Noncovalent interactions of the Apple 4 domain that mediate coagulation factor XI homodimerization.

The Apple 4 (A4) domain of human plasma factor XI (FXI) was used to investigate the process of FXI noncovalent dimer formation. Recombinant 6-histidine-tagged A4 domain proteins were prepared utilizing a bacterial expression system. Purification was accomplished under denaturing conditions, followed by a refolding protocol to facilitate correct disulfide bond formation. Analysis of the A4 domain (C321S mutant) by size exclusion chromatography indicated the presence of a slowly equilibrating reversible monomer-dimer equilibrium. The elution profiles reveal highly symmetrical peaks for both dimeric and monomeric species with elution times that were highly reproducible for varying amounts of both the dimeric and monomeric species. The monomer-dimer equilibrium was found to be dependent upon changes in both pH and salt concentration. Under conditions approximating physiologic salt concentration and pH (20 mm HEPES, 100 mm NaCl, and 1 mm EDTA, pH 7.4), it was determined that the monomer-dimer equilibrium was characterized by a dissociation constant (K(D)) value of 229 +/- 26 nm with a calculated Delta G value of 9.1 kcal/mol. This report identifies electrostatic contributions and the presence of a hydrophobic component that mediate interactions at the A4 domain interface. The rate of dissociation for the recombinant A4 domain C321S mutant was examined by monitoring the increase in 4,4'-dianilino-1,1'-binaphthyl-5,5'-disulfonic acid dipotassium salt fluorescence under dissociating conditions, giving a value for a dissociation rate constant (k(off)) of 4.3 x 10(-3) s(-1).

Human factor XI (FXI) 1 is a plasma glycoprotein (ϳ5% carbohydrate) involved in the intrinsic pathway of blood coagulation (1). FXI exists as a covalently linked two-chain homodimer, thus making it unique among coagulation factors. Dimeric FXI circulates in plasma as a zymogen in a 1:2 stoi-chiometric complex with the cofactor high molecular weight kininogen at plasma concentrations in the range of 4 -6 g/ml (ϳ30 nM) (2,3). Each subunit of the dimer contains 607 amino acids and has a molecular mass of ϳ80,000 Da based on reducing SDS gel electrophoresis (4). The activation of this zymogen to an active serine protease can be carried out in the presence of a polyanionic surface by activated factor XII and high molecular weight kininogen, thrombin, or activated FXI (FXIa) (4,5). Activation is dependent upon cleavage of the peptide bond between Arg 369 and Ile 370 (6). Each subunit of FXIa is composed of a heavy (50,000 Da) and light (30,000 Da) chain held together via disulfide bonds (4). The light chain functions as a serine protease involved in the activation of coagulation FIX (7). The heavy chain of FXIa contains four tandem repeat sequences called Apple domains consisting of 90 -91 amino acids each (8). Each of the four Apple domains contains three internal disulfide bonds, and the first and fourth Apple domains each contain one additional cysteine residue (9). Cys 321 in the A4 domain has been implicated in the formation of a covalent FXI dimer (9). The first Apple domain (A1) contains distinct binding sites for cofactor high molecular weight kininogen, for thrombin, and for the Kringle 2 domain of prothrombin (10 -12), whereas the second (A2) contains a FIX-binding site (13); the third (A3) has been implicated in interactions with platelets, with heparin, and with FIX (14 -17); and the fourth (A4) contains an activated factor XII-binding site along with its ability to mediate dimer formation (9,18,19).
One of the more provocative questions surrounding FXI is its homodimeric structure and how that relates to its physiologic function in blood coagulation. FXI is activated on the platelet surface, and the resulting FXIa subsequently activates FIX (7). The recombinant A3 domain of FXI binds to the activated platelet surface with the same stoichiometry (ϳ1500 sites/ platelet) as that found for full-length dimeric FXI, suggesting that one A3 domain/dimer is responsible for the interaction with the platelet surface (16). Since it is difficult to rationalize one A3 (17) and/or A2 (13) domain of FXI binding to both FIX and to the platelet surface simultaneously, it has been hypothesized that the A3 domain from one subunit of the dimer binds the platelet, while the A2 and/or A3 domain of the opposite subunit interacts with FIX. Gailani et al. (20) recently demonstrated that recombinantly derived FXI monomers in which the A4 domain of FXI was replaced with that of prekallikrein (FXI/PKA4) in comparison with their dimeric counterparts (FXI/PKA4-Ala 326 and wild-type FXI) were functionally defective in a partial thromboplastin time assay utilizing platelets as a surface, but normal with phospholipids. These results are consistent with the aforementioned hypothesis since the monomeric chimera bound to the platelet surface lacks an additional free A2 and/or A3 domain for interaction with FIX and would therefore be deficient in its ability to activate FIX. Additionally, FXI dimerization may be important for its ability to activate FIX since cleavage occurs at two spatially distinct sites. A detailed kinetic analysis of the activation of FIX by FXIa revealed that FIX is activated via a processive reaction mechanism without release of a singly cleaved intermediate (21). These findings suggest the possibility that each active site within the dimer carries out a single cleavage without the need for reorientation of the FIX molecule between successive cleavages, further suggesting that it may be functionally advantageous for FXI to exist as a dimeric molecule.
To fully understand the functional significance of FXI homodimer formation, we aim to define the structural determinants and mechanism involved in this process. Meijers et al. (19) demonstrated that at least part of the molecular information required for mediating dimer formation between the two identical subunits appears to reside within the A4 domain of FXI. Chimeric tissue plasminogen activator molecules with the A4 domain of FXI substituted for the finger and growth factor domains exist as dimers, as shown by gel filtration analysis (19). In the same chimeras as well as the full-length FXI protein, the replacement of Cys 321 with serine results in molecules that also exist as dimers, as determined by gel filtration (19). These results strongly suggest that the A4 domains mediate noncovalent interactions resulting in dimer formation, which is subsequently stabilized by a covalent linkage between cysteine residues in each monomer at position 321 (9,19). We have therefore prepared the rA4 domain of FXI as well as a C321S mutant to obviate covalent dimer formation and to permit determinations of binding constants under conditions of varying pH values and ionic strength. We employed a rapid gel filtration procedure similar to that described by Manning et al. (22) for the determination of natural and recombinant hemoglobin dissociation constants (K D ).

