CANCER BIOTHERAPY & RADIOPHARMACEUTICALS Volume 17, Number 2, 2002 © Mary Ann Liebert, Inc.
Modeling of the Anticancer Action for Radical Derivatives of Nitroazoles: Quantitative StructureActivity Relationship (QSAR) Study Andrei Khlebnikov,1 Igor Schepetkin,2 and Byoung Se Kwon2 1 Altai State Technical University, Barnaul, Russia; 2Immunomodulation Research Center, University of Ulsan, Ulsan, Korea
A QSAR analysis of the anti-tumor, anti-metastasis and anti-colony formation (for metastatic colonies) activities of eighteen nitroazoles (including metronidazole and hypoxic radiosensitizers RP-170, KU-2285 and sanazole (drug AK-2123)) and their nitro and nitroso anion radical derivatives against melanoma B16 in mice has been performed. The QSAR models were built with the use of the frontal polygon method. This approach has features of different 3D QSAR methodologies and belongs to the group of “indirect” methods. The procedure allows to build robust models with high predictive ability even in series of diverse and conformationally flexible compounds. Key atomic characteristics accompany the geometrical requirements in the analysis of local 3D molecular similarity. By variation of weight coefficients for hydrophobicity, refraction increments, and partial charge it is possible to achieve better quality of QSAR and evaluate the importance of each characteristic for biological activity under consideration. It was found that hydrophobicity, molar refraction and charge characteristics of nitro anion radical derivatives are more significant for interaction with molecular targets than those of the parent compounds and of the nitroso anion radical derivatives. Size and hydrophobic properties of substituents in nitro anion radicals play significant role for ligand-target interaction in the processes of inhibition of metastatic spreading and growth of metastatic colonies by nitroazoles. A scheme of competitive interaction of parent nitroazoles and their anion radicals with a target in organism is suggested. It can be concluded that the step of one-electron reduction of nitroazoles can be important for anticancer activity of these drugs. Key Words: anti-tumor drugs, nitroazoles, radiosensitizers, quantitative structure-activity relationships INTRODUCTION On the basis of nitroazoles new drugs have been created which are applied as anticancer bioreductive agents and radiosensitizers. 1–3 Nitroazole drugs are also synthesized that are proposed as markers for tumor hypoxia.4–5 One possible mechanism of cytotoxic action of these drugs results from their bioreduction to highly reactive Address reprint requests to Andrei Khlebnikov, Altai State Technical University, 46 Lenin Avenue, Barnaul 656099, Russia Tel: 17 (3852) 245513, 17(3852) 522435 Fax: 17 (3852) 367864 E-mail:
[email protected]
radical intermediates and reactive oxygen species (ROS) generation. ROS cause modification of tumor cell macromolecules and the subsequent apoptotic and necrotic death of tumor cells.6–10 Another mechanism of the anticancer action of nitroazoles may consist of receptor-mediated control of gene expression and/or immune response.11–12 A multi-step process that leads to activation of these biological responses may be as follows: 1) penetration of nitro compound into the vessel walls and cell plasmatic membranes, 2) interaction of nitro compound with active site of the nitroreductase followed by a one-electron reduction, 3) reaction of nitro derivatives with 193
oxygen (in aerobic conditions) or with cell macromolecules (in hypoxic conditions) resulting in an oxidative stress, modulation of gene expression, and a complex immune response to hapten-conjugate adducts.3,13 From these stages the last one possesses the properties of signal amplification, thus the development of specific anticancer action at the low nongenotoxic doses of nitroazole (Fig. 1). The redox-modulating activity of nitro compounds is restricted to specific cellular hydrophilic/hydrophobic sites. The cellular kinases, phosphatases, the arachidonic acid cascade, glutathione cycle, transcription factor NF-kB, hypoxia-inducible factor-1 (HIF-1), antioxidant responsive element (ARE), serum response factor (SRF), ICAM-1 expression on endothelial cells, dendritic cell surface markers such as MHC class II molecules, and the lymphocyte receptor, are considered as redox-sensitive sites and oxidative sensors responsible for modulation by bioreductive agents and ROS.14–16
Construction of QSAR models is very important for understanding the molecular mechanism of action of anticancer drugs and their design. For nitro compounds with anticancer action such analysis was carried out only for nitroacridines.17 The mechanism of the anti-tumor immune response activation by nitro compounds can be similar to the delay-type hypersensitivity reaction.12 QSAR analysis for skin sensitization potential was performed for several haptens.18–20 The integral hydrophobic descriptors of nitro compounds correlate with their transport into the sites of biological action, while integral electrophilic descriptors are important for describing the reactivity of these and some bioreductive agents in the processes like the one-electron reduction, interaction with oxygen, and conjugation with macromolecules.21–23 However, authors of these publications did not use the principle of structural interaction between ligand and target at the molecular level according to the “key-lock” princi-
Figure 1. Hypothetical scheme of nitroazole action on differential gene expression and immune response in tumor-bearing organism. Ligands: nitroazole, their nitro and nitroso radical products; targets: redox-sensitive sites and oxidative sensors; k1, K21 and K2—rate constants for intermediate and final ligand-target complex for nitro radical products, respectively.
