Screening for In Vitro Antimycobacterial Activity and Three-Dimensional Quantitative Structure–Activity Relationship (3D-QSAR) Study of 4-(arylamino)coumarin Derivatives


Corresponding authors: Anamik Shah,; Evans C. Coutinho,


The resurgence of tuberculosis and the emergence of multidrug-resistant strains of Mycobacteria necessitate the search for new classes of antimycobacterial agents. We have synthesized a small library of 50 analogues of 4-(arylamino)coumarins with various aromatic amines at the C4- position of the coumarin scaffold. The compounds were evaluated for antimycobacterial activity against Mycobacterium tuberculosis H37Rv with rifampicin as the standard. Of the molecules synthesized, compound 9 was found to be most potent with a minimum inhibitory concentration >6.25 μg/mL for 100% inhibition. In an effort to develop new and more effective molecules in this series, the relationship between structure and activity was investigated by comparative molecular field analysis. Various models were generated using comparative molecular field analysis alone and comparative molecular field analysis plus a hydropathy field (HINT). In all, eight models were generated with atom-fit and field-fit alignment strategies. The comparative molecular field analysis models (Models 3a and 4a) based on field-fit alignment were the best with statistically good correlation coefficients (r2) and cross-validated q2. The values of r2pred for the validation set were 0.469 and 0.516. Based on the comparative molecular field analysis contours, some insights into the structure–activity relationship of the compounds could be gained.


three-dimensional quantitative structure–activity relationship


Comparative molecular field analysis


Hydropathic interactions


minimum inhibitory concentration


multidrug-resistant tuberculosis


extensively drug-resistance tuberculosis


Thymidine monophosphate kinase

Control and prevention of tuberculosis (TB) is a major challenge as one-third of the world’s population is infected with Mycobacterium tuberculosis. Nearly 2 million deaths because of TB occur every year (1–3). According to a WHO report by the year 2020 nearly one billion more people will be infected, and 70 million will die from TB if proper steps are not taken to control it (4). TB is the world’s second cause of death from infectious diseases after acquired immune deficiency syndrome (AIDS) (5). Moreover, the association of TB with AIDS and the emergence of drug-resistant TB strains have made available chemotherapeutic strategies increasingly ineffective (6).

Unfortunately, no new drugs or antibiotics have been developed in the last 30 years against TB. There are three frontline drugs viz. isoniazid, rifampicin, ethambutol, and several second-tier drugs that include ethionamide, streptomycin, and p-aminosalicylic acid to treat TB. However, many of these drugs have disadvantages, such as prolonged treatment schedules, host toxicity, ineffectiveness against resistant strains. (5,7). The problem of TB is further complicated by the emergence of multidrug-resistant TB (MDR-TB) (8–15) and ‘extensively drug-resistance tuberculosis’ (XDR-TB) that threaten TB control globally (16,17).

The impact of ever-increasing drug resistance, the serious side effects of some current anti-TB drugs and the lack of efficacy of current treatments in immuno-suppressed patients combine to make the development of new antimycobacterial agents an urgent priority (8,18). There are several reports in the literature of new structural classes being developed as antitubercular agents. A series of C3 alkyl and arylalkyl-2,3-dideoxyhex-2-enopyranosides with minimum inhibitory concentration (MIC) values between 25.0 and 1.56 μg/mL against Mycobacterium tuberculosis H37Rv(7) has been reported. Similarly, 5-acetylenic derivatives of 2′,3′-dideoxyuridine and 3′-fluoro-2′,3′-dideoxyuridine have been tested for antimycobacterial activity against M. bovis, M. tuberculosis, and M. avium. The most promising compounds in this series, 5-dodecynyl-2′,3′-dideoxyuridine and 5-tetradecynyl-3′-fluoro-2′,3′-dideoxyuridine showed excellent activity and inhibited the growth of M. bovis and M. tuberculosis with MIC90 in the range of 1–2 μg/mL, which is close to the reference drug rifampicin (19). The derivatives of 5′-O-[(N-acyl)sulfamoyl]adenosine were evaluated against M. avium and M. tuberculosis, under iron-deficient and iron-rich conditions. Of the various molecules in the series, the 4-fluorosalicyl derivative was found to be the most potent (MIC99 of 0.098 μm), while the 4-aminosalicyl derivative was the first analogue that exhibited improved selectivity against M. tuberculosis under iron-rich conditions (MIC99 of 25 μm) (20). A series of 1-β-D-2′-arabinofuranosyl and 1-(2′-deoxy-2′-fluoro-β-D-ribofuranosyl)pyrimidine nucleosides with a variety of alkynyl, alkenyl, alkyl, and halo substituents at the C5 position of the uracil moiety was investigated against M. tuberculosis, M. bovis, and M. avium. Among the designed nucleosides, 1-β-D-2′-arabinofuranosyl-5-dodecynyluracil, 1-(2′-deoxy-2′-fluoro-β-D-ribofuranosyl)-5-dodecynyluracil, and 1-(2′-deoxy-2′-fluoro-β-D-ribofuranosyl)-5-tetradecynyluracil were the most potent against M. bovis and M. tuberculosis, respectively (21). Recently, thymidine monophosphate kinase (TMPK) has been proposed as an attractive target for the development of antitubercular agents. The newly developed (3-trifluoromethyl-4-chlorophenyl)thiourea has a Ki of 0.6 μm and a selectivity index of 600 against human TMPK. Moreover, it represents the first TMPK inhibitor with good inhibitory activity on growing M. bovis (MIC99 = 20 μg/mL) and M. tuberculosis (MIC50 = 6.25 μg/mL) strains (22).