EXPERIMENTAL PROCEDURES
Materials-Pfu polymerase, DNA markers ( DNA-HindIII digest, pBR322 DNA-MspI digest, and X174 DNA-HaeIII digest), restriction enzymes BamHI and PstI, and NEB buffer 2 were from New England Biolabs Inc. (Beverly, MA). All reagents used for SDS-PAGE were purchased from National Diagnostics, Inc. (Atlanta, GA). Ammonium persulfate and ␤-mercaptoethanol were purchased from Bio-Rad. Bacto-Tryptone, Bacto-yeast extract, LB broth base (Lennox L broth base), T4 DNA ligase, dNTPs, and prestained protein molecular mass markers (low range) were from Life Technologies, Inc. DIAFLO ® ultrafiltration membranes (3000-Da cutoff) and the Amicon 8200 stirred cell ultrafiltration apparatus were from Amicon, Inc. (Beverly, MA). Isopropyl-␤-D-thiogalactopyranoside (IPTG) was purchased from LabScientific, Inc. (Princeton, NJ). HEPES, CAPS, MES, Tris, cysteine, urea, guanidine, dithionitrobenzoate, and the MW-GF-70 gel filtration molecular mass marker kit were from Sigma. M15(pREP 4 ) cells, the QIAexpress ® pQE-9 vector, nickel-nitrilotriacetic acid (Ni 2ϩ -NTA) resin, the QIAEX II gel extraction kit, and the QIAquick polymerase chain reaction (PCR) purification kit were from QIAGEN Inc. (Chatsworth, CA Primers that flank either end of the A4 domain coding sequence were prepared by Life Technologies, Inc. The 5Ј-upstream primer contains an engineered BamHI restriction site (GGATCC) along with a FXa cut site (ATCGAAGGTAGA) that, when translated (Ile-Glu-Gly-Arg), was used to efficiently cleave an N-terminal 6-histidine tag. The 3Ј-downstream primer contains an engineered PstI restriction site (CTGCAG) along with a stop codon (TAA) to ensure transcription termination. The sequence of the upstream primer (5Ј-BamHI) is as follows: 5Ј-CGCG-GATCC(ATCGAAGGTAGA)TTCTGCCATTCTTCATTTTAC. The sequence of the downstream primer (3Ј-PstI) is as follows: 5Ј-AAAACTGCAG(TTA)CTCATTATCCATTTTACACAA. The PCR was set up as follows: 10 ng of FXI cDNA as template, a 1 M concentration of both the upstream (5Ј-BamHI) and downstream (3Ј-PstI) primers, 2.5 units of cloned Pfu DNA polymerase, 10 l of 10ϫ Pfu reaction buffer, and dNTPs at 0.2 mM brought to a final volume of 100 l. The reaction mixture was then overlaid with 100 l of mineral oil to prevent evaporation. The reaction mixtures were placed in a PerkinElmer Life Sciences DNA thermal cycler. The reactions were heated to 94°C for 7 min, followed by 40 cycles at 94°C for 1 min, 50°C for 1 min, and 75°C for 3 min. The reaction was completed by incubation at 75°C for 7 min and a final incubation at 4°C. The PCR was subsequently purified with the QIAquick PCR purification kit.
PCR was also performed in the generation of the C321S mutant (rA4-C321S). The PCR-based mutagenesis protocol was adapted from Picard et al. (23). Template plasmid DNA was obtained using a QIA-GEN plasmid purification kit (ϳ0.3-kilobase pair human FXI A4 domain subcloned into a pQE-9 expression vector). The mutagenesis reaction follows a three-step protocol.
Step 1 involves synthesis of a megaprimer using a mutagenic primer and a downstream primer (3Ј-PstI). The mutagenic primer is as follows: 5Ј-CCAAGCATCC(A)G-CAACGAAGG. The 11th base in the primer was switched from thymine in the wild-type sequence to adenine, thus allowing Cys 321 to be replaced by serine in the mutant amino acid sequence. The reactions contained 10 ng of DNA template, 10 pmol (100 nM) each of the mutagenic and 3Ј-PstI primers, 2.5 units of cloned Pfu DNA polymerase, 10 l of 10ϫ Pfu reaction buffer, and dNTPs at 0.2 mM brought to a final volume of 95 l. 10 cycles of amplification were performed for each step (94°C for 1 min, 48°C for 1 min, and 72°C for 2 min). The reaction was completed by incubation at 75°C for 5 min with a final incubation at 4°C. Addition of 50 pmol of the upstream primer (5Ј-BamHI) to the aqueous phase begins step 2, and addition of 50 pmol of the downstream primer (3Ј-PstI) initiates step 3. The same amplification protocol as employed in step 1 is used in steps 2 and 3. The insert was cut with BamHI and PstI restriction enzymes using the protocols supplied by New England Biolabs Inc. The insert was ligated to the QIAexpress pQE-9 vector, and transformations were carried out using competent E. coli K12-derived M15(pREP 4 ) cells. Expression from the pQE-9 vector was induced by the addition of IPTG.
Expression, Purification, and Folding of the rA4 Domain-Largescale expression cultures (1 liter) were propagated in LB broth in the presence of 100 g/ml ampicillin and 50 g/ml kanamycin at 37°C. The large-scale cultures were inoculated from small-scale growth cultures (1:40, v/v). The cultures were grown until A 600 reached 0.6 -0.7. Induction of protein expression required the addition of 0.5 mM IPTG, followed by further incubation and shaking at 37°C for 3 h. Cells were harvested by centrifugation (4000 ϫ g for 20 min) and then stored as cell pellets at Ϫ70°C. The concentration of protein was determined by absorbance at 280 nm employing an extinction coefficient of 12,420 M Ϫ1 cm Ϫ1 (per dimer). SDS-PAGE analysis was used to determine purity and to assure noncovalent association of subunits.
Harvested cell pellets stored at Ϫ70°C were resuspended in Buffer A (6 M guanidine hydrochloride, 0.1 M NaH 2 PO 4 , and 0.01 M Tris-HCl, pH 8.0) at 5 ml of buffer/g of cells and allowed to stir at room temperature for 1 h. The supernatant was collected from the cell lysate following centrifugation at 10,000 ϫ g for 15 min at 4°C. The crude extract was then mixed with a Ni 2ϩ -NTA resin slurry (50% Ni 2ϩ -NTA and 50% Buffer A, 10-ml total volume). The mixture was then allowed to stir at room temperature for 1 h. The crude extract/resin mixture was loaded onto a 10-ml polypropylene column (QIAGEN Inc.). A series of wash steps then ensued: 5 column volumes of Buffer B (8 M urea, 0.1 M NaH 2 PO 4 , and 0.01 M Tris-HCl, pH 8.0), a wash with Buffer C (8 M urea, 0.1 M NaH 2 PO 4 , and 0.01 M Tris-HCl, pH 6.3) until the eluant absorbance at 280 nm was Ͻ0.01, and 20 ml of Buffer D (8 M urea, 0.1 M NaH 2 PO 4 , and 0.01 M Tris-HCl, pH 5.9). The FXI rA4 domain was then eluted with 20 ml of buffer E (8 M urea, 0.1 M NaH 2 PO 4 , and 0.01 M Tris-HCl, pH 4.5). Fractions containing protein were visualized by 15% SDS-PAGE.
