|Home | About | Journals | Submit | Contact Us | Français|
Factor (F)VIIIa activity and stability depends on the non-covalent association of the A2 subunit to the A1/A3C1C2 dimer, but the interactions that contribute to A2 association are not well defined. Previous work had shown that D666A and Y1792F mutations at the A2-A3 interface resulted in increased FVIIIa decay, suggesting that the residues were involved in bonding interactions important for FVIIIa stability.
Several potential hydrogen bonding partners of D666 and Y1792 across the A2-A3 interface were selected from the low-resolution FVIII crystal structure, and we used mutagenesis and biochemical analysis to examine the bonding interactions occurring at D666 and Y1792.
Using a series of stability and functional analyses, we analyzed FVIII variants in which D666 and Y1792 were each swapped with the residues of potential bonding partners.
We present evidence for hydrogen bonds between D666 and S1787 and between Y1792 and T657 that are important for FVIIIa stability. D666S/S1787D and T657Y/Y1792T variants each displayed WT-like FVIIIa stability and performed like WT FVIII in a series of functional analyses, whereas D666S, S1787D, and Y1792T single variants showed increased FVIIIa decay and a T657Y variant had little FVIIIa activity. These results suggest that WT hydrogen bonds are disrupted with the single mutations but maintained in the swap variants. Furthermore, mutation of D666 and S1788 to cysteine resulted in disulfide bond formation across the A2-A3 interface, confirming the close proximity between D666 and S1787, and this covalent attachment of the A2 subunit significantly increased FVIIIa stability.
Factor (F)VIII is a plasma protein that is decreased or defective in patients with hemophilia A. FVIII is translated as a single polypeptide chain with domains organized as A1A2BA3C1C2 [1,2], and cleavage produces the FVIII heterodimer that circulates in plasma . The heterodimer consists of a heavy chain (HC), made up of the A1A2B domains, and a light chain (LC), made up of the A3C1C2 domains, which associate in a Cu2+ dependent interaction .
Thrombin and FXa cleave and activate FVIII to the FVIIIa heterotrimer, made up of the A1, A2, and A3C1C2 subunits . A1 and A3 maintain the copper dependent interaction in FVIII, and A2 remains associated with A1 and A3C1C2 through weak electrostatic interactions [6,7]. FVIIIa functions as a cofactor to FIXa to form the FXase complex during clotting. When assembled on a negatively charged phospholipid surface, FXase catalyzes the conversion of FX to FXa with a catalytic efficiency increased several orders of magnitude over FIXa alone [8,9]. FVIIIa is inactivated as the A2 subunit spontaneously dissociates at physiological concentrations [10-13](see  for review).
Since A2 dissociation inactivates FVIIIa, understanding the A2-A3 interface interactions that regulate A2 stability is crucial to understanding down-regulation of the FXase complex during clotting. However, due in part to the lack of a FVIIIa structure, there is little knowledge of important interactions between A2 and A3 after FVIII activation. Identifying these interactions could lead to better understanding of FVIIIa stability and the mechanisms by which a subset of hemophilia A mutations result in increased A2 dissociation [12,15,16]. Enhancing interactions at the A2 interface to slow FVIIIa decay could produce more stable FVIII therapeutics for hemophilia A . The low resolution crystal structures (~4 Å) [4,18] of FVIII show an extended interface between A2-A1 and A2-A3, with many potential points of contact. Biochemical approaches are required to identify bonding interactions between side-chains in FVIIIa because (1) the orientation of side-chains cannot be determined at 4 Å, (2) the atomic coordinates of a number of residues in the two FVIII crystal structures (2R7E and 3CDZ) are significantly different, and (3) the FVIII structure is almost certainly different than that of FVIIIa [19,20].
Previously, the FVIII structures and the homology model derived from ceruloplasmin  have been guides for targeted mutagenesis to identify residues important for FVIIIa stability and A2 retention, as measured by biochemical analysis [22-25]. A panel of polar residues at the A2 interface were mutated to examine the contribution that the residues made to FVIII/FVIIIa stability , providing evidence that several residues at both the A2/A1 and A2/A3 interfaces were important for FVIII and FVIIIa stability. Notably, the results suggested that D666 and Y1792 at the A2/A3 interface contributed more to FVIIIa stability than to FVIII thermal stability . As FVIIIa decays when the A2 subunit dissociates from A1/A3C1C2 [10-13], we hypothesized that D666 and Y1792 are involved in bonding interactions at the A2 interface in FVIIIa that are important for A2 retention. According to the structures of FVIII [4,18], D666 of the A2 subunit and Y1792 of the A3 subunit are within reasonable distance to several potential bonding partners that might contribute to stabilization of the A2 subunit of FVIIIa.
