Vibrational spectra and antimicrobial activity of selected bivalent cation benzoates

Article abstract of DOI:10.1016/j.molstruc.2008.09.018

Selected bands of FT-IR spectra of Mg(II), Ca(II), Cu(II) and Zn(II) benzoates of both solid state and water solution, were assigned to appropriate molecular vibrations. Next evaluation of electronic charge distribution in both carboxylic anion and aromat

Full text of DOI:10.1016/j.molstruc.2008.09.018

Journal of Molecular Structure 919 (2009) 284–289
Contents lists available at ScienceDirect
Journal of Molecular Structure
journal homepage: www.elsevier.com/locate/molstruc
Vibrational spectra and antimicrobial activity of selected bivalent cation benzoates
a
b,
a
c
c
*
´
´
M.H. Borawska , P. Koczon , J. Piekut , R. Swisłocka , W. Lewandowski
a Department of Bromatology, Medical University, Kilinskiego 1, 15-230 Białystok, Poland
b Food Chemistry Division, Department of Chemistry, Warsaw Agricultural University, Nowoursynowska 159 C, 02-728 Warsaw, Poland
c Department of Chemistry, Białystok Technical University, Zamenhofa 29, 15-435 Białystok, Poland
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 7 May 2008
Received in revised form 22 September
2008
Accepted 22 September 2008
Available online 30 September 2008
Selected bands of FT-IR spectra of Mg(II), Ca(II), Cu(II) and Zn(II) benzoates of both solid state and water
solution, were assigned to appropriate molecular vibrations. Next evaluation of electronic charge distri-
bution in both carboxylic anion and aromatic ring of studied compounds was performed. Classical plate
tests and turbidimetry measurements, monitoring growth of bacteria Escherichia coli, Bacillus subtilis and
yeasts Pichia anomala and Saccharomyces cerevisiae during 24 h of incubation, in optimal growth condi-
tions (control) and in medium with addition of studied benzoate (concentration of 0.01% expressed as
the concentration of benzoic acid), proved antimicrobial activity of studied compounds against investi-
gated micro-organisms.
Keywords:
FT-IR spectroscopy antimicrobial activity
PLS (partially least square) and PCR (principal component regression) techniques were applied to build
a model, correlating spectral data reflecting molecular structure of studied compounds, with degree of
influence of those compounds on growth of studied micro-organisms. Statistically significant correlation
within cross validation diagnostic of PLS-1 calibration was found, when log1/T of selected spectral
regions of water solution samples were used as input data. The correlation coefficients between predicted
with PLS calibration based on created 1, 2 or 3 factor models, and actual values of antimicrobial activity
were: 0.70; 0.76, 0.81 for P. anomala, B. subtilis, and E. coli, respectively. Log(PRESS) values of appropriate
models were 2.10, 2,39 and 3.23 for P. anomala, B. subtilis, and E. coli, respectively.
Benzoates
PLS
PCR
Ó 2008 Elsevier B.V. All rights reserved.
1. Introduction
ber of experimental data. Those statistical procedures are well
known as a proper and valuable tool to simplify IR spectral data,
Benzoic acid derivatives, e.g. salicylic, nicotinic, picolinic and o-,
m-, p-halogenobenzoic acids and their complexes with different
metal cations were investigated in our previous papers [1–5].
Qualitative and quantitative influence of metal cations on the elec-
tronic charge distribution in ligands which are of importance for
various industrial branches (e.g. pharmacy, cosmetics, food) was
established. Metal parameters, such as ionic potential, atomic mass
and electronegativity were proved to be factors responsible for the
effects of various metal cations on the electronic structure of car-
boxylate anion and aromatic ring of various benzoic acid deriva-
tives. The main conclusions of our previous studies were drawn
based on analysis of spectral data obtained with use of different
methods (IR and Raman, UV–vis, 1H and 13C NMR). We have also
tested survival of micro-organisms in presence of studied benzoic
acid derivatives to evaluate their antimicrobial potential [6–9].
As the vibrational spectra are complicated due to a number of
overlapping spectral bands, the techniques of PCA (principal com-
ponent analysis), PCR (principal component regression) and PLS
(partially least square) are commonly applied to reduce large num-
and therefore have been successfully used by many authors [10–
13]. PLS is particularly useful when the number of predictors is
higher than number of observations and ordinary least squares
regression either fails or produces small coefficients with large
standard errors.
It is well known that antimicrobial activity or more general bio-
logical activity of chemical compound varies significantly due to
even minimal changes in chemical structure. One clear example
is antitumor active cis-platine with its trans isomer lacking antitu-
mor properties [14]. Number of authors determine SAR (structure–
activity relation), considering various parameters of studied
compounds. One of the parameters is
p value, defined as difference
between logPX and logPH, where PX is the partition coefficient of
benzoic acid in the octanol–water system, and PH is the partition
coefficient of a given compound in the same system [15]. Also
wavenumber of simple NO stretches of series of compounds con-
taining NO group is considering as parameter responsible for anti-
microbial activity [16]. The theory of SAR lists a lot of different
parameters (e.g. atomic mass, electronegativity, dipole momen-
tum, hydrogen bonds, atomic formal charge, HOMO/LUMO, inter-
molecular interaction) determining biological activity [17]. Most
of listed parameters are undoubtedly related to molecular
* Corresponding author. Tel.: +48 22 59 37 616; fax: +48 22 59 37 635.
E-mail address: koczon@wp.pl (P. Koczon´ ).
0022-2860/$ - see front matter Ó 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.molstruc.2008.09.018

