Protective Effects of Non-Anticoagulant Activated Protein C Variant (D36A/L38D/A39V) in a Murine Model of Ischaemic Stroke

Ischaemic stroke is caused by occlusive thrombi in the cerebral vasculature. Although tissue-plasminogen activator (tPA) can be administered as thrombolytic therapy, it has major limitations, which include disruption of the blood-brain barrier and an increased risk of bleeding. Treatments that prevent or limit such deleterious effects could be of major clinical importance. Activated protein C (APC) is a natural anticoagulant that regulates thrombin generation, but also confers endothelial cytoprotective effects and improved endothelial barrier function mediated through its cell signalling properties. In murine models of stroke, although APC can limit the deleterious effects of tPA due to its cell signalling function, its anticoagulant actions can further elevate the risk of bleeding. Thus, APC variants such as APC(5A), APC(Ca-ins) and APC(36-39) with reduced anticoagulant, but normal signalling function may have therapeutic benefit. Human and murine protein C (5A), (Ca-ins) and (36-39) variants were expressed and characterised. All protein C variants were secreted normally, but 5-20% of the protein C (Ca-ins) variants were secreted as disulphide-linked dimers. Thrombin generation assays suggested reductions in anticoagulant function of 50- to 57-fold for APC(36-39), 22- to 27-fold for APC(Ca-ins) and 14- to 17-fold for APC(5A). Interestingly, whereas human wt APC, APC(36-39) and APC(Ca-ins) were inhibited similarly by protein C inhibitor (t½ - 33 to 39 mins), APC(5A) was inactivated ~9-fold faster (t½ - 4 mins). Using the murine middle cerebral artery occlusion ischaemia/repurfusion injury model, in combination with tPA, APC(36-39), which cannot be enhanced by its cofactor protein S, significantly improved neurological scores, reduced cerebral infarct area by ~50% and reduced oedema ratio. APC(36-39) also significantly reduced bleeding in the brain induced by administration of tPA, whereas wt APC did not. If our data can be extrapolated to clinical settings, then APC(36-39) could represent a feasible adjunctive therapy for ischaemic stroke.


Introduction
Ischaemic stroke is a leading cause of mortality and morbidity worldwide. Tissue-plasminogen activator (tPA) can be administered as thrombolytic therapy to promote clot dissolution and reduce further ischemia if administered within 4.5 hours of stroke onset. [1] However, tPA has limitations, including increased risk of intracerebral bleeding, as well as the effects associated with reperfusion injury. [2] In murine models of stroke, tPA treatment can activate matrix metalloproteinases [3], disrupt the blood brain barrier and cause neuronal toxicity, which can counter its therapeutic benefits. [4] Treatments that limit such deleterious effects could be of major clinical importance.
The activated protein C (APC) anticoagulant pathway regulates thrombin generation through the proteolytic inactivation of cofactors FVa and FVIIIa. [5] The ability of APC to cleave FVa and FVIIIa is dependent upon distinct functional attributes. Inactivation of FVa and FVIIIa occurs on negatively-charged phospholipid surfaces. For this reason, APC anticoagulant function is dependent upon the phospholipid-binding Gla domain of APC. [6] The recognition of FVa by the serine protease domain of APC is partly dependent upon a positively charged loop adjacent to the active-site (Fig 1). [7] Similarly, a Ca 2+ -binding site in proximity to this loop in the serine protease domain is of great importance in the anticoagulant function of APC, likely by maintaining the conformation of the exosite(s) that recognise FVa. [8] Finally, in plasma the anticoagulant function of APC is almost entirely dependent upon its cofactor, protein S. [9] Protein S functions as a cofactor by increasing the affinity of APC for phospholipid surfaces and for FVa, and also through repositioning of the active site of APC that augments FVa and FVIIIa inactivation. The influence of protein S is primarily observed in the~20 fold enhancement of the cleavage of FVa at R306. APC also exhibits cytoprotective cell signalling properties, primarily mediated through non-canonical