Design, synthesis and biological evaluation of new substituted 5-benzylideno-2-adamantylthiazol[3,2-b][1,2,4]triazol-6(5H)ones. Pharmacophore models for antifungal activity

2.1 Chemistry

Τhe synthesis of title compounds was performed by a multistep reaction as shown in Scheme 1. Adamantane thiosemicarbizide (3) was synthesized using a procedure reported earlier starting from adamantine-1-carbonyl chloride (1) upon reaction with thiosemicarbazide (2), followed by cyclization in alkaline solution under reflux to 5-adamantyl-4H-1,2,4-trizol-3-thiole (3). The third step includes the one pot condensation of 5-adamantyl-4H-1,2,4-triazol-3-thiole (4) with bromoacetic acid and appropriate substituted benzaldehydes in the presence of sodium acetate and acetic anhydride (Karthikeyan et al., 2008). Reactions proceed smoothly with good yields (55–88%).

All new structures of compounds 1–25 were characterized by IR, ¹H NMR and elemental analysis. IR spectra showed absorptions at 1724–1747 cm⁻¹ (C⚌O) and at 1578–1654 cm⁻¹ (C⚌N). In the ¹H NMR spectra the title compounds showed peaks in the region of 1.75–2.36 ppm (adamantine), 7.12–7.80 ppm (Ar—H) and 8.11–8.48 ppm (CH⚌).

During the reaction of 5-adamantyl-4H-1,2,4-trizol-3-thiole with different dielectrophiles the formation of two cyclic isomers, (a) thiazolo[3,2-b]-1,2,4-triazole and (b) thiazolo[3,2-c]-1,2,4-triazole is possible (Karthikeyan, 2008) (Scheme 2).

Close

Scheme 2

Formation of thiazolo[3,2-b]-1,2,4-triazole.

Export to PPT

Compounds 1–25 exist as potential E and Z geometrical isomers; the Z conformation of the 5 exocyclic C⚌C double bond was assigned on the basis of literature data for analogues structures (Ottanà et al., 2005) as well as on experimental data (¹H NMR) by comparing the resonance region of the hydrogens of the 5-adamantyl-4H-1,2,4-triazole-3-thiole to the corresponding ones of the final compounds (Scheme 2). The adamantane protons of thiol group resonated at 1.70–2.51 ppm, while the adamantane protons of the title compounds resonated at 1.75–2.36 ppm. It is obvious that both the title and intermediate compounds showed peaks for adamantane hydrogens in the same area, confirming the formation of the Z isomer thiazolo[3,2-b]-1,2,4-triazole. On the contrary, in case of the formation of E isomer, the aromatic ring protons will appear at higher chemical shift values, deshielded by the adjacent C⚌O.

2.2 Antimicrobial activity

The results of antibacterial activity of the compounds 1–25 are presented in Table 1. Almost all the tested compounds showed antibacterial activity but on different level. Compounds 15–25 showed higher antibacterial potential than others with MIC ranged between 4.5–26.4 ∗ 10⁻² μmol/mL and MBC 9.0–42.2 ∗ 10⁻² μmol/mL. The majority of the compounds were inactive against Micrococcus flavus. Compound 15, did not show activity at tested concentration against M. flavus, Staphylococcus aureus, Pseudomonas aeruginosa, Salmonella typhimurium and E. faecalis, 19 and 25 on P. aerugunisa and E. faecalis, 22 on Bacillus cereus, Escherichia coli and E. faecalis, while 21 on E. coli and E. faecalis. The antibacterial potential of tested compounds could be presented as follows: 18 > 24 > 23 > 17 > 16 > 19 > 20 > 25 > 21 > 22 > 15. It can be seen that compound 18 showed the best antibacterial activity with MIC in the interval 4.90–9.80 ∗ 10⁻² μmol/mL and MBC 9.80–36.6 ∗ 10⁻² μmol/mL. This compound is followed by 24 with MIC at 4.5–13.6 ∗ 10⁻² μmol/mL and MBC between 9.0–27.1 ∗ 10⁻² μmol/mL. The lowest antibacterial activity among all tested compounds was obtained for compound 15 with inhibitory activity at 5.3–26.4 ∗ 10⁻² μmol/mL and bactericidal effect at .10.5- > 42.2 ∗ 10⁻² μmol/mL. Ampicillin showed inhibitory effect at 24.8–74.4 ∗ 10⁻² μmol/mL and bactericidal at 37.2–124.0 ∗ 10⁻² μmol/mL, while Streptomycin showed MIC in range of 4.3–25.8 · 10⁻² μmol/mL and MBC of 8.6–51.6 · 10⁻² μmol/mL. It can be seen that all the compounds tested for antibacterial activity showed better effect than the commercial antibiotic ampicillin with exception of those compounds that did not show any activity in tested concentrations. On the other hand, all the compounds tested exhibited even better antibacterial activity than Streptomycin against Listeria monocytogenes and E. coli (except for 22 and 21).

