investigation of dna-binding properties of organic molecules using quantitative structure-activity...
TRANSCRIPT
Investigation of DNA-Binding Properties of OrganicMolecules Using Quantitative Structure-ActivityRelationship (QSAR) Models
RAJESHWAR P. VERMA, CORWIN HANSCH
Department of Chemistry, Pomona College, 645 North College Avenue, Claremont, California 91711
Received 2 October 2006; revised 29 December 2006; accepted 2 January 2007
Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.21087
Corresponde(909)607-4249; FE-mail: rverma@
Journal of Pharm
� 2007 Wiley-Liss
88 JOURNAL
ABSTRACT: Due to the great potential of DNA as a receptor, many classes of syntheticand naturally occurring molecules exert their anticancer activities through DNA-bind-ing. In the field of antitumor DNA-binding agents, a number of acridine and anthracy-cline derivatives are in the market as chemotherapeutic agents. However, the clinicalapplication of such classes of compounds has encountered problems such as multi-drugresistance and secondary and/or collateral effects. Thus, there has been increasinginterest in discovering and developing small molecules that are capable of DNA-binding,which will be expected to be used either in place of or in conjunction with, the existingcompounds. The interest in the application of the QSAR paradigm has steadilyincreased in recent decades and we hope it may be useful in the design and developmentof DNA-binding molecules as new anticancer agents. In the present review, an attempthas been made to understand the DNA-binding properties of different compound seriesand discussed using 27 QSAR models, which reveal a number of interesting points.The most important determinants for the activity in these models are Hammettelectronic (s and sþ), hydrophobic, molar refractivity, and Sterimol width parameters.� 2007 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci 97:88–110, 2008
Keywords: DNA; QSAR; log P; cancer; computer aided drug designINTRODUCTION
Establishing the structure of DNA by Watson andCrick1 was possibly the greatest achievement ofbiological sciences of the 20th century. In recentyears, DNA has become familiar to everyonereading a daily newspaper, which contains inter-esting stories of the people accused of a particularcrime that has been found to be innocent and vice
nce to: Rajeshwar P. Verma (Telephone:ax: (909)607-7726;pomona.edu)
aceutical Sciences, Vol. 97, 88–110 (2008)
, Inc. and the American Pharmacists Association
OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUAR
versa. A search from SciFinder Scholar (2006Edition) of the Chemical Abstract reveals thatthere are over 1487000 publications (Journalarticles, Patents, and Abstracts) on DNA madeduring the years of 1981 and 2006 (from January’1981 to September’ 2006), which includes about18,000 publications on DNA-(Q)SAR [(quantita-tive) structure-activity relationships]. A histo-gram of publications on DNA and DNA-(Q)SARduring the years of 1981 and 2006 reflectsfluctuation of interest and research intensity(see Fig. 1).
Since all the life processes originate from DNA,the carrier of the genetic information, DNA can beconsidered as a macromolecular receptor withunlimited possibilities.2 Due to the great potential
Y 2008
Figure 1. Histogram of publications on DNA andDNA-(Q)SAR during the years of 1981–2006 (FromJanuary’ 1981 to September’ 2006).
QSAR MODELS FOR DNA-BINDING 89
of DNA as a receptor, many classes of syntheticand naturally occurring molecules exert theiranticancer activities through DNA-binding andtheir effectiveness depends upon the mode andintensity of the binding. There are mainly threetypes of DNA-binding: (i) covalent binding, (ii) non-intercalative groove binding, and (iii) intercalation.The first kind, that is, the covalent binding, takesplace when the drugs are bifunctional alkylatingagents. The other two types of bindings are not sostrong but involve weak forces like van der Waalsforce or hydrogen bonding. In DNA molecules,each base pair provides two grooves: a minor gro-ove and a major groove. Typically groove-bindingmolecules are composed of several heteroaromaticrings linked together through amide or otherfunctional groups, or directly through a singlebond.3 The stronger attachment of drug moleculeswith DNA is however caused by intercalation.Intercalators are those molecules that insert per-pendicularly into DNA without forming covalentbonds. The recognized forces that maintain thestability of the DNA-intercalators complex arevan der Waals, hydrogen bonding, hydrophobic,and/or charge transfer forces. A frontier orbitalinteraction has also been considered. This sug-gests that such a process has the possibility ofbeing reversed, and as a consequence it must haveequilibrium constant.4 Measurement of DNA-binding constant and biological activity of acridinederivatives, and their QSAR studies suggest thatthere should be a quantitative relationship bet-ween cytotoxic activity and the congener’s DNA-binding constant.5
Thus, the discovery of new molecules withantitumor activity has become one of the mostimportant goals in medicinal chemistry. Researchin this area has revealed a range of DNA
DOI 10.1002/jps JOUR
recognizing ligands that act as antitumor agents,including groove binders, alkylating and inter-calator compounds. In the field of antitumorDNA-binding agents, a number of acridine andanthracycline derivatives are already in themarket as chemotherapeutic agents. However,the clinical application of such classes of com-pounds has encountered problems such as multi-drug resistance (MDR) and secondary and/orcollateral effects. Thus, there has been increasinginterest in discovering and developing smallmolecules that are capable of DNA-binding, whichwill be expected to be used either in the place of orin conjunction with, the existing compounds. Thequantitative structure-activity relationship (QSAR)paradigm may be helpful in the design anddevelopment of DNA-binding molecules as newanticancer agents.
In the present review, an attempt has beenmade to collect the data on different series ofcompounds with respect to their DNA-bindingproperties, and has been discussed in terms ofQSAR (quantitative structure-activity relation-ship) to understand the chemical-biological inter-actions. In the past 44 years, the use of QSAR (oneof the well-developed areas in computationalchemistry), since the advent of this methodology,6
has become increasingly helpful in understandingmany aspects of chemical–biological interactionsin drug and pesticide research, as well as in theareas of toxicology. This method is useful in eluci-dating the mechanisms of chemical–biologicalinteractions in various biomolecules, particularlyenzymes, membranes, organelles, and cells, aswell as in human.7–9 It has also been utilized forthe evaluation of absorption, distribution, meta-bolism, and excretion (ADME) phenomena inmanyorganisms and whole animal studies.10 The QSARapproach employs extra-thermodynamically deri-ved and computational-based descriptors to corre-late biological activity in isolated receptors, cellularsystems and in vivo. Four standard moleculardescriptors routinely used in QSAR analysis:electronic, hydrophobic, steric, and topologicalindices. These descriptors are invaluable in help-ing to delineate a large number of receptor–ligandinteractions that are critical to biological pro-cesses.7 The quality of a QSAR model dependsstrictly on the type and quality of the data, and noton the hypotheses, and is valid only for thecompound structure analogs to those used to buildthe model. QSAR models can stand alone toaugment other graphical approaches or can beexamined in tandem with equations of a similar
NAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
90 VERMA AND HANSCH
mechanistic genre to establish their authenticityand reliability.11 Potential use of QSARmodels forscreening of chemical databases or virtuallibraries before their synthesis appears equallyattractive to chemical manufacturers, pharma-ceutical companies and government agencies.
MATERIALS AND METHODS
All the data/equations have been collected fromthe literature (see individual QSAR for respectivereferences). K is the DNA-binding constant inmolar concentration. Similarly, Kapp is the appar-ent DNA-binding constant in molar concentra-tion. DTm represents DNA-binding affinity ofthe molecules (DTm¼Tm (DNA-compound)�Tm (DNA);where Tm is the thermal denaturation tempera-ture in 8C). log K, log Kapp, and log DTm are thedependent variables, which define the biologicalparameter for QSAR equations. Physicochemicaldescriptors are auto-loaded, and multi-regressionanalyses (MRA) used to derive the QSAR are ex-ecuted with the C-QSAR program (www.biobyte.com).12 Selection of descriptors is made on thebasis of permutation and correlationmatrix amongthe descriptors (to avoid collinearity problems).The details about the C-QSAR program, the searchengine, the choice of parameters and their use inthe development of QSAR models, have alreadybeen discussed.13 The parameters used in thisreview have also been discussed in detail alongwith their application.7 Briefly, Clog P is thecalculated partition coefficient in n-octanol/waterand is a measure of hydrophobicity, and p is thehydrophobic parameter for substituents. s, sþ,and s� are Hammett electronic parameters thatapply to substituent effects on aromatic systems.B1, B5, and L are Verloop’s sterimol parametersfor substituents.14 B1 is a measure of the mini-mum width of a substituent, B5 is an attempt todefine maximum width of the whole substituentand L is the substituent length.
CMR is the calculated molar refractivity forthe whole molecule. MR is calculated from theLorentz–Lorenz equation and is described asfollows: [(n2�1)/(n2þ 2)](MW/d), where n is therefractive index, MW is the molecular weight, andd is the density of the substance. MR is dependenton volume and polarizability. It can be used for asubstituent or for the whole molecule. A newpolarizability parameter, NVE, was developed,which is shown to be effective at delineatingvarious chemico-biological interactions.15–18 NVE
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
represents the total number of valence electronsand is calculated by simply summing up thevalence electrons in a molecule, that is, H¼ 1,C¼ 4, Si¼ 4, N¼ 5, P¼ 5, O¼ 6, S¼ 6 andhalogens¼ 7. It may also be represented as:NVE¼nsþnpþnn, where ns is the number ofelectrons in s-orbital, np is the number of elec-trons in p-orbitals, and nn is the number of lonepair electrons. MgVol is the molar volume for thewhole molecule.19 The indicator variable I isassigned the value of 1 or 0 for special featureswith special effects that cannot be parameterizedand has been explained wherever used.
In QSAR equations, n is the number of datapoints, r is the correlation coefficient betweenobserved values of the dependent and the valuescalculated from the equation, r2 is the square ofthe correlation coefficient represents the goodnessof fit, q2 is the cross-validated r2 (a measure of thequality of the QSAR model), and s is the standarddeviation. The cross-validated r2 (q2) is obtainedby using leave-one-out (LOO) procedure.20 Q isthe quality factor (quality ratio), where Q¼ r/s.Chance correlation, due to the excessive numberof parameter (which increases the r and s valuesalso), can, thus, be detected by the examination ofQ value. F is the Fischer statistics (Fischer ratio),F¼ fr2/[(1�r2)m], where f is the number of degreeof freedom, f¼n�(mþ 1), n¼number of datapoints, and m¼number of variables. The model-ing was taken to be optimal when Q reached amaximum together with F, even if slightly non-optimal F values have normally been accepted. Asignificant decrease in F with the introduction ofone additional variable (with increasing Q anddecreasing s) could mean that the new descriptoris not as good as expected, that is, its introductionhas endangered the statistical quality of thecombination that nevertheless can again improvewith the ulterior introduction of a more convin-cing descriptor.21–23 Compounds were assigned tobe outliers on the basis of their high deviationbetween observed and calculated activities fromthe equation.24–26 Each regression equation in-cludes 95% confidence limits for each term inparentheses.
RESULTS AND DISCUSSION
Acridines
In the field of antitumor DNA-binding agents, theclass of 9-anilinoacridine derivatives plays an
DOI 10.1002/jps
QSAR MODELS FOR DNA-BINDING 91
important role either as the number of compoundsor as the importance of their anticancer proper-ties. The DNA-intercalating 9-anilinoacridinederivative amsacrine (I) is a useful clinical anti-leukemic drug, but has been shown to have only anarrow spectrum of clinical antitumor activity.Extensive structure-activity relationships for9-anilinoacridine derivatives lead to the develop-ment of a ‘‘second-generation’’ compound (II; Cl–921; NSC 343499) with activity against a broaderspectrum of experimental tumors (including solidtumor) and have been in the clinical trials.27
N
HN
H3CO NHSO2CH3
N
HN
H3CO NHSO2CH3
CH3 CONHCH3
I (Amsacrine) II (Cl-921; NSC 343499)
Baguley et al.28 derived a simple Eq. 1 withHammett’s sp for the DNA-binding of 10-X-9-anilinoacridines (III) to poly[d(A-T)], whereKAT isthe drug-DNA binding constant measured for thecompound to poly[d(A-T)].
