Structures of plasmepsin X from Plasmodium falciparum reveal a novel inactivation mechanism of the zymogen and molecular basis for binding of inhibitors in mature enzyme

2.1 Crystal structure of PfPMX zymogen (Pfpro-PMX)

A polypeptide (Figure 1a) of PfPMX, with truncated prosegment (R28p–N234p, where “p” indicates the residues from the prosegment) and TEV cleavage site (ENLYFQ) at the C terminus was crystallized. From here onwards the zymogen obtained from our construct 1 (R28p–N573) is referred to as Pfpro-PMX and the mature enzyme activated or modeled from Pfpro-PMX is referred to as Pfm-PMX (I235–N573). Crystal structure of Pfpro-PMX has been determined at 2.1 Å resolution (Table 1). Figure 1b shows the sigma-A weighted 2F_o–F_c electron density of the prosegment (Movie S1). The first residue (I235) of Pfm-PMX is defined based on its sequence alignment with pepsinogen (Figure S1a). The polypeptide regions with residues K69p–N70p and E132p–K224p in the prosegment and the C-terminal linker region could not be modeled due to lack of features in the electron density. With the exception of a few residues (F308–S315) in the flap, E343–D352 of loop N2, and the last residue N573, the complete mature enzyme portion of the Pfpro-PMX structure could be modeled and refined. An N-acetylglucosamine molecule is visible in the well-defined electron density (Figure S2), indicating that the protein was glycosylated during its expression.

The overall fold of the prosegment of Pfpro-PMX (Figure 1c) is unique and such an arrangement of the prosegment has not been observed before for any other zymogens of pepsin-like aspartic proteases. The prosegment wraps around the elongated bean-shaped mature enzyme (Figure 1c, d). The structure of Pfpro-PMX can be divided into three segments: domain 1 (F248–D391), domain 2 (R28p–K68p and P392–K572), and domain 3 (P71p–Q247) (Figure 1d). Domain 3 also consists of an “uncharacterized domain” (UD) which contains ~90 residues (E132p–K224p) and is not visible in the crystal structure. The prosegment is folded into the first β-strand (β1, R28p–I32p), followed by an extended loop, the second β-strand (β2, S59p–L66p), an α-helix (α1, H74p–K85p), the third β-strand (β3, D88p–N94p), a second loop with a twisted conformation, the second long α-helix (α2, M114p–D131p), and finally an extended loop that connects to the mature part of the protease. The prosegment is stabilized by the mature domain of Pfpro-PMX through several polar and hydrophobic interactions (Figures 1e–h). In addition, the curved confirmation of the first loop region between β1 and β2 of the prosegment is stabilized by two disulfide linkages formed between C39p and C51p, as well as C42p and C48p (Figure S1b). Part of the prosegment (R95p–G104p) is folded in a unique twisted conformation (twisted loop), which then blocks the S1′ and S2′ pockets of the active site (Figure 2a). The C-terminal part of the twisted loop surrounds the initial segment (P239–K241) of the mature polypeptide. A salt bridge between the side chains of R95p and E433 stabilizes the initial part of this loop. The side chains of I99p and L103p are packed in the hydrophobic pocket formed by the flap region of the mature segment (Figure 2b).

The first five residues (I235–P239) of the mature polypeptide form a β-strand (β4) parallel to the β3 of the prosegment. Further downstream the mature peptide is extended into the active site that is stabilized by several interactions with the main chain and side-chain atoms. One notable interaction in the initial part of the mature region is the formation of a salt bridge by R244 with the carboxylate groups of the catalytic aspartates (D266 and D457) (Figure 2c). Sequence comparison also reveals that R244 is highly conserved in PMXs from other malaria-causing parasites (Figure 2d). The flap of the zymogen is in an open conformation due to folding of the twisted loop of the prosegment that occupies the active site region. Two other loops, N1 and N2, also assume an open conformation, likely due to the folding of the polypeptide that connects the second β-strand and the first α-helix of the prosegment (Figure 1c). Apart from the two disulfide linkages in the prosegment, Pfpro-PMX structure has three more disulfide bridges in the mature domain (Figure S1b). Two disulfide bonds (C279–C284 and C482–C521) are similar to those observed in structures of other plasmepsins and pepsin-like aspartic proteases. The disulfide bond formed by two adjacent cysteine residues (C447–C448), present in the Pfpro-PMX structure (Figure S1c) has not been previously reported for any other pepsin-like aspartic proteases. Sequence alignment shows that five disulfide bonds observed in the Pfpro-PMX structure should be conserved in other PMXs from parasites that cause malaria in humans (Figure S3a).

