Allosteric Activation of Coagulation Factor VIIa Visualized by Hydrogen Exchange*

Coagulation factor VIIa (FVIIa) is a serine protease that, after binding to tissue factor (TF), plays a pivotal role in the initiation of blood coagulation. We used hydrogen exchange monitored by mass spectrometry to visualize the details of FVIIa activation by comparing the exchange kinetics of distinct molecular states, namely zymogen FVII, endoproteolytically cleaved FVIIa, TF-bound zymogen FVII, TF-bound FVIIa, and FVIIa in complex with an active site inhibitor. The hydrogen exchange kinetics of zymogen FVII and FVIIa are identical indicating highly similar solution structures. However, upon tissue factor binding, FVIIa undergoes dramatic structural stabilization as indicated by decreased exchange rates localized throughout the protease domain and in distant parts of the light chain, spanning across 50Å and revealing a concerted interplay between functional sites in FVIIa. The results provide novel insights into the cofactor-induced activation of this important protease and reveal the potential for allosteric regulation in the trypsin family of proteases.

Coagulation factor VII (FVII) 2 circulates in the blood with ϳ1% in a two-chain form (FVIIa) and the remainder as singlechain zymogen FVII. FVIIa consists of a trypsin-like protease domain and an N-terminal light chain composed of a membrane-binding ␥-carboxyglutamate-rich domain (Gla domain) and two epidermal growth factor-like domains (EGF1 and EGF2) (1). Upon tissue injury, tissue factor (TF) becomes exposed, and the TF⅐FVIIa complex forms and serves as the initiator of the blood coagulation cascade (2). Numerous coagulation proteases function optimally only when complexed to cofactors, and substantial biochemical evidence supports the concept that several of these cofactors induce allosteric changes in the conformation of their cognate enzymes (1,2). TF functions to localize FVIIa and as an allosteric regulator, which dramatically enhances the activity of FVIIa. Similar to the trypsinogen-trypsin pair, zymogen FVII is converted to FVIIa by proteolysis of an internal peptide bond. A canonical "activation domain" was defined in trypsin, which includes the N terminus created upon endoproteolytic activation and three loops referred to as the activation loops (3). In trypsin, the newly generated N-terminal tail spontaneously inserts itself into a cavity close to the three activation loops, termed the activation pocket, resulting in the formation of a critical salt bridge between the N-terminal Ile-16 and Asp-194 of the active site or Ile-153 (16) and  in FVIIa (the chymotrypsin numbering is denoted in superscript with parentheses). This salt bridge leads to the formation of a correctly assembled S1 pocket of the active site and full activity. FVIIa, however, has very low activity following endoproteolytic activation and does not spontaneously rearrange into the active form. The three-dimensional structure of FVIIa bound to TF has been solved providing important structural details of the complex and the active form of FVIIa (4). In the complex, parts of the light chain and the protease domain of FVIIa form an extended interface with the N-terminal domain of TF (Fig. 1A). Crystal structures of FVIIa in the absence of TF had also been solved, but crystallizations were performed in the presence of active site inhibitors that locked FVIIa in the active form (5,6). No structural information is available for zymogen FVII and FVIIa in solution as they circulate in the blood prior to vascular injury, and the structural details of TF-induced activation are therefore unclear.
