Mutagenesis of a Potential Immunoglobulin-binding Protein-binding Site Enhances Secretion of Coagulation Factor VIII*

Coagulation factor VIII (FVIII) and factor V are homologous glycoproteins that have a domain structure of A1-A2-B-A3-C1-C2. FVIII is a heterodimer of the heavy chain (domains A1-A2-B) and the light chain (domains A3-C1-C2) in a metal ion-dependent association between the A1- and A3-domains. Previous studies identified a 110-amino acid region within the FVIII A1-domain that inhibits its secretion and contains multiple short peptide sequences that have potential to bind immunoglobulin-binding protein (BiP). FVIII secretion requires high levels of intracellular ATP, consistent with an ATP-dependent release from BiP. Site-directed mutagenesis was used to elucidate the importance of the potential BiP-binding sites in FVIII secretion. Mutation of Phe at position 309 to Ser or Ala enhanced the secretion of functional FVIII and reduced its ATP dependence. The F309S FVIII had a specific activity, thrombin activation profile, and heat inactivation properties similar to those of wild-type FVIII. However, F309S FVIII displayed increased sensitivity to EDTA-mediated inactivation that is known to occur through metal ion chelation-induced dissociation of the heavy and light chains of FVIII. The results support that Phe309 is important in high affinity heavy and light chain interaction, and this correlates with a high affinity BiP-binding site. Introduction of the F309S mutation into other secretion defective FVIII mutants rescued their secretion, demonstrating the ability of the this mutation to improve secretion of mutant FVIII proteins retained in the cell.

Two structurally related plasma glycoproteins, coagulation factor VIII (FVIII) 1 and factor V (FV), are essential cofactors for the proteolytic activation of factor X and prothrombin, respec-tively. FVIII and FV have similar domain structures of A1-A2-B-A3-C1-C2 (1)(2)(3)(4). The A-domains exhibit 35-40% amino acid identity to each other and to the A-domains of the copperbinding protein ceruloplasmin. Upon transit through the secretory compartment, FVIII is processed to a heterodimer consisting of a carboxyl-terminal derived light chain of 80 kDa (domains A3-C1-C2) in a metal ion-dependent association with an amino-terminal derived 200-kDa heavy chain (domains A1-A2-B) (5). FVIII is inefficiently secreted, and this correlates with its interaction with the protein chaperone immunoglobulin-binding protein (BiP), also known as the glucose-regulated protein of 78 kDa (GRP78) (6,7). BiP is a peptide-dependent ATPase that interacts with unfolded, mutant, or unassembled protein subunits within the ER (8). FVIII release from BiP requires high intracellular concentrations of ATP. Depletion of intracellular ATP by treatment with low concentrations of the protonophore carbonyl cyanide 3-chlorophenylhydrazone (CCCP) specifically inhibited FVIII secretion, whereas secretion of FV was not affected (9,10). Further studies demonstrated that FVIII secretion requires not only high intracellular ATP level but also the ATPase activity of BiP (11).
Previous studies using chimeric FVIII and FV cDNA molecules identified a 110-amino acid region within the FVIII A1domain that either actively retains FVIII in the ER or alters protein folding to reduce FVIII secretion (12). Exchange of this region with the homologous residues in FV yielded an efficiently secreted FVIII/FV chimeric protein that was not active and for which the heavy and light chains were dissociated. Based on the ability of 7-mer peptides displayed in filamentous phage to bind BiP, Gething and co-workers devised a statistical method to predict the BiP binding potential of a particular 7-mer peptide (13). Within the 110-amino acid region that inhibits FVIII secretion a hydrophobic cluster of residues from Ile 291 to Phe 309 contains multiple 7-mer peptides having a high probability of binding BiP (12). We tested the role of these residues in FVIII secretion by their mutagenesis to the respective amino acid present in FV. Mutation of a single amino acid Phe 309 was sufficient to increase the secretion efficiency of biologically active FVIII by severalfold and reduce its ATP dependence for transport out of the cell.

EXPERIMENTAL PROCEDURES
Plasmid Mutagenesis-Site-directed mutagenesis was performed by oligonucleotide overlap extension polymerase chain reaction (PCR). Partially complementary primers that contained the mutation were utilized with two primers directed at the MluI sites at 226 and 336 in the FVIII/FV chimeric cDNA V(227-336) previously described (12) to amplify two overlapping products that contain the directed mutation. Then these two fragments were isolated and fused together by PCR using the two MluI site containing primers. The resultant MluI fragment was then subcloned into the MluI-digested FVIII/FV 226 -336 chimera within the expression vector pED⌬C (14). All mutations were confirmed by DNA sequencing over the PCR-amplified region using the Sequenase kit (U. S. Biochemical Corp.). For every mutation analyzed, two independently isolated mutant clones were analyzed in transfection experiments.
