Acb

Aa

b

c

a

ARRAA

KS(ALFQ(QrQr

1

pegaprbb

h0

Journal of Pharmaceutical and Biomedical Analysis 159 (2018) 363–373

Contents lists available at ScienceDirect

Journal of Pharmaceutical and Biomedical Analysis

j o ur na l ho mepage: www.elsev ier .com/ locate / jpba

pplication of salting-out thin layer chromatography inomputational prediction of minimum inhibitory concentration andlood-brain barrier penetration of some selected fluoroquinolones

zza H. Rageha,∗, Noha N. Atiaa, Hamdy M. Abdel-Rahmanb,c

Department of Pharmaceutical Analytical Chemistry, Faculty of Pharmacy, Assiut University, Assiut, 71526, EgyptDepartment of Medicinal Chemistry, Faculty of Pharmacy, Assiut University, Assiut, 71526, EgyptDepartment of Pharmaceutical Chemistry, Faculty of Pharmacy, Nahda University, 62511, Beni Suef, Egypt

r t i c l e i n f o

rticle history:eceived 29 May 2018eceived in revised form 7 July 2018ccepted 9 July 2018vailable online 11 July 2018

eywords:alting-out thin layer chromatographySOTLC)mmonium sulphateipophilicity parametersluoroquinolones antibacterialsuantitative structure-activity relationship

QSAR)uantitative structure-property

elationship (QSPR)uantitative structure-retention

a b s t r a c t

The 2017 FDA safety review regarding the CNS (central nervous system) side effects associated withthe systemic use of fluoroquinolones antibacterials (FQs) was the key motivation to carry out this work.The main objective of this study is to investigate lipophilicity and retention parameters of some selectedfluoroquinolones antibacterials (FQs) namely; levofloxacin (LEV), ofloxacin (OFL), gatifloxacin (GAT), nor-floxacin (NOR), sparfloxacin (SPA), ciprofloxacin (CIP) and lomefloxacin (LOM) using salting-out thinlayer chromatography (SOTLC). Statistically significant correlations between the chromatographically-obtained retention parameters and experimental log P values were found and expressed as quantitativestructure retention relationship (QSRR) equations. Principal component analysis was carried out toexplain the variation between chromatographic and both experimental and computed lipophilicityparameters. In another aspect of this study, a comparison between the chromatographically-determinedretention parameters (for five of the drugs under study) obtained using SOTLC (current study) and rel-ative lipophilicity (RM0) determined using a previously reported RP (reversed-phase)-TLC method wascarried out. Statistically significant correlation between the two methods was found, although RM0 valuesobtained using SOTLC was lower than those reported using RP-TLC. Multiple linear regression analysiswas performed to predict MIC (minimum inhibitory concentration) and blood brain barrier (BBB) pene-

elationship (QSRR) tration of the examined drugs in which efficient QSAR (quantitative structure-activity relationship) andQSPR (quantitative structure-property relationship) models were generated using the calculated chro-matographic parameters (RM0 and C0). The described models can provide a useful approach to predict MICand BBB penetration of newly synthesized FQs targeting to increase their activity against Gram-positiveorganisms and to minimize the associated CNS side effects.

© 2018 Elsevier B.V. All rights reserved.

. Introduction

Lipophilicity is one of the most influential physicochemicalarameters that affect the biological activity of a drug. This prop-rty is connected to many steps of drug action because lipophilicityoverns solubility, reactivity and degradation of drugs, as wells formulation of pharmaceuticals, before the drug reaches itsharmacological target. The drug’s lipophilicity plays a crucial

ole in determining its passive transport through biological mem-ranes, including gastrointestinal absorption and transport acrosslood–brain barrier to achieve the desired biological activity [1,2].

∗ Corresponding author.E-mail addresses: azhesham@yahoo.com, azzarageh@aun.edu.eg (A.H. Rageh).

ttps://doi.org/10.1016/j.jpba.2018.07.010731-7085/© 2018 Elsevier B.V. All rights reserved.

Its impact is not only restricted to pharmacodynamic profile of thedrug, however it has the ultimate influence on ADMET (absorption,distribution, metabolism, excretion, toxicity) properties. Therefore,the partition in biphasic solvent systems is sometimes a bettermodel for simulating the in vivo process [2].

Thin layer chromatography applications in drug discovery pro-cess was the subject matter of very recent and interesting reviewarticles which extensively describe its usefulness in prediction ofphysicochemical properties and biological activity of compoundsbesides its utility in quantitative structure-activity relationshipstudies [3–5]. Generally, reversed phase (RP) chromatography

mode is used to simulate the octanol/water partitioning. Chro-matographic techniques comprising both high performance liquidchromatography (HPLC) and reversed-phase thin layer chro-matography (RP-TLC) are two major indirect approaches for

3 al and

di[RnKddfoRslot

rscocbgwucimwmmiw

VcaSbas[ral

unbabt(hiicmtmapb

Ft

64 A.H. Rageh et al. / Journal of Pharmaceutic

etermination of lipophilicity. RP-TLC for assessment of lipophilic-ty provides several advantages over classical shake flask method6]. The conventional reversed-phase approach employs RP18 orP8 plates, together with a water-modifier mobile phase [7]. Alter-ative approaches have been reported in the literature. In 2017,omsta working group has presented a review article, whichescribes the use of unconventional TLC systems in lipophilicityetermination and the authors have listed four unusual approachesor the determination of lipophilicity which are based on (1) the usef medium-polar stationary phases: CN, NH2, and DIOL instead ofP plates, together with water-based mobile phase; (2) the use ofilica gel in a typical normal-phase manner and treating extrapo-ated retention indices as the “reversed lipophilicity”; (3) the usef oil impregnated silica gel in the reversed-phase manner; and (4)he use of salting-out mobile phases [7].

Salting-out thin layer chromatography (SOTLC) is a typicaleversed-phase technique that is reliant on the use of highly polarorbent such as silica gel, cellulose, etc. . . together with highlyoncentrated aqueous solutions of inorganic salts (with or with-ut miscible organic solvents) as mobile phases [6,8]. Under theseonditions, non-specific hydrophobic interaction between the sor-ent and non-polar parts of the analyte is the key factor, whichoverns the chromatographic retention mechanism. Bij and co-orkers [9] reported the fundamental basis for this mechanismnder the broader name “solvophobic interactions”. Due to the highoncentration of the employed inorganic salt (usually > 1 mol L−1),on-exchange interaction as a contributor to the whole retention

echanism can be neglected [8]. Owing to its high solubility inater and significant salting-out effect, ammonium sulphate is theost widely used mobile phase in SOTLC. Compared to other chro-atographic methods, SOTLC is attractive for “green chemistry” as

t is based on using less-toxic organic solvent-free mobile phaseith low ecological concern [10].

As mirrored by the series of work published by Janjic, Tesic,uckovic and co-workers [11–20], SOTLC was used to examine thehromatographic behaviour of many polar inorganic and organicnalytes using different polar stationary phases [21]. Moreover,OTLC was used to study the lipophilicity and chromatographicehaviour of a variety of drugs that belong to different classes suchs macrolide antibiotics [21,22], oral hypoglycemic drugs [23,24],ulfonamides [8], myorelaxant drugs [25] and ACE inhibitors26–28]. Beside his extensive review about salting out chromatog-aphy [29], Komsta and his coworkers have presented multivariatenalysis of 35 model compounds and its retention on silica or cel-ulose by the use of 12 inorganic salts [30,31].

