Cryo-EM structure of coagulation factor V short

Coagulation factor V (fV) is a large (2224 residues, MW 330 kDa) precursor of the cofactor activated fV (fVa) that together with the enzyme fXa as well as Ca²⁺ and phospholipids defines the prothrombinase complex and rapidly converts prothrombin to thrombin in the penultimate step of the coagulation cascade.^1,2 In addition, fV is directly involved in the regulation of the tissue factor pathway inhibitor α (TFPIα) and protein C pathways that inhibit the coagulation response.^2-5 After the removal of a signal peptide of 28 residues, fV secreted to the plasma features the domain structure A1-A2-B-A3-C1-C2 (Figure 1A). The A1 domain (residues 1-316) is connected to the A2 domain (residues 317-709) by a short segment composed mainly of basic amino acids (residues 304-316). An acidic segment (residues 657-709) transitions the A2 domain to the large and poorly conserved¹¹ B domain (residues 710-1545) that continues to the A3 (residues 1546-1877), C1 (residues 1878-2036), and C2 (residues 2037-2196) domains. The structural organization of fV has been elucidated recently using cryogenic electron microscopy (cryo-EM),¹² a technique capable of solving the structure of high-molecular weight proteins and their complexes under conditions not biased by crystal contacts^10,13-15 and particularly suited for the analysis of coagulation factors.^10,12,16-24 The A1-A3-C1-C2 domains are arranged as originally revealed by the bovine inhibited fVa X-ray crystal structure,²⁵ with the C1 and C2 domains forming a membrane-binding module supporting the A1 and A3 domains. New information garnered from the cryo-EM structure of fV includes identification of the sites of activation by thrombin at R709 and R1545, the sites of inactivation by activated protein C (APC) at R306 and R506, and the architecture of the A2 domain housing determinants critical for prothrombin activation, such as the lid (residues 672-691) and the gate (residues 696-702).^10,16 The structure has also served as a starting point for the subsequent elucidation of fVa, prothrombinase-free and bound to prothrombin.^12,16 

An unexpected finding from the cryo-EM structure of fV has been the high degree of disorder in the large B domain, which limited assignments to 4 residues (⁷¹⁰SFRN⁷¹³) at the N-terminal end connecting to R709 and 10 residues (¹⁵³⁶PDNIAAWYLR¹⁵⁴⁵) at the C-terminal end preceding R1545.^10,12 These small segments do not include the basic region (BR; residues 963-1008) and the acidic region (AR; residues 1493-1537) (Figure 1A) that are assumed to interact intramolecularly to keep fV in its inactive state for the control of both procoagulant and anticoagulant activities.^1,5-9 Hence, a key property of the B domain of fV remains unresolved. Elucidation of the architecture of the B domain also bears significant clinical relevance.^26-32 The East Texas bleeding disorder is associated with a splice fV isoform, called fV short, resulting in the deletion of residues from 756 to 1458 that include the BR and 31 tandemly arranged repeats of the B domain^4,31,32 (Figure 1B). Loss of the BR unmasks a high-affinity binding site in the AR for TFPIα, a key regulator of the assembly of prothrombinase^29,30 and blood coagulation.²⁸ Carriers of the East Texas mutation have high plasma concentrations of fV short bound to TFPIα that act in concert with protein S (PS) to inhibit fXa, which is required for the assembly of prothrombinase and procoagulant activity.^5-7,26,29,33 A structural understanding of fV short free and in complex with TFPIα, PS, and fXa would significantly advance the basic knowledge on fV and its multiple roles in hemostasis.

In this study, we present a 3.2 Å resolution cryo-EM structure of fV short that reveals the architecture of the entire A1-A2-B-A3-C1-C2 assembly. The shorter B domain is fully resolved and houses several hydrophobic clusters and acidic residues in the AR that likely provide a binding site for the basic C-terminal end of TFPIα. In fV, these epitopes may bind intramolecularly to the BR. The results offer new insights in our understanding of the mechanism that keeps fV in its inactive state and represent an important first step toward the structural elucidation of the anticoagulant properties of fV short in complex with TFPIα, PS, and fXa.