DNA Transfection and Analysis-Expression vectors encoding the indicated mutants were transfected into COS-1 cells by the DEAEdextran procedure (15). Conditioned medium was harvested 60 h posttransfection in the presence of 10% heat-inactivated fetal bovine serum for FVIII assay. Protein synthesis and secretion were analyzed at 60 h post-transfection by pulse-chase labeling of cells (16). Proteins were subjected to SDS-PAGE under reducing conditions and visualized by autoradiography after treatment with En 3 Hance (DuPont Corp., * This work was supported by National Institutes of Health Grants HL53777 and HL52173. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Boston, MA). Band intensities were quantitated with NIH image software.
Derivation of CHO Cell Lines-The full-length chimeric FVIII F309S cDNA contained in the expression vector pED⌬C was linearized by digestion with NheI and introduced into dihydrofolate reductase-deficient CHO cells by lipofection (16). Clones were isolated by growth in nucleoside-free medium and propagated in increasing concentrations of methotrexate as described (17). Clone D56 selected in 1 M methotrexate secreted approximately 1 unit/ml/10 6 cells/day and was chosen for further study.
FVIII Purification-Logarithmically growing CHO cells stably expressing either F309S or wild-type FVIII were fed medium containing 10% fetal bovine serum. FVIII was purified from 200 ml of conditioned medium by immunoaffinity chromatography as described (18). The proteins eluted into the ethylene glycol-containing buffer were dialyzed and concentrated against a polyethylene glycol (molecular weight, ϳ15-20,000)-containing buffer (19) and stored at Ϫ70°C.
FVIII Activity and Antigen Assay-FVIII activity in conditioned medium samples was measured using a chromogenic factor Xa generation assay from Pharmacia Hepar (Franklin, OH). FVIII activity of purified FVIII samples was measured by a one-stage APTT clotting assay by reconstitution of human FVIII-deficient plasma (George B. King, Biomedical Inc., Overland Park, KS). For thrombin activation, protein samples were diluted into 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2.5 mM CaCl 2 , and 5% glycerol and incubated at room temperature with 1 unit/ml thrombin. After incubation for increasing periods of time, aliquots were diluted and assayed for FVIII activity. For heat or EDT inactivation, FVIII samples were maintained at 50°C or in 10 mM EDTA at room temperature, and FVIII activity was determined after increasing periods of time by the one-stage clotting assay. One unit of FVIII activity is the amount measured in 1 ml of normal human pooled plasma. FVIII antigen was quantified using a sandwich enzyme-linked immunosorbent assay method utilizing anti-light chain antibodies ESH-4 and ESH-8 (American Diagnostica, Greenwich, CT) (20). Purified recombinant FVIII protein was used as a standard.

RESULTS
Mutation F309S Improves FVIII Secretion-Residues 226 -336 within the FVIII A1-domain contains a hydrophobic cluster where 7 of 11 amino acid residues are Leu or Phe between residues Leu 299 and Phe 309 . To identify whether this region may inhibit FVIII secretion, we mutated hydrophobic residues in this potential BiP-binding site. Transfection of COS-1 cells with vectors encoding FVIII molecules containing all Leu and Phe residues in the potential BiP-binding pocket mutated to Ala (7LF-A) or to their respective residues in FV (11VIII-V or 8VIII-V) did not yield activity in the conditioned medium (Table I), and this correlated with the absence of detectable protein secreted into the conditioned medium (data not shown). Subsequently, the Phe and Leu residues were individually mutated to the respective homologous amino acid residues in FV. The activity of the F309S mutant was reproducibly increased 2.3fold in several transfection experiments (Table I). The increased activity of the F309S mutant correlated with a corresponding increase in FVIII antigen in at least five separate transfection experiments with two different independently isolated mutant clones, indicating a specific activity similar to wild-type FVIII. Mutation F309A also improved recovery of FVIII in the conditioned medium, indicating that the improved secretion was not selective to the F309S mutation. However, mutation of F309E reduced activity by 30%, which correlated with a 3-fold reduced secretion (data not shown), indicating specificity for the increase in activity that may be limited to small uncharged residues. In contrast, mutation at adjacent Phe or Leu residues had either little effect on (F306W and L299T) or significantly inhibited recovery of (F293S, L294T, and L300V) functional FVIII activity in the conditioned medium. Finally, mutation of C310S, a cysteine that is not disulfide-bonded and is a potential ligand for copper ion within the type I copper-binding site in the A1-domain, dramatically reduced recovery of activity in the conditioned medium.
