Short Communication

SAR of Cu (II) Thiosemicarbazone Complexesas Hypoxic Imaging Agents:

MM3 Analysis and Prediction of Biologic Properties

Sweta Singh,1,2 Anjani K. Tiwari,1 Himanshu Ojha,1 Nitin Kumar,1 Bachcha Singh,2 and Anil K. Mishra1

Abstract

Copper(II) bis(thiosemicarbazone) are very useful for blood flow and hypoxic imaging. The aim of this study wasto identify structure-activity relationships (SARs) within a series of analogues with different substitution patternsin the ligands, in order to design improved hypoxia imaging agents and elucidate hypoxia selectivity mechanisms.Genetic algorithms (GAs) were used to develop specific copper metal-ligand force field parameters for the MM3force-field calculations. These new parameters produced results in good agreement with experiment and previ-ously reported copper metal-ligand parameters. A successful quantitative SAR (QSAR) for predicting the severalclasses of Cu(II)-chelating ligands was built using a training set of 21 Cu(II) complexes. The QSAR exhibited acorrelation between the predicted and experimental test set. The QSAR preformed with great accuracy; r2¼ 0.95and q2¼ 0.90 utilizing a leave-one-out cross-validation with multiple linear regression analysis to find correlationbetween different calculated molecular descriptors of these complexes. The final QSAR mathematical models werefound as the following:

QSAR model for copper(II) bis(thiosemicarbazone)

Log P = f3.01698 ( – 0.0590)g - LUMO f0.1248 ( – 0.068)g+ MR f0.3219 ( – 0.086)gn¼ 21 jrj ¼ 0:972 s¼ 0:188 F¼ 98:102

The resulting models could act as an efficient strategy for estimating the hypoxic conditions through imagingand provide some insights into the structural features related to the biological activity of these compounds.

Key words: QSAR, linear regression, chelating agent, imaging, genetic algorithms

Introduction

Several semicarbazones, as well as their sulphur an-alogs, have proved the efficiency and efficacy in combat-

ing various diseases.1 It is of great interest because oftheir chemistry and, potentially, beneficial biologic activi-ties, such as antitumor, -bacterial, -viral, and -malarial ac-tivities.2–4

One of the important aspect of these compounds are thatCu(II) bis(thiosemicarbazone) complexes have been usedin vivo as radiotracers for the evaluation of blood flow invarious organs, such as the brain, kidneys, and heart, as wellas for the evaluation of hypoxia in tissue.5–7

We have been developing an approach, based on redox-active copper complexes,8–13 following the demonstrationthat the copper complex of bis(thiosemicarbazone) is selec-tively taken up in hypoxic tissue.14 These complexes have anumber of interesting properties and uses and have been in-vestigated for their potential in anticancer chemotherapy,15

superoxide dismutase–like activity,16 and in radiolabeledform as nontissue selective blood-perfusion tracers.17 Theydiffuse into cells because of its low molecular weight, pla-narity,18,19 and lipophilicity.20 It then becomes trapped byintracellular reduction to a copper (I) complex,21 followed bydissociation and entry of the copper into the normal intra-cellular copper pool.22

1Division of Cyclotron and Radiopharmaceutical Sciences, Institute of Nuclear Medicine and Allied Sciences, Delhi, India.2Division of Chemistry, Benares Hindu University, Varansi, India.

Address correspondence to: Anil K. Mishra; Division of Cyclotron and Radiopharmaceutical Sciences, Institute of Nuclear Medicine andAllied Sciences; Brig. S.K. Mazumdar Road, Timarpur, Delhi 110054, IndiaE-mail: mis_ak@rediffmail.com

CANCER BIOTHERAPY AND RADIOPHARMACEUTICALSVolume 25, Number 1, 2010ª Mary Ann Liebert, Inc.DOI: 10.1089=cbr.2009.0670

117

A particularly important property to be able to predict forin vivo use is the logarithm of the experimental n-octanol=water partition coefficient (log P). This is an importantpharmacologic descriptor of bioavailability for both organiccompounds and inorganic complexes. Predicting and alter-ing log P is a key element in rational drug design, since thecorrect incorporation of lipophilic groups to a chemicalcompound can result in an increase in biologic activity.

