Genomic screening of new putative antiviral lectins from Amazonian cyanobacteria based on a bioinformatics approach.

Lectins are proteins of nonimmune origin, which are capable of recognizing and binding to glycoconjugate moieties. Some of them can block the interaction of viral glycoproteins to the host cell receptors acting as antiviral agents. Although cyanobacterial lectins have presented broad biotechnological potential, little research has been directed to Amazonian Cyanobacterial diversity. In order to identify new antiviral lectins, we performed genomic analysis in seven cyanobacterial strains from Coleção Amazônica de Cianobactérias e Microalgas (CACIAM). We found 75 unique CDS presenting one or more lectin domains. Since almost all were annotated as hypothetical proteins, we used homology modeling and molecular dynamics simulations to evaluate the structural and functional properties of three CDS that were more similar to known antiviral lectins. Nostoc sp. CACIAM 19 as well as Tolypothrix sp. CACIAM 22 strains presented cyanovirin-N homologues whose function was confirmed by binding free energy calculations. Asn, Glu, Thr, Lys, Leu, and Gly, which were described as binding residues for cyanovirin, were also observed on those structures. As for other known cyanovirins, those residues in both our models also made favorable interactions with dimannose. Finally, Alkalinema sp. CACIAM 70d presented one CDS, which was identified as a seven-bladed beta-propeller structure with binding sites predicted for sialic acid and N-acetylglucosamine. Despite its singular structure, our analysis suggested this molecule as a new putative antiviral lectin. Overall, the identification and the characterization of new lectins and their homologues are a promising area in antiviral research, and Amazonian cyanobacteria present biotechnological potential to be explored in this regard.

INTRODUCTION

Lectins are proteins of nonimmune origin, which are capable of recognizing and binding to glycoconjugate moieties without altering the covalent structure of any of the recognized glycosyl ligands.1 They are present in most organisms, including viruses, bacteria, fungi, plants, and animals and act in many biological processes, such as host‐pathogen interactions, cell‐cell communication, induction of apoptosis, cancer metastasis and differentiation and targeting of cells, as well as recognizing and binding carbohydrates.2 Proteins classified as lectins show major structural differences, and because of that their relation is more functional than structural.

Some lectins have been evaluated as antiviral agents due to their ability to bind to specific carbohydrates in the infection context. Those lectins can block the interaction of viral glycoproteins to host cell receptors, a mechanism which has already been demonstrated for human immunodeficiency virus (HIV), Zaire Ebola virus, Hepatitis C virus, influenza virus, and others.2, 3, 4, 5, 6, 7

The Phylum Cyanobacteria has been shown to be a promising source of antiviral lectins, mainly because of their anti‐HIV activity. Currently, three lectins have gained prominence: (i) Cyanovirin‐N, which was isolated from Nostoc ellipsosporum, may inactivate HIV strains even at low nanomolar concentrations8, 9; (ii) microvirin, isolated from Microcystis aeruginosa, shares 33% of identity with cyanovirin, but it is 50 times less cytotoxic10; and, (iii) scytovirin, which was isolated from Scytonema varium, acts against Zaire Ebola virus, coronavirus, and Cryptococcus fungi besides having anti‐HIV activity.4, 11, 12 In this sense, screening of new cyanobacterial lectins as well their structural improvement is a reasonable strategy in the search for new microbicide candidates.13, 14, 15

Despite the great biological diversity and ecological importance of the region, the first Amazonian cyanobacteria genome was published only in 2014.16 Since then, other genomic studies have been performed with the aim of investigating the genetic potential and diversity of cyanobacteria from this region.17, 18, 19 However, there are few studies based on biotechnological applications for these cyanobacteria. In this sense, genomic screening has been used successfully to access the genetic potential of an individual or an organism group.20, 21, 22 Thus, the aim of this study was to investigate, by genomic analysis, homology modeling, and molecular dynamics simulations, the presence of potential antiviral lectins in cyanobacterial genomes isolated from Amazonian environments.

