Synthesis, Structure-Activity Relationships (SAR) and in Silico Studies of Coumarin Derivatives with Antifungal Activity

The increased incidence of opportunistic fungal infections, associated with greater resistance to the antifungal drugs currently in use has highlighted the need for new solutions. In this study twenty four coumarin derivatives were screened in vitro for antifungal activity against strains of Aspergillus. Some of the compounds exhibited significant antifungal activity with MICs values ranging between 16 and 32 μg/mL. The structure-activity relationships (SAR) study demonstrated that O-substitutions are essential for antifungal activity. It also showed that the presence of a short aliphatic chain and/or electron withdrawing groups (NO2 and/or acetate) favor activity. These findings were confirmed using density functional theory (DFT), when calculating the LUMO density. In Principal Component Analysis (PCA), two significant principal components (PCs) explained more than 60% of the total variance. The best Partial Least Squares Regression (PLS) model showed an r2 of 0.86 and q2cv of 0.64 corroborating the SAR observations as well as demonstrating a greater probe N1 interaction for active compounds. Descriptors generated by TIP correlogram demonstrated the importance of the molecular shape for antifungal activity.


Introduction
The increased incidence of fungal infections, especially dangerous hospital-acquired infections and infections in immunocompromised patients has highlighted the need for new antifungal treatments. Drug-resistant fungal isolates have been reported for all known classes of antifungal drugs. As a result, mortality, morbidity, and the associated cost of medical care for fungal infections are all steadily rising. In addition, because many of the currently available drugs are toxic and have other drawbacks involving spectrum, tissue distribution, central nervous system penetration, and high cost, the number of efficacious anti-mycotic drugs is limited [1]. Many of these drugs actually produce infection recurrence, for being fungistatic and not fungicidal.
In particular, Aspergillus infections have been increasing, and while the most frequent pathogen to cause aspergillosis is Aspergillus fumigatus, A. terreus, A. flavus and A. niger are becoming increasingly common [2]. A. fumigatus is an opportunistic pathogen which incites disease in hosts whose local or systemic immune response has been impaired, damaged, or is innately dysfunctional. After Aspergillus fumigatus, A. flavus is the second leading cause of invasive aspergillosis, and it is the most common cause of superficial infections. Common clinical syndromes associated with A. flavus infections include chronic granulomatous sinusitis, keratitis, cutaneous aspergillosis, wound infections, and osteomyelitis following trauma and inoculation. In addition, A. flavus produces aflatoxins, the most toxic natural hepatic-carcinogens ever characterized [3].
There is an urgent need for discovery and testing of new compounds with antifungal properties. Natural products are inexhaustible as a source of chemical structures and for more than a century have been used as scaffolds for the synthesis of new drugs. The coumarins, phenolic compounds which possess a benzopyranone nucleus [4] are one of the major classes of secondary metabolites, which has been highlighted in biological studies as being (anti-HIV [5], hepato-protective [6], anti-inflammatory [7], antimicrobial [8], antimitotic [9] and antitumor [10]), and in part, due to their antifungal properties [11][12][13] they are believed to exert a role in plant protection against herbivorous, fungal, and bacterial infections [14].
It was recently reported that coumarin derivatives may be used effectively as antifungal agents. Johann et al. [11] demonstrated that coumarins isolated from plant extracts exhibit activity against certain strains of Sporothrix schenckii (MICs of 125-250 µg/mL), and Cryptococcus gattii (MIC of 250 µg/mL). Daoubi et al. [13] showed their antifungal activity against Botrytis cinerea in concentrations ranging from 25 to 200 mg/L with IC 50 values of 33.3-157.5 mg/L. Creaven et al. [15] demonstrated that coumarin derivatives are active against Candida albicans (MIC 80 of 4.5-246.6 µM). Jurd et al. [16,17]  Pursuing our research in the field [18][19][20], we herewith describe the synthesis and in vitro antifungal evaluation of coumarin derivatives against Aspergillus species. After antifungal evaluation, the derivatives were subjected to structure-activity relationship (SAR) analysis, electronic surface analyses, molecular modelling, and chemometrics (Principal Component Analysis (PCA), and Partial Least Square Regression (PLS)), all to extract information on structure and its relation to antifungal properties [21,22].
The chemical structures of 11-22 and 24 were confirmed by comparing their physical and spectral data with the respective literatures. The purity of all synthesized compounds were >98% determined by HPLC.
Structural confirmation of 23 was based on its NMR, mass spectra and elemental analysis. In the 1 H NMR we observed only three signals related to three hydrogens which indicated a probable bi-nitration. There were two doublets coupled together (J = 10.0 Hz) at 6.75 and 7.77 ppm, which were attributed to H-3 and H-4, and one singlet at 8.23 ppm (related to H-8). These signals provide good evidence that a bi-nitration occurred on the aromatic ring at the C-5 and C-7 positions, both were ortho-positioned with respect to the 6-hydroxyl group, as also observed in previous works [38]. The unequivocal determination of the carbons was done by interpretation of their 13 C NMR, 13 C-1 H HMQC and 13 C-1 H HMBC spectra.
In the 13 C-1 H HMQC was possible to determine unambiguously the three primary carbons at 116.3 (C-8), 124.0 (C-3) and 136.7 ppm (C-4). Signals from quaternary carbons were determined by analysis of the correlations observed in the 13 C-1 H HMBC, which are described in Table 1.
The mass spectrum also confirmed the proposed structure, where it is observed as the base peak, the fragment [M-1] +• at 250.9, characteristic for phenolic compounds.

