Biochemical properties
The STK was predicted as neutral based on the isoelectric point and the almost equal number of positively and negatively charged residues (51 versus 52, respectively). The enzyme is predicted to be stable since its high stability index, and the estimated half-life for proteolysis of such protein in E. coli was long (Table 1). The pI and molecular weight predictions are consistent with the biochemical analysis of vaccinia virus STK that shares > 95% sequence identity with hMPXV 18. The composition information of the secondary structure of STK was predicted via NetSurfP − 3.0 and iStable 2 webservers. Features, including the percentage of α-helix, β-sheet, and coils, in addition to surface accessibility as well as relative surface accessibility, were calculated. According to NetSurfP − 3.0 and iStable 2, 57% of the protein was predicted to be buried, while the remaining 43% was exposed (Figure S1). In addition, the relative surface accessibility diagram validates the previous results, as most of the peaks had values less than 1.
Concerning secondary structure composition, 44% of the residues are within coils, while α-helices harbor 42% of the amino acids. The remaining amino acids (14%) are expected to be in β-sheets. Although it represented the lowest percentage, β-strands were expected to be part of the active site since it lies between the two lobes. However, except for the glycine-rich loop (a.k.a P loop) (discussed below), no conserved motifs can be detected by multiple sequence alignment of known STKs orthologs along with the hMPXV STK (Figure S2). Indeed, the sequence identity drops to ~ 15% between hMPXV F10 and its orthologs in other genera within the Poxviridae 41.
Table 1
Predicted biochemical features of the STK.
Feature | Value |
Molecular weight (Da) | 52146.40 |
pIa | 7.63 |
Positive residues | 51 |
Negative residues | 52 |
Estimated half-life | > 10 hours |
Instability index | 32 (stable) |
Aliphatic index | 93.69 |
GRAVYb | -0.151 |
a isoelectric point, b GRAVY: grand average of hydropathy |
Model quality and overall fold
The model obtained from AlphaFold2 has an excellent quality according to the Ramachandran plot (Figure S3) and physics-based evaluation tests implemented in MolProbity (Table S1) with an overall score of 1.39 (97th percentile). The MolProbity is an overall computational score equivalent to the x-ray resolution of experimentally solved models 24. Furthermore, according to ERRAT and VERIFY assessments 42, the model has an overall quality factor of 96.1, and 94.34% of the residues have an averaged 3D-1D score ≥ 0.2.
Despite the low sequence identity with known kinases, the hMPXV STK was predicted to have the common structural features, namely, two lobes with a cleft in-between and the glycine-rich loop at the substrate pocket. The STK was predicted in the apparently “open” conformation as a typical kinase with two lobes. Unlike other known STKs from bacteria and eukaryotes, the modeled kinase of hMPXV has additional helices and β-sheets in both lobes. The N-lobe has a stretch of ~ 25 residues that extend onto the C-lobe to form two ɑ-helices (ɑ5 and ɑ6 in Fig. 1) that behave as a structural part of the C-lobe. The N–lobe has seven helices and six anti-parallel β–sheets, whereas the C–lobe consists of 5 helices and four β–sheets. The first 20 residues of the N-lobe comprise a disordered arm, which may be a regulatory sequence for intracellular localization 18.
The biochemical studies on vaccinia virus F10 kinase suggested a membrane association of STK with the internal vesicles, membranous fragments derived from the Endoplasmic reticulum (ER), and the ER-Golgi intermediate compartment (ERGIC) at the assembly location of poxviruses 18. The same study also reported that deletion of the first 21, 69, or 91 residues from the N-terminal abolished the biological function in vitro as well as the enzymatic activity of the STK. It is still unclear whether these deletions induced improper folding or blockage of the association with the lipid membranes.
The active site pocket
Multiple computational tools were employed to identify binding pocket residues of the enzyme, namely, CASTp, fPocket, DoGSiteScorer, and PrankWeb. Among them, PrankWeb was the best in comparison with the other tools as it revealed 11 mutual AA residues (Ile93, Ser94, Thr95, Gly96, Gly97, Tyr98, Gly99, Val101, Val108, Lys110, and Glu127) with those obtained from CB-Dock 2 server (utilize AutoDock Vina) upon docking of adenosine triphosphate (ATP). The pocket volume (solvent accessible) was calculated to be 1717.3 Å3 by CASTp.
