Structure of factor VII Gla domain bound to EPCR

Visual Abstract

TO THE EDITOR:

Over the past few decades, many groups have explored a potential interaction of the endothelial cell protein C receptor (EPCR) with other Gla-containing proteins. The debate gained clinical relevance with activated factor VII (FVIIa),1, 2, 3, 4 a current pharmacological intervention in the treatment of hemophilia, given the high circulating levels achieved upon administration of recombinant FVIIa (rFVIIa),⁵ which surpass those of endogenous FVII/FVIIa. This scenario raised the question of whether the EPCR-FVIIa interaction could play an important role in the hemostatic potential of rFVIIa. In vitro, EPCR can inhibit tissue factor (TF)–dependent FVIIa coagulant activity.¹ In vivo, it was found that EPCR potentiates rFVIIa’s hemostatic effect by competing with protein C.⁴

Despite numerous studies addressing the EPCR-FVIIa interaction, including mutational and metal ion–binding studies at the EPCR-FVII interface,1, 2, 3^,6,7 a structure that confirms a coherent EPCR-FVIIa docking and provides the binding signature had not yet been reported and remained elusive. Further, it was not known whether nonconserved FVII Gla residues could influence EPCR recognition.

Here, we present structural and binding studies to address this missing piece. The crystal structure of EPCR with bound FVII Gla domain was solved at 3.0 Å resolution, with 4 EPCR–FVII Gla complexes in the asymmetric unit (Figure 1A; supplemental Table 1). Overall, the binding mode is indistinguishable from that previously observed for protein C Gla domain⁸ (Figure 1B). Unambiguous electron density (Figure 1C) enabled the identification of the docking mode and the binding interface, including the ω-loop and metal ions that are part of the FVII Gla domain. FVII Gla domain binds EPCR through intermolecular and Ca²⁺-mediated ionic and H-bond contacts (supplemental Figure 1). As for protein C (PC), the FVII Gla ω-loop docks with a geometry defined by residues Phe4, Leu5, and Leu8 engaging the EPCR surface within a hydrophobic environment (supplemental Figure 2). Likewise, a set of up to 7 divalent cations, mainly Ca²⁺ ions, are distributed within FVII Gla residues, coordinating and driving a stable conformation that provides, ultimately, a productive binding to EPCR. We observed strong Fo-Fc electron density for 5 ion-binding sites consistent with a scattering power for Ca²⁺ ions rather than Mg²⁺ (Figure 1D). Two additional sites with weaker Fo-Fc peaks were more ambiguous because they could represent Mg²⁺ with full occupancy, but also Ca²⁺ with partial occupancy. These results agree with previous reports that demonstrated the binding of both types of ions to PC and FVII Gla.7, 8, 9

Figure 1.

Structural basis for FVII Gla domain binding to EPCR. The FVII Gla–EPCR complex was crystallized in the presence of CaCl₂ (5mM) and MgCl₂ (3mM). (A) FVII Gla domain (purple blue) is shown bound to EPCR (yellow), with coordinating Ca²⁺ and Mg²⁺ ions displayed as cyan and orange spheres, respectively. Fixed Ca²⁺ ions are depicted as cyan spheres, whereas those sites occupied by either Ca²⁺ (partial occupancy) or Mg²⁺ (full occupancy) ions are indicated with orange ions. The phospholipid molecule, also shown in sticks, is bound inside the groove formed by the 2 EPCR α-helices. (B) Structural superposition of the FVII Gla–EPCR complex (blue) with the PC Gla–EPCR complex (PDB 1LQV; gray). (C) The FVII Gla–EPCR structure is displayed in combination with experimental electron density. The 2Fo-Fc electron density maps are contoured at 1.0 σ (blue mesh for FVII Gla and yellow for EPCR), illustrating the experimentally determined docking and well-defined signal for both the FVII Gla domain and EPCR molecules at the binding interface. (D) Positive Fo-Fc electron density (green) determined by omitting the Ca²⁺ and Mg²⁺ ions (cyan and orange spheres, respectively) that contribute to FVII Gla overall fold and EPCR binding. (E) Two views of the residues in EPCR (shown as a gray surface) that directly participate in contacts with FVII Gla are indicated and highlighted in blue color. For reference, the footprint of PC Gla is also shown.

Structural alignment of FVII Gla–EPCR and PC Gla–EPCR (PDB 1LQV) complexes with EPCR as reference frame reveals a highly accurate superposition (0.305 Å) of both Gla domains. The ω-loops in each Gla show almost perfect overlap and orientation relative to EPCR, as well as near identical footprints on the receptor’s surface (Figure 1E). The overall docking geometry presents a buried surface area of 468 Ų for the binding interface (Figure 1E), which, again, is similar with that observed in the PC Gla–EPCR complex (467 Ų), suggesting that FVII Gla–EPCR interaction is solid and energetically coherent.

