steroidal aromatase inhibitors: model receptor surfaces and...

9
Indian Joual of Chemistry Vol. 40B, November 21 , pp. 1054-1062 Steroidal aromatase inhibitors: Model receptor surfaces and 3D QSAR t Ram K R Jetti, Addlagatta Anthony, Ashwini Nangia & Gautam R Desiraju* School of Chemistry, University of Hyderabad, Hyderabad 500 046, India Fax: 040 3010567, E-mail: [email protected].in Received 10 March 2001; accepted (revised) 4 July 2001 Receptor surfaces have been generated with a training set of 50 steroids active against cythrome P450 enzyme, aromatase, using the Drug Discovery Workbench (Ceriu i ). A combination of van der Waalslectrostatic and Wyvill- partial-charge force fields together with overlay of 17- and 1 3-atoms of the steroid ligand resulted in four different receptor surface models. These models have high conventional and cross-validated ?, q 2 values (> 0.8) for 50 training set molecules with the four components, vdW-17A, vdW-1 3A, Wsc-17A, Wsc-13A. Binding energies of six synthetic 2-oxasteid analogues are evaluated with receptor surfaces and their biological activity predicted through 3D QSAR. Ligand-receptor binding is examined in relation to (I) van der Waals vs. Wyvill force fields, (2) 17- vs. 13-atoms overlay, (3) conformation of the 2-oxasteroid. Our computations show that replacement of C2-methylene group with an a-atom in the A-ring of androgens (2-oxasteroids) is accommodated during recognition by the receptor. The biochemical transformation of androstenedione 1 to estrone 4 is catalysed by cytochrome P450 enzyme, aromatase in the presence of flavoprotein, NADPH- P450 reductase and molecular O 2 • The first and second oxidative steps in the C 1 9 demethylation of androst-4-ene-3,17-dione are typical P450 hydroxylations, each requiring one mole of oxygen and NADPH 1 • The third oxidative step is not very clear: it involves abstraction of the C2-�H atom by an active site base leading to C2-C3 enolisation, or nucleophilic attack of a ferric peroxy species on the C l9-aldehyde group of 3 to give a peroxide intermediate 2 • Estrone, 4 is produced by the oxidative cleavage of C l9-methyl group of 1 followed by aromatisation of steroid A-ring with concomitant release of HC0 2 H (Scheme I). However, increased estrogen hormone levels in the body are known to cause malignant tumors and breast cancer. For this reason, specific and irreversible blockade of estrogen biosynthesis P450 enzyme, aromatase, through suicide inactivation or competitive inhibition has been intensely pursued 3 • Aromatase inhibitors constitute a logical approach in the therapy of estrogen-dependent malignancies and breast cancer for oncologists 4 and medicinal chests 5 . Both steroidal and non-steroidal compounds are available as practical drugs: tamoxifen, aminoglutethimide, ketoconazole, 4- androstene-3,6, 17 -trione, 4-hydroxyandrostenedione and testolactone. t Dedicated to Prof. U.R. Ghatak on his 70 birthday NADPH, 02 ° 2 I NADPH °2 NADPH, O2 -HC02H HO ° 4 3 Scheme I-Conversion of androstenedione 1 to estrone 4 by cytochrome P450 enzyme, aromatase. The reaction requires 3 moles each of NADPH and O2. Aromatase is a membrane-bound cytochrome P450 protein that requires detergent treatment for solubi lisation. Crystallisation of aromatase is difficult and because of its insoluble membrane-binding character, its three dimensional crystal structure has thus far been elusive. The crystal structure of cytochrome P450c a m (camphor hydroxylase), a soluble bacterial protein isolated from Pseudomonas putida which has been solved to 1.63 A resolution 6 , is used as a template for modelling cytochrome P450 aromatase (P450om). The amino acid sequence in

Upload: others

Post on 04-Sep-2020

5 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Steroidal aromatase inhibitors: Model receptor surfaces and ...nopr.niscair.res.in/bitstream/123456789/22438/1/IJCB 40B...Indian Journal of Chemistry Vol. 40B, November 2001, pp. 1054-1062

Indian Journal of Chemistry Vol. 40B, November 2001 , pp. 1054-1062

Steroidal aromatase inhibitors : Model receptor surfaces and 3D QSAR t

Ram K R Jetti, Addlagatta Anthony, Ashwini Nangia & Gautam R Desiraju* School of Chemistry, University of Hyderabad, Hyderabad 500 046, India

Fax: 040 3010567, E-mail: [email protected]

Received 1 0 March 2001; accepted (revised) 4 July 2001

Receptor surfaces have been generated with a training set of 50 steroids active against cytochrome P450 enzyme, aromatase, using the Drug Discovery Workbench (Ceriui). A combination of van der Waals-electrostatic and Wyvill­partial-charge force fields together with overlay of 17- and 1 3-atoms of the steroid ligand resulted in four different receptor surface models. These models have high conventional and cross-validated ?, q2 values (> 0.8) for 50 training set molecules with the four components, vdW-17A, vdW-1 3A, Wsc-17A, Wsc-13A. Binding energies of six synthetic 2-oxasteroid analogues are evaluated with receptor surfaces and their biological activity predicted through 3D QSAR. Ligand-receptor binding is examined in relation to ( I ) van der Waals vs. Wyvill force fields, (2) 17- vs. 1 3-atoms overlay, (3) conformation of the 2-oxasteroid. Our computations show that replacement of C2-methylene group with an a-atom in the A-ring of androgens (2-oxasteroids) is accommodated during recognition by the receptor.

