Activated protein C attenuates severe inflammation by targeting VLA-3^high neutrophil subpopulation in mice

Abstract

The host injury involved in multi-organ system failure during severe inflammation is mediated, in part, by massive infiltration and sequestration of hyperactive neutrophils in the visceral organ. A recombinant form of human activated protein C (rhAPC) has shown cyto-protective and anti-inflammatory functions in some clinical and animal studies, but the direct mechanism is not fully understood. Recently, we reported that, during endotoxemia and severe polymicrobial peritonitis, integrin VLA-3 (CD49c/CD29) is specifically up-regulated on hyper-inflammatory neutrophils and that targeting the VLA-3^high neutrophil subpopulation improved survival in mice. Here, we report that rhAPC binds to human neutrophils via integrin VLA-3 (CD49c/CD29) with a higher affinity compared to other RGD binding integrins. Similarly, there is preferential binding of APC to Gr1^highCD11b^highVLA-3^high cells isolated from the bone marrow of septic mice. Furthermore, specific binding of rhAPC to human neutrophils via VLA-3 was inhibited by an antagonistic peptide (LXY2). In addition, genetically modified mutant APC, with a high affinity for VLA-3, shows significantly improved binding to neutrophils compared to WT APC and significantly reduced neutrophil infiltration into the lungs of septic mice. These data indicate that variants of APC have a stronger affinity for integrin VLA3, which reveals novel therapeutic possibilities.

Introduction

Neutrophils play a critical role in the host defense system by participating in the phagocytosis and killing of infectious organisms with powerful antimicrobial substances (1). During a severe inflammatory response due to an infection, the activation of neutrophils with pro-inflammatory cytokines and microbial components induces reduced apoptosis, altered chemotaxis, and excessive infiltration into the visceral organs, leading to collateral tissue injury (1). Integrins are heterodimeric, transmembrane proteins that mediate adhesion and signaling between neutrophils and the extracellular matrix (ECM) proteins and facilitate their intravascular and interstitial migration. Recently, we showed that during endotoxemia and severe polymicrobial peritonitis, there is significant up-regulation of integrin VLA-3 (CD49c/CD29) on the surface of hyperactive neutrophils both in mice and patients (2). Gr1^highCD11b^highVLA-3^high neutrophils, isolated from the bone marrow of septic mice, produce significantly higher amounts of pro-inflammatory cytokines and show elevated MPO functions compared to those of the Gr1^highCD11b^highVLA-3^low subset. We further demonstrated that deletion of integrin VLA3, specifically from the neutrophils, significantly improved the survival of septic animals.

Activated protein C (APC), a vitamin K-dependent serine protease with strong anticoagulant functions, is derived from the thrombin-mediated cleavage of circulating protein C (PC). In addition to its natural anticoagulant functions, APC also possesses cyto-protective and anti-inflammatory activities, which include protecting the endothelial barrier, inhibiting cell apoptosis, reducing secretion of pro-inflammatory mediators and inhibiting leukocyte migration (3). While the anti-apoptotic and endothelial barrier functions of APC require the activation of the endothelial protein C receptor (EPCR)-dependent protease-activated receptor-1 (PAR-1), its anti-inflammatory effects are mediated by both EPCR-PAR-1-dependent and independent pathways and may involve cell adhesion receptors, such as integrins (3).

Data from Nick and Stern et al. showed that neutrophils express receptors for APC and that neutrophil chemotaxis is inhibited by exposure to rhAPC (4, 5). We also showed that rhAPC blocks integrins and inhibits neutrophil adhesion and migration on ECM proteins (6). In this study, we report that, among the leukocyte integrins, VLA-3 (α₃β₁; CD49c/CD29), which is a unique cell surface marker for the hyper-inflammatory neutrophil subpopulation arising during severe systemic inflammation, is a novel high affinity cellular receptor for rhAPC. Consistently, rhAPC preferentially binds to the VLA-3^high, a pro-inflammatory neutrophil population isolated from inflamed mice. The binding of rhAPC to human neutrophils was significantly reduced by a VLA-3 antagonistic peptide. Finally, we describe a genetic modification approach using a yeast surface-display system to develop recombinant APCs with high affinity for VLA-3 as potential therapeutic candidates to treat severe inflammation with a higher selectivity and an improved potency.