The protein was then refolded using a protocol that makes use of a thiol/disulfide exchange in which cysteine is utilized as the low molecular mass reducing molecule (24). Protein concentration was held in the range of 200 -250 g/ml. A Spectrapor dialysis membrane (9.3 ml/cm) with a 3500-Da cutoff was used. The starting buffer consisted of 2 M urea, 20 mM Tris-HCl, 100 mM NaCl, and 2 mM cysteine at pH 9.0. A series of seven buffer changes was used to reduce the urea and cysteine concentrations (1 M urea and 1 mM cysteine; 0.5 M urea and 0.5 mM cysteine; 0.25 M urea and 250 M cysteine; 0.1 M urea and 100 M cysteine; and 0 urea and 0 cysteine). Upon completion of dialysis, the FXI rA4 domain was then concentrated via an Amicon stirred cell ultrafiltration apparatus using DIAFLO ultrafiltration membranes (3000-kDa cutoff). The protein solution was typically concentrated down to a volume of 2-4 ml, and the protein concentration (5.0 to 2.5 mg/ml) was determined. SDS-PAGE under reducing conditions was used throughout to monitor purification and under nonreducing conditions to detect correctly oxidized FXI rA4 domain.
The N-terminal 6-histidine tag was removed by cleavage with bovine FXa. Bovine FXa cleaves proteins at the C-terminal side of the recognition sequence Ile-Glu-Gly-Arg, which is identical to that engineered at the N terminus of the FXI rA4 domain. Cleavage buffer consisted of 20 mM Tris-HCl and 100 mM NaCl, pH 9.0. A 1:50 molar ratio of FXa to the FXI rA4 domain was used. The reaction mixture was incubated at 37°C for 16 -18 h. Upon completion of the reaction, success of cleavage was determined by SDS-PAGE analysis. Following the cleavage reaction, separation of the His-tagged FXI rA4 domain from the non-Histagged FXI rA4 domain was carried out using the HiTrap metal-chelating column (5 ml). The HiTrap resin was charged with Ni 2ϩ ions by applying 2.5 ml of a 0.1 M NiSO 4 solution. The cleavage reaction was applied to the column, followed by a 5-column volume (25 ml) wash step (20 mM Tris-HCl and 100 mM NaCl at pH 9.0). The desired non-Histagged FXI rA4 domain along with FXa was then eluted with 20 mM imidazole in the above buffer system. The final step in the purification made use of gel filtration chromatography to separate the FXI rA4 domain from bovine FXa and to obtain protein that formed productive dimers. A Superdex 75 gel filtration column was equilibrated in 20 mM HEPES and 100 mM NaCl, pH 7.4. The proteins were resolved at a flow rate of 1 ml/min, and protein elution was monitored by absorbance at 280 nm. Fractions that eluted at a retention time representative of the dimeric FXI rA4 domain were pooled and concentrated with a Millipore concentrator. To determine the multimeric state of the eluted protein, data compiled for a standard curve were generated using the MW-GF-70 gel filtration molecular mass marker kit. To assess purity, all fractions were assayed for the presence of FXa using the chromogenic substrate S-2765. The protein was subsequently examined by SDS-PAGE.
The rA4 domain was examined by HPLC and matrix-assisted laser desorption ionization time-of-flight mass spectroscopy (analysis conducted by the Protein Chemistry Laboratory at the University of Pennsylvania, Philadelphia). The data demonstrated the presence of a single homogeneous species of protein with a molecular mass of 19,836 Da, which is consistent with the calculated molecular mass of the rA4 domain, which is 19,888 Da. Ellman's reagent (5,5Ј-dithiobis(2-nitrobenzoic acid)) was used to determine free sulfhydryl groups (25), which indicated Ͻ0.04 mol of free sulfhydryl group/mol of rA4 domain.
Size Exclusion Chromatography-An Amersham Pharmacia Biotech fast protein liquid chromatography (FPLC) system was used along with an Amersham Pharmacia Biotech FPLC high resolution Superose 12 column (10 ϫ 300 mm) to determine monomer-dimer K D for the FXI rA4 domain C321S mutant. A Superdex 75 column was also employed in purification and confirmation of dimer formation. To determine the multimeric state of the eluted proteins, data for a standard curve were compiled at varying pH values and salt concentrations using the MW-GF-70 gel filtration molecular mass marker kit. To construct standard curves, four proteins were employed with molecular masses as follows: albumin, 66,000 Da; carbonic anhydrase, 29,000 Da; cytochrome c, 12,400 Da; and aprotinin, 6500 Da. The effects of varying protein concentration, pH, and salt concentration were determined. All buffer solutions used to study pH effects contained 20 mM buffering reagent, 100 mM NaCl, and 1 mM EDTA. Buffering reagents for the various pH values tested were as follows: CAPS, pH 10.0; Tris-HCl, pH 9.0 and 8.0; HEPES, pH 7.0 and 7.4; and MES, pH 6.0. All buffer solutions used to study salt effects contained 20 mM HEPES, the appropriate NaCl concentration (0.025-2.0 M), and 1 mM EDTA, pH 7.4. For all experimental studies, the proteins were dialyzed against the appropriate buffer for 16 -18 h in Slide-A-Lyzer dialysis cassettes. The resultant protein solution was cleared of any precipitate by centrifugation, and protein concentration was determined spectrophotometrically by absorbance at 280 nm with the previously stated extinction coefficient. Samples were then prepared by dilution of concentrated stock solutions and incubated for Ն3 h at room temperature to ensure that equilibrium was established. Samples of 100 l at various protein concentrations were injected onto the pre-equilibrated column. Elution of resolved protein species was in the same buffer as the sample equilibration at a flow rate of 1 ml/min. The elution profile was followed by absorbance at 280 nm. The chromatograms were analyzed using FPLC Director software (Amersham Pharmacia Biotech). The retention times were determined by comparison of elution peaks at peak half-width and peak half-height. The calculated area under the elution peaks was used to determine the amount and percentage of dimeric and monomeric species present in the sample. A plot of the percent dimeric rA4-C321S as a function of the total protein concentration was used to determine dimer K D . All experimentally determined K D values (K D(app) ) were divided by a dilution factor of 6.5 Ϯ 0.27. The dilution factor was employed since dilution of the protein sample occurs as the protein is eluted through the column. The justification for this correction can be found under "Results" and in Refs. 22 and 36. The dilution factor equals the peak width at half-height (milliliter) divided by the sample load volume (100 l) (22,36).