194
195
Figure 2. tigation.
Compounds under inves-
ple. In the frontal polygon (FP) method developed previously, atomic increments for these characteristics are used instead of the integral characteristics of compounds.24 In the FP method it is considered that for molecular interaction the “key” and the “lock” have to be locally similar to each other in terms of charge distribution, hydrophobic characteristics of atoms and the atomic refractions. FP method can be used as a tool for structure-based design of drugs with different biological activities. 25 It should be classified as “indirect” QSAR method based on delineating regression relationships between a specified biological activity and structure of the compounds eliciting it. 26 In the present work the QSAR analysis of the antitumor, anti-metastasis and anti-colony formation (for metastatic colonies) activities of eighteen nitroazoles (Fig. 2) and their nitro and nitroso anion radical derivatives against melanoma B16 in mice has been performed. It should be noted that properties of anion radicals are often used for deriving QSAR models.27 The following well-known compounds were analyzed among nitroazoles I-XVIII: nonsubstituted 2-nitroimidazole (I), metronidazole (IV), three hypoxic radiosensitizers RP-170 (II), fluorinated 2-nitroimidazole KU-2285 (III), and sanazole (drug AK-2123) (VIII). The experimental results for QSAR analysis were taken from the work of Konovalova et al.28 The choice of nitro and nitroso radicals for investigation was conditioned by the assumption that biological action of nitroazoles is determined by products of their one-electron reduction. These products are more active for binding with biological macromolecules than parent compounds.3,10,29 METHOD The construction of QSAR models was performed by the FP method, which considers the three-dimensional (3D) similarity of molecules, making it possible to process series of conformationally flexible compounds.24,25 The method is based on the hypothesis of local 3D similarity, according to which the presence of similar sites on the “peripheral surface” of molecules makes them close in biological action.30 For the solution of the conformational flexibility problem we used the modification of FP method that foresees the partition of molecules under investigation into rigid and flexible fragments (submolecules).24 These submolecules for nitroazoles I-XVIII are
196
shown with help of dashed bonds in Fig. 2. Nitro and nitroso anion radicals were partitioned similarly. Rigid submolecules in the neutral molecules are the heterocycle, nitro (-NO2), nitro anion (-NO2-.), methyl (-CH3) and amide (-CONH) groups. Other fragments were considered to be flexible since they have internal rotational degrees of freedom or contain too little atoms for FP construction. The “imprints” produced from rigid fragments with the use of methodology and parameters from previous papers24,25 include projections of atoms which are characterized by the distance hX to the prototype atom X and by its charge qX.24,25 Projections of boundary atoms adjacent with other fragments were described also by hydrophobicity HX and molar refraction RX of the corresponding substituents at the atom X. HX and RX values were calculated according to the additive scheme.31 Geometry and charge distribution for anion radicals were calculated by semi-empirical MNDO method.32 The imprints are regarded as “key” portions that can be in complementarity with some “lock” portions according to the ligand-target interaction that is presumably responsible for biological effect under investigation. The imprints of compounds I-XVIII and of radical products of their reduction underwent pair comparison for establishing the degree of local similarity between molecular structures via the search for optimal impositions (OI). The OI reflect possible modes of “key-lock” interaction. The target is not treated explicitly in the FP method, instead it is assumed that some of the imprints resemble features of the target in a complementary manner. Thus, imprint-to-imprint comparison is performed to model key-lock similarity. The optimality criterion (1) was used in the process of the search for OI.30 1 2 F 5 }3/2 } wr^ r XY 1 wh ^ (hX 2 HY)2 n0 X,Y x,y
