You are on page 1of 5

Quantitative Structure-Retention Relationships of Antimicrobial Hydrazides Evaluated by Reverse-Phase Liquid Chromatography

M. L. C. Montanari 1 / Q. B. Cass 2 / C. A. Montanari 1. 1Nficleo de Estudos em Quimica Medicinal-NEQUIM, Departamento de Qulmica, Universidade Federal de Minas Gerais. Campus da Pampulha, 31270-901, Belo Horizonte, MG, Brazil. 2Departamento de Quimica, Universidade Federal de Sgo Carlos, Rodovia Washington Luis, Sgo Carlos/SP, Brazil

Key Words
Column liquid chromatography Structure-retention relationships Antimicrobial hydrazides

Summary
Quantitative structure-retention relationships of antimicrobial hydrazides were investigated by reversedphase liquid chromatography (RPLC), to disclose retention mechanisms operating in a C-8 column. Capacity factors (log k) were obtained using methanol-water mobile phase in volume fractions: 25-75 % v/v. Extrapolated log kw values were derived from log k versus mobile phase, volume fraction of organic modifier, q), using log k = 10g kw - S~o. The chemical validity of a linear relationship between the slopes (S) and intercepts (log kw) was established. S-log kw correlations are chemically meaningful for a non-homol0gous series of compounds. The existence of a clear separation of hydrazides into two different sets of hydrogen-bonding capabilities demonstrated the role of such relationships. Standard CoMFA fields were used to delineate those specific intermolecular interactions responsible for changes in retention of study hydrazides. Thus, a 3D QSAR (CoMFA) model showed that S exhibits electrostatic behaviour towards a C-8 HPLC column.

describe solute-solvent interactions such as ion-dipole, dipole-dipole, dipole-induced dipole, instantaneous dipole-induced dipole, hydrogen bonding, electron pair donor-electron pair acceptor, etc. Nonetheless, the way of dealing with this is by determining chromatographic retention data for quantitative use. From many approaches that have been taken to study the effects of the mobile phase in RPLC, the most commonly used are binary solutions of water with an organic modifier such as methanol [9-12]. This is so due to the fact that retention in RPLC is primarily controlled by the chromatographic strength of the mobile phase where solute-solvent interactions play a very important role [ i 3, 14]. It has been shown that a quadratic function best describes plots of log k versus mobile-phase, volume fraction of organic modifier, ~o. However, in general terms this sometime overestimates log kw and for a reasonable q~ range (e.g. 20-80% v/v) the logarithmic-linear relationship giving in Eq. (1) is a better choice [15]. log k = log kw- S(p (1)

where kis the solute capacity factor for a specific mobile phase composition ~o, and kw is the extrapolated k value for pure water, (p = 0. The chromatographic retention mechanism is dependent, basically, on two components: the solute size (measured by its volume or surface area) and its hydrogen bonding capability. Although solute size and other physicochemical descriptors are difficult to envisage from Eq. (1), it has been shown that the slope, S, versus log kw can display differences in the capacity for hydrogen bond formation for a set of compounds [16, 17]. First it has to be pointed out that it seems quite clear that S must be correlated with log k~ otherwise the slopes from Eq. (1) would not satisfy the background theory establishing the meaning of log k~ In other words, log kw would have to be obtained directly. Although this is a matter of dispute [ 18], it is still valid where chemically meaningful relationships can be established. The main purpose of this paper is to show that the "S-log kw compensation" phenomenon proposed by Tan and Carr [18], can be applied to a non-homologous series of Original

Introduction
The mechanisms of reversed-phase liquid chromatography (RPLC) are still a matter of intensive search and quantitative structure-retention relationships (QSRRs) have been established towards that goal [ 1-5]. The main purpose of such studies is the possibility of predicting retention times regardless of the type of column used as well as chiral recognition behaviour on the separation of chiral drugs [6-8]. It is, however, rather difficult to 722 0009-5893/00/06 722-05 $ 03.00/0

Chromatographia Vol.51, No. 11/12, June 2000 9 2000 Friedr. Vieweg & Solm VerlagsgesellschaflmbH

~/~/X

X = 1. Br, 2. NO2.3. H, 4. Me, 5. OMe, 6. t-Bu, 7. C5Hll, 8. C6H5


Figure 1

module of SYBYL. The most stable conformations derived from SYBYL systematic searches were regularised and the final structures were subjected to the CoMFA analysis. Regularising the structures via MOPAC version 6, as implemented in SYBYL, using AM1, gave rise to the same CoMFA results.