Materials and Methods

Sepsis mouse models

Endotoxemia and CLP were performed according to the Animal Resource Protocol approved by the Committee at University of Rochester. For the endotoxemia assay, 8–12-wk-old C57BL/6 (Harlan) male mice were weighed, and LPS (E. coli O55:B, Sigma-Aldrich) was administered by an intra peritoneal (IP) injection to achieve a LD90 mortality rate. The CLP survival surgery was performed under isoflurane inhalation anesthesia. The caecum was ligated and punctured through and through with a 21-gauge needle. All the mice were resuscitated with 1 ml of lactated Ringers injected subcutaneously. APC was administered via the tail vein (10 μg) at 1 hr and via retro orbital venous plexus at 4 hr, 24 hr and 72 hr (5 μg) following the CLP surgery. Survival was monitored for 120 hr. For the analysis of neutrophil migration into the inflamed lungs at 12 hr following endotoxemia, WT APC and R177G-APC were administered at 1 hr (via tail vein) and 4 hr (via retro orbital venous plexus) following the LPS injection.

Isolation and in vitro stimulation of neutrophils

Blood was collected from healthy volunteers via antecubital vein puncture in heparin containing vacutainers. The granulocytes and erythrocytes were separated from the whole blood by centrifugation through a 1-step polymorphs (Fresenius Kabi Norge AS) density gradient. The remaining erythrocytes were removed by hypotonic lysis, yielding a neutrophil purity of > 98%. The Human Research Studies Review Board of the University of Rochester approved this study, and informed consent was obtained in accordance with the Declaration of Helsinki. The neutrophils were stimulated with PMA (20 ng/ml), TNF-α (20 ng/ml), LPS (100 μg/ml) or fMLP (1 μM for 1 hr or 3 hr in L15 (Leibovitz) medium with glucose at 37°C.

Integrin ligand binding assay

The soluble integrin binding assay was performed using purified soluble human α₃β₁, α₅β₁, and α_Vβ₃ (United States Biological, Swampscott, MA); 1 μg/mL of the soluble integrins was mixed with a monoclonal antibody against the β₁ or β₃ subunit (mAb TS2/6 for β₁ and mAb D3 for β₃ integrins (kind gift from L. Jennings, University of Tennessee) plus rhAPC in the presence of 1 mM Mn²⁺ in L15 medium for 1 hour at room temperature. The protein complexes were precipitated with Protein A/G and were then subjected to SDS-PAGE and a western blot with a rat anti-human APC antibody (Cell Science, Norwood, MA) or silver staining.

For the APC binding inhibition assay, the neutrophils were isolated from healthy donors as described above and were incubated with 0, 0.1 μg, 1 μg and 10 μg APC along with blocking antibodies for integrin α₃, α₅, α_v or α_M (Millipore, Billerica, MA) in 1 ml of L15 medium for 30 minutes at 4°C. After washing, the cells were fixed with 3.7% formaldehyde for 10 minutes at room temperature. Following fixing, the cells were stained with rat anti-APC primary and a PE labeled goat anti-rat (Biolegend, San Diego, CA) secondary antibody for 30 minutes at 4°C in the dark and were then washed and resuspended in phosphate-buffered saline (PBS) for flow cytometry analysis. All the samples were collected on a FACS Calliber flow cytometer (BD Biosciences, San Diego, CA), and the data were analyzed using Flow Jo software.

