Synthesis, Biological Evaluation and Molecular Docking Studies of 5-Indolylmethylen-4-oxo-2-thioxothiazolidine Derivatives

Background: Infectious diseases represent a significant global strain on public health security and impact on socio-economic stability all over the world. The increasing resistance to the current antimicrobial treatment has resulted in the crucial need for the discovery and development of novel entities for the infectious treatment with different modes of action that could target both sensitive and resistant strains. Methods: Compounds were synthesized using the classical organic chemistry methods. Prediction of biological activity spectra was carried out using PASS and PASS-based web applications. Pharmacophore modeling in LigandScout software was used for quantitative modeling of the antibacterial activity. Antimicrobial activity was evaluated using the microdilution method. AutoDock 4.2® software was used to elucidate probable bacterial and fungal molecular targets of the studied compounds. Results: All compounds exhibited better antibacterial potency than ampicillin against all bacteria tested. Three compounds were tested against resistant strains MRSA, P. aeruginosa and E. coli and were found to be more potent than MRSA than reference drugs. All compounds demonstrated a higher degree of antifungal activity than the reference drugs bifonazole (6–17-fold) and ketoconazole (13–52-fold). Three of the most active compounds could be considered for further development of the new, more potent antimicrobial agents. Conclusion: Compounds 5b (Z)-3-(3-hydroxyphenyl)-5-((1-methyl-1H-indol-3-yl)methylene)-2-thioxothiazolidin-4-one and 5g (Z)-3-[5-(1H-Indol-3-ylmethylene)-4-oxo-2-thioxo-thiazolidin-3-yl]-benzoic acid as well as 5h (Z)-3-(5-((5-methoxy-1H-indol-3-yl)methylene)-4-oxo-2-thioxothiazolidin-3-yl)benzoic acid can be considered as lead compounds for further development of more potent and safe antibacterial and antifungal agents.


Introduction
During the last century, several dozen infections have grown and affected the health of millions of people all over the world [1]. In addition to emerging infections, antimicrobial resistance accounts for at least 50,000 deaths each year in Europe and the United States and it is expected that drug resistant infections will be responsible for even larger losses worldwide in the near future [2,3]. Resistant pathogens threaten patients in medical facilities and may emerge in the general population due to the irrational use of antimicrobial agents [4]. Moreover, microbes can transit to the biofilm-growing form to mitigate the harsh environmental conditions or tolerate the presence of a drug. Existing antimicrobial treatment often fails to prevent or eliminate such biofilms [5,6].
Unfortunately, only few novel classes of antibacterial agents (i.e., oxazolidinones, pleuromutilins, tiacumicins, diarylquinolines lipopeptides and streptogramins) have been marketed in the recent decades to solve these problems. Most of them are for the management of Gram-positive bacterial infections [7,8]. However, drug discovery and development is still one of the major ways to ease the burden of microbial infections. Thus, novel molecules with antimicrobial activity are needed and knowledge on their activity profile including both bacterial and molecular targets and off-targets is needed too to provide the possibility to fix such problems as emerging novel infections, drug resistance and tolerance in a rational way.