proteolytic activation of protease activated receptor-1 (PAR1). Amino acid alignment of human and murine protein C. Amino acid sequence is separated into different domains and numbered according to the human sequence. Amino acid identity is highlighted in yellow. Gla residues are denoted by *. Glycan attachment sites are highlighted in green. Residues involved in the catalytic triad are highlighted in red. Regions known to be important for protein S, factor Va and PAR1 binding are boxed. The amino acids mutated in each of the protein C variants 36-39, 5A and Ca-ins are shown in red, blue and green respectively. [10,11] APC-dependent signalling to the endothelium elicits anti-inflammatory effects involving reduction in adhesion molecule expression, improvement of endothelial barrier function and anti-apoptotic effects. Through these actions, APC can limit the deleterious effects of tPA in murine models of stroke. [12,13] These beneficial effects are attributable to its cytoprotective function, rather than its anticoagulant activity. Indeed, the anticoagulant actions of APC appear to be detrimental in this model. [14,15] However, the potential increased likelihood of bleeding complications associated with co-administration of both a fibrinolytic and anticoagulant agent means there is necessary caution over the use of APC in stroke patients. [16] Thus, APC variants with reduced anticoagulant, but normal cytoprotective function may represent attractive adjunctive therapy options. Three such variants have been engineered (Fig 1); 1) APC KKK191-193AAA/RR229-230AA-termed APC(5A)-with serine protease domain mutations that impair FVa proteolysis, but do not influence PAR1 signalling [17], 2) APC R223C-D237Ctermed APC(Ca-ins)-containing an engineered disulphide bond in a serine protease domain Ca 2+ -binding site that diminishes cleavage of FVa but not PAR1 [8], and 3) APC D36A/L38D/ A39V -termed APC(36-39)-containing substitutions that abolish the essential protein S enhancement of APC anticoagulant function, but leaves its signalling function untouched [9,18].
A further variant APC KKK191-193AAA-termed APC(3A)-contains three of the five substitutions present in the APC(5A) variant. The impairment of anticoagulant function of this variant is not as severe as for APC(5A). However, APC(3A) is currently undergoing trials for use in humans with a view to exploring its efficacy as an adjunctive therapy to tPA in the setting of ischaemic stroke. [19] We present here, the direct comparison of the anticoagulant functions and inactivation of the three non-anticoagulant APC variants, and an investigation of the therapeutic benefits of APC (36)(37)(38)(39) in a murine model of ischaemic stroke.

Production and quantitation of human and murine APC variants
The pRC/CMV/human protein C mammalian expression vector was used as template for sitedirected mutagenesis to introduce the KKK191-193AAA/RR229-230AA (termed 5A), R223C/ D237C (termed Ca-ins) and D36A/L38D/A39V (termed 36-39) mutations, as previously described (Fig 1). [20] The active site variant S360A was also generated. The mammalian expression vector pcDNA3.1 containing the murine protein C cDNA (gift from Prof John Griffin, The Scripps Research Institute, CA), was used as template for site-directed mutagenesis to introduce the corresponding mutations-KKK192-194AAA/RR230-231AA (5A), R224C/D238C (Ca-ins), E36A/L38D/A39V (36)(37)(38)(39) and S360A -into murine protein C. [21] All mutations were verified by sequencing. HEK293 and HEK293T cells were used for stable and transient expression, respectively, of human and murine protein C variants, as previously described. [20] Vitamin K was added to culture medium to enable γ-carboxylation. For expression studies, HEK293T cells were transiently transfected with all vectors in parallel using PEI and thereafter, cultured in OptiMEM containing vitamin K for 3 days. [9] Thereafter, human and murine protein C expression and secretion was assessed by Western blot analysis of conditioned media and of cell lysates using polyclonal rabbit anti-human protein C (Dako) or sheep anti-mouse protein C (Haematologic Technologies Inc.), Similar analyses were also performed upon stably transfected HEK293 cells.
