Loop electrostatics asymmetry modulates the pre-existing conformational equilibrium in thrombin

Abstract

Thrombin exists as an ensemble of active (E) and inactive (E*) conformations that differ for their accessibility to the active site. Here we show that redistribution of the E*-E equilibrium can be achieved by perturbing the electrostatic properties of the enzyme. Removal of the negative charge of the catalytic Asp102 or Asp189 in the primary specificity site destabilizes the E form and causes a shift in the 215–217 segment that compromises substrate entrance. Solution studies and existing structures of D102N document stabilization of the E* form. A new high resolution structure of D189A also reveals the mutant in the collapsed E* form. These findings establish a new paradigm for the control of the E*-E equilibrium in the trypsin fold.

Graphical Abstract

Four protease families account for more than 40% of all proteolytic enzymes in humans and are responsible for digestion, blood coagulation, fibrinolysis, development, fertilization, apoptosis, and immunity(1,2). Trypsin-like proteases constitute the largest group and utilize a canonical catalytic triad for activity, composed of the highly conserved residues His57, Asp102, and Ser195. Catalysis is assisted by the oxyanion hole, defined by the backbone N atoms of Gly193 and Ser195, the 215–217 segment shaping the wall of the primary specificity pocket, and residue Asp189 at the bottom of this pocket that engages the Arg residue at the P1 position of the substrate(3).

Thrombin is a trypsin-like protease that plays opposing functional roles in blood coagulation due to the interaction with numerous macromolecular substrates, receptors and inhibitors(4). Recent studies have demonstrated that thrombin exists in equilibrium between active (E) and inactive (E*) species, or ensembles(5–8). Conformational transitions for this enzyme unfold over different time scales: a slow time scale (45 ms) for the interconversion of the E* and E ensembles detected by stopped-flow measurements(9), and a fast time scale for the interconversion of conformers within each ensemble inferred by NMR and MD simulations(10–13). Thrombin dynamics involve structural fluctuations of flexible loops surrounding the active site, the Na⁺ binding loops and the 215–217 segment within the active site. Movement of the 215–217 segment is a structural signature of the trypsin fold that controls access to the active site in the protease and zymogen. Binding of Na⁺ shifts the pre-existing E*-E equilibrium in favor of the active E form and also stabilizes more active conformers within the E ensemble resulting in higher catalytic activity(9). Interestingly, the allosteric effect of Na⁺ on thrombin is selectively abrogated by the S195T replacement(14), establishing a linkage between residues within the active site and conformational dynamics of the enzyme. In this context, the role of the catalytic Asp 102 merits attention. One of the carboxylate oxygen atoms of Asp102 accepts hydrogen bonds from the main-chain amide groups of residues 56 and 57, and the second oxygen accepts hydrogen bonds from both the Nδ1 atom of His57 and the Oγ atom of Ser214. The negatively charged Asp102 maintains the unprotonated Nε2 with a lone pair of electrons as the general base catalyst for transfer of the proton from Oγ of Ser195 to the leaving group. Mutation of Asp to Asn does not change the number of hydrogen bonds but dramatically reverts the directionality of this network. Hence, Ser195 becomes a weak nucleophile and the catalytic activity of the mutant D102N is 4 orders of magnitude compromised(15,16). Similarly, replacement of the negatively charged Asp189 with Ala and Asn in the primary specificity site compromises the specificity of thrombin towards synthetic and physiological substrates up to 4 orders of magnitudes, and unexpectedly abrogate monovalent cation binding(17). Since Asp102 and Asp189 are negatively charged under physiological pH, modifications of their electrostatic potential could also propagate long-range within the active site and reach the flexible loops on the surface of the enzyme. Here we show that perturbation of the electrostatics within the active site and primary specificity site of thrombin has a drastic influence on the E*-E equilibrium and provide a mechanistic framework for the interpretation of key functional and structural features of the enzyme.

Material and Methods

Materials

α-thrombin purified from human plasma was obtained from Haematologic Technologies (Essex Junction, VT) and buffer exchanged using a FF G-25 HiTrap equilibrated with 5 mM Tris-HCl buffer, pH 8.0, 0.2M ChCl. Thrombin WT and mutants D102N and D189A were expressed in BHK cells and purified to homogeneity as described previously(15,17,18). The peptides Hir(1–47) and Hir(48–64) were obtained by solid-phase synthesis using standard Fluorenylmethyloxycarbonyl chloride (Fmoc)-chemistry on a model PS3 automated synthesizer from Protein Technologies (Tucson, AZ) and purified by RP-HPLC(19). Chemical identity was established by enzymatic fingerprint analysis and high-resolution mass spectrometry on a Mariner ESI-TOF instrument from Perseptive Biosystems (Stafford, TX). Nα-Fmoc protected amino acids, solvents and reagents for peptide synthesis were purchased from Applied Biosystems (Foster City, CA) or Bachem AG (Bubendorf, Switzerland). p-Aminobenzamidine (PABA), trypsin, salts, urea and organic solvents were of analytical grade (Sigma, MO).

Electrostatic Calculations

The E open conformation of thrombin was obtained from the crystallographic structure of α-thrombin (PDB entry: 1PPB) after removal of the inhibitor (PPACK), water molecules, and the coordinates of 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES)(20). After substituting Asp102 to Asn, PQR files were prepared and calculations were performed using APBS(21) and BLUUES program(22), which is based on the generalized Born (GB) models(23). Calculations were run on the nonglycosylated X-ray structure of thrombin wild-type and mutant D102N, which was obtained by mutating Asp102 with Asn while keeping all other coordinates unchanged. Calculations were performed using a solvent dielectric of 78.14 and a protein dielectric of 2.0 at 298 K in 145 mM NaCl. Final electrostatic maps were constructed by subtracting the protein self-energies from the calculated map using the dxmath utility in APBS.