MATERIAL AND METHODS

Strains

All of the strains analyzed in this study belong to the CACIAM collection (Coleção Amazônica de Cianobactérias e Microalgas) maintained by Laboratório de Tecnologia Biomolecular at Universidade Federal do Pará, Brazil. The genomes of 7 strains were employed for genomic search: Alkalinema sp. CACIAM 70d17, Cyanobium sp. CACIAM 14,16 Limnothrix sp. CACIAM 69d, Microcystis aeruginosa CACIAM 03,18 Nostoc sp. CACIAM 19, Synechococcus sp. CACIAM 66, and Tolypothrix sp. CACIAM 22. These cyanobacteria were isolated from two lakes in the Amazon region: Tucuruí hydroelectric power station reservoir (3°50′04.9″S, 49°42′32.2″W) and Bolonha Lake (1°25′00.7″S, 48°25′52.6″W) both in Pará, Brazil.

Genomic search

In order to identify putative lectin sequences in the 7 CACIAM strain genomes, a conserved domain approach search was applied. First, amino acid sequences from NCBI annotated as lectins that have solved structures were downloaded. Next, the CD‐HIT standalone version23 was used to cluster sequences with >90% identity, making the data nonredundant. The final dataset was submitted to a conserved domain database (CDD) webserver24 for a domain identification procedure. The results obtained were stored to be used as a positive indication of a lectin in the next step.

All coding sequences (CDS) annotated in the 7 CACIAM strains genomes were extracted and translated with Geneious R9,25 using bacterial genetic code. For each genome, the translated CDS were submitted to CDD. A custom Perl script was applied to parse CDD results, retrieving only sequences that had the same domains identified in lectin sequences.

Homology modeling

To perform the homology modeling, two strategies were used. Sequences having enough identity with templates of Protein Data Bank26 (PDB) were modeled with Modeller9.1627 software. Promals3D28 performed the sequence alignment of the template and the target. A total of 100 models were generated based on the target‐template alignment, considering different conformations, and ranked by molecular probability density function (Molpdf) and DOPE score. Automatic loop refinement was used after model building and the models were generated, satisfying spatial restrictions such as bond lengths, bond angles, dihedral angles, and interactions between nonbonded atoms, and then subjected to validation. Sequences that had no identity with PDB structures were modeled on a I‐TASSER server29 and the best model was selected according to C‐score and alignment quality with the templates.

After the model construction, the stereochemical quality was evaluated using a Ramachandran plot generated in the MolProbity30 server. Verify3D31 determined the quality of folding and finally, Qmean32 was computed to measure the local quality.

Molecular dynamics

All homology models generated here were submitted to a Molecular Dynamics (MD) refinement simulation at 100 ns. After that the proteins with known ligands were complexed to them for a new MD simulation of 210 ns and binding free energy calculations.

To perform these MD, a PDB2PQR server (http://nbcr-222.ucsd.edu/pdb2pqr_2.0.0) was used to determine the protonation state of the protein considering a pH level of 7.0. All steps of preparation and production of MD were produced using the AMBER 16 software package33. The force fields applied were GLYCAM_06j34 and FF14SB35 for the ligand and the protein, respectively. Counter ions Na+ or Cl− were added to neutralize the charges and TIP3P36 water molecules in an octagonal box with 10 Å in each direction of the protein. Energy minimization was performed in five steps, four of these using 3000 cycles of steepest descent and 5000 cycles of conjugate gradients for each one; the heavy atoms were restrained by a harmonic potential of 1000 Kcal/mol*Å2. In the last step, we used 5000 cycles of steepest descent and 30 000 cycles of conjugate gradients and no restraints. The heating and equilibration stage was divided into 14 steps. The temperature was gradually increased, until it reached 300 K. Langevin Dynamics (thermostat) were employed with a collision frequency of 3.0 ps − 1. A harmonic potential of 25 Kcal/mol*Å2 was employed in the initial steps and was turned off during step 13. The heating procedure lasted 650 ps until step 13 and was performed using an NVT ensemble. Afterward, a 2‐ns equilibration phase was employed in an NPT ensemble. The SHAKE algorithm was employed to restrict vibration of the ligations of all hydrogen atoms. The Particle Mesh Ewald method was used for calculating electrostatic interactions using a cutoff value of 10.0 Å.