Fluoroquinolones (FQs) are a popular class of antibiotics forse in a variety of infections including those caused by Gram-egative (especially Pseudomonas aeruginosa) and Gram-positiveacteria, Mycoplasma species, and Chlamydia species, and theyre widely used clinically. They exhibit their bactericidal activityy selective inhibition of bacterial DNA synthesis, more defini-ively by the inhibition of the two bacterial enzymes: DNA gyrasetopoisomerase II) and topoisomerase IV enzymes. Research workas been mainly dedicated to develop newer analogs of FQs with

mproved potency against Gram-positive organisms while retain-ng the favorable activity against Gram-negative organisms. Thelassification of FQs into first, second, third and fourth generation isainly based on their enhanced activity against Gram-positive bac-

eria and anaerobes and improved oral bioavailability. One of theost important physicochemical properties of FQs is lipophilicity

s there is a significant relationship between lipophilicity and theirharmacological activity against Gram-positive bacteria (reflected

y their minimum inhibitory concentration (MIC) values).

In 2012, G. Völgyi et al. [32] have stated that as most of theQs are primarily charged through the physiological pH region,heir lipophilicity tends to be substantially lower than many neutral

Biomedical Analysis 159 (2018) 363–373

and/or basic oral drugs and consequently a precise knowledge inthe protonation microequilibrium and the true lipophilicity couldbe useful to explain the antibacterial activity of the FQs. They havemeasured n-octanol/water partition coefficients at isoelectric pHs(or distribution constant) (log DiepH) by using the traditional shake-flask technique. The same technique was used later on to calculatethe true partition coefficient of some FQs at pH 7 and 7.4 [33], how-ever shake-flask technique is tedious, time-consuming and requiresrelatively large amounts of pure compounds, and it is unsuitablefor substances with high lipophilicity [6]. Furthermore, RP-TLC wasreported for estimation of lipophilicity of fifteen FQs [34], howeverin our opinion, the achievable RM0 (lipophilicity parameter) are rel-atively high if they are connected to the high polarity of the studiedcompounds. Therefore, in the present study, SOLTC is thought to bean alternative substitute for lipophilicity determination.

In 2017, T̈he U.S. Food and Drug Administration (FDA) haveadvised that the serious side effects associated with FQ antibac-terial drugs generally outweigh the benefits for patients with acutesinusitis, acute bronchitis, and uncomplicated urinary tract infec-tions who have other treatment options. For patients with theseconditions, FQs should be reserved for those who do not have alter-native treatment options. An FDA safety review has shown that FQswhen used systemically (i.e. tablets, capsules, and injectable) areassociated with disabling and potentially permanent serious sideeffects that can occur together. These side effects can involve thetendons, muscles, joints, nerves, and central nervous system′′ [35].Moreover, in July 2018, FDA adds stronger safety waninng to FQsabout dangerous drops in blood sugar and neurological side effectsthat can include delirium and memory problems. Therefore, it isvery demandable to afford a quantitative model to better describethe side effects associated with these compounds, especially CNSside effects. Based on the previous discussion, the chief aim of thisstudy is to: (1) investigate the chromatographic behaviour and esti-mate lipophilicity of seven FQs, which covers three different majorclasses of FQs given in Table 1 using SOTLC. (2) Evaluate the applica-bility of chromatographic retention parameters expressed as RM0or C0 determined by SOTLC in prediction of their physicochem-ical properties via providing a simple practical tool to estimatelipophilicity parameters (log P and log D). (3) Find the corre-lations between the SOTLC retention parameters and calculatedphysicochemical parameters obtained with the use of differentcomputational methods or experimental values (log Pexp). (4) Studythe contribution of both RM0 and C0 in the analysis of quantitativestructure-property relationship (QSPR) and quantitative structure-activity relationship (QSAR). (5) Construct quantitative models topredict their MIC and blood brain barrier (BBB) penetration basedon their chromatographically-determined retention parameters,log P, log BB and log D values.

2. Experimental

2.1. Materials and reagents

Seven fluoroquinolones standards namely; levofloxacin HCl;LEV (Al-Pharonia Pharmaceutical Co., Alexandria, Egypt), ofloxacin;OFL (Hoechst AG, Frankfurt, Germany), gatifloxacin; GAT ((Bristol-Myers Squibb Pharmaceutical Co., Cairo, Egypt), norfloxacin; NOR(Hoechst AG, Frankfurt, Germany), sparfloxacin; SPA (Global NapiPharmaceuticals, Egypt), ciprofloxacin; CIP ((Egyptian Interna-tional Pharmaceutical Industries Co., E.I.P.I.CO.) and lomefloxacinHCl; LOM (Alkan Pharma Co. 6 October City, Egypt) were obtained as

gifts and used as supplied. Their chemical structures, chemical for-mula, pKa values, generation and molecular weight are presentedin Table 1. Ammonium sulphate was from Sigma-Aldrich, Stein-heim, Germany; concentrations ranged from 0.50 to 2.45 mol L−1.

A.H. Rageh et al. / Journal of Pharmaceutical and Biomedical Analysis 159 (2018) 363–373 365

Table 1Chemical structures, Chemical formulas, pKa values, generation and molecular weight of the investigated FQs.

Name Chemical structure based on parent drug Chemical formula pKa1 pKa2 Generation Molecular weight based onparent drug

Levofloxacin (LEV) C18H20FN3O4 5.45 6.20 3rd 361.4

Ofloxacin (OFL) C18H20FN3O4 5.45 6.20 2nd 361.4

Gatifloxacin (GAT) C19H22FN3O4 5.69 8.73 4th 375.4

Norfloxacin (NOR) C16H18FN3O3 5.77 8.68 2nd 319.3

Sparfloxacin (SPA) C19H22F2N4O3 5.75 8.79 3rd 392.4

Ciprofloxacin (CIP) C17H18FN3O3 5.76 8.68 2nd 331.3

F2N3O

a

Ms(6Dt

2

Spaaaoupaa

t

Lomefloxacin (LOM) C17H19

Calculated using Chemaxon; source: www.drugbank.ca.

ethanol was employed to prepare the standard solutions of thetudied compounds. TLC was carried out on TLC aluminum sheets20 × 20 cm, 0.20 mm layer thickness) pre-coated with silica gel G0 254, which were purchased from Merck, Darmstadt, Germany.ouble-distilled water was used throughout this work (its conduc-

ivity was measured to be 6.1 �S cm-1).

.2. Chromatographic conditions and parameters

Before use, the TLC plates were divided into 20 × 5 cm pieces.tandard solutions of the investigated FQs (0.2 mg mL−1) were pre-ared in methanol and 2 �L (as bands of 4 mm width, 5 mm apartnd 10 mm from the lower edge and sides of the plate) werepplied in triplicate by a means of Linomat V sample automaticpplicator from Camag (Muttenz, Switzerland) under a streamf nitrogen. Ammonium sulphate of different concentrations wassed as mobile phases and the concentration range for each com-ound is listed in Table 2. The lower and upper concentrations of

mmonium sulphate depend on the linearity between RM valuesnd ammonium sulphate concentration (mol L-1).

The TLC plates were developed in a 27.0 × 7.0 x 26.5 cm conven-ional TLC tank at room temperature (25 ± 5 ◦C). The developing

3 5.64 8.7 2nd 351.3

distance was 5.5 cm in all cases. After development, the plateswere dried in air for 5 min and FQs spots were localized using aUV lamp (short wavelength 254/365 nm, Vilber Louranate 220 V50 Hz, Marne-la-Vallee Cedex, France). All analyses were carriedout in triplicate and the mean Rf (retardation factor) values werecalculated as follows:

Rf = a

b(1)

where, a is the distance migrated by the spot divided by b, which isthe distance migrated by the solvent front. Fig. S1 (Supplementarydata) illustrates two TLC plates located under UV lamp at 254 nmand 365 nm using two representative concentrations of ammoniumsulphate (a) 0.50 mol L−1 and (b) 2.00 mol L−1 and three replicatesfrom each of the studied FQs.

