Pharmacophore and docking-based hierarchical virtual screening for the designing of aldose reductase inhibitors: Synthesis and biological evaluation

MEDICINAL

CHEMISTRY

RESEARCH

Med Chem Res

DOI 10.1007/s00044-016-1510-5

ORIGINAL RESEARCH

Pharmacophore and docking-based hierarchical virtual screening

for the designing of aldose reductase inhibitors: synthesis

and biological evaluation

Bhawna Vyas¹Manjinder Singh²Maninder Kaur²

•

Om Silakari²Malkeet Singh Bahia²Baldev Singh¹

•

Received: 10 July 2015 / Accepted: 8 January 2016

Ó Springer Science+Business Media New York 2016

Abstract A set of 54 studied ﬂavonoid inhibitors of

aldose reductase (ALR2) enzyme has been utilized for

pharmacophore modeling and 3D-QSAR analysis using

Keywords Pharmacophore mapping Á

Pharmacophore-based 3D-QSAR Á Docking analysis Á

Aldose reductase (ALR2) Á Flavonoid derivatives Á

Virtual screening Á Diabetic complications

¨

‘‘PHASE’’ program of Schrodinger software. The gener-

ated pharmacophore model (AADRR.1109) was chal-

lenged to screen ‘‘PHASE’’ database to identify new ALR2

inhibitors. The retrieved hits were employed for docking

analysis and pharmacokinetic parameter calculation to

obtain orally active molecules. To predict the activity of

ﬁnal retrieved hits, 3D-QSAR model was developed, and

the best model was selected on the basis of various sta-

tistical parameters (R²_train0.719; Q_t²_est0.647 and SD 0.663).

Totally ﬁve screened molecules which showed better

enhanced predicted activity were synthesized and evalu-

ated for in vitro ALR2 inhibitory activity. All tested

molecules showed ALR2 inhibitory activity (IC₅₀) below

40 lM. Additionally, the free radical scavenging potential

of synthesized molecules was also determined which

played a useful role to control the progression of diabetic

complications. All molecules showed good antioxidant

potential, thus the designed molecules, in future, could be

explored to ameliorate the development of diabetic

complications.

Introduction

The diabetes mellitus is one of the most common metabolic

disorders characterized by chronic hyperglycemia (Wild

et al., 2007). The sustained chronic hyperglycemic condi-

tion in diabetes can result in various cardiovascular, renal,

neurological, and visual complications (Brownlee, 2001).

Even though various treatment strategies are there to treat

or to manage the diabetes by maintaining the blood glucose

level, but pharmacologic maintenance of normoglycemia

in diabetes cannot prevent the eventual complications

mainly due to protein glycation-induced cell death, dys-

regulated antioxidant defense, and signal transduction in

affected tissues (Giannoukakis, 2014). Besides these,

polyol pathway ﬂux, stimulation of protein kinase C iso-

mers and hike in hexosamine pathway ﬂux have also been

involved in diabetic complications.

Aldose reductase (ALR2) is rate-limiting enzyme impli-

cated in polyol pathway (Fig. 1) which has been well-ex-

plored among all reported molecular mechanisms of diabetic

complications (Kinoshita, 1990). ALR2 is distributed

throughout the tissues of the human body and is being

implicated in the etiology of long-term diabetic complica-

tions (Alexiou et al., 2009; Kador, 1998). It normally causes

NADPH-dependent reduction of glucose to sorbitol, which

accumulates and leads to excessive osmotic and oxidative

stress resulting in the destruction of cells (Giacco and

Brownlee, 2010; Srivastava et al., 2005). Subsequently, the

NAD^?-dependent sorbitol dehydrogenase oxidizes the

Electronic supplementary material The online version of this

article (doi:10.1007/s00044-016-1510-5) contains supplementary

material, which is available to authorized users.

& Baldev Singh

bhawnaom@rediffmail.com

1

Department of Chemistry, Punjabi University, Patiala,

Punjab 147002, India

2

Molecular Modeling Lab (MML), Department of

Pharmaceutical Sciences, Drug Research Punjabi University,

Patiala, Punjab 147002, India

123

Med Chem Res

Fig. 1 Polyol pathway

sorbitol to fructose. But, in diabetic condition, level of glu-

cose rises and reduction of glucose to sorbitol increases as

compared to oxidation of sorbitol to fructose (Kinoshita

et al., 1976). The over production of sorbitol, imbalance of

NADPH/NADP^?and NAD^?/NADH cofactors, and subse-

quent oxidative stress inside the cells are thought to be main

causes of cellular damage in diabetic complications. In

addition to it, advanced glycation end products (AGEs) are

also produced as a result of high blood levels of fructose

which can cause cellular damage (Szwergold et al., 1990;

Gonzalez et al., 1986; WellsKnecht et al., 1996). These

abnormal metabolic effects are the key factors which led to

diabetic complications such as cataracts (Robinson et al.,

1983; Kador et al., 2000), retinopathy (Engerman, 1989),

neuropathy (Young et al., 1983), and nephropathy (Dunlop,

2000). Thus, inhibition of ALR2 can be targeted as a possible

treatment of diabetic complications (Varma and Kinoshita,

1976).

of ROS may be more effective than ALR2 inhibitors alone

in the treatment of diabetic complications.

Flavonoids, being called vitamin P¹⁶, have shown ben-

eﬁcial effects in diabetes mellitus and associated compli-

cations (Havsteen, 2002; Varma, 1986). Moreover, these

are relatively simple to synthesize with slight toxicity which

make these potential dose for several therapeutic activities

and as chemopreventive agents (Moon et al., 2006). These

can possibly act in two ways either by increasing the

peripheral use of glucose or by boosting insulin sensitivity

in diabetic animals (Xie et al., 2006). Since the mid 1970s,

numerous studies on the inhibition of aldose reductase by

ﬂavonoids have been reported (Varma et al., 1975; Varma

and Kinoshita, 1976; Okuda et al., 1982; Mercader et al.,

2008). Additionally, it has been reported that antioxidant

action of ﬂavonoids has been attributed to their free radical

scavenging property or their inﬂuence on the intracellular

redox status (Rice-Evans, 2001).

Additionally, the involvement of oxidative stress in

diabetic complications, both microvascular and cardiovas-

cular, has been well established. The abnormal metabolic

processes in diabetes cause mitochondrial superoxide

overproduction that subsequently triggers ﬁve major path-

ways implicated in the pathogenesis of complications:

polyol pathway ﬂux, AGEs formation (advanced glycation

end products), increased expression of the RAGEs (Re-

ceptor for advanced glycation end products) and its acti-

vating ligands, activation of protein kinase C isoforms, and

the hexosamine pathway. Thus, targeting formation of

reactive oxygen species (ROS) might also be a potential

therapeutic strategy to improve diabetic complications

(Araki and Nishikawa, 2010).

Most of the pharmaceutical companies pay their atten-

tion to speed up the rational and cost-effective drug dis-

covery. Computer-aided drug design (CADD) approaches

can be employed for the prediction of the biological

activity of compounds by correlating with their structural

properties and can also be utilized to design new active

molecules. In this sense, quantitative structure–activity

relationship (QSAR) studies have been successfully

applied for modeling biological activities of natural and

synthetic chemicals. QSAR studies have been carried out

for modeling activities of several kinds of ALR2 inhibitor.

Some recent reports have linked structural features of the

ligands with their ALR2 inhibition by using topology

indexes (Mercader et al., 2008; Prabhakar et al., 2006)

three-dimensional (3D)-QSAR methodologies (Liu et al.,

2009; Aggarwal et al., 2010), CoMFA, CoMSIA analysis

In the light of above, it can be expected that ALR2

inhibitors along with capacity of preventing the formation

123

Med Chem Res

and artiﬁcial neural networks (Fernandez et al., 2005; Patra

and Singh, 2009; Hu et al., 2006), etc.

is capable of reproducing the native crystal conformations

of re-docked seven crystal ligands with minimum all atom

deviation as compared to other protein crystal structures.

Therefore, it was hypothesized that in the absence of 3D

structure co-crystallized with ﬂavone molecules, 1US0

would produce precise docking poses of the studied ﬂavone

derivatives and unknown molecules since it can signiﬁ-

cantly reproduce the native crystal structures of the

experimentally determined seven crystal ligands.

The current study involves the development of an in

silico model to predict the ALR2 inhibitory activity of a set

of 54 ﬂavonoids (Varma et al., 1975). Structures of these

compounds are mentioned in Table 1, and the common

scaffold is displayed in Fig. 2. Caballero in 2010 reported

3D-QSAR ﬂavonoids by CoMFA and CoMSIA and Phar-

macophore by Genetic Algorithm with Linear Assignment

of Hypermolecular Alignment of Database (GALAHAD)

(Caballero, 2010). In the present paper, we reported phar-

macophore- and docking-based hierarchical virtual

screening for the identiﬁcation of new molecules for ALR2

inhibitory activity. Among the retrieved hits, ﬁve mole-

cules have been synthesized and evaluated for ALR2

inhibitory activity, and antioxidant potential.

