Enhancement of Human Protein C Function by Site-directed Mutagenesis of the γ-Carboxyglutamic Acid Domain*

This study reports properties of site-directed mutants of human protein C that display enhanced calcium and/or membrane binding properties. Mutants containing the S11G modification all showed increased affinity for membranes at saturating calcium concentration. Ser-11 is unique to human protein C, whereas all other vitamin K-dependent proteins contain glycine. This site is located in a compact region of the protein, close to a suggested membrane contact site. Additional changes of H10Q or S12N resulted in proteins with lower calcium requirement for membrane contact but without further increase in membrane affinity at saturating calcium. Mutations Q32E and N33D did not, by themselves, alter membrane affinity to a significant degree. These mutations were included in other mutant proteins and may contribute somewhat to higher function in these mutants. This family of mutants helped discriminate events that are necessary for protein-membrane binding. These include calcium binding to the free protein and subsequent protein-membrane contact. Depending on conditions of the assay used, the mutants displayed increased activity of the corresponding activated protein C (APC) derivatives. The degree of enhanced activity (up to 10-fold) was dependent on the concentration of phospholipid and quality of phospholipid (± phosphatidylethanolamine) used in the assay. This was expected, because APC is active in its membrane-associated form, which can be regulated by changes in either the protein or phospholipid. As expected, the largest impact of the mutants occurred at low phospholipid concentration and in the absence of phosphatidylethanolamine. The anticoagulant activity of all proteins was stimulated by protein S, with the greatest impact on the enhanced mutants. Whereas plasma containing Factor V:R506Q was partially resistant to all forms of APC, the enhanced variants were more active than normal APC. Protein C variants with enhanced function present new reagents for study of coagulation and may offer improved materials for biomedical applications.

P is free protein, PL is protein binding sites on the membrane, and P*PL is the complex. In addition to variations in membrane affinity at saturating calcium (step 3 of Equation 1), members of the vitamin K-dependent protein family have different affinities for calcium (step 1 of Equation 1 (3)), making it possible to enhance protein from this stand point as well. In fact, it may be possible to find mutants that influence proteincalcium binding (step 1 of Equation 1) or membrane association at saturating calcium levels (step 3 of Equation 1), without influencing the other step. Such mutants would help identify the membrane contact region of the Gla domain and distinguish it from calcium binding to the free protein.
Full-length human protein C cDNA was cloned in the pRc/CMV vector using HindIII and XbaI sites. PCR amplification and mutagenesis of the target DNA was performed by the following steps: primers A and C were used to amplify the 5Ј part of protein C (up to the codon for the amino acid at position 38), whereas primers B and D were used to amplify the 3Ј part (from the amino acid 27 codon to the end of the protein C cDNA). These two fragments were used as templates to amplify the full-length E32D33 mutant protein C cDNA with primers A and B. HC-E32 was obtained in a similar way, but with primers E and F instead of primers C and D. PCR products were cleaved with HindIII and SalI, and the HindIII-SalI fragment was isolated and ligated together with a SalI-XbaI fragment from protein C cDNA into HindIII and XbaI cleaved pRc/CMV vector.
To prepare the four mutants, G11E32D33, E10G11E32D33, Q10G-11E32D33 and G11N12E32D33, first human protein C cDNA was used as template for the PCR amplification to prepare mutant G11 (using primers A and G), E10G11 (using primers A and H), Q10G11 (using primers A and I), and G11N12 (using primers A and J). The PCR products were digested by HindIII (at the 5Ј of human protein C cDNA) and BsrBI (located in the primer nucleotide sequence at codons for amino acids Ser-14 and Gly-15). These fragments were ligated to a protein C fragment (BsrBI-XbaI, around 1200 base pairs) containing the E32D33 mutation to obtain the full-length hC-G11E32D33 and hC-E10G11E32D33 mutants. All mutations were confirmed by DNA sequencing before transfection. Cell culture in human 293 cells, expression, purification, and quantitation of protein C molecules were performed as described (4).
Recombinant proteins (Ն100 g) were analyzed by high resolution ion exchange chromatography as described previously (5). All proteins showed a single primary peak, without indication of heterogeneity. Elution time varied by approximately 0.6 min per charged group, with a standard deviation of 0.1 min. For example, wild type human protein C and the hC-P10 mutant (4) eluted at 37.9 min, the hC-E32 mutant at 39.3 min, the hC-E32D33 mutant at 39.9 min, and the hC-E10G11E32D33 mutant at 40.8 min. High resolution ion exchange chromatography of chymotrypsin digests of the proteins also gave a single component from the Gla domain, suggesting full carboxylation of Glu residues in positions 1-41.