Retention data were expressed as hRf values and they werederived from Rf data (hRf = 100 Rf) and their values are listed inTable S1 (Supplementary data). The parameter; RM is given by the

following formula according to Bate-Smith and Westall [36]:

RM = log ((1Rf

) − 1) (2)

366 A.H. Rageh et al. / Journal of Pharmaceutical and Biomedical Analysis 159 (2018) 363–373

Table 2Relationship between RM values and C (concentration of ammonium sulphate) in the mobile phase of the equation RM = a(RM0)+bC.

Name a (RM0) Sa b Sb r Syx F C0 Range (M)

LEV 1.0424 0.0321 0.3289 0.0238 0.9871 0.0315 190.6 3.17 0.50-2.00OFL 0.9038 0.1145 0.4599 0.0857 0.9230 0.1134 28.8 1.96 0.50-2.00GAT 0.7225 0.0569 0.5170 0.0382 0.9840 0.0619 183.1 1.40 0.50-2.25NOR 0.1111 0.1199 0.5971 0.0627 0.9786 0.0638 90.6 0.19 1.25-2.45SPA 0.9512 0.0389 0.3177 0.0324 0.9799 0.0338 96.5 2.99 0.50-1.75CIP 0.2253 0.1589 0.5829 0.0831 0.9617 0.0844 49.2 0.48 1.25-2.45LOM 0.3978 0.1358 0.5101 0.0710 0.9634 0.0722 51.6 0.78 1.25-2.45

a , Sb standard deviation of slope, Syx standard error of estimate (SEE), F value of test F-S

It

R

wrc(ssaetmRi

C

ai[

2

saCUcAp

2(orfX(BP(

tab2

intercept; relative lipophilicity (RM0), b slope, Sa standard deviation of interceptnedecora, C0 lipophilicity parameter, r correlation coefficient, p value < 0.05.

n SOTLC, RM correlates linearly with the molar concentration ofhe salt in the mobile phase according to Eq. 3 [29]:

M = a(RM0) + bC (3)

here, a is the intercept or RM0 (the lipophilicity parameter), whichepresents extrapolated RM value to 100% water and C is the molaroncentration of the salt in the mobile phase. On the other side, bslope of the regression line) is related to the specific hydrophobicurface area of the solute and it is linearly correlated with RM0 for aeries of structurally related compounds. It has also been suggesteds an alternative measure of lipophilicity. The lipophilicity param-ter, C0 is defined as the salt concentration at which the RM is equalo 0, and can be also used as measure of lipophiliciity, analogous to

odifier concentration in classical lipophilicity estimation [29]. InP-TLC, C0 was introduced by Bieganowska et al. [37], and in SOTLC

t is simply computed as given in Eq. 4 [29]:

0 = RM0

b(4)

This parameter is believed to be more reliable in QSAR analysiss it comprises both chromatographically-measured lipophilicityndex (RM0) and the specific hydrophobic surface area of the solute6].

.3. Physicochemical calculations

All the computational studies were carried out on Dell preci-ion T3600 workstation with Intel Xeon1 CPU-1650.0 @3.20 GHznd Windows 7 operating system using the following software,S ChemDraw Ultra, (Cambridge Soft Corporation, Cambridge, MA,SA) and molecular operating environment (MOE 2014, Chemi-al Computing Group, Canada) as the computational software. Thedvanced Chemistry Development (ACD/Labs) calculations wereerformed online at (https://ilab.acdlabs.com/iLab2/).

Computational lipophilicity (clog P) was calculated by MOE014, ACD/Labs online service, molinspiration online servicemiLOGP, performed online at (http://www.molinspiration.com/)r obtained from www.drugbank.ca and http://www.hmdb.ca/. Theemaining clog P values were obtained from the reported valuesor the following algorithms ALOGPS, Chemaxon, ALOGP, XLOGP2,LOGP3, ACLOGP [33,34,38]. log D pH 7.4 (log D7.4), log D pH 2

log D2.0) and logarithm of the brain/blood concentration ratio (logB) were performed on the ACD/Labs online service. Values of logtrue (true partition coefficient of the neutral species) and log DiepH

distribution constant at isoelectric pHs) were taken from [32,33].The calculations of the remaining molecular descriptors namely;

otal hydrophobic surface area (ASA H), molar refractivity (MR),queous solubility (Log S), van der Waals volume (vol), hydropho-ic volume at -0.2 kcal/mol (vsurf D1) were carried out using MOE014 software.

Fig. 1. The plots of log D against pH predicted by ACD/Labs online service.

2.4. Data analysis

Linear multivariate regression analysis and principal compo-nent analysis were performed using Minitab 17 Statistical Software.The parameters Rf, RM and RM0 and the correlations between RM0,b or C0 and the descriptors were performed using Microsoft Excel2007 for Windows (Microsoft Office Excel 2010, Warsaw, Poland),while the QSPRs, QSRRs and QSARs models were developed usingMolecular Operating Environment (MOE 2014) software.

3. Results and discussion

3.1. Retention mechanism and lipophilicity study of theinvestigated drugs

Seven FQs were selected for the current study, which cov-ers three different generations with different antibacterial activityagainst various types of bacterial infections. The selected com-pounds share the same basic structural features, which are aquinolone nucleus substituted with carboxylic acid functionalgroup at position 3 and a basic substituted or un-substitutedpiperazine group at position 7. Therefore, these compounds arezwitterionic and exist mainly in neutral/zwitterionic (GAT, NOR,SPA, CIP and LOM) or anionic (OFL and LEV) forms at physiologicalpH. These compounds are hydrophilic as reflected by Fig. 1, whichexpresses their computed log D values (octanol–water distributioncoefficient in its logarithmic form) in the pH range from 2 to 12calculated using ACD/Lab software.

Here we have started our investigations with high ammonium−1

sulphate concentrations (≥ 0.5 mol L ) to exclude the possible

interactions of the analyte with the sorbent due to ion-exchangemechanism. In this case nonspecific hydrophobic adsorption canbe considered as the key factor determining retention mechanism.

A.H. Rageh et al. / Journal of Pharmaceutical and Biomedical Analysis 159 (2018) 363–373 367

Tab

le

3C

orre

lati

on

betw

een

RM

0or

C0

and

com

pu

ted

log

P

valu

es

(cal

cula

ted

usi

ng

dif

fere

nt

tech

niq

ues

)

or

exp

erim

enta

l log

P

valu

es.

Dru

g

RM

0C

0M

OEa

AC

D/L

absb

miL

ogPc

ALO

GPS

dC

hem

axon

dA

PLO

GPb

ALO

GPd

XLO

GP2

dX

LOG

P3d

AC

LOG

Pdlo

g

P HM

DB

exce

pt

LEV

from

dru

gba

nke

log

D7.

4b

log

D2.

0b

log

P tru

ef

log

Die

pH

f

LEV

1.04

2

3.17

0.68

0.84

−0.2

6−0

.02

0.65

0

1.64

1.80

−0.3

90.

26

2.10

−2.0

3−3

.10.

