Tài liệu Báo cáo khoa học: Mouse recombinant protein C variants with enhanced membrane affinity and hyper-anticoagulant activity in mouse plasma pptx

In a tissue factor-initiated thrombin generation assay using mouse plasma, all mouse APC variants, including wild-type, could completely inhibit throm-bin generation; however, one of the

membrane affinity and hyper-anticoagulant activity in

mouse plasma

Michael J Krisinger1, Li Jun Guo1, Gian Luca Salvagno2, Gian Cesare Guidi2, Giuseppe Lippi2 and Bjo¨rn Dahlba¨ck1

1 Department of Laboratory Medicine, Division of Clinical Chemistry, Lund University, University Hospital, Malmo¨, Sweden

2 Clinical Chemistry Section, Department of Morphological-Biomedical Sciences, University Hospital of Verona, Italy

Introduction

Protein C is a vitamin K-dependent c-carboxyglutamic

acid-containing protein (Gla protein) found in human

and mouse plasma at a concentration of approximately

70 nm [1] This zymogen is efficiently converted by the

thrombin–thrombomodulin complex to the multifunc-tional serine protease activated protein C (APC) With its cofactor, protein S, APC degrades factors Va and VIIIa on anionic phospholipid membranes, thereby

Keywords

anticoagulation; Gla domain; mouse protein C;

mouse plasma; protein–membrane

interactions

Correspondence

B Dahlba¨ck, Department of Laboratory

Medicine, Division of Clinical Chemistry,

Wallenberg Laboratory, Entrance 46, Floor

6, Lund University, University Hospital,

S-20502 Malmo¨, Sweden

Fax: +46 40 337044

Tel: +46 40 331501

E-mail: bjorn.dahlback@med.lu.se

(Received 1 July 2009, revised 4 September

2009, accepted 9 September 2009)

doi:10.1111/j.1742-4658.2009.07371.x

Mouse anticoagulant protein C (461 residues) shares 69% sequence identity with its human ortholog Interspecies experiments suggest that there is an incompatibility between mouse and human protein C, such that human protein C does not function efficiently in mouse plasma, nor does mouse protein C function efficiently in human plasma Previously, we described a series of human activated protein C (APC) Gla domain mutants (e.g QGNSEDY-APC), with enhanced membrane affinity that also served as superior anticoagulants To characterize these Gla mutants further in mouse models of diseases, the analogous mutations were now made in mouse protein C In total, seven mutants (mutated at one or more of positions P10S12D23Q32N33) and wild-type protein C were expressed and purified to homogeneity In a surface plasmon resonance-based membrane-binding assay, several high affinity protein C mutants were identified In

Ca2+ titration experiments, the high affinity variants had a significantly reduced (four-fold) Ca2+ requirement for half-maximum binding In a tissue factor-initiated thrombin generation assay using mouse plasma, all mouse APC variants, including wild-type, could completely inhibit throm-bin generation; however, one of the variants denoted mutant III (P10Q⁄ S12N⁄ D23S ⁄ Q32E ⁄ N33D) was found to be a 30- to 50-fold better anti-coagulant compared to the wild-type protein This mouse APC variant will

be attractive to use in mouse models aiming to elucidate the in vivo effects

of APC variants with enhanced anticoagulant activity

Abbreviations

APC, activated protein C; C max , maximal concentration of thrombin; ETP, endogenous thrombin potential; Gla protein, c-carboxyglutamic acid-containing protein; DOPS, 1,2-dioleoyl-sn-glycero-3-[phospho- L -serine]; FU, fluorescence units; PE, phosphatidylethanolamine; POPC, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine; POPE, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine; PS, phosphatidylserine;

R max , maximum surface coverage; RU, response units; SPR, surface plasmon resonance; T max , time required to reach maximum thrombin generation.

efficiently turning off the major driving force of

coag-ulation Although historically known for its role in

anticoagulation, APC was recently revealed to have

cytoprotective, anti-inflammatory and anti-apoptotic

functions These new functions of APC are related

to the ability of APC to bind endothelial protein C

receptor and activate protease activated receptor 1,

triggering intracellular signaling [2–4] Moreover,

recombinant human APC was recently shown to

inhi-bit integrin-mediated neutrophil migration by a direct

interaction with leukocyte b1 and b3 integrin receptors

[5] APC appears to play a central role in the

patho-genesis of sepsis and associated organ dysfunction In

patients with sepsis, the APC system malfunctions at

almost all levels First, plasma levels of the zymogen

protein C are low or very low because of impaired

syn-thesis, consumption and degradation by proteolytic

enzymes such as neutrophil elastase [6] Furthermore,

significant down-regulation of thrombomodulin caused

by pro-inflammatory cytokines such as tumor necrosis

factor-a and interleukin-1 has been demonstrated,

resulting in diminished protein C activation [7] The

protective effects of APC supplementation in patients

with severe sepsis complicated with disseminated

intra-vascular coagulation [8] remain to be fully elucidated

and are likely the result of its ability to modulate

multiple biochemical pathways [7]