Introduction
During the last century, several dozen infections have grown and affected the health of millions of people all over the world [1]. In addition to emerging infections, antimicrobial resistance accounts for at least 50,000 deaths each year in Europe and the United States and it is expected that drug resistant infections will be responsible for even larger losses worldwide in the near future [2,3]. Resistant pathogens threaten patients in medical facilities and may emerge in the general population due to the irrational use of antimicrobial agents [4]. Moreover, microbes can transit to the biofilm-growing form to mitigate the harsh environmental conditions or tolerate the presence of a drug. Existing antimicrobial treatment often fails to prevent or eliminate such biofilms [5,6].
Unfortunately, only few novel classes of antibacterial agents (i.e., oxazolidinones, pleuromutilins, tiacumicins, diarylquinolines lipopeptides and streptogramins) have been marketed in the recent decades to solve these problems. Most of them are for the management of Gram-positive bacterial infections [7,8]. However, drug discovery and development is still one of the major ways to ease the burden of microbial infections. Thus, novel molecules with antimicrobial activity are needed and knowledge on their activity profile including both bacterial and molecular targets and off-targets is needed too to provide the possibility to fix such problems as emerging novel infections, drug resistance and tolerance in a rational way.
Among the natural compounds, there are some indole alkaloids, such as echinulin (1), cristatumin A (2), cristatumin D (3) and tardioxopiperazine A (4) from Eurotium cristatum EN-220 that were able to inhibit the growth of Escherichia coli and S. aureus bacteria [9] (Figure 1). It is known that rhodanin is one of the most notorious key materials for the development of effective antibiotics. The favorable antimicrobial activity of rhodanines is due to the similarity of their structure with the chemical structure of penicillin, which has been proved by several researchers [10][11][12][13]. Rhodanine-3-alkanecarboxylic acid derivatives with p-N,N-benzylidenedialkyl (phenyl)amine moieties on benzene ring 5 were found to be active against staphylococcus, micrococcus and streptococcus strains [14]. Rhodanines bearing N-arylsulfonylindole fragment derivatives 6 exhibited inhibitory activity against S.aureus including methicillin resistant strains (MRSA) [15], while rhodanine 7 was found to be potent against methicillin-resistant Staphylococcus aureus, Staphylococcus epidermidis, Staphylococcus aureus, Enterococcus sp. and Mycobacteria ( Figure 2). It is known that rhodanin is one of the most notorious key materials for the development of effective antibiotics. The favorable antimicrobial activity of rhodanines is due to the similarity of their structure with the chemical structure of penicillin, which has been proved by several researchers [10][11][12][13]. Rhodanine-3-alkanecarboxylic acid derivatives with p-N,N-benzylidenedialkyl (phenyl)amine moieties on benzene ring 5 were found to be active against staphylococcus, micrococcus and streptococcus strains [14]. Rhodanines bearing N-arylsulfonylindole fragment derivatives 6 exhibited inhibitory activity against S. aureus including methicillin resistant strains (MRSA) [15], while rhodanine 7 was found to be potent against methicillin-resistant Staphylococcus aureus, Staphylococcus epidermidis, Staphylococcus aureus, Enterococcus sp. and Mycobacteria ( Figure 2).
Prompted by everything mentioned above, as well as based on our previous results [33,34], we designed and synthesized new derivatives incorporating two pharmacophores in the frame of one molecule, indole and thiazolidinone, using a hybridization approach. The aim of this approach is mainly to improve the activity profile and reduce undesired side effects.
As is known, rhodanine derivatives are synthesized by several methods, in particular by dithiocarbamate, bis(carboxymethyl)trithiocarbonate (the Holmberg method), and thiocyanate [35]. The starting 3-arylrhodanines were synthesized using the Holmberg method, making it possible to obtain compounds based on arylamines in high yields and sufficient purity, avoiding the formation of thiourea impurities. To obtain the target products, the interaction of 3-arylrhodanines with aldehydes under the conditions of the Knoevenagel reaction was used [36].
Prompted by everything mentioned above, as well as based on our previous results [33,34], we designed and synthesized new derivatives incorporating two pharmacophores in the frame of one molecule, indole and thiazolidinone, using a hybridization approach. The aim of this approach is mainly to improve the activity profile and reduce undesired side effects.
As is known, rhodanine derivatives are synthesized by several methods, in particular by dithiocarbamate, bis(carboxymethyl)trithiocarbonate (the Holmberg method), and thiocyanate [35]. The starting 3-arylrhodanines were synthesized using the Holmberg method, making it possible to obtain compounds based on arylamines in high yields and sufficient purity, avoiding the formation of thiourea impurities. To obtain the target products, the interaction of 3-arylrhodanines with aldehydes under the conditions of the Knoevenagel reaction was used [36].

Chemistry
The starting materials for the synthesis of the described products were 3-aryl-2thioxothiazolidin-4-ones 3a-e. They were obtained by the reaction of aromatic amines 1a-e with bis (carboxymethyl) trithiocarbonate 2 (Scheme 1). In the second step, the 3-aryl-2-thioxothiazolidin-4-ones further undergo Knovenagel condensation with 1Hindole-carbaldehyde to give the title compounds 5a-l. We used the optimized procedure to do this; the details of the procedure are given in Table 1. Unfortunately, we were not able to obtain 2-hydroxy-4-(4-oxo-2-thioxo-thiazolidin-3-yl)-benzoic acid 3f using 4-amino-2hydroxybenzoic acid as the starting material 1f, since spontaneous decarboxylation with the formation of 3a occurred under the reaction conditions.