N
HNX
3'
2'
1'
III
log KAT ¼ �0:52ð�0:10ÞsP þ 6:11ð�0:05Þ (1)
n ¼ 32; r2 ¼ 0:783; s ¼ 0:125; q2 ¼ 0:755;
Q ¼ 7:08; F1;30 ¼ 108:249
Note: Either the original equation was modifiedor added some statistical data (Eq. 1).
There is no significant relation between DNAbinding and p values of the 10-substituents.
The negative coefficient of sp suggests thatthe presence of highly electron-donating groupsat 10-position may enhance the DNA bindingproperties. On the basis of this equation, theDNA binding activity of compound III willincrease in the presence of electron-donatinggroups (e.g., NH2, NHMe, NMe2, NHC2H5,NHC3H7, NHC4H9, NHC6H13 etc.) and decreasein the presence of electron-withdrawing groups
DOI 10.1002/jps JOUR
(e.g., NO2, SO2Me, SO2NHMe, SO2NH2 etc.) at10-position.
It has been suggested that 9-anilinoacridinesbind to DNAs by intercalation of acridine chro-mophore between adjacent base pairs. Thus, theplane of 9-anilino ring should be twisted awayfrom that of the acridine, appears admirablysuited for the minor groove of the DNAs. In such abinding, 10-substituents could be placed directlyabove a DNA phosphate anion. Ion induced dipole(10-substituent) interactions would then beexpected with binding energies.28
Statistics of this equation was improvedby adding molar refractivity of X-substituents at10-position and shown by Eq. 2. In this equation,the coefficient of MRX�10 is very low. However, it ismuch larger than the 95% confidence interval.
log KAT ¼ �0:49ð�0:09Þsp
þ 0:06ð�0:04ÞMRX�10
þ 6:01ð�0:07Þ (2)
n ¼ 32; r2 ¼ 0:852; s ¼ 0:105; q2 ¼ 0:824;
Q ¼ 8:79; F2;29 ¼ 83:473
Note: Either the original equation wasmodified oradded some statistical data (Eq. 2).
These authors also obtained a good correlation(Eq. 3) between the binding constant and s andMR for their combined data, which included only10- and 20-substituted derivatives (III) and not the30-substituted ones.
log KAT ¼ �0:50ð�0:08Þs
þ 0:07ð�0:03ÞMRX�10
þ 6:01ð�0:05Þ (3)
n ¼ 41; r2 ¼ 0:846; s ¼ 0:104; q2 ¼ 0:825;
Q ¼ 8:85; F2;38 ¼ 104:377
Note: Either the original equation was modifiedor added some statistical data (Eq. 3).
In Eq. 3, MR values were used only for 10-substituent, while s values were used for the both10- and 20-substituents. This observation suggeststhat the electron-donating substituents at 10- and20-positions may strengthen the DNA-binding.The above equation does not allow any clue foradequate distinction between the s responsive-ness of DNA binding energies for 10- and 20-substituents.
NAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
92 VERMA AND HANSCH
Hansch et al.29 derived an Eq. 4 from thecombined DNA-binding data of 10,20,30-X-9-anili-noacridines (III) to poly[d(A-T)],28 which gave thebest correlation between the binding constant andsþX and B5X�30 showing the importance of electro-
nic and steric effects.
log KAT ¼ �0:28ð�0:05ÞsþX
� 0:41ð�0:05ÞB5X�30
þ 6:36ð�0:08Þ (4)
n ¼ 46; r2 ¼ 0:895; s ¼ 0:129; q2 ¼ 0:885;
Q ¼ 7:34; F2;43 ¼ 183:262
Note: Statistical data ‘‘Q’’ and ‘‘F’’ was added tothe original equation (Eq. 4).
In Eq. 4, sþX is the sum of sþ values for 10-, 20, and
30-substituents, while B5X�30 is the sterimolparameter for the largest width of 30-substituent,suggesting an unfavorable steric interaction atthis position. This is an important equation due toits high statistics (r2 is almost identical to q2), butit fails to predict the importance of sþ for thesubstituents at various positions, that is, 10, 20, or30, as similar to that of Eq. 3. Considering thesedrawback of Eq. 4, we developed Eq. 5 usingthe same data of Baguley et al.28, which gave agood correlation between the binding constantand sþ
X�10 and B5X�30 showing the importance ofelectronic and steric effects at positions 10 and 30,respectively. Biological and physicochemicalparameters used to derive QSAR 5 are shown inTable 1.
log KAT ¼ �0:30ð�0:09ÞsþX�10
� 0:24ð�0:06ÞB5X�30
þ 6:15ð�0:12Þ (5)
n ¼ 54; r2 ¼ 0:742; s ¼ 0:209; q2 ¼ 0:711;
Q ¼ 4:12; F2;51 ¼ 73:337
The negative coefficient with sþX�10 (�0.30)
implies that highly electron releasing substitu-ents at position 10 (e.g., NH2, NHMe, NMe2 etc.)enhance the DNA binding activity of compoundIII. The negative coefficient of B5X�30 suggeststhat the DNA binding of these compoundsdecreases as the width of the substituent atpositions 30 increases. It is interesting to note thatEq. 5 is more structural informative than that ofEq. 4, and has no outlier. With respect to Eq. 5,
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
there is a high mutual correlation between B5X�30
and MRX�30 (r2¼ 0.821, q2¼ 0.797). By consider-
ingMRX�30 in place ofB5X�30 , we can derive Eq. 5a(Tab. 1).
log KAT ¼ �0:31ð�0:09ÞsþX�10
� 0:40ð�0:10ÞMRX�30
þ 5:93ð�0:08Þ (5a)
n ¼ 54; r2 ¼ 0:741; s ¼ 0:209; q2 ¼ 0:716;
Q ¼ 4:12; F2;51 ¼ 72:956
MRX�30 is the molar refractivity of 30-substitu-ents. The statistics of Eqs. (5) and (5a) are almostidentical. Thus, it is very hard to predict for thisdata set, which steric parameter is more impor-tant B5X�30 or MRX�30 . We prefer sterimol widthparameter B5X�30 because it is more informativethan that ofMRX�30 . Molar refractivity (MR) is theparameter to measure the combined effect of bulkand polarizability.
Hansch et al.29 also derived Eq. 6 from theDNA-binding data of 10,20,30-X-9-anilinoacridines (III)that give 50% drop in fluorescence of ethidiumbound to DNA,28 which showed a best correlationbetween log 1/C and sþ
X and B5X�30 . Where, sþX is
the sum of sþ values for 10-, 20, and 30-substituents,while B5X�30 is the sterimol parameter for thelargest width of 30-substituent and C is the molarconcentration of drug to give a 50% drop influorescence of ethidium bound to DNA.
log 1=C ¼ �0:40ð�0:08ÞsþX
� 0:55ð�0:07ÞB5X�30
þ 5:50ð�0:11Þ (6)
n ¼ 43; r2 ¼ 0:873; s ¼ 0:169; q2 ¼ 0:857;
Q ¼ 5:53; F2;40 ¼ 137:480
Note: Statistical data ‘Q’ and ‘F’ was added tothe original equation (Eq. 6).
Eq. 6 has high statistics, but it also fails topredict the importance of sþ for the substituentsat various positions, that is, 10, 20, or 30, as similarto that of Eq. 4. Thus, we derived Eq. 7 in terms ofsþX�10 and B5X�30 using the same data (Tab. 1),
which shows the importance of electronic and
DOI 10.1002/jps
Table
1.
BiologicalandPhysicoch
emicalParametersUsedto
DeriveQSAREq.5
,Eq.5
a,E
q.7
,andEq.7
afortheBindingof
10 ,20 ,30 -X-9-A
nilinoa
cridines
(III)
toPoly[d(A
-T)]DNA
andto
Give50%
Dropin
Fluorescence
ofEthidium
Bou
ndto
DNA
No.
XlogKAT
(Obsd
.)
logKAT(E
q.5)
logKAT
(Eq.5a)
log1/C
(Obsd
.)
log1/C
(Eq.7)
log1/C
(Eq.7a)
sþ X�10
B5X�30
MRX�30
Pred.
DPred.
DPred.
DPred.
D
IIIa
10 –NO
25.48
5.67
�0.19
5.65
�0.17
4.72
4.56
0.16
4.55
0.17
0.79
1.00
0.10
IIIb
10 –SO
2Me
5.86
5.70
0.16
5.67
0.19
4.85
4.59
0.26
4.58
0.27
0.72
1.00
0.10
IIIc
10 –CN
5.70
5.71
�0.01
5.69
0.01
4.82
4.62
0.20
4.61
0.21
0.66
1.00
0.10
IIId
10 –SO
2NH
25.96
5.73
0.23
5.71
0.25
4.85
4.65
0.20
4.63
0.22
0.60
1.00
0.10
IIIe
10 –COMe
5.87
5.76
0.11
5.74
0.13
4.85
4.69
0.16
4.68
0.17
0.50
1.00
0.10
IIIf
10 –COOMe
5.88
5.76
0.12
5.74
0.14
4.82
4.70
0.12
4.68
0.14
0.49
1.00
0.10
IIIg
10 –CONH
25.83
5.80
0.03
5.78
0.05
4.75
4.76
�0.01
4.74
0.01
0.36
1.00
0.10
IIIh
10 –F
5.90
5.93
�0.03
5.91
�0.01
4.80
4.95
�0.15
4.94
�0.14
�0.07
1.00
0.10
IIIi
10 –Cl
5.99
5.88
0.11
5.86
0.13
4.85
4.87
�0.02
4.85
0.00
0.11
1.00
0.10
IIIj
10 –Br
6.02
5.87
0.15
5.85
0.17
4.96
4.85
0.11
4.84
0.12
0.15
1.00
0.10
IIIk
10 –I
6.19
5.87
0.32
5.85
0.34
5.09
4.85
0.24
4.84
0.25
0.14
1.00
0.10
IIIl
10 –NHSO
2C6H
56.20
6.21
�0.01
6.19
0.01
5.24
5.35
�0.11
5.35
�0.10
�0.98
1.00
0.10
IIIm
H5.86
5.91
�0.05
5.89
�0.03
4.80
4.92
�0.12
4.90
�0.10
0.00
1.00
0.10
IIIn
10 –NHCOMe
6.30
6.09
0.21
6.08
0.22
5.33
5.18
0.15
5.18
0.15
�0.60
1.00
0.10
IIIo
10 –Me
5.99
6.01
�0.02
5.99
0.00
5.00
5.05
�0.05
5.04
�0.04
�0.31
1.00
0.10
IIIp
10 –NHCOC6H
56.20
6.09
0.11
6.08
0.12
5.27
5.18
0.09
5.18
0.09
�0.60
1.00
0.10
IIIq
10 –OMe
6.12
6.15
�0.03
6.13
�0.01
5.13
5.26
�0.13
5.26
�0.13
�0.78
1.00
0.10
IIIr
10 –OH
6.28
6.19
0.09
6.18
0.10
5.29
5.33
�0.04
5.32
�0.03
�0.92
1.00
0.10
IIIs
10 –NHC6H
13
6.43
6.46
�0.03
6.45
�0.02
5.80
5.72
0.08
5.72
0.08
�1.81
1.00
0.10
IIIt
10 –NHC4H
96.45
6.46
�0.01
6.45
0.00
5.80
5.72
0.08
5.72
0.08
�1.81
1.00
0.10
IIIu
10 –NHC3H
76.35
6.46
�0.11
6.45
�0.10
5.70
5.72
�0.02
5.72
�0.02
�1.81
1.00
0.10
IIIv
10 –NHC2H
56.44
6.46
�0.02
6.45
�0.01
5.77
5.72
0.05
5.72
0.05
�1.81
1.00
0.10
IIIw
10 –NH
26.31
6.30
0.01
6.29
0.02
5.64
5.50
0.14
5.49
0.15
�1.30
1.00
0.10
IIIx
10 –NMe 2
6.51
6.42
0.09
6.42
0.09
5.70
5.67
0.03
5.67
0.03
�1.70
1.00
0.10
IIIy
10 –NHMe
6.55
6.46
0.09
6.45
0.10
5.92
5.72
0.20
5.72
0.20
�1.81
1.00
0.10
IIIz
20 –NO
25.56
5.91
�0.35
5.89
�0.33
4.57
4.92
�0.35
4.90
�0.33
0.00
1.00
0.10
IIIaa
20 –Cl
6.02
5.91
0.11
5.89
0.13
4.96
4.92
0.04
4.90
0.06
0.00
1.00
0.10
IIIab
20 –NHCOMe
5.90
5.91
�0.01
5.89
0.01
4.89
4.92
�0.03
4.90
�0.01
0.00
1.00
0.10
IIIac
20 –NHSO
2Me
5.73
5.91
�0.18
5.89
�0.16
4.82
4.92
�0.10
4.90
�0.08
0.00
1.00
0.10
IIIad
20 –OMe
5.86
5.91
�0.05
5.89
�0.03
4.85
4.92
�0.07
4.90
�0.05
0.00
1.00
0.10
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
QSAR MODELS FOR DNA-BINDING 93
Table
1.(C
ontinued
)
No.