Moreover, various structural classes, such as diarylmethanes (23), oxazolyl thiosemicarbazones (24), pyrazolines (25), amphiphilic galactopyranosyl diamine and amino alcohols (26), isonicotinohydrazones (27), triazoles (28) have been identified as promising antitubercular leads. Calanolide A, a naturally occurring coumarin derivative and an inhibitor of HIV-1 reverse transcriptase also have shown promising in vitro activity against M. tuberculosis with an MIC value of 3.13 μg/mL (29,30).

We have been studying for quite some time heterocyclic compounds with a variety of scaffolds as novel antitubercular agents (31–37) and very recently, we have reported the in vitro antitubercular activity and 3D-QSAR studies of coumarin-4-acetic acid benzylidene hydrazides (38). In continuation of this work, the current paper describes the syntheses of substituted 4-(arylamino)coumarins and coumarins fused with a benzothiazipine nucleus and their in vitro antitubercular activity against M. tuberculosis H37Rv strain. A three-dimensional quantitative structure–activity relationship (3D-QSAR) study has been recently reported on 4-aminocoumarins tested as antiplatelet agents (39,40). To understand the structural requirements for antitubercular versus antiplatelet activity, the 3D-QSAR technique of comparative molecular field analysis (CoMFA) (41,42) with and without inclusion of hydropathic interactions (HINT) hydropathy fields (43) was also carried out.

Methods and Materials


All chemicals and solvents were purchased from Spectorchem (Mumbai, India) and SD Fine Chemicals (Mumbai, India) and used without purification. For syntheses of the title compounds, a monomode, open-vessel microwave synthesizer (Q-ProM model; Questron Technologies Corporation, Mississauga, Ontario, Canada) was used. Melting points were determined in open glass capillaries in a paraffin bath and are uncorrected. 1H NMR and 13C NMR spectra were recorded on a Bruker AC 300 MHz FT-NMR spectrophotometer using tetramethylsilane as the internal standard. Mass spectra were recorded on a JEOL SX 102/DA-6000 by the FAB (fast atom bombardment) technique for ionization. FT-IR spectra were recorded on a Shimadzu-8400 using KBr disks. Thin layer chromatography (TLC) was carried out on Merck Kieselgel 60F254 (Merck 5549, USA) plates. The structures of the molecules synthesized are given in Table 1, while all experimental details are briefly given in the supplementary material (Table S1).