Antimicrobial hydrazides studied.

Results
(MR.,~*) ,.CF3 Eight antimicrobial hydrazides [17] were synthesized and their structures can be found in Figure 1. The chosen substituents have been selected through a Craig plot analysis [22] from which a non-linear dependence is evident between them, see Figure 2. The two descriptors, the Hammett electronic substituent constant, O-p, and the molar refractivity index, MR, of such substituents were chosen due to their ability to play role in size-polarizability and hydrogen bonding character, thus aiming to indicate any direct influences on the hydrogen-bonding donor capabilities of the hydrazides studied. The correlation coefficient, r 2, between O-pand MR is only 0.069, i.e. they are not correlated at all. Nevertheless, the results found so far indicate no correlation b e t w e e n - S versus ap or MR. The steric parameters for intermolecular interactions L, B1 and B5, and the calculation of the volume of substituents via MG-Vol and SA-Vol [23] were evaluated to see the dependence o f - S on steric problems in terms of substituent constants, but none appeared to be relevant. Table I shows log k, log kw and log Papp for these compounds. The log Pappvalues in Table I were obtained by simple calculation from Eq. (2): log kw = 0.974(+0.18)log Poet + 0.054 (+0.31) (n = 6; r 2 = 0.982; s = 0.123; F = 230.18, r2cu = 0.932)

(MR-,op+)
0.8-

0.6-

:o.~ 0.4 ~
0.2 ~
o.o

coc~

c,o' c.oo~,cqc%
N(C, sHs) Fd IISH

-q,B,
(lUR-,~-)
OC~ 9 OH

.OCF3

-0.2-' -0.4" -0.6-" -2

~CH3 i~t B:) r~ IC(C 1.1,~3 ==CsH11 OEI~NHCOCei-~H(CI'~

NHN/-LII 9 NI_~II NCHC~

(MR+,qp-) 0 MR.lff 1
r i

-0.8"
Figure 2

-~

2D Craig plot for O-pversusMR.

compounds, though congeneric depicted (some similar moieties present in molecules) as to the need of QSAR and QSRR studies [19, 20]. On the other hand, QSRR studies demonstrate that a single physicochemical descriptor is not easily found for such complex retention mechanisms, mainly when they involve chiral recognition problems. The search for uniqueness in parameter descriptors for retention mechanisms seems to be valid when partitioning plays a role [21 ].

(2)

Experimental
The RP-HPLC data were recorded on a Schimadzu instrument equipped with two bombs LC-10AD, UV detector SPD-6AV and LC-R6A, operating at 217 and 254 nm. The stationary phase was a C-8 LiChrosorb (5 #m) column (250 4.0 mm), from Merck (Darmstard, Germany). Sodium nitrate was used for measuring to at 217 urn. The mobile phase was a buffer of 0.1 M amonium acetate, apparent pH (pHapp) 4.6, and methanol as modifying agent. The methanol content of the mobile phase was 25-75 ( % v/v). Molecular modelling calculations were on a Silicon graphics workstation (OCTANE) with SYBYL 6.5 software, using molecular mechanics (Tripos force field), and Gasteiger-Hfickel partial atomic charges. CoMFA calculations were on the Advanced CoMFA Original

where log Poct is the partitioning between octanol and water for ethyl acetate, acetanilide, benzyl alcohol, benzaldehyde, acetophenone and benzophenone; r 2 is the correlation coefficient; s the standard deviation; F the F test and r2cv the cross-validation term for measuring the model's goodness of predictability.

Discussion
The - S values and log kw ((~MeOH = 0) were obtained from Eq. (1). The - S values for compounds 1-8 are: 0.0432, 0.039, 0.0376, 0.042, 0.0411, 0.034, 0.0329 and 0.0371, respectively. Figure 3 shows the result o f - S versus log k~. The dependence o f - S over log kw is clearly seen as twofold: there are two different types of behaviour for compounds 1-5 and 6-8, i.e. the t-Bu, CsH1] and C6H 5 substituents play a special role in the hydrogen bonding capabilities of the hydrazides studied. It is worthwhile pointing out that it seems quite reasonable to assume this behaviour as due to the bulki723

Chromatographia Vol. 51, No. 11/12, June 2000

Table I. log k, log kwand log Papp for hydrazides 1-8.