The biochemical transformation of androstenedione 1 to estrone 4 is catalysed by cytochrome P450 enzyme, aromatase in the presence of flavoprotein, NADPH­P450 reductase and molecular O2• The first and second oxidative steps in the C19 demethylation of androst-4-ene-3 , 17-dione are typical P450 hydroxylations, each requiring one mole of oxygen and NADPH1 • The third oxidative step is not very clear: it involves abstraction of the C2-�H atom by an active site base leading to C2-C3 enolisation, or nucleophilic attack of a ferric peroxy species on the C l9-aldehyde group of 3 to give a peroxide intermediate2• Estrone, 4 is produced by the oxidative cleavage of C l9-methyl group of 1 followed by aromatisation of steroid A-ring with concomitant release of HC02H (Scheme I). However, increased estrogen hormone levels in the body are known to cause malignant tumors and breast cancer. For this reason, specific and irreversible blockade of estrogen biosynthesis P450 enzyme, aromatase, through suicide inactivation or competitive inhibition has been intensely pursued3• Aromatase inhibitors constitute a logical approach in the therapy of estrogen-dependent malignancies and breast cancer for oncologists4 and medicinal chemists5. Both steroidal and non-steroidal compounds are available as practical drugs: tamoxifen, amino glutethimide, ketoconazole, 4-androstene-3,6, 17 -trione, 4-hydroxyandrostenedione and testolactone.

tDedicated to Prof. U.R. Ghatak on his 70'" birthday

NADPH, 02 •

° 2

I NADPH °2

NADPH, O2 • -HC02H

HO °

4 3

Scheme I-Conversion of androstenedione 1 to estrone 4 by cytochrome P450 enzyme, aromatase. The reaction requires 3 moles each of NADPH and O2.

Aromatase is a membrane-bound cytochrome P450 protein that requires detergent treatment for solubilisation. Crystallisation of aromatase is difficult and because of its insoluble membrane-binding character, its three dimensional crystal structure has thus far been elusive. The crystal structure of cytochrome P450cam (camphor hydroxylase), a soluble bacterial protein isolated from Pseudomonas putida which has been solved to 1 .63 A resolution6, is used as a template for modelling cytochrome P450 aromatase (P450arom). The amino acid sequence in

Page 2: Steroidal aromatase inhibitors: Model receptor surfaces and ...nopr.niscair.res.in/bitstream/123456789/22438/1/IJCB 40B...Indian Journal of Chemistry Vol. 40B, November 2001, pp. 1054-1062

JETTI et al.: STEROIDAL AROMATASE INHIBITORS 1 055

o o o

5 OH H 6 OH 7 OH

8 9 10

Scheme II-Synthetic 2-oxasteroids evaluated using Model Receptor Surfaces and 3D QSAR in this study.

P450cam shows some similarity ( 1 3-20%) with homologous regions in P450arom. The two regions in P450arom that are similar to the substrate binding pocket in P450cam, namely heme-binding and I-helix, have been used to construct an active site model of P450arom. Based on the P450cam crystal structure, models for aromatase have been proposed but these are unable to correctly identify the active site amino acid residues involved in binding. A possible reason for such a failure could be the approximate nature of the model that is inherent from the dissimilarities between the amino acid sequences of P450cam and P450arom.

Contemporary drug design approaches involve different computational methods for mapping the 3D structure of the binding site7•8• In the absence of an active site model developed from related macromolecular structures or NMR data, crystallographic coordinates of ligands that bind to a particular receptor can be used to map the active surface. If the binding affinities of a series of ligands to a receptor protein are known for a number of ligand-receptor complexes, this information can be used to develop QSAR (Quantitative Structure­Activity Relationship) equations. QSAR equations for competitive inhibition of aromatase have been studied for steroidal9 and non-steroidal lo molecules. The role of C6- and C7-substituents and the tolerance of steric bulk of alkyl groups in the AlB-ring region and on the cu'�-face of steroid ligands have been analysed9•