PASS-Based in Silico Assessment of Compounds' Activity
Activities of twelve chemical structures had been assessed using PASS [39] and PASS-based web applications to predict antifungal, antibacterial, and kinase inhibitory activity [37,38].
PASS predicted some activities associated with antibacterial and antifungal effects. The antibacterial activity itself has been predicted for each of the 12 structures. Pa-Pi values were in the range from 0.003 to 0.577. According to the PASS assessments, the most probable mechanism of antibacterial effect of studied chemical compounds are: (1) Inhibition of Enoyl-[acyl-carrier-protein] reductase (predicted for 12 compounds with Pa-Pi values in range from 0.136 to 0.577). (2) Inhibition of (R)-Pantolactone dehydrogenase (predicted for five compounds with Pa-Pi values in the range from 0.02 to 0.247). (3) Either with lower Pa-Pi values or for the smaller number of compounds, the several other mechanisms were predicted: inhibition of D-Ala-D-Ala ligase and histidine kinase, antagonism with para-aminobenzoic acid, and antagonism with human tumor necrosis factor-alpha. AntiBac-Pred assessed the probable activity for 12 structures against multiple bacterial strains and species with Pa-Pi values ranging from 0.0001 to 0.387. Overall, some bacteria were predicted as targets for each of the 12 molecules. However, Pa-Pi values were low, only for seven compounds out of 12 did they exceeded 0.3. These assessments were obtained for the activity of compounds against the Bacillus subtilis subsp. subtilis str. 168.  We studied the interaction of 3-aryl-2-thioxothiazolidin-4-ones 3a-e with 1H-indole-3carbaldehydes 4a-d. It was found that, when these reagents are boiled in acetic acid in the presence of ammonium acetate, 3-aryl-5-(1H-indol-3-ylmethylene)-2-thioxothiazolidin-4ones 5a-l are formed in good yields (Scheme 1).
An analysis of the results of obtaining the initial three and target five substances makes it possible to detect a correlation between the substituents' nature and the products' yields. Thus, para-substituents in the aromatic ring in position 3 of thiazolidine have a positive effect on the yields compared to the corresponding meta-substituents. In this case, the nature of the substituent also matters. Hydroxyl substituents have a more favorable effect on yields than carboxyl substituents. The structures of all synthesized compounds were confirmed by 1 H and 13 C NMR spectroscopy. In the 1 H NMR spectra, signals of all protons were present in regions that correspond to the structure of the obtained compounds. In particular, the signals of the methylene group of 3-aryl-2-thioxothiazolidin-4-ones 3a-e are in the range 4.33-4.40 ppm. Signals of the methylidene proton of 3-aryl-5-(1H-indol-3-ylmethylene)-2-thioxothiazolidin-4-ones 5a-l resonate at 7.99-8.17 ppm, which indicates the Z configuration of the double bond at the 5th position of the thiazolidine ring [37,38]. The signals of hydroxy groups were observed at 9.49-9.87 ppm, while those of the NH group were observed at 11.97-12.37 ppm. The indol ring protons appeared in the aromatic area in the range of 6.73-8.15 ppm. The detailed explanation is given in the experimental part.

PASS-Based In Silico Assessment of Compounds' Activity
Activities of twelve chemical structures had been assessed using PASS [39] and PASS-based web applications to predict antifungal, antibacterial, and kinase inhibitory activity [37,38].
PASS predicted some activities associated with antibacterial and antifungal effects. The antibacterial activity itself has been predicted for each of the 12 structures. Pa-Pi values were in the range from 0.003 to 0.577. According to the PASS assessments, the most probable mechanism of antibacterial effect of studied chemical compounds are: AntiBac-Pred assessed the probable activity for 12 structures against multiple bacterial strains and species with Pa-Pi values ranging from 0.0001 to 0.387. Overall, some bacteria were predicted as targets for each of the 12 molecules. However, Pa-Pi values were low, only for seven compounds out of 12 did they exceeded 0.3. These assessments were obtained for the activity of compounds against the Bacillus subtilis subsp. subtilis str. 168.
PASS predicted general antifungal activity as well as some putative mechanisms of antifungal action (Heat shock protein 90 antagonist, Kinase inhibitor) for all 12 compounds with Pa-Pi = 0.005 ÷ 0.612. Application of AntiFun-Pred, however, did not allow identification of the specificity of antifungal action against the particular fungi strains.
Among the PASS-predicted mechanisms of antibacterial and antifungal activities, inhibition of some kinases was identified. To estimate the action on the studied compounds on 20 kinases in more detail, we applied the KinScreen web application. As a result, for all twelve compounds, inhibitory activity was predicted against one or more kinases with Pa-Pi > 0.5. For six compounds, inhibition of three kinases (serine/threonine-protein kinase haspin, serine/threonine-protein kinase Nek11, and serine/threonine-protein kinase SRPK1) was estimated with Pa-Pi ≥ 0.7. Therefore, the studied compounds' antibacterial action may be also due to kinases' inhibition, including those belonging to the host (human) organism.
Taken together, the results of PASS-based activity evaluation suggest that the studied compounds may exhibit antibacterial and antifungal activities. Relatively low, but not negative, Pa-Pi, values obtained with PASS Online, AntiBac-Pred and AntiFun-Pred indicate that either the studied compounds (1) have a significant structural novelty compared to the compounds from the available training sets or (2) structurally similar compounds may be found among both active and inactive examples in the training set. Experimental studies in antibacterial assays could clarify the selectivity of the compounds' action on particular bacteria. In turn, docking of the studied chemical structures to the targets could clarify molecular mechanism of their action.