Molecular Dynamics Simulations

Glycosilated thrombin was solvated with 21116 explicit water molecules in a 90-Å side length cubic box. Na⁺ and Cl⁻ ions were added, with [NaCl]=0.145 M. The glycosylation was added at Asn60g according to the carbohydrate composition previously determined by Nilsson et al. (24) (figure S1). The model was completed by creating the connectivity between the eight cysteines to generate the four-disulfide bonds: Cys1–Cys122, Cys42–Cys58, Cys168–Cys182, and Cys191–Cys220. CHARMM-27 all-atom force field with CMAP correction(25) was used for the protein and ions, the TIP3P force field was used for water molecules, whereas the carbohydrate Chain Solution Force Field (CSFF) was used for the oligosaccharide tail(26). The pKa of all the charged residues was calculated using the pKaTool software(27) to determine the protonation state of the amino acids in the protein at pH 7. NAMD 2.9 software package(28) was used to run the simulations. Briefly, after an energy minimization stage, thrombin was gradually heated by linearly changing the temperature from 1 to 310 K at a rate of 1 K/ps and then equilibrated at 310 K for 15 ns. All the simulations were run in the canonical (NPT) ensemble. For the wild-type protein, a 90-ns production run was performed starting from the equilibrated system. For the D102N mutant, the same parameters were used for the last 50 ns of the simulations, which was preceded by the alchemical morphing (40 ns) of residue 102 from the negative Asp to the neutral Asn. This numerical experiment was done to dissect the effect of charge and steric variations upon Asp to Asn mutation. The alchemical morphing was conducted in five steps. In the first four steps the overall charge of D102 was incrementally changed from −1 to 0 (in units of the electron charge in absolute value) according to the linear interpolation $q_i(\xi )=(1-\xi )q_i^{(\text{Asp})}+\xi q_i^{(\text{Asn})}$. The parameter ξ assumed the values 0, 0.25, 0.5, 0.75, and 1 (figure S2). For ξ = 0, the Asp residue charges are assigned to all the Asp102 atoms. For ξ = 1, an artificially mutated Asp102 residue is created in which the atoms have the charges of Asn (the O2 oxygen of the carboxylate group of Asp102 collects the global 0 charge of the –NH₂ moiety in Asn102). After each charge change, we let the system evolve for a total of 10 ns. After completion of the charge morphing, the O2 oxygen of the carboxylate group of Asp102 was replaced into the –NH₂ moiety of Asn102; then the MD simulation of the fully mutated protein was forwarded for further 50 ns. Thus, the total runtime was identical for both WT and D102N, i.e. 90 ns, and the same simulation parameters were applied: 2 fs of integration time step; all bonds with hydrogen atoms were kept fixed; periodic boundary conditions were used; short and long range interactions were evaluated at each MD step; non-bonded cutoff of 12 Å (with function switching at 10 Å) was used, together with a pair list distance of 13.5 Å. The Particle Mesh Ewald (PME) model was used for electrostatic calculations (grid size of 90 Å in all directions of space, PME tolerance of 10⁻⁶). To simulate the canonical NPT conditions, the system was coupled to a stochastic thermostat with set point temperature of 310 K (with dumping constant of 1 ps⁻¹) and to a stochastic barostat with set point pressure of 1 atm (piston period of 100 fs, piston decay time of 50 fs). The Cα-Cα variance-covariance matrix, M, was used to map the conformational flexibility profile resulting from Asp102 to Asn mutation. The matrix elements were calculated as $M_{i,j}^{a,b}=\langle x_{a,j}-\langle x_{a,i}\rangle \rangle \langle x_{b,j}-\langle x_{b,i}\rangle \rangle$ where a, b = 1, 2, 3 represent the X, Y or Z coordinates of Cα atoms i and j and averaged over all snapshots of the last 20 ns MD trajectory as representative of equilibrium fluctuations. A transformation of reference frame was done before calculating the matrix. In particular, we applied a roto-translation to align the frame diagonalizing the inertia tensor (calculated at each snapshot) to the origin and orientation of the frame in the first snapshot. In this way, the dynamics of the system was represented in a non-inertial frame, fixed on the protein, from which only the internal dynamics was observed. The results are shown as the transformed matrix, m, with elements m_i,j = ln(|M_i,j|), using a cut-off of 10⁻⁵ for the matrix elements. This transformation is necessary to highlight the different scales of variance-covariance elements. Elements of m range from −11.5, corresponding to |M_i,j| ≤ 10⁻⁵, to ln(max_i,j{|M_i,j|}). The diagonal elements of matrix m quantify the positional constraint of single residues: high values of m_i,i mean that the Cα of residue i-th has high mobility. The off-diagonal terms m_i,j = m_j,i quantify the spatial correlation between residues i and j: large values of m_i,j mean that the two residues fluctuate in a concerted manner so that they keep a high degree of spatial linkage. To monitor the structural changes occurring to selected amino acids during the alchemical mutation, we applied the following analysis protocol: before each step, i.e. incremental charge neutralization, and after the first 10 ns of simulation of the fully mutated protein, we calculated the averaged distances among each amino acid on the ring formed by the negatively charged amino acids placed on the borders of the reaction site: Glu39, Asp60e, Glu61, Asp63, Glu97a, Glu146, Glu192, Glu217 and Asp221. Averages have been taken over the 200 ps (100 MD snapshots) preceding the next alchemical step: $d_{i,j}={\sum }_{t=1}^{100}{d}_{i,j}(t)/100$ where i and j indexes run over the selected amino acids in both D102N and WT trajectories. Then we calculated the differences between the distances for i−j amino acid pairs in D102N and WT, $\mathrm{\Delta }{d}_{i,j}=d_{i,j}^{D102N}-d_{i,j}^{WT}$. For each pair, Δd_i,j was graphically represented as described in the Results.