Materials obtained commercially were human plasma fV (#HCV-0100; Prolytix, Essex Junction, VT), human fV full-length complementary DNA (cDNA; #40515; ATCC, Manassas, VA), pDest40 expression vector (#12274015; Invitrogen, Waltham, MA), silver stain plus kit (#1610449; Bio-Rad, Hercules, CA), Gibco Expi293 expression kit (#A14635; Thermo Fisher Scientific, Frederick, MD), Quantifoil R 1.2/1.3, Cu 300 mesh grids (Q350CR1.3; Electron Microscopy Sciences, Hatfield, PA), PfuUltra high fidelity DNA polymerase (#600380; Agilent, Santa Clara, CA), NEBuilder HiFi DNA Assembly master mix (#E2621S; NEB, Ipswich, MA), simply Blue safe stain (LC6065; Invitrogen, Waltham, MA), and horseradish peroxidase-conjugated sheep anti-human fV immunoglobulin G (Enzyme Research Laboratory, South Bend, IN). fV short cDNA fragments were cloned via polymerase chain reaction, using full-length fV cDNA (#40515; Manassas, VA) as a template. The resulting fV short cDNA fragments were cloned into the pDest40 expression vector using Gibson assembly. To facilitate purification of recombinant fV short, a 12–amino acid HPC4 tag (EDQVDPRLIDGK) was fused to the C-terminus of the protein, as previously described.³⁴ fV short was then expressed transiently in Expi293 suspension cells per the manufacturer’s instruction. Eighteen or 20 hours after transfection, media was supplemented with the broad serine protease inhibitor PPACK (10 nM, final concentration) and soybean trypsin inhibitor (10 μg/mL) to protect against enzymatic degradation during production. Harvested conditioned media was spun down, further supplemented with benzamidine (10 mM), and purified initially on a Q Sepharose column (HiTrap Q Sepharose, Cytiva) with a salt gradient. Eluent from the Q-column was diluted and supplemented with CaCl₂ (5 mM) and purified via immune affinity using the Ca²⁺-dependent monoclonal antibody against HPC4³⁴ (supplemental Figure 1A-B, which is available on the Blood website). The tag fragment attached to the C-terminal of the C2 domain, used in the expression system, is clearly visible in the cryo-EM structure (supplemental Figure 1C). The recombinant construct of fV short corresponds to that of FV755-1459 in the Dahlbäck nomenclature^29,33 and carries a deletion of 703 residues compared with plasma fV, from S756 through D1458, including residue D756 (or N756) generated by a new codon across the splice junction between I755 and L1459 in the fV short transcripts from patients with East Texas bleeding disorder or healthy individuals.³² The deletion removes the BR (residues 963-1008), the site of thrombin activation at R1018, and yields a shorter B domain of 133 residues (Figure 1B). Our recombinant fV short cleaves prothrombin in a continuous assay,³⁵ with k_cat/K_m = 150 ± 3 μM^-1s^-1 vs 159 ± 1 μM^-1s^-1 for plasma-derived fVa (data not shown). This is consistent with the properties of fV short variants with residue 756 included^7,9 or carrying significantly longer deletions of the B domain, inclusive of the splice site, which retain fVa-like activity⁹ or the synergism between TFPIα and PS in the inhibition of fXa.²⁹ 

For cryo-EM studies, recombinant fV short was further purified and buffer exchanged at pH 7.4 into 20 mM HEPES, 150 mM NaCl, 5 mM CaCl₂ via size-exclusion chromatography, using a Superdex 200 increase 10/300 GL column. Peak fractions were collected and concentrated. At the time of freezing, grids were plasma cleaned on a Gatan Solarus 950 for 60 seconds; 3 μL of fV short (0.1 mg/mL) were applied onto each grid, blotted for 2 seconds, and then immediately plunge frozen in liquid ethane on a FEI Vitrobot Mark IV. Grids were loaded onto a Titan Krios G3 cryo-TEM operating at 300 kV equipped with a FEI Falcon IV (4k × 4k) direct electron detector. Screening and data collection were performed at a pixel size of 0.9 Å, using a dose of 55.09 e⁻/Ų across 46 frames, using a set defocus range of −1 to −2.4 μm. Exposures were collected from 2 grids under similar conditions. A total of 3316 micrographs were collected.