74g

n.d

.O

FL

0.90

4

1.97

0.68

0.84

−0.2

6−0

.02

0.65

0

1.64

1.80

−0.3

90.

26

2.10

−2.1

8−3

.10.

59

−0.4

4G

AT

0.72

3

1.40

1.41

1.21

−0.0

4−0

.23

−0.5

80.

01

1.77

n.d

.

−0.7

00.

31

2.60

−2.3

3−3

.09

0.54

−0.7

1N

OR

0.11

1

0.19

0.73

0.82

−0.6

9−0

.47

−0.9

2−0

.92

1.27

1.82

−1.0

3−0

.06

2.10

−3.0

9−4

.02

−0.4

3−1

.07

SPA

0.95

1

2.99

1.36

1.20

1.63

−0.0

7−0

.04

−0.2

31.

62

2.21

0.11

0.04

2.50

−2.4

1−3

.33

0.39

−0.0

9C

IP

0.22

5

0.48

1.04

0.65

−0.7

0−0

.57

−0.8

1−0

.71

1.41

1.94

−1.0

80.

13

2.30

−3.0

1−3

.81

−0.1

3−1

.07

LOM

0.39

8

0.78

1.30

1.71

0.08

0

−0.3

9−0

.67

1.85

2.44

−0.8

00.

28

2.80

−2.9

0−3

.77

0.19

g−1

.13

rhbe

twee

n

RM

0an

d

log

P−0

.013

60.

0568

0.54

06

0.76

70

0.86

27

0.92

66

0.53

27

−0.0

993

0.87

52

0.45

16

−0.0

858

0.96

36

0.92

66

0.93

68

0.90

27rh

betw

een

C0

and

log

P−0

.043

3−0

.000

50.

6128

0.69

48

0.81

68

0.79

05

0.38

15

−0.0

768

0.90

87

0.24

37

−0.1

407

0.87

14

0.79

05

0.82

03

0.92

29

n.d

.:

no

dat

a.a

Cal

cula

ted

usi

ng

MO

E

2014

soft

war

e.b

Cal

cula

ted

usi

ng

AC

D/L

abs

onli

ne

serv

ice

( htt

ps:

//il

ab.a

cdla

bs.c

om/i

Lab2

/ ).

cC

alcu

late

d

usi

ng

mol

insp

irat

ion

onli

ne

serv

ice

( htt

p:/

/ww

w.m

olin

spir

atio

n.c

om/ )

.d

Val

ues

take

n

from

refs

. [33

,34,

38].

eV

alu

es

obta

ined

from

ww

w.d

rugb

ank.

ca

and

htt

p:/

/ww

w.h

md

b.ca

/.f

Val

ues

take

n

from

[ 32 ]

.g

Val

ue

for

LEV

and

LOM

take

n

from

[33]

and

it

is

call

ed

intr

insi

c

par

titi

on

coef

fici

ent

(log

P i)

and

calc

ula

ted

for

the

neu

tral

spec

ies.

hC

orre

lati

on

coef

fici

ent.

Fig. 2. RM versus concentration of ammonium sulphate for seven of the investigatedFQs.

Different theories have been proposed for analyte retention usingSOTLC [29], however a special focus will be given to the hydrationtheory, which suggests that analyte ions possess hydration shelland when the salt is added, its ions compete with the analytes forthe solvent molecules in the hydration shell. In most cases the com-petition is won by an ion leaving the dehydrated form of the analyte.Consequently, this increases the hydrophobic interaction of ana-lyte with the sorbent by increasing salt concentration. Table 1Sillustrates that hRf values decrease with increasing ammonium sul-phate concentration, which coincide very well with “Salting-out”retention principle.

In the current work, SOTLC is employed to estimate thelipophilicity and retention parameters (expressed as either: RM0,b or C0) of the studied FQs. In practice, there is a linear dependencyof the parameter RM on the salt concentration as expressed by Eq.(3). The concentration range for ammonium sulphate used for eachdrug is provided in Table 2. Above the upper value of this range, thedrug remains on the start line. The chromatographic data obtainedfor the seven investigated compounds together with the resultsof linear regression analysis are listed in Table 2 and representedin Fig. 2. For all examined drugs, high values of correlation coef-ficients were achieved (r > 0.9230), with small values of standarderror of estimate, which proves the high significance of Eq. (3) fordetermination of lipophilicity.

The slopes of the regression lines relates to specific hydropho-bic area of the analyte. By careful inspection of the values of theslopes, it is clear that there is some variation in the slopes beingslightly higher for the more polar FQs (b = 0.5971, 0.5829 and 0.5101for NOR, CIP and LOM; respectively) than those obtained for theless polar ones (b = 0.3289, 0.4599, 0.5170, 0.3170 for LEV, OFL,GAT and SPA; respectively). These variations can be attributed tothe differences in their ionization states, molecular size or polar-ity. Nevertheless, a good correlation exists between RM0 and slopewith r ≈ -0.89 which suggests a similar chromatographic retentionmechanism for this homologous series of studied compounds. Thisrelationship can be represented by the following equation:

RM0 = − 2.984 (±0.666) b + 2.035 (±0.323) (5)

(r = 0.8948, R2 = 0.8008, F = 20.10, p value < 0.05, Syx = 0.1834, n = 7),R2 the determination coefficient, F is the value of test F- Snedecora,

Syx standard error of estimate.

Besides, it was observed that for the more polar analytes(NOR, CIP and LOM) the slope values (of the relationship of RMversus concn. ammonium sulphate) obtained in the range of 0.5

3 al and

tiinpcl≈R

(

3c

odltCocMbutcAouoswu

pw1tafatveosflditaou(pmcrts

pfi

68 A.H. Rageh et al. / Journal of Pharmaceutic

o 1.25 mmol L−1 ammonium sulphate differ from those obtainedn the range of 1.25 to 2.45 mmol L−1 ammonium sulphate. Thismplies that higher concentrations of ammonium sulphate areeeded to disrupt the hydration shell surrounding these com-ounds to initiate salting out effect (Table 1S). In addition, a goodorrelation between RM0 and the lipophilic parameter C0 (calcu-ated using Eq. 4; values are given in Table 2) was obtained with r

0.96 (Eq. 6).

M0 = 0.297 (±0.041) C0 + 0.160 (±0.079) (6)

r = 0.9552, R2 = 0.9123, F = 52.16, p value < 0.05, Syx = 0.1214, n = 7)

.2. Correlation between RM0 and C0 and experimental andalculated log P values

The next step in this study is to compare chromatographically-btained lipophilic parameters; RM0 and C0 with log P valuesetermined experimentally (using shake-flask technique) or calcu-

ated using different computational methods (that cover differentheoretical approaches) such as MOE, ACD/Labs, miLOGP, ALOGPS,hemaxon, APLOGP, ALOGP, XLOGP2, XLOGP3, ACLOGP or log Pbtained from HMDB (human metabolome data base). The prin-iple of these approaches were discussed briefly in the report ofornar et al. [39]. As given in Table 3, a very good correlation exists

etween either RM0 or C0 and experimentally-determined log P val-es (log Ptrue and logDiepH), which proofs the reliability of SOTLCechnique as a tool for lipophilicity estimation. In addition, a goodorrelation was observed between log P values calculated usingLOGPS, Chemaxon, APLOGP and XLOGP3 and RM0 or C0. On thether side, a poor correlation was found for log P values calculatedsing MOE, ACD/Lab, miLOGP, ALOGP, XLOGP2, ACLOGP or log Pbtained from HMDB and RM0 or C0. However, when the ionizationtate of FQs was taken into consideration, a very good correlationas obtained with log D values calculated at pH 2.0 and pH 7.4sing ACD/Labs.