XlogKAT
(Obsd
.)
logKAT(E
q.5)
logKAT
(Eq.5a)
log1/C
(Obsd
.)
log1/C
(Eq.7)
log1/C
(Eq.7a)
sþ X�10
B5X�30
MRX�30
Pred.
DPred.
DPred.
DPred.
D
IIIae
20 –OH
6.00
5.91
0.09
5.89
0.11
5.09
4.92
0.17
4.90
0.19
0.00
1.00
0.10
IIIaf
20 –Me
6.09
5.91
0.18
5.89
0.20
4.82
4.92
�0.10
4.90
�0.08
0.00
1.00
0.10
IIIag
20 –NH
26.11
5.91
0.20
5.89
0.22
5.07
4.92
0.15
4.90
0.17
0.00
1.00
0.10
IIIah
20 –NHMe
6.11
5.91
0.20
5.89
0.22
5.15
4.92
0.23
4.90
0.25
0.00
1.00
0.10
IIIai
30 –SO
2NH
25.67
5.42
0.25
5.89
�0.22
4.74
4.35
0.39
4.38
0.36
0.00
3.05
0.10
IIIaj
30 –CONH
25.37
5.42
�0.05
5.54
�0.17
4.42
4.34
0.08
4.49
�0.07
0.00
3.07
0.98
IIIak
30 –NO
25.36
5.57
�0.21
5.64
�0.28
4.44
4.52
�0.08
4.61
�0.17
0.00
2.44
0.74
IIIal
30 –I
5.33
5.64
�0.31
5.37
�0.04
3.90
4.60
�0.70
4.30
�0.40
0.00
2.15
1.39
IIIam
30 –Br
5.35
5.69
�0.34
5.58
�0.23
4.24
4.65
�0.41
4.54
�0.30
0.00
1.95
0.89
IIIan
30 –Cl
5.39
5.72
�0.33
5.69
�0.30
4.39
4.69
�0.30
4.67
�0.28
0.00
1.80
0.60
IIIao
30 –F
5.59
5.83
�0.24
5.90
�0.31
4.57
4.82
�0.25
4.91
�0.34
0.00
1.35
0.09
IIIap
30 –OH
5.36
5.69
�0.33
5.82
�0.46
4.41
4.66
�0.25
4.82
�0.41
0.00
1.93
0.29
IIIaq
30 –OMe
5.50
5.42
0.08
5.62
�0.12
4.47
4.34
0.13
4.59
�0.12
0.00
3.07
0.79
IIIar
30 –OC2H
55.20
5.35
�0.15
5.43
�0.23
4.07
4.26
�0.19
4.37
�0.30
0.00
3.36
1.25
IIIas
30 –Me
5.25
5.66
�0.41
5.71
�0.46
4.17
4.63
�0.46
4.69
�0.52
0.00
2.04
0.56
IIIat
30 –C2H
55.18
5.40
�0.22
5.52
�0.34
4.08
4.32
�0.24
4.47
�0.39
0.00
3.17
1.03
IIIau
30 –CHMe 2
5.08
5.40
�0.32
5.33
�0.25
3.85
4.32
�0.47
4.25
�0.40
0.00
3.17
1.50
IIIav
30 –CMe 3
5.23
5.40
�0.17
5.15
0.08
4.00
4.32
�0.32
4.04
�0.04
0.00
3.17
1.96
IIIaw
30 –NH
25.76
5.68
0.08
5.72
0.04
4.89
4.65
0.24
4.70
0.19
0.00
1.97
0.54
IIIaxa
30 –NHSO
2Me
5.49
5.17
0.32
5.20
0.29
5.14
4.05
1.09
4.10
1.04
0.00
4.13
1.82
IIIay
30 –NHCOOMe
5.56
5.20
0.36
5.29
0.27
4.77
4.09
0.68
4.22
0.55
0.00
3.99
1.60
IIIaz
30 –NHCOMe
5.84
5.29
0.55
5.33
0.51
4.38
4.19
0.19
4.26
0.12
0.00
3.61
1.49
IIIaaa
30 –COOMe
5.47
5.35
0.12
5.42
0.05
4.40
4.26
0.14
4.35
0.05
0.00
3.36
1.29
IIIaab
30 –N(M
e)SO
2Me
5.08
5.27
�0.19
5.00
0.08
4.43
4.16
0.27
3.86
0.57
0.00
3.72
2.34
aIIIaxwasnot
usedto
deriveEqs.
7and7a.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008 DOI 10.1002/jps
94 VERMA AND HANSCH
QSAR MODELS FOR DNA-BINDING 95
steric effects at positions 10 and 30 respectively.28
log 1=C ¼ �0:45ð�0:10ÞsþX�10
� 0:28ð�0:07ÞB5X�30
þ 5:19ð�0:14Þ (7)
n ¼ 53; r2 ¼ 0:773; s ¼ 0:246; q2 ¼ 0:742;
Q ¼ 3:57; F2;50 ¼ 85:132
outlier: X¼ 30–NHSO2MeEq. 7 suggests that the presence of highly
electron releasing groups at positions 10 and lowestvalue of largest width substituents at position 30
may increase the DNA binding activity of com-pound III that give 50% drop in fluorescence ofethidium bound to DNA. The importance of thisequation is to explain 53 compounds whereas Eq. 6explains only 43 compounds. One compound IIIax(X¼ 30–NHSO2Me) was not used in the derivationof Eq. 7 due to its high deviation from the observedactivity (Obsd.�Pred.¼ 5.14�4.05¼ 1.09> 3� s).With respect to Eq. 7, there is a high mutualcorrelation betweenB5X�30 andMRX�30 (r
2¼ 0.877,q2¼ 0.853). By considering MRX�30 in place ofB5X�30 , we can derive Eq. 7a (Tab. 1).
log 1=C ¼ �0:45ð�0:10ÞsþX�10
� 0:47ð�0:12ÞMRX�30
þ 4:95ð�0:09Þ (7a)
n ¼ 53; r2 ¼ 0:783; s ¼ 0:241; q2 ¼ 0:745;
Q ¼ 3:67; F2;50 ¼ 90:207
The statistics of Eqs. 7 and 7a are almostidentical. Thus, it is very hard to predict for thisdata set, which steric parameter is more impor-tant B5X�30 or MRX�30 . We prefer the sterimolwidth parameter B5X�30 because it is moreinformative than that of MRX�30 .
.In a subsequent QSAR study on a set of m-AMSA [40-(9-acridinylamino)methanesulfon-m-anisidine] analogues (IV), Baguley et al.5 deriveda correlation between log (1/D40) and log KAT andR2
m, where D40 is the molar dose of drug providinga 40% increase in the life span of leukemic mice,KAT is the drug-DNA binding constant measuredfor the compound to poly[d(A-T)], and Rm is themeasure of drug lipophilic-hydrophilic balancefrom reverse-phase partition chromatography.The published equation of Baguley et al.5 interms of logKAT andR2
m was very complex and notvery informative due to the use of ‘Rm’ descriptor
DOI 10.1002/jps JOUR
especially only in their squared (R2m) term. This is
the reason; we developed an Eq. 8 in terms of logKAT, CpX�2, and CpX�4 using the same data(Tab. 2).5
N
HN
H3CO NHSO2CH3
X 2
3
1
4
IV
log ð1=D40Þ
¼ 1:41ð�0:32Þ log KAT
� 0:90ð�0:37ÞCpX�2
� 0:32ð�0:16ÞCpX�4 � 3:01ð�1:84Þ (8)
n ¼ 45; r2 ¼ 0:750; s ¼ 0:355; q2 ¼ 0:692;
Q ¼ 2:44;F3;41 ¼ 41:000
outlier: X¼ 3 –CH(CH3)2In this equation, CpX�2, and CpX�4 are the
calculated hydrophobic parameters for the sub-stituents at positions 2 and 4, respectively. Thecontribution to the variance of CpX�2 (17.5%) islarger than that of CpX�4 (10.2%), suggesting thesubstituents at position-2 is more important thanthat of position-4. Negative coefficients of thehydrophobic parameters (CpX�2, and CpX�4) sug-gest that the less hydrophobic substituents atpositions 2- and 4- will be favored. Any role ofhydrophobicity or steric parameters for the sub-stituents at position 3 was not found. Onecompound IVi [X¼ 3–CH(CH3)2] was not used inthe derivation of Eq. 8 due to its high deviationfrom the observed activity (Obsd.�Pred.¼ 3.40�4.66¼�1.26> 3� s). By considering this com-pound, we can derive Eq. 8a.