Table 1.   The structures and activities of 4-aminoarylcoumarine derivatives
inline image
CompoundsR5R6R7R8Ar*Ro*Rm*RpbGI % (μg/mL)pALogit
 2-H-H-H-H-Ph-Ph-H-H65 (>6.25)4.97
 3-H-CH3-H-H-Ph-Ph-H-H14 (>6.25)3.93
 4-H-Cl-CH3-H-Ph-Ph-H-H38 (>6.25)4.55
 6-H-H-H-H-Naphα---81 (>6.25)5.29
 7-H-CH3-H-H-Naphα---10 (>6.25)3.73
 8-H-H-H-CH3-Naphα---54 (>6.25)4.75
 9-H-Cl-CH3-H-Naphα---100 (>6.25)ND
12-H-H-CH3-CH3-Naphα---9 (>6.25)3.70
13-CH3-H-CH3-H-Naphα---9 (>6.25)3.70
14-H-H-H-H-Ph-H-H-COOH18 (>6.25)3.99
15-H-CH3-H-H-Ph-H-H-COOH15 (>6.25)3.92
16-H-H-H-CH3-Ph-H-H-COOH33 (>6.25)4.37
17-H-Cl-CH3-H-Ph-H-H-COOH8 (>6.25)3.66
19-H-H-CH3-CH3-Ph-H-H-COOH41 (>6.25)4.54
20-CH3-H-CH3-H-Ph-H-H-COOH20 (>6.25)4.09
21-H-H-H-CH3-Ph-H-H-Br28 (>6.25)4.31
22-H-H-H-CH3-Phdi-Cl-H-H11 (>6.25)3.8
23-H-CH3-H-H-Ph-H-H-Br1 (>6.25)2.73
24-H-H-H-H-Ph-H-H-F18 (>6.25)3.95
25-H-H-H-CH3-Ph-H-H-H7 (>6.25)3.51
26-H-CH3-H-H-Ph-H-H-F1 (>6.25)2.64
27-H-CH3-CH3-H-Ph-H-H-F31 (>6.25)4.31
28-H-H-CH3-CH3-Ph-H-H-F22 (>6.25)4.11
30-H-H-H-H-Ph-H-Cl-F15 (>6.25)3.91
31-CH3-H-CH3-H-Ph-H-Cl-F4 (>6.25)3.43
32-H-H-H-H-Ph-Cl-H-Cl28 (>6.25)4.28
33-H-H-H-H-Ph-Cl-H-H11 (>6.25)3.73
34-H-H-H-H-Ph-H-H-Cl10 (>6.25)3.68
35-H-H-H-H-Ph-H-H-OCH315 (>6.25)3.88
36-H-H-H-CH3-Ph-H-H-OCH37 (>6.25)3.53
37-H-H-H-CH3-Ph-H-H-Cl10 (>6.25)3.71
38-CH3-H-CH3-H-Ph-H-H-Cl9 (>6.25)3.68
39-H-Ph-H-H-Ph-H-Cl-H23 (>6.25)4.22
40-H-H-H-H-Ph-H-NO2-H30 (>6.25)4.28
41-H-H-H-H-Ph-CH3-H-H44 (>6.25)4.5
42-H-H-H-H-Ph-H-H-CH350 (>6.25)4.6
43-H-H-H-H-Ph-OCH3-H-H39 (>6.25)4.44
44-H-H-H-H-Ph-CH3-CH3-H51 (>6.25)4.65
45-H-H-H-H-Ph-CH3-H-CH349 (>6.25)4.61
inline image
CompoundsR5R6ArbGI % (μg/mL)pALogit
11-H-H-Naphα11 (>6.25)3.82
inline image
CompoundsR7R8Ar*Ro*Rm*RpbGI % (μg/mL)pALogit
 5-H-H-Ph-Ph-H-H7 (>6.25)3.64
10-H-H-Naphα---5 (>6.25)3.45
18-H-H-Ph-H-H-COOH26 (>6.25)4.27
29-H-H-Ph-H-H-COOH53 (>6.25)4.74
inline image
CompoundsYR5R6R7R8Ar*Ro*Rm*RpbGI % (μg/mL)pALogit
46O-H-H-H-H-Ph-H-Cl-H29 (>6.25)4.29
47O-CH3-H-CH3-H-Ph-H-Cl-H36 (>6.25)4.47
48NH-Cl-H-H-Cl-Ph-H-H-H67 (>6.25)5.04
49NH-H-Cl-H-H-Ph-H-H-H8 (>6.25)3.62
inline image
CompoundsR5R8R9R10Ar*Ro*Rm*RpbGI % (μg/mL)pALogit
  1. -Naphα = α-naphthyl

  2. *Ro/m/p positions with respect to –NH.

  3. bGI = Growth inhibition of virulent strain of M. tuberculosis

  4. The values in the parenthesis denote the MIC in μg/mL. MIC of Rifampin: 0.015–0.125 μg/mL against M. tuberculosis H37Rv (97% inhibition).

50-H-H-CH3--Ph-H-H-H78 (>6.25)5
51-H-H-CH3-CH3-Ph-H-H-H54 (>12.5)4.53

General method for the synthesis of 4-(arylamino)coumarin derivatives (2–45)

Various 4 -hydroxycoumarins were prepared by earlier reported method. Substituted 4-hydroxycoumarin (0.01 mol) and corresponding amines (0.015 mol) were taken in a reaction vessel. The reaction mixture was subjected to microwave irradiation (400 W) at 140 °C (Scheme 1). The reaction was monitored by TLC (ethylacetate:hexane::3:2). After completion of the reaction as indicated by TLC, acetone was added to the reaction mixture and the solid mass obtained was filtered. The crude reaction mass was finally purified by silica gel (60–100 mesh) column chromatography employing ethyl acetate and hexane (3:2) as the mobile phase (44–45).

