Computational-approach understanding the structure-function prophecy of Fibrinolytic Protease RFEA1 from Bacillus cereus RSA1

Fibrinolytic protease RFEA1 production and identification

Fibrinolytic protease producing strain Bacillus cereus RSA1 (NCBI Accession number MK288105) was isolated from soil samples, procured from numerous garbage dumps of Noida (U.P, India). Bergey’s brochure for identification of bacteriology and 16S rDNA sequencing was employed for identification of the strain. The fibrinolytic protease production was statistically optimized in our previous study (Plackett Burman Design and Central Composite Design using Design Expert^®10.0.8.0, Stat-Ease Inc., Minneapolis, MN, USA) which comprised of peptone (10 g/L), yeast extract (5 g/L), and glucose (0.5 g/L) with final pH 8.0. The cultures were incubated for 24 h at 37 °C with 1% inoculum and agitation rate of 120 rpm. The enzyme was purified as per our standardized protocol in previous study, employing chilled ethanol precipitation and gel filtration chromatography–Sephadex G75 column (50 × 15 mm; Sigma Aldrich, St. Louis, MO, USA), and identified through MALDI-TOF Mass Spectroscopy (Sharma et al., 2020). Further, statistical compositional amino acid outcome of protein RFEA1 was determined using statistical tool SAP Application Performance Standard (SAPS: http://www.ebi.ac.uk/Tools/seqstats/saps/). SAPS utilize FASTA organized amino acid sequence of individual protein and investigates its composition, repetitive structure, charge dispersion, dividing and multiple periodicity (Brendel et al., 1992).

In-silico structural modeling and validation of RFEA1

The structural modelling of RFEA1 was performed using bioinformatic webservers such as I-TASSER (https://zhanglab.ccmb.med.umich.edu/I-TASSER/) (Roy, Kucukural & Zhang, 2010; Yang et al., 2015; Yang & Zhang, 2015), SWISS-MODEL (https://swissmodel.expasy.org/) (Waterhouse et al., 2018), RaptorX (http://raptorx.uchicago.edu/) (Xu, Mcpartlon & Li, 2020; Xu, 2019; Xu & Wang, 2019; Wang et al., 2017a; Wang, Sun & Xu, 2018; Wang et al., 2017b) and Phyre2 (http://www.sbg.bio.ic.ac.uk/phyre2) (Kelley et al., 2015). The modelled RFEA1 structures were validated by structural analysis and verification server (SAVES v6.0, https://servicesn.mbi.ucla.edu/SAVES/). SAVES metaserver runs six different programs for validating the submitted protein structure. We have used Ramachandran Plot (RC plot) (Carugo & Djinovic-Carugo, 2013), ERRAT (Colovos & Yeates, 1993) and Verify3D (Bowie, Lüthy & Eisenberg, 1991; Lüthy, Bowie & Eisenberg, 1992) for the same.

Analysis of Ca1 binding site in RFEA1

Ligplot⁺ v.2.2 (https://www.ebi.ac.uk/thornton-srv/software/LigPlus/) (Laskowski & Swindells, 2011) was employed for detection and visualization of the metal cation (Ca1) binding sites in modelled RFEA1 structure.

Determination of structural analogs of RFEA1

TM-align (https://zhanglab.ccmb.med.umich.edu/TM-align/), an algorithm for protein structure alignment and comparison was employed to identify structural analogs of validated RFEA1 model in PDB library. TM-align employ heuristic dynamic programming iterations to produce residue to residue alignment for two protein structures of unfamiliar uniformity. The algorithm reports top 10 proteins from PDB with closest structural similarity and provides TM-score (ranges from 0 to 1) determining highest degree of structural similarity. Higher the TM-score value, better and perfect is the structural match (Zhang & Skolnick, 2005).