We then tested whether insertion of additional mutations may further increase FVIII activity detected in the conditioned medium. Because the optimal peptide size for BiP binding is a 7-mer (11), we additionally mutated each of the residues in the hydrophobic pocket adjacent to F309S. The amount of secreted activity for each double mutant was close to that obtained with the F309S mutation alone (Table I). The only double mutant that yielded a slight increase over the F309S mutation alone was the F309S/L303E double mutation (3.0-fold greater than wild type in three separate experiments), which corresponds to a doubly modified 7-mer peptide. Finally, we tested whether introduction of the mutations that improve FVIII secretion could rescue the secretion of other FVIII mutations that destroyed activity recovered in the conditioned medium. Addition of the F309S/L303E double mutation to either the L294T or the L300V mutation increased their secretion close to the FVIII wild-type level (Table I). In contrast, addition of the F309S/L303E mutants into the F293S mutant did not improve secretion. These results show that the F309S/L303E mutations can increase secretion of some but not all secretion defective mutants of FVIII.
Metabolic pulse-chase labeling was performed to characterize whether the increased activity correlated with an increase in synthesis, secretion, or specific activity. Analysis of the pulse-labeled and chase cell extract samples indicated that FVIII wild type and all mutants were synthesized at the same rate and disappeared from the cell extract at similar rates (Fig.  1A). However, analysis of the conditioned medium demonstrated that the F309S mutation increased the recovery of FVIII in the conditioned medium (Fig. 1B). Immunoprecipitation with the anti-heavy chain monoclonal antibody co-precipitated the light chain for all the mutants, demonstrating an association between the heavy and light chains. Additionally, TABLE I Mutagenesis of potential BiP binding site Plasmid DNA encoding each mutant was transfected into COS-1 cells, and FVIII activity in the conditioned medium was assay at 60 h post-transfection. Each transfection was performed with two independently isolated clones, and the average is represented as a percentage of wild type within the same experiment. Several of the mutants (F309S, K303E/F309S, Q305E/F309S, and F309A) were transfected at least three times, and the variation between transfections compared with the wild-type FVIII was less than 10%. The results presented are from one experiment where cells were also pulse-chase labeled for analysis of FVIII secretion (Fig. 1 the amount of F309S mutant FVIII recovered in the conditioned medium was similar to the amount of the FVIII-FV hybrid V(227-336) (Fig. 1B, lane 1). In contrast, secretion of F306W mutant FVIII was not different from the wild-type FVIII. The relative amount of secreted FVIII protein observed correlated with the amount of activity recovered in the conditioned medium (compare Table I with Fig. 1B). Mutation of L294T inhibited FVIII secretion, in proportion to the reduced activity (Table I), and additional mutation of L303E and F309S improved secretion of the mutant FVIII to a level similar to that of wild type (Fig. 1C). Importantly, because wild type and all mutants were synthesized at the same rate, differences in transfection efficiency or expression cannot account for the differences in the amount of secreted protein recovered in the conditioned medium. These results demonstrate that the increased activity for all FVIII molecules harboring the F309S mutation resulted from increased secretion compared with wild-type FVIII. The increased secretion was almost equal to that of the FVIII-FV hybrid molecule V(227-336).
BiP Interaction and Secretion of Mutant F309S FVIII-To characterize the secretion of the F309S mutant FVIII in more detail, stably transfected CHO cell lines were derived that express the hybrid FVIII-FV V(227-336) protein or the F309S mutant FVIII. Of 35 original transfected CHO cell clones selected for dihydrofolate reductase expression, we obtained five clones that express significant levels of F309S FVIII (greater than 1 unit/ml/10 6 cells/day). Two of these clones express greater levels of FVIII compared with the original 10A1 cell line that was obtained by screening over 1000 original transfected cell clones (5). Thus, at this initial stage of selection in low concentrations of methotrexate, the F309S mutation permits high level FVIII expression to be obtained more readily. Pulse-chase labeling with [ 35 S]methionine and immunoprecipitation with anti-FVIII antibody demonstrated that FVIII wild type, V(227-336), and F309S mutant FVIII molecules were synthesized and secreted from CHO cells at similar rates with no significant difference in BiP association (data not shown).