Figure Scheme 1:.

 Microwave-assisted synthesis of substituted 4-arylaminocoumarins (2–45).

General methods for the synthesis of phenothiazine derivatives (46–51)

Substituted 4-hydroxycoumarin (0.01 mol) and corresponding 2-aminothiophenols (0.012 mol) were taken in a reaction vessel and subjected to microwave irradiation (400 W) at 140 °C for an appropriate time (Schemes 2 and 3). The reaction was monitored by TLC (chloroform:methanol::9:1). On completion, the reaction mixture was allowed to cool. The reaction mass was washed with acetone when the crude product separated out. This was purified by silica gel (60–100 mesh) column chromatography using chloroform and methanol (9:1) as the mobile phase (46,47).

Figure Scheme 2:.

 Microwave-assisted synthesis of phenothiazine derivatives (46–49).

Figure Scheme 3:.

 Microwave-assisted synthesis of pyrano-coumaro-phenothiazine derivative (50–51).

Antitubercular activity

Antitubercular activity was determined using the BACTEC 460 system modified as described below. Stock solutions of test compounds were prepared in dimethylsulfoxide (DMSO) in 1 mg/mL concentration and sterilized by passage through 0.22- μm PFTE filters (Millex-FG; Millipore, Bedford MA, USA). Volumes equivalent to 50 μL were added to 4 mL radiometric 7H12 broth (BACTEC 12B; Bectron Dickinson Diagnostic Instrument system, Sparks, MD, USA) to achieve a final concentration of 6.25 mg/mL. Controls received 50 mL DMSO. Rifampicin (Sigma Chemical Co. St. Louis, MO, USA) was included as a positive drug control. Rifampicin was solubilized and diluted in DMSO and added to BACTEC-12 broth to achieve MIC inhibiting 99% of the inoculum.

Mycobacterium tuberculosis H37Rv strain (ATCC 27294; American type culture collection; Rockville, MD, USA) was cultured at 37 °C on a rotary shaker in middle brook 7H9 broth (Difco Laboratories, Detroit, MI, USA) supplemented with 0.2% v/v glycerol and 0.05% v/v Tween 80 until the culture turbidity achieved an optical density of 0.45–0.55 at 550 nm. Bacteria were pelleted by centrifugation, washed twice, and resuspended in one-fifth of the original volume in Dulbecco’s phosphate-buffered saline [PBS, Irvine Scientific, Santa Ana, (A)]. Large bacterial clumps were removed by passage through an 8 mm filter (Malgene, Rochester, NY, USA), and aliquots were frozen at −80 °C. The cultures were thawed and an appropriate dilution was performed such that a BACTEC-12B vial inoculated with a 0.1 mL would reach a growth index (GI) of 999 in 5 days. One-tenth of the diluted inoculums were used to inoculate 4 mL fresh BACTEC 12B broth containing the test compounds. An additional control vial was included, which received a further 1:100 diluted inoculums (as well as 50 mL DMSO) for use in calculating the MIC of rifampicin, respectively by established procedures.

Cultures were incubated at 37 °C, and the growth of inhibition (GI) determined daily until control cultures achieved a GI of 999. Assays were usually completed in 5–8 days. Percent inhibition was defined as 1-(GI of test sample/GI of control) 10. MIC of compound effecting a reduction in daily change in GI, which was less than that observed with a 1:100 diluted control culture on day the later reached a GI of at least 30.

Computational details

Comparative molecular field analysis was carried out using Sybyl v.7.1 (Tripos Inc., USA) a running on a Pentium IV computer under the Linux RedHat Enterprise 2.3.1.

Ligand preparation

The antitubercular activity of the molecules considered in the QSAR studies span a 3 log unit range (Table 1). The molecules were built with the Sketch module in Sybyl. Figure 1 shows a representative structure for these molecules; for the phenothiazine derivatives, X represents a S-bridge between the coumarin and the phenyl rings. To determine the correct geometry at the 4-amino position, ab initio calculations were run on molecule 21 with the Gaussian software.b Geometry optimization was carried out at both the AM1 and HF-3/21G level of theories. Based on these calculations, the atom type for the N-atom was determined as ‘N3’ i.e., sp3-pyramidal. The conformation of the aromatic ring around the N-CAr bond was scanned using Systematic Search for representative molecules with different substituents in the ortho or meta-positions, and the global minimum conformation thus determined was adopted for these and other molecules with similar substitutions.