Analysis of the arrangement of the protein molecules in the crystals reveals that Pfpro-PMX forms an “S” shaped dimer, also observed in zymogens of other vacuolar PMs.²² The loop region of the prosegment consisting of residues N225p–K232p is swapped between the two adjacent monomers of the dimer (Figure S3b, c). Five residues (T227p–T231p) of one monomer form a β-strand that results in the formation of anti-parallel β-sheets by pairing with the first β-strand of the mature polypeptide of another protomer. It is not clear whether such a dimer would form in presence of the UD, which in the current structure is not visible. The UD from different PMXs exhibit variable lengths and very low sequence similarity. However, the sequences without UD are highly conserved, indicating that the PMX zymogens from different malaria-causing parasites might have similar overall structures, with the twisted loop from the prosegment and arginine (R244) from the mature peptide blocking the active site (Figure S3a).

2.2 Structural comparison of Pfpro-PMX with zymogens of other pepsin-like aspartic proteases

Pfpro-PMX has been compared with the representative zymogen structures of other pepsin-like aspartic proteases from parasites, plants, and animals. Previously reported crystal structures of the truncated zymogens of P. falciparum vacuolar plasmepsins (PfPMII, PfHAP, and PfPMIV) reveal that their prosegments have an initial β-strand, followed by two α-helices connected by a turn and a coiled extension with the characteristic “Tyr-Asp” loop that connects to the mature segment of the enzyme.²² The prosegment of the plant and gastric pepsin-like aspartic protease zymogen comprises the first β-strand, followed by two α-helices, a short α-helix, and a turn connecting to the mature domain of the enzyme.

Superposition of the Pfpro-PMX structure with the structures of pepsinogen (3PSG, representing plant and gastric pepsin-like aspartic protease zymogens) and PfPMII zymogen (1PFZ, representing vacuolar PM zymogens) shows that the folding of its prosegment is remarkably different as compared to the latter structures (Figure 3a, Movie S2). Pfpro-PMX has a much longer prosegment than the other pepsin-like aspartic protease zymogens (Figure S1a). The β2 of Pfpro-PMX occupies the same position as the first β-strand of pepsinogen and PfPMII zymogen structures (Figure 3b) and the remaining secondary structure elements have different orientations. The α1 of Pfpro-PMX points downwards, occupying the same space as the tip of a loop in PfPMII zymogen (115p–126p) and pepsinogen; hence, a similar loop is absent in the Pfpro-PMX structure. This loop of vacuolar PM zymogens plays an important role in their activation mechanism²²; the absence of such structural element in Pfpro-PMX may imply a different mechanism of maturation. The next secondary structural elements in pepsinogen and PfPMII zymogen are α-helices pointing in two different directions; in Pfpro-PMX it is a β-strand. The last part of the prosegment in pepsinogen forms a short helical turn that provides a lysine residue blocking the active site aspartates; an equivalent position is partly occupied by the twisted loop region of Pfpro-PMX. In vacuolar PM zymogens, the second α-helix is involved in formation of the inter-dimer contacts and the “Tyr-Asp” loop keeps the prosegment in a strained conformation.²² Inactivation of vacuolar PM zymogens is achieved by the prosegment that acts as a harness separating the two domains of the enzyme, resulting in distancing of the catalytic aspartates. High sequence conservation of the twisted loop region (R95p–G104p) among PMX zymogens from other malaria-causing parasites (Figure 2d) implies homology in their structural features. The sequence and phylogenetic analysis (Figure S4a,b) of the UD reveals that this region might be of plant origin and its structural superposition shows that this region may occupy a similar space to that observed for the plant-specific insert in phytepsin zymogen structure (1QDM). However, only determination of the crystal structure of Pfpro-PMX with this domain visible would confirm the actual fold and 3D arrangement of this insertion region. The pro-mature cleavage site (between N234p and I235) of Pfpro-PMX is distal from the active site as seen in vacuolar plasmepsin zymogens, whereas the cleavage site of pro-mature region is adjacent to the catalytic aspartates in the zymogens of plant and gastric pepsin-like aspartic proteases.