Identiﬁcation of new hits: pharmacophore modeling,

3D-QSAR analysis and hierarchical virtual

screening

Generation of pharmacophore and 3D-QSAR model

Pharmacophore modeling analysis was performed in

‘‘PHASE’’ program (Phase 3.4, 2012), using the sampled

conformational space of all dataset molecules with the

energy window of 20 kcal/mol. For this purpose, the

molecules with pIC₅₀value [7 and \3.7 were considered

as actives and inactives, respectively, while the remaining

ones as moderate actives. This included total 11 active and

5 inactive molecules. For the elucidation of common

pharmacophore models, two restrictions were applied to

the software, i.e., generate a common pharmacophore

hypothesis with minimum of four and maximum of ﬁve

structural features, and generated model should match all

eleven active molecules. The generated set of pharma-

cophore models was clustered, and the representative

pharmacophore of each cluster was selected based on high

S–I value (Table 3). The selected ﬁve different pharma-

cophore models were duly employed for the alignment of

dataset molecules, and thereafter, the obtained molecular

alignment was used to develop 3D-QSAR regression

model. To develop a 3D-QSAR model, whole dataset was

divided into training and test set of 30 and 24 molecules,

respectively. The training set molecules were used to

generate a series of 3D-QSAR models. Among the gener-

ated 3D-QSAR models, the best model (AADRR.1109₃)

was selected corresponding to pharmacophore model

AADRR.1109 and PLS factor three based on the highest

value of Q²_test(0.647).The selected 3D-QSAR model

AADRR.1109₃also displayed high values of R²_train(0.719),

F(22.3) and low value of SD (0.663). The other statistical

parameters of the selected 3D-QSAR model are listed in

Table 4, and the predicted activities of training set mole-

cules based on the selected model are described in Table 5.

The selected pharmacophore model AADRR.1109₃, based

on best results of 3D-QSAR model, consisted of ﬁve

structural features including two hydrogen bond acceptors,

one hydrogen bond donor and two ring aromatic features.

The stereoview of selected pharmacophore model, and its

Results and discussion

Pharmacophore modeling and subsequent virtual screening

using pharmacophore model is a promising computational

technique for identifying new lead molecules that demon-

strates desired biological activity on a validated molecular

target. In the present study, pharmacophore model for ﬂa-

vone-based ALR2 inhibitors was developed using

‘‘PHASE’’ program of Schrodinger software to identify

new lead molecules. The ﬂavones among the retrieved hits

were selected on the basis of their antioxidant potential,

which is beneﬁcial in maintaining reactive oxygen species,

in addition to their ALR2 inhibitory activity. Thus, this

dual action of selected ﬂavones may produce synergistic

effect in the treatment of diabetic complications.

Cross-docking experiment: selection of appropriate

protein for docking analysis

In the present analysis, docking simulations were employed

in two ways: one for the elimination of non-selective ALR2

hits (i.e., molecules docking with ALR1 protein) obtained

from pharmacophore screening and second for the deter-

mination of ligand–protein interactions of screened mole-

cules with ALR2 enzyme.

Since no crystal structure of ALR2 with ﬂavonoid

molecules is reported in protein data bank, the most

appropriate one was selected based on cross-docking

experiment among reported top seven high resolution 3D

structures co-crystallized with different ligands (Table 2).

The cross-docking experiment, using ‘‘XP mode of Glide’’

docking program, was performed with all seven crystal

structures, and the results showed the least average RMSD

˚

value of 1.27 A for protein structure 1US0 (Howard et al.,

2004). This analysis showed that the crystal structure 1US0

123

Med Chem Res

Table 1 Molecular structures and corresponding experimental biological activity values of ﬂavonoid ALR2 inhibitors

Sr. No.

3

5

6

7

8

2⁰

3⁰

4⁰

5⁰

pIC^a₅₀

1

OCH₃

OH

OCH₃

OH

OCOCH₃

OH

OCH₃

7.553

7.490

7.456

7.470

7.410

7.350

7.240

7.190

7.130

7.110

7.040

6.920

6.850

6.796

6.790

6.770

6.690

6.770

6.640

6.620

6.600

6.570

6.550

6.520

6.458

6.390

6.270

6.090

6.070

5.920

5.850

5.350

5.200

5.170

5.140

5.090

5.076

5.050

4.730

4.530

4.340

3.960

3.540

3.500

2

OCH₃

OH

OCH₃

OH

OCH₃

CH₂Ph

OCH₃

3

4

OCH₃

5

OH

OCH₃

OH

6

OCH₃

OH

7

OCH₃

OH

8

OH

OCH₃

OH

OCH₃

9

OCH₃

OH

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OCOCH₃

OH

OCOCH₃

OH

OCOCH₃

OCH₃

OH

OCOCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

COCH₃

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

OCH₃

OH

123

Med Chem Res

Table 1 continued

Sr. No.

3

5

6

7

8

2⁰

3⁰

4⁰

5⁰

pIC^a₅₀

49

50

51

52

53

54

a

OCH₃

OH

3.000

5.780

5.640

5.280

6.456

OCH₃

OH

OCH₃

OH

OCH₃

OH

-log IC₅₀; where IC₅₀represents the molar concentration of compound required to produce 50 % inhibition of Aldose Reductase enzyme

(Pearson-r) of the test set was observed to be 0.802. The

correlation graph between the experimental and predicted

activities of test set molecules is displayed in Fig. 5. The

similarity threshold of test set molecules was also observed

to be within the ‘‘APD’’ of AADRR.1109₃that described

the reliability of selected model to predict new molecules

(Table 7). The robustness of selected model, analyzed by

performing ‘‘Y-randomization’’ test, described low value

of R²_scramble(0.629) as compared to R_t²_rainof the selected

Fig. 2 Common template of phenyl-benzopyrene present in dataset

molecules

model, which described the real regression correlation of

the selected model. In addition, the calculated values of

2

mapping over the highest active (molecule 1; violet) and

least active (molecule 50; green) molecule are displayed in

Figs. 3 and 4, respectively.

k (1.003), k⁰(0.983), R₀²(0.998) and R⁰₀(0.999) were also

within the recommended range that strengthen the external

predictive ability of the selected 3D-QSAR model.

Validation of selected pharmacophore and 3D-QSAR

model

Hierarchical virtual screening protocol

A systematic hierarchical virtual screening protocol was

implemented for the screening of ‘‘PHASE’’ database to

identify ALR2 inhibitors employing pharmacophore model

screening, docking simulations and in silico pharmacoki-

netic properties calculation (Fig. 6). Initially, ‘‘PHASE’’

database was screened with selected pharmacophore model

AADRR.1109₃that retrieved 1150 molecules bearing the

structural features of selected pharmacophore model. These

1150 molecules were subsequently docked into the active

site of protein ALR1 (3H4G), and molecules binding into the

active site of this protein were removed from further anal-

ysis. The molecules surviving after ALR1 docking, i.e., 923

were subjected to ‘‘QikProp’’ program (Qikprop 3.5, 2012)

for the prediction of their in silico ‘‘drug likeness’’ and

pharmacokinetic proﬁle. In ‘‘QikProp,’’ partition coefﬁcient

(QPlogPo/w), water solubility (QPlogS), number of hydro-

gen bond donor groups (HBD), number of hydrogen bond

acceptor groups (HBA), molecular weight (MW) and per-

centage human oral absorption parameters of these mole-

cules were determined. After this screening, the molecules

complying with ‘‘Lipinski rule of ﬁve’’ for ‘‘drug likeli-

ness,’’ i.e., molecular weight \500, HBD \ 5, HBA \ 10

and QPlogP_o/w\ 5, and within the acceptable range of

pharmacokinetic parameters were selected for further

The selected models are theoretical, and can not be used as

such; thus, both models were validated based on different

recommended statistical calculations. A good pharma-

cophore model should be able to identify active molecules

and reject inactive molecules, while the good 3D-QSAR

model should be reliable to predict the activity of new

molecules. Therefore, selected pharmacophore model was

validated to check its ability to identify known actives from

the dataset of known inactive molecules, i.e., ‘‘GH score’’

and ‘‘enrichment’’ calculation. For this purpose, the

selected model was challenged to screen the dataset of

known active (A, 43) and inactive molecules (638). The

selected model retrieved 51 hits (Ht) including 39 known

actives (Ha, 76.47 % retrieval). The ‘‘enrichment’’ of this

screening protocol was calculated to be 12.11 that indi-

cated selected model has 12 times more ability to identify

active molecules than inactives. The ‘‘GH score’’ of this

screening protocol was calculated to be 0.785 ([0.7),

which indicates the signiﬁcance of selected pharmacophore

to identify active molecules (Table 6).