[
1 wq ^ (qX 2 qY)2 1 wH
^ (HX 2 HY)2 (1)
11wR
^ (RX 2 RY)2] (1)
X,Y
X,Y
X,Y
where n0 is number of assignments in the imposition; rXY 2 distance between the projections assigned to each other in a given imposition (Å); wr 5 0,2; wh 5 0,2; wq 5 12,8; wH 5 0,2; wR 5 0,00032 2 weight coefficients. Summing up in
expression (1) is conducted through the pairs of the projections assigned to each other (X,Y). The impositions that satisfy conditions (2) were considered to be optimal: F # KO ,
nO $ NO ,
(2)
where KO and NO 2 the parameters (threshold of the optimality criterion and minimal number of assignments in OI, respectively). Construction of QSAR models24 was made with the use of OI collections found with different KO and NO , thus differing in requirements to the degree of structural correspondence of molecules. Then with fixed values of these parameters giving the best (basic) model, variation of weight coefficients wq , wH , and wR with respect to the values given above was performed, for the purpose of the establishing the role that hydrophobic characteristics and size of substituents play for biological action of the compounds. Structure-activity relationships were built on the basis of the obtained OI massifs in the form of linear equations (3): H
I 5 ^ aiZi,
(3)
i5 1
where I is calculated biological activity, ai 2 regression coefficients. The reduced basis of variables Zi was determined by the partial least squares method with the procedure discussed previously.24,33 The dimensionality H of the basis was chosen to be smaller, but so that the sufficiently high values of the correlation coefficient 2 R and of cross-validation coefficient Rcv would be ensured: S 2cv 2 512 } Rcv , S 2ser
(4)
where Scv2 2 a root-mean-square inaccuracy of the prediction; Sser2 2 the standard deviation of the activity values within the series of compounds being investigated. RESULTS AND DISCUSSION The characteristics of QSAR models built for different types of biological activity are presented in Table 1. The number of mutually orthogonal variables H in equation (3) was chosen equal to 5 or 6 on the basis of the considerations given
above (see Method). Search for OI was carried out with the lowest number of assignments NO 5 3. An increase in this value to NO 5 4 led to significant decreasing in the correlation coefficients R and R2cv values. The possible reason for deterioration in the quality of QSAR models with an increase of NO is the loss of useful structural information included in OI. For example, if NO 5 4, the impositions obtained with participation of nitro group are not classified as optimal, although this substituent is one of the main structural elements in nitroazoles and their nitro anion radicals. The basic structure-activity relationships in majority of the cases possess sufficiently high values of R and R2cv (bold numbers in Table 1). For parent nitroazoles I-XVIII and for their nitroso anion radicals the QSAR models built with KO equal to 0.10, 0.12 or 0.14 appear to be basic. For nitro anion radical these relationships were found at KO 5 0.06. The fact that for nitro anion radicals the best characteristics of QSAR were obtained with smaller threshold KO of the optimality criterion F, i.e., with more specific OI, than for parent compounds I-XVIII and corresponding nitroso anion radicals, can testify about higher complementarity and higher affinity of nitro anion radicals to assumed targets as compared with the parent compounds. The optimization of basic structure-activity relationships obtained with the selected parameters KO was carried out by successive 50% decreasing the weight coefficients wq, wH and wR . The results of QSAR model optimization process in the series of parent nitroazoles and their anion radicals are shown in Table 2. For parent compounds I-XVIII the decrease of the weight coefficient wq causes deterioration of QSAR models for all the three types of the biological activity with respect to the basic QSAR. This can be regarded as the evidence that charge distribution plays a significant role in the process of molecular recognition of the compounds by assumed target. With respect to suppression of metastastic spreading and growth of metastatic colonies, furthermore, the hydrophobic characteristics of molecular groups play an important role, since the decreasing of the coefficient wH also leads to lowering in R and R2cv values in comparison with these values before variation of the weight coefficients. The decrease in the molar refraction contribution wR , which reflects influence of substituent size, causes essential deterioration of