While our previous work showed that D666 and Y1792 are important for FVIIIa stability, we continue this work by presenting biochemical evidence suggesting that in FVIIIa, D666 forms a hydrogen bond with S1787 and Y1792 forms a hydrogen bond with T657. We constructed variants in which the potential bonding partners were swapped and found that the swap variants, unlike the single variants, performed like WT FVIII/FVIIIa in a series of stability and functional analyses, suggesting that a WT hydrogen bond had been maintained. To our knowledge, these results are the first to present evidence of bonding interactions at the A2-A3 interface that have little effect on FVIII thermostability but are important for FVIIIa stability. Furthermore, we provide direct evidence for the proximity of D666 and S1787 by showing that mutation of both D666 and S1788 to cysteine results in a disulfide bond in FVIII and a significant increase in FVIIIa stability. As hemophilia A is treated with the supplement of FVIII, high stability FVIII variants could potentially be more effective in treating hemophilia A.
2D2 anti-A3 monoclonal antibody was a gift from Dr. Lisa Regan of Bayer Corporation (Berkeley, CA). Dioleoyl phospholipids phosphatidylserine/phosphatidylcholine/phosphatidylethanolamine (PSPCPE) were purchased from Avanti Polar Lipids (Alabaster, AL) and prepared as described . The reagents FVIII antibody R8B12 ((GMA012) Green Mountain Antibodies, Burlington, VT), α-thrombin, FVIIa, FIXaβ, FX, and FXa (Enzyme Research Laboratories, South Bend, IN), hirudin and chromogenic FXa substrate, Chromogenix S-2765, N-a-Z-D-Arg-Gly-Arg-pNA·2HCl (DiaPharma Group, Inc, West Chester, OH), Enhanced Chemifluorescence (ECF) reagent (GE Healthcare Bioscience, Piscataway, NJ), recombinant human tissue factor (rTF), thrombin fluorogenic substrate, Z-Gly-Gly-Arg-AMC (Calbiochem, San Diego, CA), and thrombin calibrator (Diagnostica Stago, Parsippany, NJ) were purchased from indicated vendors.
Mutagenesis (QuikChange, Stratagene, La Jolla, CA), subcloning and protein expression and purification to B-domainless FVIII were completed as described . WT B-domainless FVIII was purified and used as a comparator for all experiments. FVIII concentration was determined by western blot band quantification, and FVIII activity was determined by one-stage clotting and two-stage chromogenic FXa generation assays as described below.
FVIII preparations were resolved on an 8% polyacrylamide gel. Gels were stained with Gelcode Blue (Thermo Electron Corporation, Waltham, MA) or transferred to a polyvinylidene fluoride membrane (PVDF) for probing with anti-A2 (R8B12) and anti-A3 (2D2) monoclonal antibodies as described . Band densities from dilutions of each purified protein were quantified and compared to standards on the same gel or western blot to determine protein concentration.
One-stage clotting assays were performed by addition of FVIII to plasma chemically depleted of FVIII as described  and assayed using a Diagnostica Stago clotting instrument.
The rate of conversion of FX to FXa was monitored in a purified system at 23 °C as described . FVIII (1 nM) was activated to FVIIIa with α-thrombin (20 nM) for 1 min (or 15 seconds for rapidly decaying variants) in buffer containing 20 μM PSPCPE in FXa generation buffer (20 mM HEPES·HCl (pH 7.2), 100 mM NaCl, 5 mM CaCl2, 0.01% Tween20, and 100 μg/mL BSA). Thrombin was inhibited with hirudin (10 units/mL) for 1 minute, and the resulting FVIIIa was incubated with 40 nM FIXa for 1 minute. 300 nM FX was added to the mixture, and the reaction was stopped with 100 mM EDTA after an additional minute. The final concentrations of thrombin and hirudin were 16.7 nM and 8.3 U/mL, respectively. FXa generated was determined using the FXa chromogenic substrate, Chromogenix S-2765 (0.46 mM). Reactions were monitored for 3 minutes at 385 nm using a Vmax microtiter plate reader (Molecular Devices).