Preparation of APC-Approximately 0.1 mg of protein C (1 g/liter) was mixed with human thrombin (3 g) (Sigma) and incubated at 37°C for 3 h in standard buffer (0.05 M Tris, pH 7.5, 0.1 M NaCl). The reaction product was diluted with an equal volume of water and applied to a 1-ml column of SP-Sephadex C-50. The column was washed with standard buffer, and the flow-through was pooled as activated protein C (APC). APC activity was determined by hydrolysis of 0.1 mM S2366 (L-pyroglutamyl-L-prolyl-L-arginine-p-nitroaniline hydrochloride) and gave a constant activity per mg of protein for all preparations (Ϯ10%). Concentrations were normalized to S2366 activity.
Coagulation Assays-Coagulation was measured with APTT reagent (Sigma) (0.05 ml of APTT reagent that was diluted 1:10 or 1:100 with standard buffer containing 0.1 mM ellagic acid) or diluted TP (Sigma) (0.1 ml of TP-high sensitivity that had been diluted 1:100 with standard buffer containing 0.01 M CaCl 2 ). The total volume of assays was 0.2 ml. For APTT assays, 0.05 ml of plasma, 0.05 ml of APC and/or protein S solution, and APTT reagent were incubated at 37°C for 3 min. Coagulation was initiated with 0.05 ml of standard buffer containing 0.025 M CaCl 2 . For TP assays, plasma, APC, and/or protein S solution were incubated at 37°C for 3 min, and clotting was started with diluted TP. Time to form a clot was determined by a manual method. Clotting times were reported as the average and range of two determinations. Plasma samples used were human plasma (Sigma), protein S-deficient plasma (American Diagnostica, Greenwich, CT), and heterozygous and homozygous Factor V:R506Q plasma (generously provided by Dr. Nigel Key, Department of Medicine, University of Minnesota Medical School).
A one-stage coagulation assay was carried out using bovine factor Xa. Buffer (total volume of 112.5 l, containing 6.7 mM CaCl 2 ) was mixed with factor Xa, sufficient to give a clotting time of 30 Ϯ 2 s under the conditions of the assay, APC, and phospholipid. The mixture was incubated for 1 min at 37°C, and coagulation was initiated by addition of 37.5 l of normal human plasma (at 37°C).
Vesicle Preparation-Small unilamellar vesicles (SUVs) were prepared by methods described (5,16). Highly pure phospholipids with natural fatty acid distributions included phosphatidylserine (PS, from bovine brain, purchased from Avanti Polar Lipids) and phosphatidylcholine (PC, from egg, purchased from the Sigma) and phosphatidylethanolamine (PE, from egg, purchased from Avanti). These were mixed in organic solvent, dried by a stream of argon gas, and suspended in standard buffer (20 mg in 3 ml). SUVs were formed by sonication and gel filtration. Large unilamellar vesicles (LUVs) were prepared by extrusion as described by Hope et al. (17). Phospholipids included PS/ PC/PE (natural abundance fatty acids, described above). This PE preparation contained 60% plasmalogen. Alternatively, LUVs were made from 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoserine, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine, with and without 1,2-dilinoleoyl-snglycero-3-phosphoethanolamine. All synthetic lipids were from Avanti. Vesicles containing dilinoleoyl-PE were stored under argon and were used within 24 h of preparation. Others were used within 5 days. Sonication of LUVs was carried out essentially as described for preparation of SUVs. Vesicle size was measured by quasi-elastic light scattering with an LSA2 laser light scattering apparatus coupled with a Langley Ford 1096 autocorrelator. Average diameters for SUVs were 35-40 nm, and average diameters for LUVs were 110 -120 nm. Phospholipid concentrations were determined by organic phosphate assay assuming a phosphorous:phospholipid weight ratio of 25 (18).
Measurement of Protein-Membrane Interaction-Light scattering at 90°to the incident light was used to quantitate protein-membrane binding (19). Briefly, the intensity of phospholipid vesicles alone (I 1 ) and after addition of protein (I 2 ) were measured. Background from buffer and unbound protein was subtracted. The ratio of the molecular weight of the protein-vesicle complex (M 2 ) to that of the vesicles alone (M 1 ) was estimated from the following relationship: I 2 /I 1 ϭ (M 2 /M 1 ) 2 (Ѩn/ Ѩc 2 /Ѩn/Ѩc 1 ) 2 , where Ѩn/Ѩc is the refractive index of the respective species.
