cluster II as described [5]. Kinetic parameters were derived from the ligand association and dissociation signals using T200 v3.2 software (Cytiva).

2.10. Nanospray liquid chromatography tandem mass spectroscopy assay

The assay (Nano-LC-MS/MS) used purified WT and X6 (200-300 μg/mL of protein). The samples (in triplicates) were reduced by dithiothreitol, alkylated with iodoacetamide, and trypsin digested as described [12]; selected samples were digested using combined GluC and LysC enzymes (Promega). The resulting peptides were analyzed using Ultimate LC and Fusion Orbitrap MS instrument (Thermo Fisher Scientific) as described [13]. For identification of runs at positive mode using database search, Byos version 5.3.44 (Protein Metrics) was used to match MS/MS spectra with typical settings at the U.S. FDA facility for Biotechnology resources [13].

2.11. Peptide–MHCII binding affinities for WT FVIII and X6 variants

The binding affinities were estimated using the NetMHCIIpan 4.3 (https://services.healthtech.dtu.dk/services/NetMHCIIpan-4.3/) using the default peptide length of 15 amino acids spanning the full-length WT FVIII and X6. The results were used to generate a heat map where colors correspond to the binding affinity of the given allele/sequence pair with a 50 nM cutoff. The squares on the heat map correspond to the central amino acid (position #8) in a 15-mer peptide. Analyses used a set of 35 HLA-DRB1 alleles which represent ∼97% of the allele coverage for the North American population.

2.12. HLA-typed peripheral blood mononuclear cells

Peripheral blood mononuclear cells from healthy donors were obtained from the National Institute of Health blood bank. PBMCs were HLA-typed and stored in vapor liquid nitrogen until use [14].

2.13. Calculation of HLA-DRB1 allele population coverage

The HLA-DRB1 allele frequency in North American Population were obtained from The Allele Frequency Net Database (http://www.allelefrequencies.net/). The total population coverage for the donors used in this study was obtained by manual adding the population coverage of each donor.

2.14. In vitro stimulation of PBMCs with WT and X6

Cryopreserved PBMCs were thawed, washed twice with RPMI medium containing anti-aggregate wash (CTL) at 330 × g for 10 minutes, resuspended in CTL Test medium (CTL) supplemented with 1% L-glutamine (Gibco) and 10 μL/mL CD28/CD49d co-stimulatory antibody (BD Biosciences) and counted to determine cell number and viability. For the in vitro antigenic stimulation, 1.5 × 10⁶ cells per well were plated in 24-well plate and stimulated with 10 μg/mL of WT or X6, 1 ng/mL of Staphylococcal Enterotoxin B (SEB) (positive control) (Sigma-Aldrich), or 10 μg/mL of FIX (negative control). Unstimulated samples were treated with dialysis buffer (20 mM histidine, 500 mM NaCl, 5 mM CaCl₂, 9 g/L sucrose, and 0.005% Tween-80, adjusted to pH 7.2 and filtered [0.22 μm pore size] in CTL). Plates were incubated at 37 °C, 5% CO₂ for 4 days. Then, 0.5 mL of culture media was removed and replaced with fresh CTL supplemented with 1% L-glutamine, 10 μL/mL CD28/CD49d co-stimulatory antibody and recombinant human IL-2 (Pepro Tech) at 100 IU/mL and incubated at 37 °C, 5% CO₂. After 48 hours incubation with changing the media, cells were collected, counted, and used for plating in precoated IFN-γ enzyme-linked immunospot (ELISpot) assay plates.

2.15. Re-stimulation of cells and IFN-γ ELISpot assay

Prestimulated PMBCs, 10⁵ cells per well, were plated on precoated 96-well IFN-γ ELISpot plates (hIFNgp-2M/5, CTL) in 100 μL/well of CTL supplemented with 1% L-glutamine and co-stimulatory antibody. For samples treated with SEB, 25 × 10³ cells were used to allow the software to clearly read individual spots. Restimulation of PBMCs was done by adding 100 μL per well of supplemented media consisting of unstimulated cells, dialysis buffer (20 mM histidine, 500 mM NaCl, 5 mM CaCl2, 9 g/L sucrose, and 0.005% Tween-80, adjusted to pH 7.2 and filtered [0.22 μm pore size]), SEB-stimulated cells (positive control), 0.5 ng/mL; FIX-stimulated cells (negative control) (5 μg/mL); or either WT or X6 (5 μg/mL), bringing total volume to 200 μL/well. Plates were incubated overnight at 37 °C in 5% CO₂ and ELISPOT assay was performed according to the manufacturer’s (CTL) instruction. Stimulation index was used to evaluate the frequency of T-cells showing antigen-specific IFN-γ. The SI value is calculated by dividing the mean spots forming cells (SFC) treated with antigen by the mean SFC in the unstimulated cells. A SI ≥ 1.9 was considered significant. CTL Immunospot SC Suite/ Immunocapture analysis software was used to scan, count and quality control ELISpot wells.

