An integrated approach towards the development of novel antifungal agents containing thiadiazole: synthesis and a combined similarity search, homology modelling, molecular dynamics and molecular docking study

Background This study aims to synthesise and characterise novel compounds containing 2-amino-1,3,4-thiadiazole and their acyl derivatives and to investigate antifungal activities. Similarity search, molecular dynamics and molecular docking were also studied to find out a potential target and enlighten the inhibition mechanism. Results As a first step, 2-amino-1,3,4-thiadiazole derivatives (compounds 3 and 4) were synthesised with high yields (81 and 84%). The target compounds (6a–n and 7a–n) were then synthesised with moderate to high yields (56–87%) by reacting 3 and 4 with various acyl chloride derivatives (5a–n). The synthesized compounds were characterized using the IR, 1H-NMR, 13C-NMR, Mass, X-ray (compound 7n) and elemental analysis techniques. Later, the in vitro antifungal activities of the synthesised compounds were determined. The inhibition zones exhibited by the compounds against the tested fungi, their minimum fungicidal activities, minimum inhibitory concentration and the lethal dose values (LD50) were determined. The compounds exhibited moderate to high levels of activity against all tested pathogens. Finally, in silico modelling was used to enlighten inhibition mechanism using ligand and structure-based methods. As an initial step, similarity search was carried out and the resulting proteins that belong to Homo sapiens were used as reference in sequence similarity search to find the corresponding amino acid sequences in target organisms. Homology modelling was used to construct the protein structure. The stabilised protein structure obtained from molecular dynamics simulation was used in molecular docking. Conclusion The overall results presented here might be a good starting point for the identification of novel and more active compounds as antifungal agents. Electronic supplementary material The online version of this article (10.1186/s13065-018-0485-3) contains supplementary material, which is available to authorized users.


Background
Due to widespread of infectious diseases killing millions of people, the need for new active, safer and more potential antimicrobial agents has increased dramatically. Accordingly, researchers focus on synthesising novel compounds and their derivatives having different physiochemical properties which promises high activities with no or fewer side effects. Heterocyclic compounds are widespread in nature and are used in many fields. It has been known for many years that heterocyclic compounds, especially those containing nitrogen and sulphur atoms, have a variety of biological activities [1][2][3].
Thiadiazole is a five-membered heterocyclic ring system which contains two nitrogen and one sulphur atom with the molecular formula of C 2 H 2 N 2 S. 1,3,4-Thiadiazole and its derivatives have become the focus of attention in drug, agriculture and material chemistry due to their high activity in 2′ and 5′ positions in substitution reactions [4,5].
The two-electron donor nitrogen system (-N=C-S) and hydrogen-binding domain allow for great structural Scheme 1 Synthetic route for the synthesis of the target compounds (6a-n, 7a-n) stability and is known to be the component responsible for biological activity [6,7].
The drug design begins with the synthesis of a compound that exhibits a promising biological profile (lead compound), then the activity profile is optimised and finally ends with chemical synthesis of this final compound (drug candidate).
Computer Aided Drug Design helps to design novel and active compounds which also have fewer side effects. In that respect, in silico molecular modelling has been playing an increasingly important role in the development and synthesis of new drug substances and in understanding the basis of drug-target protein interactions [21][22][23].
In the light of the literature survey, the purpose of this study is to synthesise a number of compounds with different substituted groups containing 1,3,4-thiadiazoles ring and their acyl derivatives, to investigate their antifungal activities and finally to discuss the inhibition mechanism by means of computational tools.

Scheme 2
The formation mechanism for the target compounds (6a-n and 7a-n)