A sequence alignment and superposition of homologous modeled structures of other PMX-like protease zymogens from different apicomplexan parasites reveals that these proteases should also assume similar structural folds, with two characteristic disulfide bridges in the prosegment along with the blocking of the active site by a twisted loop (Figure S5a,b). The folding of the initial segment of the mature peptides in the zymogen structure of pepsin-like aspartic proteases are quite different (Figure 3c). The first part (I235–K241) of the mature polypeptide of Pfpro-PMX structure forms a β-strand parallel to the β3 of the prosegment. The polypeptide that connects the first and second β-strands of the mature enzyme is extended, with R244 forming salt bridge interactions with the two catalytic aspartates (Figure 2c). PMX-like proteases from other parasites also have arginine or lysine at the same position, forming salt bridge interactions with the catalytic aspartates (Figure S5a,b). Blocking of the catalytic aspartates by R244 from the mature polypeptide is unique to Pfpro-PMX. Although a lysine residue in the phytepsin zymogen (K11 from mature polypeptide, 1QDM) and pepsinogen (K36p from prosegment, 2PSG) is blocking the catalytic aspartates, there is an additional tyrosine residue (Y13 in 1QDM; Y9 in 2PSG) which also forms a hydrogen bond with the catalytic residues (Figure 3c, d, Movie S3), but the latter is absent in PMX-like proteases or in vacuolar PM zymogens. Further analysis of the full-length sequence-based structural alignment of different pepsin-like aspartic protease zymogens also places PMX-like proteases close to the plant and gastric proteases (Figure 3e).

2.3 Modeling of the structure of mature Pfm-PMX and its complex with a substrate

A model of the mature enzyme, Pfm-PMX (Figure 4a) was developed. The accuracy and the stability of Pfm-PMX structural fold are supported by MD simulation of the structures with and without inhibitors (Figure S6a–d). Pfm-PMX has pepsin-like aspartic proteases fold with two topologically similar N- and C-terminal domains, but with some striking differences as compared to the enzymes belonging to the same family. The N-terminal domain has a β-hairpin-like structure referred to as the flap (I303–T322). One notable substitution in Pfm-PMX is F311, located in the tip of the flap. An equivalent position in other pepsin-like aspartic proteases (except HAP) is occupied by a conserved tyrosine residue (Y77 in pepsin) which, upon substrate binding, forms a hydrogen bond with a tryptophan residue (W41 in pepsin) due to flap closure, thus, initiating a proton relay to maintain the protonated state of the N-terminal catalytic aspartate.²⁵ Presence of F311 in the case of Pfm-PMX would change the opening and closing dynamics of flap, implying a modified mechanism of catalysis without the assistance of a tyrosine. The mature part of Pfm-PMX has three disulfide bonds between residues C279–C284, C447–C448, and C482–C521; whereas vacuolar plasmepsins usually have only two disulfide bonds (C45–C50 and C249–C282 in PfPMII). Due to formation of the disulfide C482–C521 in Pfm-PMX, the peptide region from residues H478–K492 which has a short loop and helical conformation is stabilized (Figure S7a). An equivalent disulfide bond is not observed at the same location in other pepsin-like aspartic proteases.

Analysis of the MD simulation run of the Pfm-PMX-Apo provides insights into the active site architecture (Figure 4b) of the enzyme. The Cγ atoms of the carboxylate groups of D266 and D457 are at a distance of around 6.0 Å (Figure 4c), indicating the proximity of these two residues that are needed for catalytic competence. However, carboxylate groups of these two catalytic residues seem to rotate during simulation. The carboxylate group of D457 forms a hydrogen bond with the Oγ1 atom of the T460 side chain (Figure 4d). Also, the Oγ atom of the side chain of S269 is in a constant hydrogen-bonding distance from the carboxylate group of the catalytic D266 (Figure 4e) and around 6.0 Å away from the Nε1 atom of W273 (Figure 4f). All these distances are similar to the corresponding distances reported for the crystal structures of other pepsin-like aspartic proteases. The presence of a water molecule near the carboxylate group of D457 during the MD simulation also confirms its ability to act as a base during catalysis, as in other pepsin-like aspartic proteases. The side chains of W273 and S269 form a water-mediated interaction which is also extended to D266 somewhat similar to other pepsin-like aspartic proteases. The structure of Pfm-PMX indicates that a proton relay from W273 can occur and aid in catalysis even without the help of the flap tyrosine residue.