Prediction of enzyme commission numbers and gene ontology terms of RFEA1

COFACTOR (https://zhanglab.ccmb.med.umich.edu/COFACTOR/) online meta-server was employed for reporting molecular and biological functional annotations of RFEA1. COFACTOR subjects the query BioLiP to identify functional insights such as Enzyme Commission number (EC) and Gene Ontology (GO). COFACTOR provides detailed insights about GO from UniProt-GOA and STRING databases. Thus, structure-based function of RFEA1 was predicted using COFACTOR (Roy, Yang & Zhang, 2012; Zhang, Freddolino & Zhang, 2017), wherein the server provides C-Score values (range 0 to 1), where higher score indicates high reliability of each outcome.

Molecular docking

PATCH DOCK v1.3 (http://bioinfo3d.cs.tau.ac.il/PatchDock) (Duhovny, Nussinov & Wolfson, 2002; Schneidman-Duhovny et al., 2005) was employed for conducting docking analysis of RFEA1 with protein fibrin (PDB ID 2HLO). The structure of fibrin in PDB format was obtained from protein data bank (https://www.rcsb.org/) for the study. The modelled structure of RFEA1 by Phyre2 server was used as receptor molecule. PATCH DOCK server is a geometry-based docking algorithm based on shape complementarity principles and operates RMSD clustering to eradicate superfluous solutions. In our study, we have used default RMSD value of 4 Å. PATCH DOCK protein-protein complex results were further refined using Fast Interaction REfinement in molecular DOCKing (FIRE DOCK), (http://bioinfo3d.cs.tau.ac.il/FireDock/php.php) (Andrusier, Nussinov & Wolfson, 2007; Mashiach et al., 2008). The output file of docked complex was visualized with UCSF Chimera 1.15rc (https://www.cgl.ucsf.edu/chimera/) (Pettersen et al., 2004) whereas 2D plot was analyzed and deduced using Ligplot⁺ v.2.2.

In-vitro validation

In-vitro validation and comparative analysis (with our previous research) of the bio-informatics outcome was performed to draw conclusive decisions. The effect of calcium ion on efficacy and thermostability of RFEA1 was studied and confirmed in-vitro. RFEA1 was purified as per our standardized protocol in previous study, employing chilled ethanol precipitation and gel filtration chromatography—Sephadex G75 column (50 × 15 mm; Sigma Aldrich, St. Louis, MO, USA) (Sharma et al., 2020). The enzyme was incubated for 2 h at 37 °C with different concentrations of CaCl₂ (0.5–3.0 mM) and then tested for fibrin hydrolysis. The fibrinolytic activity assay was performed according to Sharma et al. (2020). Further, in another set of experiments we have used 2 mM of CaCl₂ and stability of RFEA1 was tested at different temperatures (20–80 °C). Solution with no CaCl₂ was considered control. All the experiments were performed in triplicates and statistically analyzed. The In-silico interaction of RFEA1 with substrate (fibrin) was compared with our previous in-vitro study.

Identification of conserved domain motifs of RFEA1

The homologous enzymes of RFEA1 were explored through protein sequence similarity search tool: PSI-BLAST (https://www.ebi.ac.uk/Tools/sss/psiblast/). The multiple sequence alignment for the best 10 homologous hits obtained from PSI-BLAST was performed using CLUSTALW (https://www.ebi.ac.uk/Tools/msa/clustalo/) (Madeira et al., 2019) and visualized with JalView version 2.11.1.0 (Waterhouse et al., 2009). Further, WebLogo 3.7.4 program (http://weblogo.threeplusone.com/) was used to generate sequence logos for a clear alignment of identified conserved domains (Crooks et al., 2004; Schneider & Stephens, 1990).

Procured amino acid sequence of fibrinolytic protease RFEA1 and SAP application performance standard analysis

The protein sequence of RFEA1 derived through MALDI-TOF mass spectrometric analysis in our previous study with Mr 39,483 Da and sequence score of 381 amino acids is mentioned underneath (Sharma et al., 2020).