For purification of fully γ-carboxylated protein C human and murine protein C variants, 1-2L of conditioned media from each variant was concentrated~20 fold by tangential flow filtration (Millipore), dialysed thoroughly into 20 mM Tris-HCl 150 mM NaCI, 10 mM EDTA, pH 7.4, and thereafter, purified using an anionic HiTrap Q Sepharose Fast Flow column (GE Healthcare). Samples were loaded in 20 mM Tris-HCl 150 mM NaCI, 10 mM EDTA, pH 7.4, washed with 20 mM Tris-HCl 150 mM NaCI, pH 7.4 and eluted with a step gradient into 20 mM Tris-HCl 150 mM NaCI, 50 mM CaCl 2 , pH 7.4, as previously published. [22] All fractions were analysed by Western blotting using polyclonal anti-human or anti-mouse protein C antibodies (1:1000; Haematologic Technologies Inc.) to assess isolation of protein C, and a mouse monoclonal anti-human Gla (1:1000; Haematologic Technologies Inc.) antibody to assess separation of γ-carboxylated and non-fully γ-carboxylated protein C.
Sandwich ELISAs for human and murine protein C were used to quantify protein C in all preparations. For this, 96-well Maxisorp plates (Nunc) were coated overnight with either sheep anti-human protein C (1:2000) or sheep anti-mouse protein C (1:1000; Haematologic Technologies Inc.) in 0.1M sodium carbonate (pH 9.6) buffer and blocked with 3% BSA in PBS for 2 hours. Human or murine protein C samples diluted in, 1% BSA in PBS, were added to wells. In parallel, a standard curve of 0-4 nM of human or murine plasma-derived protein C (Haematologic Technologies Inc.) was generated and incubated with wells. Thereafter, a rabbit antihuman-protein C-HRP antibody (1:2000; Dako) or a monoclonal rat anti-murine protein C antibody (1:1000; Haematologic Technologies Inc.) followed by a rabbit-anti-rat-HRPantibody (1:3000; Dako) were used for detection.
Human protein C variants were activated with protac, as previously described [23], and repurified with a HiTrap Q-Sepharose FF column as before [9]. Murine protein C and its variants were activated using mouse thrombin and rabbit thrombomodulin (Haematologic Technologies Inc.) coupled to Dynabeads M-280 Tosylactivated (5 mg beads/100 μg thrombin-thrombomodulin for up to 9 μg/ml of protein C) for 5h at 37°C, according to the manufacturer's instructions (Life Technologies). In both cases, confirmation of complete activation under these conditions was previously determined using an ELISA that is specific for protein C (and not APC) using monoclonal antibodies specific for either human or mouse protein C (Haematologic Technologies Inc.) as the capture antibody, and also based on the initial rate of hydrolysis of the S-2366 chromogenic substrate (Chromogenix). [24] For this, human or murine APC (and variants) were diluted in 20 mM Tris-HCl (pH 7.8), 150 mM NaCl, 2.5 mM CaCl 2 , 0.1mg/ml BSA and 0.1% polyethylene glycol (PEG 8000). To this, 400 μM (final concentration) of S-2366 substrate were added and hydrolysis of the S-2366 substrate measured by changes in absorbance at 405 nm. A standard curve was generated using 0-12 nM plasma derived human APC (Haematologic Technologies Inc.). Initial rates of proteolysis were plotted against APC concentration to generate the standard curve.
The amidolytic activity/catalytic efficiency of all human and murine APC variants was also assessed using the chromogenic substrate S2366 (Chromogenix) to determine kinetic constants k cat , K m and k cat /K m , enabling APC concentrations for each variant to be determined in activated preparations through its enzymatic activity. For this, 2 nM of each fully activated APC preparation was incubated with a range of concentrations of S-2366 (0-2000 μM) in 20 mM Tris-HCl (pH 7.8), 150 mM NaCl, 2.5 mM CaCl 2 , 0.1mg/ml BSA and 0.1% PEG. The initial rate of the S-2366 hydrolysis was measured at 405 nm. Curve fitting was performed, using the Michaelis-Menten equation in GraphPad Prism. K m , V max , and k cat parameters were calculated from which the k cat /K m was derived. All samples were tested in duplicate and experiments were repeated three to four times.