2.16. Generation of mutant variants of FL-FVIII

Genes coding FL-FVIII, FL-FVIII F309S, FL-FVIII with the 6 mutations (FL-FVIII X6), and a codon-optimized BDD-FVIII with replacement of the B domain with a V3 peptide (FVIII-V3) [4] were cloned in pUC57 plasmid under control of hepatocyte-specific ApoE/AAT promoter. HepG2 cells (ATCC), cultured in DMEM with 10% fetal bovine serum and 1% penicillin-streptomycin at 37 °C, 5% CO₂, were transfected with plasmid DNA at low or high doses (100 or 300 ng/well of 96-well plate) using Lipofectamine 3000 (Thermo Fisher Scientific) following the manufacturer’s protocol. Posttransfection, 72 hours, protein levels in culture media were measured by an enzyme-linked immunosorbent assay, specific to human FVIII.

2.17. Statistical analysis

Statistical significance was determined using one-way analysis of variance and t-test using GraphPad Prism 9.4.0. (GraphPad Software).

3. Results

3.1. Generation of FVIII variants and assessment of their expression levels

The 5-point and 6-point mutations described above were introduced in a codon-optimized BDD-FVIII (WT) using the expression system previously developed by us [5]. Recombinant lentiviruses (wt, x5, and x6) were transduced into adherent CHO cells, and secretion of the generated FVIII variants (X5 and X6) into the media was confirmed by immunoblot (Figure S1).

To compare expression efficiency, we tested protein secretion levels versus transgene copy numbers. The lentivirus titers were quantified by qPCR, and the cells were transduced with each lentivirus at MOI of 2, 4, or 8. The cell media were tested for FVIII variant content along with assessment of integrated transgene copy number. This confirmed proportionality between transgene copy number and the MOI used for wt, x5, and x6 gene variants: at MOI = 4 being respectively 2.1-fold, 2.0-fold, and 1.9-fold relative to MOI = 2, and at MOI = 8, they were 4.1-fold, 3.9-fold, and 4.0-fold relative to MOI = 2, respectively (Figure 1A). The respective media protein levels of X5 and X6 were higher than those of WT and increased proportionally with increasing MOI, whereas WT levels increased by only ∼60% when MOIs doubled (Figure 1B). X5 secretion levels were consistently higher than WT levels for all MOIs tested, and X6 secretion levels were higher than X5 levels with smaller differences upon MOI increase (Figure 1B). These data are presented in numerical form in Table S1.

Figure 1.

Correlation between relative copy numbers of B-domain factor VIII transgenes and secretion levels of respective proteins (WT, X5, and X6). Cultured CHO cells were transduced with respective lentiviruses at MOI of 2, 4, and 8 and assessed for the integrated transgene copy number by qPCR and level of secreted protein by ECLA. (A) Relative transgene copy numbers normalized to MOI = 2 for each FVIII variant. (B) Protein levels in media are expressed in RUm (Methods). (C) Ratios of respective ECLA/qPCR values. In each panel, the bars show respective relative values (means +/− SD, n = 5). Statistical significance was analyzed using paired t-test. ECLA, electrochemiluminescence assay; MOI, multiplicity of infection; RUm, relative units of mass; qPCR, quantitative polymerase chain reactions. ∗P < .05 ∗∗P < .01 ∗∗∗P < .001 as determined by the test.

Thus, we observed a progressive increase of protein secretion levels along with increased numbers of mutations. These results indicated that combining these 6 mutations in the BDD-FVIII has a cumulative effect on increasing its expression.