Hydrogen
Carbon
The introduction of electron withdrawing groups to the coumarin skeleton appears to contribute positively to the fungicidal activity, as seen with all the nitro-derivatives evaluated. Nitration of the inactive compounds (3, 4 and 5) resulted respectively in derivatives: 24 (MIC = 1024 µg/mL), 23 (MIC = 512 µg/mL), and 22 the most active compound (MIC = 16 µg/mL) each derivative being (respectively) 2, 4, and 128 times more active than its precursor. DFT confirmed these findings. In the LUMO density surfaces we highlight the differences between active (nitro-coumarins) and inactive compounds. The active compounds have stronger red regions (lower electronic concentration). Although the electrostatic potential map also represents the electronic concentration, LUMO density better demonstrates the ability of one molecule to receive electrons ( Figure 1).
The positive influence of electron withdrawing groups can also be observed for the three series (4-, 6-and 7-hydroxycoumarins) of synthesized compounds. In every case it was observed that the compounds resulting from acetylation (respectively 15, 14, and 12), are the most active from each series. Compound 15 was as active as compound 22 (the most active compound). The hypothesis that electron withdrawing groups favor antifungal activity corroborates the observations found with PLS analysis (Section 2.3.), where it was found that the interaction of compounds with the probe N1, increases with antifungal activity.
Analyzing the other replacements performed (introduction of geranyl, prenyl and allyl groups), we observed differing patterns of activity for the 6-hydroxy-and 7-hydroxy-coumarin derivatives. With the 6-hydroxy-coumarin derivatives, substitutions did not result in increased fungicidal activity, all compounds showed to be inactive (17, 18 and 19). This was also observed in previous studies [17]. However for the 7-hydroxy-coumarin series, there is a clear relationship between the size of the introduced group and the fungicidal activity. Reduction of the alkenyl side-chain length increases the activity proportionately, so that the compound having the largest radical (geranyl) is inactive (13 MIC ≥ 2048 µg/mL), the compound with an intermediate side-chain (prenyl) showed strong to moderate activity (16 MIC = 1024 µg/mL for A. fumigatus and 128 µg/mL for A. flavus), and the compound with the smallest aliphatic chain displayed the best activity profile (11 MIC = 64 µg/mL).These results are in agreement with studies observed by Jurd et al. [16] who observed that the antifungal activity of umbelliferone (7-hydroxy-coumarin) may be increased by O-alkylation, and O-acylation with shorter alkyl and acyl groups.