Binding of murine protein C variants to murine EPCR
The affinity of all murine protein C variants for murine EPCR was evaluated using a plate binding assay [63,232]. For this, an anti-myc monoclonal antibody (1.5 μg/ml) (Sigma) in 0.1 M sodium carbonate (pH 9.6) was immobilised onto a 96-well plate at 4°C overnight. The plate was washed with TBS containing 5mM CaCl 2 , and 0.6mM MgCl 2 , followed by blocking with TBS-3% BSA. Soluble murine EPCR (5 μg/ml) with a C-terminal myc-His tag (a kind gift of Prof J Hermida, University of Navarra, Spain) [25] was added to each well and incubated at room temperature for 1 hour. After two washes with TBS containing 5mM CaCl 2 , and 0.6mM MgCl 2 , 100 μl of each murine protein C variant (0 to 1200 nM) diluted in TBS-1% BSA containing 5mM CaCl 2 , and 0.6mM MgCl 2 were added in duplicate and incubated for 2 hours at 37°C with shaking. Following incubation, the plate was washed three times with TBS containing 5mM CaCl 2 , and 0.6mM MgCl 2 . Protein C-EPCR complexes were detected with addition of sheep anti-murine protein C followed by goat-anti-sheep-HRP. The amount of bound murine protein C was detected by addition of OPD and stopped by the addition of 50 μl of 3M H 2 SO 4 . Absorbance values were fitted using a one-site equation in GraphPad Prism 5.0. Apparent equilibrium binding constants (K D(app) ) were calculated. Experiments were repeated three times.

Inhibition of human APC variants by purified protein C inhibitor (PCI)
The rate of inactivation of 20nM human APC variants by 200nM human recombinant PCI (Prof J. Huntington, University of Cambridge, UK) was assessed by removing aliquots at selected time points (0-20 mins) and measuring residual APC activity based on the initial rate of proteolysis of the chromogenic substrate, S2366 (Chromogenix), to monitor active APC concentration. [26][27][28] Thrombin generation assays The anticoagulant activities of the human APC variants were determined based on their ability to inhibit tissue factor (TF)-induced thrombin generation, using calibrated automated thrombography assays, as previously described. [9,29,30] Both normal pooled human plasma and protein C-deficient plasma (Affinity Biologicals) were used. For the assays, 80 μl of plasma, were incubated with 65 μg/ml corn trypsin inhibitor (HTI), 50 μM phospholipid vesicles (DOPS:DOPC:DOPE, 20:60:20), 4 pM human TF (Innovin, Dade Behring), and 0-20 nM APC, in a final volume of 100 μl (all concentrations are final). Thrombin generation was initiated by automatic dispensation of 20 μl of 2.5 mM Z-Gly-Gly-Arg-AMCÁHCl (Bachem), 60 mg/ml BSA 100 mM CaCl 2 in 20 mM Tris-HCl pH 7.4 into each well. The reactions were performed at 37°C. Measurements were taken at 20 second intervals for 40 minutes at wavelengths 390 nm (excitation) and 460 nm (emission) with a Fluoroscan Ascent FL Plate Reader (Thermo Lab System) in combination with the thrombinoscope software (Synapse, BV). A thrombin calibration standard (Synapse) was used to correct inner filter and substrate consumption effects. Thrombin generation was quantified by deriving the endogenous thrombin potential (ETP)-i.e. the area under the curve. All samples were tested in duplicate and experiments were repeated at least three times.
The ability of the murine APC variants to inhibit TF-induced thrombin generation was assessed in murine plasma (Innovative Research). Briefly, 40 μl of murine plasma was incubated with 65 μg/ml corn trypsin inhibitor, 50 μM phospholipid vesicles (DOPS:DOPC:DOPE, 20:60:20), 4 pM human TF, and (0 to 20 nM) murine APC in a final assay volume of 120 μl (all concentrations are final). Samples were tested as for human plasmas in duplicate and experiments were performed three times.
For evaluation of the relative anticoagulant activity of the human APC variants, the ETP generated in the absence of APC was taken as 100%, and % ETP plotted as a function of APC concentration. These were fitted using a one phase exponential decay non-linear regression equation to derived IC 50 values for each APC variant.