MRSKKLWISLLFALTLIFTMAFSNMSAQAAGKSSTEKKYIVGFKQTMSAMSSAKKKDVISEKGGKVQKQFKYVNAAAATLDEKAVKELKKDPSVAYVEEDHIAHEYAQSVPYGISQIKAPALHSQGYTGSNVKVAVIDSGIDSSHPDLNVRGGASFVPSETNPYQDGSSHGTHVAGTIAALNNSIGVLGVAPSASLYAVKVLDSTGSGQYSWIINGIEWAISNNMDVINMSLGGPTGSTALKTVVDKAVSSGIVVAAAAGNEGSSGSTSTVGYPAKYPSTIAVGAVNSSNQRASFSSVGSELDVMAPGVSIQSTLPGGTYGAYNGTSMATPHVAGAAALILSKHPTWTNAQVRDRLESTATYLGNSFYYGKGLINVQAAAQ

Bioinformatic tool: SAPS confirmed the molecular weight of RFEA1 (39.5 kDa) and specified statistical compositional outcome of the protein RFEA1 as:

A : 48 (12.6%); D : 13 (3.4%); E : 12 (3.1%); F : 8 (2.1%); G: 37 (9.7%); H : 8 (2.1%); I : 21 (5.5%); K : 25 (6.6%); L : 22 (5.8%); M : 9 (2.4%); N : 18 (4.7%); P : 14 (3.7%); Q : 14 (3.7%); R- : 5 (1.3%); S+ : 50 (13.1%); T : 24 (6.3%); V : 33 (8.7%); W : 4 (1.0%); Y : 16 (4.2%). The results indicate that RFEA1 is a serine (13.1%) and alanine (12.6%) rich protein.

Structural modelling of RFEA1 using multiple servers

I-TASSER predicted five models with C-score: −0.16 (Model 1), −1.33 (Model 2), −2.42 (Model 3), −3.24 (Model 4), −0.86 (Model 5). C-score ranges between −5 and 2 where higher conviction is controlled by higher C-score. Thus, first of the five projected models by I-TASSER online server i.e., Model 1 with highest confidence score was selected (Fig. 1A). The analysis further revealed that RFEA1 exhibited maximum sequence identity (99%) with crystal structure of unautoprocessed form of IS1-inserted Pro-subtilisin E template (IS1-ProS221A) (PDB ID: 3whi.1.A) and henceforth structural imposition of RFEA1 with template protein (3whiA) is displayed in Fig. 1A. The estimated template modeling (TM) score (ranges between 0 and 1) is a projected scale for evaluating the structural resemblance between two structures whereas estimated root-mean-square deviation (RMSD) signify an average distance of all residue pairs of 3D model and its experimental structure (Zhang & Skolnick, 2004; Xu & Zhang, 2010). A higher TM-Score specify superior structural match while smaller RMSD value signifies good quality of model. TM-score of Model 1 is 0.69 ± 0.12 and RMSD is 7.1 ± 4.1 Å, which confirms precise topology. The outcome also revealed that projected model possessed single chain comprising ten alpha-helices and sixteen beta-strands. Fig. 1B represents structural modeling of RFEA1 using second server SWISS-MODEL. According to the predicted GMQE (0.83) and QMEAN (−0.82), RFEA1 showed maximum identification (99.15%) to the ‘3whi’ chain A of pro-subtilisin E of Bacillus subtilis. The modelled RFEA1 presented calcium binding site (Ca1) as a part of its structure and five residues (Asp¹⁴⁷, Leu¹⁸¹, Asn¹⁸³, Ile¹⁸⁵ and Val¹⁸⁷) were found within 4 Å for ligand contacts with chain A. Similarly, RFEA1 structural modelling was performed using RaptorX server (Fig. 1C). The tool predicted five models with estimated RMSD (Å): 6.9122 (Model 1), 8.9796 (Model 2), 8.3941 (Model 3), 8.7051 (Model 4) and 10.419 (Model 5). Model 1 with lowest RMSD value was selected and further analysed. The selected model constituted: Strand: 20.5%, Alpha Helix: 32.8%, 3₁₀ Helix: 2.9% and Other: 43.8%. Another server, Phyre2 was employed to generate RFEA1 3D model (Fig. 1D). This server projected model with 100% confidence and 99% identity with template ‘3whiA’. Alpha helix (29%), beta strand (25%) and TM helix (4%) were present in the predicted model. Even though there is high similarity between RFEA1 and template 3whiA, the analysis of unprocessed structures through model-template alignment (Fig. S1) suggests that a large segment of 29 residues from position 1–29 (Met¹–Ala²⁹) is lacking in template with respect to RFEA1 while a segment of 13 residues from position 78–90 (Gly⁷⁸–Pro⁹⁰) is lacking in enzyme RFEA1 with respect to template. Furthermore, the alignment clearly indicates huge dissimilarity in the positions of amino acid residues in both RFEA1 and template.