Pharmacophore Modelling Study
Taking into account the well obtained antibacterial results against S. aureus of the synthesized compounds of our previews work [34], we selected these molecules for the pharmacophore modelling study in order to search for more active compounds. The structures of the training set (1-9) and test set compounds (10)(11)(12)(13)(14) are displayed in Table 2. In order to evaluate the common features of these compounds, crucial for the antibacterial activity, the LigandScout program was used.                Conformation generation within 20 kcal/mol energy range were generated and submitted to the alignment procedure. Pharmacophore run resulted in the generation of 10 hypothesis models categorized by their rank score and mapping into all training set molecules (Table 3). From all models, model-1 was selected as the best hypothesis for further analysis based on its highest rank score and mapping ( Figure 4). Model-1 contains twelve features, two hydrophobes (H), four hydrogen bond acceptors (HBA), four aromatics (AR) one positive ionizable features (PI) and one hydrogen bond donor (HBD).   All compounds (training and test sets) were mapped onto model-1 with scoring of the orientation of a mapped compound within the hypothesis features using a "fit value" score. As a primary validation of model-1, mapping of all compounds revealed a good correlation between the biological activity and the fit value score ( Table 4). The com-  All compounds (training and test sets) were mapped onto model-1 with scoring of the orientation of a mapped compound within the hypothesis features using a "fit value" score. As a primary validation of model-1, mapping of all compounds revealed a good correlation between the biological activity and the fit value score ( Table 4). The compounds with the higher antibacterial activity showed a range of fit values of 123.59-108.35, while the rest showed a range of fit values of 97.23-85. 10. This primary correlation encouraged us to generate a linear model based on "fit value" in order to predict the antibacterial activity of the understudy compounds. This model (Equation (1), Figure 5) showed good statistics with R 2 = 0.807. Thus, based on these findings, we use this linear model in order to calculate the activity of the tested compounds (5a-5l) ( Table 5). where n is the number of compounds and R is the multiple correlation coefficient. The best aligned poses of the most active compounds 1 and 2 and the less active 14, superposed with model-1, are presented in Figure 6. As is obvious, some structural features play a key role for the activity. The 1H-indole moiety is thought to be critical for activity. Additionally, the absence of hydrophobic groups can partially explain their lack of activity. The other features that are common for all compounds are three HBA features and the positive ionizable interaction of carboxyl group.  The best aligned poses of the most active compounds 1 and 2 and the less active 14, superposed with model-1, are presented in Figure 6. As is obvious, some structural features play a key role for the activity. The 1H-indole moiety is thought to be critical for activity. Additionally, the absence of hydrophobic groups can partially explain their lack of activity. The other features that are common for all compounds are three HBA features and the positive ionizable interaction of carboxyl group.   As we mention above, we use this model in order to predict the antibacterial activity of synthesized compounds 5a-5l. Results are presented in Table 4 and revealed that compounds 5a-c have the highest fit value score and therefore probably the highest activity. Indeed, the experimental data revealed that the predicted by linear model and the experimental MIC of each compound were in the same range (Table 5).  As we mention above, we use this model in order to predict the antibacterial activity of synthesized compounds 5a-5l. Results are presented in Table 4 and revealed that compounds 5a-c have the highest fit value score and therefore probably the highest ac-tivity. Indeed, the experimental data revealed that the predicted by linear model and the experimental MIC of each compound were in the same range (Table 5).

Biological Evaluation 2.4.1. Antibacterial Action
Compounds 5a-5l were tested for antimicrobial activity against the panel of Grampositive, Gram-negative bacteria as well as eight fungal species.
S. aureus was found to be the most sensitive bacteria among all bacteria involved in the study, whereas E. coli was the most resistant one.
All compounds were more potent than ampicillin (MIC at 24.8-74.4 µmol/mL × 10 −2 and MBC at 37.2-124.0 µmol/mL × 10 −2 ) and almost all compounds were more potent than streptomycin. This is true for almost all bacterial species, except for B. cereus and S. typhimurium.
Four the most active compounds were tested against three resistant strains: methicillinresistant S. aureus, MRSA, P. aeruginosa and E. coli ( Table 7). Two of these strains MRSA and P. aeruginosa belong to ESKAPE pathogens (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter species). Tackling the problem of MRSA is a top priority for public health systems worldwide. As far as P. aeruginosa is concerned, the incidences of diseases caused by it are fairly low in the general population, but are higher in hospital inpatients, especially those which are immunocompromised. Table 7. Antibacterial activity against resistant strains (µmol/mL × 10 −2 ) and the effect on biofilm formation (inhibition percentage), * NE-no effect. All compounds appeared to be more potent against MRSA than ampicillin, displaying better bactericidal activity also against resistant strains P. aeruginosa and E. coli than ampicillin, which did not show any bactericidal effect. These compounds were tested also for their effect on biofilm formation. The antibiofilm effect of selected compounds was less promising compared to their potential to inhibit growth of planktonic bacterial cells. Compounds 5b and 5g had a slight effect on biofilm inhibition, 11.59 and 18.19% inhibition, respectively ( Table 7).