Quantitative structure-activity relationships (QSARs) havebecome a common tool in the field of molecular modelingsince their introduction. Indeed, they have found applicationin both the prediction of biologic activity and, more recently,in the prediction of the gK,23 absorption, distribution, me-tabolism, and excretion, (ADME) toxicological properties oforganic drug-like compounds. QSAR became more attractivefor chemists with the development of new software tools,which allowed them to discover and understand how mo-lecular structure influences properties and, very importantly,to predict and prepare the optimum structure. The software isnow more amenable for chemical and physical interpretation.

In this article, we identify physicochemical properties ofthese bis(thiosemicarbazone) complexes that control hypoxiaselectivity by comparison of a large series of analogs, withvarious alkyl substitution patterns (Fig. 1), with respect topartition coefficient (log P), which is the most importantdescriptor for lipophilicity. Some of the results described inthis article have been reported in preliminary form.

Experimental Section

Molecular modeling and QSAR study

QSAR studies are mathematic tools of prediction end-points of interest on organic compounds acting as drugs,which have not been experimentally determined.

Many physiologic activities can be related to the molecularstructure. The biologic activities are being modeled from usinga set of molecular descriptors that often include chemical,electronic, constitutional, and topologic indices. The derivedrelationship between descriptors and activity from mathe-matic equations are used to estimate the property of othermolecules and or finding the parameters affecting the biologicactivity. In QSAR studies, a linear regression model of the formY¼X bþ e may be used to describe a set of predictor variables.

Genetic algorithms (GAs) were used to develop specificcopper metal-ligand force-field parameters for the MM3 forcefield, from a combination of ab initio calculations. The MM3*force field is one of the most accurate force fields available forthe modeling of organic ligands. It is relatively convenientto successfully developed specific metal-ligand parameterscompatible with this accurate force field for a variety of metal-amide complexes.

In this study, first, the optimized three-dimensional (3D)structures of the molecules were obtained by MM1 calcula-tions. To insure not locating in local minima, geometry opti-mization was many times with different initial conformations.

FIG. 1. Compounds for quantitative structure–activity relationship studies.

118 SINGH ET AL.

Some geometric properties of molecules, including dihedralangles and atomic distances, were calculated.

The QSAR investigations were carried out by the linearfree-energy relationship (LFER) model proposed by Hanschand Leo.23 Selection of parameters is the first step in anyQSAR study. In the present study, parameters that wereconsidered relevant to the bis(thiosemicarbazone) series(electronic, hydrophobic, and steric) were selected and con-sidered as consistent, which included molar refractivity (MR),Vander Waals volume (VDW), Connolly accessible area(CAA), Connolly molecular area (CMA), Connolly solvent-excluded area (CSEV), dipole-dipole energy (DD), partitioncoefficient (log P), HOMO, and so forth. The above-mentionedparameters were calculated by MM2 studies, using Chem 3D6.0 software. Geometries of all compounds were completelyoptimized by the same software package.

A classical Hansch multivariate regression analysis, usingthe least-square method, was chosen to derive QSPR equa-tions for the dataset. The level of significance of each coef-ficient was judged by statistical procedures, such as F-tests.By employing the method of least square with the stepwiseselection and elimination procedure, a statistical analysis wascarried out. For each equation, several indices of best fit wereconsidered: the regression coefficient ‘‘r,’’ the standard de-viation ‘‘s,’’ and the measure of level of statistical significance‘‘F.’’ The structures of substituted analogs containing com-

pounds taken for the study are depicted in Figure 1 and theirdescriptors and log P values are shown in Table 1.