WT and FVIII variants (1.5 nM) were activated using thrombin for 1 min (or 15 seconds for rapidly decaying variants) at 23 °C in FXa generation buffer. Reactions were quenched with hirudin, aliquots were removed at the indicated times, and activity was determined using the FXa generation assay . The final concentrations of thrombin and hirudin were 15 nM and 7.5 U/mL, respectively.
WT and FVIII variants (4 nM) were incubated at 55°C in FXa generation buffer, aliquots were removed at the indicated time points, and activity was determined using the FXa generation assay . The final concentrations of thrombin and hirudin were 30 nM and 15 U/mL, respectively.
WT, D666C/S1787C, T657C/Y1792C, and D666C/S1788C FVIII (60 nM) were incubated with 120 nM thrombin for two hours at room temperature in 0.2 M NaCl FXa generation buffer. The samples were diluted into buffer with or without 50 mM dithiothreitol (DTT), 0.13 μg of FVIII was loaded per lane of a gel, and the samples were boiled for 1 minute. Western blot and analysis was completed as described above.
WT and FVIII variants (0.5 nM) in FXa generation buffer were activated with thrombin, quenched with hirudin, and then reacted with the indicated concentrations of FIXa. Activity was determined using a FXa generation assay .
WT and FVIII variants (0.5 nM) were activated with thrombin in FXa generation buffer and quenched with hirudin. FVIII was then reacted with FIXa, and FXa generation was initiated by adding the indicated concentrations of FX . The data were fitted to the Michaelis-Menten equation by nonlinear least squares regression, and parameter values were obtained.
The amount of thrombin generated in plasma was measured by calibrated automated thrombography  as described . Fluorescence signals were corrected by the reference signal from the thrombin calibrator samples  and actual thrombin generation in nM was calculated as described .
Values for FVIII/FVIIIa activity decay as a function of time were fitted to a single exponential decay curve by nonlinear least squares regression as described .
An estimation of FVIIIa-FIXa binding affinity (Kd) and Vmax for FXa generation were calculated from initial rate data by fitting the data using non-linear least-squares regression analysis to a single-site ligand binding model using the equation,
where AT is the total concentration of FVIIIa and BT is the total concentration of FIXa .
Since it was previously shown that D666 in the A2 domain and Y1792 in the A3 domain of FVIII contribute significantly to FVIIIa stability , we sought to identify the putative bonding partners. Using the FVIII 4 Å crystal structure , shown as a van der Waals surface display of the molecular structure model (Fig. 1A), we identified potential hydrogen bonding partners of D666 (including S1787, S1788, K1833, Y1837, Y1967) and Y1792 (including Y656, T657, N684) with α-carbons (Cα) less than 15 Å away from each target residue. A large distance was selected to account for the low resolution of the structure and conformational changes that could occur in this region upon activation. We performed swaps in which D666 and Y1792 were exchanged with the residue of a potential bonding partner. We expected that if the residues swapped were hydrogen bonding partners, a WT hydrogen bond, and therefore WT-like FVIIIa stability, would be maintained. Y1792 and Y656, both tyrosines, could not be swapped, and a D666Y/Y1837D variant was not generated, but the remaining swap variants were produced. The majority resulted in FVIII with low to no clotting activity, but D666 swapped with S1787 (9.4 Å separation) (Fig. 1B) and Y1792 swapped with T657 (13.8 Å separation) (Fig. 1C) produced FVIII variants with near WT activity that appeared to depend on both residues of the putative bonding pair.
Specific activities of the swap variants and corresponding single variants were determined by both one-stage clotting and two-stage activity assays. The D666S/S1787D swap variant had WT-like one-stage clotting activity, while the individual D666S and S1787D variants had 43% and 16% of WT activity, respectively (Table 1). Similarly, the T657Y/Y1792T swap variant had 47% of WT one-stage clotting activity, whereas activity of the T657Y and Y1792T single variants were reduced to 11% and 22% of WT, respectively. These results are consistent with the expectations of hydrogen bonding pairs.
The two-stage activity assay was completed with a short activation time to minimize A2 dissociation during activation and gave a similar result, as the D666S/S1787D and T657Y/Y1792T variants had close to WT two-stage specific activity values. Additionally, the D666S, S1787D, and Y1792T single variants had near WT two-stage specific activity values, suggesting that before A2 dissociation, these variants perform like WT FVIIIa. However, the T657Y single variant had a dramatically reduced two-stage activity value (12% of WT FVIII), and the restoration of activity of this variant by addition of the Y1792T mutation supports the presence of a hydrogen bond between the residues.