While numerous studies have appeared on steroids and their analogues as drugs and hormones, the medicinal relevance of unnatural 2-oxasteroids I I has been much less investigated. Some of us have recently

developed an efficient synthetic approach to 2-oxasteroids l2 and also examined their packing motifs in the solid statel3• With the available crystallographic data on the conformations and intermolecular interaction patterns of 2-oxasteroids (5-10, Scheme II), we have expanded the scope of this project towards model receptor surfaces and 3D QSAR. In our crystallographic analysis l3c, we noted that unsaturated 2-oxa-4-ene-3-one steroids 5, 7, 9 and 10 adopt the normal l a,2� half-chair conformation while the saturated 5a-analogue 8 has the l a sofa conformation. The 5� steroid 6 has a bent shape; the A-ring has a bowing angle of 8 1 .70 with respect to the mean plane of B/CID-rings. These six 2-oxasteroids have structural variation of functional groups in the B- and O-rings (ketone, hydroxy) and at the AlB ring junction (unsaturated, saturated cu'�). In view of the fact that these 2-oxa-analogues adopt conformations that are similar to the androgenic steroids, the oxasteroids should bind to the receptor protein in a manner similar to the steroid hormonesl4• In this paper, we report receptor surfaces using a training set of 50 steroids9c

active against human placental aromatase and evaluation of the synthetic 2-oxasteroids considered as test set with these receptor surfaces. Our modelling studies predict that 2-oxasteroids have a biological activity comparable to known aromatase inhibitors, including the potent 4-androstene-3,6, l 7-trione 34.

Computational methods All computations were performed on a Silicon

Graphics IndigoIR4400 workstation running under the IRIX 6.2 operating system. Relevant computational

Page 3: Steroidal aromatase inhibitors: Model receptor surfaces and ...nopr.niscair.res.in/bitstream/123456789/22438/1/IJCB 40B...Indian Journal of Chemistry Vol. 40B, November 2001, pp. 1054-1062

1 056 INDIAN J CHEM. SEC. B . NOVEMBER 2001

1 1 : R1=CH20H. R2.Ra=O 12: R1=CHpH. R2=OH. Ra=H 13: R1=CHO. R2.Ra=O 14: Rl=H. R2.Ra=O 15: R1=CHa• R2=OH. Ra=H 24: R1=CHa• R2.Ra=O

o

28: R1.R2=O 29: R1=OH. R2=H

� R2 R3

30: RI'R2.R2=H 33: R1=CH20H. R2.Ra=H 34: R1=CHa• R2.Ra=O

16 : R1=CHpH. R2.Ra=O 1 7: R1=CHO. R2.Ra=O 18: Rl=H. R2.Ra=O

19: R1=CHPH. R2.Ra=O. R4=H 20: R1=CH20H. R2=OH. Ra.R4=H 21 : R1=CHO. R2.Ra=O. R4=H

25: R1=CHa• R2.Ra=O 22: R1=CHa• R2.Ra=O. R4=H 26: R1=CHa• R2=OH. Ra=H 23: R1=CHa• R2.Ra=O. R4=Br

27: R1=CHa• R2=OH. Ra.R4=H

31 : R1=CHa• R2.Ra=O. R4.R5=H 32: R1=CHpH. R2.Ra=O. R4.R5=H 38: R1=CHa• R2=OH. Ra.R4.R5=H 39: R1=CH20H. R2=OH. Ra.R4.R5=H

o

40: R1=CH20C(O)CHa• R2.R3=O. R4.R5=H 41 : R1=CH3• R2.R3=O. R4=H. R5=Br 42: R1=CHa• R2.Ra=O. R4=Br. R5=H 43: R1=CHF2• R2.Ra=O. R4.R5=H

° R

45: R=CHa 47: R=C2H5

R 44: R=CHa 46: R=C2H5 48: R=n-CaH7 50: R=n-C4Hg 52: R=CH(CH3)2 54: R=CsH5 56: R=CH2CsH5 58: R=CH==CH2 59: R=C�H 60: R==CHMe

35: R1.R2=O. Ra.R4=O. R5=CF2 36: R1.R2=O. Ra.R4=H. R5=CH2 37: R1=OH. R2.Ra.R4=H. R5=CH2

49: R=n-CaH7 51 : R=n-C4Hg 53: R=CH(CHa)2 55: R=CsH5 57: R=CH2CsH5

Scheme III-Structures of 50 training set steroid molecules 11·60 (taken from ref. 9c).

modules were accessed from the Drug Discovery Workbench (DDW) of Ceriui (version 4.0) 1 5 . The 50 training set steroids (11·60) used in this study as competitive inhibitors of aromatase are shown in Scheme III and the six 2-oxasteroids (5·10) considered as test set molecules are displayed in Scheme II.

Cambridge Structural Databasel6: CSD refcode NOTESTO 1 ( 17�-hydroxy- 19-nor-androst-4-ene-3-one) for compound 30 and related 19-norsteroids; HTESBZ ( 1 9-hydroxy- 17�-benzoylandrost-4-ene-3-one) for 39 and related 19-hydroxyl and 19-carbonyl group containing compounds; ANDIONI0 (androst-4-ene-3, 17-dione) for 21 and 24; and HANDDO ( 17�-hydroxyandrost-4, 14-diene-3-one) for the remaining compounds. Each steroid structure was

Molecular geometries were determined from closely related crystal structures retrieved from the

Page 4: Steroidal aromatase inhibitors: Model receptor surfaces and ...nopr.niscair.res.in/bitstream/123456789/22438/1/IJCB 40B...Indian Journal of Chemistry Vol. 40B, November 2001, pp. 1054-1062

JETTI el al.: STEROIDAL AROMATASE INHIBITORS l OS7

Table I-Biological activity data (in pK; units) of 50 training set molecules9c• Aromatase inhibitor 4-androstene-3,6, 1 7-trione 34 is

underlined.