Murine transient middle cerebral artery occlusion (MCAO) model
All mouse procedures were performed in strict accordance with the UK Animals (Scientific Procedures) Act 1986 under a Home Office approved project license. All animal work was further approved by the Imperial College Committee on the Ethics of Animal Experiments. All surgery was performed under isoflurane anaesthesia, and all efforts were made to minimise suffering. The MCA of adult, male C57BL/6 mice, anaesthetised with 1-2.5% isoflurane in 30% oxygen enriched air, was occluded for 60 minutes using an occlusive intraluminal silicon-coated monofilament. Successful occlusion and reperfusion were confirmed by recording cerebral blood flow using laser Doppler flowmetry (Moor Instruments). Core temperature was monitored via a rectal probe and maintained at 37 ±0.5°C using a homeothermic blanket (Harvard Apparatus). Arterial oxygenation, heart rate, and respiratory rate were monitored via a femoral probe pulse oximetry (STARR Life Sciences Corp) and in all experiments these parameters were within the range expected for isoflurane-anaesthetised mice. Three hours post-MCAO, mice were randomly assigned to i.v. administration of either a) 10 mg/kg recombinant human tPA (10% bolus, 90% 30 min infusion), b) 250 μg/kg murine APC (wt or [36][37][38][39], c) both tPA and murine APC (wt or 36-39), d) saline control. Following recovery, mice were held in a 37°C cage for 2 hours, and then to a room with a 12 hour light/dark cycle.
After 24 hours, focal neurological deficit was scored. [31] Infarct and hemisphere areas of sectioned brains stained with cresyl violet were measured. Corrected infarct area was calculated and expressed as a mean±SEM percentage of contralateral hemisphere. [32] Oedema was calculated by the ratio of the area of the infarcted hemisphere vs. contralateral hemisphere. [32] Haemoglobin levels in homogenized ischemic hemispheres were determined by a spectrophotometric assay using Drabkin's reagent (Sigma). [33] Neurologic evaluation Twenty-four hours after treatment, an observer blinded to group assignment evaluated the animals individually for both focal and general neurological deficits using two separate previously validated graduated scales, with higher scores indicating more severe injury. The focal neurological deficit scale comprised observing appearance of: hair (0-2), ears (0-2), eyes (0-4), posture (0-4), spontaneous activity (0-4), and epileptiform behaviour (0-12), body symmetry (0-4), gait (0-4), circling behaviour (0-4), climbing a 45°slope platform (0-4), forelimb symmetry (0-4), compulsory circling of forelimbs (0-4), and whisker response (0-4), as previously described. [31] Quantification of cerebral infarction and oedema ratio Mice brains removed 24h post-treatment were cryoprotected before being cut in a cryostat at 30 μm and stained with cresyl violet (Sigma-Aldrich) and photographed (Olympus C2020). Digitised images of brain sections of the same plane were analysed by a blinded observer using ImageJ image software (NIH). For each animal, quantification of the infarcted area was performed on ten predetermined stereotaxic levels 300 μm apart encompassing the rostrocaudal extent of the lesion, ranging from Bregma +2.22 mm to-4.60 mm. The injury volume was calculated by multiplying the area of healthy tissue in mm 2 for each hemisphere by the thickness of brain sections (30 μm) and distance between them, with the difference between the volumes for the healthy tissue of each hemisphere representing infarct area, expressed as mm 3 , as previously reported. [34] The oedema ratio was calculated indirectly and represented as a ratio of the area of the infarcted hemisphere vs. contralateral hemisphere, as previously reported. [33] Quantification of cerebral haemoglobin To measure signs of intracerebral bleeding, haemoglobin levels in the ischaemic hemisphere of freshly extracted and homogenised brains were determined by a spectrophotometric assay using Drabkin's reagent (Sigma-Aldrich). [35] A standard curve was obtained with known amounts of haemoglobin (Sigma-Aldrich).

Statistical analysis
For the in vivo studies, Cronbach's α was used to test the reliability of data within different treatment groups. Kruskall-Wallis one way analysis of variance followed by a Mann-Whitney test was used to evaluate the efficacy of treatments in the MCAO model. P values <0.05 were considered statistically significant.