To decipher the molecular basis of substrate recognition by Pfm-PMX, the peptide Rh2N with the cleavage site (^P3HSF/IQ^P2’)¹⁹ sequence was docked into the protein active site and the complex was simulated. The results indicate that the substrate is oriented in a conformation with the phenylalanine at P1 occupying the S1 hydrophobic pocket formed by the flap as well as by loops N1 and N2 (Figure 4g, h, Movie S4). The amide backbone atom of phenylalanine (P1) is hydrogen bonded to D457 and its carbonyl group is pointing toward D266. The side chain of the serine (P2) points away from the flap and its amide backbone interacts with the side chain of the flap S313. Histidine (P3), the N-terminal residue of the substrate, occupies the S3 pocket of the protein formed by residues N353, Y462, and D529. The Nε2 atom of the P3 histidine interacts with the side chain of N353 of loop N2. The side chain of glutamine (P2′) at the C-terminal end makes an interaction with Y431 (Figure 4g, h, Movie S4). Most of these residues interacting with substrates are conserved among the PMX-like aspartic proteases implying a similar mode of substrate binding in these enzymes.

2.4 Removal of the prosegment of Pfpro-PMX to form mature Pfm-PMX

The full-length polypeptide (~65 kDa) of PfPMX zymogen is longer compared to other vacuolar PM zymogens (Figure S1a), and it also has a UD (10.1 kDa) inserted into the prosegment (Figure 1a). A construct (construct 1, R28p–N573 with prosegment R28p–N234p) of Pfpro-PMX, expressed a fusion protein (Figure 5a) with MW ~62 kDa in Expi 293 GnTI-cells and analysis of the eluted protein fractions shows four distinct bands corresponding to the molecular weights of ~62 (full length), ~50.5, ~40, and ~ 12.1 kDa (Figure 5b). N-terminal sequencing of the ~40 kDa band indicated that the putative mature domain begins at L221p and the ~12.1 kDa fragment starts at R28p. Similar pattern of cleavage was also observed when another construct (Construct 2, N190p–N573 with prosegment N190p–N234p) expressed a fusion protein with MW ~61.2 kDa in E. coli (Data S1). These results suggest that the maturation of PfPMX zymogen occurs via cleavage of the prosegment at multiple cleavage sites (CS-1 and CS-2).

Comparison of the crystal structure of Pfpro-PMX with the model of Pfm-PMX reveals that the maturation process of PfPMX zymogen would occur with significant structural changes in the protein (Movie S5). Upon cleavage of the prosegment, the β4 (I235–K241) of the zymogen folds upward to form the first β-strand (β1) of the mature enzyme. Notably, a tripeptide stretch of β1 of the mature enzyme with the sequence P239–L240–K241 is identical to the sequence P63p–L64p–K65p, which is located in the β2 of the prosegment. This indicates a possibility that the β1-strand (P239-L240–K241) of the mature enzyme will occupy the same position as that of the β2-strand (P63p–L64p–K65p) of the zymogen (Figure 5c). After the disruption of the salt bridge with the catalytic aspartates, R244 forms a cation-π interaction with the side chain of Y357 from the loop N2, as we observed in our experiment (Figure S7b). R244 and Y357 are also conserved, indicating the presence of such interactions in PMXs from other human malarial parasites (Figure S3a). Removal of the prosegment creates free space, and, as a result, the flap adopts a closed conformation (Figure 5d) with loops N1 and N2 moving downwards (Figure 5e). Such conformational changes involving these secondary structure elements are also observed in other pepsin-like aspartic proteases. Implications of these regions for inhibitor binding have been discussed in the subsequent sections.