A prerequisite for Gla protein-membrane binding is

the saturation of seven Ca2+ sites in the N-terminal

Gla domain, which changes its tertiary structure from

an unfolded and nonfunctional conformation to a

tightly folded membrane-binding domain [9,10] This

Ca2+ binding requires the presence of Gla residues

The Gla domains within the protein C family comprise

44 amino acids and contain between nine and 11 Gla

residues, which mediate the Ca2+ interaction In

human protein C, a detailed analysis of the function of

each of these Gla residues has been evaluated [11] Of

the Gla residues, nine are strictly conserved

through-out the Gla proteins From crystal structures of the

Gla domain of prothrombin and factor VIIa, the

placement of the seven Ca2+ in relation to their Gla

ligands is almost identical in the two proteins [12] The

conformational transition induced by the cooperative

binding of Ca2+ turns the N-terminal part of the Gla

domain inside out, exposing the hydrophobic x-loop

to solvent and burying the majority of the Gla

resi-dues Of the seven Ca2+, the majority are buried and

are integral to maintain the membrane-binding

confor-mation [9,13] A few of the Gla-bound Ca2+are

acces-sible to solvent and may play a role in membrane

binding A membrane-bound structure of a Gla

pro-tein does not exist, hampering our understanding of

how the Gla domain engages and reversibly binds to a membrane surface Thus, it is also unclear how the Gla domains of the previously engineered mutants (e.g human QGNSEDY-APC; see below) have been selectively altered to enhance membrane binding How-ever, it does appear that electrostatic, hydrophobic and specific lipid headgroup interactions are all involved in mediating the interaction The nature of the phospho-lipid membrane also contributes to binding efficiency, with phosphatidylserine (PS) and, to a lesser extent, phosphatidylethanolamine (PE) being generally accepted as most important membrane phospholipids

in promoting efficient binding, complex assembly and enzyme catalysis in vivo

Membrane affinity of a Gla protein often correlates with its membrane localized activity Strategies used to increase the affinity of the Gla protein–membrane interaction involve Gla-domain mutation (for human protein C) [14–16], Gla domain substitution [17] and covalent dimerization of the Gla protein [18] We have previously created several Gla-domain mutated human protein C variants with enhanced anticoagulant activ-ity One of these variants with several Gla domain mutations, QGNSEDY-human APC (H10Q⁄ S11G ⁄ S12N⁄ D23S ⁄ Q32E ⁄ N33D ⁄ H44Y), bound phospholipid membranes with increased (approximately seven-fold) affinity compared to wild-type [16] QGNSEDY-human APC was shown to be potent in both a QGNSEDY-human plasma-based clotting assay (20-fold better) [16] and a FVa-degradation assay, cleaving R306 (18-fold) and R506 (four-fold) more efficiently [19] However, the variant had no antithrombotic effect when used in a rat model of arterial thrombosis [20,21] The lack of effect was possibly a result of species–species differ-ences between human protein C and the rat hemostatic system The reason for the poor anticoagulant effect of human APC in rat plasma remains unknown but may

be a result of rat FVa⁄ FVIIIa being poor substrates for human APC [21]

APC variants with enhanced anticoagulant activity resulting from improved membrane-binding ability may prove more efficient than wild-type APC in the treatment of different diseases (e.g thromboembolism and sepsis) [13] The low-affinity binding of APC to negatively-charged phospholipid membranes may be adequate under normal healthy conditions because protein S serves as a specific cofactor to increase the membrane binding of APC at certain locations The situation may be different under pathologic conditions such as sepsis, where a higher membrane-binding abil-ity of APC could potentially be beneficial, in particular because protein S and FV may be consumed under these conditions High affinity mouse APC variants