Amide hydrogen exchange (HX) monitored by mass spectrometry (MS) has proven to be a powerful technique for investigating the conformational properties of proteins in solution (7)(8)(9). Altered HX rates between protein states can act as sensitive probes for changes in the dynamics of protein structure, and for native protein structures, a decrease in the observed exchange rate correlates with increased local structural stability and vice versa (10). Using HX coupled with MS detection, we have studied zymogen FVII and FVIIa along with the effects of TF binding to these different FVII/FVIIa forms. We also studied the HX kinetics of FVIIa in the presence of an active site inhibitor (IN18), which is derivatized from the amidinophenyl family of inhibitors (11) known to lock FVIIa in the active conformation (5). Both TF and IN18 induce the active conformation but bind at different sites on FVIIa, and the latter data set is * This work was supported by a scholarship from the Danish Ministry of Science, Technology, and Innovation (to K. D. R.) and by grants from the Danish Biotechnology Instrument Center and the Carlsberg Foundation (to T. J. D. J.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Fig. 1  important for distinguishing between HX effects (12) observed in FVIIa upon TF binding. Regions that are affected by both molecules represent regions in FVIIa responding to activation, whereas regions that respond to only one of the molecules can be disregarded as due to the steric effects at binding interfaces. With these tools, we were able to characterize the structural properties of the different states of FVII/FVIIa, which are of biological relevance during activation in the blood. Regions of FVIIa undergoing structural rearrangements or changes in dynamics upon transition to the active form are revealed, providing novel insights into the molecular nature of the allosteric activation of this cofactor-dependent trypsin-like protease.

MATERIALS AND METHODS
HX Sample Preparation-Recombinant human FVIIa and FVII R152A (13) and soluble human TF-(1-219) (14) were expressed and purified as reported previously. The amidinophenylurea-based FVIIa inhibitor IN18 was synthesized and purified essentially as described (Ref. 11; IN18 is identical to compound 18 in this reference).
FVIIa and FVII R152A were deglycosylated to enable MS detection by 5 units/l PNGase F (New England Biolabs) in 10 mM glycyl-glycine, pH 6.6, 50 mM NaCl, 10 mM EDTA at 30°C for 18 h. N-Deglycosylated FVIIa was subsequently purified by anion-exchange chromatography and eluted in 10 mM Tris-HCl, pH 8.0, 50 mM NaCl, and 15 mM CaCl 2 . All proteins were buffer-exchanged and concentrated in 20 mM bis-Tris, pH 6.0, 25 mM CaCl 2 , 200 mM NaCl, 3% glucose and stored at Ϫ80°C. Protein concentrations were determined by A 280 , and full activity of FVIIa was verified by an amidolytic assay in the presence of TF as described (15). Amide 1 H/ 2 H hydrogen exchange was initiated by a 12-fold dilution of a protiated protein stock solution in the presence or absence of cofactor/inhibitor into the corresponding deuterated buffer (i.e. 20 mM bis-Tris, 10 mM CaCl 2 , 99% D 2 O, pH 6.0 (uncorrected value)). A low salt buffer was used to avoid solubility problems during preparation of the stock solution of exchange-in buffer. Furthermore, 10 mM CaCl 2 was used in the exchange buffer rather than a more physiological concentration to compensate for the decreased affinity of FVIIa for Ca 2ϩ at pH 6.0. pH 6.0 was used in the exchange-in experiments to allow for improved resolution of fast-exchanging amides. Nondeuterated controls were prepared by dilution into an identical protiated buffer. Experiments at pH 7.4 were performed in 10 mM Tris-HCl, 150 mM NaCl, 5 mM CaCl 2 , 99% D 2 O, pH 7.4 (uncorrected value). All HX reactions were carried out at 25°C and contained 4. Rapid Desalting and MS Analysis-The apparatus for rapid desalting was configured as described (16) with minor modifications. The setup consisted of two high pressure liquid chromatography pumps (Agilent 1100 series), an injection valve (Rheodyne 7725i) with an injection loop containing a 4.3 ϫ 50-mm column packed with immobilized pepsin (Pierce), and a 10-port switching valve (Valco Instruments Inc.) equipped with a micro-column (approximately 2-l bed volume of Poros 20R2) in addition to an analytical C 18 column (Jupiter 4u, Phenomenex). The entire plumbing system was immersed in an ice bath. Samples were loaded directly onto the pepsin column and digested for 2 min. The resulting peptide mixture was flushed onto the micro-column and desalted for 3 min at 250 l/min in 0.05% trifluoroacetic acid. Subsequently, the micro-column was automatically switched in line with the analytical column, and peptides were eluted into an electrospray ion source at 50 l/min by a 10-min gradient of 20 -80% acetonitrile in 0.05% trifluoroacetic acid. Positive ion-electrospray ionization mass spectra of eluted peptides were acquired on a LCT mass spectrometer (Waters Inc.).