Figure 1.

 A general representation of the structure of molecules (X = S, Y = NH/O).

The geometries of all molecules, prior to the QSAR study, were optimized by energy minimization by the conjugate gradient method with the Tripos force field and the Gasteiger Hückel charges for all atoms, until a gradient 0.01 kcal/mol/Å was reached. Gasteiger-Hückel charges were used for the QSAR studies.

Atom-fit-based alignment

The atoms of the 4-aminocoumarin ring are a common feature in all molecules and this set was used for superimposition onto reference molecule 9. The alignment based on atom-fit strategy is shown in Figure 2.

Figure 2.

 4-Aminocoumarins aligned by the Atom-fit method.

Field-fit-based alignment

Field-fit alignment was also utilized for generating 3D-QSAR models. The steric and electrostatic fields were calculated around all molecules with a grid spacing of 0.5 Å. The molecules were then aligned using the field-fit methodology with the fields around molecule 9 as the reference. The alignment based on field-fit strategy is shown in Figure 3.

Figure 3.

 4-Aminocoumarin derivatives aligned by the Field-fit technique.

Statically rotated alignments

The molecules previously aligned by atom-fit and field-fit methodologies were rotated in 3D space to generate additional orientation sets for QSAR studies. The STATIC ROTATE command in Sybyl was used for generating such differently oriented sets. The orientation that yielded statistically significant results was the one obtained by rotating the molecules by 90° in the X, 180° in the Y, and 270° in the Z directions (models 1a, 2a, 3a and 4a in Table 2).

Table 2.   PLS statistics of CoMFA and CoMSIA 3D-QSAR models derived by atom-fit-based and field-fit alignments
PLS statisticsAtom-fit alignmentField-fit alignment
Model 1Model 1aModel 2Model 2aModel 3Model 3aModel 4Model 4a
  1. The models designated by letter ‘a’ refer to those deduced from datasets obtained with static rotation of particular alignments in 3D space.

  2. CoMFA = Comparative molecular field analysis, HINT = Hydropathic interactions, n = number of molecules in training set, r2 = correlation coefficient, q2 (LOO) = cross-validated r2 with Leave-one-out method, r2pred = Correlation coefficient for prediction of external datasets, r2BS = Bootstrap Correlation coefficient, F = F statistics, SE = standard error, PLS = partial least square analysis.

PLS components65656666
q2 (LOO)0.410.470.380.350.460.580.400.56
F 40.2 56.9 51.8 39.1 151.8 245.4 121.4 187.6
Field contribution


The set of 36 diverse molecules (Table S2 in supplementary information) was used in the CoMFA study. The dataset was divided into a training set (27 molecules) and a test set (9 molecules) by means of chemical as well as biological diversity. Daylight fingerprints of the molecules along with the pAlogit data were used to separate the molecules into training and test sets based on the Tanimoto similarity coefficient. Later on, few molecules were swapped among the two datasets to get statistically sound models. The final models were generated on a training set of 27 molecules and a test set of six molecules.

Biological Data

For the QSAR study, the activity values were transformed as follows b


where ‘c’ is molar concentration = concentration (μg/mL) * 0.001/(molecular weight); and


CoMFA studies

In all CoMFA studies, default settings for the 3D cubic lattice, the grid spacing, the probe atom, and the energy cut-offs were used.

The CoMFA descriptors (steric and electrostatic fields) were used as independent variables and the activity (as the negative logarithm) values as the dependent variable in a partial least squares (PLS) regression analysis (48) to derive 3D-QSAR models. The CoMFA models were generated from molecular alignments generated by atom fit and field fit and their orientation in the 3D-grid optimized with the static rotation command in Sybyl. The hydrophobic field calculated using HINT (43) was used in conjunction with the CoMFA fields to derive additional models.

The hydropathic fields were calculated with the HINT module in Sybyl. The region defined for calculation of the CoMFA fields was also used to describe the space in which the HINT fields were generated. HINT calculates the hydrophobic interaction between all atom pairs in a molecule using the following equation


where, bij = aiajSiSjRijTij

bij = micro-interaction constant representing the attraction/interaction between atoms i and j

ai = the hydrophobic atom constant for atom i

Si = the solvent accessible surface area for atom i

Rij = the functional distance behavior for the interaction between atoms i and j

Tij = a discriminant function designed to keep the signs of interactions consistent with the HINT convention that favorable interactions are positive and unfavorable interactions are negative.