Flow Cytometry

For the flow cytometry measurement of the expression integrins on the neutrophils, the bone marrow (BM), peritoneal Lavage (PL), peripheral blood and lungs were isolated from naïve and septic mice at the indicated time points, and subsequently, single-cell suspensions were prepared. An RBC lysis was performed using ACK lysing buffer (Invitrogen, San Diego, CA). The Fc receptors were blocked with unconjugated anti-CD16/32 (eBioscience, San Diego, CA) for 30 min. Samples were stained with Alexafluor 488 labeled anti-Gr1 (life technologies, Camarillo, CA). Allophycocyanin (APC) labeled Ly6G (BD Biosciences, San Diego, CA), PerCp-Cy5.5 labeled anti-CD11b (eBioscience, San Diego, CA,), fluorescein isothiocyanate (FITC) anti-F4/80 (eBioscience, San Diego, CA, USA), purified goat anti-mouse integrin α₃/CD49c (R&D Systems), and PE conjugated donkey anti-goat IgG (Santa Cruz) antibodies were used. For the EPCR surface expression studies, the samples were blocked with anti-CD16/32 and stained with FITC - anti-7/4 (AbD Serotec), PerCP-Cy5.5–anti-Ly6G (BD Biosciences), goat anti-mouse integrin α₃/CD49c (R&D Systems), PE conjugated donkey anti-goat IgG (Santa Cruz), and APC-EPCR (eBioscience). For comparing the WT APC R177G-APC, the lung cells were stained with Propidium Iodide (PI), Brilliant Violet anti-Ly6G (Biolegend, San Diego, CA), PerCp/Cy5.5 anti-CD11b (Biolegend, San Diego, CA) and anti-VLA3 antibody (R&D Systems) as described above. For the ex vivo activated protein C (APC) binding experiments, anti-human APC mAb (Hycult) was re-purified to remove BSA using the AbSelect purification kit (Innova Biosciences) and was directly conjugated to Allophycocyanin (APC) using the Lightning-Link labeling kit (Innova Biosciences). The BM cells were blocked with anti-CD16/32 and were then incubated with 100 ug/mL of human APC (Xigris, Eli Lilly) for 1 h on ice in staining buffer (Hanks’ Balanced Salt Solution supplemented with 1.5 mM CaCl₂, 1 mM MgCl₂, 10 mM Hepes and 3% fetal bovine serum) with 1 mM MnCl₂ to achieve integrin activation and were stained with FITC- anti-7/4 (AbD Serotec), PerCP-Cy5.5–anti-Ly6G (BD Biosciences), goat anti-mouse integrin α₃/CD49c (R&D Systems), PE conjugated donkey anti-goat IgG (Santa Cruz), and APC– anti-human APC conjugated as above. All the samples were fixed with 3.7% formaldehyde and were collected on a FACS Calliber flow cytometer (BD Biosciences, San Diego, CA). The data were analyzed using Flow Jo software.

Yeast display system

Human Protein C (coding for Gla-domain, light- and heavy chain) was used as a template for the following PCR. To randomize three residues, Lys146, Arg147, and Arg177, a two-step overlap PCR was applied. Primer Outer-For (GGT GGT GGT GGT TCT GGT GGT GGT GGT TCT GGT GGT GGT GGT TCT GCT AGC) and Focus-Rev (ATC TAC TTG GTC TTC TTG GTC TTC TGT GTC TCG TTT CAG GTG ACT GCG CTT CTT CTC CAT SNN SNN CCA GGG CCT CCC ACA AGG GAA) were used to randomize Lys147 and Arg148. Primer Focus-For (AAA CGA GAC ACA GAA GAC CAA GAA GAC CAA GTA GAT CCG CGG CTC ATT GAT GGG AAG ATG ACC NNS CGG GGA GAC AGC CCC TGG CAG) and Outer-Rev (GCC GCC GAG CTA TTA CAA GTC TTC TTC AGA AAT AAG CTT TTG TTC GGA TCC) were used to randomize Arg177, where N represents 25% of A, G, C, or T and S represents 50% of G or C. The two amplified DNA fragments were used as templates for the overlap PCR. The overlap PCR was performed with the Outer-For and Outer-Rev primers. The final PCR product was transformed into the yeast strain EBY100 with NheI/BamHI-restricted pNL6. The transformed yeasts were selected by culturing in selective media at 30°C and induced to express the library proteins by culturing them in induction media at 25°C for 20 hr (7). To activate Protein C on the yeast surface, 2.5 × 10⁷ inducing media-cultured yeast cells were washed with PBS and incubated with 10 μM bovine thrombin in 250 μl of 5 mM EDTA TBS pH 7.4 at 37°C for 3 hr.

Bio-Panning

HEK293 cells were grown to confluence on 6-well plates. Prior to panning, the HEK293 cells and yeast cells were washed twice with Hepes buffer, containing 1 mM Ca²⁺/Mg²⁺/Mn²⁺, 5 μg/ml of TS2/16 antibody, and 0.1% BSA. The mutant APC yeast cells were resuspended in the same buffer and were pipetted onto the HEK293 cells. The plates containing the yeasts and HEK293 cells were incubated at 25°C for 3 h. After the incubation, each well was washed twice with pH 7.5 Hepes buffer, containing 1 mM Ca²⁺/Mg²⁺/Mn²⁺, by gently swirling the plate 10 times in each direction. Then, the yeast-bound HEK293 cells were scraped off, resuspended in selective medium, and cultured at 30°C overnight to expand the recovered yeast clones. In parallel, a small aliquot of the mixture was plated on selective medium to quantify the number of recovered yeast cells. The yeast mixtures were then induced again for the next round of panning. This process was repeated until the number of binder yeast cells reached a plateau (8).