On the other hand, selected 3D-QSAR model was val-

idated for its ability to correctly predict the known mole-

cules that were not included for the generation of QSAR

model, i.e., test set. The Pearson coefﬁcient of correlation

123

Med Chem Res

Table 2 Results of cross-docking analysis for selected seven crystal structures of ALR2

Ligand from PDB

Ligand structure

Protein structures

1US0

1PWM

1Z8A

2PZN

RMSD between crystal ligand and re-docked structure (A)

3UZC

˚

1Z3N

1T41

3.50

Protein data bank codes

1US0

0.26

3.68

0.64

0.67

1.38

0.41

1.87

1.27

6.50

0.40

7.10

6.42

4.90

6.50

5.56

3.80

0.93

4.67

4.38

2.90

4.95

5.04

3.81

2.30

1.61

3.34

0.12

1.67

0.49

0.59

1.44

4.80

1.41

6.20

6.03

0.25

5.70

9.50

4.84

2.40

1.00

0.71

0.82

3.70

0.35

1.01

1.42

1PWM

1Z8A

2PZN

3U2C

1Z3 N

1T41

0.8 6

1.24

0.25

4.10

0.37

0.20

1.50

˚

Average RMSD (A)

Bold value indicates the lowest value of average RMSD

a

Root mean square deviations between ligand crystal structure and re-docked structure

analysis. ‘‘QikProp’’ ﬁlter retrieved 521 molecules, which

were docked into the active site of validated crystal structure

of ALR2 (1US0) using XP precision mode of ‘‘Glide’’

program. Finally, 18 molecules bearing structural features of

selected pharmacophore model, having acceptable in silico

pharmacokinetic properties and displaying signiﬁcant

ligand–protein interaction with ALR2 but not with ALR1,

were retrieved. The molecular structures of these 18 com-

pounds along with their calculated pharmacokinetic proﬁle

and predicted activity based on selected QSAR model are

depicted in Table S (Supplementary information).

Synthesis and biological evaluation

Out of the retrieved 18 molecules, ﬁve ﬂavonoid deriva-

tives were synthesized based on their synthetic ease. The

detailed description of the synthetic protocols is mentioned

in Schemes 1 and 2. Thereafter, the synthesized molecules

123

Med Chem Res

Table 3 PHASE scoring parameters for generated pharmacophore hypotheses

Hypothesis ID

Survival

S–I^a

Site

Vector

Volume

Matches

AADRR.1109

ADRRR.76

ADDRR.80

DDRRR.61

AADR.1090

a

62.188

62.266

62.165

62.268

62.194

60.756

60.481

60.508

60.427

60.420

0.79

0.89

0.77

0.89

0.86

0.980

0.969

0.976

0.970

0.972

0.779

0.769

0.778

0.769

0.721

11

S–I survival minus inactives that corresponds to score inactives and higher the value of S–I indicates that the generated hypothesis is more

likely to pick active molecules than inactives

Table 4 Statistical results of the selected 3D-QSAR regression

model (AADRR.1109₃)

comparatively good scavenging potential, i.e., 21.0, 22.1,

Parameter

Values

and 24.5 lM respectively (Fig. 8).

PLS Factor

3

Standard deviation (SD)

Coefﬁcient of determination for training set molecules (R_t²_rain

Fisher test (F value)

0.663

0.719

22.3

Conclusion

)

Herein, a statistically reliable and predictive pharmacophore

model was developed with ﬁve structural features and was

validated rigorously. The selected pharmacophore model

was used to align the dataset molecules and to develop a

predictive 3D-QSAR model. The predictive 3D-QSAR

model was statistically validated based on various recom-

mended criteria. The validated pharmacophore model was

employed for the screening of ‘‘PHASE’’ database, and

retrieved molecules were docked with ALR1 and ALR2

proteins to identify selective ALR2 hits. For the identiﬁed

new hits, in silico pharmacokinetic properties were also

calculated, and at the end 18 molecules were selected.

Among ﬁnal 18 molecules, ﬁve ﬂavone derivatives were

selected for synthesis based on their synthetic feasibility.

Thereafter, the synthesized molecules were tested for their

in vitro ALR2 inhibitory potential. Since clinical reports

described the beneﬁcial potential of antioxidants in the

management of diabetic complications, all synthesized

molecules were also tested for their free radical scavenging

ability. All molecules resulted in good ALR2 and free rad-

ical scavenging activity thus could be used for the man-

agement of diabetic complication in future.

Stability

0.651

0.705

0.647

0.802

Root means squared error (RMSE)

Coefﬁcient of determination for test set molecules (Q_t²_est

)

Pearson coefﬁcient of correlation for test set molecules

(Pearson-r)

were tested for their in vitro ALR2 inhibitory potential. In

ALR2 inhibitory assay method, all molecules displayed

good micromolar activity (IC₅₀, \40 lM) (Table 8), and

compound 5a and 5b showed comparatively high inhibi-

tory activity, i.e., IC₅₀value 9.76 and 8.33 lM, respec-

tively. The graph displaying predicted and experimental

activities of the designed molecules is mentioned in Fig. 7.

Interestingly, it was observed that the ﬂavone deriva-

tives having electron withdrawing groups at ring-B display

micromolar ALR2 inhibitory activity and are compara-

tively more potent than other derivatives, e.g., compound

5a with ﬂuoro group substituted at meta position of ring-B

displayed the highest ALR2 inhibitory potential, i.e.,

IC₅₀= 9.76 lM.

The beneﬁcial role of antioxidant therapy in the man-

agement of diabetic complications is well evidenced from

clinical studies. The use of antioxidants is reported to

prevent and even reverse many early changes in the vas-

cular and neurological tissues of diabetic animals (Scott

and King, 2004). In the light of these reports, the free

radical scavenging potential of synthesized compounds was

also determined using in vitro 2,2-diphenyl-1-picrylhy-

drazyl (DPPH) assay method. The free radical scavenging

potential of molecules is expressed as IC₅₀values, which

corresponds to the concentration of molecules required to

produce 50 % scavenging of free radicals. All compounds

exhibited free radical scavenging activity in micromolar

range (Table 8), and compounds 5e, 5d and 5c showed

Materials and methods

Molecular dataset: selection and preparation

A dataset of 54 literature reported ALR2 inhibitors was

selected for molecular modeling (Varma et al., 1975). The

reported biological activity, i.e., pIC₅₀(-log IC₅₀) of the

selected molecules corresponds to negative logarithm of

the molar concentration of molecule required to produce

50 % inhibition of enzyme. The activity values of dataset

123

Med Chem Res

Table 5 Dataset of molecules used for the generation of pharmacophore model and 3D-QSAR models and their corresponding predicted activity

values

a

Sr. No.

Biological activity (pIC₅₀

)

Sr. No.

Biological activity (pIC₅₀

)

Experimental^b

Predicted^c

Experimental^b

Predicted^c

1^e

2^d,e

3^e

4^e

5^d,e

6^e

7^d,e

8^e

9^d,e

10^e

11^d,e

12

7.55

7.49

7.46

7.47

7.41

7.35

7.24

7.19

7.13

7.11

7.04

6.92

6.85

6.79

6.77

6.69

6.77

6.64

6.62

6.60

6.57

6.55

6.52

6.69

7.06

6.69

7.06

6.97

7.06

5.64

7.06

6.41

7.06

6.41

5.64

6.69

7.06

6.41

5.64

7.06

6.69

28

6.46

6.39

6.27

6.09

6.07

5.92

5.85

5.35

5.20

5.17

5.14

5.09

5.08

5.05

4.53

4.34

3.96

3.54

3.50

3.00

5.78

5.64

5.28

6.46

6.69

5.64

6.69

5.64

5.03

5.64

5.03

4.82

5.64

4.77

4.71

4.99

5.03

5.64

5.03

4.81

4.59

5.64

5.18

3.01

5.42

5.56

5.64

5.56

29

30^d

31

32^d

33

34^d

35

36

37

38

39^d

13^d

40

14

41^d

42

43^d

15^d

16

17

44

18^d

19^d

20

45^d

46^f

47^d,f

48^f

49^d,f

50^f

51^d

52

21^d

22

23

24

25^d

26

53^d

54^d

27^d

a

-log IC₅₀, where IC₅₀is molar concentration of compound required to produce 50 % inhibition of Aldose Reductase enzyme

b

c

d

e

f

pIC₅₀reported in the literature (Mercader et al., 2008)

Predicted pIC₅₀from pharmacophores-based QSAR model

Test set molecules for 3D-QSAR model (pharmacophore alignment based)

Active molecules for pharmacophore generation

Inactive molecules for pharmacophore generation

Generation of pharmacophore and 3D-QSAR model

molecules spanned over a wide range, i.e., 7.553–3.000 ([4

log order of magnitude), and for each molecule, activity

was determined using same experimental protocol. The

chemical structures of all selected molecules along with

their corresponding pIC₅₀values are displayed in Table 1.