Conversion of K D to the Change in Binding Energy (⌬G 0 )-The Gibbs free energy of dissociation was calculated using the equation ⌬G 0 ϭ ϪRT ln K D , where R is the gas constant (1.987 cal ϫ mol Ϫ1 ϫ K Ϫ1 ), T is the absolute temperature at which experiments were done (298 K), and K D is the dissociation constant or the concentration at which equal amounts of dimer and monomer exist in solution.
Fluorescence Spectroscopy-Fluorescence measurements were performed on an SLM-AMINCO/Bowman Series 2 luminescence spectrometer. The stock solutions of bis-ANS were filtered, and the concentration was determined by absorbance at 385 nm using an extinction coefficient of ⑀ 385 ϭ 16,790 cm Ϫ1 M Ϫ1 (30). Varying amounts of the apolar probe were used to titrate the rA4 domain while monitoring extrinsic probe fluorescence. Protein concentrations were held constant at 0.5 M. The samples were excited at 385 nm, and emission was monitored at 497 nm. The excitation and emission slit widths were set to 4 and 8 nm, respectively. All results were corrected with respect to spectra of buffer and protein background.
The establishment of dissociation kinetics was followed in a time-dependent manner by monitoring the increase in bis-ANS fluorescence as well. For dissociation kinetics, the protein was diluted to a concentration of 0.5 M in the presence of 20 M bis-ANS. The data were evaluated using the program Kaleidagraph TM (Abelbeck Software) and fitted using a first-order equation with a single rate constant. The association rate constant (k on ) for the conversion of monomer to dimer was calculated using the following equation, K D ϭ k off /k on , where K D represents the dissociation constant determined by size exclusion chromatography and k off is the dissociation rate constant determined by fluorescence spectroscopy.

Expression and Purification of the rA4 Domain of FXI-Since
the FXI rA4 domain is expressed as insoluble inclusion bodies, its recovery requires solubilization, folding, and oxidation. To facilitate the purification of the rA4 domain, we generated an N-terminal 6-histidine-tagged fusion protein expressed in an E. coli K12-derived strain (M15(pREP 4 )). DNA corresponding to the A4 domain of FXI was PCR-amplified and placed in the pQE-9 expression vector. The vector contains a regulatable promoter/operator element allowing expression to be controlled by the addition of IPTG. Fig. 1A shows the expression of the rA4 domain compared with uninduced cells and cells induced with IPTG.
Solubilization and purification took place in the presence of the denaturants 6 M guanidine hydrochloride and 8 M urea. We utilized the N-terminal 6-His tag as a molecular handle to facilitate binding of the recombinant protein by a high affinity interaction for the Ni 2ϩ -NTA resin (K D ϭ 10 Ϫ13 M at pH 8.0). Wash steps with subsequent stepwise decreases in pH allowed for the removal of contaminating proteins (Buffers B-D). The protein of interest was eluted at pH 4.5 (Buffer E), a condition favoring the protonation of the 6-His tag and thereby diminishing the interaction made with the Ni 2ϩ -NTA resin. The eluant was collected in 1-ml fractions and analyzed by both absorbance at 280 nm and SDS-PAGE. The FXI rA4 domain eluted in the first 10 ml of buffer E. This first step proved to be quite advantageous, allowing us to purify the fusion protein from the crude extract in one step with considerable purity. Purity was assessed by resolving the fractions by 15% SDS-PAGE and staining with Coomassie Blue (Fig. 1). The typical yield after the Ni 2ϩ affinity chromatography step from a 500-ml culture was ϳ10 mg of denatured fusion protein.
The protein was then refolded using a protocol that makes use of a thiol/disulfide exchange in which cysteine is utilized as the low molecular mass reducing molecule (24). By stepwise removal of the urea, the protein slowly renatures, allowing the correct cysteine residues to come in close proximity of one another. Furthermore, the presence of cysteine molecules allows for a rapid disulfide exchange or reshuffling of incorrect disulfide bonds, which ultimately allows for formation of the most energetically favorable disulfide linkages. Even though the protein concentration during the folding step was held relatively low (200 -250 g/ml), a considerable amount of protein precipitate was observed at this step due to nonspecific aggregate formation. Approximately 30 -50% of the protein was lost, making it the most inefficient step in the purification scheme. Protein folding, especially dependent upon disulfide bond formation, is highly sensitive to fluctuations in pH, temperature, and protein concentration (25,26), which may account for the variability between preparations and the substantial loss of protein at this step. The protein that remained soluble was separated from the precipitate by centrifugation and ultrafiltration. Success of folding was examined by SDS-PAGE analysis under reducing and nonreducing conditions (Fig. 2) and gel filtration. Despite the considerable loss of protein at this step, protein that remained soluble was capable of forming productive dimers exclusively, as evidenced by distinct bands resolved by SDS-PAGE and single peaks on gel filtration chromatograms. Exclusive dimer formation at this step provides some evidence that protein folding was successful since improperly folded rA4 domains would conceivably produce a wide range of multimers. Mixed disulfide-bonded species tend to resolve as diffuse bands and multimeric ladders upon nonreducing SDS-PAGE as well as broad elution peaks upon gel filtration chromatography (25,26). Total recovery of protein post-folding based on yields achieved from the initial Ni 2ϩ -NTA chromatography step were in the neighborhood of 50 -70% or 5-7 mg.
The removal of the 6-His tag plus four amino-terminal residues (Met-Arg-Gly-Ser) was accomplished by cleavage with bovine FXa. Success of cleavage includes the removal of the 6-His tag plus the four amino acids (Ile-Glu-Gly-Arg) composing the FXa cut site. The molecular mass of the protein was reduced by 1854 Da to give a calculated molecular mass of 9943 Da/monomer. The cleavage reaction was monitored by SDS-PAGE analysis.