Spectroscopic measurements

Measurements were carried out at 25 ± 0.1 °C in 5 mM Tris-HCl buffer, pH 8.0, containing 0.1% (w/v) PEG-8000 and 0.2M choline chloride (ChCl). Temperature correction was applied for Tris buffer(29). All measurements were carried out at least in duplicate and the spectra were subtracted for the corresponding baselines. Protein concentration was determined by UV absorption at 280 nm on a Lambda-2 spectrophotometer from Perkin-Elmer (Norwalk, CT) or Nanodrop using a molar absorptivity value of 65770 M⁻¹·cm⁻¹ for natural α-thrombin and mutant D102N and 2920 M⁻¹·cm⁻¹ for Hir(1–47). The concentration of Hir(48–64), having only one aromatic chromophore in the sequence (i.e., Phe-54), was taken as 200 M⁻¹·cm⁻¹ at 257 nm. The active-site concentration of thrombin was also determined by titration with hirudin HM2 in the presence of FPR as a chromogenic substrate, using a procedure similar to that reported elsewhere(30) and found identical (within 5% error) to that determined by UV absorption. The concentration of PABA was determined using a molar absorptivity value of 15000 M⁻¹cm⁻¹ at 293 nm(31). Far-UV CD spectra were recorded on a Jasco (Tokyo, Japan) model J-810 spectropolarimeter equipped with a water-jacketed cell holder, connected to a model RTE-111 (NesLab) water-circulating bath. The spectra were recorded in a 1-mm cell, at a scan-speed of 10 nm/min, with a response time of 16 sec, and resulted from the average of four accumulations. CD data were expressed as the mean of the residue ellipticity [θ] = θ_obs·MRW/(10·l·c), where θ_obs is the observed signal in degrees, MRW is the mean residue weight taken as 114.6 Da, l is the cuvette pathlength in cm, and c is the protein concentration in g/ml. Fluorescence spectra were recorded at a scan speed of 200 nm/min in a 1-cm pathlength cuvette (2 ml internal volume) on a Jasco model FP-6500 spectropolarimeter, equipped with Peltier model ETC-273T temperature control system from Jasco. Protein samples (5 – 50 nM) were excited at 280, using excitation and emission slits of 5 and 10 nm, respectively.

Stability measurements

Urea-induced denaturation experiments were carried out in 20 mM Tris-HCl buffer, 0.1% PEG₈₀₀₀ pH 8.0 at 25 ± 0.1 °C, in the presence of 0.2 M chloride salts. After 1-hour incubation at the specified urea concentration, protein samples (2 ml, 50 nM) were excited at 280 nm and the fluorescence intensity was recorded at 334 nm. At each urea concentration, the fluorescence signal was subtracted for that of the corresponding base line. Data were analyzed within the framework of a two-state process (32) and the value of [urea]_1/2 was estimated as described elsewhere (33).

Ligand binding to thrombin

The equilibrium association constants of Hir(1–47) and Hir(48–64) for thrombin wild-type and mutant D102N were determined at 25 ± 0.1 °C in 5 mM Tris-HCl buffer, pH 8.0, containing 0.1% (w/v) PEG-8000 and the specified concentration of chloride ion, by exciting the protein samples at 280 nm and measuring the increase of fluorescence intensity at 334 nm as a function of ligand concentration. Under all conditions tested, the optical density of the solution was always much lower than 0.02 units both at λ_ex and λ_em and therefore no inner filter effect was observed(34). Aliquots (2–5 µl) of ligand stock solutions were added, under gentle magnetic stirring, to a thrombin solution (5–50 nM, 2 ml) in a 1-cm pathlength cuvette. After stirring (30 s), the solution was allowed to equilibrate for 2 min and the signal intensity was recorded at 334 nm. Fluorescence data were corrected for sample dilution, which was always lower than 2% at the end of the titration. Photobleaching was reduced to less than 2%, even after prolonged light exposure, by using a 1-cm pathlength quartz cuvette (2 ml internal volume) with two opaque walls that are able to diffuse the incident light inside the sample, thus preventing photodegradation of Trp-residues. Fluorescence data were fitted to a simple one-site binding equation. For Na⁺ binding to thrombin, a solution containing 50 nM enzyme in 5 mM Tris-HCl, pH 8.0, 0.1% PEG 8000 and 400 mM NaCl was incrementally added to a solution containing 50 nM thrombin in 5 mM Tris-HCl, pH 8.0, 0.1% PEG 8000 and 400 mM ChCl. The ionic strength and enzyme concentration were held constant, while the Na⁺ concentration was increased(35). The interaction of p-Aminobenzamidine (PABA) with thrombin was monitored by adding, under gentle magnetic stirring in a 1 cm pathlength cuvette, aliquots (2–10 µl) of PABA stock solution (12.5–50 mM) to a solution of thrombin (1.4 ml, 376 nM). At each PABA concentration, thrombin sample was equilibrated for 2 min at 25±0.1°C and excited at 325 nm, using an excitation/emission slit of 5 and 10 nm, respectively. The increase in fluorescence intensity of PABA in the presence of the enzyme was recorded at the 376 nm as a function of PABA concentration. Fluorescence data were corrected for IFE, both at λ_ex and λ_em(34,36).