The three-dimensional geometries of molecules were

drawn and cleaned in ‘‘builder tools’’ option of ‘‘Maestro’’

workspace (Maestro 9.3, 2012), and thereafter, energy

optimized in ‘‘LigPrep’’ program using ‘‘OPLS_2005’’

force ﬁeld. Furthermore, the ionic states of these molecules

were generated at pH value of 7.0 0.5 (Ligprep

2.5, 2012).

Pharmacophore modeling was performed using energy

optimized molecules in ‘‘PHASE’’ program (PHASE

3.4, 2012). The conformational space of each molecule

was sampled within an energy window of 20 kcal/mol,

using Monte Carlo Minimization Method (MCMM)/Low

Mode Docking (LMOD) algorithm. Thereafter, the sam-

pled conformational space of each molecule was utilized

for the development of pharmacophore models. ‘‘PHASE’’

generates various different pharmacophore hypotheses

which were ranked based on various software-dependent

123

Med Chem Res

Fig. 3 The stereoview of

selected pharmacophore

hypothesis AADRR.1109

scoring parameters including survival (S) and survival

minus inactives (S–I) score. Survival score corresponds to

score actives and S–I to score inactives, where the phar-

macophore model with high value of score actives corre-

spond to model with high ability to identify active ligands,

and the model with high value of S–I to identify active

from inactives, i.e., discrimination between active and

inactive molecules. Various generated pharmacophore

models were clustered, and the representative model was

selected from each cluster based on high S–I value. Phar-

macophore model clustering was done to select signiﬁ-

cantly different models and to decrease the probability of

selecting similar models from generated series of models

(Kaur et al., 2013). The selected pharmacophore models

were used to align dataset molecules and to generate a

predictive 3D-QSAR model. For the generation of 3D-

QSAR model, ‘‘t-value’’ and grid spacing was set as less

Fig. 4 Mapping of AADRR.1109 over the highest active molecule 1

(violet colored compound with ﬁt value 3.00) and the least active

molecule 50 (Green colored compound with ﬁt value 1.23) (Color

ﬁgure online)

˚

Table 6 Statistical results of GH score and enrichment calculations

for selected model AADRR.1109

than 2 (\2) and 1 A, respectively. The PLS factors were

incorporated not more than 1/5th the number of molecules

in training set. The best 3D-QSAR model was selected

based on the highest value of Q²_test(coefﬁcient of prediction

of test set molecules). The selected model must also

describe high values of R²_train(coefﬁcient of prediction for

training set molecules) and F-value, but low standard

deviation (SD) (Kaur et al., 2013).

Parameters

Values

Total molecules in database (D)

Total number of actives in database (A)

Total hits (Ht)

681

43

51

Active hits (Ha)

39

% Yield of actives [(Ha/Ht) 9 100]

% Ratio of actives [(Ha/A) 9 100]

Enrichment factor (E) [(Ha 9 D)/(Ht 9 A)]

False negatives [A - Ha]

False positives [Ht - Ha]

Goodness of hit score (GH)^a

a

76.47

90.69

12.11

4

Validation of selected pharmacophore and 3D-

QSAR models

11

The selected pharmacophore and 3D-QSAR models are

theoretical thus need rigorous validation. Accordingly, the

selected pharmacophore model was validated by calculat-

ing ‘‘goodness of hit’’ (GH score) (Henry et al., 1997) and

0.785

[(Ha/4HtA) (3A ? Ht)) 9 (1 - ((Ht - Ha)/(D - A))]; GH score

of [0.7 indicates a statistically good model

123

Med Chem Res

Fig. 5 Correlation graph

between experimental and

predicted activities of training

set and test set molecules from

AADRR.1109₃

Table 7 Applicability domain for test set molecules of pharma-

cophore-based 3D-QSAR model

Compound Distance

(APD = 3.5)

Compound Distance

(APD = 3.5)

2

3.1

30

32

34

39

41

43

45

47

49

51

53

54

1.7

5

2.0

2.2

2.4

1.7

2.4

2.0

3.3

2.4

1.7

2.0

1.7

2.2

2.0

1.7

3.0

1.4

2.6

0.0

7

9

11

13

15

18

19

21

25

27

‘‘enrichment factor’’ (Nagarajan et al., 2005), while the

selected 3D-QSAR model was validated by calculating test

set prediction (Pearson-r), ‘‘applicability domain’’ (Kaur

et al., 2013; Zhang et al., 2006), R₀, R⁰₀, k and i⁰(Golbraikh

and Tropsha, 2002), and performing ‘‘Y-randomization’’

test (Kaur et al., 2013; Richard et al., 1988).

Fig. 6 Hierarchical virtual screening (HVS) protocol for identiﬁca-

tion of new lead molecules

The calculation of ‘‘GH score’’ and ‘‘enrichment factor’’

determines the ability of selected pharmacophore model for

the identiﬁcation of known active molecules that are spiked

into the dataset of known inactive molecules. For this

purpose, a speciﬁed number of known active molecules

were spiked in the dataset of known inactives, and the

selected pharmacophore model was challenged to screen

the total dataset. From the screening results, ‘‘GH score’’

and ‘‘enrichment factor’’ was calculated as [(Ha 7 4HtA)

(3A ? Ht)) 9 (1 - ((Ht - Ha) 7 (D - A))] and [(Ha 9

D)7(Ht 9 A)], respectively. Herein, Ha corresponds to

123

Med Chem Res

Scheme 1 Synthetic protocol

for the synthesis of designed

molecules (5a–d)

active hits retrieved, D to total molecules in database, A to

total number of actives spiked in database and Ht to total

hits retrieved upon screening of dataset (Kaur et al., 2013;

Henry et al., 1997; Zhang et al., 2006).

(observed vs. predicted activities) should be close to R².

2

Moreover, [(R²- R_o²)/R²] and [R²- R⁰_o/R²] should be

\0.1, and the corresponding slopes of regression lines with

2

R²_oand R⁰_ocorrelation coefﬁcient, i.e., k and k’ should be

On the other hand, for the validation of selected 3D-

QSAR model, ‘‘applicability domain’’ (APD) was calcu-

lated to describe the reliability of selected model to predict

new molecules. For ‘‘APD’’ calculation, similarity mea-

surements were calculated based on Euclidean distances

between all pairs of training and test set compounds sep-

arately (Kaur et al., 2013; Henry et al., 1997). Initially,

average (l) of all Euclidean distances between the training

set compounds were calculated, and thereafter new average

\d[ and standard deviation (r) of training set molecules

with distances lower than the value of l were computed. Z

is an empirical cutoff with default value of 0.5.

between 0.85 and 1.15 (0.85 B k, k⁰B 1.15) (Golbraikh

and Tropsha, 2002).

‘‘Y-randomization’’ test was employed to determine the

regression quality of generated model. For this, the activity

data of training set molecules was scrambled to generate

different new training sets which were used to determine

R²_trainvalues (Kaur et al., 2013; Richard et al., 1988). An

average value of R²_trainobtained from scrambled training sets,

denotedbyR²_scramble, mustdescribelowvalueincomparisonto

original R²_trainvalue of the selected model; otherwise, the

selected model would be meaningless and generated by

chance.

Finally, the selected and duly validated pharmacophore

model and 3D-QSAR model were employed for the

screening of ‘‘PHASE’’ database of about 1.5 million

compounds and to predict the biological activity of

screened hits, respectively. The retrieved hits were sub-

jected further to docking simulations to remove false-pos-

itive hits identiﬁed by pharmacophore model, and

remaining molecules (screening hits minus false negatives)

were analyzed in ‘‘QikProp’’ program for the determination

of in silico pharmacokinetic properties (Qikprop 3.5,

2012).

APD ¼ hdi þ rZ

In 2002, professor Golbraikh and Tropsha described

criteria of few statistical parameters that any developed

QSAR model should accomplish in order to get considered

as a reliable model. A good QSAR model should always

have coefﬁcient of correlation close to the ideal model

(R²= 1) for having high predictive reliability. In addition

to original R²of the generated model, the correlation

coefﬁcients of the regression lines passing through the

origin, i.e., R²_o(predicted vs. observed activities) and R⁰_o

2

123

Med Chem Res

Scheme 2 Synthetic protocol

for the synthesis of designed

molecules (5e)

Docking analysis

selected protein structures were isolated, energy minimized

to change internal co-ordinates, and re-docked into the

active sites of each protein structure. After re-docking, the

docked conformation of each crystal ligand was compared

with its respective native crystal conformation, and RMSD

between these two conformations was determined in terms

of all atoms. Finally, an average RMSD value was calcu-

lated from the RMSD values of all crystal ligands in each

protein system, and a protein describing the lowest value of

average RMSD was selected for ligand docking.