Steroid pK; Steroid pK; Steroid pKI

1 1 - 1 .04 28 1 .23 45 2.25

12 - 1 .65 29 -0.30 46 2.85

13 - 1 . 1 1 30 0.85 47 2.32

14 -0.38 3 1 1 .88 48 2.33

15 -0.74 32 1 .90 49 2. 1 7

16 - 1 . 1 7 33 0.59 50 2 .05

17 -0.25 34 1 .58 5 1 1 .92

18 -0.47 35 1 .30 52 1 .65

19 0.00 36 2 .30 53 1 .50

20 -0.99 37 0.92 54 1 .43

21 -0. 1 4 38 0.08 55 1 .67

22 0.92 39 0.77 56 1 .20

23 - 1 .04 40 0.53 57 2.00

24 0.60 41 0.64 58 2 .29

25 0.65 42 0.34 59 1 .20

26 -0.72 43 0.35 60 2.30

27 -0.47 44 1 .95

optimised with MOPAC 6.0 (MNDO). Alkyl and aryl residues were modelled in the all-trans extended conformation; they did not show much variation during minimisation. Since conformational flexibility of the tetracyclic steroid skeleton is minimal 14, an exhaustive conformational search of the molecules was not deemed necessary prior to alignment. Ligands were aligned by rigid-body, least-squares fitting of common carbon/oxygen atoms (A/B/C/D-rings, 1 7-atom model) or the B/CID-rings ( I 3-atom model) of each molecule to those in 3-deoxy-4-androstene- 1 7-one (steroid 31 In Scheme III). I n our crystallographic study of 2-oxa-steroid conformations 13c, overlay of 2-oxa-4-androstene-3, 17-dione 9 with 4-androstene-3, 17 -dione (ANDSEO) showed a modest r.m.s. deviation of 0. 1 0 A with the 17-atom model which improved to 0.05 A. when only the B/C/D rings were considered ( I 3-atom model), suggesting that the A-ring geometry can vary considerably in steroid analogues. Therefore, we have used both overlay methods to study relationship between conformations of A-ring and receptor maps. Thus, overlay plots were built from 50 training set molecules with both 17- and 1 3-atom overlay ( 1 7 A, 1 3A) using van der Waals surface, electrostatic charge (vdW) and Wyvill surface, partial charge (Wsc, Wyvil l soft contour) complementarity criteria. These force fields were selected from DDW modules

Table 11- Conventional and cross-validated ?, l values for four receptor models.

Model VdW-1 7A

r2 0.84

l 0.8 1

VdW- 1 3A Wsc-17A Wsc- 1 3A

0.84 0.88 0.83

0.80 0.85 0.79

VdW: van der Waals force field; Wsc: Wyvill soft contour 1 7A: 1 7 atom overlay; 1 3A: 1 3 atom overlay

and default parameters were used for the computations .

Receptor surfaces were generated with weights based on biological activity data (Table I) taken from a previous CoMFA (Comparative Molecular Field Analysis) study9c. Since f:..G is proportional to binding affinity, -logKj (f:..G "" 1 .4 pKj, in kcal/mol) and QSAR analyses are based on l inear free energy relationships, the biological activity of 50 training set steroids (expressed in 1lM) were converted to 10g lo( I /Kj) values1 7• The receptor surface descriptors, expressed as field values based on the probes of methyl group and a proton, were added to the study table in DDW along with interaction energies for QSAR. Regression analysis was carried out using the GFA (genetic function algorithm) method consisting of over 20000 generations and with specific inclusion of constant, linear, spline, quadratic, offset quadratic and quadratic spline variable terms in QSAR equation with no fixed length and scaling.

Results and Discussion

Model receptor surfaces The four different receptor surfaces computed are

vdW-17A, vdW-1 3A, Wsc-17A, Wsc-1 3A. The idea of using different atom overlay and force field combinations is to accommodate the flexible steroid A-ring and assess the nature of drug-receptor intermolecular interactions with the putative surface. Cross-validation was carried out using the LGO method (leave-group-out) 18 by random removal of one out of every five molecules in the training set. This procedure gave good conventional and cross-validated coefficients, I and l, respectively (> 0.8, Table II). LGO calculation of I and l with randomly selected training and test sets of 40 and 10 molecules respectively gave values within 5% of those in Table II, showing the robustness of our models. The binding of six novel 2-oxasteroid analogues to these receptor surfaces and the generation of 3D QSAR equations with Wyvill and van der Waals force functions are discussed further.

Page 5: Steroidal aromatase inhibitors: Model receptor surfaces and ...nopr.niscair.res.in/bitstream/123456789/22438/1/IJCB 40B...Indian Journal of Chemistry Vol. 40B, November 2001, pp. 1054-1062

1 058 INDIAN J CHEM, SEC. B , NOVEMBER 200 1

Figure l a Figure lb Figure l a and lb-Stereoview of overlay plot of 50 train ing se t molecules 1 1-60 with (a) 1 7-atol11 overlay and (b) 1 3-atol11 overlay. Note the more spread out spatial d i stribution of the A-ring in (b). The spread of atoms on the left s ide comprises the C6-alkyl group containing steroids (e.g. 44-60).