Crystallization

Crystallization for the human thrombin mutant D189A was achieved at 22° C by the vapor diffusion technique, with 10 mg/ml protein 3µl mixed with 3µl of the solution containing 200 mM KCl and 20% PEG 3350. Crystals were grown in 1–2 weeks and were cryoprotected in a solution of 200 mM KCl, 25% PEG 3350 and 20% glycerol prior to flash freezing. X-ray diffraction data were collected at beamline 19ID at the Advanced Photon Source, Argonne, IL and were indexed, integrated and scaled with the HKL2000 software package(37). Structure were solved by molecular replacement using MOLREP from the CCP4 suite(38) and the structure of thrombin bound to PPACK (PDB ID code 1SHH) as starting model. Refinement and electron density generation were performed with REFMAC5 from CCP4 package(38). 5% of the reflections were randomly selected as a test set for cross-validation. Model building and analysis of the structure were carried out using COOT(39). In the final refinement stage, TLS tensors modeling rigid-body anisotropic temperature factors were calculated and applied to the model(40). Ramachandran plots were calculated using PROCHECK(41). Statistics for data collection and refinement are summarized in Table 2. Atomic coordinates and structure factors have been deposited in Protein Data Bank (accession code: 5JDU).

Table 2.

Crystallographic data for human thrombin mutant D189A

Buffer/salt	200 mM KCl
PEG	3350 (20%)
Data collection:
Wavelength (Å)	0.90
Space group	P212121
Unit cell dimensions (Å)	a=55.4, b=81.3, c=146.4
Molecules/asymmetric unit	2
Resolution range (Å)	40-1.7
Observations	626477
Unique observations	74112
Completeness (%)	99.9 (99.8)
Rsym (%)	10.5 (39.4)
I/s(I)	18.2 (3.2)
Refinement:
Resolution (Å)	40-1.7
|F|/s(|F|)	No cutoff
Rcryst, Rfree	0.181, 0.206
Reflections (working/test)	70070/3729
Protein atoms	4473
NAG/Cl−	4/1
Solvent molecules	622
Rmsd bond lengthsa (Å)	0.011
Rmsd anglesa (°)	1.4
Rmsd DB (Å2) (mm/ms/ss)b	1.85/1.44/2.73
<B> protein (Å2)	26.3
<B> NAG/Cl− (Å2)	44.5/24.0
<B> Solvent (Å2)	37.9
Ramachandran plot:
Most favored (%)	99.6
Generously allowed (%)	0.0
Disallowed (%)	0.4

^aRoot-mean-squared deviation (Rmsd) from ideal bond lengths and angles and Rmsd in B-factors of bonded atoms.

^bmm, main chain-main chain; ms, main chain-side chain; ss, side chain-side chain.

Results

Electrostatics of thrombin wild-type and D102N

Thrombin displays a non-uniform electrostatic potential, generated by a highly asymmetric distribution of positive and negative amino acids (figure 1A)(20). Two positively charged patches, exosite I and exosite II, oppose the active site pocket, which is negatively charged because of Asp102 and Asp189. Interestingly, this arrangement is complemented by nine Asp- and Glu-residues (i.e. Glu39, Asp60e, Glu61, Asp63, Glu97a, Glu146, Glu192, Glu217 and Asp221a) that form a negatively charged ring around the active site (figure 1B). Early work on rat trypsin showed that the overall structure of the mutant D102N crystallized in the presence of benzamidine was superimposable with the wild-type, both sharing a wide open active site primed for ligand binding(16). However, the very same mutation in thrombin unexpectedly resulted in an occluded form of the active site due to collapse of the 215–217 segment(15). Long-range electrostatic forces may be responsible for this conformational rearrangement. We therefore tested the hypothesis that, in thrombin, the loss of a negative charge caused by isosteric substitution of Asp to Asn may alter the electrostatic potential of regions surrounding the active site causing the 215–217 to reposition.

Figure 1. Negative ring surrounding the active site and electrostatic potential for thrombin wild-type (A, B) and D102N (C).

(A) Strong electropositive potential (blue) dominates the exosite I and II and opposes the negative potential of the Na⁺ binding site (Na⁺) and the active site. (B) View from the top of the active site of thrombin (1PPB). The conserved catalytic triad is composed of H57, D102, and S195 (green sticks), which is located at the bottom of the active site. Movement of the 215−217 segment (orange sticks) ensures productive interaction of the incoming substrate with the residue D189 (green stick) that forms the primary specificity pocket. Nine Asp and Glu-residues (magenta sticks) surround the active site and generate a negative electrostatic gradient that steers positively charged substrates into the active site pocket. (C) Substitution of Asp102 to Asn weakens the electronegative potential of the active site pocket. (D) Difference of the electrostatic potential Δζ_(D102N-WT) calculated for each residue reported as kJ/(mol·q). A positive value of Δζ indicates that a given amino acid in D102N has a less negative or a more positive ζ value compared to WT. Arrows point to protein segments that are affected by the mutation.