Patch-based motion correction, dose weighting, and patch-based contrast transfer function estimation were performed sequentially in cryoSPARC.³⁶ Particles were initially picked using Blob Picker job. Multiple rounds of 2D classification were performed to clean up the particle stack. Particles from the previous step were used to create an ab initio 3D construction volume. Initial models similar to fV protein data bank (PDB) 7KVE were combined into a single volume to create 2D templates for the template picker. Particles selected by the template picker were cleaned via multiple rounds of 2D classification and used to train Topaz neural network-based picker.³⁷ Particles from different types of picker jobs were combined, duplicate particles removed (using 0.75 particle diameter as threshold), and then cleaned with multiple rounds of 2D classification to produce the final stack. Particles were re-extracted by recentering using a 336 pixel box, and new ab initio 3D construction with multiple classes implementing a high similarity score was carried out to capture potential heterogeneity. 3D volumes were compared in UCSF Chimera.³⁸ The resultant 3D volumes deemed similar were pooled into a single nonuniform refinement, giving a consensus overall resolution of 3.2 Å (EMDB-29010). Masked local refinement of the A2 (EMDB-29008) and C1/C2 (EMDB-29009) with and without parts of the A1 and A3 domains of fV short was carried out to improve the density of mobile components and the disordered loops. To aid with model building, a composite map of the consensus refinement and local refined maps was generated by aligning the 3 maps in UCSF Chimera and taking the maximum value at each voxel (EMDB-29011). Maps were sharpened in Phenix,³⁹ and resolution was estimated using a gold-standard Fourier shell correlation cutoff of 0.143. An initial model for fV short was generated in COOT,⁴⁰ using PDB 7KVE and AlphaFold⁴¹ AF-P12259 for human fV. Initial models were docked and fit as a rigid body using UCSF Chimera. All atoms were refined with restraints, and model building was completed manually in COOT, using the composite map. Manual placement of residues in the B domain was partially guided by models generated using the CR-I-Tasser web server.⁴² The final model was refined against the composite map, with AlphaFold AF-P12259 and 7KVE as reference models for secondary structure restraints, using Phenix. Relevant parameters of the cryo-EM structure of fV short are summarized in Table 1. Representative 2D class averages and gold-standard Fourier shell correlation of the masked refinement of fV short maps are reported in supplemental Figure 2A-D.

The cryo-EM structure of fV short was solved at atomic (3.2 Å) resolution and reveals the A1-A2-B-A3-C1-C2 assembly (1493 residues) in its entirety for the first time (Figure 2; supplemental Figure 3). Overall, the structure is similar (root-mean-square-deviation (rmsd) = 2.22 Å over 1134 Cα atoms) to that of fV solved recently.¹² Unlike the highly disordered B domain of fV, the shorter B domain of fV short fits well in the electron density map, with a backbone and bulky side chains (Figure 3), and reveals new information of functional relevance. The sites of thrombin activation at R709 and R1545 are clearly visible in the A2 and B domains and are >80% exposed to solvent for proteolytic attack (Figure 2A). The sites of APC cleavage at R306 and R506 associated with fV^{Leiden4,43-47} are 46% and 35% exposed to solvent, respectively, as observed in fV.¹² The C domains are similar to each other (rmsd = 1.19 Å over 120 Cα atoms) and shaped as distorted jelly-roll β-barrels, as reported previously in the cryo-EM structure of fV,¹² the X-ray crystal structures of bovine inhibited fVa,²⁵ and the recombinant C2 domain of human fVa and fVIIIa.^48,49 Together, they define a membrane-binding module and support the A1-A2-B-A3 domains (Figure 2A). The new I755-L1459 junction at the splice site of our fV short construct has a well-defined density (Figure 3B) and sits 10 Å away from R1545 and 57 Å from R709. The C domains are almost identical to those of fV (rmsd = 0.88 Å over 115 Cα atoms for the C1 domain and rmsd = 1.19 Å over 130 Cα atoms for the C2 domain), and so are the A1 (rmsd = 1.02 Å over 291 Cα atoms) and A3 (rmsd = 1.07 Å over 262 Cα atoms) domains. Most of the differences between fV short and fV reside in the A2 domain (rmsd = 2.05 Å over 274 Cα atoms) that, in addition to the site of thrombin activation at R709 and the site of APC cleavage at R506, houses the gate (⁶⁹⁶YDYQNRL⁷⁰²) and the lid (⁶⁷²ESTVMATRKMHDRLEPEDEE⁶⁹¹) that play an important role in the prothrombinase complex.¹⁶ In fV short, the gate shifts 18 Å toward R709 and changes its conformation, with the side chains of Y696 and R701 pointing to opposite directions, as observed in fVa in the prothrombinase complex.¹⁶ More substantial rearrangements cause the lid to flip almost 90° counterclockwise and to expose the acidic segment ⁶⁵⁸PDDDEDSYEIFEPP⁶⁷¹ in close proximity to the AR (Figures 2A and 4A). The conformational change defines a locale for new interactions with positively charged moieties (see further) and the protease domain of fXa,^51,52 consistent with the role of the segment ⁶⁵⁹DDDED⁶⁶³ in controlling the rate of prothrombin activation.⁵⁰ 