3.2. Assessment of FVIII variants expression from single-copy transgene

The data shown in Figure 1B indicate that the difference in protein expression among the tested constructs is generally more pronounced at low MOI, suggesting that one of the factors contributing to this pattern is saturation of the cell’s biosynthetic machinery by protein overexpression upon increasing transgene copy number. For a more realistic comparison of gene expression efficacy, we compared protein expression from the single-copy transgenes, assumed to place the lowest burden on the cellular machinery. Cell lines with a single copy of respective transgenes (wt, x5, and x6) were generated, and the culture media were analyzed for protein secretion levels. Compared to WT, the levels of X5 and X6 increased 1.6-fold and 2.3-fold, respectively (Figure 2). Thus, combining the 6 mutations resulted in increased protein secretion.

Figure 2.

Secretion levels of WT, X5, and X6 in media of cultured CHO cells with a single copy of transgene. Protein concentrations were measured by electrochemiluminescence assay (Methods), where the background signal from untransfected cells (a negative control) was subtracted from the values. Bars show relative protein concentrations (means +/− SD, n = 6) expressed in RUm (Methods). Statistical significance was analyzed using a paired t-test. RUm, relative units of mass. ∗P < .05 ∗∗P < .01 ∗∗∗P < .001 as determined by the test.

In further study, we tested X6 versus WT to assess whether general protein structure and properties were preserved. We considered the specific activity test comprehensive for this purpose, as discussed above. The X5 variant was omitted from the analysis because its analogue had previously been found to have similar specific activity to the nonmutated protein [9].

3.3. Generation of purified X6 and WT

To test whether the 6 mutations affect functional properties of protein (X6), we aimed to compare them with WT properties in subsequent experiments. To generate the proteins, we established a CHO cell line expressing X6, and both proteins were expressed from the respective clonal cell lines and purified. Testing both preparations by PAGE (Figure 3) showed that the major bands corresponded to the heavy chain (HCh; A1-A2-truncated B domains) and light chain (LCh, A3-C1-C2 domains) of the BDD-FVIII, while minor bands of its single chain (SCh; A1-A2-A3-C1-C2 domains) and smaller fragments were also present. The band identities were assessed by molecular weight, WB (Figure S2), and MS. Protein purity was assessed by gel densitometry, resulting in ∼ 97% (WT) and ∼87% (X6) purity. We attribute this difference to apparently higher cleavage of X6 by intracellular or secreted proteases, likely due to mutations affecting the expression pathway (Discussion).

Figure 3.

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis of purified WT and X6. The preparations containing 0.6 RUm, 1.2 RUm, 2.4 RUm, and 4.8 RUm of protein (from left to right respective lanes) were loaded on 4% to 12% (gradient) gel and subjected to electrophoresis followed by Coomassie blue staining. Single chain, heavy chain, light chain, A2-A3-C1-C2, and A1/A2 derived correspond to respective FVIII fragments (domains); von Willebrand factor is bovine protein from cell culture medium; S1, a B-domain-FVIII standard (moroctocog alfa); M, molecular weight markers. Multiplicity of heavy chain/light chain bands is related to structural variability (in glycosylation, etc.). The identity of protein bands was confirmed by using anti-FVIII polyclonal antibodies (Figure S2), MS, and a monoclonal anti-A2 antibody (GMA-012). FVIII, factor VIII; RUm, relative units of mass.

3.5. Assessment of FVIII specific activity of X6 and WT

The specific activity (IU/RUm) of purified X6 was compared with those of WT and 2 commercial BDD-FVIII drug products (controls) using a CS assay to measure FVIII activity, and an ECLA assay to measure protein concentration. Specific activities were found similar for all proteins (Figure 4). When the specific activity was assessed using an OC assay, specific activities of WT and X6 were also similar, though higher than the control samples (Figure S3) and consistent with our previous data for WT [5]. It should be noted that the combination of anti-FVIII antibodies used for protein quantitation was apparently less sensitive to isolated FVIII fragments, assumed to contain the A2 domain, truncated, or with the B domain-like linker, as products of cell culture-derived protease cleavage. Thus, our ECLA assay would be more sensitive to whole FVIII molecule, which would explain the similarity of WT and X6 specific activities. This similarity, in turn, demonstrated that the 6 mutations did not affect the protein cofactor function and the general structure of the molecule.

Figure 4.

Assessment of WT and X6 specific activity using CS assay. Factor VIII activity of X6, WT, and 2 commercial B-domain factor VIII products (S1, moroctocog alfa and S2, turoctocog alfa; controls) was measured by CS assay. Protein concentrations were measured by electrochemiluminescence assay and are expressed in RUm (Methods). Bars show relative specific activities (IU/RUm, means +/− SD, n = 4). Statistical significance was analyzed using paired t-test.