After mutating the negatively charged Asp102 to Asn in the E open form of thrombin, we calculated the electrostatic potential for the two proteins at the atomic level. Thrombin WT (figure 1A) and D102N (figure 1C) have similar electrostatic properties for exosite I, exosite II and the Na⁺ binding site but substitution of Asp to Asn causes a sharp reduction of the electronegative potential around the active site. In particular, low (Δζ_(D102N-WT) < 3 kJmol⁻¹q⁻¹) to moderate (3 <Δζ_(D102N-WT) < 6 kJmol⁻¹q⁻¹) change is experienced by residues Trp60d, Glu60f, Leu99 and Ile174, which are part of the S2 and S3 recognition subsites. Residues Leu40-Cys42, Arg173-Asp178 and Glu146-Trp148 become more positive, too. High-level perturbation (Δζ_(D102N-WT) > 6 kJmol⁻¹q⁻¹) is seen for Asp189 and Ser195 that are 13 and 6Å away from the site of the mutation. Finally, the most significant variation strikes the 213–217 strand where the Δζ reaches a value of 12.5 kJmol⁻¹q⁻¹.

Molecular Dynamics Simulations

All-atom molecular dynamics MD simulations were performed on the E forms of the WT and D102N to further understand the structural consequences of this perturbation. Overall, the mutation does not change the tertiary and secondary structure of the enzyme, as expected(15). The effect is rather local; it affects flexible loops and overlaps with regions identified previously by our electrostatic calculations. Relevant to the catalytic mechanism, His57 shows an enhanced flexibility in the mutant. This is due to a different protonation state of Nε2. The altered protonation state results in increased mobility of Ser195. The residue Arg221a, whose carbonyl oxygen offers hydrogen bonding to coordinate Na⁺, shows a much higher RMSD upon mutation, from 0.17 to 0.35 nm. The larger conformational space explored by the side chain of Arg221a in the D102N structure makes the formation of such an interaction unlikely. Relevant to the E*-E conformational equilibrium, the side chains of Trp60d, Trp148 and Trp215 experience similar or lower dynamics but on average assume a position closer to the catalytic triad. Some differences are also notable at the N-terminal region but the high RMSD in both protein systems discourage facile conclusions. A more comprehensive view of the structural and dynamic changes introduced by the mutation comes from analysis of the variance-covariance matrix obtained by averaging the MD trajectories over the last 20 ns of the simulation (figure 2). Thrombin WT and D102N show similar magnitude of the diagonal elements indicating unaltered backbone flexibility upon mutation. On the other hand, the magnitude of most of the out-of-diagonal elements is reduced in D102N suggesting that the spatial correlation between residues has been weakened by the mutation. This effect is particularly evident for the 141–155 segment linking the flexible autolysis loop to the E*-E equilibrium. Indeed, deletion of nine residues of this loop, ¹⁴⁶ETWTANVGK^149e, stabilizes thrombin in the E* form (42) whereas replacement of the autolysis loop with the homologous loop of the murine enzyme, ¹⁴⁶ETWTTNINEI¹⁵⁰, results in a significant improvement of catalytic activity via stabilization of the E form(43). The deletion Δ146–149e eliminates one negative charge, Glu146, which is part of the outer ring. On the other hand, in the murine/human swap, the replacement of ^149dGKG¹⁵⁰ with ^149dNEI¹⁵⁰ rigidifies the loop and inverts the charge ratio of the polypeptide sequence from + 1 to −1 increasing electrostatic repulsion.

Figure 2. Transformed variance-covariance matrices, m, for WT (A) and D102N (B).

The transformed matrix, m, with elements m_i,j = ln(|M_i,j|) was used to map the conformational flexibility profile resulting from Asp102 to Asn mutation. The degree of variance-covariance is color-coded; no variance-covariance is shown as dark blue; high variance-covariance is shown as dark red. Numbering follows the chymotrypsinogen scheme and the secondary structure elements are reported as it follows: α-helices in red, β-sheets in black, loops/random coils in green. The most striking difference concerns the segment 141–155, which is centered on the residue Gly149e in the autolysis loop. In the wild-type (A), this segment is strongly correlated with all other residues of the protein (intense red cross) but this spatial correlation is lost upon mutation.

Further insights on the role of Asp102 come from simulations in which the negative charge on the carboxylate-group was gradually turned off before substituting Asp to Asn. Moving from panel A (−0.75) to panel D (0) to panel E (-NH₂) of figure 3, there is almost a complete reversal of the colors for residues Glu192, Trp215, Glu217 and Asp221, from blue to red. Residue Glu146 goes from red to blue. Of note, residues Glu192, Trp215, Glu217 and Asp221 move towards Tyr60a and Trp60d in the 60-loop, which is located on the opposite side of the active site pocket. This movement is expected to narrow the aperture of the active site, thereby decreasing the docking efficiency of any ligand. Considering the solvent accessible surface area (SASA) of the active site pocket an appropriate structural marker of the occlusion, we continuously monitored its fluctuation during the simulations. Results are shown in figure 3F. From step 1 to 3, i.e. Asp(−0.75), Asp(−0.5) and Asp(−0.25), SASA^D102N remains on average similar to SASA^WT. In the penultimate step, when the charge is completely annihilated, the two traces start differentiating with SASA^D102N becoming lower than SASA^WT. The same trend continues after mutating Asp(0) to Asn. Interestingly, there is a drop at 50 ns in the trace of the mutant. However, this is a transient state and not statistically significant. In fact, the system rapidly relaxes and stabilizes to an average SASA of 1050Ų, which is the center of the Gaussian distribution of the values of SASA^D102N obtained over the trajectory of the last 50 ns of the simulations. On average, SASA^D102N is 100Ų smaller than SASA^WT and this is due to the movement of residues 192, 215, 217 and 221 towards the 60-loop. Remarkably, the charge annihilation of Asp102 is necessary and sufficient to trigger the conformational transition in thrombin. Substitution of the hydroxyl group with an amine simply completes the stabilization by offering perhaps a favorable environment to form hydrogen bonds.