Indeed, the structure of fV short and its electrostatic properties (supplemental Figure 4) define a suitable scaffold for the interaction with fXa and prothrombin that could explain the fVa-like activity of this variant.^1,5,8,9,14 A possible complex (Figure 5) built by replacing fVa with fV short in the recent structure of the prothrombin-prothrombinase complex¹⁶ would require only small conformational adjustments for productive interactions. Interestingly, the B domain provides no significant hindrance to the binding of prothrombin or fXa aligned against fV short, as observed in the prothrombin-prothrombinase complex.¹⁶ Repositioning of the lid in fV short pushes the gate against the backbone of the segment ²⁶⁰LDEDSDRAIE²⁶⁹ of the A chain of prothrombin.¹⁰ The clash removes the important interaction between the site of cleavage of prothrombin at R271 with D697 of the gate that directs the A chain to the active site of fXa to promote cleavage at R320 but would be corrected by small adjustments of the gate and the A chain. Repositioning of the lid also removes numerous interactions with fXa observed in the prothrombinase complex¹⁶ but exposes the acidic segment ⁶⁵⁸PDDDEDSYEIFEPP⁶⁷¹ (Figures 2A and 4A) for possible new contacts with the enzyme. A backbone clash between the AR segment ¹⁵³⁰INSSRDPDNIAAVYLR¹⁵⁴⁵ with the segment ⁸⁶RKLCSLDN⁹³ of the EGF2 domain of fXa may be removed by a modest (<3 Å) relative rearrangement of the 2 segments. No significant clashes involve fXa and other regions of the B domain, including the I755-L1479 junction at the splice site (Figure 3B).

Although much shorter than that of fV, the B domain of fV short reveals the architecture of the entire AR (residues 1493-1537) and its preceding segment preAR (residues 1458-1492) housing the hydrophobic patch ¹⁴⁸¹PLVIVG¹⁴⁸⁶ important for the recognition of TFPIα.^26,29,33 Overall, the overlap with the 4 residues at the N-terminal end connecting to R709 and 10 residues at the C-terminal end preceding R1545 detected in the structure of fV¹² is very poor (rmsd = 25.2 Å over 14 Cα atoms) because of a very distinct orientation of these segments in the 2 proteins. The B domain of fV short starts after R709 at the C-terminal end of the A2 domain and stretches across the entire width of the protein, making contacts with the A1, A2, and A3 domains (Table 2) but being suspended over and not in contact with the C1 and C2 domains (Figure 2A). After reaching the opposite side, near the site of cleavage at R1545, the B domain loops over the A3 domain in the backbone of the protein, changing direction and pointing upward toward the lid in the A2 domain (Figure 2B-C). The segment from 1503 to 1511 of the AR occupies the top of the loop where the B domain reverses course (Figure 4B) downward, toward R1545 (Figure 2B). The segment penetrates a crevice in the A2 domain with the ¹⁵⁰⁷EDDY¹⁵¹⁰ wedge, sandwiched between residues from 505 to 516 and from 665 to 672, and pushes the segment ⁶⁶⁸FEPP⁶⁷¹ to trigger relocation of the lid from its position in fV. The shift unmasks the acidic segment ⁶⁵⁸PDDDEDSYEIFEPP⁶⁷¹ (Figures 2A and 4A) next to the AR and generates an extended surface of negative electrostatic potential (supplemental Figure 4) for the engagement of fXa (Figure 5) or positively charged moieties, such as the BR or the C-terminal end of TFPIα. This provides a structural explanation of how TFPIα suppresses fVa-like activity in fV short⁷ and how intramolecular binding of the BR to this region in fV may keep the cofactor in its inactive state.^1,5-7,9,53 