log ð1=D40Þ
¼ 1:47ð�0:36Þ log KAT
� 0:86ð�0:41ÞCpX�2
� 0:34ð�0:18ÞCpX�4 � 3:39ð�2:08Þ (8a)
n ¼ 46; r2 ¼ 0:709; s ¼ 0:399; q2 ¼ 0:653;
Q ¼ 2:11; F3;42 ¼ 34:110
The compound IVi [X¼ 3–CH(CH3)2] is not wellpredicted by the above Eq. 8a (Obsd.�Pred.¼
NAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
Table 2. Biological and Physicochemical Parameters Used to Derive QSAR Eq. 8 for the Binding ofm-AMSA [40-(9-Acridinylamino)Methanesulfon-m-Anisidine] Analogues (IV) to Poly[d(A-T)] DNA and Molar Dose of Drug Providinga 40% Increase in the Life Span of Leukemic Mice (D40)
No. X
log 1/D40 (Eq. 8)
log KAT CpX�2 CpX�4Obsd. Pred. D
IVa H 5.29 4.82 0.47 5.57 0.00 0.00IVb 3–NHCOCH3 5.98 5.76 0.22 6.24 0.00 0.00IVc 3–NH2 5.52 5.72 �0.20 6.21 0.00 0.00IVd 3–NHCOOCH3 5.48 5.94 �0.46 6.37 0.00 0.00IVe 3–NHCH3 6.38 5.66 0.72 6.17 0.00 0.00IVf 3–NO2 5.38 4.93 0.45 5.65 0.00 0.00IVg 3–CH3 5.52 5.35 0.17 5.95 0.00 0.00IVh 3–CH2CH3 4.77 4.94 �0.17 5.66 0.00 0.00IVia 3–CH(CH3)2 3.40 4.66 �1.26 5.46 0.00 0.00IVj 3–OCH3 5.66 5.18 0.48 5.83 0.00 0.00IVk 3–F 5.07 4.77 0.30 5.54 0.00 0.00IVl 3–Cl 5.42 5.50 �0.08 6.06 0.00 0.00IVm 3–Br 5.36 5.83 �0.47 6.29 0.00 0.00IVn 3–I 5.56 5.91 �0.35 6.35 0.00 0.00IVo 3–CN 4.26 4.93 �0.67 5.65 0.00 0.00IVp 3–CONH2 4.83 4.94 �0.11 5.66 0.00 0.00IVq 3–CF3 4.17 4.35 �0.18 5.24 0.00 0.00IVr 2–NH2 5.87 5.84 0.03 5.95 �0.55 0.00IVs 2–CH3 3.69 4.06 �0.37 5.35 0.50 0.00IVt 2–CH(CH3)2 3.58 3.29 0.29 5.39 1.43 0.00IVu 2–F 4.73 4.23 0.50 5.28 0.20 0.00IVv 2–I 3.67 3.94 �0.27 5.70 1.18 0.00IVw 4–OCH3 5.34 5.24 0.10 5.94 0.00 0.28IVx 4–OCH2CH3 4.80 4.84 �0.04 5.77 0.00 0.81IVy 4–OCH2CH2OH 5.41 5.24 0.17 5.74 0.00 �0.59IVz 4–OCH2CHOHCH2OH 5.60 5.73 �0.13 5.90 0.00 �1.41IVaa 4–OCH2CONHCH3 5.51 5.45 0.06 5.80 0.00 �0.95IVab 4–O(CH2)3CONH2 5.26 5.21 0.05 5.74 0.00 �0.47IVac 4–CONH2 4.65 5.00 �0.35 5.47 0.00 �1.01IVad 4–CONHCH3 5.23 5.01 0.22 5.54 0.00 �0.72IVae 4–CON(CH3)2 4.74 4.46 0.28 5.04 0.00 �1.22IVaf 4–CO-cyclo–NC4H8O 3.91 4.30 �0.39 4.96 0.00 �1.05IVag 4–CONH(CH2)3CH3 4.38 4.30 0.08 5.40 0.00 0.86IVah 4–CONH(CH2)2OH 5.11 5.02 0.09 5.42 0.00 �1.29IVai 4–CONH(CH2)3OH 4.84 4.89 �0.05 5.40 0.00 �0.96IVaj 4–CONHCH2CHOHCH3 5.53 4.81 0.72 5.34 0.00 �0.98IVak 4–CONHCH2CHOHCH2OH 4.52 4.84 �0.32 5.26 0.00 �1.42IVal 4–CONHCH2CONH2 4.65 5.16 �0.51 5.39 0.00 �1.85IVam 4–CH3 5.74 5.30 0.44 6.03 0.00 0.50IVan 4–(CH2)2CONHCH3 5.56 5.29 0.27 5.65 0.00 �1.12IVao 4–C6H5 3.92 4.25 �0.33 5.60 0.00 1.89IVap 4–F 5.10 4.86 0.24 5.65 0.00 0.20IVaq 4–Cl 4.43 4.84 �0.41 5.76 0.00 0.77IVar 4–Br 4.50 4.52 �0.02 5.57 0.00 0.92IVas 4–CN 4.08 4.15 �0.07 5.01 0.00 �0.38IVat 4–NO2 3.90 4.32 �0.42 5.20 0.00 �0.09
aIVi was not used to derive Eq. 8.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008 DOI 10.1002/jps
96 VERMA AND HANSCH
QSAR MODELS FOR DNA-BINDING 97
3.40�4.62¼�1.22> 3� s). Thus, the considera-tion of this compound as an outlier is justified. Thereason for this outlier is not very clear. It may beattributed due to some experimental error.
For an another series of 3-X-5-Y-amsacrinederivatives (V) studied by Denny et al.,30 the DNAbinding constants were found to be correlatedwith electronic and steric parameters as shown byHansch et al.29 (Eqs. 9 and 10).
N
HN
H3CO NHSO2CH3
3
5 X
Y
V
log KAT ¼ �0:41ð�0:12Þsþ þ 0:94ð�0:22ÞB1X
þ 0:56ð�0:28ÞB1Y þ 3:86ð�0:57Þ (9)
n ¼ 21; r2 ¼ 0:850; s ¼ 0:139; q2 ¼ 0:770;
Q ¼ 6:63; F3;17 ¼ 32:111
log KGC ¼ �0:19ð�0:12Þsþ þ 0:54ð�0:21ÞB1Xþ 1:28ð�0:27ÞB1Y þ 3:86ð�0:55Þ
(10)
n ¼ 21; r2 ¼ 0:864; s ¼ 0:134; q2 ¼ 0:800;
Q ¼ 6:94; F3;17 ¼ 36:000
Note: Statistical data ‘Q’ and ‘F’ was added to theoriginal equation (Eqs. 9 and 10), where KAT andKGC are the drug-DNA binding constants mea-sured for the compound to poly[d(A-T)] andpoly[d(G-C)], respectively. B1X and B1Y are thesterimol parameters for the minimum width of Xand Y substituents, respectively. The positivecoefficient of B1X and B1Y suggests that anincrease in the value of B1 at X and Y positionwill increase the DNA binding activity for thecompounds (V) to poly[d(A-T)] and poly[d(G-C)].In both of the Eqs. 9 and 10, sþ is the sum of sþ
values for X and Y-substituents. Thus, it is veryhard to predict that the presence of highly electronreleasing substituents at which position X or Y orboth will enhance the DNA binding activity of thecompound V. To solve this problem, we derivedEqs. 11 and 12 from the same data used for Eqs. 9
DOI 10.1002/jps JOUR
and 10, respectively (Tab. 3).30
log KAT ¼ �0:41ð�0:13ÞsþX
þ 0:93ð�0:22ÞB1X
þ 0:31ð�0:26ÞB1Y þ 4:13ð�0:55Þ (11)
n ¼ 21; r2 ¼ 0:843; s ¼ 0:142; q2 ¼ 0:762;
Q ¼ 6:46; F3;17 ¼ 30:427
log KGC ¼ �0:19ð�0:12ÞsþX
þ 0:54ð�0:21ÞB1X
þ 1:16ð�0:24ÞB1Y þ 3:97ð�0:51Þ (12)
n ¼ 21; r2 ¼ 0:869; s ¼ 0:131; q2 ¼ 0:807;
Q ¼ 7:11; F3;17 ¼ 37:590
In these equations, sþX is the value of sþ of the
substituents at X-position only. On the basis ofEqs. 11 and 12, we can say that the presence ofhighly electron releasing substituents (e.g., NH2,OMe etc.) at position X as well as increase in thevalue of B1 at X and Y positions will increase theDNA binding activity for the compounds (V) topoly[d(A-T)] and poly[d(G-C)].
The same data of Denny et al.30 (used in thederivation of Eqs. 10 and 12) was also used byThakur et al.31 to derive a total number of elevenequations.We considered only one Eq. 13, which isstatistically the best.
log KGC ¼ 0:03ð�0:01ÞMR
þ 0:21ð�0:09ÞIDS
þ 0:11ð�0:08ÞIX þ 2:55 (13)
n ¼ 21; r2 ¼ 0:897; s ¼ 0:116; q2 ¼ 0:828;
Q ¼ 8:14; F3;17 ¼ 49:350
Note: Statistical data ‘r2’, ‘q2’, and ‘F’ was added tothe original equation (Eq. 13).
In the above equations, MR is the molarrefractivity, IDS is the indicator variable thataccounts for the presence of substituents at bothX- and Y-positions, while IX is for the presence ofsubstituents at only X-position. From the aboveequation, it may be concluded that the DNA-binding affinity of 3-X-5-Y-amsacrine derivatives(V) is structure specific in nature. The positivecoefficient of indicator variables IDS and IXsuggests that the di-substitution and substitution
NAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
Table 3. Biological and Physicochemical Parameters Used to Derive QSAR Eq. 11 and Eq. 12 for the Binding of3-X-5-Y-Amsacrine Derivatives (V) to Poly[d(A-T)] and Poly[d(G-C)] DNA
No. X Y
log KAT (Eq. 11) log KGC (Eq. 12)
sþX B1X B1YObsd. Pred. D Obsd. Pred. D
Va H H 5.57 5.37 0.20 5.65 5.67 �0.02 0.00 1.0 1.00Vb NH2 H 6.21 6.22 �0.01 6.13 6.11 0.02 �1.30 1.35 1.00Vc NO2 H 5.65 5.70 �0.05 6.13 5.90 0.23 0.79 1.70 1.00Vd Me H 5.95 5.98 �0.03 6.08 6.01 0.07 �0.31 1.52 1.00Ve OMe H 5.83 6.01 �0.18 5.97 6.01 �0.04 �0.78 1.35 1.00Vf Cl H 6.06 6.07 �0.01 5.98 6.08 �0.10 0.11 1.80 1.00Vg Br H 6.29 6.20 0.09 6.12 6.16 �0.04 0.15 1.95 1.00Vh I H 6.35 6.39 �0.04 6.20 6.27 �0.07 0.14 2.15 1.00Vi H CONH2 5.47 5.52 �0.05 6.13 6.25 �0.12 0.00 1.00 1.50Vj H CONHMe 5.54 5.53 0.01 6.18 6.30 �0.12 0.00 1.00 1.54Vk H CONHCH2CONH2 5.39 5.53 �0.14 6.40 6.30 0.10 0.00 1.00 1.54Vl NH2 CONHMe 6.29 6.39 �0.10 6.82 6.74 0.08 �1.30 1.35 1.54Vm NO2 CONH2 5.96 5.86 0.10 6.40 6.48 �0.08 0.79 1.70 1.50Vn NO2 CONHMe 5.71 5.87 �0.16 6.65 6.52 0.13 0.79 1.70 1.54Vo Me CONH2 6.40 6.13 0.27 6.30 6.59 �0.29 �0.31 1.52 1.50Vp Me CONHMe 6.22 6.15 0.07 6.68 6.64 0.04 �0.31 1.52 1.54Vq Me CONHCH2CONH2 6.00 6.15 �0.15 6.69 6.64 0.05 �0.31 1.52 1.54Vr OMe CONHMe 6.38 6.18 0.20 6.82 6.64 0.18 �0.78 1.35 1.54Vs Cl CONH2 6.33 6.23 0.10 6.58 6.66 �0.08 0.11 1.80 1.50Vt Cl CONHMe 6.29 6.24 0.05 6.65 6.71 �0.06 0.11 1.80 1.54Vu Cl CONHCH2CONH2 6.06 6.24 �0.18 6.83 6.71 0.12 0.11 1.80 1.54
98 VERMA AND HANSCH
at 3rd position will enhance their DNA-bindingaffinity.
In one of their study, Hansch et al.29 developedEq. 14 from the DNA-binding data of 10-X-3,6-Y-9-anilinoacridines (VI) to give 50% drop in fluores-cence of ethidium bound to DNA.32Where,C is themolar concentration of drug to give a 50% drop influorescence of ethidium bound to DNA.