Results and Discussion

The squared correlation matrix of the physicochemicalproperties used in the stepwise regression analysis is shown inTable 2. From Table 2, one can see that LUMO, VDW, and MRare highly interdependent, and there is no covariance. Table 2also describes the corelation of other parameters with log P.

All the parameters showed significant correlation withbiologic activity (r< 0.8) (Table 2), but the LUMO andMR exhibited the best correlation (r> 0.9) of high statisticalsignificance >99.9%. The statistical quality of the resultingmodels depicted in Equation 1 is determined by r2 (r2> 0.9).The calculated parameters and correlation matrix needed forMRA (multiple regression analysis) are shown in Figure 2.

QSAR model for copper(II) bis(thiosemicarbazone)

Log P¼f3:01698ð – 0:0590Þg�LUMOf0:1248ð – 0:068ÞgþMRf0:3219ð – 0:086Þg

n¼ 21 jrj ¼ 0:972 s¼ 0:188 F¼ 98:102 ð1Þ

The graphs between observed -log P and calculated -log Pare shown in Figure 2. The calculated -log P were calculatedby putting the MR values in the equations, and the predicted

Table 1. Values of Descriptors Calculated for Copper(II) Bis(thiosemicarbazone) Analogs

Name Log P MR CMR DD CLogP C.T LUMO VDW t PSA

ETS �0.33 60.27 6.3401 792.64 �1.00906 865.83 43.74 371.6 88.46KTS �0.17 75.88 7.8846 883.26 �0.74486 882.48 30.56 172.18 97.69n-prTS 0.58 69.47 7.2677 838.4 �0.04894 874.85 34.44 330.32 88.46i-Pr TS 0.56 69.65 7.2677 837.96 �0.08106 869.59 34.77 325.04 88.46n-BuTS 0.99 74.07 7.7315 861.28 0.57794 880.42 30.86 309.68 86.46PhTS 1.74 80.02 8.4635 933.72 0.50084 937.55 36.25 504.93 88.46PTSM 0.71 70.27 7.2677 793.68 0.72044 850.68 31.14 369.68 60.84n-PrTS 1.62 79.47 8.1953 839.44 1.77844 863.06 26.4 328.4 60.48i-PrTSM 1.6 79.65 8.1953 839 1.64844 857.23 26.61 323.12 60.48Bu-TSM 2.03 84.07 8.6591 862.32 2.30744 870.25 23.11 307.76 60.48PhTSM 2.78 90.02 9.3911 934.76 2.26284 924.79 26.54 503.01 60.48PTSE] 1.39 79.87 8.1953 839.44 1.77844 863.06 25.4 928.4 60.48PTSM2 1.47 80.35 6.1953 763.98 1.13294 852.26 24.65 356.52 42.9ETSM 1.2 74.87 7.7315 816.56 1.24944 856.54 28.05 349.04 60.48KTSM 1.63 95.96 9.7398 854.6 1.39714 875.06 18.72 157.1 52.13ATSM 0.34 76.06 7.7315 816.44 1.02944 855.28 28.2 339.25 60.48ATS 1.2 74.87 7.7315 816.56 1.24944 856.54 28.05 349.04 60.48CTS �0.22 70.62 7.2677 838.28 �0.17106 873.91 34.64 320.53 88.46DTSM 1.31 85.22 8.6591 862.2 2.08744 868.55 23.23 297.97 60.48GTSM 1.2 74.87 7.7315 816.56 1.24944 856.54 28.05 349.04 60.48PTS 4.04 109.4 11.3625 1075.84 3.96688 999.95 22.89 636.34 60.48