The B domain starts a 180° turn at the ¹⁵⁰⁷EDDY¹⁵¹⁰ wedge to descend along the A3 domain toward R1545, thereby positioning the entire preAR and AR in the back of fV short (Figure 2C). Intramolecular interactions within the B domain include L1544 with I755, near R1545, and R1534 with N1480 (Figure 6). A strong H-bond connects R1534 to E1668 in the A3 domain and anchors the R1534-N1480 interaction to the surface of the A3 domain. The entire portion from I755 to L1544 defines 4 hydrophobic clusters (HCs) and a patch of acidic residues. The long HC1, from P753 to P1468, occupies the bottom of the B domain and contains 7 residues in a linear arrangement. HC1 continues with the hydrophobic patch ¹⁴⁸¹PLVIVG¹⁴⁸⁶, identified recently as a locale for the synergistic action of TFPIα and PS in mediating fXa inhibition.²⁹ The patch is part of a more extended cluster (HC2) that also includes F1474, L1475, F1479, and L1487. Residues P1663, W1665, and F1666 are nearby on the surface of the A3 domain (Figure 6), with F1666 interacting with V1487 in HC2. The B domain then continues into HC3, comprising I1497, I1498, V1516, P1517, and Y1518, sealing the ascending and descending portions of the B domain through an intramolecular interaction between I1498 and Y1518. Finally, the B domain reaches R1545 with HC4 defined by V1526, I1530, I1539, A1541, W1542, and L1544, which comes in contact with I755 in HC1. The HCs may contribute to the synergistic recognition of TFPIα and PS. Extensive deletions of fV short support the conclusion that the hydrophobic patch ¹⁴⁸¹PLVIVG¹⁴⁸⁶ in HC2, but not the segments preceding the patch, are required for this synergism.²⁹ The observation rules out involvement of HC1, but not HC3 and HC4, in TFPIα and PS recognition. The HCs encircle a group of negatively charged residues in the AR that define an acidic cluster (AC) with D1493 and E1496 in the ascending portion, D1519 and D1520 in the descending portion, and D1660, D1661, E1664, and E1668 on the surface of the A3 domain (Figure 5). The AC may work in conjunction with HC2 and the neighboring HC3 and HC4 to provide a locale for interaction with the C-terminal basic region of TFPIα in fV short and the BR in fV.

The cryo-EM structure of fV short advances the knowledge recently acquired on fV¹² by revealing new features of the B domain, especially the AR that, in fV, likely interacts with the BR to keep the cofactor in its inactive state.^1,5-9 The shape of the preAR and AR segments (Figure 6) and the electrostatics in this region (supplemental Figure 4) suggest that the BR may couple with negatively charged residues but also engage several HCs through hydrophobic interactions. These possible BR-AR intramolecular interactions have eluded the recent cryo-EM structure of fV because of the intrinsic disorder of the entire B domain.¹² The residues of the preAR and AR regions responsible for these interactions represent new targets for mutagenesis studies that will validate the current structure of fV short and future, high-resolution structures of fV with the B domain resolved in their entirety.

The cryo-EM structure of fV short offers a plausible mechanism for the interaction with TFPIα and PS. The AR has been implicated in the synergistic interaction of TFPIα and PS, leading to sequestration of fXa from the circulation and the mild bleeding phenotype associated with the East Texas bleeding disorder.^29,31-33 The preAR and AR of the B domain are organized in 4 HCs of various size that encircle several acidic residues in the AC. Interestingly, a computational model of the AR also identified some hydrophobic and acidic patches,⁵⁶ but most of their predicted architecture is not confirmed with the cryo-EM structure (Figure 6). HC2 includes the hydrophobic patch ¹⁴⁸¹PLVIVG¹⁴⁸⁶ responsible for the high-affinity binding of TFPIα and its synergistic cross talk with PS in the inhibition of fXa.²⁹ The effect is retained after deletion of the entire sequence from 713 to 756 or from 1458 to 1480 (67 residues) in the FV712-1481 construct,²⁹ in which only 3 residues separate the site of thrombin cleavage at R709 from the hydrophobic patch. The Cα atom of R709 is >75 Å away from residues of the hydrophobic patch (Figure 2A,B), but that separation may become <10 Å in the FV712-1481 construct. That should bring the hydrophobic patch much closer to R709 and significantly affect the architecture of the B domain. This scenario merits future structural investigation.