Figure 3. Alchemical mutation.

Variation (Δ d_i,j) of selected residues for each alchemical step: (A) −0.75, (B) −0.5, (C) −0.25, (D), 0 and (E) Asn. The distance between a pair of residues (i,j) was calculated for the wild-type (d_i,j^WT) in each step of the simulation and then subtracted from the distance of the same pair calculated in the D102N (d_i,j^D102N). The difference between these two quantities, $\mathrm{\Delta }{d}_{i,j}=d_{i,j}^{D102N}-d_{i,j}^{WT}$, was used to visualize structural changes for those residues whose electrostatic potential was affected by the mutation. The dimension of the circle located at the intersection of each quadrant directly informs on the magnitude and sign of the vector associated with Δd_i,jj. When two amino acids i and j are closer in D102N with respect to WT, Δ d_i,j < 0 and the circle is colored in red. On the contrary, when Δd_i,jj>0, the circle is shown in blue. (F) Evolution of the solvent accessible surface area (SASA) of the active site pocket identified using CASTp for wild-type (black) and D102N (red). Vertical dashed lines mark the end of each step of the alchemical mutation. (G) SASA distribution (P(SASA)) for thrombin WT (black) and D102N (red) fitted to a single Gaussian function.

Spectroscopic Studies

The solution properties of D102N were investigated by circular dichroism (CD) and fluorescence spectroscopy and compared to wild-type (figures 4A,B). In the free form (0.2M ChCl), thrombin wild-type has a very distinguishing far-UV CD spectrum characterized by a shallow transition between the two minima located at 208 nm and 225 nm(44). The far-UV CD spectrum of D102N differs both in shape and intensity. The first minimum at 208 nm is conserved whereas the second minimum shifts 4 nm to the right, from 225 nm to 229 nm. The transition between these two new minima is sharp and reaches a maximum at 220 nm. Although the shape of this spectrum may recall profiles obtained for alpha-helix containing peptides, both the intensity and position of the minima are not compatible with such secondary structure. Instead, the spectrum of thrombin is dominated by the contribution of 31 aromatic amino acids (11 Phe, 9 Trp and 11 Tyr)(44). The predominance of these residues is why the movement of the 215–217 segment in D102N, and in particular relocation of the side chain of Trp215 into the active site, leads to such a different far-UV CD spectrum. Similarly, the more hydrophobic environment experienced by some aromatic residues explains why there is a small but significant increase (6%) of the fluorescence quantum yield (figure 4B)(34). Assuming a collapsed conformation for the mutant D102N, we would also expect a higher resistance to chemical denaturation. Since chemicals must diffuse into the protein core to promote unfolding, disordered and flexible proteins typically require lower concentration of denaturant to unfold whereas globular and compact proteins are generally more stable. In our experiments, the fluorescence intensity at λ_max was monitored while increasing the concentration of urea (figure 4C)(32,44,45). In both proteins, the curves followed a single sharp transition, suggestive of a highly cooperative unfolding process. The concentration of urea that unfolds 50% of the protein, i.e. [urea]_1/2, was then used to compare the stability of thrombin wild-type and D102N. Under all the experimental condition tested, the mutant was more resistant to chemical denaturation displaying a value of [urea]_1/2 of 3.1±0.1M, which is 0.5M higher than the wild-type in the free form but similar to the Na⁺-bound form of thrombin (E: Na⁺).

Figure 4. Spectroscopic studies.

Far-UV CD (A) and fluorescence (B) spectra of thrombin wild-type (black) and D102N (red). (C) Urea-induced denaturation experiments were carried out at protein concentration of 50 nM by following the decrease of fluorescence intensity at 334 nm after excitation at 280 nm. Data are reported as F/F₀ where F is the fluorescence intensity at any given concentration of urea and F₀ is the intensity of thrombin wild-type at pH 8.0 in 200 mM NaCl. Experimental conditions are 20 mM Tris-HCl buffer, pH 8.0, 0.1% PEG8000, 200 mM NaCl (black circles) or 200 mM ChCl (empty squares). The values of [urea]_1/2 extrapolated from the fit are: WT_ChCl = 2.6±0.1 M; WT_NaCl = 3.2±0.1 M; D102N_ChCl = 3.1±0.1 M.