Compounds
1 2 3 4 5 6 7 8

25
1.742 1.34 1.20 1.54 1.38

35
1.218 0.811 0.67 0.985 0,85

45
0.796 0.473 0.292 0.61 0.48 1.254 1.48 1.23

55

q~MeOH/lOg k 65
-0.051 -0.237 -0.315 0.203 -0.310 0.414 0.749 0.378

75
-0.45 4).688 -0.776 -0.598 -0.714 0.187 0.509 0.140

log kw
2.754 2.250 2.053 2.516 2.344 2.726 2.936 2.871

log Papp
2.772 2.255 2.052 2.528 2.351 2.743 2.959 2.892

0.338 0.073 -0.018 0.165 0.06 0.74 1.110 0.698

ness ofsubstituents in compounds 6-8. Compounds 1-5 have the same hydrogen bonding pattern. The slopes of-Svalues are obtained from a linear energy relationship between log k, the capacity factor for hydrazides studied against mobile-phase, volume-fraction organic modifier, q~0. Log kw values are obtained from such relationships by assuming ~00 = 0. All eight equations are statistically highly significant and have the cross-validated term, r2c~ in the range: 0.808-0.994, that is, the predicting power of the model is well established [24]. The - S values are meaningful as representing the contribution of (P0 to log k~, i.e., increasing the water content of the mobile phase does get closer to log k~, and that is reflected in the slope of the linear solvation relationship of solutes. This is the reason for the type of plot shown in Figure 3. If intermolecular interactions play a different role in the mobile phase, they will be found in the - S values, thus modifying the mechanism of retention due to the differing intermolecular forces taking part in RP-HPLC measurements. That can be easily seen from Eq. (3): log kw(compound 1) ~- 0.548(+0.32)1og k(oMeOH, 75% +2.720(+0.18) (n = 8, r ~ = 0.745, s = 0.13, F = 17.48,

0,044 -

Br
Me

0,0420,0400,0380,036H

OMe

NO 2

oHo
t-Bu

0,0340,032

CsH~
i i i i I i

2,0

2,2

2,4

2,6

2,8

3,0

log
Figure 3

Plot of-S versus log kwfor hydrazides studied.

/cv=0.588)

(3)

The slope in Eq. (3) will approach unity only when the water content is 100%, i. e., (Po = 0. The same trend has been found so far for all other compounds studied. Thus, there is a clear dependency of log kw on (Po, but only for solutes controlled by the same retention mechanism, not by a different one. To check the validity of the above statement we analysed standards used for obtaining log Papp for hydrazides under study. Eq. (4) shows the result: -S = 1.001 (+0.06)log kw + 0.899(+0.11) (n = 6, r 2 = 0.998, s = 0.04, F = 1,960.32, r2cv = 0.991) .

(4)

Equation 4 states that there is a single homo-energetic retention-governing solute property for the standards. However, it can be clearly seen that - S e 0 when log kw assumes this value. Accordingly, -S values do empha724

size the nature of retention that takes place in such RPHPLC measurements. Next, we have attempted the plot of S versus log kave [18], but r2cv for hydrazides 1-5 were in the range 0.666-0.744, thus much poorer in their predictability power than that for normal (p values. However, the range found so far for hydrazides 6-8 was 0.955-0.994, thus better than the corresponding ones indicated above. As a result, the graph found in Figure 3 changes if the average values are on the plot: hydrazides 1-5 will have scattered data points whereas those of hydrazides 6-8 will have a linear relationship, but having the same trend as depicted in Figure 3. We also examined the ability of the comparative molecular field analysis method, CoMFA, to reproduce S in terms of its steric and electrostatic character within the hydrazides studied. Molecular fields calculated with a C.3 probe and Gasteiger-Hfickel partial atomic charges gave a good fit and cross-validated estimates of steric and electrostatic fields [25]. The aim of such a study was to devise any 3D relationship according to the fact that S values may also be connected to solute sizes as well as the solvent-staOriginal

Chromatographia Vol. 51, No. 11/12, June 2000

Table II. Predicted, actual and residual values for S values, from CoMFA model.