For further examination of the differences among the com-utational and experimental lipophilicity, the lipophilicity valuesere arranged in a matrix of dimensions seven (compounds) x

7 (two chromatographic lipophilicity (RM0 and C0), 13 compu-ational lipophilicity and two experimental lipophilicity (log Ptrue

nd logDiepH)), then principal component analysis PCA was per-ormed on the whole matrix. The first principal component PC1ccounts for 54.8% of the data variation, whereas the second andhe third principal components explain 27.1% and 11.8% of the dataariation, respectively (93.8% total). As can be seen in Fig. 3, PC1xplains the differences between the mean (overall lipophilicty)f the studied drugs, which has its largest value for LEV and itsmallest value for NOR. On the other hand, PC2 explains the dif-erences between the computational methods. The computationalog P values have larger variation (they described the majority ofataset variance) than the experimental ones and their mean value

s modelled by the first PC. The Chemaxon and XLOP2 methods havehe greatest impact on this PC. The third PC describes the variationmong the experimental and chromatographic lipophilicity. More-ver, it is obvious from Fig. 3 that log Ptrue and logDiepH (obtainedsing shake-flask technique) [32,33] are highly correlated withi) chromatographically-obtained lipophilicities (RM0 and C0), (ii)redicted values of log D2.0 and log D7.4 and (iii) those deter-ined using Chemaxon and APLOGP, while with weak or nearly no

orrelation with other computational techniques, which intenselyationalize the use of chromatographic techniques and not compu-ational ones for lipophilicity estimation as a possible alternative

ubstitute to experimental methods.

The lipophilicity matrix was further reduced to seven (com-ounds) x nine (two chromatographic lipophilicity (RM0 and C0),ve computational lipophilicity and two experimental lipophilic-

Biomedical Analysis 159 (2018) 363–373

ity (log Ptrue and logDiepH)), then principal component analysisPCA was again performed on the whole matrix. The first principalcomponent PC1 explains 86.5% of the data variation, whereas PC2accounts for 8.2% of the data variation, respectively (94.6% total). Asgiven in Fig. 4, the variation in the lipophilicity of the investigatedFQs can be mainly explained by PC1, while PC2 explains differ-ences among chromatographic, computational and experimentaltechniques.

3.3. Correlation between RM0, b and C0 values using SOTLC andRP-TLC

The chromatographically obtained lipophilicity parameters;RM0, b and C0 obtained in the current study are compared tothose previously reported using RP-TLC method for five of theinvestigated compounds (OFL, NOR, SPA, CIP and LOM) using fivetypes of organic modifier/water systems as mobile phases [34].A good correlation exists between RM0, b, C0 determined usingmethanol/water or acetone/water and RM0, b, C0 obtained in thecurrent study, which highlights the usefulness of the two tech-niques in estimation of lipophilicity (Table 4). Poor correlationexists between RM0, b, C0 obtained using either acetonitrile/water(Table 4) or 2-propanol/water or tetrahydrofuran/water (resultsnot shown) and RM0, b, C0 achieved in the present study. How-ever, in all systems, the obtained RM0 values are much higher thanreported in the current study, despite the fact of the high polar-ity of the studied FQs. This can be attributed to the high solubilityof the studied compounds in organic solvents. Moreover in RP-TLCtechnique, the ionization states of FQs have not been taken intoconsideration and weak correlations exists between RM0 (in all thefive systems for five of the studied drugs) and log D7.4 or log D2.0 (r< 0.7832).

3.4. QSRR analysis

The principle of QSRR approach was previously described byKalizan [40,41]. Some physicochemical parameters calculated forthe studied FQs in addition to the correlation matrix of lipophilicityparameter (RM0) obtained experimentally and calculated moleculardescriptors are listed in Table 5 and 6.

The structural and molecular size-related descriptors: ASA H,total hydrophobic surface area; MR, molar refractivity; vol, van derWaals volume; vsurf D1, hydrophobic volume at -0.2 kcal/mol andlog Ptrue (Table 3) are in a good correlation with chromatographicretention parameter (RM0). Each of the previously mentioneddescriptors is related to the size and lipophilicity of investi-gated drugs. This is consistent with the theory that non-specifichydrophobic interactions are decisive in this chromatographictechnique. Thus, the QSRR analysis could describe very well salting-out process in thin layer chromatography where the primarydriving force is hydrophobic interactions [8].

Moreover, two QSRR (quantitative-structure retention relation-ship) models were derived using multiple linear regression analysisbetween either RM0 or C0 of the studied FQs and log Ptrue and logD7.4 (predicted using ACD/Labs online service).

RM0 = 2.27 + 0.195 log Ptrue + 0.662 log D7.4 (Model I)(r = 0.9660, R2 = 0.9332, n = 7).C0 = 8.36 - 0.22 log Ptrue + 2.63 log D7.4 (Model II)(r = 0. 8718, R2 = 0.7600, n = 7).As given from the two models, RM0 and C0 are strongly influ-

enced by log D7.4 and subsequently the ionization state of the drug,which necessitate the importance of ratio of ionized/non-ionizedspecies of the molecule in estimation of the true lipophilicity of thecompound.

A.H. Rageh et al. / Journal of Pharmaceutical and Biomedical Analysis 159 (2018) 363–373 369

Fig. 3. Scores and loadings of Principal Component Analysis of the lipophilicity matrix. The missing values of XLOP2 of GAT and log DiepH of LEV were replaced by a meanvalue during the analysis.

Fig. 4. Scores and loadings of Principal Component Analysis of the lipophilicity matrix. The missing value log DiepH of LEV was replaced by a mean value during the analysis.

Table 4Correlation between RM0 values obtained using RP-TLC and those obtained using SOTLC (current study).

Drug RM0

(this study)b(this study)

C0

(this study)RM0 MeOHa RM0 ACNa RM0 acetonea b MeOHa b ACNa b acetonea C0 MeOHb C0 ACNb C0 acetoneb

LEV 1.042 0.3289 3.17 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.OFL 0.904 0.4599 1.97 2.632 2.444 2.566 −0.016 −0.021 −0.022 169.8 119.2 115.1GAT 0.723 0.5170 1.40 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.NOR 0.111 0.5971 0.19 1.592 1.979 1.658 −0.006 −0.018 −0.013 269.81 111.2 129.5SPA 0.951 0.3177 2.99 3.533 3.131 2.897 −0.032 −0.032 −0.029 112.1 97.8 99.5CIP 0.225 0.5829 0.48 2.175 2.526 2.371 −0.010 −0.022 −0.019 219.7 113.3 126.8LOM 0.398 0.5101 0.78 2.492 2.534 2.381 −0.017 −0.024 −0.021 149.2 106.1 116.1r between RM0 (this study) and previous study 0.8755 0.7104 0.8410r between b (this study) and previous study 0.9735 0.8599 0.9484r between C0 (this study) and previous study 0.8227 0.4391 0.9453

n.d.: no data.a Values taken from ref. [34].b Calculated by using the following formula C0 in RP-TLC = - RM0/b [37].

370 A.H. Rageh et al. / Journal of Pharmaceutical and Biomedical Analysis 159 (2018) 363–373

Table 5Physicochemical parameters, MIC, 1/MIC values, Log 1/MIC and log BB calculated and reported for the studied FQs.