The results of antifungal activity of tested compounds against eight fungi are presented in Table 2. It can be seen that all compounds exhibited antifungal effect. MIC is in range of 3.67–34.6 ∗ 10⁻² μmol/mL and MFC in 7.35–39.6 ∗ 10⁻² μmol/mL. The antifungal potential could be presented as: 18 > 24 > 19 > 20 > 23 > 17 > 25 > 16 > 22 > 21 > 15. Compound 18 again showed the best antifungal activity among other tested compounds with MIC at 3.67–9.80 ∗ 10⁻² μmol/mL and MFC at 7.35–19.6 ∗ 10⁻² μmol/mL. Compound 24 is the next one with strong antifungal activity with MIC at 11.3–22.6 ∗ 10⁻² μmol/mL and MFC 11.3–31.6 ∗ 10⁻² μmol/mL, while compound 15 possessed the lowest antifungal potential with inhibitory activity at 13.2–21.1 ∗ 10⁻² μmol/mL and fungicidal activity at 21.1–39.6 ∗ 10⁻² μmol/mL. The majority of the compounds presented the best activity against Aspergillus ochraceus, Aspergillus versicolor and Aspergillus fumigatus, while Candida albicans was the most resistant species to the compounds. The commercial antifungal agent, bifonazole, showed MIC at 32.0–64.0 ∗ 10⁻² μmol/mL and MFC at 48.0–80.0 ∗ 10⁻² μmol/mL. Ketoconazole showed fungistatic activity at 38.0–475.0 ∗ 10⁻² μmol/mL and fungicidal effect at 95.0–570.0 ∗ 10⁻² μmol/mL. All the tested compounds showed better fungistatic effect than bifonazole and ketoconazole. Some of the compounds did not exhibit fungicidal activity in tested concentration against Aspergillus niger, Trichoderma and C. albicans. Compound 18 showed the best effect against bacteria and fungi, while compound 2 exhibited the worst activity against all the tested microorganisms.

It was observed that compounds that exhibited the lowest antibacterial activity are 1–9 (Table 1). The antibacterial potential of this group could be presented as follows: 4 > 8 > 9 > 7 > 5 > 3 > 6 > 1 and 2. The majority of compounds did not affect some bacteria, such as M. flavus, E. coli, and Enterobacter cloacae. The compound 4 with highest antibacterial activity showed inhibitory effect at 98.0–490.0 ∗ 10⁻² μmol/mL bactericidal at 490.0 ∗ 10⁻² μmol/mL. Compound 2 inhibited bacterial growth with lowest potential at 100.72–377.7 ∗ 10⁻² μmol/mL, while bactericidal effect was achieved at 214.4–805-76 ∗ 10⁻² μmol/mL. This compound showed the lowest antibacterial activity among all the others. Streptomycin and Ampicillin showed better antibacterial activity than this group of compounds. Antifungal potential of compounds from this group could be presented as follows: 5 > 4 > 2 > 3 > 8 > 1 > 7 > 9 > 6 (Table 2). Compound 5 showed the best antifungal potential with inhibitory concentration at 6.15–24.6 ∗ 10⁻² μmol/mL and fungicidal at 24.6–98.4 ∗ 10⁻² μmol/mL. Compound 4 showed also very good antifungal potential with inhibitory concentration at 6.12–24.5 ∗ 10⁻² μmol/mL and fungicidal at 49.0–171.5 ∗ 10⁻² μmol/mL. The lowest antifungal effect was achieved for compound 6 with MIC at 6.3–75.9 ∗ 10⁻² μmol/mL and MFC at 75.9–177.1 ∗ 10⁻² μmol/mL. Only compounds 7, 8, 9, 6, 2, and 4 exhibited antifungal potential against C. albicans. The majority of compounds showed higher antifungal potential than commercial fungicides (Table 2).