Computational Studies
For performing chemometric analysis (Principal Component Analysis (PCA)), and Partial Least Square Regression (PLS) the MIC values for Aspergillus fumigatus ATCC-16913 were used [21,39,40]. Inhibitory activity data of the investigated compounds determined as micrograms per milliliter were converted to the negative logarithms of molar MICs (log1/c MIC ) ( Table 2) which was used as a dependent variables set in this study. Principal component analysis (PCA) and partial least squares (PLS) are chemometric tools for extracting and rationalizing the information from any multivariate description of a biological system. Complexity reduction and data simplification are two of the most important features of such tools. These chemometric tools were developed in the Pentacle software [41]. The Pentacle software is a computational tool for computing alignment-free molecular descriptors, also called GRid-INdependent descriptors or GRIND. The software is based on Molecular Interaction Fields, describe the ability of the molecules to interact with other molecules and do not require to superimpose the compounds. GRIND descriptors are highly relevant for describing biological properties of compounds [42] (see item 3.4. Molecular modeling and electronic surfaces).

Principal Component Analysis (PCA) and Partial Lest Squares Regression (PLS)
The PCA is a technique used to reduce the principal matrix information, splitting into two smaller matrices called loading and score: the matrix loading (P) contains information about the variables and is composed of vectors (principal components, PCs) which are obtained from the original variables; the matrix score (T) contains information about the objects. Each object is described in terms of the projections from the PC instead of the original variables.
Preliminary exploratory analysis was developed using PCA and considering 405 independent variables or descriptors. Two significant principal components (PCs) explain more than 60% of the total variance (Table 3). Score and loading plots are interconnected until any descriptor change in the loading plot is reflected by changes in the position of compounds in the score plot [43]. The score plot exhibits satisfactory discrimination between more potent (blue) and less potent (red) compounds. The less active compounds are concentrated in the upper left quadrand (Figure 2). The best model from PLS was obtained with three LVs and 72 descriptors selected from 405. The variable selection via fractional factorial design (FFD), which evaluates the effect on the model standard derivation of error of prediction (SDEP) of every single variable and variable combination. The GRIND descriptors condensed represent a small number of principal properties (GRIND-PP), requiring a biologically relevant description of the molecular similarity. Two significant latent variables emerged from PLS model and leave-one-out [44]. The best model showed an r 2 of 0.86, and a q 2 cv of 0.64. Figure 3 represent the score plot obtained with PLS. The principal descriptors highlighted in the PLS model were: 350 (5.6-6 A), 531 (10.8-11.2 A), (shape), the molecular contour also appears to be important. The molecular descriptors obtained can be observed in graphical diagrams called "correlograms". We observed the largest number of interactions of active compounds with the probe N1. This may indicate that coumarins devoid of electrons have greater attraction with electronegative atom of this probe. Otherwise occurs in inactive compounds. The Figure 4 shows the active compounds 12 and 15 and their interactions with N1. Comparatively, we can see the inactive compounds (3 and 4). Inactive interactions are smaller than active ones, corroborating the fact that electron withdrawing groups increases the antifungal activity.

General Alkylation Procedure
A mixture of commercial coumarins (3, 4 or 5) (3 mmol), K 2 CO 3 (3.9 mmol), and an alkyl halides (3.9 mmol) in anhydrous acetonitrile (20 mL) was stirred in reflux for 22-46 h, and monitored by TLC. The reaction was filtered and washed two times with ethyl acetate (10 mL). The organic phase was washed with water (50 mL), dried over Na 2 SO 4 , and then evaporated in vacuum. The residue obtained was recrystallized using Methanol/THF mixtures to give the compounds (by % yield/time reaction): 11 [25]