Analysis of FVIIIa stability showed that both the D666S/S1787D and T657Y/Y1792T variants had WT-like FVIIIa decay rates, which were substantially improved over those of the corresponding single variants. FVIIIa decay was measured as described in Materials and Methods. Our results (Fig. 2A, Table 1) show that the D666S/S1787D and T657Y/Y1792T swaps had WT-like FVIIIa decay rates. By contrast, the D666S, S1787D, and Y1792T variants decayed at rates that were 3.5-fold, 8-fold, and 7.6-fold faster than that of WT, respectively, and T657Y decayed too rapidly to measure, indicating that the WT stability observed in the swap variants is not due to any one of the individual mutations being stabilizing. Since FVIIIa decays as A2 dissociates , these results are consistent with WT hydrogen bonds disrupted with the individual mutations being maintained in the swap mutations.
By contrast, we note that although a D666S/S1788D swap variant had significant, but reduced, one-stage and two-stage activities (28% and 32% of WT, respectively), its FVIIIa decay was too rapid to measure. Thus, it does not appear that the D666S/S1788D swap variant has maintained a WT hydrogen bond that stabilizes FVIIIa (Table 1). The S1788D single variant, which had 5% of WT one-stage and two-stage activities, also had FVIIIa decay that was too fast to measure.
The D666S/S1787D and T657Y/Y1792T variants also displayed WT-like FVIII thermal stability. FVIII thermal decay at 55°C is a measure of the overall FVIII protein stability. FVIII (4 nM) was incubated at 55°C, and thermal decay was measured as described in Materials and Methods. The D666S/S1787D and T657Y/Y1792T variants had WT-like FVIII thermal stability (Fig. 2B, Table 1), while the corresponding S1787D and Y1792T variants had dramatically increased thermal decay rates (8.9-fold and 3.1-fold, respectively) compared to WT. The D666S mutation only marginally increased FVIII thermal decay (1.3-fold), consistent with the previously reported 1.4-fold increase in FVIII thermal decay of the D666A variant , and the T657Y variant had too little activity to measure decay. Because both of the swap variants rescued the decreased FVIII thermal stability of at least one of their corresponding single variants, this provides evidence for a WT hydrogen bond between the residues of each pair.
The D666S/S1787D and T657Y/Y1792T variants displayed WT-like affinity for FIXa, the protease with which FVIIIa interacts as a cofactor. To measure affinity (Kd) for FIXa, thrombin-activated FVIII (0.5 nM) was titrated with variable concentrations of FIXa, and FXase activity was measured by addition of FX. Since FXase activity of WT FVIII and the swap variants reached saturation, we could obtain an apparent Kd value for the FVIIIa/FIXa interaction (Fig. 3A, Table 2). The D666S/S1787D variant had a modestly increased Kd (1.9-fold) for FIXa compared to WT FVIII, with no significant effect on Vmax of the reaction (1.1-fold increase). Similarly, the T657Y/Y1792T variant had a modestly increased Kd (1.6-fold) and Vmax (1.5-fold). The near WT affinity and Vmax values of the swap variants provides evidence that these swap mutations at the A2-A3 interface do not affect FIXa interaction or FXase activity.
The FXase complexes formed with the D666S/S1787D and T657Y/Y1792T variants converted FX to FXa with Km values indistinguishable from WT FVIII. To examine the Michaelis-Menten kinetics of the FXase complex, thrombin activated FVIII (0.5 nM) and an excess of FIXa (40 nM) were titrated with variable concentrations of FX until saturation of FXase activity was observed (Fig. 3B, Table 2). The D666S/S1787D and T657Y/Y1792T variants had a WT-like Km values for FX (74% and 103% of WT, respectively), and for both variants the Vmax values were similar to WT (1.04-fold and 1.49-fold increased respectively). These results further demonstrate that the swap mutations do not affect the ability of the FVIIIa/FIXa complex to convert FX to FXa and are consistent with the maintenance of a hydrogen bond between the swapped residues. The WT-like stability and cofactor activity of the swap variants translated into near WT thrombin generation in a thrombin generation assay (supporting information Fig. 1A,B,C, Table 1). While the D666S/S1787D FVIII variant displayed WT-like peak thrombin generation and endogenous thrombin potential (ETP), the T657Y/Y1792T FVIII variant had slightly reduced thrombin generation values, consistent with the reduced one-stage activity.