Compounds
The structure-activity relationship revealed that the presence of the methyl group in the indole ring as well as the hydroxy group in position 3 of benzene ring (5b) of (Z)-3-(3-hydroxyphenyl)-5-((1-methyl-1H-indol-3-yl)methylene)-2-thioxothiazolidin-4-one is beneficial for antibacterial activity. Displacement of the methyl group by hydrogen in the indole ring and at the same time introduction of the 5-MeO group to the indole ring from one hand and replacement of the 3-OH group in the benzene ring by 3-COOH gave compound 5h, second in the order of activity. On the other hand, removal of methyl as a substituent on the nitrogen and the 5-OMe group of the indole ring resulted in compound 5g, with decreased activity being the third one in the order of activity of compounds. In general, among 5-OMe-indole derivatives more favorable for activity is the presence of 3-COOH (5h) than the 3-OH group (5c) in benzene ring. On the other hand, shifting of the 3-OH group to position 4 of benzene ring was negative, leading to one of the less potent compounds 5f, whereas removal of the 5-MeO group of compounds 5f appeared to be detrimental, resulting in the less active compound 5e. In case of 6-OMe derivatives, the presence of 4-OH, 3-COOH substitution is (5l) beneficial, while the opposite 3-OH, 4-COOH (5d) was negative. On the other hand, substitution in position 4 of the benzene ring with the hydroxy group resulted in compound (5j), with increased activity compared to 5d. From all these compounds mentioned above, it is clear that the antibacterial activity of designed and synthesized compounds depends not only on substituents and their position in the indole ring but also on substituents and their position in the benzene ring.
All the compounds, except 5g, were additive with streptomycin (FICI 1.5, Table 8), implying that efficient combination with this antibiotic might be further developed. Compound 5g was indifferent in combination with streptomycin (FICI 2). Application of the selected compounds has significantly reduced the number of viable P. aeruginosa colonies (Figure 7). After 2 h of application, the number of viable P. aeruginosa colonies was reduced for more than 84% (5b and 5h) and more than 74% (5g). All of the examined compounds, after 6h of treatment, reduced the number of P. aeruginosa CFU by more than 98%. It should be mentioned that even the less active compound 5b appeared to be able to reduce the number of P. aeruginosa CFUs.

Antifungal Action
The ability of compounds to inhibit fungi growth is presented in Table 9 and follows the order 5k > 5l > 5f > 5g > 5a > 5h > 5i > 5c > 5j > 5b > 5d > 5e. The minimum inhibitory concentration of compounds was in the range of 0.97-34.05 mmol/mL ×10 −2 , while the minimal fungicidal concentration (MFC) varied from 1.95 to 68.1 µ mol/mL ×10 −2 . The best an- It should be mentioned that even the less active compound 5b appeared to be able to reduce the number of P. aeruginosa CFUs.
Our results indicate that the described compounds are 13-52 times more potent than ketoconazole and 6-to 17-fold more potent than bifonazole.
According to the study of structure-activity relationships, it seems that the presence of 3-COOH, 3-OH groups on the benzene ring of indole derivatives (5k) is beneficial for antifungal activity, whereas attachment of the 6-methoxy group to the indole ring of compound 5k results in the less potent derivative 5l.
The order of activity of indole derivatives can be presented as 5k > 5g > 5a > 5i > 5e. Thus, the combination of 3-COOH, 4-OH groups attached to benzene ring (5k) leads to the increase in activity of indole derivatives, while the opposite (5a) is less favorable. Additionally, the 3-COOH group on the benzene ring produced a positive effect on\ antifungal activity of indole derivatives. In the case of 5-methoxy indole derivatives, the 4-OH substitution (5f) was the most favorable, while 3-COOH (5h) and 3-OH substitution (5c) were much less favorable. Concerning the 6-methoxy indole derivatives: the combination of 3-COOH with 4-OH substituents in the benzene ring is favorable, whereas the combination of 3-OH and 4-COOH produced a negative effect on antifungal activity. Thus, the antifungal activity as, in case of antibacterial compounds, depends not only on the nature of substituents, but also on their relative positions.
Our results indicate that antibacterial and antifungal activities have different relationships with the chemical structure: the compound that showed the best antibacterial activity (5b) appeared to be one of the least active as antifungals. On the other hand, compounds 5g, 5d and 5e expressed almost the same behavior against bacteria and fungi.