Data Analysis-Peptic peptides were identified in separate experiments using standard MS/MS methods. Average masses of peptide isotopic envelopes were determined from lockmasscorrected centroided data (processed using MassLynx software, Waters Inc.) using an Excel spreadsheet. Complete deuteration of control samples was achieved by incubation for 70 h at 37°C in the presence of 6 M guanidinium chloride (deuterated). Average back-exchange (i.e. deuterium loss) was measured to be 14% for the analyzed peptides. However, no corrections were made for this deuterium loss because only the relative levels of deuterium incorporation of all samples were compared. Mass determination of replicate samples of FVIIa at 100 s of deuteration showed a standard deviation in mass measurements, Ͻ 0.1 Da (n ϭ 4). Representative HX curves of deuterium incorporation versus time were fitted by nonlinear least squares regression to a triexponential equation as described (16). In all fits, k fast was set to a value of 20 s Ϫ1 . Fits to the triexponential equation were not significantly influenced by the random error of the experimental data (S.D. Ͻ 0.1 Da).

RESULTS
HX kinetics of zymogen FVII (inactivable FVII R152A ) and FVIIa in the presence or absence of TF along with FVIIa in the presence of the active site inhibitor IN18 were studied. HX experiments were initiated by dilution of proteins into deuterium, and the subsequent time-resolved deuterium incorporation was localized to various regions of FVII/FVIIa by mass analysis of peptides produced by peptic proteolysis. In this manner, 57 peptides of FVIIa ( Fig. 1B; supplemental Fig. 1, A-C) were identified and used to monitor HX kinetics of 311 of 407 amide hydrogens (44% FVIIa light chain, 98% FVIIa protease domain). Corresponding peptides were similarly identified for zymogen FVII R152A (supplemental Fig. 1, A-C), allowing comparison of HX kinetics relative to FVIIa. HX time course plots of representative peptides were fitted to a triexponential model to categorize exchanging amides into a fast kinetic component and two slower kinetic components (see supplemental Table 1). All exchange-in experiments were carried out at pH 6.0 to better resolve the fast exchanging amides and in the presence of Ca 2ϩ . To ensure a correlation between the experimental conditions and those found in the blood, the HX kinetics of FVIIa and TF-bound FVIIa were also monitored at physiologi-cal pH and ionic strength (pH 7.4, 150 mM NaCl) and found to reliably reflect those observed at pH 6.0 (data not shown). Furthermore, an exchange-in experiment using fully glycosylated FVIIa in the presence and absence of TF showed similar effects upon TF binding as observed for N-deglycosylated FVIIa (data not shown).
Visualization of the Allosteric Response of FVIIa-The transition of FVIIa into an active form following binding of TF (or binding of the active site inhibitor IN18) results in dramatically decreased deuterium incorporation (i.e. HX rates) as shown in Fig. 1C for a peptide spanning ␤-strand A2. These effects are primarily observed in peptides of the second ␤-barrel and several surface loops, and thus the majority of peptides displaying reduced HX rates are seen to localize across one face of the protease domain of FVIIa (Fig. 2). In all the affected regions of the FVIIa structure, the active (TF-bound) form displays decreased HX rates compared with the free form, indicating that the active form is more stable and less dynamic than the free form. As illustrated in Fig. 2, decreased HX rates of the active form are seen in the TF-binding helix (␣1) and the neighboring 170-loop (peptide 306 -325 (164 -178) ). Regions around the active site cleft also show decreased HX rates as seen in activation loop 3 (localized to residues 370 -377 (221-228) in peptide 361-377 (212-228) ) and in the 94-shunt (peptide 228 -245 (88 -105) ). However, several areas remote from the active site and the TF-binding region also display decreased HX rates demonstrating more widespread allosteric effects on the FVIIa structure. The N-terminal tail of the protease domain shows significantly decreased HX rates (peptide 153-169 (16 -32) ), and this effect is mirrored in ␤-strands A2 (peptide 275-287 (129F-144)) and B2 (peptide 297-302 (155-160) ), which form part of the cavity for N-terminal insertion referred to as the activation pocket. Helix ␣0 (peptide 264 -275 (124 -129F) ) that extends C-terminally from ␤-strand A2 likewise shows reduced HX rates, and a dramatic reduction in HX rates is seen in the Ca 2ϩbinding loop (peptide 207-220 (67-80) ). Interestingly, a region of the light chain (peptide 133-141), which contains the disulfide interchain linkage between the protease domain and the light chain of FVIIa, displays reduced HX rates in the active form (Fig. 2). Furthermore, the first EGF-like domain in the light chain (peptide 63-80) has decreased HX rates in the active form.