To directly examine the proximity of D666 to S1787 and Y1792 to T657, we attempted to change the proposed bonding interaction to a disulfide bond. D666 of the A2 subunit and S1787 of the A3 subunit were mutated to cysteine with the expectation that if D666 interacts with S1787, the cysteine residues would form a disulfide bond across the A2-A3 interface. Similarly, Y1792 and T657 were mutated to cysteine.
The presence of a disulfide bond was then examined by western blot, looking for evidence of a peptide of ~120 kDa after thrombin activation, demonstrating covalent attachment of the A2 subunit to the A3C1C2 subunit. The D666C/S1787C and T657C/Y1792C FVIII variants did not show disulfide bond formation (Fig. 4A,B), and although a faint 120 kDa band was observed for the T657C/Y1792C variant in non-reducing conditions by the anti-A2 antibody, this band was not detected with the anti-A3 antibody. However, we found that a D666C/S1788C mutation did produce a FVIII variant with a disulfide bond, based on a prominent 120 kDa polypeptide observed in non-reducing conditions (but not in reducing conditions), which was detected with both anti-A2 antibody and anti-A3 antibody (Fig. 4C, compare lanes 4 and 8). This 120 kDa polypeptide was not found in the WT control (lanes 3 and 7), and its presence correlates with the complete absence of the corresponding A2 subunit (lane 4) and decreased amounts of the A3C1C2 subunit (lane 8). The fact that all of the A2 subunit is covalently attached to A3C1C2 may suggest that disulfide bond formation is nearly 100% efficient, however, as some A3C1C2 remains, the lack of free A2 subunit in lane 4 may be due to lower sensitivity of the western blot for the A2 subunit.
Like the corresponding D666S/S1788D swap variant, the D666C/S1788C variant had similarly reduced one-stage and two-stage activities (31% and 46% of WT respectively, Table 1). However, the D666C/S1788C variant had almost no detectable FVIIIa decay. While WT FVIIIa decayed to ~20% activity within 13 minutes of thrombin activation, the D666C/S1788C variant retained ~90% activity 60 minutes after activation, consistent with little or no A2 dissociation (Fig. 2C). Indeed, the calculated FVIIIa decay rate was 1% of WT. Covalent attachment of A2 to A3 also contributed to the thermal stability of the D666C/S1788C variant, as the variant decayed at a decreased rate (66%) compared to WT FVIII (Fig. 2D). The extraordinary stability of the D666C/S1788C variant supports the conclusion that D666 is within close proximity of S1788, and therefore S1787, in FVIIIa. Investigation of the interaction of the D666C/S1788C variant with FIXa and FX suggests that activation to FVIIIa results in a slightly different conformation as this variant had reduced affinity for FIXa upon activation and a reduced Km for FX and Vmax for FXase activity (supporting information Fig. 2A,B,C, Table 2). Consistent with high stability, the D666C/S1788C variant produced more thrombin in plasma (supporting information Fig. 1D,E, Table 1), and given the reduced one-stage specific activity of the D666C/S1788C variant (31% of WT), the increase in thrombin generation testifies to the stability added due to covalent attachment of the A2 subunit to the A3 subunit.
This study provides experimental evidence that D666 forms a hydrogen bond with S1787 and Y1792 forms a hydrogen bond with T657 at the A2-A3 interface in FVIIIa. Our earlier study suggested that D666 and Y1792 were involved in bonding interactions important for FVIIIa stability . Although detailed atomic coordinates are different among the structures and homology model of FVIII [4,18,21], they showed D666 and Y1792 within close proximity to S1787 and T657, respectively. Swapping the side-chains of D666 with S1787 and Y1792 with T657 produced variants with WT FVIIIa decay rates, whereas the D666S, S1787D, T657Y and Y1792T single variants decayed at least 3.5-fold faster than WT. The WT FVIIIa decay displayed by each swap variant is consistent with the maintenance of a hydrogen bond at the A2 interface, a bond that is disrupted with the individual mutations. Additionally, the D666S/S1787D and T657Y/Y1792T variants showed WT-like FVIII thermostability, affinity for FIXa, Michaelis-Menten kinetics for FX activation, and thrombin generation in plasma demonstrating that they are normal in all functions.