All docking simulations were performed in ‘‘Glide’’ pro-

gram (Friesner et al., 2004; Glide 5.8, 2012), which

include two precision modes for conformational sampling

of docking ligands named extra precision (XP) and stan-

dard precision (SP). Extra precision is comparatively more

extensive and produces accurate conformations as com-

pared to SP thus used for present study. Total seven 3D

structures of protein ALR2 and one of ALR1 co-crystal-

lized with different inhibitor molecules were downloaded

from protein data bank (PDB) based on their high resolu-

tion. The downloaded crystal structures of ALR2 are 1US0

(Howard et al., 2004), 1PWM (El-Kabbani et al., 2004),

1Z8A (Steuber et al., 2006), 2PZN (Ruiz et al., 2015),

3U2C (Steuber 2011), 1Z3 N (Van Zandt et al., 2005), and

1T41 (Ruiz et al., 2004) while of ALR1 is 3H4G (El-

Kabbani et al., 2005). These downloaded crystal structures

were devoid of hydrogen atoms and contain solvent

molecules thus required to be prepared prior to docking

analysis. These were corrected in ‘‘protein preparation

wizard,’’ and their preparation included addition of

hydrogen atoms, completion of bond order, removal of

solvent molecules, construction of missing side chains and

loops. Thereafter, restricted minimization was performed

on the whole complex of crystal protein ‘‘impref’’ program,

using ‘‘OPLS_2005’’ force ﬁeld (Silakari et al., 2012), and

the resulting new protein complex was within the RMSD

Chemical synthesis

Materials required

All chemicals required for synthetic protocol were pur-

chased from Sigma Aldrich and were [99 % pure as cer-

tiﬁed by manufacturer thus used without any further

puriﬁcation. The completion of chemical reaction at each

step was screened by thin layer chromatography {DC-

Alufolien (20 9 20 cm) Kieselgel 60 F₂₅₄chromato

plates} using hexane/ethylacetate (6:4) as TLC develop-

ment solvent system. All products were puriﬁed either by

recrystallization from appropriate solvent or eluting on

silica columns. Melting point of each compound was noted

on Labtronics digital automatic melting point apparatus.

The characterization of all compounds was performed

based on their respective spectral data. Infrared (IR) spectra

of compounds was recorded on Bruker^Ò(Alpha E) FTIR

spectrometer, mass spectra on Waters^ÒQ-TOF micromass

(LC–MS) using ESI (?) mode, and proton (¹H) NMR

spectra on Bruker^ÒAdvance-II 400 MHz NMR spec-

trometer using deuterated chloroform (CDCl₃) or

dimethylsulfoxide (DMSO-d₆) as solvents.

˚

threshold of 0.15 A for all atoms.

Among the selected proteins, a universal crystal struc-

ture which could serve as a representative for ALR2 pro-

tein, to analyze the interactions of its ligands, was elected

based on cross-validation results of all proteins (Kaur et al.,

2012). In cross-docking protocol, the crystal ligands of all

123

Med Chem Res

Table 8 Molecular structures, predicted and experimental activities of the synthesized ALR2 inhibitors

Compound code

Molecular structure

pIC₅₀

*

Exp. IC₅₀(lM)

DPPH radical scavenging

activity (IC^#₅₀SEM)

Predicted

4.72

Experimental

5.02 0.05

5a

9.76

47.5 0.52

90.0 0.65

24.5 0.34

22.1 0.70

21.0 0.60

5b

5c

5d

5e

4.70

5.00

5.56

6.94

–

5.08 0.07

4.83 0.11

4.43 0.36

4.41 0.20

7.14 0.07

8.33

14.90

37.00

38.87

0.07

Standard

Epalrestat

Ascorbic acid

20.0 0.28

*-log IC₅₀; where IC₅₀represents the dose of compound in mole required to produce 50 % inhibition of Aldose Reductase enzyme

#

IC₅₀was deﬁned as the concentration resulting in 50 % scavenging activity

General procedure for the synthesis of 2-Phenyl-4H-

pure o-benzoyloxyacetophenone derivatives. In the second

step, the solution of above-synthesized compounds (3a–d,

0.083 mol) was prepared in dry pyridine. To this warm

solution at 50 °C, 7 g of hot pulverized potassium

hydroxide (85 %) was added in one lot. The reaction

mixture was mechanically stirred for 15 min which resul-

ted in the appearance of yellow precipitates of potassium

salt of o-hydroxydibenzoylmethane (4a–d). The reaction

mixture was cooled, and acidiﬁed with 100 mL of 10 %

acetic acid to desalt the compounds. The pure diketone

compounds (4a–e) were obtained as light-yellow colored

solids which were ﬁltered, recrystallized from alcohol and

air-dried. In the ﬁnal step, these diketones (4a–d) were

reﬂuxed (1 h) on water bath, in the mixture glacial acetic

acid and concentrated sulfuric acid. After the completion of

reaction, the reaction mixture was poured on crushed ice

with vigorous stirring that resulted in the precipitation of

crude product. The crude product was ﬁltered, washed with

water to remove acid traces, and air-dried. The crude

product was eluted on silica columns using hexane:ethyl

chromen-4-one derivatives (5a–e)

The detailed description of synthetic protocol for com-

pounds 5a-d is displayed in Scheme 1 (Baker, 1933; Mahal

and Venkataraman, 1934). Variously, substituted o-hy-

droxyacetophenones (1a–d, 0.1 mol) and substituted ben-

zoyl chlorides (2a–d, 0.15 mol) were made to react using

dry pyridine as catalyst in well-dried ﬂask under anhydrous

conditions to obtain o-benzoyloxyacetophenone (3a–

d) analogues. Initially, the reaction being exothermic, ﬂask

was cooled occasionally under the tap water. The reaction

was completed in about 15 min. The reaction mixture was

brought down to room temperature, subsequently it was

poured into 600 mL of 3 % hydrochloric acid containing

crushed ice (200 g) with constant and vigorous stirring

which resulted in the appearance of crude solid product.

The crude product was ﬁltered, washed with methanol

followed by water, and air-dried. The crude product was

recrystallized from methanol to obtain white crystals of

123

Med Chem Res

added hydroalcoholic solution (95 % alcohol v/v) and

sodiumhydroxide followed by substituted o-hydroxyace-

tophenone (1e, 0.43 mol) and substituted benzaldehyde

(2e, 0.43 mol) with constant stirring in a cooling bath. The

reaction temperature was maintained between 15 and

30 °C, and after 2–3 h of stirring, a thick paste was

obtained. The apparatus was dismantled and the thick

reaction mixture was washed with water until ﬁltrate was

neutral to litmus. The crude solid was washed with 95 %

alcohol and air-dried. The crude solid was recrystallized

from 95 % alcohol to get pure compound (3e) (Kohler and

Chadwell, 1941; Menezes et al., 2009). In the next step,

compound 3e (3e, 0.006 mol) was dissolved in a mixture of

pyridine (10 mL) and aqueous KOH (10 %, 24 mL). The

reaction mixture was warmed to 60–70 °C in a water bath,

and hydrogen peroxide (10 %, 12 mL) solution was added

to it with stirring during a period of 5 min. The insoluble

fraction obtained from this reaction was separated by

extraction with ether. The remaining alkaline solution on

acidiﬁcation with hydrochloric acid provided 5e (Kurzw-

ernhart et al., 2012; Ahluwalia and Aggarwal, 2004).

Fig. 7 Graphical representation of the experimental and predicted

ALR2 inhibitory activity of compounds (5a–e)

2-(3⁰-Flourophenyl)-4H-chromen-4-one (5a) Gray amor-

phous powder The synthesis was performed according to

the general procedure given above, o-hydroxyacetophe-

none (1.2 mL, 0.1 mol) and 4-nitrobenzoyl chloride

(2.37 g, 0.15 mol) were stirred in 2 mL of dry pyridine

afford compound 5a. 78 %; mp 211–215 °C; FTIR (neat)

acetate (6:4) as solvent system to provide obtain pure ﬁnal

cyclized products (5a–d).

For the synthesis of compound 5e, a different procedure

was followed as summarized in Scheme 2. To the round

bottom ﬂask (RBF) ﬁtted with a mechanical stirrer were

m

_max: 1621 cm^-1, ketonic (C=O) stretching (s); 1250 cm^-1

,

C–F stretching (s); 1382–1001 cm^-1, (C–O) stretching; ¹H

Fig. 8 Graphical representation

of the experimental antioxidant

activity of compounds (5a–e)

123

Med Chem Res

NMR (CDCl₃, 400 MHz) d (ppm) = 6.83 (1H, s),

6.90–6.93 (1H, dd, J = 2.20, 8.72 Hz), 6.98 (1H, d,

J = 2.16 Hz) 7.33 (1H, m), 7.54–7.59 (2H, m), 7.75–7.79

(1H, dt, J = 3.96, 7.92 Hz), 7.83 (1H, d, J = 7.88 Hz),

7.92 (1H, d, J = 8.72 Hz); ¹³C NMR (100 MHz,

d₆-DMSO) d = 176.5, 163.6, 162.8, 160.4, 157.4, 133.6,

130.6, 126.2, 121.7, 117.9, 116.1, 114.9, 112.8, 107.2,

102.3. MS [ESI (?), m/e]: 241.08 (M ? H, 100 %).

(1H, s, ArH), 6.58 (1H, s, ArH); ¹³C NMR (100 MHz,

d₆-DMSO) d = 181.9, 162.9, 153.1, 150.0, 146.6, 131.1,

131.0, 128.9, 128.6, 125.7, 104.3, 104.2, 93.8. MS [ESI

(?), m/e]: 271 [M? ?1].