The next group of compounds are compounds 10–14 (Table 1). They showed higher antimicrobial activity than previous one. The best activity in this group is seen for compounds 13 and 14 with MIC at18.6–92.8 ∗ 10⁻² μmol/mL × 10⁻² and 23.2–92.8 ∗ 10⁻² μmol/mL × 10⁻² and MBC at 946.4–185.6 ∗ 10⁻² μmol/mL. Compounds 10 and 11 showed slightly higher antibacterial activity than previous one, while 12 was the less active with MIC and MBC values (125.9–377.7 · 10⁻² μmol/mL and 200.8–503.6 · 10⁻² μmol/ml ∗ 10⁻²). Antifungal activity for these compounds presented in Table 2 could be presented as follows: 10 > 14 > 13 > 11 > 12. The best antifungal activity is seen for compound 10 which possessed MIC at 6.1–170.8 ∗ 10⁻² μmol/mL and MFC at 24.4–97.6 ∗ 10⁻² μmol/mL. Compound 12 as in the case of antibacterial activity showed the worst antifungal potential with MIC at 6.28–175.7 ∗ 10⁻² μmol/mL and MFC at 25.1–175.7 ∗ 10⁻² μmol/mL. Almost all compounds in this group showed better antifungal potential than ketoconazole, with exception of compounds 11 and 12 against A. fumigatus, compounds 10, 12, 13, 14 against A. niger and C. albicans. Bifonazole showed lower antifungal effect than compounds from this group toward all fungi, except A. fumigatus, A. niger and C. albicans.

It can be seen that the third group of compounds 15–25 possessed the highest antibacterial and antifungal potential and did not affect E. faecalis at all, and only 22 against C. albicans. Second group of compounds (10–14) showed lower activity than previous one, and did not exhibit potential on E. cloacae and C. albicans. The lowest antimicrobial potential could be seen for third group of compounds 1–9 which showed high MIC and MBC/MFC. These compounds did not exhibit activity toward E. coli and E. cloacae, and C. albicans from fungi. It is obvious that all three groups of compounds showed at first different level of antibacterial and antifungal activity and then they were inactive against different bacterial and fungal species. The reason of such activity could be because of different mechanisms of action of different derivatives influenced by different chemical groups.

2.3 Elucidating the relation of molecular properties to antimicrobial activity

Physically significant descriptors and pharmaceutically relevant properties were predicted using the QikProp software. Then multivariate data analysis was implemented to elucidate the relation between molecular properties to antimicrobial and antifungal activities of the synthesized compounds. Specifically, we based our approach on the “a priori” knowledge, that compound 18 is the optimum against most of the tested strains as it yields the highest antibacterial and antifungal activities. Therefore, the objective was to determine the physically significant descriptors and pharmaceutically relevant properties of any other synthesized compound that resembles its antibacterial and antifungal fingerprints.