The models were internally evaluated by leave-one-out (LOO) cross-validation. The optimal number of components was determined by the SAMPLS (49) method and this was subsequently used to derive the final QSAR models. In addition to the q2, the conventional correlation coefficient r2 and standard errors (SE) were also computed, the statistics of the models generated are shown in Table 2. The models obtained were internally validated by bootstrapping (50–52), Y-scrambling (53,54), and externally validated by calculating the r2pred for the validation set (test set).

Design of new molecules

New molecules were designed based on a protocol published recently (55). A library of 913 molecules was generated using the CombiChem I utility in Cerius2 (Accelrys Inc., San Diego, CA, USA). The generation of this library of molecules was guided by the CoMFA contours. Appropriate side chains were attached to molecules 6, 26 and 50. Molecules generated in this way were optimized with the same protocol of energy minimization and then aligned as described earlier. The activities of these molecules were predicted using the best models 3a and 4a, respectively (Table 3). The ADME and toxicity properties of the molecules were predicted with QuickPropc and the web-based PreADMET (56) tools to identify drugable compounds.

Table 3.   The predicted pALogit values for the designed molecules (A) Derivatives of molecules 6 (52–64) (B) Derivatives of molecules 26 (65–67) and (C) Derivatives of molecules 50 (68–79)
inline image
MoleculesSeriesPredicted activitySubstitution
Model 3aModel 4aR1R2R3

Results and Discussion


The synthesis of 4-(arylamino)coumarins as reported in the literature involves two steps. 4-Hydroxycoumarin is first converted to 4-chlorocoumarin and this is then converted to the respective amines (57). The 4-chlorocoumarins are strong skin irritants with several toxic effects, so even a small-scale preparation is environmentally hazardous. With the use of microwaves, it is possible to carry out this reaction in a single step with a drastic reduction in reaction time. This facile method has been optimized to prepare a small library of 50 compounds reported in this paper. The use of microwaves represents a far simpler and quicker mode of synthesis of 4-(arylamino)coumarins over methods reported in the literature. The method involves irradiation of the reactants, the 4-hydroxycoumarin derivatives, and the aryl amines in appropriate proportion, with 400 W microwave power at 140 °C in the absence of a solvent for an appropriate time (Schemes 1, 2 and 3). All reactions were carried out in a monomode open-vessel microwave synthesizer. The structures of the synthesized molecules were confirmed by 1H NMR, IR spectra and mass spectra; details are given in supplementary data (Table S1).

Antitubercular activity

The compounds were screened against M. tuberculosis H37Rv strain using the BECTEC 460 radiometric system. The results are summarized in Table 1 with rifampin, as the standard. Of the fifty compounds, compound 9 is the most potent showing 100% inhibition of M. tuberculosis at 6.25 μg/mL, while compound 6 is moderately active with 81% inhibition at the same concentration. It is interesting to note that both bear a bulky naphthyl group at the 4-position of the coumarin ring. The benzenoid part of the coumarin skeleton in 9 is substituted with a 6-chloro and a 7-methyl group, while in case of 6, it is unsubstituted. In contrast, compounds with methyl substitutions at both the 5- and 8-positions on the coumarin ring (e.g., compounds 7, 12 and 13) are nearly inactive. Compound 2 with a biaryl functionality at the 4-position shows 65% inhibition and is more potent than closely related analogues. When the biphenyl derivatives are additionally substituted with methyl groups on the coumarin ring, the activity falls drastically e.g., compounds 3 and 4. No worthwhile activity was observed in the series possessing an electron withdrawing –COOH group (molecules 14–20). The activity of compounds bearing a halogen substituent in the aromatic ring joined at the 4-position of the coumarin ring is variable with no clear-cut pattern. In the phenothiazine series of compounds (molecules 46–50), compounds 50 and 48 show 78% and 67% inhibition, respectively. From the activity range, it can be concluded that the presence of the naphthyl group enhances potency, while methyl substituents on the coumarin ring diminishes potency. Because of the very interesting pattern of activity shown by these novel compounds, the molecules were subjected to CoMFA studies to probe the fine relationship between structure and activity that exists for these compounds.