197
QSAR models only in the case of tumor growth suppression activity. The same considerations applied to the series of radical products of bioreduction, make it possible conclude about the importance of charge distribution for the nitroso anion radicals interaction with targets. Both for nitro and nitroso anion radicals with respect to anti-tumor and anticolony formation activities of the compounds, significant deterioration in the quality of QSAR models occurs after decreasing the wH weight coefficient (Table 2). This testifies about noticeable 198
contribution of hydrophobic characteristics to the molecular recognition process if one assumes that anion radicals directly interact with targets. Furthermore, during the optimization of the basic structure-activity relationships for the products of nitroazole one-electron reduction in a number of cases it was possible to reach an improvement in QSAR quality. The experimental biological activities in comparison with the activities calculated in optimized QSAR models for parent nitroazoles and their nitro and nitroso radical derivatives are presented in Table 3. We have
performed pair regression analysis on the basis of these values. In the majority of cases the pair linear correlation coefficients between the selected types of biological activity proved to be small (r , 0.8) both for parent nitroazoles and for their nitro and nitroso radical derivatives (Table 4). This may signify that each form of biological activity investigated is determined by individual collection of factors. In the details, metastasis is a complicated multi-step process that involves interactions between tumor cells, extracellular matrix and vessel walls.34 Consequently, anti-metastasis action of a drug will depend both on its influence upon the metastastic spreading process itself and on cell proliferation in primary tumor focus. The suppression of tumor growth in the primary tumor focus can be determined mainly by anti-proliferating action of drug under the hypoxia conditions. On the contrary, the suppression of melanoma B16 metastatic colonies growth in pulmonary tissue can be determined by anti-proliferating action of a drug under conditions of high-level oxygen supply and corresponding significant contribution of ROS to this biological activity development mechanism. The
analysis of experimental data for nitroazoles showed that 2-methyl-5-nitroimidazole derivatives (compounds IV (Metronidazole), V, and VI) suppressed the metastasis regardless of their antiproliferative action in metastatic foci. Elimination of these three compounds led to appearance of good correlation (r 5 0.91) between the antimetastasis and anti-colony formation activities for nitroso radicals derived from fifteen remaining drugs (Fig. 3). Thus, anti-metastasis activity of compounds IV, V, and VI can be connected to their influence on interactions between endothelial cells and extracellular matrix, whereas anti-metastasis activity of the remaining fifteen compounds should be attributed to their antiproliferative activity both in primary and metastatic foci of tumor growth. Table 4 also contains values of multiple regression coefficient R for each type of biological activity. The R value is one of the main characteristics of optimized QSAR models. Examination of these values for all the QSAR models built by us shows that the best quality was obtained with the use of nitro anion radical structures (bold numbers in Table 2). Consequently, it can be as199