3.6. Assessment of tyrosine sulfation in X6 and WT

Considering the relatively high expression of X6, we investigated whether its overexpression might result in saturation of the PTM machinery, particularly, tyrosine sulfation, which is important for FVIII cofactor activity [15]. This was measured by MS, comparing the degree of sulfation (occupancy) of 6 specific tyrosines, known to be sulfated in FVIII, between WT and X6. We found that sulfation of Y346, Y1664, and Y1680 was noticeably reduced in X6 compared with WT (Figure 5A). Sulfation of each of Y718, Y719, and Y723, located close to each other, could not be measured due to technical limitation of the assay. It was only possible to quantify the respective proteolytic peptide (NTGDYYEDSYEDISAYLLSK) variants bearing 0, 1, or 2 sulfation modifications, but not the tri-sulfated peptide because its peak was not fully resolved. These results showed that cumulative sulfation of Y718, Y719, and Y723 was moderately reduced in X6 compared to WT (Figure 5B). Thus, overexpression of FVIII resulted in a decrease of its tyrosine sulfation under the conditions used.

Figure 5.

Assessment of tyrosine sulfation in WT and X6 by MS. (A) Assessment of sulfation degree of Y346, Y1664, and Y1680. (B) Assessment of cumulative sulfation degree (amount of non-sulfation, mono-sulfation, and di-sulfation; Non-S, Mono-S, and Di-S) at Y718, Y719, and Y723 of proteolytic peptide NTGDYYEDSYEDISAYLLSK. The tri-sulfated version is detectable, but not qualifiable. Bars show relative amount of the peptide bearing variable sulfation sites (means +/− SD, n = 3). Statistical significance was analyzed using a paired t-test. ∗P < .05 ∗∗P < 0.01 as determined by the test.

3.7. Assessment of X6 and WT interactions with VWF and LRP cluster II

To assess possible structural differences between X6 and WT, we examined their interactions with VWF and Cluster II of LRP1, a major FVIII clearance receptor [[16], [17], [18]]. Both ligands were immobilized and tested for binding to WT and X6 using surface plasmon resonance (Figure 6). The results showed that WT and X6 displayed comparable binding affinities for both ligands, suggesting that their respective binding sites are structurally similar. At the same time, the K_D values for VWF binding by both WT and X6 were higher than previously reported K_Ds for WT and plasma-derived FVIII [5,19], and the responses showed some differences. We attribute these variations to differences in experimental conditions.

Figure 6.

Analysis of WT and X6 binding to von Willebrand factor (VWF) and lipoprotein receptor-related protein (LRP) 1 Cluster II using surface plasmon resonance. LRP1 Cluster II (A and B) and von Willebrand factor (VWF) (C and D) were immobilized on a sensor chip at ∼250 Resonance Units. Binding of WT (A and C) and X6 (B and D) was recorded using protein concentrations of 0.5, 1.9, 7.5, 30, and 120 nM. Protein dissociation was monitored by injecting buffer alone. Between each cycle, the sensor surfaces were regenerated as described previously [5]. The red curves represent fits to the data using a 1:1 Langmuir binding model.

3.8. In silico and in vitro assessment of immunogenicity of X6

In additional experiments, we assessed whether the six mutations resulted in increased protein immunogenicity. First, WT and X6 were compared for peptide-MHC class II (MHCII) binding affinities in silico. Both sequences exhibited comparable binding affinity profiles across the panel of tested MHCII alleles (Figure S4).

Next, the immunogenicity of X6 was assessed in vitro by evaluation of human T-cell (antigen-reactive) immune response to WT and X6 using the ELISpot assay with PBMCs derived from 10 individuals carrying MHC-II DRB1 alleles found in ∼25% of the North American population. Upon treatment of PBMCs with either WT or X6, T cell stimulation was evaluated based on the release of IFN-γ. We found that the mean numbers of IFN-γ SFC upon the treatment by WT and X6 were 9.3 and 5.9 per 10⁵ cells respectively, compared with 4.6 SFC per 10⁵ for unstimulated cells. When comparing unstimulated cells with those stimulated by WT and X6, there were no significant differences in response (Figure 7A). Representative IFN-γ positive spots for each condition are shown in Figure 7B. We also compared the percentage of donors showing positive response to WT and X6 by calculating the stimulation index and found that 60% of the donors showed the response when stimulated with WT and 50% of the donors showed the response when stimulated with X6. Thus, we did not observe a significant difference between WT-induced and X6-induced T cell responses (Figure 7C). The results suggest that the 6 mutations may not affect FVIII immunogenicity.