3D-QSAR study

The most interesting feature in these molecules is the geometry of the 4-amino group. In an ideal situation, it must have a planar trigonal geometry as it is backed by resonance to an aromatic system as well as to the coumarin carbonyl group. However, preliminary molecular mechanics calculations showed lack of planarity as the hydrogen atom attached to the 4-amino group experiences severe steric clash with the C5 hydrogen. To gain a better perspective of the correct geometry, ab initio calculations were run on molecule 21 with the Gaussian software.b Calculations were run on structures with both pyramidal and planar trigonal geometries for the 4-amino group. The calculations revealed that the N-atom adopts a state that is between these two ideal geometries. As no such atom type has been defined in the Tripos force field, the atom type for the N-atom was set to be ‘N3’ i.e., sp3-pyramidal.

The next issue was to fix the geometry at the N-C4 bond between the amine N and the C4 atom on the coumarin ring (Figure 1). As mentioned previously, certain molecules have an S-bridge between the aromatic ring and the 3-coumarin position. The activities of these constrained cyclic amines (e.g., molecule 50) are higher than the corresponding non-cyclic amines; this indicates that the conformation attained by such systems must be considered in building the conformation of the other molecules in the series. This fact was used to fix the conformation around the N-C4 bond for the open chain molecules. The constrained arylaminocoumarins were first subjected to complete minimization, and the final conformations achieved for these molecules were adopted for the other non-cyclic arylaminocoumarins. Further, Systematic conformational search was executed on representative molecules to derive the optimal conformation to be used for the CoMFA studies.

Comparative molecular field analysis alone and in conjunction with HINT hydropathy fields was carried out. CoMFA models 3a and 4a with additional HINT fields were the most robust, with correlation coefficients (r2) of 0.987 and 0.983 and cross-validated r2 (q2) of 0.578 and 0.556, respectively (Table 2). Both these models were tested for their predictive ability on the validation set. The r2pred for the various models are presented in Table 2 and the best value is 0.522, indicating that most models have decent predictive powers. Further the robustness of the models was gauged by cross-validation using bootstrapping (50–52) carried out with 100 runs. The values of r2BS are close to the related r2 values. To check the redundancy and chance correlation, the activity values were randomized (Y-scrambling) (53–55) and the regression coefficients were calculated for models 3a and 4a. These r2YS calculated on randomized activity were significantly smaller (0.377 and 0.305 for models 3a and 4a, respectively) indicating low degree of redundancy in the training set.

Molecules 8, 21 and 39 were found to be outliers; their activities were not well predicted by any of the models. One of the reasons for the poor prediction may be an inadequate consideration of the conformation of these molecules in the development of the models.

Incorporation of the HINT hydropathy field to the CoMFA models gave a significant improvement in the quality of the models. The statistical data in Table 2 indicate that the hydrophobic property along with steric and electrostatic fields is necessary to explain the activity of these molecules. The contribution of the HINT hydropathy field was seen to be equal to the electrostatic field.

Graphical interpretation of the results

The CoMFA contour maps were generated using scalar products of coefficients and standard deviation (STDEV*COEFF) set at 80% and 20% for favored and disfavored levels, respectively.

Analysis of the contours for CoMFA models

The CoMFA contours (Figures 4, 5, and 6) were analyzed around the most active, moderately active, and the least active molecules 50, 3 and 26, respectively. The steric contours indicate bulky groups attached to the 7-position on the aromatic ring of the coumarin ring are strongly favored. The C6-position on the coumarin ring should be unsubstituted as indicated by the small yellow contour near this position (Figure 6). This is aptly borne by the fact that the alkyl and aryl groups decrease the activity of molecules 3 and 26, respectively compared to molecule 50, which does not have a substituent at the C6-position (Figure 5).

Figure 4.

 Comparative molecular field analysis electrostatic contours displayed around molecule 50.

Figure 5.

 Comparative molecular field analysis contours displayed around molecules 3 and 26.

Figure 6.

 Comparative molecular field analysis steric contours displayed around molecule 50.

The electrostatic contours (Figure 4) depict that electronegative groups on the aromatic ring will improve activity. For molecules with a phenothiazine ring (e.g., molecule 50), electronegative groups are favored at the meta-positions to the amino functionality i.e., 5′ and 7′-positions, while the para-position i.e., 6′-position needs an electropositive group to improve activity. The same pattern is followed in the case of molecules with N-phenyl groups. It is interesting to find that electronegative group para to the amino group have a detrimental effect on the activity and indeed this position needs to be filled with electropositive groups. This is the main reason why molecules 3 and 26 with electronegative carboxyl and fluoro groups have greatly diminished activities.