For this purpose a PCA model was extracted in order to discern from the scores plot the synthesized compounds that localize near compound 18 and identify from the loading plot the molecular properties that mainly characterize them. The PCA model (A = 2, N = 25, R²X(cum) = 0.92, Q²(cum) = 0.83) was extracted based only on the physicochemical properties of the synthesized compounds (Fig. 1). The formation of two groups is evident along the PC1 principal component. Compound 18 belongs to Group 1 together with 17, 5, 10, 11, 4, 7, 9,19, 8, 6, 16 and 17. This suggests that these molecules probably exhibit similarly potent biological activities explained by some common descriptors. The loading plot highlights the descriptors which contribute to the clustering between the two groups. From these, the most significant are the (a) number of non-trivial (not CX3), non-hindered (not alkene, amide, small ring) rotatable bonds (#rotor), (b) total solvent accessible surface area (SASA) in square angstroms using a probe with a 1.4 A radius, (c) hydrophobic component of the SASA(FOSA), (d) hydrophilic component of the SASA (FISA), (e) total solvent-accessible volume in cubic angstroms using a probe with a 1.4 Å radius (volume), (f) estimated average number of hydrogen bonds (taken over a number of configurations) that would be donated by the solute to water molecules in an aqueous solution (HBd), (g) estimated average number of hydrogen bonds (taken over a number of configurations)that would be accepted by the solute from water molecules in an aqueous solution (HBa), (h) hexadecane/gas partition coefficient (QPlogPC16), (i) octanol/gas partition coefficient (QPlogPoct), (j) water/gas partition coefficient (QPlogPw), (k) predicted brain/blood partition coefficient (QPlogBB), (l) predicted skin permeability (QPlogKp), (m) number of likely metabolic reactions (#metab), (n) Van der Waals surface area of polar nitrogen and oxygen atoms and carbonyl carbon atoms (PSA), (o) number of nitrogen and oxygen atoms (#NandO), (p) number of heavy atoms (#nonHatm). These produced values can be further used as descriptors for QSAR and in silico screening techniques.

Close

Figure 1

PCA model A = 2, N = 25, R²X(cum) = 0.92, Q²(cum) = 0.83. A. Scores scatter plot depicting a clear separation into two groups (Group1: circles; Group 2: squares) B. Loading scatter plot displaying the variables that characterize each group.

Export to PPT

2.4 Generation of pharmacophore model for antifungal activity

The generation of the pharmacophore model using a training set of six compounds (8, 16, 18, 19, 23 and 25) resulted in top ten hypotheses with individual ranking scores and pharmacophore features as depicted in Table 3 for A. fumigatus and C. albicans.

It was interesting to note that the top two hypothesis (Hypo1 and Hypo2) in both species (A. fumigatus & C. albicans) had exactly the same pharmacophoric features (RHHAAA) suggesting possibly the same target in both these species. Both these hypothesis (hypo1 and Hypo2) in both species (A. fumigatus & C. albicans) also showed same ranking score of 94.84 and were constituted by two HAl(HAl1 and HAl2), three HBA(HBA1, HBA2 and HBA3) and one (RA) features. All the ten generated hypotheses in case of A. fumigatus were found to be uniform regarding the composition of chemical features. However in case of C. albicans the first five hypotheses were found to be uniform in terms of pharmacophore feature composition. The Hypo1 for both A. fumigatus and C. albicans having same chemical feature composition (RHHAAA) showed similar chemical feature mapping for the most active compound 18. The preliminary pharmacophore models in both the species (A. fumigatus and C. albicans) were further refined by addition of excluded volumes for identification of the least active compounds in the series properly. As an optimization strategy the HAl1 feature was also modified to have a location sphere radius of 2.5 Å in both cases (A. fumigatus and C. albicans) to make improvements in prediction. The location sphere radius for the excluded volumes were optimized and reduced to lower values so as to assign better fit values for the active compounds and lower fit values for the least active compounds in the series. In both CFPM compound 18 (Fig. 2a & c) was observed to map the HAl1 with the 3-methoxy group attached to the phenyl ring while the adamantyl group present at the other terminal mapped the HAl2 feature. The threes HBA features generated mapped the thiazole sulfur atom (HBA1), the ketooxygen atom (HBA2) and the triazole nitrogen atom (HBA3) while the RA feature generated mapped the phenyl ring of compound 18 (Fig. 2a & c). The distance matrix of pharmacophore features for Hypo1 for both A. fumigatus and C. albicans were found to be the same and has been presented in supporting information (Table 3). The least active compound 6 in the series of A. fumigatus mapped all the features of the pharmacophore (CFPM of A. fumigatus) with the same functional groups as observed for compound 18 except HAl1 as the compound lacks a hydrophobic feature at this position (Fig. 2b). A similar mismatch with the HAl1 was also observed for the least active compound 2 in the series for C. albicans (CFPM of C. albicans) (Fig. 2d). The pharmacophore fit values (Supplementary information: Table 2) obtained from the CFPM were further correlated with biological activity (pMIC) of the compounds for each species. The generated CFPM for A. fumigatus was found to have a correlation value (R) of 0.51(n = 25) between observed activity (pMIC) and pharmacophore fit values that improved to 0.74 (n = 23) when two outliers (5, 23) were removed from the series (Fig. 2e). In case of C. albicans similar improvement (n = 18, R = 0.48; n = 16, R = 0.74) (Fig. 2f) was observed when two outliers (20, 23) were removed from the series that indicate the generated pharmacophore models have good ability to predict the individual antifungal activity in these series of compounds.