Binding free energy calculations

For the systems complexed with a carbohydrate ligand, Molecular Mechanics Generalized Born Surface Area (MM‐GBSA), Molecular Mechanics Poisson‐Boltzmann Surface Area (MM‐PBSA)37 and Solvated Interaction Energy (SIE)38 methods were employed to calculate the binding free energy of the protein‐ligand complexation (ΔGbind). It used the HIV envelope dimannose (MAN‐MAN) to evaluate the lectin affinity for this virus. These calculations were based on five thousand snapshots from the last 10 ns of the molecular dynamics simulations.

RESULTS AND DISCUSSION

A total of 75 unique CDS were returned from the genomic search in the 7 CACIAM strains presenting one or more lectin domains, according to the comparison with the NCBI solved lectin structures dataset (Table 1). More than 50% of these sequences are annotated as hypothetical proteins, which reinforces the importance of structural analysis in discovering new proteins and their functions.

Table 1

Results of genomic screening showing the number of CDSs predicted with lectins domains

Strain	Sequences	Smaller	Largest
Alkalinema sp. CACIAM70d	2	177 aa	264 aa
Cyanobium sp. CACIAM14	4	219 aa	1257 aa
Limnothrix sp. CACIAM69d	9	289 aa	1652 aa
Microcystis sp. CACIAM03	20	61 aa	1267 aa
Nostoc sp. CACIAM19	24	128 aa	1480 aa
Synechococcus sp. CACIAM66	4	358 aa	784 aa
Tolypothrix sp. CACIAM22	22	99 aa	1507 aa

Known antiviral lectins produced by cyanobacteria present less than 200 aa and 2 or more disulfide bonds in their structures.14 Therefore, the best candidates for homology modeling and molecular dynamics analysis were chosen following these criteria. Additionally, two sequences were chosen due to the presence of the CVNH cyanovirin domain; they are from Nostoc sp. CACIAM 19 and Tolypothrix sp. CACIAM 22 strains. Another sequence of 177 aa identified in Alkalinema sp. CACIAM 70d was selected to be modeled because of its length, the presence of 6 cysteine residues, and the CDD results, which identified the VCBS domain (pfam13517), a well characterized lectin domain. The BlastP tool39 classified this sequence as an integrin‐like fungal protein that adopts a seven‐bladed beta‐propeller structure and interacts with monosaccharides and calcium.

Therefore, these three sequences (the two cyanovirin homologues and the Alkalinema sp. CACIAM 70d sequence) were modeled with the aim of investigating their structural properties.

Cyanovirin tridimensional structures of Nostoc sp. CACIAM 19 and Tolypothrix sp. CACIAM 22 were modeled with Modeller9.16. Nostoc sp. CACIAM 19 cyanovirin homologue presented 103 amino acid residues after the cleavage of a signal peptide predicted by the server SignalP.40 The best alignment with PDB structures showed 60% of identity with the cyanovirin of Cyanothece sp. PCC 7424 (PDB ID: 5 K79),41 which was chosen as template. The root‐mean‐square deviation (RMSD) of these structures was 0.1 Å (Figure 1(A)). The model presented two disulfide bonds between Cys8–Cys22 and Cys59–Cys74.

Figure 1

– Structural alignment of Nostoc sp. CACIAM 19 cyanovirin and its template (a) and Tolypothrix sp. CACIAM 22 cyanovirin and its template (B) [Color figure can be viewed at http://wileyonlinelibrary.com]

Tolypothrix sp. CACIAM 22 cyanovirin homologue presented 105 amino acid residues with no signal peptide. Its alignment with PDB structures showed 39% of identity with a potent variant of cyanovirin from Nostoc ellipsosporum (PDB ID: 1 N02).42 RMSD of target and template structures was 0.4 Å (Figure 1(B)). This model showed only one disulfide bond between Cys9 and Cys23.