Figure 2a Figure 2b Figure 2a and 2b-Receptor surface generated with van der Waals force field, 50 molecules training set and 17 -atom overlay of l igands (vdW- 1 7 A). (a) Hydrophobic contours of the surface in shades of brown and polar regions as white patches: (b) Electrostatic map of training set l igands with the receptor showing the attractive i nteractions in magenta.

A stereoview of 50 training steroids 11-60 with 1 7-and 1 3-atom overlay methods i s shown as in Figure 1. The overlay is tighter with 1 3-atom match of B/C/D-rings but this is offset by a larger spread of A­ring atoms. Energies of i nteraction, Einteract , with two of the four receptor surfaces, vdW - 1 7 A and Wsc-17 A, are l isted in Table III . I nteraction energies with the 1 3-atom overlay are in the same range (- 1 3 to -3 kcal/mol) and show simi lar trends. In the receptor surface (Figure 2a), shades of brown colour indicate the extent of hydrophobici ty and white patches represent polar domains. The electrostatic interaction map (Figure 2b) has attracti ve hydrogen bonding regions on the receptor, displayed in magenta contours around 03 and 0 1 7 atoms (white patches i n Figure 2a). Einteract values for 2-oxasteroids with the four model receptor surfaces (- 1 5 to -5 kcaUmol, Table IV) are comparable to energies calculated with the training set steroids (Table III), except for oxasteroid 6 which is an outl ier. An examination of the receptor surface bound to 2-oxasteroid l igands shows that the i nteraction is repUlsive in the vicin ity of 02 because the newly intj-oduced O-atom makes an

unfavourable contact with the receptor (Figure 3a, green patch; compare with Figure 2b, grey/magenta). However, the total binding energy between 2-oxasteroid l igands and receptor surface i s favourable, as revealed by magenta regions in Figure 3b. This shows that the change i n steroid structure from C2 ---7 0 is accommodated i n binding of l igand to the receptor.

Eintcruct values for 6 are large positive numbers ( 1 02

to 106 kcallmol , Table IV) ; the binding of 5� oxasteroid to the receptor is very repulsive. The soft Wyv i l l receptor is better able to accommodate changes i n the shape and conformation of l igand compared to the van der Waals potential with i ts rigid contours. The five oxasteroids with a trans or an unsaturated AlB-ring junction (5, 7-10) are generally flat and bind to the receptor l i ke the training set aromatase inhibitors. When the geometry of 6 i n MRS (molecular receptor surface) was al lowed to vary during energy min imisation, a value of - 1 1 .3 1 kcallmol was obtained with the vdW-17 A model compared to + 1 45.34 kcal/mol when the conformation of the l igand was fixed. In this minimum energy

Page 6: Steroidal aromatase inhibitors: Model receptor surfaces and ...nopr.niscair.res.in/bitstream/123456789/22438/1/IJCB 40B...Indian Journal of Chemistry Vol. 40B, November 2001, pp. 1054-1062

JEITI el al.: STEROIDAL AROMATASE INHI BITORS 1 059

Table I I I-Energy of interaction, Ein,cr:oc, ( in kcallmol), of the 50 training set molecules with overlay of 17 atoms. Receptor surfaces were generated using van der Waals and Wyvi l l foree fields.

Steroid vdW- 1 7A Wsc-17A Steroid VdW- 1 7A Wsc- 1 7A Steroid vdW- I 7A Wsc- 1 7A

1 1 -9.24 - 1 0.27 28 -6.66

1 2 -9.80 - 1 1 . 1 2 29 -4.7 1

13 -8.96 - 1 1 .44 30 -8.01

1 4 -9.22 -1 1 .50 31 -9. 1 9

1 5 -9.87 - 1 1 .38 32 -9.28

16 -9. 1 6 -1 0.62 33 - 1 1 .98

1 7 -8.5 1 - 10.79 34 -7.95

1 8 -8.99 - 1 1 .55 35 -8.7 1

19 -9.02 - 1 0.98 36 -6.53

20 -9.72 -1 0.88 37 -5.52

2 1 -8.99 -1 1 .23 38 -9.82

22 -9.3 1 -1 1 .44 39 -9.85

23 -9.40 -1 1 .00 40 -7.56

24 -6.60 -6.58 4 1 -9. 19

2 5 -6.64 -6.62 42 -9. 1 8

26 -4.37 -3.00 43 -8.44

27 -4.56 -2.80 44 - 1 2. 1 3

Table IV-Eintc"", ( in kcallmol) of 2-oxasteroids with different model receptor surfaces.