will allow the in vivo elucidation of the biologic

conse-quences of the enhanced membrane-binding ability of

protein C and may open a path for the development

of APC variants with improved therapeutic potential

in sepsis, as well as other thromboembolic disorders

Although human protein C variants with enhanced

affinity and function can be created, interspecies

incompatibility in functional assays using human

pro-tein C in animal model systems, prompted us to

char-acterize several mouse protein C variants in the

present study Individual amino acids residues within

the Gla domain contribute to the membrane affinity

differences reported for the Gla proteins In the

44-res-idue Gla domain of human and mouse protein C,

there are eight amino acid differences Thus, wild-type

mouse protein C as well as seven variants mutated at

three regions (positions 10, 12, 23, 32 and 33) of the

Gla domain were purified and characterized The

results obtained indicate that the functional

improve-ments were closely related to enhanced membrane

affinity The mutant with highest function, mutant III

(P10Q⁄ S12N ⁄ D23S ⁄ Q32E ⁄ N33D), showed reduced

Ca2+ dependence for membrane binding and a

30-50-fold inhibition improvement over wild-type in

tissue-factor-dependent thrombin generation in mouse

plasma Overall, the proteins described in the present

study provide insight into the Gla protein–membrane

interaction and identify new reagents with varying

degrees of anticoagulant potency that may be of use

for testing in murine models of sepsis and

thrombo-embolic disorders

Results

Expression and characterization of mouse protein

C variants

To determine whether the mutations previously made

in human protein C result in a similar enhancement of both membrane affinity and anticoagulant activity in a mouse system, the analogous mutations were made in mouse protein C Wild-type and seven variants of mouse protein C (Fig 1) were expressed and purified SDS-PAGE analysis of the purified proteins (Fig 2) demonstrated slightly different mobilities of the light chains, an effect caused by the mutations, whereas the

Fig 1 Gla domain sequence alignment from different species and mouse protein C variants used in the present study N-terminal Gla sequence (1–44) is shown and defined between the propeptidase and chymotrypsin cleavage sites Positions in the sequence at which c-carboxylation of glutamic acid residues is either known to occur or may occur are indicated by X The numbering at the top refers to the mouse protein C sequence Highlighted residues are different with respect to wild-type mouse protein C Sequences used for comparison were obtained from NCBI with accession numbers: protein C for mouse (NP_032960.2), human (NP_000303.1), rat (NP_036935.1), bovine (XP_585990.3) and human prothrombin (NP_000497.1).

Fig 2 SDS-PAGE analysis of recombinant mouse protein C and APC Purified proteins (8 lg) were incubated with human thrombin-thrombomodulin for 0 h (odd lanes) or 24 h (even lanes) Thrombin catalysis was stopped with excess hirudin and subjected to 12% SDS-PAGE under reducing conditions Approximately 0.1 lg of pro-tein C (odd numbered lanes) or APC (even numbered lanes) was applied to each lane and visualized by silver staining Protein C vari-ants and molecular weight markers (MWM) ran in each lane are indicated The location of heavy chain (HC), light chain (LC) and thrombin (IIa) is also indicated.

heavy chains migrated to similar positions All proteins

were fully activated by the human thrombin–TM

com-plex, as demonstrated by the shift of the heavy chains

to slightly lower molecular weight positions The

amid-olytic activities of activated protein C mutants were

comparable with that of wild-type protein C (data not

shown) The proteins bound Ca2+ similar to their

human counterparts, as judged by the shift in

mobili-ties in native agarose gel electrophoresis in the

pres-ence of Ca2+ compared to EDTA (data not shown)

The proteins were found to be c-carboxylated, as

judged by western blotting using a Gla-specific

anti-body (Fig S1)

Membrane binding ability of wild-type and

variants of mouse protein C

To determine the functional significance of the

substi-tuted Gla domain residues, we measured membrane

binding properties by surface plasmon resonance (SPR)

Chips were coated with 0-20-80, 0-10-90 and 20-10-70

1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine⁄

1,2-dioleoyl-sn-glycero-3-[phospho-l-serine]⁄

1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPE-DOPS-POPC)