N
HN
X
3
5
2
1
4
6
1'
YY
VI
log 1=C ¼ �0:49ð�0:07ÞsþX
� 0:36ð�0:07ÞsþY þ 5:19ð�0:08Þ (14)
n ¼ 39; r2 ¼ 0:872; s ¼ 0:178; q2 ¼ 0:850;
Q ¼ 5:25; F2;36 ¼ 122:625
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
Note: Statistical data ‘Q’ and ‘F’ was added to theoriginal equation (Eq. 14).
sþX values were used only for 10-substituent,
while sþY values were used for the both 3- and 6-
substituents. This equation suggests that thepresence of highly electron releasing substituentat position X and Y (e.g., NH2, OMe etc.) willenhance the DNA binding activity of compoundVI. Again, it is very hard to predict the presence ofhighly electron releasing substituents at whichposition of Y (3 or 6 or both) will enhance the DNAbinding activity. It is interesting to note that thedata set has a total number of 43 compounds andout of which, 24 compounds are unsubstitued atpositions-3 and 6; 14 compounds have –NH2 groupat position-3 but unsubstitued at positions-6; andthe rest 5 compounds have –NH2 group at bothpositions-3 and 6. Thus, the main observation forthis data set is to determine the presence of –NH2
group at which position of Y is beneficial. To solvethis problem, we derived Eq. 15 from the samedata (Tab. 4).29,32
log 1=C ¼ �0:46ð�0:08ÞsþX þ 0:52ð�0:13ÞI
þ 5:22ð�0:09Þ (15)
DOI 10.1002/jps
Table 4. Biological and Physicochemical Parameters Used to Derive QSAR Eq. (15) for the Binding of 10-X-3,6-Y-9-Anilinoacridines (VI) to Give 50% Drop in Fluorescence of Ethidium Bound to DNA
No. X Y
log 1/C (Eq. 15)
sþX IObsd. Pred. D
VIa OH H 5.54 5.64 �0.10 �0.92 0VIb NHC3H7 H 6.00 5.81 0.19 �1.30 0VIc NHC2H5 H 6.00 5.81 0.19 �1.30 0VId NH2 H 5.85 5.81 0.04 �1.30 0VIe NHC4H9 H 6.00 5.81 0.19 �1.30 0VIf NHCOMe H 5.57 5.50 0.07 �0.60 0VIg NHMe H 6.16 6.05 0.11 �1.81 0VIh NHSO2C6H4NH2 H 5.62 5.22 0.40 0.01 0VIi NHCONHMe H 5.70 5.34 0.36 �0.25 0VIj NHSO2Me H 5.43 5.21 0.22 0.03 0VIk NMe2 H 6.05 6.00 0.05 �1.70 0VIl NHCOC6H5 H 5.55 5.50 0.05 �0.60 0VIm NHSO2C6H5 H 5.43 5.67 �0.24 �0.98 0VIn OMe H 5.36 5.58 �0.22 �0.78 0VIo Br H 5.14 5.15 �0.01 0.15 0VIp NH(Me)SO2Me H 5.10 5.11 �0.01 0.24 0VIq SO2NH2 H 5.00 4.95 0.05 0.60 0VIr Cl H 4.96 5.17 �0.21 0.11 0VIs COMe H 4.92 4.99 �0.07 0.50 0VIt CONH2 H 4.89 5.06 �0.17 0.36 0VIu SO2NHMe H 4.85 4.96 �0.11 0.57 0VIv H H 4.82 5.22 �0.40 0.00 0VIw CN H 4.77 4.92 �0.15 0.66 0VIx NO2 H 4.64 4.86 �0.22 0.79 0VIy OH 3–MH2 5.89 6.16 �0.27 �0.92 1VIz NH2 3–NH2 6.22 6.33 �0.11 �1.30 1VIaa NHMe 3–NH2 6.52 6.56 �0.04 �1.81 1VIab H 3–NH2 5.41 5.74 �0.33 0.00 1VIac SO2NH2 3–MH2 5.57 5.47 0.10 0.60 1VIad CN 3–NH2 5.22 5.44 �0.22 0.66 1VIae SO2NHMe 3–NH2 5.64 5.48 0.16 0.57 1VIaf NHC4H9 3–NH2 6.52 6.33 0.19 �1.30 1VIag COMe 3–MH2 5.52 5.51 0.01 0.50 1VIah N(Me)SO2Me 3–NH2 5.70 5.63 0.07 0.24 1VIai NHCOC6H5 3–NH2 5.77 6.01 �0.24 �0.60 1VIaj Br 3–NH2 5.77 5.67 0.10 0.15 1VIak NHSO2Me 3–NH2 5.85 5.73 0.12 0.03 1VIal NO2 3–NH2 5.46 5.38 0.08 0.79 1VIam H 3,6–(NH2)2 5.80 5.74 0.06 0.00 1VIan OH 3,6–(NH2)2 6.05 6.16 �0.11 �0.92 1VIao NH2 3,6–(NH2)2 6.10 6.33 �0.23 �1.30 1VIap SO2NH2 3,6–(NH2)2 5.92 5.47 0.45 0.60 1VIaq NHSO2Me 3,6–(NH2)2 5.96 5.73 0.23 0.03 1
QSAR MODELS FOR DNA-BINDING 99
n ¼ 43; r2 ¼ 0:823; s ¼ 0:204; q2 ¼ 0:798;
Q ¼ 4:45; F2;40 ¼ 92:994
The indicator variable I is assigned the value of1 and 0 for the presence and absent of –NH2 group
DOI 10.1002/jps JOUR
at position-3. Negative coefficient of sþX (�0.46)
and positive coefficient of I (þ0.52) suggest thatthe presence of highly electron releasing sub-stituent at position X (e.g., NH2, OMe etc.)followed by –NH2 group at position-3 will enhancethe DNA binding activity of compoundVI. It must
NAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
NNH
N HN NH2
NH
HNH2N
Me
O
NH
O
O NH
Me
O
HN
NH2
YO
Netropsin (VII) VIIIa-d
O
HN
NH
YO
IXa-d
N
Cl
ClO
Figure 2. Structure of amidine analogs (VII–IX) used to derive Eqs. 16–18.
100 VERMA AND HANSCH
be note that the above Eq. 15 has no outlierwhereas the Eq. 14 has four outliers.
Amidine Analogs
A series of amidine analogs of chlorambucil wassynthesized by Bielawska et al.,33 where 5-[4-(N-alkylamidino)phenyl]-2-furancarboxamide andchlorambucil moiety are linked by a NH(CH2)2NHlinkage. The data from ethidium displacementassay indicated that these compounds bind in theminor groove of DNA and show moderate speci-ficity for AT base pair. From the DNA-bindingdata of these analogs (Fig. 2) to Calf thymus, T4,and poly(dG-dC)2 DNA, we developed QSAR 16,17, and 18, respectively. Biological and physico-chemical parameters used to derive these equa-tions are given in Table 5.
Table 5. Biological and Physicochemical Parameters UsedAnalogs (VII–IX) to Calf Thymus, T4, and Poly(dG-dC) DNA
No. Y
log Kapp (Eq. 16) lo
Obsd. Pred. D Obs
VII Netropsin 0.94 0.95 �0.01 0.VIIIa C(––NH)NH2 0.30 0.32 �0.02 0.VIIIb C(––NH)NH–cyclopropane 0.26 0.18 0.08 0.VIIIc C(––NH)NHCH(CH3)2 0.15 0.15 0.00 0.VIIId C(––NH)NH–cyclopentane 0.04 0.10 �0.06 0.IXa C(––NH)NH2 0.26 0.25 0.01 0.IXb C(––NH)NH–cyclopropane 0.20 0.11 0.09 0.IXc C(––NH)NHCH(CH3)2 0.08 0.08 0.00 0.IXd C(––NH)NH–cyclopentane �0.05 0.03 �0.08 �0.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
Binding of amidine analogs (VII–IX) to Calfthymus DNA
log Kapp ¼ �0:15ð�0:03ÞC log P
þ 0:07ð�0:02ÞCMR
� 0:23ð�0:15Þ (16)
n ¼ 9; r2 ¼ 0:962; s ¼ 0:064; q2 ¼ 0:890;
Q ¼ 15:328; F2;6 ¼ 75:947
Clog P versus CMR: r¼ 0.778Kapp is the apparent binding constant in molar
concentration. The Clog P parameter is of criticalimportance in describing the calf thymus DNA-binding. It alone accounts for 64.2% of thevariance in the data. Negative Clog P suggeststhat DNA-binding activity of these moleculedecreases with the increase of their hydrophobi-city. On the contrary, the increase in the molar
to Derive QSAR Eqs. 16–18 for the Binding of Amidine
g Kapp (Eq. 17) log Kapp (Eq. 17)
Clog P CMRd. Pred. D Obsd. Pred. D
92 0.93 �0.01 0.40 0.40 0.00 �2.28 11.4034 0.31 0.03 0.08 0.11 �0.03 0.20 7.6020 0.15 0.05 �0.05 �0.06 0.01 1.65 8.8608 0.12 �0.04 0.00 �0.09 0.09 1.91 8.9900 0.06 �0.06 �0.10 �0.16 0.06 2.54 9.7415 0.18 �0.03 �0.52 �0.33 �0.19 3.92 15.2115 0.02 0.13 �0.52 �0.50 �0.02 5.37 16.4600 �0.01 0.01 �0.52 �0.53 0.01 5.62 16.6015 �0.07 �0.08 �0.52 �0.60 0.08 6.25 17.35
DOI 10.1002/jps
QSAR MODELS FOR DNA-BINDING 101
refractivity of the whole molecule (CMR)increases DNA-binding activity of these com-pounds (positive coefficient).
Binding of amidine analogs (VII–IX) to T4 DNA
log Kapp ¼ �0:16ð�0:04ÞC log P
þ 0:06ð�0:03ÞCMR� 0:27ð�0:13Þ (17)
n ¼ 9; r2 ¼ 0:955; s ¼ 0:076; q2 ¼ 0:843;
Q ¼ 12:855; F2;6 ¼ 63:667
Eqs. 16 and 17 are very similar to each otherand expressing that these compounds (VII–IX)may target the same mode of binding in each ofthese two DNA types.
Binding of amidine analogs (VII–IX) topoly(dG-dC)2 DNA
log Kapp ¼ �0:12ð�0:03ÞC log P
þ 0:13ð�0:11Þ (18)
n ¼ 9; r2 ¼ 0:936; s ¼ 0:092; q2 ¼ 0:910;
Q ¼ 10:511; F1;7 ¼ 102:375
Hydrophobicity is found to be the single mostimportant parameter. The linear Clog P modelsuggests that the less hydrophobic molecules willbe more active. Eq. 18 explains 93.6% of thevariance in log Kapp.
In the above three Eqs. 16–18, the hydrophobicparameter (Clog P) is found to be of criticalimportance and explain a higher percentage of thevariance (64.2–93.6%) in log Kapp but havenegative coefficients (�0.12 to �0.16). This seemsto be strange because the Clog P values of thecompounds (VII–IX) are in a wide range (�2.28 to6.25). This may be due to the fact that thesecompound (VII–IX) explaining only the halfsecond part of the parabolic or bilinear relation.Thus, more compounds are needed to explain anddetermine the optimum hydrophobicity.