Validation of the predicted 3D structures of RFEA1

The modelled RFEA1 structures were further validated through online SAVES v6.0 scrutinizing RC plot (Fig. 2), ERRAT (Fig. S2) and Verify3D (Fig. S3). RC plot for modeled structures of RFEA1 by I-TASSER, SWISS-MODEL, RaptorX and Phyre2 servers showed 72.90, 86.10, 87.80 and 84.60% of residues in most favored region with 2.10, 0.30, 0.60 and 0.00% residues in disallowed regions, respectively (Figs. 2A, 2B, 2C & 2D). Table 1 reveals validation statistics which indicated that Phyre2 server predicted better RFEA1 model than other servers, with maximum (100%) of residues in acceptable region and no (0.00%) residues in disallowed region of RC plot, overall quality factor of 89.64 and 99.42% residues with averaged 3D-ID score >=0.2. Henceforth, validation scores suggest that Phyre2 modelled RFEA1 can be used for further structure-based molecular docking analysis.

Assessment of predicted high-affinity calcium binding site in RFEA1

SWISS-MODEL predicted Ca1 binding site of RFEA1 which was analyzed using Ligplot tool. Ligplot v.2.2. analyzed the interaction of Ca1 with chain A (Complex: Ca, tetrahedral) of RFEA1 with residues Asp¹⁴⁷, Leu¹⁸¹, Ile¹⁸⁵ and Val¹⁸⁷. Figure 3 presents the 2D view of Ca1 binding sites of RFEA1 along with the distance (Å) between metal and target atom. The Ca1 site of RFEA1 homolog ‘3whi’ was found as ‘Complex: Ca, trigonal.bipyramidal’ with dissimilar amino acid residue positions (in comparison to RFEA1) in addition to radius between Ca1 and target atom at –Asp¹³¹ (2.18 Å), Asp¹³¹ (2.57 Å), Leu¹⁶⁵ (2.23 Å), Ile¹⁶⁹ (2.27 Å) and Val¹⁷¹ (2.43 Å). According to the modelled structure of AprE176, residues Gly¹⁶⁹, Tyr¹⁷¹ and Val¹⁷⁴ were located within distance of 3.0 Å from Ca ion. Also, for M179, the hydroxyl group of Thr¹⁷⁶ was located closely (2.9 Å) to Ca ion (Jeong et al., 2015). Several other Bacillus enzymes (BPN and Carlsberg) with fibrinolytic potential have been reported with Ca1 and Ca2 sites (Bryan et al., 1992; McPhalen & James, 1988). Furthermore, calcium binding sites (Ca1) Asp⁴¹, Leu⁷⁵, Ile⁷⁹ and Val⁸¹ were present in crystal structure of subtilisin-propeptide complex (PDB ID: 1scj) (Jain et al., 1998), complex between subtilisin from a mesophilic bacterium and leech inhibitor eglin-c (PDB ID: 1mee) (Dauter et al., 1991) and crystal structure of nattokinase from Bacillus subtilis natto (PDB ID: 4dww) (Yanagisawa et al., 2010). The comparison of RFEA1 with different available crystallographic structures of homologous therapeutic enzymes for Ca1 binding residues is detailed in Table 2.