Chemistry
In the first part of the study, the thiadiazole compounds (3 and 4) were synthesised from the reaction of the compounds 1 or 2 (purchased) with the thiosemicarbazide in trifluoroacetic acid (TFA) at 60 °C. The compounds 3 and 4 were obtained as specified in the literature [24,25].
The acyl derivatives of thiadiazole, which are the target compounds of the study, were obtained from the reactions of acyl derivatives (5a-n) with the compounds 3 and 4. All the synthesised 28 compounds (6a-n and 7an) were obtained in moderate to good yields (56-87%). The synthetic route employed to synthesise these compounds is given in Scheme 1 and the formation mechanism is shown in Scheme 2.
As can be seen from the reaction mechanism in Scheme 2, the main reaction proceeds through a typical nucleophilic acyl substitution reaction.
The structure of the compounds obtained were elucidated using the FT-IR, 1 H NMR, 13  analysis and mass spectroscopy techniques. The results are given in detail in "Experimental" section, and the relevant spectra are given in Additional file 1. In addition, the structure of the compound 7n, obtained as a single crystal, was explained with X-ray spectroscopy. The crystal structure of the compound 7n and all X-ray data are provided in Additional file 1.
The target compounds in our study (6a-n and 7a-n) were synthesised in moderate to high yields (56-87%) from the reaction of the acyl chloride derivatives (5a-n) with the 2-amino-1,3,4-thiadiazole derivatives (3 and 4) in the presence of dry benzene.
In the IR spectra of the compounds 6a-n and 7a-n, the symmetric and asymmetric absorption bands corresponding to -NH 2 group (3261-3098 cm −1 ) disappear and instead, the -NH absorption bands at 3186-3092 cm −1 are observed which are the most significant evidences that the compounds were acylated.
Another significant evidence is the C=O absorption band peaks seen at 1720-1624 cm −1 . The appearance of the -NH and C=O absorption bands in the IR spectra is another indication that the compounds (6a-n and 7a-n) were acylated. Other spectrum data of the compounds are presented in detail in "Experimental" section. Also, when the 1 H NMR spectrums of these compounds are examined, the disappearance of the -NH 2 proton signals observed at 7.69 and 7.08 ppm for the compounds 3 and 4 and appearance of -NH signals as a singlet which shift at 13.40-12.09 ppm due to the electron withdrawing property of the carbonyl group, are the most significant evidence that these compounds (6a-n and 7a-n) were acylated. This data is consistent with findings in the literature [24,25]. Other 1 H NMR spectrum data for the compounds are presented in "Experimental" section, and the relevant spectra are given in Additional file 1.
Similarly, when we examine the 13 C NMR spectra of the target compounds (6a-n and 7a-n), the appearance of the C=O carbonyl group peaks at 169.03-162. 49 ppm also supports that the amino group in the thiadiazole ring was acylated. The C-2 carbon signals corresponding to the thiadiazole ring in the compounds 6a-n and 7a-n were observed in the range 161.12-150.26 ppm, and the peaks corresponding to the C-5 carbon were observed between 169.01 and 161.78 ppm. Other 13 C NMR spectrum data of the compounds are presented in detail in "Experimental" section.
In addition, the mass spectra of all the synthesised compounds were obtained and the products were also confirmed with the molecular ion peaks.

In vitro antimicrobial activity studies
The activity values of the compounds against the tested fungus species (inhibition zones and percentage inhibition values) are presented in Tables 1 and 2. At the used doses of the compounds (500 and 1000 µg/ml), varying levels of activity were observed for each fungus species. For all compounds and doses used, the most sensitive fungus species was found to be the Monillia fructigena pathogen. This is followed by Fusarium oxysporum f. sp. lycopersici (FOL) and Alternaria solani pathogens. For thiram, which was used for positive control purposes, a 25 mm inhibition zone was observed for all pathogens, and it inhibited their development at 100%. DMSO, which was used in negative control, did not affect the development of pathogens. According to the results obtained, the smallest inhibition zones at 1000 µg/ml for FOL was found in compound 3 (12.35 mm), and the greatest was in compound 7c (19.51 mm). In case of M. fructigena, the smallest inhibition zones was found in compound 4 (14.25 mm), and the greatest was in compound 7g (21.23 mm); the smallest for A. solani was in compound 6l (10.73 mm), and the greatest in compound 7n with 18.19 mm. In addition, the percent inhibition values that the compounds exhibit against the pathogens were between 49 and 77% at the 1000 µg/ml dose for FOL, between 61 and 85% for M. fructigena, and between 43 and 73% for A. solani (Table 2). It is clear that by increasing the doses used, 100% inhibition rates would be observed. The LD 50 , minimum fungicidal activity (MFC) and minimum inhibitory concentration (MIC) values of the compounds against the fungi were calculated (Table 3). Accordingly, the LD 50 values were calculated to be between 350 and 797 µg/ml for FOL; between 312 and 679 µg/ml for M. fructigena, and between 414 and 1392 µg/ml for A. solani. Despite the overall variation according to the fungus type, the MFC values varied between > 250 and > 1000 µg/ml, and the MIC values varied between < 31.25 and 500. According to the results, it was found that the compounds exhibited high to moderate levels of activity against the tested organisms.