liposomes, whereas a control surface was either left

blank or coated with 100% POPC We first measured

the binding of each protein at equi-molar concentration

(100 nm) to estimate their relative membrane binding

abilities (Fig 3A–C) Noticeably, mutants II and III

stand out from the other proteins analyzed, obtaining

the highest responses for all membrane types Mutants

V and VII also show a significant binding-response

enhancement, whereas mutations introduced into

mutants IV and VI had little effect relative to the

wild-type protein (Fig 3A–C, insets) Figure 3D–F

shows the equilibrium binding analysis of the protein–

membrane interactions, and the KD values determined

from the curve fitting are summarized in Table 1 The

affinities of wild-type mouse protein C for 0-20-80 and

20-10-70 liposomes are comparable (KD 8 lm), with

a value lower than that of the human ortholog

(KD= 2.1 lm) [15] and comparable with the bovine

ortholog (KD= 9.2 lm) [14], as assessed previously

under similar experimental conditions Using 0-20-80

or 20-10-70 membranes, mouse protein C variants that

show a considerable improvement in membrane

affin-ity over wild-type are mutant II (12-fold KDdecrease),

mutant III (six-fold KD decrease) and mutants V and

VII (three- to four-fold KD decrease) Equilibrium

binding dissociation constants, using 0-10-90

mem-branes, could only be determined for the high affinity

proteins A further improvement in membrane binding

of the variants is shown in terms of membrane

bind-ing occupancy at the saturatbind-ing protein concentration,

a parameter experimentally determined as Rmax For 0-20-80 membranes, the respective binding Rmax deter-mined for wild-type [722 response units (RU)], and mutants II (3569 RU), III (4380 RU) and VII (2060 RU), is clearly different, as is also evident from

an inspection of Fig 3D (or the other membranes in Fig 3E,F) All variants were tested using the same immobilized membrane preparation Thus, different variants are able to utilize a different number of bind-ing sites on the membrane surface For example, mutant II can utilize approximately five times as many binding sites on a 0-20-80 membrane as wild-type protein C

Importance of the liposome phospholipid composition on membrane binding Simple model membranes composed of one, two or three synthetically-derived phospholipids were used to assess membrane binding Membranes composed entirely of POPC were inert to binding, whereas DOPS

or DOPS with POPE-containing liposomes were neces-sary to obtain a binding response By varying the DOPS composition, we were able to show binding specificity in terms of DOPS content Doubling the DOPS content from 10 to 20 mol % resulted in increased binding sites with enhanced affinity (Fig 3D,E and Table 1) For example, mutant II binds

to 0-10-90 with KD= 2.45 lm⁄ Rmax = 3005 RU, whereas binding to 0-20-80 is improved with

KD= 0.66 lm⁄ Rmax= 3569 RU PE has been shown

to enhance the assembly and function of several clot-ting factor complexes [22,23] We also show that POPE influences the binding of mouse protein C Comparing the binding data of 20-10-70 and 0-10-90 membranes (Fig 3E,F and Table 1), we observe the effect that POPE has on membrane binding when holding DOPS

at a fixed concentration Substantial improvements in both the number of binding sites and average affinity are observed with the POPE containing membrane

Importance of Ca2+on membrane binding Because the Gla protein–membrane interaction is highly dependent on Ca2+, we also investigated how the introduced mutations affect membrane binding as

a function of Ca2+ concentration Figure 4A presents

a representative sensorgram showing the effect of Ca2+

on the interaction of mutant II, at fixed concentration, with 0-10-90 membranes Wild-type protein C at

20 mm Ca2+is included as a standard for comparison

As expected, the Gla protein–membrane interaction

is highly dependent on Ca2+, with maximum binding

occurring at approximately 10 mm Ca2+ Figure 4B–

D shows the equilibrium binding analysis of the

protein–membrane interactions at different Ca2+

con-centrations, and [Ca2+]1⁄ 2 max determined from the

curve fitting are summarized in Table 2 Employing a

20-10-70 membrane, half-maximum Ca2+ concentra-tions required for mutant II (3.7 mm) and III (3.7 mm) are much improved compared to wild-type ( 11 mm) and approach that of plasma-derived human prothrombin (1.8 mm), comprising an efficient membrane-binding Gla protein

Fig 3 Protein–membrane interaction of wild-type and variant mouse protein C to liposomes of varying phospholipid composition Protein C variants (wild-type, I–VII at 0.1 l M ) or running buffer was injected for 8 min (association), over either the (A) 0-20-80 or (B) 0-10-90 or (C) 20-10-70 POPE-DOPS-POPC membrane bilayer surface, to determine their relative binding efficiencies Dissociation under running buffer conditions was followed for an additional 8 min Ca2+concentrations used throughout were 5 m M The SPR response curves are shown after background correction using a blank control flow cell Binding to the control surface was not apparent and no evidence of nonspecific binding was evident from an injection of Gla-less, prethrombin-1 (10 l M , not shown) Similar amounts of 0-20-80 (5539 RU), 0-10-90 (5738 RU) and 20-10-70 (5997 RU) liposomes were immobilized allowing comparisons Protein labeling is shown Insets show the same data

on a smaller scale highlighting the low affinity binders Note y-axis scale differences in (B) Steady-state binding of mouse protein C wild-type ( ), mutant II ( ), III (.) and VII ( ), over either the (D) 0-20-80 or (E) 0-10-90 or (F) 20-10-70 POPE-DOPS-POPC membrane bilayer surface, was measured using the indicated protein concentrations Responses obtained at equilibrium were used to generate a binding isotherm fitted to a one-site binding hyperbola using nonlinear least squares analysis Binding isotherms were used to determine K D reported in Table 1 and Rmax Additional details are provide in the Experimental procedures.