Docking to Antibacterial Targets
In order to elucidate the probable mechanism of action of designed compound docking studies to several antibacterial targets, such as E. coli DNA Gyrase, S. aureus Thymidylate kinase, E. coli Primase and E. coli MurA and MurB, were performed. The docking studies showed that the assessments of binding free energy to E. coli DNA Gyrase, Thymidylate kinase, E. coli Primase and E. coli MurA were higher than that to E. coli MurB; therefore, it may be resolved that the inhibition of E. coli MurB enzyme is the most probable, among the considered mechanism of action of the compounds, where binding scores were consistent with biological activity (Table 10).
The most active compound 5b in E. coli MurB enzyme showed three favorable hydrogen bond interactions. Two hydrogen bonds were observed between the hydroxyl substituent of the benzene ring of the compound and the residues Tyr157 and Lys261 (distance 1.72 Å and 2.58 Å, respectively), and another hydrogen bond interaction between the O atom of the carbonic group of the thiazolidine ring of the compound and Ser228 (distance 2.65 Å). The benzene ring interacts hydrophobically with the residues Tyr124, Gly122, Asn232 and Arg158, while the thizolidine ring interacts hydrophobically with the residues Tyr189 and Leu289 (Figures 8 and 9). These interactions stabilize the complex compound enzyme and play a crucial role in the increased inhibitory action of the compound 5b. Moreover, the hydrogen bond formation with the residue Ser228 is crucial for the inhibitory action of this compound, because this residue takes part in the proton transfer at the second stage of peptidoglycan synthesis [41]. Hydrogen bond interactions with the residue Ser228 were also observed for the rest of the studied compounds.

Docking to Antifungal Targets
All the synthesized compounds and the reference drug ketoconazole were d lanosterol 14a-demethylase of C. albicans and DNA topoisomerase IV (Table 11).

Docking to Antifungal Targets
All the synthesized compounds and the reference drug ketoconazole were docked to lanosterol 14a-demethylase of C. albicans and DNA topoisomerase IV (Table 11). Docking results showed that the most active compound 5k take place inside the active site of the enzyme interacting with the heme group of CYP51 Ca throughout its -COOH substituent of the benzene ring forming negative ionizable interactions with the heme group. Furthermore, the oxygen of the -OH substituent interacts with the Fe iron of the heme group and with the N atom of Heme, forming a hydrogen bond. Another hydrogen bond is formed between the N atom of indole moiety and Ser378. Hydrophobic interactions were also detected with the residues Thr311, Ley376, Phe233, Phe380 and Met308. Interaction with the heme group was also observed with the benzene ring of ketoconazole, which forms hydrophobic and aromatic interactions (Figures 10 and 11). However, compound 5k forms more and stronger interactions than ketoconazole and a more stable complex of the ligand with the enzyme. This is probably the reason why compound 5k has better antifungal activity than ketoconazole.
Docking results showed that the most active compound 5k take place inside the active site of the enzyme interacting with the heme group of CYP51Ca throughout its -COOH substituent of the benzene ring forming negative ionizable interactions with the heme group. Furthermore, the oxygen of the -OH substituent interacts with the Fe iron of the heme group and with the N atom of Heme, forming a hydrogen bond. Another hydrogen bond is formed between the N atom of indole moiety and Ser378. Hydrophobic interactions were also detected with the residues Thr311, Ley376, Phe233, Phe380 and Met308. Interaction with the heme group was also observed with the benzene ring of ketoconazole, which forms hydrophobic and aromatic interactions (Figures 10 and 11). However, compound 5k forms more and stronger interactions than ketoconazole and a more stable complex of the ligand with the enzyme. This is probably the reason why compound 5k has better antifungal activity than ketoconazole.

Cytotoxicity Assessment
The assessment of cellular cytotoxicity of the compounds in normal human MRC cells was evaluated at two concentrations in culture, i.e., 1 ×10 −5 M (Figure 12A,B) and ×10 −6 M ( Figure 12B, C). No substantial effect on cell proliferation after 48 h exposure h been observed in cultures, since the growth was ≥80% for all the tested agents compar to control untreated cultures ( Figure 12A,C). Moreover, the percentage of dead cells acc mulated in cultures was very low, since the maximum number did not exceed that of 2.5% ( Figure 12B,D).

Cytotoxicity Assessment
The assessment of cellular cytotoxicity of the compounds in normal human MRC-5 cells was evaluated at two concentrations in culture, i.e., 1 × 10 −5 M (Figure 12A,B) and 1 × 10 −6 M ( Figure 12B,C). No substantial effect on cell proliferation after 48 h exposure has been observed in cultures, since the growth was ≥80% for all the tested agents compared to control untreated cultures ( Figure 12A,C). Moreover, the percentage of dead cells accumulated in cultures was very low, since the maximum number did not exceed that of 2-2.5% ( Figure 12B,D).
The assessment of cellular cytotoxicity of the compounds in normal human MRC-5 cells was evaluated at two concentrations in culture, i.e., 1 ×10 −5 M (Figure 12A,B) and 1 ×10 −6 M ( Figure 12B, C). No substantial effect on cell proliferation after 48 h exposure has been observed in cultures, since the growth was ≥80% for all the tested agents compared to control untreated cultures ( Figure 12A,C). Moreover, the percentage of dead cells accumulated in cultures was very low, since the maximum number did not exceed that of 2-2.5% ( Figure 12B,D).  (panels A,C). Moreover, the evaluation of cell death was assessed by the trypan-blue exclusion-dye assay, as shown in " Methods" (panels B,D). The results shown above indicate the mean numbers ± SD of two independent biological experiments (n ≥ 3). The diagrams shown above and the t-test statistical analysis were carried out using the GraphPad Prism 6.0 program. Notably, no statistical significance between the control-untreated culture with each one of treated compounds was obtained.