Ligand binding to thrombin wild-type and D102N

Conclusive evidence that mutation of Asp102 to Asn shifts the conformational equilibrium towards the E* form of thrombin comes from binding studies aimed at probing the accessibility of the active site, i.e. PABA and Hir(1–47), exosite I, i.e. Hir(48–64), and Na⁺ binding site (Table 1, figure 5A–D). As expected, in the free form, D102N binds weakly to PABA (>10 mM) (figure 6A) and Hir(1–47) (>150 µM) (figure 6B). The affinity for Na⁺ is also compromised (figure 5C) and decreases 6-fold, from 18±1 mM in the wild-type to 120±5 mM in the mutant. In contrast, binding of the C-terminal fragment of hirudin Hir(48–64) is minimally affected by the mutation (figure 5D). Na⁺ is an allosteric activator of thrombin (46,47) and it is well established that positive linkage exists between Na⁺ binding and transition to the E form(48). This property is conserved upon mutation. In the presence of 0.8M NaCl, PABA binds to D102N with an affinity constant of 354±9 µM, which is similar to 182±5 µM of thrombin wild-type in the free form and only 7-fold lower than the E: Na⁺ form. Similar results were obtained with the N-terminal fragment of hirudin Hir(1–47). Titration of Na⁺ (figure 5B) progressively rescues the opening of the active site and the affinity for Hir(1–47) at 0.8M NaCl is 5.2 µM, which is only 4 times lower with respect to the wild-type in ChCl, i.e. 1.4 µM but still 150-fold worse compared to the E: Na⁺ bound form of the wild-type. The difference between PABA and Hir(1–47) requires attention. Both ligands bind to the active site but interact with different subsites that do not overlap. PABA binds to the D189 in the primary specificity pocket (31) and its ability to penetrate the active site is dominated by diffusion. Hir(1–47) engages multiple subsites including S2 and S3(19,49) whose conformational dynamics are likely affected independently by the adjacent mutation and not linked to Na⁺ binding. This would explain why under almost saturating concentration of Na⁺, the affinity for the larger Hir(1–47) is more compromised compared to smaller PABA. Another class of physiological allosteric activator of thrombin includes molecules that bind to exosite I such as Hir(48–64)(50,51). In the presence of saturating concentrations of Hir(48–64), the affinity for Hir(1–47) increases 10-fold, from 1.4 µM to 140 nM and becomes similar to the affinity of Hir(1–47) obtained in the presence of the Na⁺, i.e. 40 nM. Unexpectedly, this enhancement is not seen in the mutant (figure 6A). Saturating concentrations of Hir(48–64) slightly improves the affinity of Hir(1–47) but are not enough to shift the equilibrium towards a conformation of the enzyme that is fully competent for binding. This can be achieved with minimal concentrations of Na⁺, i.e. 5 mM, suggesting that 1) Na⁺ is necessary to change the distribution between E* and E forms of the enzyme and 2) a strong coupling between exosite I and Na⁺-binding site must exist. Indeed, Hir(48–64) improves 10-fold the affinity for Na⁺, which was measured from the linkage effect on Hir(1–47) binding (figure 6B)(52,53).

Table 1.

Equilibrium dissociation constants (µM) for active site and exosite-I binders to thrombin wild-type and mutant D102N

	WT		D102N
	NaCl	ChCl	NaCl	ChCl
PABA	49 ± 2	182 ± 5	354 ± 9	>10000
Hir(1–47)	0.040 ± 0.002	1.44 ± 0.02	5.2 ± 0.2	>150
Hir(48–64)	2.3± 0.3*	8.1 ± 0.3	2.2 ± 0.3	19.3 ± 0.9
Hir(1–47)+Hir(48–64)	0.037 ± 0.002	0.14± 0.01	2.9 ± 0.1	>120

Experimental conditions are: 5 mM Tris, 0.1% PEG 8000, pH 8.0, at 25 °C. The ionic strength (IS) was kept constant at 200 or 800 mM with ChCl. When the fluorescence signal was too low to describe a titration curve only the lower limit for the K_d was extrapolated from the fit.

^*No fluorescence change was detected upon binding of fragment Hir(48–64) to thrombin wild-type in the presence of NaCl since saturation of the Na⁺-binding site causes a similar fluorescence(58). The value is from Vindingi et al, 1997(50).

Figure 5. Ligand binding to thrombin wild-type and D102N.

Equilibrium binding curves for PABA (A), Hir(1–47) (B), Na⁺ (C) and Hir(48–64) (D) binding to thrombin wild-type (black) and D102N (red). Data are presented as fractional saturation according to the normalization (F−F₀)/ΔF_max where F₀ is the fluorescence intensity at 0 ligand and F_max is the fluorescence intensity at saturating concentration of ligand extrapolated from the fit. Solid lines were drawn according to a simple binding equation with best fit-parameters listed in Table 1. Experimental conditions are: 5 mM Tris-HCl pH 8.0, 0.1% PEG8000, 800 mM NaCl (filled circles) or ChCl (empty squares) for PABA and Hir(48–64) and 5 mM Tris-HCl pH 8.0, 0.1% PEG8000, 0 – 800 mM NaCl for Hir(1–47), IS was kept constant with ChCl.

Figure 6. Linkage between Na⁺, active site and exosite–I.

(A) Equilibrium binding curves for Hir(1–47) binding to thrombin D102N under saturating concentrations of Hir(48–64) and different concentrations of Na⁺. (B) Linkage between Na⁺ and Hir(1–47) binding to D102N in the absence (red circles) and presence (green circles) of saturating concentration of Hir(48–64). Shown is the effect of saturating Hir(48–64) (200 µM) on the equilibrium association constants for Hir(1–47) binding to the thrombin D102N. Continuous lines were drawn according to the linkage equation $K_a=(K_a^0+K_a^1\frac{[N{a}^{+}]}{K_d})/(1+\frac{[N{a}^{+}]}{K_d})$, where $K_a^0$ and $K_a^1$ are the values of K_a for Hir(1–47) binding in the absence and under saturating [Na⁺] and K_d is the equilibrium dissociation constant for Na⁺. Best-fit parameter values are Hir(1–47): $K_a^0=5.7\times {10}^3{\mathrm{M}}^{-1}$ (fixed), $K_a^1=2.5\times {10}^5{\mathrm{M}}^{-1}$ (fixed), K_d (Na⁺) = 578 mM; Hir(1–47) + Hir(48–64): $K_a^0=8.3\times {10}^3{\mathrm{M}}^{-1}$ (fixed), $K_a^1=5.1\times {10}^5{\mathrm{M}}^{-1}$, K_d (Na⁺)= 57±10 mM; Experimental conditions are: 5 mM Tris-HCl pH 8.0, 0.1% PEG8000, at 25°C.