Wild-type protein C and mutant II have different

maximum surface coverage (Rmax) and therefore any

attempt to draw conclusions based on absolute

response values is erroneous Furthermore, because

human prothrombin (72 kDa) has a different

molecu-lar mass than that of mouse protein C (56 kDa),

abso-lute response values cannot be directly compared

However, calculating the fraction of binding relative to

Rmax (equilibrium response at indicated Ca2+ divided

by Rmax at saturating Ca2+, i.e  20 mm Ca2+), as

shown in Fig 5, reveals the Ca2+-dependent

mem-brane binding differences amongst the proteins

Mutant II and III display a much improved fractional

membrane occupancy at physiologically relevant Ca2+

concentrations (1–5 mm) [24,25] compared to

wild-type For example, at 2 mm Ca2+, prothrombin

already obtains over 70% of its potential binding and

protein C mutants II and III each have approximately

50%, whereas wild-type protein C has obtained a mere

19% of its potential binding This indicates that the

mutations introduced in mutants II and III lower the

Ca2+ concentration requirement for effective binding,

thereby improving membrane affinity A similar trend

in fractional binding site occupancy is observed for the

0-20-80 and 0-10-90 membranes (data not shown)

Mouse APC variants with hyper-anticoagulant

activity in mouse plasma

The generation of thrombin is severely diminished in

mouse plasma when an APC variant with high affinity

for membranes is included in the reaction (Fig 6A and Table 1) Compared at equivalent concentrations (0.5 nm), mutant III completely abolished thrombin generation, whereas mutants I, II, V and VII and, to a lesser extent, mutant IV caused a down-regulation of thrombin generation compared to wild-type APC, which did not have an anticoagulant effect at this con-centration Although wild-type recombinant APC can function as an effective anticoagulant (Fig 6B), as also shown by Tchaikovski et al [26], strikingly lower con-centrations of APC mutants II and III (Fig 6C,D) were required to achieve an identical anticoagulant result For example, to observe a similar measurable down-regulation difference of thrombin generation rel-ative to thrombin generation in the absence of added APC [as assessed by either maximal concentration of thrombin (Cmax) or endogenous thrombin potential (ETP)], the concentrations required for wild-type, mutant II and mutant III were 1, 0.08 and 0.02 nm, respectively Similarly, thrombin generation was com-pletely inhibited at the concentrations tested for wild-type (16 nm), mutant II (1.28 nm) and mutant III (0.5 nm) Figure 7 reflects these findings and summa-rizes how each of the recombinant APC variants at several concentrations influences the generation of thrombin in mouse plasma Interestingly, thrombin generation parameters of lag-phase and time required

to reach maximum thrombin generation (Tmax) are not significantly altered by the addition of the tested APC molecules Furthermore, a three-fold higher concentra-tion of mutant III was required to obtain a similar

Table 1 Effect of Gla-domain mutations on mouse protein C ⁄ APC Membrane dissociation constants (K D ) at various membrane composi-tions were determined by SPR for mouse protein C C max and ETP generated in mouse plasma were determined using mouse APC Further details on methodology and experimental conditions are provided in Fig 3 (membrane binding) and Fig 6 (thrombin generation).

Protein C Membrane affinity (POPE-DOPS-POPC)

Activated protein C Thrombin generation in mouse plasma

a

K D is a representative determination from three experiments. bMembranes used were 100 nm extruded liposomes with synthetic phospholipids: POPE-DOPS-POPC (mol %) c SE from one-site binding hyperbola fitting d NA, not available Concentrations tested did not allow determination of KD e SD from three independent experiments f ETP determined after 60 min g Concentration of APC added to assay.

anticoagulant activity (assessed by either Cmax or ETP)

in human plasma compared to mouse plasma (data not shown), further highlighting the importance of using proteins from the same species

Discussion

Recombinant APC has been used to treat patients with reduced protein C levels suffering from severe sepsis (PROWNESS study) [8] The results of this and several other proceeding clinical trials [27] lead us to develop

an APC molecule with enhanced anticoagulant activity with the purpose of investigating the effect of APCs with increased anticoagulant activity in vivo on differ-ent thromboembolic diseases APC has species specific-ity in its anticoagulant function that may, to a certain extent, be a result of its interaction with protein S [21,28,29] Therefore, we were unable to continue work

in mouse models of sepsis or disseminated intravascu-lar coagulation with our previously developed human APC variants (e.g QGNSEDY-APC) In the present study, we created several mouse APC variants with improved membrane binding characteristics and hyper-anticoagulant activity Binding ability was established using a SPR membrane binding assay and anticoagu-lant activity was assessed by a thrombin generation assay The most active variant mutant III (P10Q⁄ S12N⁄ D23S ⁄ Q32E ⁄ N33D or QNSED), which is the

