Understanding the Autocatalytic Process of Pro-kumamolisin Activation from Molecular Dynamics and QM/MM Free Energy Simulations

Graphical Abstract

MD simulations show that protonation of Asp164 triggers conformational changes and leads to reconstruction of the functional active site for the cleavage of the prodomain. QM/MM free-energy simulations reveal the catalytic mechanism and energetics for the acylation reaction of the autocatalytic process.

Proteolytic enzymes are synthesized as inactive precursors or zymogens to promote folding, prevent unwanted protein degradation, and provide a mechanism for regulating protein function through proteolytic activation. Understanding the ways by which Nature prevents unwanted activation of proteases and mechanisms for conversion of zymogens to active enzymes is of considerable interest.^[1] Nevertheless, a detailed understanding of the activation processes and energetics involved is still lacking. Here we applied molecular dynamics (MD) and quantum mechanical/molecular mechanical (QM/MM) free-energy simulations to study the process and energetics of the conversion of the kumamolisin zymogen to the active enzyme. It is shown that the protonation of Asp164 would trigger conformational changes and generate the functional active site for autocatalysis. The mechanism of acylation for autocatalytic cleavage of prodomain is also derived from the two-dimensional QM/MM free energy simulations. The results seem to indicate that one of the reasons for sedolisins to use an aspartate as a catalyst (e.g., Asp164 in kumamolisin) instead of asparagine (the oxyanion-hole residue in classical serine peptidase) might be due in part to the requirement for the creation of a built-in switch that delays the self-activation until secretion into acidic medium.

Kumamolisin belongs to a recently characterized family of serine-carboxyl peptidases (sedolisins)^[2][3] which are present in a wide variety of organisms and most active at low pH. The members of the family also include tripeptidyl-peptidase 1 (TPP1) for which the loss of the activity as a result of mutations in the TPP1 gene is believed to be the cause of a fatal neurodegenerative disease.^[4] Like other proteolytic enzymes, sedolisins are synthesized as inactive precursors to prevent unwanted activation and proteolysis. The inactive precursors for sedolisins include propeptides of ~200 amino acids in length which may play a role in the folding of the proteins as well as protecting them from being prematurely activated. The activation cleavage of the prodomains occurs after the release of zymogens into the acidic environment, leading to the production of the active enzymes.^[3b]

The crystal structures have been determined for the inactive Ser278Ala pro-kumamolisin mutant^[5] as well as for the active kumamolisin.^[6] A superposition of the structure of the Ser278Ala pro-kumamolisin mutant with that of kumamolisin showed that the catalytic domain of the mutant exhibits a virtually identical structure compared to the active enzyme.^[5] The pro-kumamolisin structure has therefore proved that the catalytic domain has basically adopted a mature-like conformation already in the zymogen form. One of the key structural features of the proenzyme is the presence of a salt bridge between the P₃-Arg169p linker residue (here the suffix ‘p’ designates the residues in the pro-domain) and Asp164 from the catalytic domain.^[5] The existence of this salt bridge seems to prevent the formation of the functional configuration for Asp164.^[7] An interesting question is whether the protonation of Asp164 and subsequently breaking of this salt bridge during secretion into an acidic medium would trigger conformational changes and lead to generation of well-positioned general acid/base catalyst (Asp164) and creation of a functional active site in the preparation of the self-activation.^[5]

The X-ray structure of the Ser278Ala pro-kumamolisin mutant was used to generate the model for the wild-type pro-kumamolisin through a manual change of Ala278 to Ser278. Fig. 1A shows that the catalytic Ser residue, along with Glu78 and Asp82, are located at the correct position and would attack the carbonyl carbon of P₁-His171p during the autocatalytic cleavage. This agrees with the earlier suggestion that the catalytic triad in the pro-kumamolisin structure has already adopted a mature-like conformation.^[5] By contrast, Asp164 is located at a different position compared to that observed in the active enzyme^[6] and forms a salt bridge with the P₃-Arg169p (Fig. 1A). Asp164 is therefore unable to participate in the stabilization of the TI during the autocatalytic cleavage of the peptide bond between P₁-His171p and P₁^’-Phe172p.

Figure 1.

(A) The active site of pro-kumamolisin with Ala278 changed to Ser (wild-type) before the protonation of Asp164. The residues from the catalytic domain are shown in ball-and-stick and those from the pro-domain are in stick. For clarity, most of hydrogen atoms are not added. (B) The average structure after the protonation of O_δ2 and 15ns MD simulations. A similar result was obtained with the protonation of the other oxygen of Asp164.

The average active-site structure for pro-kumamolisin after the protonation of Asp164 and 20ns MD simulations is given in Fig. 1B. It is evident that the salt bridge between P₃-Arg169p and Asp164 has already been broken; Arg169p moved away from its position in pro-kumamolisin and interacts instead with solvent molecules. Interestingly, after the MD simulations the active site has changed to the functional configuration with Asp164 well aligned to act as a general acid/base catalyst. This was confirmed with four independent 20 ns simulations with different initial velocities. As mentioned earlier, the catalytic triad in the pro-kumamolisin structure has already adopted a mature-like conformation. Fig. 1B shows that this catalytic machinery has not been significantly altered by the MD simulations and remains functional. Thus, our simulations have confirmed the hypothesis that the protonation of Asp164 would trigger conformational changes to generate the well-positioned general acid/base catalyst (Asp164) and a functional active site for the next step of the autocatalytic pro-kumamolisin activation.