sumed that nitro anion radicals generated after one-electron reduction of parent molecules IXVIII, participate in key stages of the anti-tumor action of nitroazoles both via anti-proliferative and anti-metastasis mechanisms. Increased oxygenation of lungs as compared with other tissues in organisms can lead to the rapid oxidation of nitro radicals and ROS formation. Thus, high quality of the QSAR models based on nitro anion radical structures can testify for the important contribution of ROS to anti-metastasis action of nitroazoles. Targets for nitroazole radical products and ROS can be glutathione (GSH).35 High intracellular level of GSH protects B16 melanoma cells from in vivo and in vitro oxidative stress, contributing to mechanism of metastatic cell survival.36,37 200
In general, the interaction of parent nitroazole (ligand L) and its radical product (ligand in a form of radical, L.-) with the target can be presented in the following form (see also Fig. 1). Let us assume that L reversibly interacts with target (T) with the formation of an intermediate complex TL*, which is then converted into TL complex with high binding affinity. T1L
k
k2 1 — * TL* —R TL — ) k 21
(A)
For ligand in a form of radical: k2 . —k1 T 1 L 2) * TL.2 —R TL 2e k— 21
(B)
As it was obtained from QSAR analysis, the interaction of L.2 with the target can be more
preferable than that of the parent compound L. According to Fig. 1, the complex TL causes differential gene expression and final biological effect. The rate of the whole process (B) is expressed by equation (5): v 5 k2 [T]0[L.2 ]0/(Km 1 [L.2 ]0),
(5)
where K m 5 (k21 1 k2 )/k1 2 Michaelis constant. If Km is small (i.e., radicals have high affinity for the target, and formation of conjugate product TL from TL.2 is the slow stage) then at sufficiently high concentration of L.2 the
Figure 3. Plot of anti-metastasis activity (% from control) versus anti-colony formation activity (% from control) for nitroso radical derivatives. The regression line corresponds to the subset of 15 compounds excluding IV (metronidazole), V, and VI.
Michaelis-Menten equation is reduced to the simplified form: v » k2[T]0
(6)
Significant concentrations of nitro radicals from nitroazoles can be achieved in the conditions of tissue hypoxia when reoxidation of these radicals by oxygen with the formation of parent nitroazoles and O2.2 is limited (see Fig. 1).38,39 In this case the rate of the process is determined mainly by the initial concentration of the target active centers [T]0 in organism and by rate constant k2 . Since value of k2 depends on molecular structure of the intermediate complex TL.2 and, hence, on the structure of initial radical L.2 , QSAR models may reflect, in particular, the relation of this value to radical structure. The rate constant k2 is a very important characteristic in this model, since it can determine the preference of gene expression, caused by ligand radical binding to a certain redox-sensitive site. It can be concluded that high quality of the QSAR models built with the use of nitro anion radical structures gives indirect evidence that these radicals participate to great extent in molecular recognition responsible for investigated biological activities of nitroazoles I-XVIII. Size and hydrophobic properties of substituents in nitro anion radicals play significant role for the processes of inhibition of metastatic spreading and growth of metastatic colonies by nitroazoles. It should be mentioned that rate of one-electron reduction may not influence the resulting biologi201
cal effect if this rate is high, thus conditions are achieved when radical concentration in comparison to that of the target is sufficiently high (under hypoxic conditions). In this case the concentration [L.2 ]0 becomes greater than Km , and Eq. (5) can be simplified to Eq. (6). Better quality of QSAR models obtained with the structures of nitro anion radicals may testify that these radicals play the role of agonists, competing successfully with the parent nitroazoles for binding with the target. ACKNOWLEDGMENTS
10.
11.
12.
13.
The work has been supported in part by SRC Fund to IRC at UOU from the KOSEF and the Korean Ministry of Sciences and Technology.
14.
REFERENCES
15.