2.5 Mode of binding of 49c in the Pfm-PMX active site

The compound 49c (Figure 6a), containing a hydroxylethylamine scaffold, was designed as a peptidomimetic inhibitor of plasmepsins²⁶ and is reported to inhibit PfPMX.^{12, 18} The effect of 49c on PfPMX stability at different pH conditions was monitored using nanoDSF (Figure 6b). Difference in melting temperature (ΔTm°C) of Apo- and 49c bound PfPMX suggests that the binding of 49c increases thermal stability of the enzyme. Sequential dip in the difference in melting temperature (ΔTm°C) is observed with an increase in pH, with the highest ΔTm°C observed at pH 5.5, suggesting that binding of 49c is pH dependent.

To understand the molecular basis of the binding of 49c to Pfm-PMX, the compound was computationally docked into the enzyme active site and the inhibitor-bound complex was simulated. Here we present, for the first time, a detailed picture of the interactions of 49c with Pfm-PMX (Figure 6c). Several polar and hydrophobic interactions stabilize the binding of 49c in the active site of Pfm-PMX (Movie S6). The inhibitor binds in a conformation such that its phenylalanine group occupies the S1 pocket, making π–π stacking interactions with F311, similar to those observed for the substrate (Figure S8a). The central hydroxyl group of the inhibitor points toward D266. As the phenylalanine group of 49c mimics the P1 group of a substrate, inhibitors with identical or similar groups at this position could possibly have a binding mode similar to 49c or the substrate. The 49c interacts with the catalytic aspartates D266 and D457, as well as with the residues (F311–G314) from the tip of the flap (Figure S8a). The carboxylate group of D266 is in constant hydrogen bonding with the hydroxyl group (atom number 28) of phenylalanine moiety (Figure 6d) and the carboxylate group of D457 side chain is interacting with the N30 atom of the phenylethylamine moiety of 49c (Figure 6e). Also, the carbonyl group of benzamide (O17) interacts with S313 from the flap during most of the time of simulation, holding the flap in a closed state (Figure 6f). Hydrogen bonding of D266 and D457 carboxylate groups with the hydroxyl groups of the nearby residues S269 (Figure S8b) and T460 (Figure S8c), respectively, hints at the correct geometry of the catalytic residues in the inhibitor-bound state. Based on sequence analysis, it appears that the residues which are involved in binding of 49c in the active site of Pfm-PMX are also conserved or similar (Figure 6g) among PMX-like proteases from various apicomplexan parasites. Therefore, it is likely that 49c would inhibit PMX-like proteases and arrest invasion and egress processes in apicomplexan parasites. Such molecular insights unraveling the interactions of 49c with Pfm-PMX would allow development of potent inhibitors for PMX-like proteases.

Although our data suggest that Pfm-PMX active site can accommodate inhibitors which are transition state mimics with a central hydroxyl group, notably this enzyme is not inhibited by pepstatin A, a potent inhibitor of a majority of pepsin-like aspartic proteases.¹² Structural superposition of Pfm-PMX and pepstatin A-bound PfPMII (1M43) shows that the ligand-binding pocket in both proteases is quite dissimilar (Figure S9). The presence of Y462 in Pfm-PMX, compared to A219 in PfPMII, narrows the binding pocket, leading to possible steric hindrance and blocks the binding of pepstatin A. Screening of pepsin-like aspartic protease inhibitors against Pfm-PMX model suggests that the enzyme has higher affinity toward the compounds which inhibit β-secretase (BACE-I) and HIV-1 proteases (Data S2).