Importantly, most peptides displaying decreased HX rates in the presence of the physiological cofactor TF also display reduced HX rates upon binding of the reversible active site inhibitor, IN18 (supplemental Fig. 1, A-C). The only exceptions are peptides 40 -44 and 361-369 (212-221A) , which selectively show decreased HX rates upon either TF or IN18 binding, respectively. In agreement with structures of FVIIa inhibitor (5,18) or FVIIa⅐TF (4) complexes, these effects are strictly due to the steric solvent exclusion by IN18 or TF at the binding interfaces. In contrast, all of the other peptides displaying decreased HX rates upon binding of both TF and IN18 must be, to some extent, subject to allostery in the FVIIa structure originating from the TF-binding region and/or the active site.
Structural Properties of Zymogen FVII-To gain further insight into the events surrounding FVII activation, we studied the HX kinetics of zymogen FVII in the presence and absence of TF. Because FVII can be autoactivated to FVIIa during purification, we constructed the zymogen variant FVII R152A . Similar to a previously reported R152E variant (19), the R152A mutation in FVII abolishes the proteolytic cleavage site (Arg-152/Ile-153 (16) ), which is required for conversion of zymogen FVII to FVIIa (data not shown). As shown in Fig. 3, the HX behavior of zymogen FVII R152A is highly similar to that of FVIIa in the sense that corresponding peptides of the two proteins show identical HX rates.

DISCUSSION
Recently, the role of even subtle changes in structural dynamics in regulating enzymatic function has become increasingly evident (8,21,22). High-resolution, sensitive techniques are needed to detect the elusive signals that mediate the dynamic interplay of distant sites in a protein structure. Amide HX offers a highly sensitive probe for monitoring these important fluctuations of protein structure in solution during function and regulation (12). In the present work, we studied the HX kinetics of zymogen FVII, FVIIa, TF-bound zymogen FVII, and active FVIIa (TF-or IN18-bound) to characterize the various activation states of FVII/FVIIa. Substantial indirect biochemical evidence has documented an allosteric linkage between the TF-binding region, the active site, and the activation region of FVIIa (23). The present study directly correlated localized changes in dynamics throughout the FVIIa structure with the allosteric activation afforded by TF upon initiation of blood coagulation.
Allosteric Response of FVIIa to TF Is Revealed by HX-Our results demonstrate that when free FVIIa attains the active form (either by TF or IN18 binding), decreased HX rates can be observed primarily in the second ␤-barrel of the protease domain and several interconnecting loops, spanning 45 Å from the TF-binding region to the Ca 2ϩ -binding loop, straight across one face of the protease domain. Apart from two interconnecting loops, the secondary structure of the first ␤-barrel seems to be largely unaffected by these allosteric effects. Remarkably, the allosteric activation signal is not confined to the protease domain as structural stabilization is observed upon either TF or IN18 binding even in remote regions of the light chain of FVIIa located approximately 50 Å from the active site.