I‐TASSER server generated the Alkalinema sp. CACIAM 70d lectin model. The sequence submitted presented 147 amino acid residues after the identification of a signal peptide by the server SignalP.40 This model has 3 disulfide bonds between the cysteine pairs Cys4–Cys26, Cys51–Cys73, and Cys105–Cys127. COACH meta‐server43 present in the I‐TASSER server predicted two relevant putative bind sites for this model around residue Lys140 for sialic acid in C‐terminal portion and another one around residue Cys105 for N‐acetylglucosamine (Figure 2). Sialic acid is derived from neuraminic acid that occurs in polysaccharides, glycoproteins, and glycolipids in bacteria and animals. N‐acetylglusamine is a common carbohydrate in mammal glycoproteins added by cotranslational or posttranslational modifications acting in different biological process.44 Both ligands are related to viral infections in humans; for example, influenza virus hemagglutinin uses sialic acid receptors to attach to human cells,45 and dengue virus uses N‐acetylglucosamine receptors during the infection process.46

Figure 2

– (A) Alkalinema sp. CACIAM 70d lectin model showing two regions predicted as putative binding sites. (B) Electrostatic surface of the model showing the cavities predicted by COACH. Sialic acid binding sites are on the left and N‐acetylglucosamine binding sites are at the center [Color figure can be viewed at http://wileyonlinelibrary.com]

A 100‐ns MD simulation was performed for refining the models. After that, they were validated and the results are presented in Table 2. Conformational changes observed in these simulations were fundamental for improving the validation tests of the models, which were constructed based on crystallographic data. Alkalinema sp. CACIAM 70d lectin showed the highest RMSD values but its Ramachandran evaluation went up from 74% to 96.5% after the 100 ns simulation (Figure 3(A)). The CDD identification, the validated model, and the binding site prediction suggest that this ORF (OUC12179.1) annotated as a hypothetical protein in the Alkalinema sp. CACIAM 70d genome is probably a new putative antiviral lectin, the first described for this new genus.17

Table 2

– Models identification and validation. Ramachandran values show the residues in favorable regions. SignalP server was used to predict the presence of signal peptide (http://www.cbs.dtu.dk/services/SignalP)

Model	Ramachandran	Verify3D	Qmean	Signal P	Domain	Annotation
Alkalinema sp. CACIAM70d	96.5%	100.0%	−2.87	Yes	VCBS	Hypothetical protein
Nostoc sp. CACIAM19	96.0%	100.0%	−1.12	Yes	CVNH	Cyanovirin‐N (CV‐N)
Tolypothrix sp. CACIAM22	94.9%	89.4%	0.39	No	CVNH	Hypothetical protein

Figure 3

– RMSD graphs of MD simulations. (A) 100 ns refinement simulation of the models. (B) 210 ns simulation of Nostoc sp. CACIAM 19 cyanovirin complexed with MAN–MAN. (C) 210 ns simulation of Tolypothrix sp. CACIAM 22 cyanovirin complexed with MAN–MAN

The structural coordinates of crystallographic dimannose (MAN‐MAN) were obtained from cyanovirin 2RDK deposited in PDB47 and were employed as a template to complex this ligand to the cyanovirin models and their respective templates used in the homology modeling step. MD simulations of 210 ns were produced for these four systems and the structural stability of the models is presented in Figure 3(B) and (C). All complexes remained stable during the simulation and the ligand MAN‐MAN showed RMSD values less than 1 Å in the four cases. The final state of the MD simulation for each system is presented in Figure 4. The binding site affinity and the individual contribution of residues were estimated through binding free energy calculations.

Figure 4

– Final coordinates of MD simulation of (A) Nostoc ellipsosporum (1 N02) template cyanovirin, (B) Cyanothece sp. PCC 7424 (5 K79) template cyanovirin, (C) Tolypothrix sp. CACIAM 22 model cyanovirin, and (D) Nostoc sp. CACIAM 19 model cyanovirin. The residues labeled were the ones most important to the binding process according to the binding free energy calculations

The last MD 5000 frames were used for calculating the binding free energy by MM‐GBSA, MM‐PBSA, and SIE methods. The results are presented in Table 3. According to these results, Tolypothrix sp. CACIAM 22 cyanovirin and its template Cyanothece sp. PCC 7424 (5 K79) cyanovirin presented higher affinity to the MAN‐MAN ligand. The individual contribution of residues was evaluated by decomposition of the MM‐GBSA method and the results are presented in Figure 5. Asn, Glu, Thr, Lys, Leu, and Gly, which were described as binding residues for cyanovirin, were also observed on those structures. As to other known cyanovirins, those residues in both of our models also made favorable interactions with dimannose.