2-(2⁰,4⁰-dihydroxyphenyl)-3,5,7-trihydroxy-4H-chromen-4-

one (5e) Light Brown amorphous powder The synthesis

was performed according to the general procedure given in

section 8.1, 2,4,6-trihydroxyacetophenone (1, 1.80 g,

4.3 mol) was stirred in the solution of sodium hydroxide in

water (40 mL) and 95 per cent alcohol (20 mL) with sub-

sequently addition of benzaldehyde (2, 1.48 g, 4.3 mol). The

resulted precipitates of 2⁰-hydroxychalcone (3j, 0.006 mol)

were dissolved in a mixture of pyridine (10 mL) and aqueous

KOH (10 %, 24 mL) and were warmed to 60–70 °C in a

water bath. By the addition of hydrogen peroxide (10 %,

12 mL) with stirring gave the insoluble fraction, which was

separated by extraction with ether. The alkaline solution on

2-(3⁰-Chlorophenyl)-4H-chromen-4-one (5b) Light yel-

lowish brown, amorphous powder The synthesis was per-

formed according to the general procedure given in

section 8.1, o-hydroxyacetophenone (1.2 mL, 0.1 mol) and

3-chlorobenzoyl chloride (2.62 g, 0.15 mol) were stirred in

2 mL of dry pyridine to afford the compound 5b. 83 %; mp

°

221–227 C; FTIR (neat) m_max: 1633 cm^-1(s), (C–Cl)

1092 cm^-1(w), (C–O) 1363–1006 cm^-1

;

¹H NMR

(CDCl₃, 400 MHz) d (ppm) = 8.23 (1H, dd, J = 1.56,

7.96 Hz, ArH), 7.94 (1H, t, J = 1.4 Hz, ArH), 7.81 (1H, dt,

J = 1.36, 7.68 Hz, ArH), 7.71–7.75 (1H, m, ArH), 7.59

(1H, d, J = 8.04 Hz, ArH), 7.42–7.53 (3H, m, ArH), 6.82

(1H, s, ArH); ¹³C NMR (100 MHz, d₆-DMSO) d = 178.1,

160.6, 156.2, 148.8, 134.4, 133.7, 131.8, 130.3, 126.0,

125.9, 125.8, 124.0, 121.3, 118.2, 108.9. MS [ESI (?),

m/e]: 257.00 (M ? H, 100 %).

acidiﬁcation with hydrochloric acid gave substituted 3-hy-

°

droxyﬂavone (5e). 64 %; mp 301–307 C; FTIR (neat) m_max

:

(C = C) 1614 cm^-1(s), (C=O) 1661 cm^-1(s), (O–H)

3375 cm^-1(w); ¹H NMR (DMSO-d₆, 400 MHz) d

(ppm) = 12.48 (1H, s, ArOH), 10.45 (1H, s, ArOH), 9.63

(2H, s, ArOH), 8.05 (1H, s, ArOH), 7.25 (1H, d,

J = 8.44 Hz, ArH), 6.39 (1H, d, J = 2.2 Hz, ArH), 6.36

(1H, dd, J = 8.48, 2.2 Hz, ArH), 6.25 (1H, d, J = 2.04 Hz,

ArH), 6.12 (1H, d, J = 2.08 Hz, ArH); ¹³C NMR (100 MHz,

d₆-DMSO) d = 175.9, 163.5, 160.9, 160.5, 156.8, 156.4,

148.4, 135.6, 131.1, 109.3, 107.1, 103.5, 103.1, 97.9, 93.2.

MS [ESI (?), m/e]: 303 [M? ?1].

2-(3⁰-Nitrophenyl)-4H-chromen-4-one (5c) Light yellow,

crystalline solid The synthesis was performed according to

the general procedure given above, o-hydroxyacetophe-

none (1.2 mL, 0.1 mol) and 4-nitrobenzoyl chloride (2.7 g,

0.15 mol) were stirred in 2 mL of dry pyridine afford

°

compound 5c. 90 %; mp 251–256 C; FTIR (neat) m_max

:

(C=O) 1631.49 cm^-1(m), (N=O) 1516, 1372 cm^-1(w),

Biological evaluation

1

(C–O) 1363–1006 cm^-1; H NMR (CDCl₃, 400 MHz) d

(ppm) = 8.83 (t, 1H, J = 1.84 Hz, ArH), 8.40 (2H, dd,

J = 0.92, 2.28 Hz, ArH), 8.24–8.27 (1H, m, ArH),

7.74–7.78 (2H, m, ArH), 7.65 (1H, d, J = 7.76 Hz, ArH),

7.46–7.50 (1H, m, ArH), 6.92 (1H, s, ArH); ¹³C NMR

(100 MHz, d₆-DMSO) d = 178.3, 161.8, 156.2, 135.3,

134.0, 133.6, 131.5, 130.4, 126.4, 125.8, 125.5, 124.4,

123.9, 118.1, 108.2. MS [ESI (?), m/e]: 268.07 (M ? H,

100 %).

Aldose reductase inhibitory activity

In order to perform ALR2 inhibitory activity assay, fol-

lowing chemicals and materials were procured and used for

study. DL-glyceraldehyde (Sigma Aldrich), NADPH and

2-mercaptoethanol (Himedia Laboratories Pvt. Ltd.,

Mumbai), epalrestat (Microlabs Ltd.), sodium carbonate,

dipotassium hydrogen orthophosphate, ammonium sulfate,

lithium sulfate and sodium hydroxide (Loba Chemicals

Pvt. Ltd., Mumbai). All required solutions for study were

freshly prepared, and goat eyes were obtained from a local

abattoir soon after slaughtering.

5,6,7-trihydroxy-2-phenyl-4H-chromen-4-one (5d) Yel-

lowish crystalline solid The synthesis was performed

according to the general procedure using given in sec-

tion 8.1,

2,4,5,6-tetrahydroxyacetophenone

(1.84 g,

0.1 mol) and benzoyl chloride (2.10 mL, 0.15 mol) were

stirred in 2 mL of dry pyridine to afford compound 5d.

68 %; mp 254–258 °C; FTIR (neat) m_max: (C = C)

1611 cm^-1(m), (C=O) 1652 cm^-1(s), (O–H) 3398 cm^-1

(w); ¹H NMR (DMSO-d₆, 400 MHz) d (ppm) = 10.11

(1H, s, ArOH), 8.59 (1H, s, ArOH), 7.93–7.95 (2H, m,

ArH), 7.90 (1H, s, ArOH), 7.51–7.55 (3H, m, ArH), 6.68

The eye lenses of a freshly slaughtered goat were surgi-

cally removed, washed with cold distilled water and

homogenized in cold water (distilled) containing 10 mM

2-mercaptoethanol (1:3 w/v). The homogenate was cen-

°

trifuged (10,000 rpm) for 30 min at 4 C, and thereafter, the

supernatant liquid was saturated by adding 40 % ammonium

sulfate. The saturated solution was ﬁltered, and the ﬁltrate

123

Med Chem Res

was again centrifuged (10,000 rpm) for 30 min at 4 °C to

get a clear lens solution containing ALR2 enzyme which

was (supernatant) utilized for ALR2 enzyme assay (Hayman

and Kinoshita, 1965; Gupta et al., 2013).

activity of the test compounds (Okawa et al., 2001).

0.1 mM solution of DPPH was prepared freshly in

methanol (39.4 mg in 1000 mL). Different concentrations

of test compounds were added with an equal volume to

methanolic solution of DPPH. After 30 min at room tem-

perature, the absorbance was recorded at 517 nm, using

ascorbic acid as standard. IC₅₀values denote the concen-

tration of sample, which is required to scavenge 50 % of

DPPH free radicals. IC₅₀value was determined from the

plotted graph of scavenging activity against the different

concentrations of test compounds.

The assay reaction mixture contained 1 mL of super-

natant (ALR2), 1 mL of 0.0067 M phosphate buffer con-

taining 0.104 mM NADPH, 10 mM glyceraldehyde, 0.4 M

Li₂SO₄and 1 mL of different concentrations of inhibitors

(2.5, 5, 10, 20, 40 lg/mL). The enzyme reaction starts

upon the addition of a substrate, i.e., glyceraldehyde

(10 mM), and the absorbance of assay reaction mixture

was recorded for 3 min on a UV spectrophotometer at

340 nm. The enzyme inhibitory activity of each compound

was calculated which was based on reduction in absor-

bance of reaction mixture due to decreased enzymatic

oxidation of NADPH to NADP^?by ALR2. The activity of

each compound at different concentrations was calculated

using following formula:

Acknowledgments Authors thank Dr. Ravikumar Muttineni (Ap-

plication Scientist), Er. Anirban Banerjee (IT Consultant), and Mr.