In subsequent research, two sets of amidinederivatives of non-planar tetracyclic system:tetrahydroquino[4,3-b][1]benzazepines (X) andtetrahydrobenzo[k]naphthyridine (XI), bearingthree types of side chains (hydroxyl, amine andalkyl) were synthesized by Eifler-Lima et al.34
These compounds were studied in DNA thermaldenaturation experiments using calf thymusDNA. All of these compounds were found to havesignificant DNA-binding affinity. The three-dimensional ‘‘envelope’’ shape of these moleculessuggests that they may bind to DNA via the
DOI 10.1002/jps JOUR
minor-groove rather than through an interactiveprocess, although no specific interaction with theDNA backbone is feasible. From the DNA-bindingaffinity data of these molecules in Table 6, wederived QSAR 19, which showed a linear correla-tion between the DNA-binding affinity and PSA ofthese molecules. PSA is the polar surface area ofthe molecule in A2 unit and calculated by usingSpartan’06 program.35
NH3C
H3C
N
CH3
H
NH
Y
NHN
H3CH
H3CH3C
HNY
Xa-d XIa-d
logDTm ¼ 0:02ð�0:01ÞPSA � 1:13ð�0:47Þ (19)
n ¼ 8; r2 ¼ 0:741; s ¼ 0:155; q2 ¼ 0:577;
Q ¼ 5:56; F1;6 ¼ 17:166
DTm represents DNA-binding affinity ofthe molecules (DTm¼Tm(DNA-compound)�Tm(DNA);where Tm is the thermal denaturation tempera-ture in 8C). The polar surface area (PSA) of amolecule is a good parameter to predict its(passive) absorption into the body. It is definedas the surface area of nitrogen and oxygen atomsin a molecule plus the surface of the hydrogenattached to these hetero-atoms. This has beenestablished that a PSA value of over 140 A2
generally yields molecules that are poorlyabsorbed from stomach and gastro-intestinaltract, whereas values below 60–70 A2 suggestpotential penetration of the blood-brain barrier,which, of course, is needed for CNS-activedrugs.36,37 It is interesting to note that the PSAvalues of all the compounds (X and XI) for thisdata set are 16.29–52.88 A2. To focus on thesubstituents contribution for this data set, we canuse an identity variable (I) to differentiate thesetwo ring systems series (X and XI) and deriveEq. 20 (Tab. 6).
logDTm ¼ 0:69ð�0:33ÞMRY � 0:25ð�0:21ÞI
� 1:44ð�0:60Þ (20)
n ¼ 8; r2 ¼ 0:881; s ¼ 0:115; q2 ¼ 0:726;
Q ¼ 8:17; F2;5 ¼ 18:508
NAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
Table 6. Biological and Physicochemical Parameters Used to Derive QSAR Eqs. (19) and (20) for the Binding ofAmidine Derivatives (X and XI) to Calf Thymus DNA
No. Y log DTm (Obsd.)
log DTm (Eq. 19) log DTm (Eq. 20)
PSA MRY IPred. D Pred. D
Xa (CH2)2CH3 �0.77 �0.82 0.05 �0.72 �0.05 16.29 1.39 1Xb (CH2)3OH �0.62 �0.43 �0.19 �0.62 0.00 36.08 1.54 1Xc (CH2)3NH2 �0.34 �0.34 0.00 �0.47 0.13 41.19 1.76 1Xd (CH2)4NH2 �0.24 �0.33 0.09 �0.15 �0.09 41.69 2.22 1XIa (CH2)2CH3 �0.54 �0.60 0.06 �0.48 �0.06 27.45 1.39 0XIb (CH2)3OH �0.46 �0.21 �0.25 �0.38 �0.08 47.71 1.54 0XIc (CH2)3NH2 �0.06 �0.11 0.05 �0.23 0.17 52.88 1.76 0XId (CH2)4NH2 0.06 �0.12 0.18 0.09 �0.03 52.21 2.22 0
102 VERMA AND HANSCH
MRY is the molar refractivity of the Y-sub-stituents. Positive MRY suggests that DNA-binding affinity of the molecule increaseswith the increase in their molar refractivity,which supports the author’s finding. Eifler-Lima et al.34 provided a rank order of DNA-binding for both series, which is exactly similar tothat of the molar refractivity of the Y-substitu-ents: Y¼ (CH2)4NH2> (CH2)3NH2> (CH2)3OH>(CH2)2CH3
The identity variable I is assigned the value of 1and 0 for the presence of the ring systemX andXI,respectively. Negative coefficient of the identityvariable suggests that the ring system XI will bepreferred over that ofX. With respect to QSAR 20,there is no significant correlation between CpY
(calculated hydrophobicity of Y-substituents)and MRY (r¼ 0.360). Thus, CpY cannot re-place MRY. It is interesting to note that Eq. 20is more informative and statistically better thanthat of Eq. 19.
Anthrapyrazole Derivatives
The binding constants of a series of anthrapyr-azole derivatives (XII) for the interaction withDNA have been determined by Hartley et al.38
using an ethidium displacement assay, whichindicates that the binding is influenced not only bythe nature of the side chains but also by thenumber and position of hydroxyl groups on thechromophore. From the DNA-binding data ofthese analogues (XII) in Table 7, we developedQSAR 21.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
N
O
N
R2R3
R6
R5
R4
R1
XIIa-l
log Kapp ¼ 0:46ð�0:21ÞCMR�R1
þ 0:47ð�0:35ÞI þ 6:62ð�0:43Þ (21)
n ¼ 11; r2 ¼ 0:874; s ¼ 0:206; q2 ¼ 0:797;
Q ¼ 4:54; F2;8 ¼ 27:746
outlier: XIIlIndicator variable I¼ 1 and 0 is for the presence
and absent of –OH group at position R3. Thepositive coefficient of the indicator variable(þ0.47) suggests that the presence of the –OHgroup at R3 position will increase the bindingactivity of anthrapyrazole derivatives (XII) toDNA. CMR-R1 is the calculated molar refractivityof the R1 substituents and can not be replaced byhydrophobic parameter of these substituents. Bysubstitutiong Cp-R1 for CMR-R1 in Eq. 21 gave avery poor fit (r2¼ 0.810, q2¼�6.278), the presenceof negative q2 will never be acceptable. Thepositive coefficient of CMR-R1 (þ0.46) suggeststhat the R1 substituents having high molarrefractivity/polarizability will be more active.One compound (XIIl) was not used in the deri-vation of Eq. 21 due to its high deviation from theobserved activity (Obsd.�Pred.¼ 7.81�6.83¼0.98> 3� s). The reason for this outlier (XIIl) is
DOI 10.1002/jps
Table
7.
BiologicalandPhysicoch
emicalParametersusedto
DeriveQSAR
Eq.21fortheBindingof
Anthrapyrazole
Derivatives
(XII)to
DNA
No.
R1
R2
R3
R4
R5
R6
logKapp(E
q.21)
CMR-R
1I
Obsd
.Pred.
D
XIIa
(CH
2) 2NH(C
H2) 2OH
NH(C
H2) 2NH(C
H2) 2OH
HH
HH
7.87
7.71
0.16
2.38
0XIIb
HNH(C
H2) 2NH(C
H2) 2OH
HH
HH
6.60
6.62
�0.02
0.00
0XIIc
(CH
2) 2NH(C
H2) 2OH
NH(C
H2) 2NH(C
H2) 2OH
OH
HH
H8.00
8.18
�0.18
2.38
1XIId
(CH
2) 2NH(C
H2) 2OH
NH(C
H2) 2NH(C
H2) 2OH
HH
HOH
7.57
7.71
�0.14
2.38
0XIIe
(CH
2) 2NH(C
H2) 2OH
NH(C
H2) 2NH(C
H2) 2OH
OH
HH
OH
8.30
8.18
0.12
2.38
1XIIf
CH
2CH
2OH
NH(C
H2) 2NH(C
H2) 2OH
OH
HH
OH
7.61
7.58
0.03
1.08
1XIIg
(CH
2) 2NH(C
H2) 2OH
NH(C
H2) 5NH
2OH
HH
OH
8.43
8.18
0.25
2.38
1
XIIh
(CH
2) 2NH(C
H2) 2OH
OH
HH
OH
8.38
8.18
0.20
2.38
1
XIIi
(CH
2) 2NH(C
H2) 2OH
NH(C
H2) 2NH(C
H2) 2OH
OH
OH
HOH
8.28
8.18
0.10
2.38
1XIIj
(CH
2) 2NH(C
H2) 2OH
NH(C
H2) 2NH(C
H2) 2OH
OH
HOH
OH
7.90
8.18
�0.28
2.38
1XIIk
(CH
2) 2NH(C
H2) 2OH
NH(C
H2) 2NHCH
3OH
OH
HOH
7.94
8.18
�0.24
2.38
1XIIla
CH
3H
NH(C
H2) 2NH–(C
H2) 2OH
HH
H7.81
6.83
0.98
0.46
0
aXIIlwasnot
usedto
deriveEq.21.
DOI 10.1002/jps JOUR
QSAR MODELS FOR DNA-BINDING 103
not very clear, possibly it is due to the presence ofa large groupR3¼NH(CH2)2NH(CH2)2OH insteadof R3¼H or OH. Considering this outlier (XIIl) inthe derivation of Eq. 21, gave a very poor fit(r2¼ 0.648, q2¼ 0.105), which is not acceptable.
Bis-Guanylhydrazones
In a QSAR study from the published data of Daveet al.,39 on a series of aliphatic and aromaticbis(guanylhydrazones) (XIII), Denny and Cain40
showed that the drug concentration necessary toinhibit L1210 DNA-dependent DNA polymerasein vitro by 50% (IC50) was linearly related to themeasure of drug-DNA binding (C50) with nopreference for a particular primary sequence ofDNA as evidenced by Eqs. 22 and 23.
R2 NH
NN
XN
N NH
R2
NH
R1 R1
NH
XIII
log IC50 ¼ �0:49ð�0:12Þ log½1=C50ðA� TÞ�
þ 2:00 (22)
n ¼ 13; r2 ¼ 0:846; s ¼ 0:220; q2 ¼ 0:810;
Q ¼ 4:18; F1;11 ¼ 64:6
log IC50 ¼ �0:47ð�0:11Þ log½1=C50ðG� CÞ�
þ 1:74 (23)
n ¼ 13; r2 ¼ 0:828; s ¼ 0:240; q2 ¼ 0:774;
Q ¼ 3:79; F1;11 ¼ 51:5
Note: Statistical data ‘r2’, ‘q2’, and ‘Q’ was added tothe original equation (Eqs. 22 and 23).
Polyamines
An ethidium bromide displacement assay wasemployed by Rege et al.41 to evaluate the DNA-binding affinity of an aminoglycoside-polyaminelibrary (XIV–XXXI) in a 96-well formate. Thepercent fluorescence decreased value (%F) wasused as a parameter to rank the DNA-binding
NAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
Table 8. Biological and Physicochemical Parameters used to Derive QSAR Eq. 24 for the Binding of Polyamines(XIV–XXXI) to DNA
No. Compound/Substituent
log (%F) (Eq. 24)
CMRObsd. Pred. D
XIVa Spermidine �0.62 0.68 �1.30 4.53XV Ethylenediamine 0.05 0.56 �0.51 1.84XVI N8–Acetyl-spermidine 0.31 0.73 �0.42 5.49XVII Diethylenetriamine 0.52 0.62 �0.10 3.14XVIII Paromomycin 0.79 1.13 �0.34 14.12XIX Bekanamycin 0.79 1.00 �0.21 11.37XX Spermine 0.86 0.77 0.09 6.29XXI Apramycin 0.90 1.07 �0.17 12.74XXIIa Neamine (R––H) 0.94 0.84 0.10 7.78XXIIb R––CONHCH2CH2NH2 1.47 1.24 0.23 16.44XXIIc R––(CH2)3NH(CH2)4NH(CH2)3NH2 1.84 2.07 �0.23 34.23XXIII Streptomycin 0.96 1.10 �0.14 13.33XXIV Pentaethylenehexamine 1.05 0.80 0.25 7.03XXVa Neomycin(R––H) 1.17 1.14 0.03 14.34XXVb R––CONHCH2CH2NH2 1.73 1.75 �0.02 27.32XXVIa R1––R2––H, R3––CONH(CH2)3NH(CH2)4NH(CH2)3NH2 1.10 1.07 0.03 12.75XXVIb R1––R3––H, R2––CONH(CH2)3NH(CH2)4NH(CH2)3NH2 1.24 1.07 0.17 12.75XXVIc R1––R2––R3––CONH(CH2)3NH(CH2)4NH(CH2)3NH2 1.51 1.69 �0.18 25.98XXVIIa R––CONHCH2CH2NHCH2CH2NH2 1.29 1.14 0.15 14.27XXVIIb R––CONH(CH2)3NH(CH2)4NH(CH2)3NH2 1.75 1.43 0.32 20.57XXVIIIa R1––NH2, R2––CONHCH2CH2NH2 1.51 1.41 0.10 20.03XXVIIIb R1––OH, R2––CONHCH2CH2NHCH2CH2NH2 1.69 1.64 0.05 25.00XXVIIIc R1––OH, R2––CONH(CH2)3NH(CH2)4NH(CH2)3NH2 1.92 2.23 �0.32 37.60XXIX Kanamycin-based derivative 1.82 1.94 �0.12 31.31XXX R1––R2––NH(CH2)3NH(CH2)4NH(CH2)3NH2 1.67 1.23 0.44 16.14XXXIa R––a–OCH2C6H5 1.71 1.41 0.30 20.12XXXIb R––a–OCH3 1.77 1.30 0.47 17.61
aSpermidine (XIV) was not used to derive Eq. 24.