To separate the cleaved FXI rA4 domain (without the 6-His tag) from the noncleaved FXI rA4 domain, the digested sample was applied to a HiTrap metal-chelating column charged with Ni 2ϩ ions. The digested sample contained primarily three species: bovine FXa, the 6-His-tagged A4 domain, and the non-6-His-tagged A4 domain. Following an initial wash step (20 mM Tris-HCl and 100 mM NaCl, pH 9.0), all three species remained on the column, which was somewhat unexpected since FXa and the nontagged A4 domain are devoid of histidine tags. However, both FXa and the A4 domain are capable of binding to a Mono Q TM anion exchange column (data not shown) at pH 9.0, suggesting that these proteins are somewhat negatively charged at this pH. This may account for a nonspecific charge interaction with the positive Ni 2ϩ ions bound to the column. Elution of the nontagged A4 domain was accomplished by addition of 20 mM imidazole to the buffer. Once again, 1-ml fractions were collected and assessed by absorbance at 280 nm and SDS-PAGE. Bovine FXa and the nontagged A4 domain coeluted in fractions 10 -20 (10 -20 ml). As a final step in the purification, the remaining protein was resolved on a Superdex 75 gel filtration column. The gel filtration step accomplished two goals: removal of bovine FXa (ϳ45 kDa) and purification of only the A4 domain protein (ϳ20 kDa) that is capable of forming productive dimers. Fractions that eluted at volumes corresponding to the dimer form of the FXI rA4 domain (ϳ76 ml) were collected. Elution volumes were compared with a standard curve of log molecular mass versus elution volume generated by chromatography of the gel filtration molecular mass marker kit (data not shown). The collected fractions were also assayed with the chromogenic substrate S2765 to confirm separation of FXa from the A4 domain. Final purity was determined by SDS-PAGE analysis. It was conservatively estimated from the gel that the A4 domain was purified to Ͼ95% homogeneity (Fig. 1B). Based on the protein yield achieved after the Ni 2ϩ -NTA chromatography step, a final yield of ϳ1-2 mg (10 - 20% recovery) of pure FXI rA4 domain remained per 500 ml of expression culture.
Reducing and Nonreducing SDS-PAGE-The SDS-PAGE results provided additional confirmation for both covalent (rA4) and noncovalent (rA4-C321S) dimer formation (Fig. 2). Size comparisons were made against low molecular mass markers (Life Technologies, Inc.). Samples of both rA4 and rA4-C321S (Ϯ␤-mercaptoethanol) were resolved by SDS-PAGE. Both reduced samples resolved corresponding to monomer molecular masses migrating between markers of 14 and 6 kDa (expected size of 9944 Da). The nonreduced rA4 domain resolved as a dimer migrating between molecular mass markers of 29 and 18 (expected size of 19,888 Da). The nonreduced rA4-C321S mutant resolved as a monomer identical to that of the reduced samples (Fig. 2). The results indicate that the wild-type rA4 domain exists as a dimer and that dimer formation is mediated by a disulfide linkage, as would be expected. The rA4-C321S mutant dimer does not contain a cysteine residue at position 321, but was still capable of forming dimers under native conditions, as seen from the gel filtration results (Fig. 3). Under nonreducing conditions, SDS was capable of disrupting the C321S dimer interaction, confirming that the mutant dimer interaction is mediated via noncovalent intersubunit amino acid interactions.
Dimerization Assay-A high resolution Superose 12 column (10 ϫ 300 mm) was used to develop a dimerization assay to determine monomer-dimer K D for mutant rA4 domains of FXI. Similar assays have been described by both Manning et al. (22) and Gallagher and Huber (27) to measure tetramer-dimer dissociation constants of natural and recombinant hemoglobins and the monomer-dimer equilibrium of M15 ␤-galactosidase from E. coli, respectively. To determine the multimeric state of the eluted protein, data for a standard curve were compiled under varying conditions using the MW-GF-70 gel filtration molecular mass marker kit (see "Experimental Procedures"). Peak area, determined by FPLC Director software, was used to calculate the percent dimeric and monomeric species for each sample. A plot of percent dimeric rA4-C321S as a function of the total rA4-C321S concentration prior to application to the column gave a hyperbolic curve. A representative dissociation curve for the C321S mutant along with elution profiles (chromatograms) are presented in Fig. 3. The elution volumes of the two peaks labeled dimer (14.4 ml) and monomer (16.7 ml) in Fig. 3 correspond to molecular masses of 21,060 and 8291 Da, respectively. These masses correspond well to the theoretical calculated monomer (9944 Da) and dimer (19,888 Da) forms of the protein. Under conditions approximating physiologic salt concentration and pH (20 mM HEPES, 100 mM NaCl, and 1 mM EDTA, pH 7.4), it was determined that the monomer-dimer equilibrium was characterized by a K D(app) of 1.5 Ϯ 0.2 M. This value of K D(app) was corrected for a 6.5-fold dilution factor during the gel filtration procedure (22,36) to give a K D value of 229 nM. The justification for this correction is given in the last paragraph under "Results" and in Refs. 22 and 36.
The results indicate that the monomer-dimer equilibrium of rA4-C321S is reversible, as evidenced by a shift in the ratio of dimer to monomer as a function of decreasing protein concentration. The reversibility of the equilibrium was also demonstrated by subjecting the dimer peak to a second round of chromatography and observing the presence of two peaks corresponding to dimer and monomer (data not shown). The equilibrium can be driven to completion, generating either nearly all dimeric or monomeric species, indicating that all the protein present in the reaction is functional and participates in the equilibrium process. The elution times were highly reproducible for both varying amounts of the dimeric and monomeric species, as evidenced by peak width at half-height determinations. All elution profiles resolved as highly symmetric peaks, indicative of a mixture of monomeric and dimeric rA4-C321S domains in slow equilibrium between species (28). The lack of peak broadening also rules out potentially nonspecific interactions with the column material.
These results indicate that the A4 domain by itself is sufficient to cause dimerization and that the interchain disulfide bond at C321S is not necessary for dimer formation. These findings are consistent with the results of Meijers et al. (19) for the Cys 321 mutant of full-length FXI, which migrated as a dimeric molecule upon gel filtration. Conclusively, the analysis indicates that the C321S mutant exists in a reversible monomer-dimer equilibrium that can be measured and used to determine the K D(app) for this interaction under varying conditions.
A time course for dissociation using the same size exclusion chromatography method was examined (Fig. 4). A stock solution of the rA4-C321S domain (ϳ25 M) was diluted to 2.0, 0.75, and 0.35 M in buffer containing 20 mM HEPES, 100 mM NaCl, and 1 mM EDTA, pH 7.4. Aliquots were then resolved on a Superose 12 column at various times post-dilution. The data were subjected to linear regression analysis. We observed that the equilibrium was essentially stable over the time scale tested, demonstrating that equilibrium had already been achieved prior to our measurements. The final monomer concentration for all three trials (2.0, 0.75, and 0.35 M) after equilibrium was established and was internally consistent with our experimentally determined K D(app) of 1.5 M. A solution of rA4-C321S at concentrations of 2.0, 0.75, and 0.35 M should be ϳ40, 65, and 80% monomeric, respectively. This is exactly what we observed. This observation leads us to conclude that all the protein participates in the equilibrium process. However, this method of analysis does not allow us to effectively study the dissociation kinetics of the dimer interaction.