Identification of target protein
Molecular docking is a value added tool in computer aided drug design. It helps us to understand inhibition mechanism of a drug or drug candidate against its target. Ligand similarity search is one of the techniques used for target prediction. This method compares structures of the studied compounds to the compounds with known targets in the databases. For the cases where experimental crystal structures are not available, homology modelling is used to build protein structure based on a template. Optimization or refinement of protein structures are done through molecular dynamics (MD) simulations.
Here we followed a multi-stage computational strategy in order to find a potential target. Initially, the most active two structures based on their LD 50 values for each fungus (Alternaria solani; 6e, 7n. Monilia fructigena; 6k, 7b. Fusarium oxysporum f. sp. lycopersici; 7c, 7h) were selected and used for the similarity search. A number of similar compounds corresponding to our structures was retrieved from NCBI's PubChem database. The resulting structures and their targets are listed in Table 4.
Proteins belonging to the most active compounds in Table 4 were selected for the BLAST search (SRC and ABL1). These proteins are members of non-receptor protein tyrosine kinases family. Besides, FAK1 in Table 4 is also a member of this protein family. On the other hand, some of protein kinases have been shown to be antifungal targets [26,27]. As seen in Fig. 1, a highly conserved kinase domain is present in those three proteins. This region contains a ligand binding site targeted to design anticancer drugs in human, and many protein structures of this domain are available in the Protein Data Bank (PDB) [28][29][30]. Thus, these proteins (SRC, ABL1 and FAK1) were selected for further modelling of the target of our compounds.
As a first step, the amino acid sequences of SRC, ABL1 and FAK1 were retrieved from the Universal Protein Resource (UniProt) and then submitted to the BLAST search to find similar protein sequences present in our target organisms [31][32][33]. Although the BLAST is available for only two species, Fusarium oxysporum f. sp. lycopersici (FOL) and Alternaria solani (AS), the size of proteome information for AS is not adequate. Thus, the BLAST search was performed for only FOL to identify similar protein sequences (Table 5). Finally, two different proteins were identified (Table 5). A remarkable alignment with 33 identical and 61 similar positions was obtained at the protein kinase domain (Fig. 1). After the comparison of those two FOL's proteins with human proteins, A0A0D2XXP0 was found to be more similar to human proteins than A0A0D2XZI2. Besides, additional insertion sites were observed in the A0A0D2XZI2 which can cause a different conformational change at the tertiary structure, and also affect the ligand binding site (Fig. 1). Hence, A0A0D2XXP0 was

Homology modelling
The 3D structure of STE/STE20/YSK protein kinase is currently not available in the Protein Data Bank (PDB). In such cases, homology modelling has been found as an effective method for 3D structure prediction of proteins. Therefore, homology modelling was performed through the Automated Comparative Protein Modelling Server (SWISS-MODEL) [34]. The STE/STE20/YSK protein kinase sequence was retrieved from Uniprot (Uniprot ID: A0A0D2XXP0). A sequence similarity search against other sequences with available structural information in PBD was applied to determine the template structure. A high resolution (1.6 Å) crystal structure of Serine/threonine-protein kinase 24 (MST3) (PDB ID: 4U8Z) was selected as template which shows 66.93% sequence identity (GMQE: 0.80) with the target.
The backbone of the model was validated using Ramachandran plot obtained through RAMPAGE server [35]. The Ramachandran plot for our model structure indicated that 92.1% of the residues were located in the most favourable region, 5.9% of the residues were in the allowed regions, and 2.0% of the residues were in the outlier regions. This suggests that the STE/STE20/YSK protein kinase model is of good stereo chemical quality (Fig. 2). The measurement of the structural error at each amino acid residue in the 3D structural model was measured by the ERRAT plot [36]. The overall quality factor of the model was computed as 97.38% (Fig. 3).

Molecular dynamic simulation
The validated protein model was used in the molecular dynamics simulations. Root-mean-square-deviation (RMSD) and radius of gyration (Rg) were used to check the stability of protein. The RMSD is a crucial parameter to analyse the stability of MD trajectories. To check the stability of protein during the simulation, RMSD of the protein backbone atoms were plotted as a function of time (Fig. 4). The analysis of the RMSD values indicates that the equilibration was reached after 7 ns simulation time.
The radius of gyration, Rg, was also carried out to give us insight into the overall dimensions of the protein.