A

B

C

D

Fig 4 Effect of Ca 2+ on the mouse protein C–membrane inter-action (A) Protein C mutant II (1 l M ) was injected for 4 min (associ-ation) over a 0-10-90 POPE-DOPS-POPC membrane bilayer surface

at variable Ca 2+ concentration to determine binding efficiency as a function of Ca2+concentration for this high affinity binder Dissoci-ation under running buffer conditions was followed for an additional

6 min Ca 2+ concentration (m M ) is indicated near the appropriate curve The SPR response curves are shown after background cor-rection using a 100% POPC control flow cell Binding to the control surface was not apparent and no evidence of nonspecific binding was evident from an injection of Gla-less, prethrombin-1 (10 l M , not shown) Wild-type protein C (1 l M ) at 20 m M Ca2+ was also included for comparison (indicated with an arrow: 20 wild-type) (B–D) Steady-state binding of mouse protein C wild-type ( ), mutant II ( ), III (.), V (+) and human prothrombin (h), each at

1 l M , over either the (B) 0-20-80 or (C) 0-10-90 or (D) 20-10-70 POPE-DOPS-POPC membrane bilayer surface was measured using the indicated Ca2+ concentrations Responses obtained at equilib-rium were used to generate a binding isotherm fitted to a one-site binding hyperbola using nonlinear least squares analysis Note that the 50 m M Ca2+ data points were excluded from fitting because high Ca 2+ concentrations appear to inhibit the membrane inter-action Binding isotherms were used to determine half-maximal

Ca2+ concentration or [Ca2+] ½ max reported in Table 2 Data are representative of three experiments Additional details are provide

in the Experimental procedures.

mouse equivalent of the human QGNSEDY-APC

vari-ant, are discussed in more detail below

In the ETP assay using mouse plasma, mutant III

served as a far superior anticoagulant compared to

wild-type APC Mutant III had a 50-fold higher

activ-ity, thereby efficiently down-regulating thrombin

gener-ation, such that a mere 0.02 nm concentration was

required to reduce either Cmaxor ETP Complete

inhi-bition of thrombin generation was obtained at a

mutant III concentration of 0.5 nm, which is 30-fold lower than the concentration of wild-type APC required to give similar inhibition Previous work based on a tissue factor-dependent clot-based assay in normal human plasma showed that human APC vari-ant QGNSEDY had a 20-fold higher vari-anticoagulvari-ant potential than human wild-type APC [16] Although this was a simple end point assay, the anticoagulant potency of human APC variant QGNSEDY in human plasma parallels that of mouse APC mutant III in mouse plasma Mutant III, although only containing five mutations, is equivalent to the human QGNSEDY variant because the wild-type mouse Gla domain already has G at position 11 and Y at position 44 Thus, the seven Gla domain residues introduced in human APC (Q10, G11, N12, S23, E32, D33 and Y44) are all present in mouse APC mutant III

The high sensitivity of SPR detection allowed us to accurately analyze even low affinity proteins, such as wild-type mouse protein C, to membrane binding site saturation The results obtained are thus based on the combined analysis of binding affinity, Rmax and quali-tative kinetics Efficient binding of protein C, as well

as other Gla proteins, to membrane is dependent on three complimenting factors: an optimal Ca2+ concen-tration, an optimal phospholipid composition and, lastly, structures intrinsic to the protein (e.g optimal arrangement of residues in the Gla domain) For any given Gla protein, a good membrane can compensate for binding at sub-optimal Ca2+ concentrations Conversely, an optimal Ca2+ concentration can com-pensate for binding at sub-optimal membrane compo-sitions Thus, the results obtained in the present study indicate that any one of these three factors can com-pensate for two of the others if they are presented in a sub-optimal manner

The protein C mutations act as excellent reporters

of how these residues influence the membrane interac-tion Single and double amino acid substitutions at three separate regions of the protein C Gla domain were introduced at positions 10⁄ 12, 23 and 32 ⁄ 33 Mutagenesis at position 10⁄ 12, as in mutant I (QN), caused a small (1.5-fold) gain in affinity compared to wild-type The major affinity improvement came from combined mutagenesis at position 32⁄ 33 with mutant