Figure 7.

Determination of T-cells response to WT and X6. (A) Representation of IFN-γ release from 10 individuals stimulated with WT and X6 (US, unstimulated cells). Graph shows the spot-forming cells (SFC/10⁵ cells). Statistical significance was analyzed using paired t-test. Data are presented as mean ± SD. (B) Representative images from ELISpot assay for Unstimulated, SEB (positive control), WT, X6, and FIX (negative control) test results from a particular donor. (C) Response plot of donors showing positive response (highlighted in green) to WT and X6 when SI is ≥ 1.9. ∗P < .05 ∗∗P < .01 ∗∗∗P < .001 as determined.

3.9. Generation and testing expression of a full-length FVIII analogue of X6

To evaluate the effect of the 6 mutations on FL-FVIII expression, we introduced them into the gene, as well as only the single F309S mutation for comparison. HepG2 cells were transfected with plasmids encoding FL-FVIII, FL-FVIII F309S, FL-FVIII X6, and a BDD-FVIII (FVIII-V3 [4], which is known to have a relatively high expression level and was therefore used as a positive control). Seventy-two hours posttransfection, FVIII protein levels in culture media were measured. As shown in Figure 8, expression levels of FL-FVIII F309S and FL-FVIII X6 were comparable with FL-FVIII, whereas expression of FVIII-V3 was markedly higher, consistent with its codon optimization. These data show that the 6 mutations in FL-FVIII do not increase its secretion levels in human hepatocytes.

Figure 8.

Secretion levels of FL-FVIII variants in cultured HepG2 cells. The cells were transfected with a low (A) and high plasmid dose (B) to express FL-FVIII, FL-FVIII F309S, FL-FVIII X6, and FVIII-V3 (a BDD-FVIII, a positive control) or untransfected (Mock, a negative control), and media were measured for FVIII levels by enzyme-linked immunosorbent assay after 72 hours. Protein concentrations are expressed in “RUm” (Methods) (means +/− SD, n = 3).

4. Discussion

Previous studies described the five amino acid substitutions in BDD-FVIII and the single amino acid substitution in FL-FVIII which increased both proteins’ expression. In our study, combining these point mutations in BDD-FVIII (X6) resulted in increased protein expression compared to each parental construct (by 1.6-fold vs X5 and by 2.3-fold vs WT from the single copy transgenes, respectively). Most likely, for a non-codon-optimized gene, such differences would be even higher due to the lower burden of protein expression on the cell’s expression machinery.

Upon assessment of the structural functions of the 6 mutations, we found that 4 positions (I86V, G132K, M147T, and F309S) are located internally within the A1 domain, and 2 positions (A108S and L152P) are surface-exposed (Table S2). The assessment indicates that internal residues’ interactions may result in formation of additional or tighter intramolecular contacts that may, in turn, aid in better protein folding kinetics by favoring the proper conformation, in addition to the known effect of the F309S mutation on FVIII interaction with BiP [7].

At the same time, combining these 6 mutations in FL-FVIII did not result in increased expression. A plausible explanation for this discrepancy is that the FL-FVIII and BDD-FVIII have a difference in biosynthetic mechanisms due to the contributions of the heavily glycosylated B-domain. This hypothesis is supported by studies that demonstrated expression of native FVIII depends on the interactions of B-domain glycans with ER chaperones LMAN1 (lectin and mannose-binding) and MCFD2, which altogether provide cargo transportation of FVIII to the Golgi system [20]. In contrast, expression of BDD-FVIII is LMAN1-independent and does not require this transportation [10,20,21]. Our study results are consistent with these data and suggest that the 6 mutations are favorable only for FVIII expression via the LMAN1/MCFD2-independent pathway.