104 VERMA AND HANSCH
efficacy of the library constituents. We developedQSAR 24 from the data in Table 8 (Fig. 3).
logð%FÞ ¼ 0:05ð�0:01ÞCMRþ 0:47ð�0:22Þ (24)
n ¼ 26; r2 ¼ 0:742; s ¼ 0:263; q2 ¼ 0:689;
Q ¼ 3:27; F1;24 ¼ 69:023
outlier: SpermidineCMR is one of the most important single
parameters for this dataset. Positive CMR sug-gests that the molecules having high molarrefractivity/polarizability will be more active.With respect to QSAR 24, there is not a significantcorrelation between Clog P and CMR (r¼ 0.670).Thus, ClogP cannot replace CMR. By substitutingClog P for CMR in Eq. 24 gave a very poor fit(r2¼ 0.204, q2¼ 0.074), which is not acceptable.One compound (Spermidine) has been omitted dueto its high deviation from the observed value
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
(Obsd.�Pred.¼�0.62�0.68¼�1.30> 3� s). Byconsidering spermidine, we can derive Eq. 24a.
logð%FÞ ¼ 0:05ð�0:02ÞCMR
þ 0:32ð�0:28Þ (24a)
n ¼ 27; r2 ¼ 0:673; s ¼ 0:358; q2 ¼ 0:605;
Q ¼ 2:29; F1;25 ¼ 51:453
Again for spermidine, the deviation (Obsd.�Pred.¼�0.62�0.56¼�1.18> 3� s) is very high.This is the reason, spermidine was considered tobe an outlier and we kept Eq. 24, which has alsobetter statistics than that of Eq. 24a.
Sandramycin Analogs
Sandramycin (XXXIIa) is a potent antitumorantibiotic. It possesses a twofold axis of symmetry
DOI 10.1002/jps
H2N
HN NH2
XIV
H2NNH2
XV
H2N NH
HN
XVIO
XXIII
O
HO
HO
NH2
O
H2N NH2
H
O
OH
OH
HOO
OH
O
OH
H2N
H
OH
HO
NH2
XVIII
O
O
O
O
OH
HO
OH
NH2
HOHH2N
H
NH2
OH
NH2HO
H2N
HN
NH2
NH2
XVII
XIX
NH2
HN
HN
NH2
XX
OO
OO NH2
OHH
H
HOOH
OH
H
OH
NH
Me
HO
NH2
H2NNH2
H
XXI
NHR
HO
OHO
HO
O
NHR
NHR
NHRHO
XXIIa-c
O
HN
O O
O
NH2
NH
OH
HO
OH
HN
OH
OH
H
HOOH
HN Me
H2N
HN
H
OHCMe
NH
NH
NH
NH
NH2
NH2
XXIV
NHR
O
OHO
HO
O
NHR
NHR
NHRHOO
OHO
HO
O
NHRRHN
HOHO XXVa,b
N
NN
N
NHR1
O
OR2
R3O
XXVIa-c
O
OHO
OMeHO
O
NHRRHN
HOHO
XXVIIa-b
OHO
HO
OR2
OHO
O
OH
OHR2HN
R2HN
OHO
NHR2
R1HO
HOO
NHR2
XXVIIIa-c
OHO
O
OH
OHHN
HN
OHO
NH
OHHO
HOO
NH
HN
NH
O
O
HNHN
NH
NH
O
HN NH2
NH
NH
HN
O
NH2
XXIX
NHR1
OH
XXX
O
HO NH
O
NH
HN
O
NHHN
HN
NH2
O
XXXIa-b
RHO
NH NH2
Figure 3. Structure of polyamines (XIV–XXXI) used in the formulation of QSAR 24.
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
QSAR MODELS FOR DNA-BINDING 105
106 VERMA AND HANSCH
and two heteroaromatic chromophores thatresults in sequence-selective DNA bis-intercala-tion spanning two base pairs, preferentially at 50-AT sites. A series of sandramycin analogs wassynthesized by Boger et al.,42 where each analogcontained a deep-seated structural change in thechromophore including the deletion of key func-tional groups or core structural elements capableof revealing its role in the high affinity bis-intercalation binding. Fluorescence quenchingstudies were employed to establish the DNA-binding affinity of sandramycin (XXXIIa) and itschromophore analogs (XXXIIb-k) for calf thymusDNA and 50-d(GCATGC)2. From the DNA-bindingdata of these analogs, we developed QSAR 25 and26.
Binding of sandramycin and its analogs(XXXIIa–k) to Calf thymus DNA (Tab. 9)
N
HN
OMeN
N
HN
OH
R
OH
O Me
O
O
HN
NNO
O
OMe
MeH
O O
N
ONH
R
H
XXXIIa-k
log Kapp ¼ 3:73ð�0:93ÞCp�R
� 0:52ð�0:12Þ ðCp�RÞ2
þ 1:65ð�0:36Þ (25)
n ¼ 10; r2 ¼ 0:939; s ¼ 0:166; q2 ¼ 0:815;
Q ¼ 5:837; F ¼ 53:877
optimum Cp�R¼ 3.56 (3.42�3.69)It is a parabolic correlation in terms of Cp�R
(calculated hydrophobicity of R-groups), whichsuggests that the DNA-binding affinity of san-dramycin analogs (XXXIIa–k) to calf thymusDNA first increases with an increase in hydro-phobicity of theR-groups up to an optimum Cp�Rof 3.56 and then decreases.
Binding of sandramycin and its analogs(XXXIIa–k) to 50-d(GCATGC)2 DNA (Tab. 9).
log Kapp ¼ 4:67ð�1:15ÞCp�R
� 0:64ð�0:15Þ ðCp�RÞ2
� 2:00ð�1:00Þ (26)
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
n ¼ 11; r2 ¼ 0:918; s ¼ 0:216; q2 ¼ 0:763;
Q ¼ 4:435; F2;8 ¼ 44:780
optimum Cp�R¼ 3.68 (3.55�3.80)QSAR Eqs. 25 and 26 are very similar to each
other suggesting that the role of the R-groups ofsandramycin analogs may be identical for bis-intercalation binding to both calf thymus DNAand 50-d(GCATGC)2.
Thioacridone Derivatives
It has already been established that acronycine(acronine, XXXIII) possess anticancer activityagainst a wide variety of solid human tumors,including ovarian cancer and metastatic tumor ofunknown origin. A series of thioacridone deriva-tives (XXXIV) related to anticancer alkaloidacronycine was synthesized by Dheyongeraet al.43 The binding constant of these compoundswith DNA was measured to determine theirdegree of intercalation with DNA in vitro. Wedeveloped QSAR 27 from the data in Table 10.
N O
CH3
CH3
OCH3
CH3
O
XXXIII(Acronycine)
Binding of thioacridone derivatives (XXXIV) toDNA (Tab. 10).
N
R2
R1
R3
S
XXXIVa-o
log K ¼ �0:97ð�0:38ÞC log P
þ 0:97ð�0:22ÞCMR�R1
þ 3:19ð�1:58Þ (27)
n ¼ 15; r2 ¼ 0:927; s ¼ 0:443; q2 ¼ 0:854;
Q ¼ 2:17; F2;12 ¼ 76:192
K is the DNA-binding constant in molarconcentration. The CMR-R1 is the calculatedmolar refractivity of R1 groups and is foundto be of critical importance in describing the
DOI 10.1002/jps
Table 9. Biological and Physicochemical Parameters Used to Derive QSAR Eq. 25 and Eq. 26 for the Binding ofSandramycin Analogs (XXXIIa–k) to Calf Thymus and 50-d(GCATGC)2 DNA
No. R
log Kapp (Eq. 25) log Kapp (Eq. 26)
Cp�RObsd. Pred. D Obsd. Pred. D
XXXIIa
N
OH
O(Sandramycin)
7.20 7.00 0.20 7.81 7.57 0.24 3.54
XXXIIb
N
OMe
O
6.57 6.80 �0.23 7.11 7.23 �0.12 2.94
XXXIIc
N
O
6.51 6.52 �0.01 7.15 6.85 0.30 2.60
XXXIId
N
OH
O
5.92 5.97 �0.05 6.11 6.11 0.00 2.16
XXXIIe
N
OH
O
MeO 6.92 6.97 �0.05 7.41 7.57 �0.16 3.82
XXXIIf
N
O
MeO 6.68 6.75 �0.07 6.81 7.17 �0.36 2.87
XXXIIg
N
OH
O
Me 6.76 6.89 �0.13 7.40 7.50 �0.10 4.04
XXXIIh
N
OH
O
Cl
6.88 6.72 0.16 7.56 7.33 �0.23 4.30
XXXIIi
N
OCH2Ph
O
Cl
5.11 5.15 �0.04 5.54 5.59 �0.05 5.45
XXXIIj N
O
ND 6.29 — 6.54 6.53 0.01 2.39
XXXIIk
N
O
6.51 6.29 0.22 6.53 6.53 0.00 2.39
ND, not determined.
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
QSAR MODELS FOR DNA-BINDING 107
Table 10. Biological and Physicochemical Parameters Used to Derive QSAR Eq. 27 for the Binding of ThioacridoneDerivatives (XXXIVa–o) to DNA
No. R1 R2 R3
log K (Eq. 27)
Clog P CMR-R1Obsd. Pred. D
XXXIVa NH(CH2)2N(CH3)2 H CH3 2.93 2.79 0.14 2.99 2.59XXXIVb NH(CH2)2N(CH3)2 Cl CH3 2.04 1.94 0.10 3.86 2.59XXXIVc NH(CH2)2N(CH3)2 CH3 CH3 2.36 2.30 0.06 3.49 2.59XXXIVd NH(CH2)2N(CH3)2 H H 4.41 3.48 0.93 2.28 2.59XXXIVe NH(CH2)2N(CH3)2 Cl H 2.08 2.44 �0.36 3.35 2.59XXXIVf NH(CH2)2N(CH3)2 CH3 H 2.43 2.99 �0.56 2.78 2.59XXXIVg N(CH2CH2Cl)2 H H 2.04 2.47 �0.43 3.93 3.21XXXIVh N(CH2CH2Cl)2 Cl H 1.95 1.56 0.39 4.86 3.21XXXIVi N(CH2CH2Cl)2 CH3 H 1.78 1.99 �0.21 4.42 3.21XXXIVj Cl H H �0.10 0.30 �0.40 3.46 0.49XXXIVk Cl Cl H �0.22 �0.45 0.23 4.23 0.49XXXIVl Cl CH3 H �0.15 �0.19 0.04 3.96 0.49XXXIVm Cl H CH3 �0.52 �0.03 �0.49 3.80 0.49XXXIVn Cl Cl CH3 �0.40 �0.74 0.34 4.52 0.49XXXIVo Cl CH3 CH3 �0.30 �0.52 0.22 4.30 0.49
108 VERMA AND HANSCH
DNA-binding. It alone accounts for 73.4% of thevarience in the data.
CONCLUSIONS
In the present review, an attempt has been madeto collect the data and present a total number of 27QSAR models (11 published and 16 newlyformulated QSARs) on seven different compoundseries that are acridine, amidine, anthrapyrazole,bis-guanylhydrazone, polyamine, sandramycin,and thioacridone for their DNA-binding activities,which have been found to be well correlated with anumber of physicochemical and structural para-meters. The important parameters for thesecorrelations are Hammett electronic (s and sþ),hydrophobic, molar refractivity, and Sterimolwidth parameters. The difference in the QSARequations may be either due to the presence ofdifferent compound series or the involvement ofdifferent DNA-binding mechanism.