4.1 Expression and purification of recombinant Pfpro-PMX (R28p-N573) (Construct 1) in HEK cells

The synthetic gene encoding the Pfpro-PMX (R28p–N573) of P. falciparum plasmepsin X (XP_001349441) (Figure 5a) was engineered for expression in a mammalian host (Human Embryonic Kidney cells, HEK cells) by GeneArt (Thermo Fisher, Waltham, MA, USA) and cloned into a variant of pTT5 with the mouse Ig secretion signal peptide on the N terminus and a TEV-cleavable 8x Histidine tag on the C terminus. The recombinant protein was expressed in Expi 293 GnTI- (Thermo Fisher, Waltham, MA, USA) at 37°C, using Expi293 expression media (Thermo Fisher, Waltham, MA, USA). After 5 days of culture, the medium was clarified by centrifugation, filtered, and applied to 2 × 5 ml HisTrap Ni Excel columns equilibrated with Buffer A (25 mM Tris–HCl, pH 8.0, 200 mM NaCl, 1 mM TCEP, 50 mM Arginine, 0.25% Glycerol). After the protein was loaded, the column was washed with 50 ml of 2% Buffer B (25 mM Tris HCl, pH 8.0, 200 mM NaCl, 1 mM TCEP, 500 mM imidazole), followed by 50 ml of 4% Buffer B. The protein was eluted with a linear gradient of 4–60% Buffer B. Fractions were analyzed by SDS-PAGE using 4–12% gradient gels in the MOPS buffer. Fractions containing the Pfpro-PMX were pooled and dialyzed overnight at room temperature into Buffer A containing 10 mg/ml TEV and 170,000 U of EndoH. The extent of tag removal and deglycosylation was assessed by SDS-PAGE. The untagged and deglycosylated protein was applied to 1 × 5 ml HiTrap Ni Chelating Column equilibrated with Buffer A. The column was washed 50 ml of Buffer A and the protein was eluted with a linear gradient of 0–60% Buffer B. Fractions containing the untagged and deglycosylated Pfpro-PMX were observed in the flow thru and fraction 1 of the gradient elution. Fractions were pooled and concentrated to ~12 mg/ml for size exclusion chromatography (SEC). The concentrated protein was applied to a 1 × 120 ml Superdex 200 column in the SEC buffer (25 mM Tris–HCl, pH 8.0, 200 mM NaCl, 1 mM TCEP, 1% glycerol). Fractions containing the protein of interest were pooled and concentrated to 11.4 mg/ml for crystallization and a second pool of protein, concentrated to 13.3 mg/ml, was used for assays.

4.2 Study of 49c binding by nanoDSF

A label-free nanoscale differential scanning fluorimetry (nanoDSF) method was used to assess the effect of compound 49c on PfPMX stability. PfPMX was diluted to 0.5 mg/ml (7.9 μM) in the MES buffer in varying pH conditions (25 mM Tris/HCl, 25 mM MES pH 5.5, 6.5, 7.5, 8.5 and 9.5). Thermal stability of PfPMX was characterized at each pH condition with and without 49c at 160 μM. All samples were tested at the final 0.16% concentration of DMSO. Samples were loaded in standard capillaries (PR-C002, NanoTemper Technologies) with a final volume of 10 μl. PfPMX intrinsic fluorescence (aromatic residues fluorescence) was measured over a linear thermal gradient 20–95°C at 1°C min⁻¹ with an excitation power set to 40%. Fluorescent measurements were made at excitation 280 nm (UV-LED) and at emission wavelengths of 330 and 350 nm using Prometheus NT.48 (NanoTemper Technologies). Data from duplicate measurements of each sample were processed using PR.ThermControl Software (NanoTemper Technologies).

4.3 Crystallization, diffraction data collection, and structure solution of Pfpro-PMX

Purified Pfpro-PMX (SSGCID batch ID PlfaA.17789.b.HE11.PD28363) at 11.4 mg/ml was crystallized via vapor diffusion using a condition derived from JCSG+ screen (RigakuReagents), condition A7 (100 mM TrisHCl/NaOH pH 9.4, 18% [w/V] PEG 8000) with 0.4 μl protein + 0.4 μl reservoir as sitting drops in 96-well XJR trays (Rigaku Reagents) at 14°C. Crystals were cryoprotected with 20% ethylene glycol as cryoprotectant added to the reservoir solution. Diffraction data were collected at the Life Sciences Collaborative Access Team beamline 21-ID-F at the Advanced Photon Source, Argonne National Laboratory using a Rayonix MX-300 detector. The diffraction data were reduced with XDS/XSCALE³³ to 2.1 Å resolution. The intensities were converted to structure factors with the program modules F2MTZ and CAD of CCP4.³⁴