V (ED) having a 3.5-fold higher affinity than wild-type Similarly, this double mutagenesis caused an eight-fold higher affinity in mutant II (QNED) relative

to mutant I (QN) A Glu residue (converted to Gla residue upon post-translational modification) intro-duced at position 32 in factor VII has been suggested

to bind an additional Ca2+during membrane binding [30], although it does not appear to serve the same role

Fig 5 Fractional membrane binding site occupancy by the various

protein C variants at different Ca 2+ concentrations Equilibrium

binding responses for mouse protein C wild-type ( ), mutant II ( ),

III (.), V (+) and human prothrombin (h), using a 20-10-70

POPE-DOPS-POPC membrane, were obtained at the indicated Ca 2+

con-centration as described in Fig 4D Equilibrium binding responses

were normalized for fractional occupancy for each individual

protein % R max is expressed as the equilibrium response at the

indicated Ca 2+ concentration divided by Rmax at saturating Ca 2+

concentration (20 m M Ca 2+ ).

Table 2 Half-maximal binding [Ca2+] as functions of various

mem-brane compositions were determined by SPR for mouse protein C.

Ca 2+ titration determined at fixed (1 l M ) protein concentration.

Membrane (POPE-DOPS-POPC) a

[Ca 2+ ]1⁄ 2 max(m M ) b

mPC wild type 15.9 ± 7.0 c 15.2 ± 11.1 10.8 ± 3.43

mPC mutant II 3.7 ± 1.6 7.2 ± 3.4 3.7 ± 1.5

mPC mutant III 3.6 ± 1.6 6.7 ± 3.5 3.7 ± 1.6

mPC mutant V 9.0 ± 5.0 16.7 ± 12.0 8.0 ± 3.7

Human prothrombin 1.4 ± 0.6 2.9 ± 0.9 1.8 ± 0.6

a Membranes used were 100 nm extruded liposomes with

synthetic phospholipids: POPE-DOPS-POPC (mol % indicated).

b Half-maximum binding [Ca 2+ ] determined from one-site binding

hyperbola fitting Representative determination from three

experi-ments.cSE determined from one-site binding hyperbola fitting.

in mouse protein C (see discussion on Ca2+ below).

An acidic residue at position 23, such as Asp in pro-tein C, has been speculated to instill low affinity to Gla proteins However, the D23S mutation had little effect on binding affinity; for example, when compar-ing mutant VI (S) and wild-type The D23S mutation appears even inhibitory if mutant II (QNED) and III (QNSED) are compared Similarly, the gain of func-tion observed with mutant I (QN) was reversed by having the D23S mutation present, as in mutant IV (QNS) It is noteworthy that the multi-site mutations introduced often resulted in synergistic affinity effects and were not simply the additive sum of individual mutations As such, there appears to be an intra-molecular synergism between the 10⁄ 12 (QN) and

32⁄ 33 (ED) sites in mouse protein C Our previous work with human protein C showed that variant QGNSEDY had increased membrane affinity (3.5- to seven-fold) and was more potent as an anticoagulant

in a TF-dependent clotting (PT) assay (approximately five-fold longer prolonged clot time) than wild-type [16] Mutagenesis at less sites in human protein C, as

in variants GNED (approximately one-fold⁄ 1.5-fold), QGN (1.1-fold⁄ fold) and SEDY (1.6-fold ⁄ one-fold), had negligible or minor improvements in affinity and anticoagulant activity, respectively [16] Although

a single amino acid in some Gla proteins can signifi-cantly influence affinity upon mutagenesis, there appears to be other additional mechanisms that can control the affinity of Gla proteins, as illustrated with Pro10 of the human, bovine and mouse protein C orthologs A Pro at position 10, occurring naturally or introduced by site-directed mutagenesis was shown to significantly lower the membrane affinity of Gla pro-teins In bovine protein C, P10H mutagenesis results in

a ten-fold affinity increase, whereas, with human pro-tein C, H10P mutagenesis results in a five-fold affinity

A

B

C

D

Fig 6 Inhibition of thrombin generation in mouse plasma by mouse APC and variants (A) Mouse plasma was incubated with either 0.5 n M extrinsically added wild-type mouse APC, or variant (I–VII), or in the absence of added protein (as indicated on curves) Thrombin generation was initiated with 0.25 p M tissue factor,

10 l M phospholipid liposomes and 16.7 m M CaCl2 and followed continuously with the fluorogenic substrate I-1140 (Z-Gly-Gly-Arg-7-amino-4-methylcoumarinÆHCl, 300 l M ) in 25 m M Hepes, 175 m M NaCl (pH 7.4) containing 0.5% BSA at 37 C All indicated concen-trations are final concenconcen-trations Mouse plasma (10 lL) was used

in a final reaction volume of 120 lL The first derivative of a typical experiment (n = 3) is shown (B–D) Concentration-dependent inhibi-tion of thrombin generainhibi-tion in mouse plasma by mouse APC is shown Mouse plasma was incubated with the indicated concentra-tion (n M ) of extrinsically added mouse APC (B) wild-type, (C) mutant