17 atom overla,t 1 3 atom overla,t Steroid vdW Wsc vdW Wsc

5 - 1 0.72 - 1 3 .85 -1 1 .48 -1 4.39

6 145.34 - 1 2.53 2794.57 -2.0 x 1 06

7 -8.46 -9.28 -8.88 -9.99

8 -10.46 - 1 3.86 -1 1 .02 - 1 3.5 1

9 -10.70 - 1 4. 1 5 -1 1 .59 -1 4.83

10 -6.05 -5 . 1 1 -6.24 -5.43

conformation, 03 of the carbonyl group moves towards the �-face by l .29 A and 06-H moves by 0.54 A (Figure 4). The planar geometry of A-ring becomes a half-chair after minimisation while the C2-C3-C4-C5 torsion angle changes from 14.0° to 5 1 .4°. Although the geometry and interaction energy of 6 change upon binding to the receptor, its single point energy (SPE) in the bound and free states is nearly the same (- 1 74.40 VS. - 1 70.78 kcallmol, respectively) . Since SPE is energy of a frozen conformation and not the minimum energy of a molecule, small differences of a few kcallmol should not be over-interpreted. In terms of recognition, 6 can adapt to fit with the complementary MRS without much of an energy penalty because of its numerous low energy conformations. Similar results were obtained when this calculation was repeated with vdW-1 3A model. These computations give an idea of the conformational changes required in the l igand for

-6.88 45 -1 2.04 - 1 5.79

-2.99 46 - 1 2.27 - 1 6 . 1 4

-9.40 47 - 1 2.45 - 1 6. 1 8

- 1 0.78 48 - 1 2.40 - 1 6. 1 7

- 1 0.98 49 -1 2.78 - 1 6.40

- 1 6.01 50 - 1 2.70 -1 6.44

-8.63 5 1 - 1 3 .24 -1 6.73

-9: 1 5 52 - 1 2.44 - 1 6.20

-5.99 53 -1 2 .34 - 1 5.88

-4.22 54 - 1 3 .29 - 1 6.59

- 1 1 .25 55 - 1 3 . 1 9 - 1 6.54

- 1 1 .38 56 - 1 3 .66 -1 6.79

-6.46 57 -1 3.25 - 1 6.29

- 1 0.55 58 - 1 2. 1 3 - 1 5 .80

-9.62 59 - 1 2 . 1 6 - 1 5.96

-9.85 60 -8.20 -8.79

- 1 5.80

binding to the receptor and the energIes associated with such events.

Two general comments will c larify matters further: ( 1 ) We have generated receptor surfaces with a 50 molecules training set in which many steroids have alkyl groups at C6 (e.g. 44-60, Scheme III). When a subset of 29 training steroids was selected which are devoid of C6-substituents (1 1-34, 38, 39, 41-43), it was found that the receptor surfaces thus generated were too tight to accommodate three out of six 2-ox asteroids with a C6-0H group (5-7, Scheme II). This resulted in a program overflow for the calculation of ligand-receptor interaction energies because the values obtained are too high (» 1 06

kcallmol) . Because the receptor surfaces generated with C6-unsubstituted training steroids were over­constrained for the 2-oxasteroid test molecules they are not discussed. (2) It may be noted that results obtained from both van der Waals and the Wyvill force field functions are discussed together. We found the vdW contours to be more visually appealing because of their sharpness and better contrast. So al l receptor surfaces displayed in this paper are generated using the vdW function . Receptor models generated with Wsc function show similar contours and colour schemes but have a gradual profile because of the softer intermolecular potential . However, results based on molecular modelling with both vdW and Wsc force fields are discussed.

Page 7: Steroidal aromatase inhibitors: Model receptor surfaces and ...nopr.niscair.res.in/bitstream/123456789/22438/1/IJCB 40B...Indian Journal of Chemistry Vol. 40B, November 2001, pp. 1054-1062

1 060 INDIAN J CHEM, SEC. B, NOVEMBER 200 1

��re � F�re Th Figure 3a and 3b-Binding of 2-oxasteroids 5-10 with the receptor surface shown in Figure 2. (a) Energy of interaction; (b) Total

binding energy. In these plots, magenta indicates attractive and green repulsive l igand-receptor interactions.

Figure 4-Stereoview of 5� steroid 6 before (red) and after (blue) minimisation with the receptor. Notice the variation in AlB-ring geometry and the change in orientation of pendant groups on these rings.

3D QSAR

3D QSAR equations generated by genetic function approximation (GFA) statistical method correlated better compared to those derived from principal component analysis (PCA) and partial least squares (PLS) methods. Therefore, only the GFA results are discussed. Equations generated using receptor surface descriptors from Wyvill force field had better predictive power than van del' Waals field. This is because the 'soft' nature of Wyvill intermolecular potential is better able to tolerate changes in ligand structures being tested compared to the 'hard' van del' Waals potential which produces wide fluctuations i n values, be they energy of interaction or biological activity.

The predicted activity of six 2-oxasteroids obtained from 3D QSAR equations generated with Wyvill potential, Wsc-1 7A and Wsc- 1 3A, are given i n Table V. These values may be compared with the actual and predicted biological activity data displayed in Figure 5. The predicted activity of 2-oxasteroids with Wsc- 13A equation is comparable to the activity of many training set molecules listed in Table I, with the well-known aromatase inhibitor 34 (4-androsten-

3,6, 1 7-trione, pKi = 1 .S8 , Ki = 0.026 �M) being a benchmark for comparison. Our QSAR studies show that the novel 2-oxasteroids are active ligands against cytochrome P4S0 aromatase with pKi values in the range 1 .3-3.8 (Table V). Therefore they should bind to the enzyme with good affinity.