From the initial visual analysis, the groove between lobes was expected to be the pocket for substrate binding. Molecular docking of ATP (by GNINA 1.0) has predicted a reasonable binding pose in which the purine ring has an H-bond with Asn342, while the pentose ring oxygen is interacting via another two hydrogen bonds with Ser94 and Ser259 (Fig. 2). Similarly, the terminal gamma phosphate has an H-bond with the amino group of Gly96 at the glycine-rich loop. However, these interactions were slightly modified after 100ns of MD simulation (discussed below). The glycine-rich-loop (a.k.a. P-loop) is a stretch of 7–10 residues interacting with the ATP to facilitate phosphoryl transfer to the other peptide substrate during catalysis 12. The consensus sequence for non-viral kinases is GX1GX2ΦGX3V, where X is any residue and Φ is a hydrophobic residue (tyrosine or phenylalanine). The glycine-rich loop consensus sequence in poxvirus is STGGYGIV.
In nearly all known kinases, the active form is marked by a salt bridge between two conserved residues, lysine and glutamate, near the active site 43–47. The Lysine residue is located in a β-sheet near the glycine-rich loop, whereas the glutamate is part of the activation helix. In the model of hMPXV, the glutamate (E127) is projecting from a helix in the C-lobe (α13 in Fig. 1), while lysine (Lys110) is at a similar location as seen in eukaryotic and prokaryotic kinases, and the two residues were well oriented to form salt bridge even in the absence of the ATP.
The real function of this pair in hMPXV kinase can be deciphered by the construction of mutant STK. The catalytic loop (His305 - Asn312) comprises a short stretch of charged residues that interact with the phosphate moiety and Mg2+ ions. The orientation and folding of the catalytic loop are highly similar to the ATP-bound eukaryotic MEK1 structure 48. The Asp311 in the catalytic loop is expected to facilitate the nucleophilic attack of the oxygen on the ATP gamma phosphate group. The active conformation of MEK1/2 dual kinases is characterized by the formation of the salt bridge between the K-E conserved pair; however, the salt bridge in STK of hMPXV is formed and sustained irrespective of ATP binding to its pocket.
MD simulation
The apo model shows good stability over the simulation period (100 ns) without experiencing unfolding events. The presence of ATP in the active site enhanced STK model stability. The deviations of backbone atoms from the initial structure were limited, roughly within 1 Å in both systems (Fig. 3A). However, the backbone deviations of the ATP-bound system were milder and fastly equilibrated. The backbone-based clustering of the ATP-bound trajectory showed that all frames after 20ns are grouped into three clusters with an RMSD difference of 1.54 Å between their representatives (Figure S4).
The stability of the model was also inferred by the calculations of the radius of gyration, the solvent-accessible surface area (SASA), and the total number of H-bonds (Figs. 3B, 3C, and 3D). In the presence of the ATP, the average radius of gyration decreased slightly as a response to the transition from open conformation to the closed (or partially closed) state in which the ATP becomes surrounded by residues and loops originally positioned away from the active site in the apo conformation. The radius of gyration fluctuates within 2Å for the Apo and 1Å for ATP-bound owing to the restriction of N lobe movement by the presence of the ATP. Similarly, the SASA decreased in the ATP-bound system because the presence of ATP occupied an area accessible to solvent molecules in the apo system.
The all-atoms residue motions during the simulation were calculated by the root-mean-square fluctuation (RMSF). The most fluctuating residues were those in disordered loops between α-helices or β-sheets. This is normal behavior as the residues are highly affected by the motion of the solvent. ATP binding was also stable, as evident by the little changes in its RMSD compared to its initial binding pose predicted by molecular docking.