Close

Figure 2

(a). Mapping of the most active compound 18 with the CFPM of A. fumigatus. (b) Mapping of the least active compound 6 with the CFPM of A. fumigatus. (c) Mapping of the most active compound 18 with the CFPM of C. albicans. (d) Mapping of the least active compound 18 with the CFPM of C. albicans. (e) Graph displaying correlation between observed activity (pMIC) and pharmacophore fit value for (e) A. fumigatus (f) C. albicans. All the excluded volumes have location sphere of radius of 0.8 Å except those mentioned in the figure. The HAl1 has a location sphere of radius 2.5 Å.

Export to PPT

4.1 Chemistry-general aspects

Melting points (°C) were determined with a MELTEMP II capillary apparatus (LAB Devices, Holliston, MA, USA) without correction. Elemental analyses were performed on a Perkin–Elmer 2400 CHN elemental analyzer and for all compounds synthesized were within a 0.4% of theoretical values. IR spectra were recorded, as Nujol mulls, on a Perkin Elmer Spectrum BX. Wave numbers in the IR spectra are given in cm⁻¹. ¹H NMR and ¹³C NMR spectra of the newly synthesized compounds, in DMSO-d₆ or CDCl₃ solutions, were recorded on a Bruker AC 300 instrument (Bruker, Karlsruhe, Germany) at 298 K. Chemical shifts are reported as δ (ppm) relative to TMS as internal standard. Coupling constants J are expressed in Hertz (Center of Instrumental Analysis of the University of Thessaloniki). The reactions were monitored by TLC on F₂₅₄ silica-gel precoated sheets (Merck, Darmstadt, Germany) and each of the purified compounds showed a single spot. Solvents, unless otherwise specified were of analytical reagent grade or of the highest quality commercially available. Synthetic starting materials, reagents and solvents were purchased from Aldrich Chemie (Steinheimm, Germany).

4.2 Synthesis of 1-(1-adamantanocarbonylo)thiosemicarbazide (Tozkoparan et al., 2000)

To a solution of thiosemicarbazide (0.1 mol, 9.1 g) in dry pyridine (150 ml), a solution of 1-adamantylcrabonyl chloride (0.1 mol, 19.869 g) in dry benzene (150 ml) under stirring and temperature at −5 °C was added. The reaction mixture was stirred for another 0.5 h at −5 °C and after left at room temperature for 12 h. After removal of solvents, to the remain solid water was added and precipitate was filtered and recrystallized from methanol: dichloromethane. Yield: 70%.

4.3 Synthesis of 5-adamantyl-4H-1,2,4-triazol-3-thiol ((Tozkoparan et al., 2000)

To a solution of NaOH 5% (5 ml) 1-(1-adamantanocarbonylo)thiosemicarbazide (0.01 mol) was added and reaction mixture was refluxed for 2 h. After cooling to reaction mixture HCl 10% was added till pH = 6. The precipitate washed with water and filtered. Dry product recrystallized from methanol. Yield, 93%.IR: (cm⁻¹, Nujol): 3554 (NH), 2670 (SH), 1579 (C⚌N). ¹H NMR: (δ ppm, DMSO-d₆, 300 MHz): 1.70–2.51 (m, 15H, adamantane).

4.4 General method for preparation of 5-aryliden -2-adamatylthiazol[3,3-b]triazol-6(5H)-ones (Vicini et al., 2008)

To a solution of 5-adamantyl-4H-1,2,4-triazol-3-thiol (4 mol) in acetic acid (10 ml), bromoacetic acid (6 mol), appropriate aromatic aldehyde (4 mol), sodium acetate (0.54 gr, 6.58 mmol) and acetic anhydride (8 ml) were added. The reaction mixture was refluxed under stirring for 6 h. The obtained product left for some hours at room temperature and after pawed to the ice. The obtained precipitate filtered and recrystallized from dioxane.