Conclusions

The biological activity of steroidal aromatase inhibitors together with the crystallographic data on selected molecules have been utilised to evaluate the binding of some synthetic 2-oxasteroid analogues with receptor surfaces generated in the Drug Discovery Workbench (Ceriui) . Ligand-receptor interaction energies have been calculated for 2-oxasteroids starting with a training set of SO steroids using van der Waals and Wyvill force fields and 1 7-and 1 3-atom overlay methods. The initial conformations of l igands were obtained from related crystal structures and then optimised with Minimiser. Replacement of C2-methylene with an O-atom, a bioisosteric substitution in the A-ring, is well tolerated in terms of ligand-receptor recognition and binding.

Future studies will focus on the nature of favourable residues in the B-ring. Since C6-0H is introduced during the synthesis of functionalised 2-oxasteroids, 12 hydrophobic O-alkyl and O-phenyl derivatives could be interesting candidates for future synthesis. The recently reported crystal structure of a Fab fragment bound to cytochrome P4S0 aromatase provides insight i nto antibody-antigen recognition 19. In the absence of P4S0arom enzyme crystal structure, inhibitor design will still have to rely on model structures or computational approaches. This new and

Page 8: Steroidal aromatase inhibitors: Model receptor surfaces and ...nopr.niscair.res.in/bitstream/123456789/22438/1/IJCB 40B...Indian Journal of Chemistry Vol. 40B, November 2001, pp. 1054-1062

JEITI et al.: STEROIDAL AROMATASE INHmITORS 1061

Wsc-SOM-17 A 3.0..,-------------------,

2.S

2.0

� 1 .S is « 1 .0

i O.S 'C £ 0.0

tf. -O.S (!) - 1 .0

- 1 .S

+ +

+ + + + + ++

• + + .... + + +

-2.0 +---.,,-----.---r--.----r--.-.--r-�-_l -2.0 -1 .0 0.0 1 .0

Observed Biological Activity

Figure Sa

2.0 3.0

Wsc-SOM-13A 3.0..,-------------------,

2.S

2.0

� 1 .S is « 1 .0

i O.S 'C £ 0.0

tf. -O.S (!) -1 .0

-1.S

+ ++ + +

+ +

++ + + + + \ . t * +

-2.0 +--,-----,--,-,----,---,-..,------,,--....,-----1 -2.0 -1 .0 0.0 1 .0

Observed Biological Activity Figure sb

2.0 3.0

Figure Sa and 5b-Plots of Observed vs. GFA predicted biological activity of 50 training set molecules with Wyvill force field. (a) 17-atom ligand overlay; (b) 1 3-atom ligand overlay. The slope indicates the goodness of fit.

Table V-Predicted biological activity (in pKi units) from 3D QSAR of six 2-oxasteroids. Compare the test set values with

training set data in Table I. Oxasteroid

5

6

7

8

9

10

QSAR equations 17-atom overlay

Wsc-17A

-0.56

-0.79

1 .58

-1 .00

-0.64

1 .29

Wsc-13A

1 .8 1

0.35

1 .96

1 .36

1 .89

3.83

Activity = 1 .47991 + 64.3589 [(Ele/1279) - 0.26 1981 ] + 1 . 1 6 143 [-27.0589 - (Ele/2791 ) - 0.002939 (Ele/3272)2] + 0.003575 [0.035489 - (EleI2379)]2 - 59.3079 (vdWI1 805)2

13-atom overlay Activity = 0.930677 - 0. 1 84173 (Ele/1985) + 1 .07088 [(-28. 1449) - (Ele/2736)] + 0. 142121 (EIe/221 2) + 7 1 .6904 [(vdWI1681) + 0.044613] - 0.01793 (Ele/301 9)2

active lead of 2-oxasteroids as aromatase inhibitors will be optimised through in silico methods

20•

Acknowledgements We thank the CSIR for financial assistance

(90/008/99/EMR-II) and for fellowship support to RKRJ and AA. We acknowledge the interest and co­operation of Dr. Osman Guner (Molecular Simulations Inc., San Diego) and Prof. R. Kumar (I. I . Sc. , Bangalore).

References 1 Cole P A & Robinson, C H, J Med Chem, 33, 1990, 2933.

2 (a) Akhtar M, Calder M R, Corina D L & Wright J N, Biochem J, 210, 1982, 569. (b) Oh S S & Robinson C H, J. Steroid Biochem Malec Bioi, 44, 1993, 389.

3 Van Wauwe J P & Janssen P A J, J Med Chem, 32, 1989, 223 1 .

4 (a) Goss, P E & Gwyn, K M E H , J Clin Oncol, 12, 1994, 2460. (b) Ibrahim N K & Buzdar A U, Am J Clin Oncol (CCT), 1 8, 1995, 407.