Activation of Protein C and Catalytic Activity against Small Substrates

Purified WT protein C or R177G protein C variant were incubated with human-thrombin (10:1 mol/mol) in 20 mM Tris-HCl, 100 mM NaCl, pH 7.4, at 37°C for 2 h in the presence of 5 mM EDTA, followed by the addition of hirudin (Sigma) to inactivate the thrombin. Q sepharose fast flow chromatography was used to remove thrombin. APC concentrations were estimated by measurement of absorbance at 280 nm. Amidolytic activities were determined using chromogenic substrate S-2366 (Chromogenix, Sweden) in a microplate reader and expressed in absorbance change at 405 nm. S-2366 (ranged between 33 and 667 μM) were added to APCs (67 nM) in 50 mM Tris-HCl, 130 mM NaCl, 10mM CaCl₂, pH 8.0 at room temperature. K_m and V_max values were obtained from Lineweaver-Burk plots. The experiments were performed in triplicate wells for each condition.

Neutrophil adhesion assay

The adhesion assay was carried out essentially as previously described (6). Cover slips were coated with LXY2 (10 μg/ml). The residual binding sites were blocked by incubating the wells with 0.1% (wt/vol) polyvinylpyrrolidone in PBS for 30 minutes at room temperature; 2.5×10⁵/250 μl of neutrophils was suspended in L15 medium plus 2 mg/mL glucose and was pretreated for 15 minutes at 37°C with rhAPC or R177G-APC. The cover slips were aspirated and washed with L15 medium; 250 μl of L15/2 mg/mL of the glucose medium containing rhAPC or R177G-APC, with/without 20 nM fMLP, was placed on each cover slip and prewarmed for 15 minutes at 37°C. The cells (250 μl) were then immediately added and were further incubated at 37°C for 15 minutes. The unbound cells were then washed with warm L15 medium. The bound cells were then fixed with formaldehyde. For each experimental condition from 3 independent donors, 5 random phase-contrast images were obtained, and the number of cells in each well was scored from the printed micrographs.

Data analysis

All the values are expressed as the mean ± SEM. All the statistics were performed using the GraphPad Prism 4.0 software. P value <0.05 was considered significant.

Results

APC binds to integrin VLA-3 on human neutrophils

Emerging evidence suggests that APC inhibits neutrophil migration both in vitro and in vivo, at least in part, by directly interacting with cell surface integrins through an Arg-Gly-Asp (RGD) motif (4–6). Human neutrophils express several RGD-binding integrins, including VLA-3 (very late antigen-3; α₃β₁), VLA-5 (α₅β₁), and α_Vβ₃ (9, 10). To determine the specific neutrophil integrin that interacts with APC with a high affinity, we performed a soluble integrin-binding assay. To this end, increasing concentrations of APC were incubated with equal amounts of activated, soluble VLA-3, VLA-5, or α_Vβ₃ integrins. An analysis of the integrin immunoprecipitates revealed that VLA-3 manifested a significantly higher affinity for APC, compared to the other two integrins (Fig. 1A). To further validate the strong and selective interaction between APC and VLA-3 on human neutrophils, flow cytometry-based APC-binding assays were performed. The amount of surface-bound APC was determined by measuring the mean florescence intensity (MFI) of a labeled antibody. The assay buffer, containing 1 mM Ca²⁺ and 1 mM Mg²⁺, was supplemented with 1 mM Mn²⁺ to activate all of the cell surface integrins, and this condition significantly increased the binding of APC to the neutrophil surface (Fig. 1B). The addition of blocking mAbs against VLA-5, α_Vβ₃, or Mac-1 (CD11b/CD18; α_Mβ₂) did not influence the binding of rhAPC to the neutrophil surface. Consistent with the soluble integrin-binding assay, the presence of a blocking mAb against VLA-3 significantly displaced APC from the neutrophil surface (Fig. 1B). Collectively, these results suggest that the major mechanism controlling the binding of APC to human neutrophils is through a direct interaction between APC and VLA-3.

Figure 1. Integrin VLA-3 is a receptor for APC.