The close proximity of D666 to S1787 was confirmed by generating a disulfide bond between D666 and S1788. A disulfide bond did not form in D666C/S1787C or T657C/Y1792C variants, consistent with the Cα distances between these pairs (9.4 Å and 13.8 Å respectively ) exceeding the range of typical cysteine Cα distances (4-7 Å) for disulfide bond formation . However, the Cα distance between D666 and S1788 does fall within this range (6.8 Å), and the D666C/S1788C variant formed a disulfide bond across the A2-A3 interface and had substantial activity. D666 and S1788 do not appear to be hydrogen bonding partners in FVIIIa as a D666S/S1788D swap decayed rapidly.
Covalent attachment of A2 to A3 resulted in a FVIIIa decay rate only 1% of the WT decay rate and contributed to FVIII thermal stability, reducing it 66% of the WT rate. The increase in FVIII and FVIIIa stability resulted in prolonged thrombin generation in a thrombin generation assay. Three other stable disulfide variants of FVIII have been described in which A2 is crosslinked to A1 or A3, and these variants also display significantly increased FVIIIa stability [22,33,34].
The substantial activity of the D666C/S1788C variant has implications for the range of conformational changes that occur during FVIII activation. FVIII undergoes a conformational change upon activation to enable FIXa binding . However, the activity of the D666C/S1788C variant implies that there is only at most a modest change in this region, since a stable disulfide bond would prevent more extensive conformational changes. Nonetheless, the disulfide bond FVIIIa variant had altered FIXa affinity and FX Michaelis-Menten parameters, suggesting that the constraints of the disulfide bond do limit or alter some of the changes occurring. In contrast, a H281C/S524C variant linking A2 to A1 had only marginal effects on FIXa affinity and FX Michaelis-Menten parameters , perhaps indicating even less of a conformational change in this region upon activation.
There are a significant number of residues lining the A2 interfaces that are stabilizing or destabilizing to A2 after FVIIIa activation (Fig. 5 summarizes). A number of polar and charged residues contribute to FVIIIa stability, likely through A2 association, as mutations to these residues result in greater than 2-fold increases in FVIIIa decay rates or increased dissociation of A2 (Fig. 5, green) [25,35]. Several charged residues at the A2 interface are destabilizing in WT FVIIIa as mutations to these residues result in increased FVIIIa stability or A2 affinity (Fig. 5, black) [30,35-37].
There is also a significant set of FVIII variants with increased A2 dissociation that were originally identified based on patients with hemophilia A, whose plasma had a characteristic discrepancy between one-stage clotting and two-stage clotting or chromogenic assays implicating a role for these residues in A2 association and FVIIIa stability (Fig. 5, orange) [38-42]. However, as with all destabilizing mutations, the increased dissociation of A2 may be due to the polarity or charge changes conferred by the mutations themselves, rather than being due to the specific stabilizing effect of the WT residue. For instance, the hemophilia A mutations S1791P and R1966Q localized to the A2-A3 interface have a one-stage/two-stage discrepancy according to the Hemophilia A Database, but we have previously shown that S1791A and R1966A mutations have little effect on FVIIIa stability .
Overall these studies emphasize an intricate network of both stabilizing and destabilizing electrostatic interactions at the A2 interface of FVIIIa. The data presented here provide evidence of two hydrogen bond interactions between A2 and A3 in FVIIIa, one between D666 and S1787, and the other between T657 and Y1792. In many of the other cases cited above, it is not known how the mutations affect A2 interactions. Future experiments will undoubtedly reveal these mechanisms to elucidate how A2 dissociation is balanced to promote optimal FXase activity.
We thank Lisa M. Regan of Bayer Corporation for the 2D2 antibody, Pete Lollar and John Healey for the FVIII cloning and expression vector, and Eric M. Phizicky for insights and assistance with writing the manuscript. This work was supported by NIH Grant HL038199 originally to P.J.F and currently to H.W. M.M. was supported by NIH Training Grant in Cellular, Biochemical and Molecular Sciences T32 GM068411.
Author contributions: P. J. Fay, H. Wakabayashi and M. Monaghan conceived and designed the study. M. Monaghan performed and analyzed the experiments and wrote the paper. A. E. Griffiths provided technical assistance. M. Monaghan, H. Wakabayashi and A. E. Griffiths reviewed the results and approved the final version of the manuscript.
Conflict of Interest: M. Monaghan, H. Wakabayashi and A. E. Griffiths report grants from NIH during the conduct of the study.