Results and Discussion
Expression and activation of human and murine protein C variants Three APC variants (human and murine) with reduced anticoagulant function, but normal, or near normal, cell signalling properties-APC(5A), APC(Ca-ins) and APC (36)(37)(38)(39) were generated (Fig 1). All human and murine protein C variants were secreted well. However, consistent with a previous report [8], an estimated 5-20% of the human and murine protein C(Ca-ins) variant was secreted as disulphide-linked dimers as evidenced by detection of an additional 120kDa band by Western blotting (which was not present for the other variants), and that disappeared upon reduction. These results suggest that a proportion of this protein C(Ca-ins) variant does not contain the intended engineered disulphide bond, which leaves surface exposed free cysteines available that can lead to dimerisation.
Large scale expression of all variants was performed followed by isolation of fully γ-carboxylated protein C by anion exchange chromatography. Efficient separation of γ-carboxylated and non-γ-carboxylated protein C was confirmed using an anti-Gla mAb, which revealed that the protein C that did not bind efficiently to the ion exchange column was recognised by anti-protein C antibodies, but was not immunoreactive with the anti-Gla antibody. Protein C eluted from the column with Ca 2+ was immunoreactive with both anti-protein C and anti-Gla antibodies.

Affinity of murine protein C variants for murine EPCR
Many of the important cytoprotective functions of APC are dependent upon its binding to EPCR prior to proteolytic activation of PAR-1. [36] Previous studies have revealed that the human APC(5A) and APC(36-39) variants have normal signalling function and/or EPCR binding. [17,18] Although the human APC(Ca-ins) variant harbours substitutions in the protease domain, which is spatially separated form the EPCR-binding site in the Gla domain, it has been reported that this confers reduced affinity for EPCR. [8] There are differences in the amino acid sequence of the ω-loop in the Gla domains of human and murine protein C that span the region that binds EPCR. This has revealed differences in the affinity of human protein C for murine EPCR, which may influence the potency of human APC in transducing cytoprotective signalling to the endothelium in mice. [37,38] This could warrant some caution when using human APC variants in murine models. To examine the effect of the non-anticoagulant substitutions upon the affinity of the murine protein C variants with murine EPCR, we performed a plate binding assay using the soluble extracellular domain of murine EPCR that was C-terminally tagged to enable immunocapture onto plates (Fig 2). Using this approach, we derived K D(app) of 94.3nM ±18.3 nM of plasma-derived murine protein C for murine soluble EPCR. For the recombinant murine protein C variants, we derived K D(app) of 112 ±17.6 nM for wt protein C, 110 ± 16.7 nM for protein C(5A), 111 ± 13.2 nM for protein C(36-39), 129 ± 17.1 nM for protein C(S360A), and 214 ± 30.4 nM for protein C(Ca-ins) (Fig 2). These results demonstrated that all murine variants exhibited normal affinity for murine soluble EPCR with the exception of the protein C(Ca-ins), which, similar to its human counterpart, had an approximately 2 fold reduced affinity for its receptor.

Comparison of proteolytic function of human and murine APC variants
The activation of human protein C variants was performed using protac. All variants were activated normally, with the exception of protein C(Ca-ins), which was activated less efficiently (>2 fold less efficiently). This finding was similar to the findings of a previous report, which suggested that this protein C(Ca-ins) variant was also activated less well than wt protein C by the thrombin-thrombomodulin complex (although 60-80-fold better by thrombin alone). [8] Activating the murine protein variants was not possible with protac, as this activator does not efficiently convert murine protein C. We attempted several different activation strategies including human thrombin, human thrombin in conjunction with rabbit soluble thrombomodulin, and murine thrombin with rabbit thrombomodulin. Of these, murine thrombin/rabbit thrombomodulin was the most efficient in converting the murine variants to APC, with the exception of the protein C(Ca-ins), which was more efficiently activated by murine thrombin alone in the presence of EDTA.
All human and murine protein C variants were fully activated prior to kinetic analysis of their proteolytic function using the APC chromogenic substrate S2366. Complete activation was confirmed using a protein C-specific ELISA capable of detecting any residual unactivated protein C in preparation. All human and murine APC variants exhibited very similar kinetic parameters (k cat and K m ) by comparison to wt APC. Of all the variants, only APC(Ca-ins) exhibited a slight (~25% increase) difference in K m for proteolysis of the short substrate, S2366 (Table 1).