Antifungal Activity
The in vitro antifungal activity of the studied compounds was investigated using eight Aspergillus strains, including four strains of Aspergillus fumigatus (ATCC 16913,  Stock cultures were kept on sterile Sabouraud Dextrose Agar (SDA) slants under 7 °C (±1 °C). For preparing the inoculum of the anti-mold assays, we used 7 day-old cultures grown on sterile SDA at 25-28 °C. After the incubation period, the mold conidia were removed by adding sterile NaCl 0.85% to the growth media followed by gentle shaking for 30 s. Mold conidia were counted using a hemocytometer. The conidial suspension inoculum was adjusted using sterile NaCl 0.85%, for approximately 5 × 10 6 conidia/mL. MIC values were determined by the microdilution broth method using 96-wells microplates [45,46]. Conidial suspension from 7-day-old A. flavus culture was prepared and standardized by hemocytometer in sterile NaCl 0.85% for susceptibility testing as described previously. To a 96-well plate Sabouraud broth and coumarins were added at concentrations of 2048 to 8 μg/mL. The MIC determination was conducted with an inoculum of approximately 2.5 × 10 5 conidia/mL microorganism in each well. The plates were incubated at 25-28 °C for 72 h. Within 72 h (and confirmed at 7 days) there was visible fungal growth. The MIC was defined as the lowest concentration of antifungal agent that completely inhibited the growth of the fungi, as detected by the unaided eye [47]. Amphotericin B was used as the reference fungicide.

Molecular Modeling and Electronic Surfaces
Using the program Hyperchem version 8.0 [48], the chemical structures of the compounds of interest were drawn and their geometry was optimized using MM+ force field [49]. Afterwards, we performed a new geometry optimization using the semi-empiric method RM1 (Recife Model 1) [50]. The RHF (Restricted Hartree-Fock) level was used. The optimised structures were subjected to conformational analyses using the random search method [22,51] with 1000 interactions, 100 cycles of optimisation, and 10 lowest minimum energy conformers. The selected dihedrals were evaluated by rotation in accordance with the standard (default) conditions of the program, in which the number of simultaneous variations was 1 to 8, and acyclic chains were submitted to rotations from 60 to 180 o , while torsion rings were in the range of 30° to 120°.
The lowest energy conformers selected were saved in .sdf format with the Spartan for Windows 8.0 program [52], and then exportedto the Pentacle 1.5 program [41], PCA and PLS methodologies were carried out. The Pentacle software is a computational tool for computing alignment-free molecular descriptors, also called GRid-Independent descriptors or GRIND. The software is based on Molecular Interaction Fields and describes the ability of the molecules to interact with other molecules without requiring compound superimposition. Aims to find the underlying relationship between the structure of a molecule and its binding affinity (or other biological properties) using information extracted from molecular descriptors. The calculation of descriptors includes the hydrophobic probe (DRY), the hydrogen bond acceptor probe (O), the shape probe (TIP), and the hydrogen bond donor probe (N1) [42].

Conclusions
Synthesis and antifungal evaluation of twenty-four coumarin derivatives against Aspergillus fumigatus and A. flavus are described. Some derivatives showed significant antifungal activities with MICs values ranging between 16 and 32 µg/mL, including seven compounds (2, 11, 12, 14, 15, 20 and   22) which were more active than the reference drug Amphotericin B for the LM-26 strain (A. flavus).
SAR study permitted two conclusions: O-substitution is essential for antifungal activity and the presence of a short aliphatic chain (geranyl<prenyl<allyl), and/or electron withdrawing groups (NO 2 and/or acetate) favors activity.
In parallel, the compounds were submitted to the chemometric tools: PCA and PLS, using the program Pentacle. The results were satisfactory and corroborated the SAR analysis. The most active compounds showed greater interaction with the probe N1, which shows greater electronic concentration. These findings were also confirmed by density functional theory (DFT), calculating the LUMO density. Descriptors generated by the probe TIP were also highlighted, indicating that the molecular contour is indeed important for the antifungal activity, as observed in the SAR analysis.
These results suggest derivatives of coumarin as promising compounds for the development of new anti-Aspergillus agents.