PASS and PASS-Based Web Applications
PASS predictions are based on the structure-activity relationships derived from the data on over eight thousand biological activities of more than one million molecules included in the training set [37,[42][43][44]. Structure-activity relationships are examined using MNA (Multilevel Neighborhoods of Atoms) structural descriptors [45] and modified the Naïve Bayes approach [46]. Structural formulae presented as MDL MOL or SDF files [47] are used as input information. PASS output is the list of predicted activities with two assessments: Pa is the estimate of the probability of belonging to the class of active compounds, and Pi is the estimate of the probability of belonging to the class of inactive ones [46]. The higher the probability difference Pa-Pi, the higher the chance to confirm the prediction in the following experiment.
Since 1999 [48], the PASS Online web application has been freely available via the Internet. It provides an opportunity to predict several thousand biological activities with an average probability of about 95% [39]. Comparing PASS Online with some other freely available web services predicting biological activity profiles demonstrated its superiority in performance [49]. Using the special training sets created based on ChEMBL data [50] we developed several specialized PASS-based web applications: AntiBac-Pred [38,51], AntiFun-Pred [52], KinScreen [53,54], which predict the detailed antibacterial, antifungal and kinase inhibitory activity profiles, respectively. These web applications differ from the standard PASS version only by the training sets focused on the particular pharmacotherapeutic fields. Therefore, the interpretation of the prediction results is the same as described above.

Ligand-Based Pharmacophore Modeling
The LigandScout program (Advanced version 4.4.7) [55] with default settings was used to perform the pharmacophore modeling studies. Prior to the generation of pharmacophore hypotheses, all tested dataset compounds were built using ChemDraw Ultra (CambridgeSoft, version 12.0) and converted into the 3D format. The lowest energy conformations were generated using MMFF94 (Merck molecular force field) and the BEST conformation model generation method was used during conformer generation with a maximum number of 250 conformations, an energy threshold value of 20 kcal/mol above the global energy minimum and an RMS threshold of 0.8.
Compounds 1-9 were selected for training. The training set molecules play a key role in determining the quality of the pharmacophore models generated, while the test set compounds serve to validate the resultant pharmacophore solutions.
Based on the feature mapping results, five matching features were selected, including hydrophobic features (H, yellow), aromatic rings (AR, blue), hydrogen bond donors (HBD, green), hydrogen bond acceptors (HBA, red) and negative ionizable features (red star). The quality of generated hypotheses was ranked based on the pharmacophore fit score, which indicates the modality of the mapping between a molecule and a model. A value of 1 reflects the best prediction [56]. The highest rank score hypotheses for antibacterial activity were considered statistically the best hypotheses and selected for the further analysis.

Pharmacophore Validation
The generated pharmacophore hypothesis was validated using a test set and leaveone-out methods.

Pharmacophore Validation Using Test Set
The test set method is used to clarify whether the generated pharmacophore model is capable to predict the antibacterial activity of compounds other than the training set compounds and categorize them properly in their activity scale. For the test set, compounds 10-14 were selected. For the test set compounds, the conformation generation was performed using default values and BEST conformation analysis algorithms [57,58].
Pharmacophore Validation Using Leave-One-Out The pharmacophore model was cross-validated by the leave-one-out method. In this method, pharmacophore models are recomputing again by leaving one compound at a time from the training set compounds, until each compound was left out once, and its affinity is predicted using that new model [59]. This validation is performed to verify that the correlation of the original pharmacophore model does not depend only on one particular compound [57,60]. By leaving each one of the 9 training set compounds, 9 new models were generated. Thus, we did not obtain any meaningful differences between Model-1 and each model generated from this method, validating our pharmacophore model. The minimum inhibitory (MIC) and minimum bactericidal (MBC) concentrations were determined by the modified microdilution method, as previously reported [34,60].
Resistant strains used in microdilution assay were isolates of S. aureus (strain isolated from cow), E. coli (strain isolated form pig) and P. aeruginosa (strain isolated from cat) obtained as described in Kartsev et al. [61].