The major study of this review is for acridinederivatives (9-anilinoacridines) that contains 15equations. Hammett electronic parameter (s orsþ) is one of the most important determinants forthe binding of 9-anilinoacridines to DNAs. Out of15 QSAR, 13 contain a correlation between DNA-binding activity and Hammett electronic para-meter (s or sþ). It may be due to the fact that 9-anilinoacridines bind to DNAs by intercalation ofacridine chromophore between adjacent basepairs. Thus, the plane of 9-anilino ring should
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
be twisted away from that of the acridine, appearsadmirably suited for the minor groove of theDNAs. In such a binding, 10-substituents (parasubstituents of the aniline ring) could be placeddirectly above a DNA phosphate anion. Ioninduced dipole (10-substituent) interactions wouldthen be expected with binding energies. This maybe the result of strong correlation between DNAbinding and Hammett electronic parameter (s orsþ) of the 10-substituents, which is found in sixequations (Eqs. 1, 2, 5, 7, 14, and 15) with negativecoefficient (�0.30 to �0.52) suggesting that thepresence of highly electron releasing substituentsat position 10 (e.g., NH2, NHMe, NMe2 etc.) willenhance the DNA binding activity of 10-substi-tuted-9-anilinoacridines. A negative linear corre-lation between DNA binding and Sterimol widthparameter (B5) of the 30-substituents (orthosubstituents of the aniline ring) is found in fourequations (Eqs. 4, 5, 6, and 7), and the coefficientrange from �0.24 (Eq. 5) to �0.55 (Eq. 6). Thissuggests that the presence of largest widthsubstituents at the ortho position of the anilinering will decrease the DNA binding activity of10,20,3-substituted-9-anilinoacridines. The aboveobservations are not true for the amsacrinederivatives (IV and V) because the ortho andpara positions of the aniline ring are alreadyoccupied by –OCH3 and –NHSO2CH3 groupsrespectively.
The DNA-binding activities of the rest sixcompound series (amidine, anthrapyrazole, bis-guanylhydrazone, polyamine, sandramycin, and
DOI 10.1002/jps
QSAR MODELS FOR DNA-BINDING 109
thioacridone) contain 12 equations. The mostimportant parameter for these compound seriesis the hydrophobic parameter, which is one of themost important determinants of DNA-binding. Outof 12 QSAR, 6 contain a correlation between DNA-binding activity and hydrophobicity of the mole-cules/substituents. A negative linear correlation isfound in 4 equations (Eqs. 16, 17, 18, and 27), andthe coefficient range from �0.12 (Eq. 18) to �0.97(Eq. 27). Less hydrophobic congeners in thesecompound families might display enhanced activ-ity. Parabolic correlations with hydrophobic para-meter of the substituents are found in twoequations (Eqs. 25 and 26), which reflect situationswhere activity increases with increasing hydro-phobicity of the substituents up to an optimal valueand then decreases. The optimal values of p
(hydrophobic parameter of the substituents) forEqs. 25 and 26 are 3.56 and 3.68, respectively. Thesecond important parameter is the molar refrac-tivity of the molecules/substituents with positivecoefficients that are present in 6 QSARs (Eqs. 16,17, 20, 21, 24, and 27). This suggests that themolecules having high molar refractivity/polariz-ability may display high activity.
REFERENCES
1. Watson JD, Crick FHC. 1953. The structure ofdeoxyribonucleic acid (DNA). Cold Spring HarbSymp Quant Biol 18:123–131.
2. Hurley LH, Boyd FL. 1988. DNA as a target fordrug action. Trends Pharmacol Sci 9:402–407.
3. Gupta SP. 1994. Quantitative structure-activityrelationship studies on anticancer drugs. ChemRev 94:1507–1551.
4. Martinez R, Chacon-Garcıa L. 2005. The search ofDNA-intercalators as antitumoral drugs: What isworked and what did not work. Curr Med Chem12:127–151.
5. Baguley BC, Denny WA, Atwell GJ, Cain BF.1981. Potential antitumor agents. 35. Quantitativerelationships between antitumor (L1210) potencyand DNA binding for 40-(9-acridinylamino)metha-nesulfon-m-anisidide analogues. J Med Chem 24:520–525.
6. Hansch C, Maloney PP, Fujita T, Muir RM. 1962.Correlation of biological activity of phenoxyaceticacids with Hammett substituent constants and par-tition coefficients. Nature 194:178–180.
7. Hansch C, Leo A. 1995. Exploring QSAR: Funda-mentals and applications in chemistry and biology.Washington, DC: American Chemical Society.
8. Selassie CD, Garg R, Kapur S, Kurup A, VermaRP, Mekapati SB, Hansch C. 2002. Comparative
DOI 10.1002/jps JOUR
QSAR and the radical toxicity of various functionalgroups. Chem Rev 102:2585–2605.
9. Verma RP, Kurup A, Mekapati SB, Hansch C.2005. Chemical-biological interactions in human.Bioorg Med Chem 13:933–948.
10. Hansch C, Leo A, Mekapati SB, Kurup A. 2004.QSAR and ADME. Bioorg Med Chem 12:3391–3400.
11. Selassie CD, Mekapati SB, Verma RP. 2002.QSAR: Then and now. Curr Topics Med Chem 2:1357–1379.
12. C-QSAR Program, BioByte Corp., 201W. 4th st.,Suit 204, Claremont, CA 91711, USA www.biobyte.com.
13. Hansch C, Hoekman D, Leo A, Weininger D,Selassie CD. 2002. Chem-bioinformatics: Compara-tive QSAR at the interface between chemistry andbiology. Chem Rev 102:783–812.
14. Verloop A. 1987. The sterimol approach to drugdesign. New York: Marcel Dekker.
15. Hansch C, Steinmetz WE, Leo AJ, Mekapati SB,Kurup A, Hoekman D. 2003. On the role ofpolarizability in chemical-biological interactions.J Chem Inf Comput Sci 43:120–125.
16. Hansch C, Kurup A. 2003. QSAR of chemicalpolarizability and nerve toxicity. 2. J Chem InfComput Sci 43:1647–1651.
17. Verma RP, Kurup A, Hansch C. 2005. On the roleof polarizability in QSAR. Bioorg Med Chem 13:237–255.
18. Verma RP, Hansch C. 2005. A comparison betweentwo polarizability parameters in chemical-biologi-cal interactions. Bioorg Med Chem 13:2355–2372.
19. Abraham MH, McGowan JC. 1987. The use ofcharacteristic volumes to measure cavity terms inreversed phase liquid chromatography. Chromato-graphia 23:243–246.
20. Cramer RD III, Bunce JD, Patterson DE, FrankIE. 1988. Cross validation, Bootstrapping and par-tial least squares compared with multiple regres-sion in conventional QSAR studies. Quant StructAct Relat 7:18–25.
21. Pogliani L. 1996. Modeling with special descriptorsderived from a medium-sized set of connectivityindices. J Phys Chem 100:18065–18077.
22. Pogliani L. 2000. From molecular connectivityindices to semiempirical connectivity terms: Recenttrends in graph theoretical descriptors. Chem Rev100:3827–3858.
23. Agrawal V, Singh J, Khadikar PV, Supuran CT.2006. QSAR study on topically acting sulfonamidesincorporating GABA moieties: A molecular connec-tivity approach. Bioorg Med Chem Lett 16:2044–2051.
24. Selassie CD, Kapur S, Verma RP, Rosario M.2005. Cellular apoptosis and cytotoxicity of phenoliccompounds: A quantitative structure-activity rela-tionship study. J Med Chem 48:7234–7242.
NAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
110 VERMA AND HANSCH
25. Verma RP, Hansch C. 2006. Cytotoxicity of organiccompounds against ovarian cancer cells: A quanti-tative structure-activity relationship study. MolPharm 3:441–450.
26. Verma RP, Hansch C. 2007. Understandinghuman rhinovirus infections in terms of QSAR.Virology 359:152–161.
27. Atwell GJ, Rewcastle GW, Baguley BC, DennyWA. 1987. Potential antitumor agents. 48.30-Dimethylamino derivatives of amsacrine: Redoxchemistry and in vivo solid tumor activity. J MedChem 30:652–658.
28. Baguley BC, Denny WA, Atwell GJ, Cain BF.1981. Potential antitumor agents. 34. Quantitativerelationships between DNA binding and molecularstructure for 9-anilinoacridines substituted in theanilino ring. J Med Chem 24:170–177.
29. Hansch C, Kurup A, Garg R, Gao H. 2001. Chem-bioinformatics and QSAR: A review of QSAR lack-ing positive hydrophobic terms. ChemRev 101:619–672.
30. Denny WA, Atwell GJ, Baguley BC. 1983. Poten-cial antitumor agents. 38.3-substituted 5-carboxa-mido derivatives of amsacrine. J Med Chem 26:1619–1625.
31. Thakur A, Thakur M, Kakani N, Joshi A,Thakur S, Gupta A. 2004. Application of topo-logical and physiochemical descriptors: QSARstudy of phenylamino-acridine derivatives. ARKI-VOC: 36-43. (The article is available freely atthe following address: http://arkat-usa.org/ARKI-VOC%5CJOURNAL_CONTENT%5Cmanuscripts-%5C2004%5C04-1151JP%20as%20published%20-mainmanuscript.pdf).
32. Robertson IGC, Denny WA, Baguley BC. 1980.Inhibition of T4 bacteriophage yield by 9-anilinoa-cridines; comparison with in vivo antitumor activ-ity. Eur J Cancer 16:1133–1139.
33. Bielawska A, Bielawski K, Wołczynski S, AnchimT. 2003. Structure-activity studies of novel amidineanalogues of chlorambucil: Correlation of cytotoxicactivity with DNA-binding affinity and topoisome-rase II inhibition. Arch Pharm Pharm Med Chem336:293–299.
34. Eifler-Lima VL, Uriac P, Huet J, Jenkins TC,Thurston DE. 1995. Relationship between cyto-
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 97, NO. 1, JANUARY 2008
toxicity and DNA-binding affinity of amidinederivatives of tetrahydroquino[4,3-b]benzazepinesand tetrahydrobenzo[k]naphthyridines. BioorgMed Chem Lett 5:3003–3006.
35. Spartan’06 Program, Wavefunction, Inc., 18401Von Karman Avenue, Suite 370, Irvine, CA92612, USA www.wavefun.com.
36. Clark DE. 1999. Rapid calculation of polarmolecular surface area and its application to theprediction of transport phenomena. 1. Predictionof intestinal absorption. J Pharm Sci 88:807–814.
37. Kelder J, Grootenhuis PDJ, Bayada DM,Delbressine LPC, Ploemen JP. 1999. Polar mole-cular surface as a dominating determinant for oralabsorption and brain penetration of drugs. PharmRes 16:1514–1519.
38. Hartley JA, Reszka K, Zuo ET, Wilson WD,Morgan AR, Lown JW. 1988. Characteristics ofthe interaction of anthrapyrazole anticancer agentswith deoxyribonucleic acids: Structural require-ments for DNA binding, intercalation, and photo-sensitization. Mol Pharmacol 33:265–271.
39. Dave C, Ehrke MJ, Mihich E. 1977. Studies on thestructure-activity relationship among aliphatic andaromatic bisguanylhydrazones and some relatedcompounds. Chem-biol interact 16:57–68.
40. Denny WA, Cain BF. 1979. Potential antitumoragents. 31. quantitative structure-activity relation-ships for the antileukemic bis(guanylhydrazones).J Med Chem 22:1234–1238.
41. Rege K, Ladiwala A, Hu S, Breneman CM,Dordick JS, Cramer SM. 2005. Investigation ofDNA-binding properties of an aminoglycoside-poly-amine library using quantitative structure-activityrelationship (QSAR) models. J Chem Inf Model 45:1854–1863.
42. Boger DL, Chen J-H, Saionz KW, Jin Q. 1998.Synthesis of key sandramycin analogs: Systematicexamination of the intercalation chromophore.Bioorg Med Chem 6:85–102.
43. Dheyongera JP, Geldenhuys WJ, Dekker TG, Vander Schyf CJ. 2005. Synthesis, biological evalua-tion, and molecular modeling of novel thioacridonederivatives related to the anticancer alkaloid acro-nycine. Bioorg Med Chem 13:689–698.
DOI 10.1002/jps