We also performed similar analyses using the murine APC(5A), (Ca-ins) and (36)(37)(38)(39) variants in murine plasma and titrating up to 20nM. These experiments revealed very similar results to those obtained using the equivalent human APC variants in human plasma and indicated reductions in anticoagulant function of 19-, 27 and 57-fold by comparison to murine wt APC (Fig 3C).
Taken together, these results demonstrate that all non-anticoagulant APC variants (human and murine) exhibit large reductions in their ability to function as anticoagulants, and suggest that, of these, the APC(36-39) has the most marked reduction in its ability to inhibit thrombin generation in plasma.
We originally aimed to generate and characterise the non-anticoagulant forms of APC with a view to exploring their therapeutic benefit in protecting against the deleterious effects of tPA in a murine model of stroke, and to examine whether they reduce the risk of bleeding associated with the use of wt APC. Similar studies have been carried out using APC(5A) and another variant, termed APC(3A), that contains just three (KKK191-193AAA) of the five substitutions present in the 5A form. Although murine APC(3A) exhibits a reduction in anticoagulant function (~10-fold in vivo), this is not as severe as for the murine APC(5A) variant in murine plasma. Both APC (5A) and APC(3A) have been shown to be efficacious in rodent models of stroke. Based on these studies APC(3A) is currently undergoing development and testing in humans. [12,14,19,37]

Direct comparison of APC variant inactivation
There exist some important species differences in the protein C system between humans and mice (Fig 1). For example, murine protein S does not appear to function efficiently as a cofactor  Non-Anticoagulant APC and Stroke for human APC. [39] Consequently, as APC function in plasma appears to be highly dependent upon protein S for its anticoagulant function, this suggests that the use of human APC (or variants thereof) in mouse models may naturally confer appreciably reduced anticoagulant activity when used in this setting. Studies using human APC in murine plasma have indicated that it exhibits significantly reduced anticoagulant function when compared to murine APC. A further potential difference between species may be in the inactivation of APC. In humans, APC has a plasma half-life of~23 minutes. [40] PCI is just one of the inhibitors in human plasma capable of inhibiting APC. Intriguingly, this serpin is not present in murine plasma and, therefore, is not a determinant of its inactivation or functional half-life in mice. A previous report demonstrated that the positively charged loop that has been mutated in the APC(5A) and APC(3A) variants has a negative effect upon inhibition of APC by PCI. We therefore examined the inactivation rates of the human APC variants (20nM) by 200nM human recombinant PCI (Fig 4). Under these conditions, the inactivation of wt APC and variants APC (36)(37)(38)(39) and APC(Ca-ins) was comparable (t 1/2 33 to 39 minutes), and very similar to previous reports for wt APC. However, the inactivation of APC(5A) was appreciably accelerated (t 1/2~4 minutes)-consistent with a previous report that explored the influence of K191, K192 and K193 upon PCI-mediated inactivation. [26] The other major inhibitor of APC in plasma, α1-antitrypsin, is unlikely influenced by the mutations in the APC variants, based on the reported structural requirements for its inactivation of APC. [41,42] The enhanced inactivation rate of APC(5A) by PCI is due to the positively charged residues adjacent to the active site (KKK191-193) in wt APC normally repelling the bait loop of PCI, making inactivation inefficient. However, when these residues are substituted for non-charged alanine in the APC(5A) and APC(3A) variants, this repulsive effect is lost leading to enhanced inactivation. These findings suggest that, in humans, the APC(5A) and APC(3A) variants may have the potential to get inactivated by PCI more efficiently than wt APC, APC (36)(37)(38)(39) and APC(Ca-ins). In mice however, due to the lack of PCI in plasma, the influence of enhanced PCI inactivation would have no effect. A recent report detailing findings from a Phase I study of APC(3A) in humans suggested a half-life (a combination of both inactivation and clearance) of 16 mins (range 13 to 18 mins), which may be moderately shorter than a previous estimate of 23 mins for wt APC. [19,40] These findings likely reflect a modest influence of PCI upon the normal inactivation/clearance of APC in humans. Evaluation of murine APC (36)(37)(38)(39)

in a mouse model of ischaemic stroke
The in vitro analyses of wt APC and the APC(5A), APC (36)(37)(38)(39) and APC(Ca-ins) variants suggested that APC(Ca-ins) exhibited certain disadvantages as a non-anticoagulant form of APC. This was based on the findings that it is expressed partly in a dimeric form, and it also exhibited a modest reduction in affinity for EPCR suggesting that its cytoprotective function may be similarly compromised. APC(5A) exhibited normal expression, activation, EPCR binding and enzymatic function. However, based on the titration of this variant into thrombin generation assays, although it exhibited severely reduced anticoagulant function, APC(Ca-ins) and APC (36)(37)(38)(39) appeared to be more profoundly compromised. Therefore, based on our in vitro findings, we compared the therapeutic efficacy of wt APC and APC (36)(37)(38)(39) both with and without tPA in the murine MCAO model. At 24-hours post treatment, mice were scored by a blinded observer using a defined focal neurological deficit scale. Following MCAO, mice receiving saline had a mean neurological score of 15.2 ±1.4 (n = 19). Mice that received 250μg/kg APC (36)(37)(38)(39) alone, or in combination with 10mg/kg tPA, had significantly improved neurological scores (6.1 ±0.6; n = 13, and 7.9 ±1.3; n = 14, p <0.05, respectively), that was at least as effective as wt APC in improving this parameter (Fig 5A).