If phospholipid and protein concentrations are known, the concentration of bound P*PL and free protein P can be estimated. These values, together with the maximum protein binding capacity ([P*PL max ]) of the vesicles (assumed to be 1.0 g/g for all proteins) was used to obtain the dissociation constant for protein-membrane interaction from the following relationship: where all values are in molar concentrations of protein or protein binding sites. Equilibrium binding constants are reported as K D values and are primarily valuable for comparisons. Values at 5 mM calcium were calculated at a total protein:phospholipid ratio of 2.0:1.0 (w/w). Values at 2 mM calcium were determined at a total protein:phospholipid ratio of 0.7:1.0 (w/w).
Protein binding to LUVs is reported as the light scattering intensity ratio (I 2 /I 1 ). Vesicle size is a significant fraction of the wavelength of light so that nonideal light scattering may occur. However, when all solution parameters, vesicle sizes, and distributions were the same for titrations of two similar proteins, the light scattering intensity ratio provided an indication of relative binding by the two proteins. For several reasons, direct comparison of two proteins by this method is more accurate than absolute binding constants obtained (19). The variation of membrane binding constants reported in separate studies (reviewed in Ref. 7) may partially reflect these factors. Direct comparisons between proteins in a single study is an important aspect of this study.
Other Proteins-The bovine proteins factor X (20) and prothrombin (21) were purified by published procedures. Activation of factor X was performed as described (22). Human protein S was generously provided by Dr. Walter Kisiel, Department of Pathology, University of New Mexico.

RESULTS
Membrane Association-Protein-membrane interaction was measured at two phospholipid compositions. Lower PS content provided better comparison of high affinity proteins (Fig. 1B), whereas higher PS content gave better evaluation of low affinity proteins (Fig. 1A). Bovine factor X is included as a standard of comparison. Membrane binding by the hC-E32 mutant was indistinguishable from that of wild type human protein C (Fig.  1), in agreement with earlier reports of a hC-L31E32 mutant (15). However, further mutation of Ser-11 to Gly (hC-G11E32D33) resulted in substantial improvement (3-7-fold decrease in K D ; Table I). Fig. 2 shows the calcium titration of membrane binding by the various proteins. The midpoint of the titration curve for the hC-G11E32D33 mutant was quite similar to that of wild type protein. However, further modifications of H10Q or S12N resulted in lowered calcium requirement. Thus, the relative efficacy of these proteins should differ with the calcium concentration used in the assay. The impact of calcium concentration is illustrated in Table I. For example, the difference between hC-G11E32D33 and hC-Q10G11E32D33 is greater at 2 mM calcium than at 5 mM calcium.
Activity of APC in Standard Coagulation Assays-The activity of APC-E32 was similar to that wild type APC in a diluted thromboplastin assay (Fig. 3A, also see Ref. 13). This assay showed that APC-G11E32D33 was about 10-fold superior to wild type protein and that the APC-Q10G11E32D33 and APC-G11N12E32D33 mutants had approximately 12-fold higher activity.
The relative activity was dependent on the commercial batch of thromboplastin used. Fig. 3B shows approximately 5.5-fold difference between the APC-G11E32D33 and APC-E323 mutants. In keeping with the slight differences of membrane binding by other mutants, APC-E32D33 was approximately equal to APC-E32 and APC-E10G11E32D33 was approximately equal to APC-G11E32D33 (Fig. 3B). Plasma containing Factor V:R506Q (heterozygote and homozygote; for review on APC resistance see Ref. 23) showed resistance to both APC-E32 and APC-G11E32D33 (Fig. 3B, inset). Nevertheless, APC-G11E32D33 was more effective than APC-E32 in these plasmas.
Activity in a diluted APTT was also determined (data not shown). This assay relies on the intrinsic coagulation pathway and therefore utilizes factor VIIIa. APC inhibits both Va and VIIIa, and it was of interest to determine whether a difference was observed for the mutants in these situations. As for the thromboplastin assay, larger differences between APC-G11E32D33 and APC-E32 occurred in diluted APTT reagent and/or at longer coagulation times. There was minimal impact of the mutations when full strength APTT reagent was used. Greatest impact of enhanced membrane affinity at low coagulation stimulus was common to other proteins with improved membrane affinity, including bovine protein C (4) and  a Determined at a protein:phospholipid ratio of 0.70 as described under "Experimental Procedures" (data not shown). Triplicate determinations gave a maximum range of determinations of Ϯ10% for low affinity proteins and less for higher affinity proteins.
b Determined from bound and free protein at a protein:phospholipid ratio of 2.0 (data shown in Fig. 1A). Triplicate determinations gave a range of determinations of Ϯ10% for low affinity proteins and less for higher affinity proteins. ND, not determined.