1. Kagiya VT, Nishimoto S, Ito T, Wang J, Abe M and Sasai K. Fluorine modification of nitroazole radiosensitizers for the enhancement of sensitizing activity with lowering toxicity: a pharmacokinetic characterization. Int J Radiat Oncol Biol Phys 1992; 22: 601–605. 2. Kagiya T, Nishimoto S, Shibamoto Y, Wang J, Zhou L, He YL, Sasai K, Takahashi M and Abe M. Importance of tumor affinity of nitroazoles in hypoxic radiosensitization. Int J Radiat Oncol Biol Phys 1989; 16: 1033–1037. 3. Rauth AM, Melo T and Misra V. Bioreductive therapies: an overview of drugs and their mechanisms of action. Int J Radiat Oncol Biol Phys 1998; 42: 755–762. 4. Melo T, Ballinger JR and Rauth AM. Role of NADPH:cytochrome P450 reductase in the hypoxic accumulation and metabolism of BRU59-21, a technetium-99m-nitroimidazole for imaging tumor hypoxia. Biochem Pharmacol 2000; 60: 625–634. 5. Murugesan S, Shetty SJ, Noronha OP, Samuel AM, Srivastava TS, Nair CK and Kothari L. Technetium-99mcyclam AK 2123: a novel marker for tumor hypoxia. Appl Radiat Isot 2001; 54: 81–88. 6. Brezden CB, McClelland RA and Rauth AM. Apoptosis and 1-methyl-2-nitroimidazole toxicity in CHO cells. Br J Cancer 1997; 76:180–188. 7. Kedderis GL, Argenbright LS and Miwa GT. Covalent interaction of 5-nitroimidazoles with DNA and protein in vitro: mechanism of reductive activation. Chem Res Toxicol 1989; 2: 146–149. 8. Kuno Y and Shinomiya N. PR-000350, a novel hypoxic radiosensitizer, enhances tumor cell killing by promoting apoptosis preferentially in the S-phase fraction. Apoptosis 2000; 5: 69–77. 9. Raleigh JA, Shum FY and Liu SF. Nitroreductase-induced binding of nitroaromatic radiosensitizers to un-
202
16.
17.
18.
19.
20.
21.
22.
23.
saturated lipids. Nitroxyl adducts. Biochem Pharmacol 1981; 30: 2921–2925. Varghese AJ and Whitmore GF. Properties of 2-hydroxylaminoimidazoles and their implications for the biological effects of 2-nitroimidazoles. Chem Biol Interact 1985; 56: 269–287. Inomata T, Ogawa Y, Itoh S, Kariya S, Hamada N, Nishioka A, Yoshida S, Nagasawa H, Hori H and Inayama S. Lung metastasis suppression of the bifunctional new radiosensitizer KIN-806. Int J Mol Med 1999; 4: 257–260. Schepetkin IA. Immune response to haptenized tumor antigen as possible mechanism of anticancer action of hypoxic bioreductive agents at low doses. Cancer Biother Radiopharm 1999; 14: 291–296. Roberts DW and Basketter DA. Further evaluation of the quantitative structure-activity relationship for skinsensitizing alkyl transfer agents. Contact Dermatitis 1997; 37: 107–112. Fuchs J, Zolner TM, Kaufmann R and Podda M. Redox-modulated pathways in inflammatory skin diseases. Free Radic Biol Med 2001; 30: 337–353. Dhakshinamoorthy S, Long DJ 2nd and Jaiswal AK. Antioxidant regulation of genes encoding enzymes that detoxify xenobiotics and carcinogens. Curr Top Cell Regul 2000; 36: 201–16. Hayes JD and McLellan LI. Glutathione and glutathione-dependent enzymes represent a co-ordinately regulated defence against oxidative stress. Free Radic Res 1999; 31: 273–300. Mazerska Z, Mazerski J and Ledochowski A. QSAR of acridines. II. Features of nitracrine analogs for high antitumor activity and selectivity on mice, searched by PCA and MRA methods. Anticancer Drug Des 1990; 5: 169–187. Cronin MT and Basketter DA. Multivariate QSAR analysis of a skin sensitization database. SAR QSAR Environ Res 1994; 2: 159–179. Franot C, Roberts DW, Basketter DA, Benezra C and Lepoittevin JP. Structure-activity relationships for contact allergenic potential of g,g-dimethyl-g-butyrolactone derivatives. 2. Quantitative structure-skin sensitization relationships for a-substituted-a-methyl-g, g-dimethyl-g-butyrolactones. Chem Res Toxicol 1994; 7: 307–312. Roberts DW and Basketter DA. Quantitative structureactivity relationships: sulfonate esters in the local lymph node assay. Contact Dermatitis 2000; 42: 154–161. Debnath AK, de Compadre RLL, Debnath G, Shusterman AJ and Hansch C. Structure-activity relationship of mutagenic aromatic and heteroaromatic nitro compounds. Correlation with molecular orbital energies and hydrophobicity. J Med Chem 1991; 34: 786–797. Hansch C, McKarns SC, Smith CJ and Doolittle DJ. Comparative QSAR evidence for a free-radical mechanism of phenol-induced toxicity. Chem Biol Interact 2000; 127: 61–72. Kawakami Y and Hopfinger AJ. Prediction of initial reduction potentials of compounds related to anthracy-
24.