Decreased HX rates in activation loop 3 and the 94-shunt demonstrate an allosteric linkage between the TF-binding region and these two loops of the active site. Activation loop 3 encompasses the bottom of the S1 pocket (Gly-375 (226) and Val-376 (227) ) and the Cys-368 (220) -Cys-340 (191) disulfide bond, both of which are an integral part of the S1 pocket of the active site, and stabilization of this loop could facilitate maturation of the active site. Interestingly, the 94-shunt of FIXa contains a Tyr residue at the position corresponding to Thr-239 (99) in FVIIa, which blocks access to the active site but is rearranged upon activation (24). Furthermore, mutation of Thr-239 (99) or the neighboring Val-240 (100) in FVIIa has been shown to result in decreased activity towards the macromolecular substrate FX (17). The decreased HX rates in the 94-shunt of FVIIa observed upon TF binding correlate well with previous findings and strongly suggest that allosteric stabilization of this active site loop in FVIIa is important for activation. Surprisingly, HX rates of activation loop 2 (localized in peptide 332-356 (184 -207) ) are unaltered between FVIIa and TF-bound FVIIa. Activation loops 1-3 are seen to undergo structural rearrangements during activation of trypsinogen to trypsin (25), and it suggests that in FVIIa, activation loop 2 does not play a major role in allosteric activation.
The decreased HX rates observed in the light chain of FVIIa in the active form (following binding of TF or the active site inhibitor IN18) present new intriguing insights into the activation response of FVIIa. The functional consequences of this interchain transfer of an activation signal are at present unknown. Interestingly, the decreased HX rates in the EGF1 domain correlate with studies of an antibody specific for the EGF1 domain, which binds to FVIIa with increased affinity in the presence of a covalent active site inhibitor known to induce the active conformation (26). Furthermore, Ca 2ϩbinding to the EGF1 domain has been shown to increase the amidolytic activity (27). The results presented in this study provide a structural rationale for these previously inexplicable observations. The decreased HX rates observed in the EGF1 domain are likely to be linked to the structural stabilization of the interchain linkage region. The effects of activation observed in the light chain might also contribute to the increased affinity with which TF has been shown to bind to active site-inhibited FVIIa (28).
In general, decreased HX rates upon transition to the active form include both fast and slow kinetic components in peptides of FVIIa but as indicated in Fig. 2, they mainly concern slow kinetic components. These effects are due to the stabilization of existing main-chain secondary or tertiary structure because amides categorized as fast kinetic components are expected to be fully solvent-exposed and unstructured in solution, whereas slower kinetic components engage in some type of secondary structure or intramolecular interactions. However, several peptides (e.g. 153-169 (16 -32) , 207-220 (67-80) , 306 -325 (164 -178) , or 361-377 (212-228) ) also display decreased HX rates for fast kinetic components, and such effects indicate structural rearrangements involving the formation of stable intramolecular interactions in previously unstructured or very flexible solventexposed regions (29).
Zymogen FVII Is Structurally Very Similar to FVIIa but Has a Crippled Allosteric Response-Remarkably, the HX rates of zymogen FVII R152A are virtually identical to those of FVIIa. This strongly indicates that the solution structures of zymogen FVII and FVIIa are highly similar, save for differences in the N-terminal tail. The HX rates of TF-bound zymogen FVII R152A show that TF binding to the zymogen induces only a crippled allosteric response. Evidently, several of the incremental rearrangements and dynamic effects in the FVIIa structure following TF binding are energetically favorable by themselves and do not require a complete global, concerted transition into an active FVIIa form to occur. This observation is supported by findings with a FVIIa variant, which displays enhanced activity without an accompanying increased burial of the N terminus as assessed by carbamylation experiments (30). The HX rates of the Ca 2ϩ -binding loop and the interchain linkage region of the light chain are unaltered in zymogen FVII R152A upon TF binding, and thus insertion of the N-terminal tail into the activation pocket seems to be necessary for induction of an allosteric response in these two regions. These results demonstrate a tight linkage between the dynamics of the N-terminal tail and the Ca 2ϩ -binding loop and identify the N-terminal tail as a structural relay switch in FVIIa, capable of transmitting the allosteric signal to the Ca 2ϩ -binding loop and to the interchain linkage region of the light chain. Interestingly, HX rates in peptide 297-302 (155-160) comprising ␤-strand B2 are similarly unaltered in TF-bound zymogen FVII R152A as opposed to TFbound FVIIa. As seen in the FVIIa/TF crystal structure, Val-299 (157) of ␤-strand B2 makes a main chain hydrogen bond to Lys-157 (20) of the N-terminal tail, suggesting that the presence of this interaction is required for allosteric stabilization of the Ca 2ϩ -binding loop and the interchain linkage region and that this interaction cannot form in TF-bound zymogen FVII. The Ca 2ϩ -binding loop is a conserved feature of trypsin and several related proteases. Results presented here open the possibility that in addition to the known Ca 2ϩ stimulation of FVIIa activ-ity, stabilization of the Ca 2ϩ -binding loop via the N-terminal tail might be a common allosteric feature during the activation of several trypsin-like proteases.