Identification of structural analogs of RFEA1 in Protein Data Bank

The top 10 TM-align identified structural analogs of RFEA1 are detailed in Table 3, in which the first model (Rank 1) with PDB hit: 3whiA holds highest TM-score 0.905. RMSD, IDEN and Cov score of the homolog were 0.81, 0.983 and 0.911 respectively, which designates an acceptable outcome. ‘3whiA’ was observed to be the crystal structure of unautoprocessed form of IS1-inserted Pro-subtilisin E with Classification: Hydrolase from Organism Bacillus subtilis subsp. subtilis str. 168. Gene Names: aprE, apr, aprA, sprE, BSU10300 (EC: 3.4.21.62). Subtilisin catalyzes protein and peptide amides hydrolysis and is an extracellular alkaline serine protease (Graycar et al., 2013). Also, aprE gene of Bacillus sp. is reported to have strong fibrinolytic activity which strongly supports fibrinolytic attribute of our RFEA1 protein (Yao et al., 2017). However, in some cases proteins with similar folds might have distinct functions (Skolnick et al., 2015; Kinyanyi et al., 2018).

Prediction of enzyme commission number and gene ontology terms of RFEA1

COFACTOR server predicted both Enzyme Commission Number (EC) and Gene Ontology term by structural comparisons (local and global) of RFEA1 model with proteins in BioLiP database (Roy, Yang & Zhang, 2012). The server suggested five Enzyme Commission (EC) numbers and active sites with Cscore^EC 0.618, 0.562, 0.557, 0.539 and 0.502. The model with highest and reliable Cscore^EC 0.618 based on PDB ID: 3bx1B with EC number 3.4.21.62 corresponds to enzyme subtilisin, a serine endopeptidase produced by various Bacillus species. Also, the predicted GO terminologies were used to infer modelled RFEA1 molecular and biological functional annotation. The consensus prediction of RFEA1 function suggests that it exhibits serine type-endopeptidase (GO: 0004252, GO score: 0.99) and proteolysis (GO: 0006508, GO score: 0.99) activity. The cellular component suggested extracellular region (GO: 0005576, GO score: 0.65). These findings support the serine endopeptidase and proteolytic attribute of RFEA1.

In-silico interaction of RFEA1 with Fibrin

PATCH DOCK server is a docking algorithm based on shape complementarity principles and was employed to study molecular interactions between fibrinolytic enzyme RFEA1 and fibrin. Figure S4 illustrates the details of output files of predicted RFEA1-fibrin interactive complexes. PATCH DOCK results were further refined and investigated for binding energy and hydrogen bonding in the complex with FIRE DOCK server. FIRE DOCK refined top 10 solutions (Fig. S5) and indicated that solution No. 7 of PATCH DOCK results has highest binding energy of −21.36 kcal/ mol with hydrogen bond contribution of −5.11. The binding energy between protein-protein complexes is demarcated by their contact region and interface, and determines their interaction strength. The more negative value of binding energy signifies stronger interaction between both the proteins. The 3D view of RFEA1-fibrin interaction was visualized using Chimera software (Fig. S6) and the interactive site residues of RFEA1 and fibrin were examined using Ligplot (Fig. 4). Total 9 amino acid residues of fibrin (Cys¹³⁵, Cys¹³⁹, Gln¹³⁴, Lys¹²⁵, Glu¹³⁷, Gln¹¹¹, Glu¹³², Tyr¹¹⁴ and Asn¹¹⁷) and 8 residues of RFEA1 (Cys¹⁶¹, Cys¹⁶⁵, Gly¹⁶³, Ser¹⁶⁴, His¹³², Asp¹⁴⁶, Lys¹⁵⁷ and Gln¹⁴³) served as hotspots which are involved in enzyme-substrate binding. The results thus showed that catalytic triad of RFEA1 ‘Asp¹⁴⁶, His¹³², and Ser¹⁶⁴’ was involved in interaction with fibrin residues ‘Lys¹²⁵, Gln¹¹¹, Gln¹³⁴ and Glu¹³⁷’. The In-silico results regarding binding affinity and substrate specificity of RFEA1 were then compared with the in-vitro results obtained in our previous study (Sharma et al., 2020).