Inhibition of Biofilm Formation
The method was performed as described [62] with some modifications. Briefly, the P. aeruginosa resistant strain was incubated with MIC and subMIC of tested compounds in Triptic soy broth enriched with 2% glucose at 37 • C for 24 h. After 24 h, each well was washed twice with sterile PBS (phosphate buffered saline, pH 7.4) and fixed with methanol for 10 min. Methanol was then removed and the plate was air dried. Biofilm was stained with 0.1% crystal violet (Bio-Merieux, Lyon, France) for 30 min. Wells were washed with water, air dried and 100 µL of 96% ethanol (Zorka, Serbia) was added. The absorbance was read at 620 nm on a Multiskan™ FC Microplate Photometer, Thermo Scientific™. The percentage of inhibition of biofilm formation was calculated by the formula: [(A 620 control − A 620 sample)/A 620 control] × 100.

Checkboard Assay
It was carried out with 96-well microplates containing TSB medium for the resistant P. aeruginosa strain, supplemented with examined compounds in concentrations ranging from 1/16 to 4 × MIC as described previously [60] in the checkboard manner. The fractional inhibitory concentration index (FICI) was calculated by the following equation as described in our previous paper [63]: FICI = FIC1 0 /MIC1 0 + FIC2 0 /MIC2 0 FIC1 0 and FIC2 0 are the MICs of a combination of tested compounds and antibiotics, and MIC1 0 and MIC2 0 , represent the MIC values of individual agents. The following cut-offs: FIC ≤ 0.5 synergistic, >0.5 <2 additive, ≥2 <4 indifferent, and FIC > 4 antagonistic effects were used for the discussion of obtained results.

Docking Studies
The AutoDock 4.2 ® software was used for the docking stimulation. The free energy of binding (∆G) of E. coli DNA GyrB, Thymidylate kinase, E. coli MurA, E. coli primase, E. coli MurB, DNA topoIV and CYP51 of C. albicans in complex with the inhibitors were generated using this molecular docking program. Regarding the X-ray crystal structures, data of all the enzymes used were obtained from the Protein Data Bank (PDB ID: 1KZN, AQGG, 1DDE, JV4T, 2Q85, 1S16 and 5V5Z, respectively). All procedures were performed according to our previous paper [66].

Assessment of Cytotoxicity
The normal human lung fibroblast MRC-5 cell line is stored and used in our laboratory in a routine manner (passage < 40). MRC-5 cells were grown in culture (37 • C, humidified atmosphere containing 5% v/v CO 2 ) in DMEM medium supplemented with 10% v/v FBS, 1% PS penicillin-streptomycin). The compounds tested were dissolved in DMSO and stored in 4 • C. For the assessment of cytotoxicity, the cells were seeded in a 96-well plate at an initial concentration of 5 × 10 4 cells/mL and allowed to attach for at least 3h before the addition of the compounds at two different concentrations: 1 × 10 −5 M (10 µM) and 1 × 10 −6 M (1 µM). Note that the concentration of DMSO in culture was ≤0.2% v/v, in which no detectable effect on cell proliferation is observed [67]. To assess the cytotoxicity of each compound, the cells were allowed to grow for additional 48 h before their number is estimated in culture using the Neubauer counting chamber under an optical microscope. Cell growth in each treated culture is expressed as the percentage compared to that seen for the untreated control cells. Moreover, the number of dead cells was also measured using the Trypan-blue method, as previously described [65][66][67]. Statistical t-test analysis was performed via the use of GraphPad Prism 6.0 program.
It should be mentioned that all compounds appeared to be more potent than ampicillin against all bacteria tested and streptomycin against all bacteria except B. cereus, and En. Cloacae. The most sensitive bacteria were found to be S. aureus, while L. monocytogenes was the most resistant one. Compounds also appeared to be active against three resistant strains MRSA, E. coli and P. aeruginosa, showing better activity against MRSA than both reference drugs, while showing better activity against the other two resistant strains than ampicillin.
Concerning antifungal action, the tested compounds exhibited very good activity against all the fungal species tested, being more active than ketoconazole and bifonazole. The most sensitive fungal strain appeared to be T. viride, while the most resistant filamentous A. fumigatus.
It can be observed that the growth of both Gram-negative and Gram-positive bacteria and fungi responded differently to the tested compounds, which indicates that different substituents may lead to different modes of action or that the metabolism of some bacteria/fungi was better able to overcome the effect of the compounds or adapt to it.
Docking analysis to DNA Gyrase, Thymidylate kinase and E. coli MurB indicated a probable involvement of MurB inhibition in the antibacterial mechanism of compounds tested while docking analysis to 14α-lanosterol demethylase (CYP51) and tetrahydrofolate reductase of Candida albicans indicated a probable implication of CYP51 reductase at the antifungal activity of the compounds and secondary involvement of dihydrofolate reductase inhibition at the mechanism of action of the most active compounds.