Dimer Dissociation (rA4-C321S) as a Function of pH-To analyze the dimer interaction, size exclusion chromatography at various concentrations of rA4-C321S was used to determine The protein was diluted to the appropriate concentration in 20 mM HEPES, 100 mM NaCl, and 1 mM EDTA at pH 7.4 and incubated at room temperature (ϳ25°C) for Ͼ3 h to establish equilibrium. The elution profiles were followed by the absorbance at 280 nm, and the amounts of protein present as monomer and dimer were assessed by peak area determined by FPLC Director software. With a flow rate of 1 ml/min, the retention times are equivalent to elution volume in milliliters. The peak labeled Dimer refers to the elution of dimer, and that labeled Monomer refers to that of monomer.
K D as a function of pH in a range between 6.0 and 9.0. A composite of the dissociation curves is represented in Fig. 5. The results are presented in Table I. The tightest interaction (K D ϳ 229 nM) for the dimer was seen in the acidic pH range tested (6.0 -7.4). From pH 6.0 (K D ϭ 283 Ϯ 32 nM, ⌬G ϭ 8.9 kcal/mol) to pH 7.4 (K D ϭ 229 Ϯ 26 nM, ⌬G ϭ 9.1 kcal/mol), little or no change (⌬K D ϭ 54 nM, ⌬⌬G ϭ 0.2 kcal/mol) was observed for the binding interaction. In contrast, a marked change was observed when comparing pH 7.4 with pH 8.0 (K D ϭ 229 Ϯ 26 nM and ⌬G ϭ 9.1 kcal/mol versus K D ϭ 614 Ϯ 146 nM and ⌬G ϭ 8.5 kcal/mol). An ϳ2.7-fold increase in K D was observed, followed by a further 1.5-fold increase in K D at pH 9.0 (K D ϭ 955 Ϯ 122 nM, ⌬G ϭ 8.2 kcal/mol). Conclusively, an increase in pH causes a general destabilization of the rA4-C321S dimer interaction, reflected in an overall ϳ4.0-fold increase in K D (⌬⌬G ϭ 0.8 kcal/mol) from pH 6.0 to 9.0. The destabilization trend observed as a function of pH provides evidence for a charge interaction at the dimer interface that aids in mediating dimer formation. A marked shift in binding affinity between pH 7.4 and 8.0 suggests the involvement of amino acid residue(s) with a pK a of ϳ8.0.
Chromatography was also attempted for pH values below 6.0 and above 10.0 that are outside the range shown in Table II. The resultant chromatograms revealed evidence of peak broadening and inconsistencies in peak elution times. Therefore, it was not possible to interpret the results. However, these results suggest that the A4 domain (rA4-C321S) may be somewhat conformationally unstable in pH conditions below 6.0 and above 10.0. These results suggest three possibilities: global structural changes in the protein and/or changes in the rate of dissociation or interaction with the column material.
Dimer Dissociation (rA4-C321S) as a Function of Salt Concentration-To analyze the dimer interaction, size exclusion chromatography at various concentrations of rA4-C321S was used to determine dissociation constants as a function of salt concentration between 0.025 and 2.0 M. A composite of the dissociation curves is represented in Fig. 6. The results are presented in the Table II Bis-ANS Titration-We conducted studies to assess the role of hydrophobic interactions in A4 dimer formation. The apolar probe bis-ANS was used to titrate the A4 domain while monitoring extrinsic probe fluorescence. Upon binding to a hydro- represents the concentration of dimer that would be present if the protein were all dimer. Percent dimer is the percentage of protein that is actually dimer at various C321S concentrations.

TABLE I Effect of pH on dimer dissociation
Shown is the effect of pH on dimer dissociation represented in terms of K D (M) and Gibbs free energy (⌬G, kcal/mol). The K D values were determined from plots of percent dimer versus [C321S] utilizing Kaleidagraph software and were corrected for 6.5-fold dilution of samples on the column as described under "Experimental Procedures." The Gibbs free energy of dissociation was calculated using the following equation: ⌬G ϭ ϪRT ln K D , where R is the gas constant (1.987 cal ϫ mol Ϫ1 ϫ K Ϫ1 ), T is the absolute temperature (298 K), and K D is the dimer dissociation constant experimentally determined.

Interactions of the FXI A4 Domain Mediate Homodimerization
phobic moiety or hydrophobic cluster on the surface of a protein, one observes a severalfold increase in bis-ANS fluorescence intensity (29). In comparison with bis-ANS emission maximum (553 nm) when the dye is unbound and in aqueous medium, one also observes a blue-shifted emission maximum upon binding of the dye to a hydrophobic moiety (30). The extent of both the fluorescence intensity increase and the shift in emission maxima is strictly dependent upon the environment of the dye-binding site (31). An increase in the bis-ANS fluorescence and a shift in the emission maximum from 533 to 497 nm were observed upon addition of saturating levels of bis-ANS (10 M) to rA4-C321S (0.5 M) (data not shown).
When higher concentrations of dye were added, we observed no further shift in the emission maximum, which is suggestive of a single dye-binding site (32). As a control, the fluorescence intensity of 10 M bis-ANS in buffer (in the absence of protein) was negligible compared with the fluorescence of dye in the presence of protein (data not shown). We compared the wildtype rA4 domain (0.5 M) and the rA4-C321S domain (0.5 M) in the presence of increasing amounts of bis-ANS and noted the change in relative bis-ANS fluorescence (Fig. 7). Binding of the probe to the mutant rA4-C321S domain showed saturable binding, along with a marked increase in probe fluorescence, as evidenced by a plot that was hyperbolic in character (Fig. 7). The hyperbolic nature of the curve is indicative of a single dye-binding site with a K D equal to ϳ2 M. In contrast, the covalently linked, fully dimeric wild-type rA4 domain showed little or no significant ability to bind the apolar probe, as evidenced by the absence of a detectable increase in relative bis-ANS fluorescence. These results suggest the absence of hydrophobic sites on the surface of the covalently linked rA4 domain. To test the specificity of the dye-binding site on the rA4-C321S domain, a denatured rA4-C321S domain incubated in 6 M guanidine hydrochloride was titrated with bis-ANS. The results demonstrated a small amount of nonspecific binding of the dye, as evidenced by the linear plot shown in Fig. 7. The dye-binding site exposed on the surface of the partially dissociated native rA4-C321S domain can be disrupted by exposure of the protein to denaturant. These findings suggest that the hydrophobic site exposed only in the presence of the dissociated dimer must contain some conformational specificity.