Using the ratio of the total area of the infarcted to non-infarcted brain hemisphere as a measure of brain oedema, APC (36)(37)(38)(39) in combination with tPA significantly reduced brain oedema (n = 7; Fig 6A) to a similar extent as wt APC in combination with tPA (n = 6). These findings, with respect to the extent of oedema reduction, are similar to those reported in previous murine studies using the APC(3A) variant [15,37].
Taken together, our data suggest that the APC(36-39) variant may represent potential adjunctive therapy to tPA for the treatment of ischaemic stroke. They also demonstrate that targeting the protein S cofactor enhancement of APC is a viable strategy for producing nonanticoagulant APC. APC(3A) has currently undergone phase I testing in humans. The results from this small study revealed that APC(3A) was well tolerated in healthy individuals up to doses of 540μg/kg. At this dose, plasma concentrations of~65nM APC were measured at 15 mins. Based on the reported functional half-life of 15 mins for APC(3A), the initial plasma concentrations of this dose might be predicted to be approximately 130nM. Although the APC (3A) variant has impaired anticoagulant function, this is reportedly about 10-fold reduced. Based on this figure, a concentration of 130nM APC(3A) may exert an anticoagulant action that approximates to that of 13nM wt APC. In plasma-based thrombin generation assays, 13nM APC is enough to completely abolish thrombin generation and could be considered to represent quite potent anticoagulation. In the phase I study, the degree of anticoagulation was measured at 1 hour post administration of APC(3A). Although significant dose-dependent increases in aPTT were reported, these increases at 1 hour were not considered likely to elevate the risk of bleeding. However, by 1 hour post-administration, APC(3A) has already been through four half-lives, which, by extrapolation, could suggest that the level of anticoagulation that may occur during the period immediately after infusion of the drug may be appreciably higher. It should be pointed out though, that whether this is likely to significantly alter bleeding risk, or bleeding severity is unclear at this time. However, this may provide a rationale to exploring the use of an APC variant with a more profound reduction in anticoagulant function than that of the APC(3A) variant.
At this time, it is difficult to ascertain the optimal dosing of human APC (and variants thereof) that are required for effective therapeutic cytoprotection. Based on our findings though, APC(36-39) may enable higher dosing than that of APC(3A) without appreciably perturbing the anticoagulant profile of the patient. Interestingly, a recent study by Ni Ainle et al suggested that alteration of N-linked glycosylation of the protein C serine protease domain through introduction of an Asn329Gln substitution significantly augmented APC cytoprotective signalling. [43] This finding may provide a further opportunity to create variants not only with reduced anticoagulant function, but also selectively enhanced cell signalling function. This later approach may, in turn, enable reduced levels of APC to be administered whilst maintaining therapeutic benefit. Although further studies are now required to explore the full potential of APC as an adjunct therapy in the setting of ischaemic stroke in human, our data, in combination with those from others, provide potential strategies to further augment the therapeutic efficacy of APC.