However, the discrepancy between FL-FVIII F309S expression levels in the study of Swaroop et al. [7] and our study can also be attributed to differences in experimental conditions. In particular, (i) that study used COS-1 cells for protein expression, whereas we used HepG2 cells and (ii) had a different expression strategy for BDD- and FL-FVIII. For BDD-FVIII, we used a lentivirus-mediated delivery of the transgene, stably expressed in a clonal cell line, while for FL-FVIII, we used conventional transfection delivery (due to the inability of a lentivirus to encompass the large gene) followed by transient expression. Also, (iii) the expression of the BDD- and FL-FVIII was conducted in different cell lines that could have affected its mechanisms and outcomes. Although a well-established hepatic model (HepG2 cells) was utilized, comparing BDD- and FL-FVIII under the same conditions would also be worthwhile, considering variations in mechanisms between the species and cell lines. Nevertheless, the results are still in accordance with data showing that the efficiency of FVIII cargo transport via the LMAN1/MCFD2 pathway varies in different cell types [10].

We demonstrated that mutated BDD-FVIII (X6) preserves its functional properties, such as specific activity and binding to VWF and LRP1 Cluster II, which were similar to nonmutated protein. This indicates that the overall structure of X6 is preserved, as these interactions involve large surface areas on the molecule. Indeed, manifestation of FVIII activity requires interactions with multiple ligands (platelets, FIX, and FX [1]) via FVIII sites located on its heavy chain and light chain, and the binding to VWF and LRP1 involves extended areas on the LCh [22,23]. Notably, both WT and X6 demonstrated higher activity (and respectively, specific activity) in the OC assay than in the CS assay resulting in a reversed OC/CS ratio (>1) for both proteins compared with BDD-FVIII therapeutic products (which typically have a ratio of ∼0.7). This reversed ratio of OC/CS has also been described (i) for hemophilia A patients who underwent gene therapy with a BDD-FVIII [24], (ii) for nonmutated protein (WT) [5] and (iii) for another BDD-FVIII variant [25]. Notably, both of the latter studies showed that such reversed assay ratio was not caused by codon optimizations, but rather by unknown structural properties. We recently proposed a general mechanism that causes the OC/CS assay discrepancy for FVIII [18].

It should be noted that at high transgene copy number and thus higher expression level, sulfation of some tyrosines in X6 was moderately lower than in the nonmutated protein, indicating sub-saturation of the cell biosynthetic machinery due to protein overexpression. However, this difference was not found to affect the activity and VWF-binding affinity of X6, whereas both properties were previously shown to depend on FVIII tyrosine sulfation [15]. We attribute this to small differences in the degree of tyrosine sulfation between WT and X6 (10-15%), so their expected differences in both properties were undetectable under the conditions used.

Likely due to the higher expression level, we observed apparently increased cleavage of X6 compared to WT, which we attribute to cell culture proteases. CHO cells are known to express various types of proteases, such as metalloproteases and serine proteases, which have been shown to degrade recombinant proteins [[26], [27], [28]], including FVIII [29]. Indeed, porcine FVIII expression has been shown to be independent of BiP interaction and biosynthetically distinct from human FVIII expression [29]. An additional factor that may further contribute to higher cleavage of X6 is its reduced interaction with BiP due to the F309S mutation.

The assessment of the X6 immunogenicity was performed by (i) an in-silico assessment of the binding affinities of its and WT peptides to 30 HLA Class II variants found in > 90% of the North American population and (ii) estimating T-cell stimulation in PBMCs. Both approaches suggested that X6 is likely not more immunogenic than the nonmutated protein. However, it is important to note that correlations between in silico and in vitro assessments and clinical immunogenicity have not been tested. These data agree with results from Cao et al. [9], who tested the immunogenicity of another BDD-FVIII variant with the 5 mutations (i.e, X5 analogue) in mouse and rat models and in human T-cell culture and did not find significant differences with the nonmutated protein. If further testing confirms the low immunological risk of X6, it will enhance the potential for its use in both FVIII replacement and gene therapies for Hemophilia A. Notably, while the X6 used was based on a codon-optimized gene, none of the current recombinant therapeutic FVIII products is produced using codon-optimization, whereas a recent gene therapy construct approved for treatment of hemophilia A is based on a codon-optimized gene [6].

Furthermore, the 6 mutations in a BDD-FVIII could be combined with other favorable designs to further improve the therapeutic properties of FVIII. For example, the molecule could be made single-chained (which is also useful for gene therapy) or fused with other protein portions such as Fc antibody fragments, XTEN sequences, and VWF D’D3 domains, which are known to extend the protein’s plasma half-life [18] and provide more efficient of treatment of hemophilia A.