In conclusion, these results give clear indication that a saturable hydrophobic site(s) is exposed only upon dimer dissociation and that this site(s) can be disrupted by denaturation of the protein. A hydrophobic patch present only on the monomeric subunit gives credence to the argument that a hydrophobic component exists at the homodimer interface, which may be in part responsible for mediating A4 dimer formation.
Kinetics of rA4-C321S Dissociation-The change in the fluorescence of bis-ANS upon binding to the dissociated rA4-C321S domain was useful for monitoring the kinetics of dissociation (Fig. 8). Concentrated stock solutions (25-100 M) of both the rA4-C321S and rA4 domains were diluted to a final concentration of 0.5 M in the presence of 10 M bis-ANS and monitored for increases in bis-ANS fluorescence. We observed an increase in bis-ANS fluorescence for rA4-C321S, which follows a single exponential reaction with a dissociation rate constant (k off ) of 4.3 ϫ 10 3 s Ϫ1 for the dissociation of dimer to monomer (Fig. 8). Utilizing the previously determined K D of 229 Ϯ 26 nM (size exclusion chromatography), we were able to TABLE II Effect of salt on dimer dissociation Shown is the effect of salt on dimer dissociation represented in terms of K D (M) and Gibbs free energy (⌬G, kcal/mol). The K D values were determined from plots of percent dimer versus [C321S] utilizing Kaleidagraph software and were corrected for 6.5-fold dilution of samples on the column as described under "Experimental Procedures." The Gibbs free energy of dissociation was calculated using the following equation: ⌬G ϭ ϪRT ln K D , where R is the gas constant (1.987 cal ϫ mol Ϫ1 ϫ K Ϫ1 ), T is the absolute temperature (298 K), and K D is the dimer dissociation constant experimentally determined. calculate an association rate constant (k on ) for the conversion of monomer to dimer of 1.9 ϫ 10 4 M Ϫ1 s Ϫ1 (see "Experimental Procedures"). The determination of the kinetic parameters associated with the rA4 domain monomer-dimer equilibrium provides a detailed look at the kinetic mechanism behind the dimer interaction of the rA4 domain as well as the noncovalent interaction of FXI as a whole. It is of critical importance to point out that this assay is based on the assumption that bis-ANS binds to the dissociated monomer at a much faster rate compared with the rate of dissociation of the dimer. We can make this statement since the rate of encounter of a small molecule with a large molecule for a diffusion-controlled reaction gives a second-order rate constant in the range of 10 9 -10 11 M Ϫ1 s Ϫ1 (33). If we hypothesize that the true rate of association for the apolar probe were altered by 3 orders magnitude (10 6 M Ϫ1 s Ϫ1 ), accounting for changes in protein conformation upon association of the probe with the protein, the rate of binding of the probe (10 4 s Ϫ1 at 10 M bis-ANS) to the protein would still be much faster compared with our observed rate for dimer dissociation of 4.3 ϫ 10 Ϫ3 s Ϫ1 . Therefore, in this experiment, we are observing the dissociation of the rA4-C321S dimer and not the binding of bis-ANS to the monomer. Moreover, the covalently linked wild-type protein (rA4) does not bind the apolar probe, demonstrating the specificity of the probe for the dimer interface. These observations are in accordance with the bis-ANS titration experiments, in which we observed no increase in bis-ANS fluorescence in the presence of the wild-type rA4 domain.
The findings of both the fluorescence assay (kinetics of dissociation) and the gel filtration assay are mutually consistent. Based on the experimentally determined rate of dissociation (k off ϭ 4.3 ϫ 10 Ϫ3 s Ϫ1 ), we calculated a half-life (t1 ⁄2 for the dissociation reaction of 160 s. In the time scale of the gel filtration experiment in which the majority of the determinations were made 10 min (3.7 half-lives) after dilution, we would expect at least 92% of the equilibrium to have been achieved (Fig. 4). This provides an explanation as to why in this experiment we observed no change in monomer concentration over time and also provides a justification for correcting values of K D(app) to account for the 6.5-fold dilution of protein samples during the gel filtration dimerization assay (see "Experimental Procedures").

DISCUSSION
The results presented in this paper demonstrate the feasibility of studying the noncovalent dimer interaction of FXI by use of a recombinantly generated A4 domain. Herein we report the K D (ϳ230 nM) and the Gibbs free energy of dissociation (⌬G ϳ 9 kcal/mol) for the rA4-C321S noncovalent mutant under physiologic conditions, as determined by size exclusion chromatography. We also report a k off of 4.3 ϫ 10 Ϫ3 s Ϫ1 and a calculated k on of 1.9 ϫ 10 4 M Ϫ1 s Ϫ1 using fluorescence spectroscopy.
The FXI rA4 domain and the rA4-C321S mutant have been purified and characterized, including their capacity to form both covalent and noncovalent dimers, DNA sequencing of the recombinant DNA, mass spectroscopic analysis of the protein product, and N-terminal sequence analysis. Considering the enormous number of disulfide bond combinations (135,135) that are theoretically possible in a protein like the rA4 domain with a total of seven cysteines (25,26), we took great pains in characterizing the rA4 domains. The single most important line of evidence that these proteins are conformationally intact is the fact that they maintain their capacity to form dimers. Moreover, the rA4 domain exists as a single species upon reverse-phase HPLC, mass spectroscopy, and gel filtration, supporting the conclusion that the protein is homogeneous and conformationally intact (34). We also tested our proteins for their susceptibility to protease cleavage. Both trypsin and endoproteinase Lys-C were unable to cleave the rA4 domains, as monitored by SDS-PAGE (data not shown). While elucidating the disulfide bond arrangement in FXI, McMullen et al. (9) similarly observed resistance to trypsin proteolysis for an isolated fragment containing the entire A4 domain (His 267 -Met 358 ). In the native conformation, a majority of proteins possess some protease resistance, and proteolysis typically takes place in regions that are typically exposed or disordered (35). Therefore, one might expect that if the rA4 domain were folded incorrectly, cleavage by the previously mentioned proteases would occur more readily. We conclude that all of the rA4 domain polypeptides exist in native conformation.