As mentioned earlier, some 3D-QSAR studies have been carried out on 4-aminocoumarins with regard to antiplatelet activity (39,40). One of these studies was carried out on a diverse set of 2-/4-aminocoumarins, coumarins, and benzocoumarins substituted with 1-piperazine, and pyrimidine, pyrido[1,2-a]pyrimidine and thiazo[1,2-a]pyrimidines analogues. The difference between these molecules and the present series is the presence of an aromatic group on 4-amino functionality. For antiplatelet activity, bulky substituents are required at the 7- and 8-positions of the coumarin ring but are disfavored in the 4′-position of the 4-piperazine group. In strong contrast, for antitubercular activity, bulky groups are preferred at the 7-position, and further, this site seems to be encompassed by a large hydrophobic pocket. Similarly, comparison of the electrostatic contours reveals a favor for electronegative groups at the 4-position for antitubercular activity. Thus, there are clear distinctions between the steric and electrostatic attributes in 4-aminocoumarins that separate antiplatelet from antitubercular activity.

Design of new molecules

The end purpose of all medicinal chemistry endeavors is to design and synthesize better molecules. The molecules should not only be potent but also have a good safety profile. Thus, it is necessary to consider ADMET properties right at the design stage itself. With this idea, around 913 molecules were designed and their activity predicted using the best CoMFA models. The 135 molecules predicted to be highly active were then screened for various ADMET properties like aqueous solubility (QPlogS), partition coefficient in different solvents/media (QPlogPoct, QPlogPw, QPlogPo/w, CIQPlogS), partition coefficient in biological fluids, and membranes like skin and Blood Brain Barrier (BBB), apparent Caco-2 cell permeability, MDCK cell permeability, binding to human serum albumin, metabolism, human oral absorption, IC50 value for blockage of HERG K+ channels, and central nervous system activity. For details of the essential ADMET properties, the reader is referred to the supplementary material. A few molecules with the best activity and the satisfactory ADMET profile are enlisted in Table 3. Their respective ADMET data (Table S3 in supplementary data) show that most molecules have satisfactory drug-like attributes. Toxicity data show that only 28 molecules are predicted to be non-carcinogenic and six were devoid of mutagenic effects.


A small library of 4-(arylamino)coumarins was synthesized using microwave-assisted synthesis in good yields and short reaction times. All compounds were screened for activity against M. tuberculosis H37Rv strain. A 3D-QSAR analysis of this library of compounds was carried out using CoMFA alone and CoMFA in conjunction with a hydrophobic field evaluated with HINT to map the structural features responsible for antitubercular activity. Inclusion of the HINT hydropathy field to the CoMFA fields improved the quality of the models. Models obtained by field-fit alignment were better in terms of statistics than those with database/atom-fit alignment. Analysis of the CoMFA contours provided details on the fine relationship linking structure and activity and offered ideas for structural modifications that can improve the activity. There is still a good scope for optimization of these molecules.


  • a

    Sybyl version 7.1 Tripos Inc. USA.

  • b

    Gaussian 03, Gaussian Inc., Wallingford CT, USA.

  • c

    CoMFA and QSAR Manual S, Associates Inc., 1699 S Hanley Rd., St. Louis, MO 631444, USA.

  • d

    QuickProp. Schrodinger LLC.


The authors are thankful to DST (program FIST, Sanction No. SR/FST/CSI-072/2003, Dt. 24/12/2003) and to UGC (SAP program, Sanction No. 540/6/DRS/2004, SAP-I, Dt. 26/03/2004) for their generous financial and instrumentation support. Special thanks are due to ‘National Facility for Drug Discovery through New Chemical entities (NCE’s) Development & Instrumentation Support to Small Manufacturing Pharma Enterprises’ Program under Drug & Pharma Research Support (DPRS) jointly funded by Department of Science & Technology (DST, sanction letter no. VI/D&P/188/06-07/, TDT, dated 30/03/07) New Delhi, Government of Gujarat Industries Commissionerate (Sanction Letter No. IC/SSI/R&D/SU/07/3279/168, dated 17/01/07) & Saurashtra University (Sanction Letter No.PLG/UGC/MS/238/20006, dated 19/09/06), Rajkot. The authors are also grateful to Government of Gujarat for PG Diploma’s for Instrumentation & Patentization Financial Grant (Sanction Letter KVT/UNI/CHABA/2008/20690-207).

Authors are also thankful to the Dr. Cecil Kwong, Tuberculosis Antimicrobial Acquisition and Coordinating Facility (TAACF), USA, for antitubercular screening. The computational facilities at BCP were provided by the All India Council for Technical Education through grant F. No. 8022/RID/NPROJ/RPS-5/2003-04/. The Department of Science and Technology is also thanked for infrastructure facilities through the FIST program (SR/FST/LS1-163/2003).