Drug M.wta Log Pa ASA-Ha MRa vola vsurf D1a Log Sa Log BBb MIC(SA-1199c) 1/MIC (SA-1199) Log 1/MIC SA-1199 Log 1/MIC S.aureusd

LEV 361.4 0.677 568.4 9.15 321 759.3750 −2.84 −0.11 n.d. n.d. n.d. 6.5000OFL 361.4 0.677 569.4 9.15 320.9 761.8750 −2.84 −0.11 0.2 5 0.6990 n.dGAT 375.4 1.414 600.8 9.63 343.6 712.2500 −3.11 −0.1 0.1 10 1.0000 n.d.NOR 319.3 0.725 530.6 8.28 290.6 657.6250 −2.51 −0.09 0.78 1.2821 0.1079 5.5500SPA 392.4 1.36 612.8 9.74 346.5 771.8750 −3.4 −0.14 0.1 10 1.0000 6.8000CIP 331.3 1.037 548.4 8.53 302 687.3750 −2.73 −0.13 0.39 2.5641 0.4089 6.1600LOM 351.3 1.299 556.6 8.77 310.8 704.1250 −3.13 −0.06 0.78 1.2821 0.1079 5.8500

Abbreviations: M.wt, molecular weight; LogP(o/w), octanol–water partition coefficient; ASA H, total hydrophobic surface area; MR, molar refractivity; vol, van der Waalsvolume; vsurf D1, hydrophobic volume at -0.2 kcal/mol; Log S, aqueous solubility; Log BB, logarithm of the brain/blood concentration ratio. n.d.: no data.

a Calculated using MOE 2014 software.b Calculated using ACD/Labs online service.c Values taken from [42].d Values taken from [43].

Table 6Correlation matrix of lipophilicity parameter (RM0) obtained experimentally and calculated molecular descriptors using MOE 2014 and ACD/Labs software.

Name a (RM0) M.wt Log P (o/w) ASA-H MR vol vsurf D1 Log S Log BB

a (RM0) 1.000 0.843 −0.011 0.741 0.833 0.779 0.956 −0.539 −0.377M.wt 1.000 0.507 0.970 0.981 0.975 0.825 −0.881 −0.348Log P 1.000 0.605 0.491 0.570 0.000 −0.788 0.023ASA-H 1.000 0.984 0.993 0.705 −0.861 −0.436MR 1.000 0.995 0.773 −0.811 −0.389vol 1.000 0.721 −0.841 −0.378vsurf D1 1.000 −0.587 −0.451Log S 1.000 0.148Log BB 1.000

M.wt, molecular weight; LogP(o/w), octanol–water partition coefficient; ASA H, total hydrophobic surface area; MR, molar refractivity; Log S, aqueous solubility; vol, vander Waals volume; vsurf D1, hydrophobic volume at -0.2 kcal/mol; Log BB, logarithm of the brain/blood concentration ratio.

Table 7QSAR and QSPR models for the studied FQs calculated using MOE 2014 software.

Model n R2 RMSE

Antibacterial activitya

Log 1/MIC (SA-1199) = −4.55999 + 0.0008 RM0 + 0.0070 vsurf D1 6 0.8488 0.1453Log 1/MIC (SA-1199) = −4.78721 + 0.01674 vol 6 0.8456 0.1469

Antibacterial activityb

Log 1/MIC (S.aureus) = −6.1293-0.97071 RM0 + 0.01792 vsurf D1 5 0.8980 0.1423Log 1/MIC (S.aureus)= −1.15631 + 0.0102 vsurf D1 5 0.8611 0.1228Log 1/MIC (S.aureus) = 5.7048 + 0.3070 C0 5 0.7863 0.2060

Blood-Brain Barrier PenetrationC0 = 6.1761 – 10.4336 logBB + 2.2269 logD7.4 7 0.8097 0.4802C0 = -6.6876 – 15.8440 logBB + 2.1528 logPtrue 7 0.7912 0.4995

Abbreviations: n, number of compounds; R2, correlation coefficient; RMSE, root mean standard error; RM0, C0, chromatographic retention parameters; vsurf D1, hydrophobicvolume at -0.2 kcal/mol; vol, van der Waals volume; Log BB, logarithm of the brain/blood concentration ratio; logD7.4, distribution coefficient at physiological pH (pH = 7.4);l

3

fm[

ebeMdoivRtw

ogPtrue, experimental lipophilicity.a MIC values taken from [42].b MIC values taken from [43].

.5. QSAR and QSPR analysis

The values for minimum inhibitory concentration values (MIC)or wild type S. aureus (SA-1199) were reported in [42], while the

olar MIC values for the Bayer AG S. aureus strain were reported in43].

The quantitative structure–activity relationship (QSAR) mod-ls of the studied FQs are constructed to determine correlationsetween their antibacterial activity and the physicochemical prop-rties. This could be done by calculating 2D or 3D descriptors usingOE 2014 software. A highly predictive QSAR model with a high

etermination coefficient (R2) value can be used in the estimationf the activities of new compounds. According to the results (shownn Table 7), models having the 3D descriptors for the surface area,

2

olume, and shape exhibited the highest R values and the lowestMSE even when using only one descriptor to build the model. Fur-hermore, the chromatographic retention parameters (RM0 and C0)ere efficiently used as variables descriptors in these models. As

also shown in Table 7, the hydrophobicity volume at −0.2 kcal/mol(vsurf D1) and the van der Waals volume (vol.) are the most cor-related descriptors and are directly proportional to their activity.This emphasizes the importance of compound volume (dependenton the bulkiness of C-7 and C-8 substituents) on the bactericidalactivity of FQs against S. aureus. These finding are also in agreementwith previous QSAR models [42,43] and can explain the lowest MICvalue of SPA with the bulky cis 2,6-dimethylpiperazine substituentat C-7 (Fig. 5).

Validation of the QSAR model was conducted to evaluate thepredicted activities and the residuals for the studied molecules. Forexample on using (Log 1/MIC SA-1199 = -4.55999 + 0.0008 RM0 +0.0070 vsurf D1) QSAR model to predict the antibacterial activitiesof the studied FQs, minimal residuals between the predicted log

1/MIC values compared to the experimentally measured data wereobtained (Table 8). This observation was an indication of the pre-dictive power of this model. Furthermore, no z-score values morethan 2.5 are obtained indicating the absence of outliers.

A.H. Rageh et al. / Journal of Pharmaceutical and Biomedical Analysis 159 (2018) 363–373 371

F ied FQu

wTataa

Fu

ig. 5. 3D plot between antibacterial activity (Log 1/MIC against SA-1199) of the studsing MOE software).

According to the FDA safety review, systemic FQs are associatedith disabling and potentially permanent serious CNS side effects.

hese side effects can involve the tendons, muscles, joints, nerves,

nd central nervous system [35]. In trying to explain the cause ofhese CNS side effects, most researchers are focusing on the shapend size of the C-7 substituents (as piperazine and pyrrolidine) thatllow FQs to bind and displace Gamma-aminobutyric acid (GABA)

ig. 6. 3D plot between logarithm of the brain/blood concentration ratio of the studied FQsing MOE software).

s with their RM0 and hydrophobicity volume (vsurf D1) (calculation was performed

from its receptor leading to the general CNS stimulation causingthese side effects [44–46].

Herein, we tried to construct a QSPR model that would predict

BBB penetration associated with systemic FQs. As a general concept,drugs having CNS side effects must firstly be able to pass the BBB toreach their target. However, in our case only crossing the BBB can’tbe the sole decisive factor for FQs CNS side effects. This is because

s (log BB) with their C0 and the ACD calculated logD 7.4 (calculation was performed

372 A.H. Rageh et al. / Journal of Pharmaceutical and

Table 8Validation of a QSAR model for prediction of antibacterial activity.