The stabilized binding pose for ATP shows that the purine ring is buried into the hydrophobic region and stabilized by two hydrogen bonds linking the purine ring to the carboxyl amide group of Pro245 (Fig. 4). In addition, the ribose ring has another H-bond contact with Ser94, whereas the triphosphate end established four H-bonds with Lys309, Lys110, and Asp343 (two bonds). The later interactions are expected to be stabilized for subsequent phosphoryl transfer to serine or threonine residue on the substrate peptide. Additional hydrogen bonds were also seen between the phosphate moiety and water molecules coordinating the Mg2+ (Fig. 5). The phosphate moiety was stabilized via coordinating Mg2+ ions with Asp311, Asn312, Asp343, and possibly Asp345. Hence, the DFD (343–345) in hMPXV is believed to be the equivalent to the conserved DFG motif in eukaryotic and bacterial kinases 18.
The segment (Ser347 to His365) after the DFD motif containing a short helix is expected to be the activation loop. In S. aureus PknB kinase (PDB: 4EQM), this loop occludes the formation of the conserved salt bridge between K39-E58 in the active form by displacing the activation helix away from the ATP-binding site 49; thus, it regulates the transition to the active state. In contrast, the corresponding loop in hMPXV kinase is away from the E127 residue; therefore, it is expected to modulate the peptide substrate for receiving the phosphate group. Similarly, another short loop near the ATP pocket (Ala115-Thr126) may play a crucial role in phosphate transfer to the protein substrate. Indeed, serine and threonine residues on each loop are close to each other and to the phosphate moiety of the ATP pocket. Both loops seem responsible for phosphate transfer reaction to the protein substrate.
MD simulation predicted octahedral coordination of Mg2+ ions by D311, N312, D343, and water molecules (Fig. 5). In most kinases, the purine ring is stacked by a phenylalanine residue at the active site 43–47. However, during the simulation, the nearest phenylalanine residue (F253) sustained its orientation. Only the first aspartate residue of the DFD motif was involved in interactions with ATP or the divalent ion (Mg2+). The perpendicular orientation of the H305 side chain traps the movement of the following phenylalanine (F344) sidechain ring. On the other hand, the D345 side chain is always in a salt bridge interaction (2.77 ± 0.2 Å) with the terminal amino group of the K356 side chain (Figure S5).
A decomposition scheme was followed to calculate the binding free energy via MM/GBSA for the last 50 ns of the simulation trajectory to evaluate the energy contributions of different residues at the active site. The most contributing residues are graphically summarized in Fig. 6. Residues involved in hydrogen bonding with the ATP have an average binding free energy of − 3.36 kcal/mol. These results are helpful for the identification or de novo design of inhibitors against hMPXV.
Conformational changes
Proteins are dynamic macromolecules whose subtle conformational changes (native ensemble) can be sampled by unbiased atomistic molecular dynamics simulations 50,51. The native ensemble was explored in the two simulation runs (Apo & ATP-bound). The distance analyses between certain residues showed that both lobes of the Apo form structure had experienced large-scale motions (Fig. 7A-C). The motion of the N-lobe was reduced when complexed with ATP, whereas the C-lobe retained roughly the same translocation moment pattern in both systems (Apo versus ATP-bound). The motion of lobes is facilitated by three disordered loops (A115-S120, I146-T150, and L255-S295) (Fig. 7C).
Interestingly, the lateral motion of the C-lobe was sustained in both states (apo & ATP-bound). This motion is likely due to the availability of additional conformational space to be occupied when interacting with the peptide substrate(s). The most changes were seen in the residues corresponding to the activation segment (dashed circle in Fig. 7A) that translated towards the phosphate in the ATP-bound state (Fig. 7B). This observation supports the functional role of these residues as the activation segment (via autophosphorylation) or it may play roles in phosphoryl transfer reaction to other protein substrates, most likely via S347 residue following the DFD motif. Despite the larger changes experienced by the triphosphate moiety during the simulation, the glycine-rich loop showed limited conformational changes (Fig. 7D).
The distance/ length of the salt bridge between K110 and E127 in the apo and the ATP-bound is comparable and seems not to be affected by ATP absence (Fig. 7E). In the ATP-bound model, the abrupt increase in the distance at 30-35ns was due to the temporary deviation of the K110 side chain towards the phosphates of ATP to establish temporary hydrogen bonds. However, in the apo system, a transient hydrogen bond was formed between K110 and D343 in the DFD motif.