The structure was solved via Molecular Replacement with MR-ROSETTA³⁵ using PDB entry 3PSG as the search model. The structure was further modeled in iterative cycles of real-space refinement in Coot,^{36, 37} and through reciprocal space refinement in phenix.refine.³⁸ The structure was subsequently refined for 10 cycles using REFMAC5.³⁹ The resulting electron density map visualized in Coot indicated better features of the missing residues and ligands of the Pfpro-PMX model. Subsequently, manual model building was performed by visual inspection and refinement using REFMAC5. The ligands, solvent molecules, and ions were progressively added at peaks of electron density higher than 3σ in sigma-A weighted F_o–F_c electron density maps while monitoring the decrease of R_free and improvement of the overall stereochemistry. The polypeptide region with residues K69p–N70p and E132p–K224p in the prosegment and the C-terminal linker region could not be modeled due to the lack of features in the electron density. With the exception of a few residues (H308–S315) in the flap, (E343–D352) in N2 loop and the last residue N573, the complete mature enzyme portion of the Pfpro-PMX structure could be modeled and refined. The side chains of the residues with weak or missing electron density have not been added in the structure. The adjacent disulfide bond (447–448) is modeled as cis peptide bond conformation.⁴⁰ The data collection and refinement statistics are presented in Table 1. Quality assessment tools built into Coot, phenix, and MolProbity⁴¹ were used to assess the quality of the structure during refinement.

The structure was deposited in the PDB with code 7RY7. Diffraction images were made available to the Integrated Resource for Reproducibility in Macromolecular Crystallography (www.proteindiffraction.org).⁴²

4.4 Modeling of mature Pfm-PMX

A model of Pfm-PMX (I235–K572) was generated by comparing the crystal structure of Pfpro-PMX and the model obtained from SwissModel (https://swissmodel.expasy.org/). The quality of the model was assessed by the Ramachandran plot and visual inspection of the active site. Furthermore, the stability of the Pfm-PMX model was monitored by molecular dynamics (MD) simulation as discussed in the next section.

4.5 Molecular dynamics (MD) simulations of Pfm-PMX as Apo and its complexes

All the molecular dynamics (MD) simulations in this study were performed by GROMACS 2019.1 with “Amber99SB” force field.⁴³ The force field parameters for the ligands were generated using Antechamber. The “xleap” module of AmberTools was used to generate topologies for Pfm-PMX Apo as well as complexes taking care of the disulfide bonds. The protein molecules (as apo or complexed) were placed at the center of the box with a distance of 20 Å from the wall surrounded by TIP3P water molecules with periodic boundary conditions. The complexes were neutralized by adding the appropriate number of required counter ions. The ParmED (https://parmed.github.io/ParmEd/html/parmed.html) tool was used to convert the topology files obtained from xleap module of AmberTools for the use of those by Gromacs. Energy minimizations were performed for 50,000 steps. The energy minimized systems were further equilibrated using canonical ensemble (NVT) followed by isothermal-isobaric ensemble (NPT). In the NVT equilibration, systems were heated to 300 K using V-rescale, a modified Berendsen thermostat for 1 ns. In NPT, all these heated systems were equilibrated using the Parrinello Rahman barostat for 1 ns to maintain a constant pressure of 1 bar. The unrestrained production MD simulations were performed for 100 ns for all structures. The covalent bonds involving H-atoms were constrained using the “LINCS” algorithm and the long-range electrostatic interactions with particle mesh Ewald (PME) method. The time step for integration was set to 2 fs during the MD simulation.

MD simulation trajectories were further analyzed by using the inbuilt tools in GROMACS 2019.1 and visual molecular dynamics (VMD).⁴⁴ The images were generated using the Discovery studio visualizer (https://www.3ds.com/products-services/biovia/) and PyMol (https://pymol.org/2/). All simulations were performed on the Spacetime High-Performance Computing (HPC) resource at IIT Bombay.

4.6 Sequence alignment

All the sequence alignments have been performed using PROMALS3D multiple sequence and structure alignment server (http://prodata.swmed.edu/promals3d/promals3d.php) and T-COFFEE server (https://www.ebi.ac.uk/Tools/msa/tcoffee/). These alignments were curated manually and the figures were prepared using ESPRIPT server (https://espript.ibcp.fr/ESPript/ESPript/).