Molecular docking
Prior to molecular docking of the synthesized thiadiazole derivatives to STE/STE20/YSK protein kinase using our homology-modelled protein, the reliability and accuracy of LeDock was analysed using crystal structures of MST3 which has the best sequence identity with STE/STE20/ YSK protein kinase. Accordingly, 17 PDB structures of MST3 were downloaded from PDB database. Then, crystal binding poses, and binding affinities of native ligands were predicted using LeDock in triplicate. LeDock scores and RMSD values are listed in Table 6. According to the results, LeDock displayed 0.74 ± 0.02 pearson correlation (score vs IC 50 ) and predicted the experimental binding mode with 78.43 ± 3.40% (RMSD) success rate. These results suggested that LeDock can be used as a reliable docking tool for the STE/STE20/YSK protein kinase.
Finally, we performed molecular docking to our studied compounds using stabilized structure of STE/ STE20/YSK protein kinase model. Each docking was carried out in triplicate. The results are listed in Table 7. Ligands are ranked according to the LD 50 values.
The Pearson correlation was also calculated. Although the scoring functions of the docking software available at the market have low success rate in discriminating between active and inactive compounds [24,[37][38][39] the Pearson correlation was found to be 0.63 ± 0.03 which shows a good agreement between experimental LD50 values and calculated docking scores.
In an effort to investigate the differences between the binding modes of the active and non-active compounds, we aligned the most two active compounds (7c and 7h) and also the least two activate compounds (6d and 6i) which share similar scaffolds. As can clearly be seen in Fig. 5, the most active compounds (7c and 7h) adopt similar binding orientations and burry deep into Finally, as can be seen from Table 7, practically all compounds obey Lipinski's rule of five and all have drug-like pharmacokinetic profile.

Table 7 Docking scores, LD 50 values and calculated molecular properties of the studied compounds
a ADMET values could not be calculated due to the steric clashes between methoxy substituents b Total solvent accessible surface area (SASA) in square angstroms using a probe with a 1.4 Å radius (recommended value: 300.0-1000.0) c Logarithm of the partition coefficient of the compound between n-octanol and water (recommended value < 5) d Predicted brain/blood partition coefficient (recommended value: − 3.0 to 1.2) e Predicted apparent MDCK cell permeability in nm/sec (< 25 poor, > 500 great) f Percentage of human oral absorption (< 25% is weak and > 80% is strong) g Polar surface area (recommended value ≤ 140 Å 2 ) [40] h Violations to the Lipinski's rule of five [41] Comp. LD

Conclusion
In the present study, 2-amino-1,3,4-thiadiazole and its acyl derivatives were synthesised with moderate to high yields using simple and applicable methods. The structures of all synthesized compounds were characterised by various spectroscopic methods such as IR, 1 H NMR, 13 C NMR, MS. The in vitro antifungal activity of the synthesised compounds was also evaluated against plant pathogens which revealed promising activities against all tested pathogens.
The combination of several computational tools such as similarity search, homology modelling, molecular dynamics and molecular docking helped in finding a potential target, constructing its 3D model and finally enlighten a possible inhibition mechanism.
In the light of in vitro and in silico results, the studied compounds promise as antifungal candidates worthy of further development in the future.

Materials and methods
The 1 H NMR and 13 C NMR spectra of the compounds were recorded in DMSO-d 6 using an Agilent NMR VNMRS spectrometer at 400 MHz and 100 MHz, respectively. TMS was used as an internal standard.
The IR spectra were measured in ATR using a Perkin Elmer FT-IR Spectrometer Frontier. The mass spectra were measured with a Thermo TSQ Quantum Access Max LC-MS/MS spectrometer. The elemental analysis of the compounds was performed using a LECO 932 CHNS device and the results were within ± 0.4% of the theoretical values. Melting points were recorded on a Thermo Scientific IA9000 series apparatus and were uncorrected. All of the chemicals were obtained from Sigma-Aldrich Chemicals.

General synthesis of 2-amino-1,3,4-thiadiazole derivatives (3,4)
In a round-bottomed flask, compounds 1 or 2 (0.075 mol) and thiosemicarbazide (0.100 mol) in trifluoroacetic acid (5 ml) at 60 °C were stirred for 3-5 h. After completion of the reaction, the reaction mixture was poured into 250 ml ice-water mixture and neutralized with diluted ammonia. The solution was filtered, and solid substance was obtained. The solid substance was washed with water, ethyl alcohol, and diethyl ether, respectively. The solid was recrystallized from the appropriate solvent. The pure substance is dried with P 2 O 5 vacuum oven. Finally, the structures of the synthesized compounds were elucidated with FT-IR, 1 H NMR, 13 C NMR, mass spectroscopy, and elemental analysis. The spectral data and the physical properties of the products are listed below.

General acylation reactions of 2-amino-1,3,4-thiadiazole derivatives (6a-n, 7a-n)
In a two-necked flask, compounds 3 or 4 (0.004 mol) were solved in dry benzene (40 ml) and added pyridine (1 ml) to this solution. Acyl chloride derivatives (5a-n) (0.004 mol) were added drop-wise to this solution at room temperature with the assistance of a dropping funnel. The mixture was then refluxed and stirred for 4-6 h. The progress of the reaction was monitored by TLC at appropriate time intervals. After completion of the reaction, the solution was filtered, and the solid matter was obtained. It was washed with deionized water, ethanol and diethyl ether, respectively. The solid matter was recrystallized from the appropriate solvent. All physical properties and spectral data derived from the obtained products are given below.