To further characterize the BiP interaction we analyzed the effect of ATP depletion on the transport of FVIII wild type, V(227-336), and F309S mutant FVIII from CHO cells. Cells were pulse-labeled with [ 35 S]methionine and chased for 6 h in the presence of increasing concentrations of CCCP. Cell extracts were harvested and immunoprecipitated with anti-FVIII or anti-FV antibody for analysis by SDS-PAGE. Whereas transport of the majority of wild-type FVIII was inhibited by 10 M CCCP, FV required higher concentrations to significantly inhibit transport. The V(227-336) hybrid molecule also required high levels of CCCP (250 M) to inhibit its transport (Fig. 2). The single point mutant F309S displayed an ATP dependence that was intermediate between wild-type FVIII and FV. These results are consistent with the interpretation that the F309S mutant FVIII binds BiP but with an apparent lesser affinity and that this interaction does not require high levels of ATP for release.
F309S Mutant FVIII Displays Increased Sensitivity to EDTA Inactivation-The properties of the F309S mutant FVIII were studied by purification of this mutant from conditioned medium from CHO cells. Wild-type FVIII was purified in parallel for control. The specific activity determined by the factor Xa generation assay and enzyme-linked immunosorbent assay of wild-type and F309S FVIII were not significantly different (approximately 2500 units/mg). F309S mutant FVIII displayed a thrombin activation very similar to that of wild-type FVIII (Fig. 3A). The thermal stabilities of wild-type and F309S FVIII, tested by treating the purified proteins at 50°C for increasing periods of time, were similar (Fig. 3B). However, F309S FVIII was inactivated by EDTA treatment at an approximately 10fold greater rate than wild-type FVIII (Fig. 3C). These results show that F309S mutation did not alter FVIII specific activity, thrombin activation, or thermal inactivation. However, this mutation did significantly increase the sensitivity to EDTAinduced inactivation.

DISCUSSION
Analysis of expression of FVIII, FV, and deletion and chimeric proteins demonstrated that sequences within the carboxyl-terminal half of the A1-domain in FVIII inhibit secretion (12). This region is also required for heavy and light chain association and contains a hydrophobic pocket having peptide sequences that have the potential to bind BiP. Mutagenesis of a single amino acid residue, Phe 309 , improved FVIII secretion to a level comparable with that of the FVIII/FV chimeric protein V(227-336). In addition, the secretion of the resultant molecule displayed a reduced requirement for ATP, following reduction of intracellular ATP levels by treatment of cells with the protonophore CCCP. We interpret these findings in either one of two ways. First, the Phe 309 mutant FVIII may exhibit a weaker interaction for BiP, and its release may therefore require fewer molecules of ATP. Alternatively, the mutant molecule may exhibit improved folding properties so that it more quickly attains its final conformation required for transport from the cell and therefore is bound to BiP for a lesser period of time. Dramatically, introduction of the F309S mutation into FVIII molecules that were defective for secretion was able to rescue their secretion defect. This suggests that hemophilia A associated with mutations that reduce FVIII secretion may be improved by introduction of the F309S mutation into the FVIII gene in these patients.
The F309S mutant FVIII was characterized to exhibit thrombin activation and thermal lability indistinguishable from wildtype FVIII. However, the molecule did exhibit a 10-fold increased sensitivity to EDTA-induced inactivation. EDTA is known to inactivate FVIII through chelation of a metal ion that is required for association of the A1-domain in the heavy chain with the A3-domain in the light chain. FVIII contains 1 mol of copper/mole of protein, and copper ion is only detected when the heavy chain is associated with the light chain (21). Mutation of C310S in wild-type FVIII produced an inactive molecule in which the heavy and light chains were dissociated (22 ). The requirement for Cys 310 in FVIII and its absence in FV indicate that the requirements for folding and final structures of these two homologous molecules may be distinctively different. It is intriquing that F309S is adjacent to a ligand in the type I copper-binding site that may be essential for copper ion-dependent A1-and A3-domain interaction.
Recently, a structural model of the FVIII A-domains was proposed based on the recently elucidated structure of ceruloplasmin (23,24). The proposed structure places the phenyl side chain of Phe 309 3.7 Å from Phe 536 in the A2-domain, suggesting a potential hydrophobic interaction between the A1-and A2-domains. If this model is correct, mutation of F309S should weaken the A1-and A2-domain interaction. Our experimental data on the characterization of the F309S mutant FVIII show that thrombin inactivation kinetics are similar for wild-type and F309S mutant FVIII. Because inactivation of thrombinactivated FVIII occurs as a consequence of A2-domain dissociation (25), our experimental evidence does not suggest a weakened interaction between the A1-and A2-domains. Further experimental evidence should elucidate the validity of the structural model.