Raghu Rangaswamy from Schrodinger, Bangalore, for their constant

scientiﬁc and technical support to handle Schrodinger software and

work smoothly. Author, Bhawna Vyas, thank acknowledge Indian

Council of Medical Research (ICMR), New Delhi, for providing

Senior Research Fellowship (SRF); Grant No. 45/15/2011/BIF/BMS.

DA test=min Â Total volume of assay mixture

Activity U=mL ¼

ð6:2lM; extinction coefficient of NADPH at 340 nmÞ Â Volume of enzyme taken for analysis

The percentage inhibition of all compounds was calcu-

lated by considering 100 % ALR2 activity in the absence

of inhibitor. The concentration of each test compound

giving 50 % inhibition (IC₅₀) was determined by plotting

log concentration of test compounds versus percentage

inhibition.

References

Aggarwal TS, Bhardwaj TR, Kumar M (2010) 3D-QSAR studies on a

series of 5-arylidine-2, 4-thiazolidinediones as aldose reductase

inhibitors: a self-organizing molecular ﬁeld analysis approach.

Med Chem 6:30–36. doi:10.2174/157340610791208718

Ahluwalia VK, Aggarwal R (2004) Comprehensive practical organic

chemistry: preparation and quantitative analysis, chapter: Algar–

Flynn–Oyamada oxidation. Universities Press (India) Private

Limited, pp 23–24

Alexiou P, Pegklidou K, Chatzopoulou M, Nicolaou I, Demopoulos V

(2009) Aldose reductase enzyme and its implication to major

health problems of the 21st century. Curr Med Chem

16:734–752. doi:10.1517/17425255.2014.916277

DPPH scavenging activity

The stable 1,1-diphenyl-2-picryl hydrazyl radical (DPPH)

was used for the determination of free radical scavenging

123

Med Chem Res

Araki E, Nishikawa T (2010) Oxidative stress: a cause and therapeutic

target of diabetic complications. J Diabetes Investig 1:90–96.

doi:10.1111/j.2040-1124.2010.00013.x

Baker W (1933) Molecular rearrangement of some o-acyloxyace-

tophenones and the mechanism of the production of 3-acylchro-

mones. J Chem Soc 10:1381–1389. doi:10.1039/JR9330001381

Brownlee M (2001) Biochemistry and molecular cell biology of

diabetic complications. Nature 414:813–820. doi:10.1038/

414813a

Caballero J (2010) 3D-QSAR (CoMFA and CoMSIA) and pharma-

cophore (GALAHAD) studies on the differential inhibition of

aldose reductase by ﬂavonoid compounds. J Mol Gr Model

29:363–371. doi:10.1016/j.jmgm.2010.08.005

Dunlop M (2000) Aldose reductase and the role of the polyol pathway

in diabetic nephropathy. Kidney Int 58:S3–S12. doi:10.1046/j.

1523-1755.2000.07702.x

El-Kabbani O, Darmanin C, Schneider TR, Hazemann I, Ruiz F, Oka

M, Joachimiak A, Schulze-Briese C, Tomizaki T, Mitschler A,

Podjarny A (2004) Ultrahigh resolution drug design. II. Atomic

resolution structures of human aldose reductase holoenzyme

complexed with Fidarestat and Minalrestat: implications for the

binding of cyclic imide inhibitors. Proteins 55:805–813. doi:10.

1002/prot.20001

El-Kabbani O, Carbone V, Darmanin C, Oka M, Mitschler A,

Podjarny A, Schulze-Briese C, Chung RP (2005) Structure of

aldehyde reductase holoenzyme in complex with the potent

aldose reductase inhibitor ﬁdarestat: implications for inhibitor

binding and selectivity. J Med Chem 48:536–5542. doi:10.1021/

jm050412o

Henry DR, Miller MA, Gu¨ner OF (1997) American Chemical Society,

213th National Meeting, April 13–17: Paper COMP–39

Howard EI, Sanishvili R, Cachau RE, Mitschler A, Chevrier B, Barth

P, Lamour V, Van Zandt M, Sibley E, Bon C, Moras D,

Schneider TR, Joachimiak A, Podjarny A (2004) Ultrahigh

resolution drug design I: details of interactions in human aldose

reductase-inhibitor complex at 0.66 A. Proteins 55:792–804.

doi:10.1002/prot.20015

Hu L, Chen G, Chau RM (2006) A neural networks-based drug

discovery approach and its application for designing aldose

reductase inhibitors. J Mol Gr Model 24:244–253. doi:10.1016/j.

jmgm.2005.09.002

Kador PF (1998) The role of aldose reductase in the development of

diabetic complications. Med Res Rev 8:325–352

Kador PF, Lee JW, Fujisawa S, Blessing K, Lou MF (2000) Relative

importance of aldose reductase versus nonenzymatic glycosyla-

tion on sugar cataract formation in diabetic rats. J Ocul

Pharmacol Ther 16:149–160

Kaur M, Bahia MS, Silakari O (2012) Exploring the role of water

molecules for docking and receptor guided 3D-QSAR analysis of

naphthyridine derivatives as spleen tyrosine kinase (Syk) inhibi-

tors. J Chem Inf Model 52:2619–2630. doi:10.1021/ci300227f

Kaur M, Kumari A, Bahia MS, Silakari O (2013) Designing of new

multi-targeted inhibitors of spleen tyrosine kinase (Syk) and

zeta-associated protein of 70 kDa (ZAP-70) using hierarchical

virtual screening protocol. J Mol Gr Model 39:165–175. doi:10.

1016/j.jmgm.2012.11.011

Kinoshita JH (1990) A thirty year journey in the polyol pathway. Exp

Eye Res 50:567–573. doi:10.1016/0014-4835(90)90096-D

Kinoshita JH, Varma SD, Fukui NH (1976) Aldose reductase in

diabetes. Jpn J Ophthalmol 20:399–410

Engerman RL (1989) Pathogenesis of diabetic retinopathy. Diabetes

38:1203–1206. doi:10.2337/diab.38.10.1203

Fernandez M, Caballero J, Helguera AM, Castro EA, Gonzalez MP

(2005) Quantitative structure-activity relationship to predict

differential inhibition of aldose reductase by ﬂavonoid com-

pounds. Bioorg Med Chem 13:3269–3277. doi:10.1016/j.bmc.

2005.02.038

Friesner RA, Banks J, Murphy RB, Halgren TA, Klicic JJ, Mainz DT,

Repasky MP, Knoll EH, Shaw DE, Shelley M, Perry JK, Francis

P, Shenkin PS (2004) Glide: a new approach for rapid, accurate

docking and scoring. 1. Method and assessment of docking

accuracy. J Med Chem 47:1739–1749. doi:10.1021/jm0306430

Giacco F, Brownlee M (2010) Oxidative stress and diabetic compli-

cations. Circ Res 107:1058–1070. doi:10.1161/CIRCRESAHA.

110.223545

Kohler EP, Chadwell HM (1941) Benzalacetophenone. Org Synth

2:1. doi:10.15227/orgsyn.002.0001

Kurzwernhart A, Kandioller W, Bartel C, Bachler S, Trondl R,

Mu¨hlgassner G, Jakupec MA, Arion VB, Marko D, Keppler BK,

Hartinger CG (2012) Targeting the DNA-topoisomerase com-

plex in a double-strike approach with a topoisomerase inhibiting

moiety and covalent DNA binder. Chem Commun

48:4839–4841. doi:10.1039/C2CC31040F

¨

Ligprep, version 2.5 (2012) Schrodinger, LLC, New York

Liu H, Liu S, Qin L, Mo L (2009) CoMFA and CoMSIA analysis of

2,4-thiazolidinediones derivatives as aldose reductase inhibitors.

J Mol Model 15(7):837–845. doi:10.1007/s00894-008-0439-0

¨

Maestro, version 9.3 (2012) Schrodinger, LLC, New York

Giannoukakis N (2014) Evaluation of ranirestat for the treatment of

diabetic neuropathy. Expert Opin Drug Metab Toxicol 10:1051–

1059

Mahal HS, Venkataraman KJ (1934) Synthetical experiments in the

chromone group. Part XIV. The action of sodamide on

1-acyloxy-2-acetonaphthones. J Chem Soc. doi:10.1039/JR934

0001767

¨

Glide, version 5.8 (2012) Schrodinger, LLC, New York

Golbraikh A, Tropsha A (2002) Beware of q2! J Mol Gr Model

20:269–276. doi:10.1016/S1093-3263(01)00123-1

Gonzalez AM, Sochor M, Hothersall JS, McLean P (1986) Effect of

aldose reductase inhibitor (sorbinil) on integration of polyol

pathway, pentose phosphate pathway, and glycolytic route in

diabetic rat lens. Diabetes 35:1200–1205. doi:10.2337/diab.35.