The two-dimensional free energy map for the acylation step of the autocatalytic cleavage is given in Fig. 2A. The free energy curve along the minimum free energy path in Fig. 2A is shown in Fig. 2B. A comparison of the curve in Fig. 2B with the one for the acylation reaction involving kumamolisin_As published earlier^[7d] shows that the two curves are rather similar, even though the barriers in Fig. 2B are slightly lower (by about 2 kcal/mol). We also determined the free energy curve without the protonation of Asp164, and the free energy barrier was found to be considerably higher (by more than 10 kcal/mol). This result suggests that the protonation of Asp164 and breaking of the salt bridge are the necessary steps for autocatalysis, consistent with experimental observations. Some average active-site structures obtained from the QM/MM simulations are given in Figure 3A–C. As is evident from Figure 3, Ser278 is the nucleophile that attacks the carbonyl carbon atom C(His171p), whereas Glu78 and Asp164 act as the general base and acid catalysts, respectively.

Figure 2.

(A) Two-dimensional free energy (potential of mean force) map for the acylation step of self-activation obtained from the QM/MM simulations. (B) The free energy curve along the minimum free energy path in A. Here ES is used for the initial state of pro-Kum before acylation (i.e., after the conformational changes to generate the active configuration), TSa1 is the first TS, TI1 is the tetrahedral intermediate for acylation, TSa2 is the second TS, and AE1 is the state corresponding to the acylenzyme in the enzyme-catalyzed reaction.

Figure 3.

Certain active-site structures along the reactant path (See Fig. 2B). (A) Initial structure before acylation. (B) TSa₁. (C) TI.

In conclusion, we were able to confirm the hypothesis that the protonation of Asp164 would trigger conformational changes to position this residue as a general acid/base catalyst, thus reconstructing the functional active site. We have also studied the mechanism of the acylation reaction for the autocatalytic cleavage of the prodomain and obtained the free energy map and catalytic mechanism for the acylation reaction.

Computational Section

The initial coordinates for pro-kumamolisin were based on the X-ray structure (PDB code: 1T1E) of the S278A mutant;^[11]Ala287 was converted back to Ser278 manually. The structure was constructed under acidic (pH = 3.0) condition based on calculated pKa values. The whole system was solvated in a TIP3P ^[12] water box of 85 × 106 × 93 Å using the NAMD2.9 program.^[13] Fifty three chloride ions (Cl⁻) were added to neutralize the charged system. Four independent MD simulations of 20 ns each with different initial velocities were performed on the TACC Ranger at the University of Texas at Austin using 128 processes, periodic boundary conditions and a time step of 0.002 ps. The CHARMM force field (PARAM27)^[14] was used along with the SHAKE algorithm.^[15] The system was gradually heated to 300 K from 100 K within a time period of 100 ps, followed by 10 ns equilibration and 10 ns product run (NPT) with temperature controlled by Langevin thermostat.^[16] A typical reactant structure (close to the average structure) was selected to perform the QM/MM MD simulation.

The stochastic boundary molecular dynamics^[17] was used for the QM/MM MD and free energy simulations. The reaction region was a sphere with radius r of 20 Å, and the buffer region extended over 20 Å ≤ r ≤ 22 Å. The reference center for partitioning the system was chosen to be the carbonyl carbon atom of P1-His171p. The QM part includes the side chains of Glu78, Asp82, Asp164, and Ser287, as well as a part of the linker-peptide containing the carbonyl of P2-Pro170p, the backbone of P1-His171p and the C_α and amide group of P1’-Phe172p. The atoms in the QM region were simulated with the Self-Consistent Charge Density Functional Tight Binding (SCC-DFTB) method implemented in the CHARMM program. ^[9] The rest of the system was treated with the MM method using the all-hydrogen potential function (PARAM27). The link-atom approach^[18] was used to separate the QM and MM regions.

The 2D free energy map (potential of mean force, or PMF) from the substrate complex to the acyl-enzyme was determined with the umbrella sampling method^[19] and two-dimensional Weighted Histogram Analysis method^[20]. The time step for QM/MM MD simulation is 0.001 ps. The first reaction coordinate, ξ₁ = R(C-N) – R(C…Oγ), is the distance difference between the scissile peptide bond R[C(His171p)-N(Phe172p)] and nucleophilic attack distance [C(His171p)…O_γ(Ser278)]. The second reaction coordinate is ξ₂ = r(O_δ2-H_a)-r(O-H_a). 1296 windows were used in the 2D PMF calculations, and free energy mesh-contour plot was created with a bin size of 0.05 × 0.05 Ų. For each window, we performed 50 ps heating, 50 ps equilibration 50 ps production run with a force constant of 1000 kcal∙Mol⁻¹∙Å⁻².