Crystal structure of thrombin D189A

The proposed electrostatic model of thrombin also predicts that removal of the negatively charged Asp189 would favor collapse of the 215–217 segment. Replacement of Asp189 with Ala, Asn, Glu and Ser compromises the specificity of thrombin towards synthetic and physiological substrates up to 4 orders of magnitudes and the mutants D189A and D189N share comparable functional properties(17). The X-ray crystal structure of the thrombin mutant D189A was solved at 1.7Å resolution in the free form, in the absence of any ligands (Table 2). The asymmetric unit contains two molecules that overlap with an RMSD of 0.204Å. The high-resolution allowed us to assign with confidence the site of mutation, but not the 144–152 segment in the flexible autolysis loop. As predicted, D189A crystallizes in the E* form. The strands 215–217 and 190–193 connected by the disulfide bond Cys191–Cys220 collapse into the active site. The Cα-Cα distance between Gly216 and Gly193 is only 7.9Å, as expected for the E* form (figure 7)(6,9). In this conformation, substrate cannot penetrate the primary specificity pocket. The same Cα-Cα distance increases from 7.9 to 12.1Å upon transition to the E form or binding of PPACK. Perturbation of the electrostatic potential caused by substituting Asp102 to Asn and Asp189 to Ala turns thrombin into E*.

Figure 7. Crystal structure of thrombin D189A.

X-ray crystal structure of thrombin mutant D189A compared to D102N free (E*, 2GP9) and D102N bound to extracellular fragment of PAR-1 (EL, 3BEF). Thrombin D189A crystallizes in the E* form. Electron density is 2F₀−F_c map contoured at 1 σ. The Cα-Cα distance between G216 and G193 is 7.9Å. Such a distance is conserved in the crystal structures of D102N(15), Y225P(5), N143P(55), Δ146–149e(42), E217K(56) and thus is a hallmark of the E* conformation in thrombin. E and EL structures require a wider aperture, 12.2Å in order to accommodate any ligand.

Discussion

Recent analysis of the protein data bank has shown that trypsin-like proteases exist in closed (E*) and open (E) forms at equilibrium and points to the 215–217 segment as the gatekeeper of the active site(6). A high resolution structure of chymotrypsinogen was the first to reveal two distinct conformations of the 215–217 segment in two molecules in the asymmetric unit, consistent with the E*-E equilibrium(54). More recently, thrombin mutants Y225P and N143P have been trapped in alternative forms with open and closed conformations of the active site(5,55).

In this study, we offer compelling evidence as to why mutation of Asp102 or Asp189 leads to a dramatic structural rearrangement and stabilization of the E* form, and propose a model that explains the results from previous studies. The active site pocket of thrombin can be represented as a funnel in which Asp102 in the catalytic site and Asp189 in the primary specificity site are positioned below a negative ring surrounding the active site (figure 1B). It is therefore conceivable that neutralization of Asp102 and Asp189 could lead to a rearrangement of the electrostatic potential. Electrostatics, molecular dynamics simulations and structural biology point to the 215–217 strand as the structural element that is the most affected by the charge neutralization and show how the active site pocket becomes stabilized into a collapsed conformation incapable of ligand binding.

Structural highlights of the E* forms are the collapse of the 215–217 segment to occlude the active site, increased flexibility of residue Arg-221a that can be found either exposed to the solvent or pointing towards the protein core occupying the primary specificity pocket, and relocation or disorder of the residues in the 186-loop forming the Na⁺-binding site(5,6,15,56). In these conformations, thrombin would not be able to bind active site ligands such as PABA or the N-terminal fragment of hirudin Hir(1–47) as well as Na⁺. In agreement with structural biology, D102N binds weakly to PABA and the N-terminal fragment of hirudin Hir(1–47) and the affinity for Na⁺ is compromised. Positive linkage exists between the Na⁺-binding site and the opening of the active site as well as the exosite-I and the Na⁺-binding site, as expected from rapid kinetics and structural biology(48,55). Interestingly, the coupling between exosite-I and the active site is weakened by the mutation. In the wild-type, binding to exosite-I favors more active conformers within the E ensemble without affecting significantly the E*-E distribution(9). Since D102N mostly exists in the E* form, the effect of hirugen is small because of the reduced population of E, which has been trapped in the crystal structure of D102N bound to the fragment of the protease-activated receptor PAR-1(57).

Further insight on the conformational and dynamic properties of E* in solution comes from spectroscopic studies. The far-UV CD and fluorescence spectra of D102N are different compared to WT and reflect the enrichment of E* caused by the mutation. The stability studies show that the E* form of thrombin, as represented by the mutant D102N, maybe thermodynamically favored with respect to the E form but not E:Na⁺. This defines an energy landscape where the E form is an intermediate between E* and E:bound forms of the enzyme. The negative charge of Asp102 and Asp189 therefore has a structural role because it reduces the probability of thrombin to be in the collapsed form.