11.1200

Menezes MJ, Manjreker S, Pai V, Patre RE, Tilwe SG (2009) A facile

microwave assisted synthesis of ﬂavones. Indian J Chem

48:1311–1314

Mercader AG, Duchowicz PR, Fernandez FM, Castro EA, Bennardi

DO, Autino JC, Romanelli GP (2008) QSAR prediction of

inhibition of aldose reductase for ﬂavonoids. Bioorg Med Chem

16:7470–7476. doi:10.1016/j.bmc.2008.06.004

Gupta S, Singh N, Jaggi AS (2013) Evaluation of in vitro aldose

reductase inhibitory potential of alkaloidal fractions of Piper

nigrum, Murraya koenigii, Argemone mexicana, and Nelumbo

nucifera. J Basic Clin Physiol Pharmacol 24:1–11. doi:10.1515/

jbcpp-2013-0071

Havsteen BH (2002) The biochemistry and medical signiﬁcance of

the ﬂavonoids. Pharmacol Ther 96:67–202. doi:10.1016/S0163-

7258(02)00298-X

Moon YJ, Wang X, Morris ME (2006) Dietary ﬂavonoids: effects on

xenobiotic and carcinogen metabolism. Toxicol Vitro 20:187–

210. doi:10.1016/j.tiv.2005.06.048

Nagarajan K, Zauhar R, Welsh WJ (2005) Enrichment of ligands for

the serotonin receptor using the Shape Signatures approach.

J Chem Inf Model 45:49–57. doi:10.1021/ci049746x

Okawa M, Kinjo J, Nohara T, Ono M (2001) DPPH (1,1-diphenyl-2-

picrylhydrazyl) radical scavenging activity of ﬂavonoids

obtained from some medicinal plants. Biol Pharm Bull

24:1202–1205. doi:10.1248/bpb.24.1202

Hayman S, Kinoshita JH (1965) Isolation and properties of lens

aldose reductase. J Biol Chem 240:877–882

123

Med Chem Res

Okuda J, Miwa I, Inagaki K, Horie T, Nakayama M (1982) Inhibition

of aldose reductases from rat and bovine lenses by ﬂavonoids.

Biochem Pharmacol 31:3807–3822. doi:10.1016/0006-2952(82)

90297-0

Patra JC, Singh O (2009) Artiﬁcial neural networks-based approach to

design ARIs using QSAR for diabetes mellitus. J Comput Chem

30:2494–2508. doi:10.1002/jcc.21240

Steuber H, Zentgraf M, Podjarny A, Heine A, Klebe G (2006) High-

resolution crystal structure of aldose reductase complexed with

the novel sulfonyl-pyridazinone inhibitor exhibiting an alterna-

tive active site anchoring group. J Mol Biol 356:45–56. doi:10.

1016/j.jmb.2005.10.067

Szwergold BS, Kappler F, Brown TR (1990) Identiﬁcation of fructose

3-phosphate in the lens of diabetic rats. Science 247:451–454.

doi:10.1126/science.2300805

¨

PHASE, version 3.4 (2012) Schrodinger, LLC, New York

Prabhakar YS, Gupta MK, Roy N, Venkateswarlu Y (2006) A high

dimensional QSAR study on the aldose reductase inhibitory

activity of some ﬂavones: topological descriptors in modeling

the activity. J Chem Inf Model 46:86–92. doi:10.1021/ci050060u

Van Zandt MC, Jones ML, Gunn DE, Geraci LS, Jones JH, Sawicki

DR, Sredy J, Jacot JL, Dicioccio AT, Petrova T, Mitschler A,

Podjarny AD (2005) Discovery of 3-[(4,5,7-triﬂuorobenzothia-

zol-2-yl)methyl]indole-N-acetic acid (lidorestat) and congeners

as highly potent and selective inhibitors of aldose reductase for

treatment of chronic diabetic complications. J Med Chem

48:3141–3152. doi:10.1021/jm0492094

¨

QikProp, version 3.5 (2012) Schrodinger, LLC, New York

Rice-Evans C (2001) Flavonoid antioxidants. Curr Med Chem

8:797–807. doi:10.2174/0929867013373011

Richard DC, Jeffrey DB, David EP, Ildiko EF (1988) Crossvalidation,

bootstrapping, and partial least squares compared with multiple

regression in conventional QSAR studies. QSAR 7:18–25.

doi:10.1002/qsar.19880070105

Robinson WG, Kador PF, Kinoshita JH (1983) Retinal capillaries:

basement membrane thickening by galactosemia prevented with

aldose reductase inhibitor. Science 221:1177–1179. doi:10.1126/

science.6612330

Ruiz F, Hazemann I, Mitschler A, Joachimiak A, Schneider T,

Karplus M, Podjarny A (2004) The crystallographic structure of

the aldose reductase-IDD552 complex shows direct proton

donation from tyrosine 48. Acta Crystallogr D Biol Crystallogr

60:1347–1354. doi:10.1107/S09074449040113

Ruiz F, Hazemann I, Darmanin C, Mitschler A, Van Zandt M,

Joachimiak A, El-Kabbani O, Podjarny A (2015) The crystal-

lographic structure of Aldose Reductase IDD393 complex

conﬁrms Leu300 as a speciﬁcity determinant. doi:10.2210/

pdb2pzn/pdb

Scott JA, King GL (2004) Oxidative stress and antioxidant treatment

in diabetes. Ann N Y Acad Sci 1031:204–213. doi:10.1196/

annals.1331.020

Silakari O, Chand S, Bahia MS (2012) Structural basis of amino

pyrimidine derivatives for inhibitory activity of PKC-h: 3D-

QSAR and molecular docking studies. Mol Inf 31:659–668.

doi:10.1002/minf.201100123

Srivastava SK, Ramana KV, Bhatnagar A (2005) Role of aldose

reductase and oxidative damage in diabetes and the consequent

potential for therapeutic options. Endocr Rev 26:380–392.

doi:10.1210/er.2004-0028

Steuber H (2011) An old NSAID revisited: crystal structure of aldose

˚

reductase in complex with sulindac at 1.0 A supports a novel

Varma SD (1986) Inhibition of aldose reductase by ﬂavonoids:

possible attenuation of diabetic complications. In: Cody V,

Middleton E, Harborne JB (eds) Plant ﬂavonoids in biology and

medicine: biochemical, pharmacological, and structure-activity

relationships. Alan Liss R, New York, pp 343–358

Varma SD, Kinoshita JH (1976) Inhibition of lens aldose reductase by

ﬂavonoids—their possible role in the prevention of diabetic

cataracts. Biochem Pharmacol 25:2505–2513. doi:10.1016/0006-

2952(76)90457-3

Varma SD, Mikuni I, Kinoshita JH (1975) Flavonoids as inhibitors of

lens aldose reductase. Science 188:1215–1216. doi:10.1126/

science.1145193

WellsKnecht KJ, Brinkmann E, Wells-Knecht MC, Litchﬁeld JE,

Ahmed MU, Reddy S, Zyzak DV, Thorpe SR, Baynes JW (1996)

New biomarkers of Maillard reaction damage to proteins.

Nephrol Dial Transpl 11:41–47. doi:10.1093/ndt/11.supp5.41

Wild S, Roglic G, Green A, Sicree R, King H (2007) Global

prevalence of diabetes: estimates for the year 2000 and

projections for 2030. Diabetes Care 27:1047–1053. doi:10.

2337/diacare.27.5.1047

Xie W, Wang W, Su H, Xing D, Pan Y, Du L (2006) Effect of

ethanolic extracts of Ananas comosus L. leaves on insulin

sensitivity in rats and HepG2. Biochem Physiol C Toxicol

Pharmacol 143:429–435. doi:10.1016/j.cbpc.2006.04.002

Young RJ, Ewing DJ, Clarke BF (1983) A controlled trial of sorbinil,

an aldose reductase inhibitor, in chronic painful diabetic

neuropathy. Diabetes 32:938–942. doi:10.2337/diab.32.10.938

Zhang S, Golbraikh A, Oloff S, Kohn H, Tropsha A (2006) A novel

automated lazy learning QSAR (ALL-QSAR) approach: method

development, applications, and virtual screening of chemical

databases using validated ALL-QSAR models. J Chem Inf

Model 46:1984–1995. doi:10.1021/ci060132x

mechanism for its anticancer and antiproliferative effects. Chem

Med Chem 6:2155–2157. doi:10.1002/cmdc.201100374

123

Article Doi

Pharmacophore and docking-based hierarchical virtual screening for the designing of aldose reductase inhibitors: Synthesis and biological evaluation

DOI: 10.1007/s00044-016-1510-5

Source and publish data:

Authors:

Article abstract of DOI:10.1007/s00044-016-1510-5

Full text of DOI:10.1007/s00044-016-1510-5

Products guided by the article

R&D Labs maybe for 145370-30-5

Relevant to this article

Hot Product