Drug MICa Log 1/MIC $ Pred $ Res $Z-score

OFL 0.2 0.6990 0.7827 −0.0837 0.5757GAT 0.1 1.0000 1.0000 0.0000 0.0000NOR 0.78 0.1079 0.0511 0.0568 0.3909SPA 0.1 1.0000 0.8528 0.1472 1.0127CIP 0.39 0.4089 0.2598 0.1491 1.0261LOM 0.78 0.1079 0.3774 -o.2695 1.8540

Abbreviations: $Pred, predicted log 1/MIC value; $Res, residual value from exper-ip

at(ps

btvs

4

wsdemttcoetmsn

MQcdcFu

C

A

i0

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

mental data; $Z-score, the number of standard deviations from the mean a dataoint is.a MIC values taken from [42].

fter crossing BBB, FQs need to have structure similarity to GABAo bind to its receptor [44]. Therefore, in-silico estimation of Log BBlogarithm of the brain/blood concentration ratio) could be a keyredictor of the amount of FQs crossing the BBB regardless of itsimilarity to GABA.

As shown in Table 7, the log BB values are highly influencedy the FQs lipophilicity. Good correlations were obtained betweenhe estimated C0 values with both FQs logBB and logD7.4 or logPtrue

alues. The use of logD7.4 highlights the effect of FQs ionizationtate on their BBB permeability (Fig. 6).

. Conclusions

In the current work, lipophilicity and retention parametersere estimated for seven FQs using SOTLC based on ammonium

ulphate-containing mobile phase. The chromatographically-etermined lipophilicity parameters are highly correlated withxperimentally-reported log P values obtained using shake-flaskethod, however, SOLTC is very simple, fast and environmen-

ally benign with no consumption of organic solvents if comparedo RP-TLC or shake flask method. On the other side, weakorrelation exists between most of the computational meth-ds for lipophilicity estimation and chromatographically- orxperimentally-determined lipophilicity which confirms the facthat it is still early to employ computational techniques as sole

ethods for determination of lipophilicity and highlights at theame time the importance of chromatographic methods as alter-ative substitute.

Multiple linear regression analysis was performed to predictIC and BBB penetration of the examined drugs. Several efficientSAR and QSPR models were constructed using the calculatedhromatographic parameters (RM0, and C0) that are predictive foretermination of MIC and BBB penetration of some FQs. The out-omes of this work can be further expanded to newly-synthesizedQs with insufficient information about their lipophilicity, MIC val-es or BBB penetration.

onflict of interest

There are no conflicts of interest to declare

ppendix A. Supplementary data

Supplementary material related to this article can be found,n the online version, at doi:https://doi.org/10.1016/j.jpba.2018.07.10.

eferences

[1] Y. Henchoz, B. Bard, D. Guillarme, P.A. Carrupt, J.L. Veuthey, S. Martel,Analytical tools for the physicochemical profiling of drug candidates topredict absorption/distribution, Anal. Bioanal. Chem. 394 (2009) 707–729.

[

[

Biomedical Analysis 159 (2018) 363–373

[2] M.P. Gleeson, Generation of a set of simple, interpretable ADMET rules ofthumb, J. Med. Chem. 51 (2008) 817–834.

[3] D. Milojkovic-Opsenica, F. Andric, S. Segan, J. Trifkovic, Z. Tesic, Thin-layerchromatography in quantitative structure-activity relationship studies, J. Liq.Chromatogr. Relat. Technol. 41 (2018) 272–281.

[4] A.W. Sobanska, Application of planar chromatographic descriptors to theprediction of physicochemical properties and biological activity ofcompounds, J. Liq. Chromatogr. Relat. Technol. 41 (2018) 255–271.

[5] K. Ciura, S. Dziomba, J. Nowakowska, M.J. Markuszewski, Thin layerchromatography in drug discovery process, J. Chromatogr. A 1520 (2017)9–22.

[6] Z.L. Tesic, D.M. Milojkovic-Opsenica, TLC determination of drug lipophilicity,in: L. Komsta, M. Waksmundzka-Hajnos, J. Sherma (Eds.), Thin LayerChromatography in Drug Analysis, CRC Press, Taylor & Francis Group, BocaRaton, 2013, pp. 225–246.

[7] K. Wicha-Komsta, L. Komsta, Unconventional TLC systems in lipophilicitydetermination: a review, J. Liq. Chromatogr. Relat. Technol. 40 (2017)219–225.

[8] J. Flieger, R. Swieboda, M. Tatarczak, Chemometric analysis of retention datafrom salting-out thin-layer chromatography in relation to structuralparameters and biological activity of chosen sulphonamides, J. Chromatogr. BAnalyt. Technol. Biomed. Life Sci. 846 (2007) 334–340.

[9] K.E. Bij, C. Horvath, W.R. Melander, A. Nahum, Surface silanols in silica-bondedhydrocarbonaceous stationary phases, J. Chromatogr. A 203 (1981) 65–84.

10] K. Ciura, M. Belka, P. Kawczak, T. Baczek, M.J. Markuszewski, J. Nowakowska,Combined computational-experimental approach to predict blood brainbarrier (BBB) permeation based on g̈reens̈alting-out thin layerchromatography supported by simple molecular descriptors, J. Pharm.Biomed. Anal. 143 (2017) 214–221.

11] T.J. Janjic, V. Zivkovic, G. Vuckovic, M.B. Celap, Salting-out thin-layerchromatography of transition metal complexes. II. Mixedaminocarboxylatocobalt(III) complexes on a polyacrylonitrile sorbent, J.Chromatogr. 626 (1992) 305–309.

12] T.J. Janjic, V. Zivkovic, G. Vuckovic, M.B. Celap, Salting-out thin-layerchromatography of transition metal complexes. III: investigation of thebehavior of mixed aminocarboxylato cobalt(III) complexes on cellulose,Chromatographia 37 (1993) 534–538.

13] T.J. Janjic, V. Zivkovic, M.B. Celap, Salting-out thin-layer chromatography oftransition metal complexes. IV: effect of sorbent on the behavior of mixedaminocarboxylato cobalt(III) complexes, Chromatographia 38 (1994)447–452.

14] T.J. Janjic, V. Zivkovic, M.B. Celap, Linear relation between RM values obtainedin salting-out planar chromatography and mol % ammonium sulfate inaqueous solvent systems, Chromatographia 38 (1994) 355–358.

15] T.J. Janjic, G. Vuckovic, M.B. Celap, Linear interdependence of RM pair valuesof substances in salting-out planar chromatography, Chromatographia 42(1996) 675–679.

16] D.M. Milojkovic-Opsenica, M.J. Malinar, Z.L. Tesic, Effect of the chelate ringsize of diamine-Co(III) complexes on their hRF values obtained undersalting-out thin-layer chromatography conditions, J. Chromatogr. A 847(1999) 291–296.

17] G. Vuckovic, D. Miljevic, T.J. Janjic, M.I. Djuran, M.B. Celap, Salting-out thinlayer chromatography of transition metal complexes: a comparative study ofthe effect of increased number of CH2 groups in chelate rings,Chromatographia 40 (1995) 445–447.

18] G. Vuckovic, D. Miljevic, T.J. Janjic, M.B. Celap, Salting-out thin-layerchromatography of transition metal complexes. I. Investigation of thebehavior of mixed aminocarboxylato cobalt(III) complexes on silica gel, J.Chromatogr. 609 (1992) 427–431.