(A) Binding assay of soluble VLA-3, VLA-5, and α_Vβ₃ to APC. Soluble integrins were incubated with the indicated amount of rhAPC in the presence of 1 mM MnCl₂ + integrin activating Abs. The protein complexes were precipitated and subjected to SDS-PAGE and western blotting with an APC Ab (upper panel). The total amounts of soluble integrins precipitated were shown by silver staining (lower panels). (B) rhAPC was incubated with human neutrophils in the presence of 1 mM MnCl₂. The cells were pretreated with control IgG (Con IgG) or blocking antibodies against integrins VLA-3 (α₃), VLA-5 (α₅), α_Vβ₃ (β₃), and Mac-1 (α_M). Bound APC was detected by flow cytometry. APC binding in the presence of integrin blocking antibodies is expressed as the percentage increase in MFI, compared with the cells not treated with APC (control) in the presence of the same blocking antibody. The data shows the mean ± SEM of three donors. * indicates significance between Con IgG vs. anti-VLA-3 (α₃). * p<0.05 (Student t-test).

Preferential binding of APC to the Gr1^highCD11b^highVLA-3^high neutrophil subtype

We recently showed that, in septic mice, there is a significant up-regulation of integrin VLA-3 on CD11b^highGr1^high neutrophils (2). Consistently, both mouse APC (mAPC) and LXY2, a specific blocking peptide for VLA-3, reduced the levels of the IL-6 cytokine, a measure of sepsis severity, in the serum from septic mice (Fig. 2A and 2B). The fact that APC binds to VLA-3 with a high affinity and that the expression of VLA-3 dramatically increases in the hyper-inflammatory neutrophil subtype suggests that APC may selectively target Gr1^highCD11b^highVLA-3^high pro-inflammatory neutrophils during sepsis. To assess the relative binding of APC to Gr1^highCD11b^highVLA-3^high vs. Gr1^highCD11b^highVLA-3^low neutrophils, we performed ex vivo binding assays using neutrophils isolated from septic mice. As shown in Fig. 2C, the cell-surface binding of APC to Gr1^highCD11b^highVLA-3^high cells was significantly enhanced, suggesting the preferential binding of APC to the pro-inflammatory neutrophil subtype. We previously showed that APC binds to EPCR and integrins simultaneously at the neutrophil surface, where EPCR provides support for integrin binding (6). The cell-surface expression of EPCR was measured by flow cytometry in the neutrophils from septic mice, thereby revealing that the Gr1^highCD11b^highVLA-3^high neutrophils expressed significantly higher levels of surface EPCR compared to the Gr1^highCD11b^highVLA-3^low cells (Fig. 2D). Thus, these data suggest that the enhanced expression of both VLA-3 and EPCR at the neutrophil surface allows APC to selectively target the Gr1^highCD11b^highVLA-3^high pro-inflammatory neutrophil subpopulation during sepsis.

Figure 2. Preferential binding of APC to Gr1^highCD11b^highVLA-3^high neutrophils.

(A) and (B) Serum levels of IL-6 were measured by a sandwich ELISA. The graph shows the OD values at 450 nm. * indicates significance vs. Con (PBS). * p<0.05 (Mann-Whitney test). (C) BM cells from LPS-induced septic mice were isolated at 12 hr and were incubated with rhAPC in the presence of 1 mM MnCl₂. Surface bound APC on the Ly6G^highVLA-3^high and Ly6G^highVLA-3^low populations among the mature and stimulated neutrophils was detected by flow cytometry. APC binding is expressed as the percentage increase in MFI compared with a no-APC control. The data are the mean ± SEM of three mice. * indicates significance between VLA-3^high vs. VLA-3^low. *p<0.05 (Student t-test). (D) BM and peripheral blood cells from LPS-induced septic mice were isolated at 12 hr, and the surface expression of EPCR was measured in the Ly6G^highVLA-3^high and Ly6G^highVLA-3^low populations by flow cytometry. The data are the mean ± SEM of six mice. * indicates significance between VLA-3^high vs. VLA-3^low. * p<0.01 (Mann-Whitney test). (E) & (F) The mice were injected with PBS or rhAPC (10 μg, IV). The lungs and PL were harvested after 8 hr and were digested with collagenase to prepare a single cell suspension. The cells were stained with CD11b, Ly6G and VLA-3. The bar graphs show the frequencies in the total granulocyte population (gated on Gr1^highCD11b^high) and the total number of VLA-3^high and VLA-3^low cells in the PL (E) and lungs (F). n=3/group. The data are the mean ± SEM of six mice. * indicates significance between PBS vs. APC. * p<0.05 (Mann-Whitney test).