In-vitro validation

The in-vitro analysis revealed a gradual increase in fibrinolytic activity of RFEA1 in the presence of 0.5–2.00 mM of Ca²⁺ ions after 2 h of incubation (Table 4). An increase of approximately 30% in the activity of RFEA1 (130.87 ± 1.76%) was observed with 2 mM CaCl₂. However, slight reduction in the activity (125.71 ± 1.82 and 119.98 ± 1.87%) was detected with 2.5 and 3.0 mM of CaCl₂. Also, with 2 mM of CaCl₂, a significant increase in the stability of RFEA1 was testified at different temperatures. An increase of 45.65, 41.25, 36.98, 30.17, 26.71, 21.12 and 10.58% in fibrinolytic activity was observed at temperature 20, 30, 40, 50, 60, 70 and 80 °C, respectively (Table 5).

Further, the in-vitro efficacy of RFEA1 has been evaluated using both fibrin and mammalian blood clot as substrate in our previous study (Sharma et al., 2020). The study reported higher affinity of RFEA1 towards fibrin with K_m and V_max values of 1.093 mg/mL and 52.39 ug/mL/min. The thrombolytic potential of RFEA1 was evaluated in comparison to a commercial thrombolytic agent streptokinase/myokinase (Biocon, India), using mammalian blood clot. The endogenous fibrinolytic factors such as plasmin and plasminogen were deactivated by thermal treatment of blood clots. Complete clot dissolution was observed within 4 h with RFEA1 and streptokinase.

Identification of conserved motifs of RFEA1

Conserved domain analysis of RFEA1 was predicted using top 10 homologous enzymes, as depicted by PSI-BLAST, through multiple sequence alignment constructed using ClustalW (Fig. 5). UNIPROT: SUBN_BACNA P35835 Subtilisin NAT (3.4.21.62) (Nattokinase) (Precursor) with Length: 381, 100% Identity, 742 Score (bits), 100% Positives and E-value = 0 was observed to exhibit maximum homology with RFEA1 sequence. Furthermore, moderate similarity of RFEA1 sequence was examined with UNIPROT: SUBT_BACSU P04189 Subtilisin E (3.4.21.62) (Precursor); Length: 381, 99% Identity, 740 Score (bits), 99% Positives and E-value = 0. Multiple sequence alignment suggests numerous conserved columns with a score of 11 indicated by asterisk (^∗) and mutations (Score:10) but conserved properties marked with plus (+). The analysis suggests that RFEA1 is a subtilisin-like fibrinolytic serine protease obtained from novel Bacillus cereus RSA1, with homologs from other sources as well. Besides, the sequence logos of homologous enzymes were generated using Weblogo server to further construct a clear alignment of the identified conserved domains. The results indicated that very low/no significant conservation was found at positions 1, 27, 57–62 and 70–72, whereas highly conserved domains were present at 181–188, 207–213, 303–307, 335–340 and 342–350 residue positions (Fig. 6).

Conclusion

Fibrinolytic enzyme RFEA1 from Bacillus cereus RSA1 with in-vitro thrombus hydrolysis potential might have tremendous possibilities towards industrial/therapeutic deployment in blood clot removal and treatment of cardiovascular thrombosis, respectively. Therefore, comprehending the structural attributes of RFEA1 is imperative to obtain further insights into its molecular and biological functional characteristics. Prediction of the in-silico 3D structural model is exceedingly challenging but beneficial for examination of structure-function aspects of a protein. The presented work is thus an effort to analyse structure based functional aspects of RFEA1. SAPS statistical compositional outcome has evidently presented RFEA1 as a serine (13.1%) and alanine (12.6%) rich protein with molecular weight 39.5 kDa. Validation statistics of modelled structure revealed that the Phyre2 server predicted the RFEA1 model better than other servers. Further study testified the presence of a high affinity calcium binding site in RFEA1. The in-silico molecular docking and in-vitro characterization reflects high binding affinity (−21.36 kcal/ mol) and substrate specificity of RFEA1 towards fibrin. Conclusively, this study provides an insight into structural functional characteristics of RFEA1 and might be a significant contribution in computational analysis for detection/identification of such fibrinolytic enzymes. Nevertheless, in-vitro analysis in our previous study (Sharma et al., 2020) and present study has reported similar characteristics of RFEA1, but in-vivo experimentations would be essential to confirm the claims.