Table 2. Stepwise Regression Equation

No. Equation n R2 Q2 s Overall F-test

1 Log P¼ {0.01592 (� 0.0199)} – VDW {0.1040 (�0.062)}þMR{0.1048 (�0.023)} 21 0.24 0.15 0.68 38.4522 Log P¼ {1.01504 (� 0.0648)} – LUMO {0.1026 (�0.142)}þ (VDW)2{0.1641 (�.058)} 21 0.40 0.25 0.62 69.1953 Log P¼ {1.2104 (� 0.1102)} – DD {0.0028 (�0.019)}þMR{0.1058 (�0.102)} 21 0.58 0.39 0.52 46.8764 Log P¼ {1.02708 (� 0.0812)} – DD {0.0102 (�0.092)}þ (VDW)2{0.1004 (�.014)} 20 0.65 0.91 0.19 92.1805 Log P¼ {2.41352 (� 0.0184)} – DD {0.0136 (�0.021)}þMR{0.8023 (�0.042)} 19 0.86 0.92 0.35 90.5326 Log P¼ {3.01698 (� 0.0590)} – LUMO {0.1248 (�0.068)}þ MR{0.3219 (�0.086)} 21 0.95 0.90 0.18 98.102

QSAR OF HYPOXIC IMAGING AGENT 119

-log P were calculated by the NCSS software, taking all thevalues between the upper limits and lower limits.

Since MR is a ‘‘corrected’’ form of the molar volume, itreflects the effect of size and polarizability, as indicated byEquation 1, suggesting that MR plays a significant role towardthe expressed biologic activities, which is probably due tosteric interactions occurring in the polar spaces. It has gener-ally been assumed that a positive coefficient with an MRterm in a correlation equation suggests a binding action viadispersion forces. Such binding could produce a concomi-tant conformational change in a macromolecular binding site;however, if the conformational change is detrimental, a neg-ative coefficient could result for the MR term. Negative coef-ficients with MR have also been assumed to reflect sterichindrance of one kind or another.

The log P values have been very useful in understandingthe distribution of a drug between different mediums. Thereare different methods of calculations of log P.

All the compounds are with low log P values. It mustbe noted that molecules having log P values below 5 are foundto have good drug absorption and permeability of com-pounds.

Conclusions

The results illustrate the effectiveness of the application ofGA to the method of developing metal ligand parametersfor the MM3* force field. These new parameters were uti-lized in the development of the QSPR model to predict, first,the lipophilicity of a series of common copper(II)-chelatingagents, which are specific for hypoxic imaging.

A multivariate regression analysis of the application ofthe QSPR methodology to the prediction of biodistributionwould be facilitated by a larger, more diverse dataset, anissue which we hope to address in future work.

Disclosure Statement

There is no conflict of interest among the authors and nocompeting financial interests exist.

References

1. Fahmi N, Singh RV. Spectroscopic, antifungal, and antibac-terial studies of some manganese heterochelates. J Ind ChemSoc 1996;73:257.

2. Klayman JF, Bartosevich TS, Griffin CJ, et al. 2-acetylpyridinethiosemicarbazone: A new class of potential antimalarialagent. J Med Chem 1979;22:855.

3. Pandeya SN, Ponnilavarasan I, Pandey A, et al. Evaluationof p-nitro phenyl substituted semicarbazide for anticonvul-sant properties. Pharmazie 1999;54:923.

4. Pandeya SN, Manjula H, Stables JP. Design of semi-carbazones and their isosteric analogues as potent anticon-vulsants. Pharmazie 2001;56:121.

5. Green MA. A potential copper radiopharmaceutical for im-aging the heart and brain: Copper-labeled pyruvalde-hyde bis(N4-ethylthiosemicarbazone). Nucl Med Biol 1987;14:59.

6. Fujibayashi Y, Taniuchi H, Yonekura Y, et al. Copper-62-ATSM: A new hypoxia imaging agent with high membranepermeability and low redox potential. J Nucl Med 1997;38:1155.

7. Dearling JLJ, Lewis JS, McCarthy DW, et al. Redox-activemetal complexes for imaging hypoxic tissues: Structure-activity relationships in copper(II) bis(thiosemicarbazone)complexes. J Chem Soc Chem Commun 1998;2531.

8. Dearling JLJ, Lewis JS, Mullen GED, et al. Copper-64-pyruvaldehyde-bis(N4-methylthiosemicarbazone) for theprevention of tumor growth at wound sites following La-paroscopic Surgery, Eur J Nucl Med 1998;25:788.