25.
26.
27.
28.
29.
30.
31.
clines and implications for estimating cardiotoxicity. Chem Res Toxicol 1990; 3: 244–247. Khlebnikov AI. The frontal polygon method for conformationally flexible molecules. Structure-activity relationship in the series of Baker triazines - dihydrofolate reductase inhibitors. Khim-Farm Zhurn 1997; 31(3): 41–48. Khlebnikov AI. Frontal polygon method: a new approach to the analysis of structure-biological activity relationships. Khim-Farm Zhurn 1994; 28(11): 32–35. Loew GH, Villar HO and Alkorta I. Strategies for indirect computer-aided drug design. Pharm Res 1993; 10: 475–486. Schmitt H, Altenburger R, Jastorff B and Schuurmann G. Quantitative structure-activity analysis of the algae toxicity of nitroaromatic compounds. Chem Res Toxicol 2000; 13: 441–450. Konovalova NP, Volkova LM and Kagiya VT. Side chain structure of nitroasoles and anti-tumor activities in B16 melanoma bearing mouse by ultra low dose administration. Sensitisation Newsletter 1996; 3(4): 3–5. Varghese AJ and Whitmore GF. Binding to cellular macromolecules as a possible mechanism for the cytotoxicity of misonidazole. Cancer Res 1980; 40: 2165–2169. Khlebnikov AI. An algorithm for determining local similarity of molecules. Zh Strukt Khim 1995; 36: 1083–1087. Ghose AK and Crippen GM. Use of physicochemical parameters in distance geometry and related three-dimensional quantitative structure-activity relationships. A demonstration using Escherichia coli dihydrofolate reductase inhibitors. J Med Chem 1985; 28: 333–346.
32. Dewar MJS and Thiel W. Ground states of molecules. 38. The MNDO method. Approximation and parameters. J Am Chem Soc 1977; 99: 4899–4907. 33. Glen WG, Dunn III WJ and Scott DR. Principal component analysis and partial least squares regression. Tetrahedron Comput Methodol 1989; 2: 349–376. 34. Bittinger F, Gonzalez-Garcia JL, Klein CL, Brochhausen C, Offner F and Kirkpartick CJ. Production of superoxide by human malignant melanoma cells. Melanoma Res 1998; 8: 381–387. 35. Varghese AJ and Whitmore GF. Misonidazole-glutathione conjugates in CHO cells. Int J Radiat Oncol Biol Phys 1984; 10: 1341–1345. 36. Anasagasti MJ, Martin JJ, Mendoza L, Obrador E, Estrela JM, McCuskey RS and Vidal-Vanaclocha F. Glutathione protects metastatic melanoma cells against oxidative stress in the murine hepatic microvasculature. Hepatology 1998; 27: 1249–1256. 37. Carretero J, Obrador E, Anasagasti MJ, Martin JJ, Vidal-Vanaclocha F and Estrela JM. Growth-associated changes in glutathione content correlate with liver metastatic activity of B16 melanoma cells. Clin Exp Metastasis 1999; 17: 567–574. 38. Re JL, De Meo MP, Laget M, Guiraud H, Castegnaro M, Vanelle P and Dumenil G. Evaluation of the genotoxic activity of metronidazole and dimetridazole in human lymphocytes by the comet assay. Mutat Res 1997; 375: 147–155. 39. Schepetkin IA, Cherdyntseva NV and Kagiya VT. Sanazole as substrate of xanthine oxidase and microsomal NADPH/cytochrome P450 reductase. Pathophysiology 2001; 8: 119–127.
203