To investigate whether the preferential binding of APC to the Gr1^highCD11b^highVLA-3^high cells results in delayed tissue infiltration by these cells during sepsis, we measured the percentage and number of cells in the peritoneum and lungs of septic animals after APC injection. The administration of APC significantly inhibited the infiltration of Gr1^highCD11b^highVLA-3^high cells into the peritoneal tissues, while the migration of Gr1^highCD11b^highVLA-3^low cells was not altered in the presence of APC (Fig. 2E). In the lungs, a reduction in the number of both infiltrating Gr1^highCD11b^highVLA-3^high and Gr1^highCD11b^highVLA-3^low cells in the presence of APC was observed (Fig. 2F). The APC-induced decrease in the number of Gr1^highCD11b^highVLA-3^high cells in the lung was more pronounced compared to the Gr1^highCD11b^highVLA-3^low cell population (Fig. 2F). These data indicate that the specific binding of APC to VLA-3^high granulocytes is sufficient to cause a delay in tissue infiltration by the pro-inflammatory Gr1^highCD11b^highVLA-3^high neutrophil population during sepsis.

A genetically modified mutant APC with a high affinity for VLA-3

In the crystal structure of the integrin α_Vβ₃ bound to the RGD peptide, the calculated buried surface area of the RGD motif is 373 Å (11). The solvent accessible surface area of the rhAPC RGD motif, which is comparable to the buried area of the complex, is 140 Å. This small-buried area is mainly attributed to the fact that the light chain covers a portion of the RGD motif in the catalytic domain (Fig. 3A). This property suggests that the accessibility of the RGD motif in wild-type (WT) rhAPC is not optimal for integrin binding. We also compared the structures of the rhAPC RGD motif and the RGD peptide in complex with α_Vβ₃. We found that the rhAPC RGD motif has different amino acid side-chain orientations than the integrin-bound RGD motif, suggesting the possibility of a steric clash when the WT rhAPC RGD motif binds to VLA-3.

Figure 3. Development of a mutant APC variant with a higher affinity for integrin VLA-3.

(A) The crystal structure of APC without the Gla domain (1AUT.pbd) is depicted in a ribbon diagram showing the EGF-like domain 1 (EGF1) in red, EGF2 in blue, and the heavy-chain in green (left). The region encircled by dots from the bottom-view is enlarged (right), representing the RGD sequence (residues 178–180) in the sphere and the mutational target residues (Arg-177 and Lys-146) on the mesh surface. (B) Schematic depiction of the yeast display. Aga2, HA, Protein C, and cMyc are expressed in order from the pNL6 vector. (C) Flow cytometry analysis of the expression of HA, c-Myc, and protein C. The samples were compared before and after activation with bovine thrombin. Note that the binding of anti-protein C Ab decreased after the surface protein C was activated by thrombin. (D) Microscopic analysis of each round of bio-panning on a confluent HEK293 cell monolayer during the enrichment of yeasts expressing the high-affinity mutant APC. The yeast cells were labeled with a fluorescent conjugated anti-HA antibody (left), and the fluorescence signal (red) was overlapped with the bright field image of HEK293 cell monolayer. (E) The recovered yeast cell numbers from each round of biopanning. (F) Flow cytometry analysis of WT and mutant APCs in the yeast cell surface with an anti-APC antibody. (G) The recovered yeast cell numbers from the binding assay (similar to bio-panning) with HEK293 cells using WT APC- and mutant APC-expressing yeast cells. The control condition was Hepes buffer without Mn²⁺/TS2–16. The binding was measured in three independent experiments and is presented as the mean ± SEM. * indicates significance between Con. vs. Mn²⁺/TS2–16. * p<0.05 (Student t-test).

To increase the affinity of APC for VLA-3 by altering the protein structure surrounding the RGD motif, we employed a yeast surface display system that was based on random mutagenesis and coupled it with a functional affinity screen (12). First, mutagenesis of the three amino acids adjacent to the RGD motif (Lys-146, Arg-147, and Arg-177) of WT human protein C was performed using site-specific randomized primers and two-step overlap PCR (Fig. 3A). Second, the mutant protein C variants were expressed and activated APC on the yeast surface (Fig. 3B & 3C). Third, the progressive enrichment of the cells encoding high-affinity APC was achieved by biopanning (8) (Fig. 3D and 3E) using HEK293T cells that do not express EPCR but express high levels VLA-3 (data not shown). Fourth, the genes isolated from the remaining yeast colonies were sequenced. A total of 68 yeast colonies, selected by panning, were sequenced. Based on the amino-acid substitution frequency at residue 177 and in an effort to minimize a possible steric hindrance by bulky positively charged side-chains at residues 146 and 147 (Table 1), we designed three mutant APC variants (R177G, K146G/R147G/R177G, and K146G/R177G) for further analysis. The yeast surface expression levels of these three mutant APCs were identical (Fig. 3F). However, only the R177G-APC expressing yeast showed a significantly increased binding to HEK293 cells in the presence of Mn²⁺ and mAb TS2/16 (β₁ integrin-activating Ab) (Fig. 3G).