Molecular sieve chromatography or size exclusion chromatography has been well described in the literature and has become a valuable tool for determining equilibrium constants for reversibly associating systems (28,36,37). Values obtained by this technique have been shown to be reliably comparable to sedimentation equilibrium and fluorescence depolarization assays, among other techniques designed to scrutinize multimer equilibria (22,27). The results presented in Figs. 3-6 provide evidence for a process in which a mixture of both dimeric and monomeric rA4 domains is unaffected by the chromatography process and exists in a slowly exchanging equilibrium (28). In a slowly exchanging system like ours, the separation process (column chromatography) is faster than the reversible exchange of dimer and monomer. This explains why we observed two well resolved, highly symmetrical peaks in our study (28,37).
We examined the influence of pH on the dissociation equilibrium using the freely dissociable mutant (rA4-C321S) and observed an increase in K D at the more alkaline pH values tested (Table I). Although the pH effect could be due to deprotonation of an ionizable group (or groups) on the protein resulting in a structural distortion of the protein, we observed a high degree of peak uniformity and consistency of retention times, suggesting that the proteins are behaving uniformly both in respect to their Stokes radii and lack of interaction with the column matrix (28). We can therefore theorize that the pH effects are attributed to a deprotonation of a critical amino acid side chain involved in the dimer interaction. The most signifi- The increase in bis-ANS fluorescence was measured at a fixed wavelength of 497 nm (excitation at 385 nm). The curve was fit according to a first-order reaction. cant changes in K D occur in the pH range between 7.4 to 8.0 (ϳ2.7-fold increase in K D ), suggesting deprotonation of a side chain with a pK a value of ϳ7.4 -8.0. There are only two possible candidates: cysteine (pK a ϭ 8.5) and the N-terminal amino group (pK a ϭ 8.0) (38,39). Since all cysteine residues within the A4 domain in both the full-length FXI molecule (9) and our own protein constructs (see "Results") are engaged in disulfide linkages, there are no free sulfhydryl groups to mediate dimer formation. The N-terminal amino group of the A4 domain (Phe 271 ) is engaged within FXI in a peptide bond with its neighboring residue (Val 270 ). Moreover, both the FXI rA4 and rA4-C321S domains are capable of forming dimers with and without the histidine tag present (data not shown). Furthermore, if the N terminus were located at the dimer interface, bovine Xa probably would not have steric access to the cleavage site, and this would present a problem for removal of the 6-His tag. Therefore, it is highly unlikely that the N terminus of the rA4 subunit plays a role in the dimer interaction. It is difficult to predict accurately the ionization behavior of any one group without additional information. However, we can conclude from our analysis that an increase in pH affects the ability of the A4 domain to associate, suggesting the possibility that deprotonation of an amino acid side chain with a perturbed pK a due to its microenvironment directly weakens the monomer interactions. The obvious implication from this analysis is that an electrostatic component mediates dimer formation; whether we are observing a specific side chain contribution or global charge repulsion remains to be seen. In any event, it is this type of information that will provide the basic knowledge about the oligomerization of proteins within a cell and the rationale for future experimentation designed to answer questions about the nature of the FXI dimer interface.
Electrostatic complementarity has been proposed to play a significant role in the formation of a large number of protein complexes (43). For example, homodimers of MyoD are subsequently destabilized by increasing salt, which, in turn, confirmed an electrostatic contribution to subunit interactions first suggested by crystallographic studies (44,45). The study of cytochrome P-450 and P-450 reductase has revealed an interaction that is also held together by complementary charged residues (46 -48). Electrostatic interactions play an important role in the dimerization of FXI mediated by the A4 domain. The shift in equilibrium toward monomer at the high concentrations of salt (Table II) is thought to be a result of charge shielding by the addition of ions and a reduction in the productive charge interactions at the interface (49,50). When small diffusible ions such as Na ϩ and Cl Ϫ are present in an aqueous environment, the ions tend to concentrate in the vicinity of charges that are opposite in sign (positive versus negative) (51). Therefore, the increase in salt concentration would gradually neutralize the charge interactions by subsequently weakening the resultant protein-protein interaction at the A4 dimer interface. Under conditions of high salt, water molecules tend to be excluded at protein interfaces (51,52), intensifying an existing hydrophobic interaction resulting in lower K D values. Since we did not observe a large change in K D or ⌬G due to increasing salt concentrations, we can speculate that the increase in salt may intensify an existing hydrophobic interaction at the dimer interface compensating for the destabilizing effect seen due to shielding of charges at the interface.
The specificity of the bis-ANS probe for hydrophobic moieties on the surface of proteins (53) has led to many studies, including the characterization of protein interface contacts between subunits found in coagulation cofactor VIII (32). The probe permitted the identification of a surface-exposed hydrophobic site on dissociated rA4-C321S as well as the absence of hydro-phobic sites on the covalently linked rA4 domain. The presence of a site only on the dissociated subunits is suggestive of its involvement in homodimer formation. These results provide clear evidence of a hydrophobic component present at the A4 dimer interface and demonstrate that the bis-ANS probe binds only to a single affinity dye-binding site. These findings are also in agreement with the salt-dependent changes seen in binding affinity, which suggest that the hydrophobic component can compensate for the charge destabilization caused by increasing salt concentrations.
Our study has clear implications for the assembly mechanism of FXI as well as for the study of oligomeric proteins in general. Our results are consistent with published studies of homodimer interfaces that indicate that the overall proportions of nonpolar interactions (van der Waals and hydrophobic) and polar interactions (hydrogen bonds) tend to vary greatly from one oligomer interface to another (54). The majority of the interfaces surveyed contained a mixture of small hydrophobic patches, polar interactions, and water molecules scattered over the entire interface (55).
This study provides an initial characterization of the interaction between the two monomeric subunits of FXI and provides seminal information that will guide future investigations of platelet-associated coagulation reactions and function of FXI (56). Since the dimeric nature of FXI may be critical for its ability to function properly on the platelet surface (16,20,21), it is important to understand the mechanism of FXI dimerization. A future goal is to compare full-length FXI with the rA4 proteins to determine whether all of the binding information resides within the A4 domain. It will be interesting to compare the rA4 domain with the full-length FXI C321S mutant to determine whether another region of the FXI molecule could harbor additional binding energy. Such studies would be facilitated by the development of dimerization assays such as fluorescence polarization (44,57) and isothermal titration calorimetry (58), solution-based methods that provide a true equilibrium measurement that does not require separation of free and bound species. Such studies, together with structural information from NMR and x-ray crystallography combined with site-directed mutational analysis, will allow us to examine individual amino acid contributions made at the dimer interface.