Compotmds Br H Me MeO NO2 Pent Ph t-Bu

Actual 4.32 3.76 4.20 4.11 3.90 3.29 3.71 3.40

Calculated 4.329 3.757 4.192 4.108 3.904 3.292 3.706 3.402

Residual -0.009 0.003 0.008 0.002 -0.004 -0.002 0.004 -0.002

Figure 4

CoMFA superimposition rule for hydrazides studied.

It

Figure 5

CoMFA electrostatic field drawn at 80 % contribution.

9 Br 4.2 ~ 4.0 ~
liNC~z

9 Me
mMeO

-soo "~)

3.8 ~

IIH Ph

predicted 3.6-"
3.4" 9 t-Bu

3.2

'=Pent 3.2

3'.6

3'.8

410

4'.2

4'.4

-S(10.2 )actual
Figure 6

S values of test compounds predicted by PLS no validation model (Table II).

tionary phase parameter (electrostatics, including hydrogen bonding) [ 18]. In the CoMFA analysis we first superimposed the hydrazides using the most stable conformation o f a systematic search procedure in S Y B Y L [25] at C ( O ) N H N H C ( O ) moiety, Figure 4. The standard CoMFA characteristics were used: 2 A regular grid spacing in all three dimensions within the defined region; a disOriginal

tance dependent dielectric constant. The models were estimated by the " l e a v e - o n e - o u t " procedure in the S A M P L S [26] procedure. The 3D Q S A R model derived from the training on hydrazides S: crossvalidated: r2cv= 0.706, scv= 0.375, at 5 components; non-crossvalidated: r 2 = 1.00; s = 0.010, F = 1,962.62; steric field = 49.6 % and electrostatic field = 50.4%. Figure 5 represents the CoMFA STDEV*C O E F F contour plot obtained from the PLS no validation model: more bulk near bigger contour map and more positive charge near smaller map. First o f all, it has to be pointed out that the electrostatic field slightly predominates over the steric field in describing S-character. This is new in terms o f CoMFA fields for derived S meaning. Moreover, it also has to be made clear that the solely electrostatic field is represented by a large region o f low electron density (blue graph not shown) where a lower electron density is favoured around the C ( O ) N H N H C ( O ) moiety than that o f position 4 phenyl-ring substitutions; that is, greater positive charge near blue increases electrostatic interactions with solvent-stationary phase. Table II shows the predicted, actual and residual values obtained from the CoMFA analysis, while F i g u r e 6 shows fitting data. From the above 3D Q S A R model it seems reasonable to assume that electrostatics do play an important role over S, thus not especially emphasising solute size. However, the CoMFA field calculation using a C.3 probe ignores the possibility o f hydrogen bonding to the probe atom. Hydrogen bonding m a y be found either for H o r 0.3 I probes [25]. We have tried both o f these. The H +1 did give rise to electrostatic interaction (52.1%) with hydrogen-bond accepting atoms within the C ( O ) N H N H C ( O ) moiety to the extent o f r ~ = 0.732, but the r2cv was very poor, viz. 0.215. Similar behaviour was found when an 0.3 -1 probe was used: r 2 = 0.731, r2c~ = 0.244, both o f them with 3 components. The electrostatic relative contribution was 54.4%. We also examined the importance o f 3 D Q S A R CoMFA fields over log Papp. The indicator field used did resemble that o f a CoMFA standard, but r2c~ is m u c h poorer (0.340). The electrostatic field contribution is 51.9%. 725

Chromatographia Vol. 51, No. 11/12, June 2000

Conclusions
The correlation shown in Figure 3 does establish differentiation between the role substituents play in obtaining log k~ The "slope-intercept c o m p e n s a t i o n " found in Eq. (4) for a congeneric set o f compounds demonstrates similar behaviour in terms o f their capabilities for forming hydrogen bonds. However, this does not hold true for the hydrazides stuided where two different characteristics have been found so far. Besides, there is no evidence that - S or log kw correlate with molecular (or substituent) size. There is also no "isoelution p o i n t " for hydrazides 1-5 and 6-8 (graph not shown) but there is among them, i.e. log k versus ~0MeOHis parallel between the two sets. It seems quite reasonable to assume that there is no reason to examine plots o f S versus log kave instead o r S versus log k~, Both o f them have shown the same trend towards elution o f the hydrazides by the mobile phase studied. Furthermore, the predicting power o f the latter seems to be better than the former. The 3D Q S A R model can be useful for depicting the electrostatic character o f S, as the larger CoMFA field needed to show its relationship towards solvent- C - 8 column stationary phase.