Table 1.

Frequency of mutant amino-acid substitutions in the target residue. (*) represents wild-type amino acid.

Lys146		Arg147		Arg177
Mutation	Counts	Mutation	Counts	Mutation	Counts
R	11	R*	14	G	16
P	8	A	8	L, R*	7
L	7	P	7	A, S	6
A, G	6	S	6	H, M	4
T	5	L, T	5	T, V	3
S	4	N	4	I, N, P, Q, W	2
C, W, Y	3	F, Q, W	3	E, Y	1
D, I, K*, V	2	G, H, V	2
F, H, N, Q	1	C, E, I, K	1

68 clones were sequenced.

R177G-APC was purified from a mammalian cell culture system (HEK293), and enzyme assay using the chromogenic substrate, S-2366, indicated that the R177G mutation did not perturb amidolytic functions of APC, although it showed a slight decreased enzymatic activity comparing to that of WT APC (Suppl. Fig. 1). It is important to note, however, that Kerschen et al. showed that APC variants with reduced anticoagulant activities retain efficacy while reducing the risk of bleeding in severe sepsis (13). To determine whether R177G-APC binds to VLA-3 with a high affinity, we performed a soluble integrin-binding assay. An analysis of the immunoprecipitates by silver staining revealed that R177G-APC had a significantly higher affinity for VLA-3 compared to WT APC (Fig. 4A). To further demonstrate the strong and specific interaction between R177G-APC and VLA-3 on human neutrophils, we assessed cell adhesion to a cover glass coated with the VLA-3 specific peptide LXY2. Human neutrophils were allowed to adhere to immobilized LXY2 in the presence of fMLP. Compared to WT APC, the addition of R177G-APC significantly enhanced the inhibition of fMLP-induced adhesion (Fig. 4B). Furthermore, in the LPS-induced endotoxemia model, the systemic administration of R177G-APC significantly reduced the infiltration of CD11b^hiLy6G^hi neutrophils into the inflamed lungs compared to WT APC (Fig. 4C). These data further support a specific interaction between APC and VLA-3 and suggest that our approach using a yeast surface display system can serve as a blueprint for engineering better-targeted and safer APC variants to treat septic patients.

Figure 4. R177G-APC inhibits VLA-3-mediated neutrophil adhesion.

(A) Binding assay of soluble VLA-3 to WT APC and R177G-APC. Soluble integrins were incubated with the indicated amounts of WT APC or R177G-APC in the presence of 1 mM MnCl₂ and the mAb TS2/16 (β₁ integrin-activating Ab). The protein complexes were precipitated and subjected to SDS-PAGE and silver staining. The total amounts of APC (upper panel) and soluble VLA-3 (β₁ subunit) (lower panel) precipitated are shown. (B) The binding of 10 nM fMLP-treated neutrophils to immobilized LXY2 in the presence of various concentrations of WT rhAPC or R177G-APC. For each condition, the binding was measured in triplicate and is presented as the mean ± SEM. * p<0.05 versus fMLP-treated cells in the presence of WT APC (Student t-test). The data are representative of at least 3 independent experiments. (C) The mice were injected with PBS, WT APC (5μg or 10 μg, IV) or R177G-APC (5μg or 10 μg, IV) 1 hr and 4 hr after the LPS injection. The lungs were harvested after 12 hr, and a single cell suspension was prepared following a digestion with collagenase-1. The cells were stained with Propidium Iodide and CD11b, Ly6G and VLA-3 antibodies. The line diagram shows the total neutrophils (gated on Gr1^highCD11b^high granulocytes) in the lungs (n=3–4 mice/group, * p<0.05 (two-way ANOVA with a Bonferonni post-test)).