19] G. Vuckovic, S.B. Tanaskovic, T.J. Janjic, D. Milojkovic-Opsenica, M.B. Celap,Effect of increasing the bidentate chelate ring size in Co(III) complexes ontheir behavior in salting-out TLC on different adsorbents, J. PlanarChromatogr. Mod. TLC 12 (1999) 461–465.

20] V. Zivkovic-Radovanovic, G. Vuckovic, Poly(ethylene glycol) as impregnatorfor silica gel in salting-out thin-layer chromatography of some Co(III)complexes, Chromatographia 62 (2005) 91–97.

21] K. Ciura, J. Nowakowska, K. Rudnicka-Litka, P. Kawczak, T. Baczek, M.J.Markuszewski, The study of salting-out thin-layer chromatography and theirapplication on QSRR/QSAR of some macrolide antibiotics, Monatsh. Chem.147 (2016) 301–310.

22] T.B. Tosti, K. Drljevic, D.M. Milojkovic-Opsenica, Z.L. Tesic, Salting-outthin-layer chromatography of some macrolide antibiotics, J. PlanarChromatogr. - Mod. TLC 18 (2005) 415–418.

23] A.M.I. Mohamed, F.A.-F. Mohamed, S.A.-R. Ahmed, A.S. Aboraia, Y.A.S.Mohamed, Salting-out thin-layer chromatography and computationalanalysis of some oral hypoglycemic drugs, Biomed. Chromatogr. 28 (2014)1156–1162.

24] Y.A.S. Mohamed, A.M.I. Mohamed, F.A.-F. Mohamed, S.A.-R. Ahmed, Newsalting out stability-indicating and kinetic thin layer chromatographicmethod for determination of glimepiride and metformin HCl binary mixture,J. Chromatogr. Sci. 53 (2015) 1603–1610.

25] M. Aleksic, J. Odovic, D.M. Milojkovic-Opsenica, Z.L. Tesic, Salting-outthin-layer chromatography of several myorelaxants, J. Planar Chromatogr.Mod. TLC 16 (2003) 144–146.

26] J. Odovic, B. Markovic, S. Vladimirov, K. Karljikovic-Rajic, In vitro modeling ofangiotensin-converting enzyme inhibitor’s absorption with chromatographic

al and

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[45] A. De Sarro, G. De Sarro, Adverse reactions to fluoroquinolones. An overviewon mechanistic aspects, Curr. Med. Chem. 8 (2001) 371–384.

[46] S. Munjal, Y. Smolin, Ciprofloxacin associated exacerbation of anxiety in an

A.H. Rageh et al. / Journal of Pharmaceutic

retention data and selected molecular descriptors, J. Chromatogr. B Analyt.Technol. Biomed. Life Sci. 953–954 (2014) 102–107.

27] J.V. Odovic, B.B. Stojimirovic, M.B. Aleksic, D.M. Milojkovic-Opsenica, Z.L.Tesic, Examination of the hydrophobicity of ACE inhibitors and their activemetabolites by salting-out thin-layer chromatography, J. Planar Chromatogr.Mod. TLC 18 (2005) 98–103.

28] M.B. Aleksic, D.G. Agbaba, D.M. Milojkovic-Opsenica, Z.L. Tesic, Salting-outthin-layer chromatography of some angiotensin converting enzymeinhibitors on different sorbents, Chromatographia 53 (2001) S442–S444.

29] L. Komsta, R. Skibinski, A. Bojarczuk, M. Radon, Salting-out chromatography -a practical review, Acta Chromatogr. 23 (2011) 191–203.

30] L. Komsta, R. Skibinski, A. Bojarczuk, Chemometric characterization of modelcompounds retention in salting-out thin layer chromatography on cellulose, J.Liq. Chromatogr. Relat. Technol. 35 (2012) 1298–1305.

31] L. Komsta, R. Skibinski, M. Radon, Chemometric characterization of modelcompounds retention in salting-out thin layer chromatography on silica, J.Liq. Chromatogr. Relat. Technol. 34 (2011) 776–784.

32] G. Volgyi, G. Vizseralek, K. Takacs-Novak, A. Avdeef, K.Y. Tam, Predicting theexposure and antibacterial activity of fluoroquinolones based onphysicochemical properties, Eur. J. Pharm. Sci. 47 (2012) 21–27.

33] E. Klosinska-Szmurlo, M. Grudzien, K. Betlejewska-Kielak, F.A. Plucinski, J.Biernacka, A.P. Mazurek, Physico-chemical properties of lomefloxacin,levofloxacin and moxifloxacin relevant to biopharmaceutics classificationsystem, Acta Chim. Slov. 61 (No 4) (2014) 827–834.

34] U. Hubicka, B. Zuromska-Witek, L. Komsta, J. Krzek, Lipophilicity study offifteen fluoroquinolones by reversed-phase thin-layer chromatography, Anal.Methods 7 (2015) 3841–3848.

35] FDA Drug Safety Communication: FDA updates warnings for oral and

injectable fluoroquinolone antibiotics due to disabling side effects. https://www.fda.gov/Drugs/DrugSafety/ucm511530.htm (Accessed 07.04.2018).

36] E.C. Bate-Smith, R.G. Westall, Chromatographic behaviour and chemicalstructure I. Some naturally occuring phenolic substances, Biochim. Biophys.Acta 4 (1950) 427–440.

Biomedical Analysis 159 (2018) 363–373 373

37] M.L. Bieganowska, A. Doraczynska-Szopa, A. Petruczynik, The retentionbehaviour of some sulphonamides on different TLC plates. 2. Comparison ofthe selectivity of the systems and quantitative determination ofhydrophobicity parameters, J. Planar Chromatogr. Mod. TLC 8 (1995) 122–128.

38] E. Klosinska-Szmurlo, F.A. Plucinski, M. Grudzien, K. Betlejewska-Kielak, J.Biernacka, A.P. Mazurek, Experimental and theoretical studies on themolecular properties of ciprofloxacin, norfloxacin, pefloxacin, sparfloxacin,and gatifloxacin in determining bioavailability, J. Biol. Phys. 40 (2014)335–345.

39] A. Mornar, M. Damic, B. Nigovic, Pharmacokinetic parameters of statin drugscharacterized by reversed phase high-performance liquid chromatography,Anal. Lett. 44 (2011) 1009–1020.

40] R. Kaliszan, Quantitative structure-retention relationships, Anal. Chem. 64(1992) 619A–631A.

41] R. Kaliszan, QSRR: quantitative structure-(chromatographic) retentionrelationships, Chem. Rev. 107 (2007) 3212–3246.

42] T. Takenouchi, F.F. Tabata, Y.F. Iwata, H.F. Hanzawa, M.F. Sugawara, S. Ohya,Hydrophilicity of quinolones is not an exclusive factor for decreased activityin efflux-mediated resistant mutants of Staphylococcus aureus, Antimicrob.Agents Chemother. 40 (1996) 1835–1842.

43] G. Cianchetta, R.F. Mannhold, G.F. Cruciani, M.F. Baroni, V. Cecchetti,Chemometric studies on the bactericidal activity of quinolones via anextended VolSurf approach, J. Med. Chem. 47 (2004) 3193–3201.

44] A. Kamath, Fluoroquinolone induced neurotoxicity: a review, Journal ofAdvanced Pharmacy Education & Research 3 (2013) 16–19.

elderly patient: a review of anxiogenic potential of fluoroquinolones, J. Clin.Psychopharmacol. 37 (2017) 370–372.