References
[1] M. Hui, C. H. Lochmiiller, Abstr Pap Am Chem, S 217, U114 (1999). [2] J Christon, E. E Healy, Abstr Pap Am Chem. S 217, U455 (1999) [3] R. Kaliszan, M. Markuszewski, P. Haber, A. Nasal, T Cserhati, E. Forgacs, R.M. GadzaIa-Kopeiuch, B. Buszewski, Chem Anal-Warsaw 43, 547 (1998) [4] M. L. C. Montanari, D. P. Veloso, Q. B. Cass, C. A. Montanari, J. Liq. Chromat. & Relat. Technol. 20, 1703 (1997) [5] M. M. Britto, C. A. Montanari, C. L. Donnici, Q. B. Cass, J. Liq. Chromat. & Relat. Technol. 22,357 (1999)

[6] Q. B. Cass, A. L. Bassi, A. L. Marlin, Chirality 11, 46 (1999) [7] Y.H. Zhang, Z H. Yun, Chinese J. Anal. Chem. 27,309 (1999) [8] E. Tesarova, Z Bosakova, V. Pecakova, J. Chromatogr. A 838, 121 (1999) [9] J W. Li, P. W. Carr, Anal. Chem. 69, 2550 (1997) [10] J G. Dorsey, W. T. Cooper, Anal. Chem. 66, A857 (1994) [11] J. G. Dorsey, M. G. Khaledi, J. Chromatogr. A 656, 485 (1993) [12] D. J. Minick, J. H. Frenz, M. A. Patrick, D. A. Brent, J. Med. Chem. 31,1923 (1988) [13] A. Vailaya, Horvath, C. J., Phys. Chem. B 101, 5875 (1997) [14] A. Vailaya, C. Horvath, J. Chromatogr. A 829, 1 (1998) [15] J. J Michels, J. G. Dorsey, J. Chromatog~ 457, 85 (1988) [16] N El Tayar, H. van de Waterbeemd, B. Testa, Quant. Struct.Act. Relat. 4, 69 (1985) [17] H. Van de Wateerbemd, M. Kansy, B. Wagner, H. Fischer, Lipophilicity in Drug Action and Toxicology. in "Methods and Principles in Medicinal Chemistry"; R. Mannhold, H. Kubinyi, H. Timmerman, Eds., Vol. 4, Weiheim, 1996, p. 73 [18] L. C. Tan, P. W. Carr, J. Chromatogr. A, 656, 521 (1993) [19] M.L.C. Montanari, Ph.D Thesis, Federal University of Minas Gerais, 1998, Brazil [20] A. Vaitaya, C. Horvath, J. Phys. Chem. B 102, 701 (1998) [21] A. Vailaya, C. Horvath, Ind. Engin. Chem. Res. 35, 2964 (1996) [22] H. Kubinyi, QSAR: HanschAnalysis andRelatedApproaches, Vol. 1 in "Methods and Principles in Medicinal Chemistry", R. Mannhold, R Krogsgaard-Larsen, H. Timmerman, Eds., VCH, Weinheim, 1993, p. 109 [23] C. Hansch, A. Leo, in "Exploring QSAR: Fundamentals and Applications in Chemistry and Biology". ACS, Washington, 1995, p. 76 [24] (a) S. Wold, Quant. Struct.-Act. Relat. 10, 191 (1991) (b) H. Kubinyi, QSAR: Hansch Analysis and Related Approaches, Vol. 1 in "Methods and Principles in Medicinal Chemistry", R. Mannhold, P. Krogsgaard-Larsen, H. Timmerman, Eds., VCH, Weinheim, 1993, p. 102 [25] Sybyl, Version 6.5, Tripos, Inc., 1998 [26] E Lindgren, S. Riinnar, Alternative Partial Least-Squares (PLS) Algorithms, in "3D QSAR in Drug Design % Vol. 3. H. Kubinyi, G. Folkers, Y. C. Martin, Eds., Kluwer/Escom, London, 1998, p. 105 Received: Aug 9, 1999 Revised manuscript received: Nov 4, 1999 Accepted: Nov 17, 1999

726

Chromatographia Vol. 51, No. 11/12, June 2000

Original

You might also like