Factor VII (5).
Sensitivity to Protein S-Both coagulation assays were sensitive to protein S but showed somewhat different requirements (Fig. 4). The diluted APTT test showed a midpoint of about 7 nM protein S, whereas the diluted TP assay showed a midpoint at 30 -40 nM protein S. In both assays and in the protein S titrations (Fig. 4), APC-G11E32D33 showed continual increase of inhibition at higher protein S concentrations. A difference in APC concentration requirement was also apparent; inhibition in the diluted APTT assay occurred at 0 -2 nM APC but required 0 -15 nM APC in the diluted TP reaction (Fig. 3).
Impact of Phospholipid Type-Phospholipid concentration and composition have an impact on coagulation. Although attention is normally paid to the need for anionic lipids, an interesting effect was observed for neutral phospholipids, PC versus PE. The presence of PE in the membrane has a large synergistic impact on function of PS for membrane binding by vitamin K-dependent proteins (24). Similar effects were seen for diverse proteins such as protein kinase C, annexins (25), pentraxins (26), myristoylated alanine-rich C kinase substrate (27), and blood clotting factor VIII (28).
Enhanced protein binding to membranes containing PE was evident from the comparison shown in Fig. 5A. Phospholipids containing natural fatty acid distributions gave similar results for both wild type and the h C-E32 mutant (not shown). The impact of PE was apparent for hC-G11E32D33 (Fig. 5A).
PE also enhances coagulation reactions, such as those involving factor VIIa (29). The importance of PE was shown in this study by the fact that membranes of PS/PC (5/95) were not competent to support coagulation in the assay shown in Fig.  5B, whereas PS/PE/PC (5/50/45) was an excellent membrane source (data not shown).
Phosphatidylethanolamine also impacts on anticoagulation by APC. It supports higher activity of APC against factor Va in a purified system (30). More interesting is a selective promotion of anticoagulation. That is, APC prolongs coagulation to a higher extent in membranes containing PE, under conditions where procoagulation (clotting time) in the absence of APC is constant (30). This study supported that finding, although with a smaller impact of PE. Synthetic phospholipids (Fig. 5, B and C), as well as purified phospholipids (data not shown), showed higher anticoagulant activity for APC in membranes containing PE than in those containing only PC (Fig. 5, B and C). The difference varied somewhat with phospholipid preparation and storage time. The largest difference (about 4-fold) was observed in preparations that had been stored at 4°C for 2 months.
An important implication of these comparisons was that the enhanced activity of the site-directed mutants differed with the phospholipid concentration and composition used in an assay. The effect of PE on wild type APC was highest at low phospholipid concentration (Fig. 5C), whereas the APC-G11E32D33 mutant showed small differences at the different phospholipid concentrations (compare clotting time versus APC-G11E32D33 concentrations in Fig. 5, B and C). APC-G11E32D33 was also relatively insensitive to the presence of PE (Fig. 5, B and C). As a result, the enhanced function of mutants of higher activity will differ with the conditions selected for the assay. High membranes concentrations or the presence of PE reduced the apparent effect of the mutations. This outcome was qualitatively similar to the impact of a chimera protein C that contained residues 1-44 of human prothrombin (30), except that PE virtually eliminated the difference between wild type and chimera APC proteins. DISCUSSION Earlier predictions (2) and the outcome for other proteins (4,5) suggest that enhanced vitamin K-dependent plasma proteins can be created by site-directed mutation of the Gla domain to enhance membrane binding affinity. This study showed success in generation of human protein C with enhanced membrane binding. Several results indicated that human protein C has unique features among this family of proteins. For example, bovine protein C responded to incorpo- The open symbols used the diluted APTT assay, and the filled symbols used the diluted TP assay (described in Fig. 3). APC and S were added 3 min prior to the initiation of coagulation. Inhibition of coagulation was reported as the ratio of clot time in the presence of protein S to an assay in the absence of protein S. Clotting times without protein S for APC-G11E32D33 and APC-E32 were 105 and 89 s (TP assay) and 210 and 140 s (APTT assay), respectively. ration of E32D33 (from Q32N33 in the wild type protein) by a large increase in membrane binding affinity, 2 whereas human protein C showed no change for the Q32E mutation (14) and only a small change for addition of the N33D mutation (this study). Human factor VII showed greatly improved membranebinding affinity upon modification of Pro-10 to Gln and Gln-32 to Glu (5). Other evidence of uniqueness of human protein C is the replacement of Gla-29 by Asp, which had low impact on human protein C interaction with membranes (14) but caused major loss of membrane binding by human prothrombin (31) and human factor X (32). Thus, identification of residues that can be altered to improve membrane interaction by human protein C is an important step to understanding the unique properties of this protein. It also provides methods to potentiate this important regulator of blood coagulation for research and biomedical applications.