5 (a) Recanatini M & Cavalli A, Bioorg Med Chem, 6, 1998, 377. (b) Rao S, Aoyama R, Schrag M, Trager W F, Rettie A & Jones J P, J Med Chem, 43, 2000, 2789. (c) Hartmann R W, Hector M, Haider S, Ehmer P B, Reichert W & Jose J, J Med Chem, 43, 2000, 4266. (d) Hartmann R W, Hector M, Wachall B G, Palusczak A, Palzer M, Huch V & Veith M, J Med Chem, 43, 2000, 4437. (e) Recanatini M, Bisi A, Cavalli A, Belluti F, Gobbi S, Rampa A, Valenti P, Palzer M, Palusczak A & Hartmann R W, J Med Chem, 44, 2001, 672.

6 (a) Graham-Lorence S, Khalil M W, Lorence M C, Mendelson C R & Simpson E R, J Bioi Chem, 266, 1991, 1 1939. (b) Laughton C A, Zvelebil M J J M & Neidle S, J Steroid Biochem Malec Bioi, 44, 1993, 399. (c) Chen S, Zhou D, Swiderek K M, Kadohama N, Osawa Y & Hall P F, J Steroid Biochem Malec Bioi, 44, 1993, 347.

7 (a) Beddel C R (Ed) The Design of Drugs to Macromolecular Targets, (Wiley, Chichester), 1992. (b) Medicinal Chemistry for the 21" Century, edited by C G Wermuth, N Koga, H Konig & B W Metcalf (Blackwell , Oxford), 1 994. (c) 3D QSAR in Drug Design: Theory, Methods and Applications, edited by H Kubinyi (ESCOM, Leiden), 1993. (d) Boyd D B in Encyclopedia of Computer Science and Technology, Vol 33 (Suppl 1 8), edited by A Kent & J G Williams, (Marcel Dekker, New York), 1995, pp 4 1-7 1 .

8 (a) Desiraju G R, Gopalakrishnan B, Jetti R K R, Raveendra D, Sarma J A R P & Subramanya H S, Molecules, 5, 2000, 945. (b) Desiraju G R, Sarma J A R P, Raveendra D, Gopalakrishnan B, Thilagavathi R, Sobhia M E & Subramanya H S, J Phy Org Chem, 14, 2001, 48 1 .

Page 9: Steroidal aromatase inhibitors: Model receptor surfaces and ...nopr.niscair.res.in/bitstream/123456789/22438/1/IJCB 40B...Indian Journal of Chemistry Vol. 40B, November 2001, pp. 1054-1062

1 062 INDIAN J CHEM, SEC. B, NOVEMBER 200 1

9 (a) O'Reil ly J M, Li N, Duax W L & Brueggemeir R W, J Med Chem, 38, 1995, 2842. (b) Numazawa M & Oshibe M , J Med Chem, 37, 1994, 1 3 1 2. (c) Oprea T I & Garcia A E, J Computer-Aided Mol Des, 1 0, 1996, 1 86.

10 (a) Baroudi M, Robert J & Luu-Duc C, J Steroid Biochem Malec Bioi, 57, 1996, 73. (b) Koymans L M H, Moereels H & Bossche H V, J Steroid Biochem Malec Bioi, 53, 1995, 1 9 1 .

1 1 (a) Shibata K, Takegawa S, Koizumi N, Yamakoshi N & Shimazawa E, Chem Pharm Bull, 4C, 1992, 935. (b) Koizuma N, Takegawa S , Mieda M & Shibata K, Chem Pharm Bull, 44, 1996, 2 162.

1 2 Nangia A & Anthony A , Indian J Chem, 36B, 1997, 1 1 1 3 . 1 3 (a) Anthony A, Jaskolski M, Nangia A & Desiraju G R,

Chem Commun, 1998, 2537. (b) Anthony A, Jaskolski M & Nangia A, Acta Crystallogr, Sect C, 55, 1999, 787. (c) Anthony A, Jaskolski M & Nangia A, Acta Crystallogr, Sect B, 56, 2000, 5 1 2.

1 4 Duax W L, Griffin J F & Ghosh D in Structure Correlation, Vol 2, edited by H-B B iirgi & J D Dunitz, (VCH, New York), 1994, pp 605-633.

1 5 Cerius2 Program, Molecular S imulations Inc, 9685 Scranton Road, San Diego, CA 92 1 2 1 , USA.

16 Allen F H & Kennard 0, Chem Des Autom News, 8, 1993, 3 1 .

1 7 Oprea T I , Waller C L & Marshall G R , J Med Chem, 37, 1994, 2206.

1 8 (a) Keimowitz A R, Martin B R, Razdan R K, Crocker P J, Mascarel la S W & Thomas B F, J Med Chem, 43, 2000, 59. (b) Robarge M J, Agoston G E, Izenwasser S, Kopajtic T, George C, Katz J L & Newman A H, J Med Chem, 43, 2000, 1 085.

19 Sawicki M W, Ng P C, B urkhart B M, Pletnev V, Higashiyama T, Osawa Y & Ghosh D, Mol Immunol, 36, 1999, 423 (PDB code 32C2).

20 Sanseau P, Drug Discovery Today, 6, 2001 , 3 16.