II or (D) mutant III using the conditions described above.

decrease [14] Mouse protein C contains the ‘low

affin-ity’ Pro10, but did not gain a significant increase in

affinity by mutagenesis to P10Q because mutant I

(QN) had only a modest 1.5-fold affinity increase

com-pared to wild-type protein Thus, work on human,

bovine and mouse protein C mutagenesis illustrates the

intricacy of improving the Gla protein–membrane

interaction

Mutant II has a modest (two-fold), but significant,

affinity enhancement compared to mutant III,

although, unexpectedly, mutant III has better

antico-agulant activity The only differences between mutant

II and mutant III is the amino acid at position 23 (D

in mutant II and S in mutant III) It is conceivable

that the amino acid at position 23 may affect

anticoag-ulant activity that is not dependent on membrane

binding In human proteins, protein S binds to the

protein C Gla domain Although this interaction has

been mapped to Gla regions C-terminal to the 23 site

[31,32], it may still influence the interaction to the

cofactor and thus influence the anticoagulant activity

of APC An influence of position 23 on the interaction

to APC substrates FVa and FVIIIa also cannot be

ruled out The same reasoning can be applied when

comparing membrane binding and anticoagulant

activ-ity data of mutant V and mutant VII, which also only

vary at position 23 Thus, 23S appears to be better for

membrane binding, but 23D appears to instill a better

anticoagulant property in mouse APC

Membrane binding capacity is a measure of the Gla protein packing density to sites provided by the mem-brane at saturating protein concentrations Of rele-vance for this discussion is the variation of membrane binding site occupancy amongst the different mouse protein C variants (Fig 3D–F) For example, steady-state Rmax for wild-type and mutant II differ by approximately five-fold to a 0-20-80 membrane sur-face In the classical view of a simple bimolecular (1 : 1) interaction, a receptor site can be saturated with different affinity analytes (of equal molecular weight), which, by definition, will all have an equivalent satura-tion response (i.e occurring when all receptor sites are occupied) that will be approached by a specific analyte concentration specified by the KDof the receptor–ana-lyte interaction This was clearly not observed when the saturation binding levels of the protein C variants were compared, for any of the phospholipid mem-branes tested, implying a different mode of binding between the proteins This indicates that the mem-brane, even in simple model membranes, provides a number of binding sites that are not isolated and homogenous in nature, but rather heterogeneous For example, it can be envisioned that a Gla protein can engage with a membrane site containing a variable number of PS molecules each displaying a different affinity The results obtained in the present study indi-cate that the high affinity mutants (e.g II and III) can utilize ‘poor’ binding sites that low affinity mutants (e.g wild-type) cannot engage with, and, thus, the high affinity mutants have access to more total binding sites resulting in a higher Rmax The membrane binding capacity difference observed amongst the variants argues for the existence of several binding sites com-posed of a variable number of PS molecules Thus, we conclude that two processes make the binding of mouse protein C mutants more efficient to membrane than wild-type First, variants such as mutants II and III are able to utilize a higher number of membrane binding sites Second, these variants are also able to engage with these membrane binding sites with an overall higher averaged affinity then wild-type mouse protein C These two processes allow substantially more protein to bind membrane when eqimolar con-centrations of these proteins are compared

Membrane binding on- and off-rates are probably important specifications for the functions of protein

C⁄ APC as well as the other Gla proteins Association with the membrane is likely a dynamic process involv-ing several bound intermediates An initial membrane engagement step (termed electrostatic docking) [33], followed by the association of an unknown number of

PS headgroups, hydrophobic x-loop insertion into the

Fig 7 Concentration-dependent inhibition of thrombin generation

in mouse plasma by mouse APC Mouse plasma was incubated in

the absence of protein or with the indicated concentration of

extrin-sically added wild-type mouse APC or variant (I–VII) Thrombin

gen-eration was initiated as described in Fig 6 % Cmaxis expressed as

the maximum concentration of thrombin generated in the presence

of APC divided by the maximum concentration of thrombin

gener-ated in the absence of APC Mean values from duplicate

determi-nations from a single experiment (data for 0.1, 5 and 10 n M ) and

the mean ± SD from three independent experiments each with

duplicate determinations (data for 0.5 and 1 n M ) are shown (log2

scale).