The S11G mutation generated protein C with higher membrane affinity at saturating calcium levels. This mutation therefore appeared to alter the actual membrane contact site on the protein. This could occur by a direct impact of glycine-11 on membrane interaction or by indirect improvement of other positions in the protein. Other residues could include placement of E32D33 to a more optimum location that allowed them to enhance membrane contact. Although the actual site of membrane contact is still unknown, it can be noted that this site is adjacent to position 10, one of the three residues that appeared to regulate most of the diversity in membrane affinity among the naturally occurring proteins (sites 10, 32, and 33 (2)). Position 11 is in a compact region of the protein that is closely placed to Ca-2 in the x-ray crystal structure (33). The Ser-11 hydroxyl may function as a ligand to the calcium ion, thereby modifying structure in this region. Alternatively, the serine side chain may not allow optimum protein folding due to steric factors. An importance of position 11 is also suggested by low activity of a Gly (position 12 of factor IX) to Arg mutation of factor IX (34).
Other site-directed mutations produced proteins that displayed a different type of enhanced function. For example, an additional change of H10Q or S12N resulted in proteins with a lower calcium requirement for membrane association, but with nearly identical membrane binding at saturating calcium levels. This suggested that Gln-10 and Asn-12 modify calcium binding to the protein without participation in actual membrane contact. These mutants should show higher activity at low calcium concentrations, but identical activity at saturating calcium. In the assays used, these mutants were similar to APC-G11E32D33, indicating that calcium levels were saturating in these assays. The calcium concentration under in vivo conditions is not known, and whole animal experiments may be needed to determine whether improvements in calcium binding to the protein can aid function.
In all cases and with all membranes used, improved enzyme activity of APC correlated with increased membrane affinity. In fact, the behavior appeared to fit to a general equilibrium expression for assembly of the active APC complex. Although the correct binding expression for APC function as an anticoagulant may be quite complex, it should contain protein and phospholipid components such as those given in boldface in Equation 2. will enhance the activity of APC. This is a common proposal for protein S function; by enhancing membrane bound APC, activity of APC is increased. Other ways in which the ratio of [P free ]/[P*PL] can be decreased include the addition of more PL to the reaction. Indeed, higher phospholipid levels in the assay gave higher APC activity (compare APC-wild type activity in Fig. 5A versus Fig. 5B). An alternative way to lower [P free ]/ [P*PL] is with membrane content that results in higher affinity. This can include an increase in PS or use of PE as a portion of the neutral phospholipid. Again, improvements in the phos- pholipid, by inclusion of PE,enhanced the activity of APC (Fig.  5). A third way of reducing [P free ]/[P*PL] is by improvements of the membrane binding site on the protein. Higher activity by APC-G11E32D33 illustrated this property. Finally, these three approaches to higher activity of APC impact on the same phenomenon, the ratio of [P free ]/[P*PL], and a maximum limit for activity is reached when [P free ] becomes zero. Lowered [P free ]/ [P*PL] by one mechanism will decrease the potential impact of other mechanisms. As expected from this interrelationship, the highest impact of the enhanced mutants was observed at low concentrations of membranes that did not contain PE (Fig. 5C). The largest impact of PE was observed with low affinity proteins (Fig. 5, B and C) and was almost undetected for high affinity proteins. The impact of phospholipid concentration was greatest for low affinity proteins and was minimal for the APC mutants with high membrane affinity (compare Fig. 5B with Fig. 5C). Due to these properties, it is possible that neither the impact of PE nor the increased activity of the mutants will be detected in assays that use high phospholipid concentrations. Standard clotting assays often use an optimum phospholipid concentration, and that may mask the benefit of both PE and the mutant proteins.
The pattern of protein S impact may suggest a function more complex than simple enhancement of membrane-bound APC. Rather than reducing the importance of protein S, APC mutants with increased membrane affinity showed a higher response to protein S that did wild type protein. In addition, the two coagulation assays showed different requirements for protein S. The thromboplast in assay is sensitive to components of the prothrombinase complex and the major site of APC function is factor Va. This assay required a higher protein S concentration for maximum activity than did the APTT assay. The latter relies on the intrinsic pathway, including factor VIIIa/IXa. Thus, the result of protein S titrations with APTT versus thromboplastin assays (Fig. 4) suggested that protein S interacted somewhat differently for APC action on factor VIIIa versus factor Va.