Discussion

The mechanism involved in the beneficial effects of rhAPC in severe inflammation is associated primarily with its cyto-protective attributes via EPCR and PAR-1 receptors (14). Previously, we showed that hyper-inflammatory subsets of neutrophils, with the potential to cause tissue injury during sepsis, express high levels of VLA-3 on their surface (Gr1^highCD11b^highVLA-3^high) (2). In this study, we showed that the enhanced expression of both VLA-3 and EPCR at the neutrophil surface allows APC to selectively target the Gr1^highCD11b^highVLA-3^high pro-inflammatory neutrophil subpopulation during sepsis (Fig. 2C & 2D). Therefore, we hypothesized that APC might bind more strongly to Gr1^highCD11b^highVLA-3^high pro-inflammatory neutrophil subset in vivo and thus show a superior suppression of migration in this population than in Gr1^highCD11b^highVLA-3^low neutrophils. As expected, administration of APC significantly inhibited the infiltration of Gr1^highCD11b^highVLA-3^high cells into the peritoneum and lung (Fig. 2E & 2F). Thus, our data suggest that rhAPC selectively targets VLA-3-expressing hyperactive neutrophils and improves survival in sepsis possibly by limiting their ability to infiltrate into the tissues of visceral organs.

Loss of endothelial barrier function is a prominent feature in the pathogenesis of severe inflammatory diseases. An increase in vascular permeability results in the massive infiltration of inflammatory cells into adjacent tissues, ultimately leading to organ failure during sepsis (15). EPCR was the first APC receptor identified on the endothelium (16). The anti-apoptotic and vascular-protective effects of APC are mediated by EPCR engagement and the secondary activation of G protein–coupled PAR-1 by the APC–EPCR complex (17–19). Importantly, many cell types that express EPCR, including endothelial cells, also express VLA-3. We showed that APC binds to EPCR and integrins simultaneously on the cell surface because the integrin-binding site on APC is distinct from the EPCR- binding site (6). Thus, APC may bind to EPCR and VLA-3 simultaneously at the endothelial cell surface, and this interaction could synergistically impact APC-mediated EPCR-PAR-1 signaling. It is also important to note that neutrophils express EPCR at their surface (Fig. 2D) (6). Second, these observations suggest that APC may not only prevent the massive infiltration of hyper-inflammatory neutrophils but may also improve EPCR-mediated vascular barrier protection through a simultaneous interaction with VLA-3.

Consistent with our findings in human neutrophils, a study in mice using murine APC demonstrated that APC binds to integrins on macrophages (20). In addition, the interaction between APC and integrin suppresses pro-inflammatory cytokine production. The authors proposed that Mac-1 is a novel APC receptor on macrophages that mediates the anti-inflammatory functions. In stark contrast, our study revealed that blocking Mac-1 in human neutrophils had no effect on APC binding (Fig. 1B). This discrepancy may be due to the different binding affinity of mouse Mac-1 to APC or due to the use of a different APC (human vs. mouse) in the experiments. In the aforementioned study, cell adhesion assays were performed mainly using surfaces coated with mouse APC, in which non-specific binding was blocked by pretreating the plates with bovine serum albumin (BSA) which is a strong Mac-1 ligand (21).

Our mutational studies demonstrated that R177G-APC showed a higher binding affinity to VLA-3 and inhibited the specific interaction between VLA-3 and its peptide ligand, LXY2, more effectively than WT APC did. According to the crystal structure of APC lacking the Gla-domain, there are hydrogen bonds between Arg-177 and Asp-180 (22). The bulky and positively charged side chains of the two adjacent arginines (residue Arg-177 and Arg-178) are positioned apart from each other, possibly because of the repulsive electrical force. Therefore, it is likely that the substitution of Arg-177 with Gly may release the electrical tension between the two arginines and provide enough flexibility to cause a conformational change in the RGD motif that is more favorable for a high-affinity binding to VLA-3.

The combination of the anticoagulant and cyto-protective properties of APC has made it an important clinical adjuvant for the treatment of severe sepsis. However, the anticoagulant properties have also led to increased bleeding events and resulted in a very narrow therapeutic window for this drug, ultimately leading to a voluntary withdrawal of the drug by the manufacturer. In this study, we demonstrated that VLA-3, which is specifically up regulated in a hyper-inflammatory neutrophil subpopulation in both human and murine sepsis, is a novel cellular receptor for APC. Moreover, the specific interaction of APC with VLA-3 selectively inhibits the migration of the pro-inflammatory neutrophil subpopulation during sepsis. In support, R177G-APC was more efficient at reducing the infiltration of neutrophils (CD11b^hiLy6G^hi) into the lungs of LPS-treated endotoxemic mice. Pursuing this alternative mechanism and optimizing APC binding to neutrophils will enable the development of better targeted, more effective, and safer sepsis therapies.