9. Dearling JLJ, Lewis JS, McCarthy DW, et al. Redox-activemetal complexes for imaging hypoxic tissues: Structure-activity relationships in copper(II) bis(thiosemicarbazone)complexes. Chem Commun 1998;22:2531.

10. Dearling J, Mullen G, Lewis J, et al. In: Nicolini M, Mazzi U(eds.), Technetium, Rhenium, and Other Metals in Chem-istry and Nuclear Medicine, Vol 5. Verona, Italy: CortinaInternational, 1990;599.

11. Lewis JS, McCarthy DW, McCarthy TJ, et al. Evaluation of64Cu-ATSM in vitro and in vivo in a hypoxic tumor model.J Nucl Med 1999;40:177.

12. Fujibayashi Y, Cutler CS, Anderson CJ, et al. Comparativestudies of Cu-64-ATSM and C-11-acetate in an acute myo-cardial infarction model: Ex vivo imaging of hypoxia in rats.Nucl Med Biol 1999;26:117.

13. Lewis JS, Sharp TL, Laforest R, et al. Tumor uptake ofcopper-diacetyl-bis(N(4)-methylthiosemicarbazone): Effect ofchanges in tissue oxygenation. J Nucl Med 2001;42:655.

14. Fujibayashi Y, Taniuchi H, Yonekura Y, et al. Hyperfixationof copper-62-PTSM in rat brain after transient global ische-mia. J Nucl Med 1997;38:1155.

FIG. 2. Relationship between experimental versus calculated log P.

120 SINGH ET AL.

15. McCarthy DW, Bass LA, Cutler PD, et al. High-purity pro-duction and potential applications of copper-60 and copper-61. Nucl Med Biol 1999;26:351.

16. Petering DH. In: Sigel H (ed.), Metal Ions in Biological Sys-tems, Vol 11. New York: Dekker, 1980;197.

17. Shibuya K, Fujibayashi Y, Yoshimi E, et al. Cytosolic=microsomal redox pathway: A reductive retention mecha-nism of a PET-oncology tracer, cu-Pyruvaldehyde-bis(N4-methylthiosemicarbazone) (cu-PTSM). Ann Nucl Med 1999;13:287.

18. Mathias CJ, Welch MJ, Raichle M, et al. Evaluation of a po-tential generator-produced PET tracer for cerebral perfusionimaging: Single-pass cerebral extraction measurements andimaging with radiolabelled Cu-PTSM. J Nucl Med 1990;31:351.

19. Taylor MR, Gabe EJ, Glusker JP, et al. The crystal struc-tures of compounds with antitumor activity. 2-keto-3-ethoxybutyraldehyde bis(thiosemicarbazone) and its cupriccomplex. J Am Chem Soc 1966;88:1845.

20. John E, Fanwick PE, McKenzie AT, et al. Structural charac-terization of a metal-based perfusion tracer: Copper(II)pyruvaldehyde bis(N4-methylthiosemicarbazone. Int J RadAppl Instrum B. 1989;16:791.

21. John EK, Green MA. Structure-activity relationshipsfor metal-labeled blood flow tracers: Comparison of keto-aldehyde bis(thiosemicarbazonato) copper(II) derivatives.J Med Chem 1990;33:1764.

22. Fujibayashi Y, Wada K, Taniuchi H, et al. Mitochondria-selective reduction of 62Cu-pyruvaldehyde bis(N4-methyl-thiosemicarbazone) (62Cu-PTSM) in the murine brain:A novel radiopharmaceutical for brain positron emissiontomography (PET) imaging. Biol Pharm Bull 1993;16:146.

23. Tiwari AK, Ojha H, Mishra AK. Quantitative structure–property relationship (correlation–analysis) of phosphonicacid–based chelates in design of MRI contrast agent. ChemBiol Drug Des 2009;74:87.

QSAR OF HYPOXIC IMAGING AGENT 121