Notably, structural alignment of the EPCR-bound and TF-bound (PDB ID 2A2Q) FVII Gla domains reveals 2 Gla states defined by alternative architectures of the N-terminal ω-loop region (Figure 2A). We observed that EPCR competence is restricted to a repositioning of residues Phe4, Leu5, and Leu8 to enable a distinct orientation of the ω-loop relative to the receptor surface. The alternative conformation takes place without global rearrangement of the Gla fold and is spatially coupled to central (position 4) Ca²⁺ coordination. Following this line, although the first FVIIa-TF complex structure (PDB 1DAN) was determined in the presence of CaCl₂ with 7 bound Ca²⁺ ions within the Gla domain,¹⁰ a later study demonstrated the replacement of 3 Ca²⁺ ions by Mg²⁺ (PDB 2A2Q).⁶ Importantly, one such Ca²⁺ with Mg²⁺ substitution at position 4 leads the key conformation switch that dictates EPCR competence. Because under physiological conditions both Ca²⁺ and Mg²⁺ are present in human blood, mixed Ca²⁺/Mg²⁺ FVIIa assemblies could be more representative of predominant species in solution. However, the 4-Ca²⁺/3-Mg²⁺ configuration is not consistent with the EPCR-bound state observed in the present FVII Gla–EPCR structure. Structurally and chemically, the coordination geometry supports Ca²⁺ occupancy at position 4, yielding a metal ion arrangement similar to that found for the EPCR–protein C interface, that is, 5 Ca²⁺ and 2 Mg²⁺ ions.⁸ This assignment is consistent with additional coordination involving Asn2 and Gla7, which adopts a substantially different conformation relative to the EPCR-free structure (Figure 2B).⁶ In the EPCR–FVIIa Gla complex structure the resulting coordination environment is inconsistent with the canonical octahedral geometry normally observed for a 6-liganded Mg²⁺ ion. Instead, our data indicate a 7- or 8-coordinate Ca²⁺ geometry, depending on whether the Ca²⁺-ligand distance thresholds are set to 2.9 Å or 3.2 Å, respectively. In the latter case, coordination by the Ala1 backbone carbonyl oxygen would contribute to the eighth ligand. Improved R-factor and R-free values, consistent B-factors, and metal-oxygen distances characteristic of Ca²⁺ rather than Mg²⁺ coordination further suggests that position 4 is occupied with Ca²⁺. Taken together, these findings indicate that FVIIa/Mg²⁺/Ca²⁺ complexes with different stoichiometries could coexist in solution, but only those in which site 4 of the Gla domain is occupied by Ca²⁺ sustain EPCR binding.

Figure 2.

EPCR binding determines FVIIa Gla conformation in a fine-tuned divalent cation manner. (A) Structural alignment of the FVII Gla domain in its unbound (light blue), TF-bound (gray, PDB 2A2Q), and EPCR-bound (yellow) states, highlighting 2 different views the conformational differences concentrated in the ω-loop. The EPCR platform is shown in pale yellow for reference. (B) Close-up view illustrating the EPCR-dependent arrangement of the metal (magenta sphere) coordination environment at position 4. Coordinating interactions are indicated by dashed lines. Left panel corresponds to PDB 2A2Q. (C) rFVIIa was diluted in buffer containing defined divalent ion conditions (1.3mM CaCl₂, 0.6mM MgCl_2, or a combination of both), and the binding profiles to immobilized EPCR were monitored in real time by biolayer interferometry.

In addition, we solved the crystal structure of the unbound FVII Gla domain at 1.55 Å resolution. Crystals were obtained in high CaCl₂ concentration (100mM), then biasing all ion sites toward full Ca²⁺ occupancy. We observed that the crystallized unbound FVII Gla domain exhibits complete Ca²⁺-occupancy and an overall fold and ω-loop geometries near identical to those observed in the EPCR-bound complex (Figure 2A). Although the high Ca²⁺ concentrations do not align with the physiological concentrations found in plasma, the geometry obtained in the unbound FVII Gla structure closely matches that of the FVII Gla co-crystallized with TF in the presence of 2.5mM CaCl₂ and in the absence of MgCl₂.¹⁰ Collectively, these results demonstrate that EPCR competence is independent of receptor engagement.

Recombinant FVIIa is formulated in a Mg²⁺-free, Ca²⁺-containing solution (∼10mM CaCl₂), thus populating the EPCR-competent state. Dilution into plasma could challenge this state as physiologic Mg²⁺ levels could suppress EPCR engagement by rFVIIa.⁶ To this end, we performed real-time FVIIa-EPCR binding assays replicating plasma Ca²⁺ and Mg²⁺ levels. Under these conditions, FVIIa diluted in MgCl₂ (0.6mM) yield an association equivalent to that of FVIIa diluted in either CaCl₂ (1.3mM) or MgCl₂/CaCl₂ (0.6mM/1.3mM) (Figure 2C). However, Ca²⁺ dropped the dissociation rate, with the MgCl₂/CaCl₂ condition providing an optimal configuration that renders the most stable binding (slowest off-rate), in line with previous findings.^9,11 These findings indicate that physiologic Mg²⁺ concentrations do not prevent EPCR engagement when Ca²⁺ is present, therefore both circulating FVII and therapeutic FVIIa species associate with the EPCR-competent conformation.

In conclusion, Ca²⁺ occupancy at FVII Gla site 4 appears to stabilize the EPCR-competent conformation observed in the structure. Based on these results, we suggest that the EPCR-competent, Ca²⁺-loaded Gla domain conformation is an intrinsic solution-accessible state regardless of partner proximity, ruling out receptor-induced ion exchange. Together with previous structures, our findings point to the coexistence of a diverse population of FVIIa-Ca²⁺/Mg²⁺ assemblies in solution, with EPCR binding being favored by those where site 4 in the Gla domain is filled with Ca²⁺. These results highlight a fine-tuned metal-driven modulation of FVIIa biology, extending this behavior to therapeutic rFVIIa.

Conflict-of-interest disclosure: The authors declare no competing financial interests.