The HCs may work in conjunction with the AC in the AR to provide a locale for interaction with TFPIα and account for the much higher (100-fold) affinity compared with fV.^6,7,57 TFPIα binds several coagulation factors through its Kunitz domains. The first Kunitz domain inhibits fVIIa bound to tissue factor, and the second inhibits fXa.^6,29 Inhibition of fXa is enhanced by PS that binds to the third Kunitz domain⁵⁸ and functions synergistically with TFPIα in the inhibition of fXa on the surface of negatively charged phospholipids and platelets.⁵⁹ The third Kunitz domain is also required for optimal cell binding, together with the positively charged C-terminal domain.⁶⁰ The C-terminal BR of TFPIα has a sequence nearly identical to that of the B domain of fV that maintains the inactive state of the cofactor by binding to the AR.^1,5-7,53 Consequently, the C-terminal end of TFPIα binds to the AR of fV and protects it from catalytic attack at R1545.⁶¹ The cofactor activity of TFPIα requires the integrity of the R1545 site because cleavage at this site by thrombin abolishes TFPIα binding.⁶ Peptides corresponding to the C-terminal end of TFPIα competitively inhibit thrombin cleavage at R1545.^57,61 Interestingly, binding of TFPIα also takes place in variants of fVa that retain the AR, such as those released from activated platelets.⁶² The interaction inhibits prothrombinase assembled with these variants on the activated platelet surface by engaging fXa through the second Kunitz domain.⁵⁷ The tandem engagement of fXa and the AR by TFPIα may be critical to stabilization of the fV short/TFPIα/PS/fXa complex as well, with the third Kunitz domain binding to PS. Indeed, the complex must ensure tight interactions among its components on the surface of negatively charged phospholipids to compensate for the subnanomolar concentrations of fXa, TFPIα, and fV short in the blood.⁶ The cryo-EM structure of fV short suggests that TFPIα binds to the AR and the nearby acidic segment ⁶⁵⁸PDDDEDSYEIFEPP⁶⁷¹ through its basic C-terminal end and hinders the accessibility of R1545 for thrombin cleavage. In this position, TFPIα would also encroach on R506 (Figure 2AB), thereby explaining the difficulty of interaction with fV^Leiden (R506Q)^54,55 and the ability to protect R506 from APC cleavage.²⁷ Similar structural interactions are expected of the BR in fV and account for how intramolecular binding to the AR may keep the cofactor in its inactive state.^1,5-7,9,53 The third Kunitz domain engages PS likely bound to HC1, HC2, and HC4 of the AR (Figure 6) and in contact with fXa, bound along the A2, A3, C1 domains as observed in the prothrombinase complex¹⁶ (see Figure 5) and inhibited by the second Kunitz domain. In this scenario, TFPIα would strongly stabilize the fV short/TFPIα/PS/fXa complex by wrapping around fV short, fXa, and PS, with fXa bound as in the prothrombinase complex and next to PS bound to the back of fV short. A cryo-EM structure of the fV short/TFPIα/PS/fXa complex will test the validity of this hypothesis.

The authors gratefully acknowledge Tracey Baird for her assistance with illustrations.

This study was supported in part by the National Institutes of Health National Heart, Lung, and Blood Institute grants HL049413, HL139554, and HL147821 (E.D.C). M.J.R. is supported by the Washington University Center for Cellular Imaging, which is partly funded by Washington University School of Medicine, the Children’s Discovery Institute of Washington University, and St. Louis Children’s Hospital (CDI-CORE-2015-505 and CDI-CORE-2019-813), the Foundation for Barnes-Jewish Hospital (3770), the Washington University Diabetes Research Center (DK020579), and The Alvin J. Siteman Cancer Center at Barnes-Jewish Hospital and Washington University School of Medicine (CA091842).

Contribution: B.M.M. and L.A.P. performed biochemical preparations, structure determination, and analysis; M.J.R. performed cryo-EM data acquisition; and B.M.M., L.A.P., M.J.R., and E.D.C. analyzed the results and prepared the manuscript.

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

Correspondence: Enrico Di Cera, Edward A. Doisy Department of Biochemistry and Molecular Biology, Saint Louis University School of Medicine, 1100 South Grand Blvd, St. Louis, MO 63104; e-mail: enrico@slu.edu.