Table 3

– Bind free energy calculations results based on last 10 ns of MD simulation. All energy values are in kcal Mol−1

	MM‐GBSA	Std. dev.	Std. error	MM‐PBSA	Std. dev.	Std. error	SIE	Std. dev.	Std. error
Nostoc ellipsosporum (1 N02)	−35.02	4.41	0.14	−31.88	5.74	0.18	−7.97	0.41	0.03
Cyanothece sp. PCC 7424 (5 K79)	−26.74	3.61	0.11	−30.50	3.60	0.11	−7.59	0.33	0.02
Tolypothrix sp. CACIAM 22	−31.31	3.39	0.11	−31.97	3.58	0.11	−7.93	0.30	0.02
Nostoc sp. CACIAM 19	−18.01	2.91	0.09	−19.75	3.20	0.10	−6.85	0.25	0.02

Figure 5

– Individual energy contribution by residues according to the MM‐GBSA method in cyanovirin‐ligand complexation [Color figure can be viewed at http://wileyonlinelibrary.com]

Despite the structural differences among four cyanovirins evaluated here, it was possible to observe the structural conservation of a group of residues. The aspartate, asparagine, and threonine triad appeared in three of the four systems and the Nostoc sp. CACIAM 19 cyanovirin system, which had not conserved the aspartate residue, showed the worst results in binding free energy calculations (Figure 4 and Table 3). Besides that, this triad seems to be fundamental to complexation with dimannose of microvirin, another cyanobacterial lectin.48 Arginine residues were also conserved at the binding site and they presented relevant energy values in decomposition analysis (Figures 4 and 5). In general, polar residues (Glu, Asp, Arg, Lys), capable of making favorable electrostatic interactions with dimannose, collaborated together with the aspartate, asparagine, and threonine triad in the cyanovirin‐ligand complexation. Lys3, Glu80, and Glu94 were the most contributing residues in Nostoc ellipsosporum (1 N02), Cyanothece sp. PCC 7424 (5 K79), and Tolypothrix sp. CACIAM 22 cyanovirin, respectively (Figure 5).

Cyanovirin has been tested against different viral targets and promising results have already been found for human and simian immunodeficiency virus, parainfluenza virus, herpes simplex virus, Epstein‐Barr virus, human herpes virus, bovine viral diarrhea virus, influenza A virus, and others.5, 6, 7, 9, 14 Due to this fact, several studies have been developed with the aim of creating an efficient heterologous expression system for obtaining and purifying cyanovirin. Soy seeds, Pichia pastoris, Nicotiana tabacum, and Althaea officinalis are examples of eukaryotic hosts used to express cyanovirin and its homologues.49, 50, 51, 52

In this sense, the search for new forms of cyanovirin obtained from other cyanobacteria could reveal new applications for this protein, including the reduction of adverse reactions caused by some protein variants. In fact, an identity of approximately 33% is capable of reducing the cytotoxic 50‐fold and maintaining the antiviral activity in microvirin10. Cyanovirin is more active than microvirin due its bivalent iterations with the viral envelopes53, so new cyanovirin forms may present combined properties to be potent inhibitors and less cytotoxic than current variants at the same time. Additionally, the detailed study of residue conservation and binding interactions helps to select the best candidates for antiviral applications.

CONCLUSION

This work was the first attempt to identify antiviral lectins in Amazonian cyanobacterial diversity. A genomic search approach returned a reasonable number of sequences with lectin domains and this pipeline may be replicated in different organisms to help in lectin identification, given that the number of sequences annotated as hypothetical protein was superior to 50%. It was also possible to identify two cyanovirin homologues with average sequence identity with PDB proteins, one annotated as cyanovirin and the other as hypothetical protein, both with the binding property demonstrated by molecular dynamics analysis. Additionally, the genomic search allied with homology modeling suggests that the Alkalinema sp. CACIAM 70d lectin model validated here is a binding sugar lectin, the first one reported for this cyanobacteria genus. Thus, the identification and the characterization of new lectins and their homologues are a promising area in antiviral research, and Amazonian cyanobacteria present biotechnological potential to be explored in this regard.

CONFLICT OF INTERESTS

The authors declare that there is no conflict of interest.