Allosteric Linkage-A crystal structure of zymogen FVII in complex with an exosite inhibitor, A183, has been solved (31). The most striking feature of this structure is a three-residue reregistration of ␤-strand B2. The hypothesis that this reregistration plays a fundamental role in the activation of FVII has been challenged recently by the demonstration of plausible crystal packing effects on the structure of the FVII⅐A183 complex (32) and by mutagenesis studies (33). To investigate this further, we identified a peptide (residues 302-308 (160 -166) ) in zymogen FVII R152A , comprising the residues of ␤-strand B2 involved in the alleged reregistration. The conformation of this region in the FVII-A183 crystal structure rotates residues Arg-304 (162) and Leu-305 (163) such that their amide hydrogens are placed in a fully solvent-exposed environment and no longer participate in hydrogen bonds with the neighboring ␤-strand A2 as in previous structures of FVIIa (4,5). This would cause the amide hydrogens of Arg-304 (162) and Leu-305 (163) to exchange several orders of magnitude faster in zymogen FVII. However, peptide 302-308 (160 -166) of zymogen FVII R152A shows HX rates identical to those seen for free and TF-bound FVIIa. In fact, the HX rates of peptide 302-308 (160 -166) of both FVIIa and zymogen FVII R152A are slow and characteristic of amides participating in strong hydrogen bonding. In conjunction with the work of Perera and Pedersen (32), the results presented here strongly indicate that the solution structure of this region is conserved in zymogen FVII, free and TF-bound FVIIa.
If the allosteric activation of FVIIa by TF is not propagated via reregistration of ␤-strand B2, what is the pathway? In thrombin, residues of activation loop 3 are critical for mediating allosteric activation (34). The proximity of allosterically stabilized segments in helix ␣1, the 170-loop, and activation loop 3 of FVIIa imply that one or more hydrogen bonds between these regions are absent in free FVIIa but present in TF-bound FVIIa. This is supported by the decreased number of fast kinetic components in peptide 361-377 (212-228) of activation loop 3 upon TF binding. Importantly, however, the number of fast kinetic components of peptide 361-377 (212)(213)(214)(215)(216)(217)(218)(219)(220)(221)(222)(223)(224)(225)(226)(227)(228) in zymogen FVII R152A are unaltered upon TF binding (in contrast to FVIIa). Evidently, a crippled stabilization of activation loop 3 seems to correlate to the absence of N-terminal insertion and active site maturation in TF-bound zymogen FVII. Taken in conjunction, these observations suggest that activation loop 3 constitutes an essential part of the structural element through which a TF-induced allosteric change is conveyed to the active site and that this loop could function as a conformational switch necessary for TF to turn on the activity of FVIIa.
In summary, many of the coagulation proteases function optimally only when bound to cofactors, and considerable biochemical evidence supports the concept that several of these cofactors induce allosteric changes in their cognate enzymes (1,2). Through the use of HX monitored by MS, we detected extensive allostery and dynamic interplay between functional sites in FVIIa upon TF binding, revealing the potential for allosteric regulation in the trypsin family of proteases. In the framework of these findings, the zymogen form displays a crippled allosteric response to TF, despite having a highly similar solution structure to FVIIa. The results imply the strength of the experimental approach to investigate structural dynamics in solution of the numerous proteins and protein systems subject to structural transitions during function.