4.5 Biological evaluation

4.6 Prediction of molecular properties

A next step was to proceed with the prediction of physically significant descriptors and pharmaceutically relevant properties of the synthesized molecules. The QikProp (Maestro 9.8) as a fast and accurate prediction software was used.

4.7 Multivariate data analysis

The aforementioned molecular properties in relation to their antifungal and antibacterial activities constituted the dataset that was further subjected to multivariate data analysis. In particular the SIMCA-P version 13.0 software (Umetrics, Umeå, Sweden) was utilized in order to implement Principal Component Analysis (PCA).

A PCA model estimates the systematic variation in a data matrix by a low dimensional model plane. The unsupervised PCA pattern recognition method manages to extract the dominant patterns in the data matrix in terms of a complementary set of scores and loading plots, thus enabling a reduction of dimensionality, a data exploration finding relationship between objects, an estimation of the correlation structure of the variables and investigation on the necessary components (a linear combination of original features) to explain the greater part of variance with a minimum loss of information (Trygg et al., 2007).

The PCA model were UV scaled, extracted at a confidence level of 95% and any observations at <5% were considered to be outliers.

Loading plots were extracted to reveal the variables that bear class discriminating power.

The quality of models was described by the goodness-of-fit R² (0 ⩽ R² ⩽ 1) and the predictive ability Q² (0 ⩽ Q² ⩽ 1) values. The R² explains the variation, thus constitutes a quantitative measure of how well the data of the training set is mathematically reproduced. The overall predictive ability of the model is assessed by the cumulative Q² representing the fraction of the variation of Y that can be predicted by the model and was extracted according to the internal cross validation default method of SIMCA-P software (Eriksson et al., 2006).

In particular, all models demonstrated high statistical values (R² > 0.7 and Q² ⩾ 0.6), the difference between the goodness-of-fit and the predictive ability remained always lower than 0.2 (R²X(cum) – Q²(cum) < 0.2) and the goodness-of-fit never equalled to one (R²X(cum) ≠ 1). Therefore, since the extracted models abide by these rules then their robustness and predictive response are enhanced (Geladi et al., 1989).

5.1 Selection of training set

In order to generate pharmacophore models six compounds 8, 16, 18, 19, 23 and 25 having promising inhibitory activity (IC₅₀ < 35 μM) toward Aspergillus fumigatus and C. albicans were selected from the dataset of 25 molecules and 18 molecules respectively and rest of the molecules were kept for extending the pharmacophore model. As observed, the dataset of synthesized molecules has variation in functionality regarding the substituent at the phenyl ring, hence the criteria for training set selection was based on the active molecules with good diversity in functionality at this position.

5.2 Generation of pharmacophore model

The selected training set was utilized to develop CFPM for both A. fumigatus and C. albicans for detection of important chemical functionalities guiding activity. The HipHop module in DS2.0 identifies common chemical feature pattern by over laying molecules in the training set. The chemical functions namely hydrogen bonding acceptor (HBA), ring aromatic (RA), and hydrophobic aliphatic group (HAl), was selected based on optimization procedure and the chemical features present in the training set. Variations in the parameters related to Maximum Omitted Features, Misses, and Complete Misses were done due to the possibility of presence and absence of chemical features in some compounds of the training set. In this context, the “principal number” was set to 2 that ensure that all chemical features in the compounds are considered during generation of hypothesis space. The “maximum omitting features” was set to 0 that force the mapping of chemical features with the pharmacophore features. All the parameters were kept default for the generation CFPM in both cases (A. fumigatus and C. albicans). In hypothesis generation run top 10 possible pharmacophore hypotheses were sorted on the basis of ranking score.

5.3 Conformation generation

The structures in the dataset were built using the 2D editor ISIS draw 2.5 and imported to (DS2.0) window for the generation of 3D structures. The conformational search for each molecule was next performed utilizing the best quality conformational search option in DS2.0 keeping the energy threshold constraint to 20 kcal mol⁻¹ above the global energy minimum. To ensure proper conformation sampling, a maximum of 255 conformations in BEST mode were generated for each structure.