M.H. Borawska et al. / Journal of Molecular Structure 919 (2009) 284–289
285
structure of a given compound. In fact molecular structure and
electronic charge distribution determine micro and macro proper-
ties, including biological activity. The task is to construct proper
mathematical formula connecting features of both structure and
activity. In our study [6–8], as well as in study of other authors
[16], data provided by IR spectra are considered as extremely
important, as they include information on movement of atoms
and amount of bonds energy. Latter features are in turn responsible
for molecular structure (e.g. bond length or angels).
(Perkin Elmer) with 10 T of pressure for 1 min. Transparent pellets
were placed in the holder and spectra were registered against
background spectrum. Background spectrum was registered with
pellet prepared in the same way as described above, but with
KBr powder alone.
Two different techniques: attenuated total reflectance and
transmittance were used to register water solution spectra. Reflec-
tance technique spectra were registered using Perkin Elmer HATR
accessory (ZnSe, 60° crystal), while transmittance spectra were
registered in film, using KRS plates. In both cases System 2000 Per-
kin Elmer spectrophotometer was used, however spectral range
was 2000ꢀ600 and 2000ꢀ400 cmꢀ1 for reflectance and transmit-
tance techniques, respectively. Two techniques had to be applied
as intensities of some bands were extremely low when using
reflectance technique alone.
Two species of bacteria: E. coli, and B. subtilis as well as two spe-
cies of yeasts: S. cerevisiae and P. anomala were used for antimicro-
bial tests. (S. cerevisiae – ATCC 9763, E. coli – ATCC 25922, B. subtilis
– ATCC 6633, P. anomala – the collection of Institute of Agricultural
and Food Biotechnology in Warsaw). The studied micro-organisms
were inoculated on broth medium and stored at 35 °C, for 24 h. The
growth of the tested cells was standardized using turbidimetry
analysis by measuring optical density at 600 nm for B. subtilis,
E. coli, P. anomala, and S. cerevisiae with a V-2001 HITACHI spectro-
photometer [23]. The microbiological tests of the studied com-
pounds (the final concentration was 0.01% expressed as the
concentration of benzoic acid) were carried out in water medium.
The ultra pure water was used (deionised water, resistance
In our previous studies we have proved the relation between IR
spectral data and antimicrobial activity of some benzoic acid deriv-
atives. Continuing our research in this paper we propose PLS tech-
nique to correlate IR data with antimicrobial properties of selected
bivalent cation benzoates. Our investigation is aimed at finding ro-
bust model to establish correlation between molecular structure
characterised by spectral data reduced with PCA or PLS, and anti-
microbial activity of series of compounds, which is quite new idea.
In this paper we have studied bivalent cation benzoates, namely
Mg(II), Ca(II), Cu(II) and Zn(II) benzoates to compare their elec-
tronic structures characterised by IR data, evaluate antimicrobial
activity against two bacteria and two yeast species, and finally test
the existence of the structure–antimicrobial activity relation. Pro-
posed compounds are water-soluble (especially Mg and Ca are of
good water solubility), which is an advantage in their potential
practical application. Moreover, studied metals are of great interest
from the nutritional point of view [18,19] which makes studied
salts particularly interesting due to potential application as food
preservatives. Although there are spectral data of herein proposed
complexes in literature [20–22], there is a lack of in-depth discus-
sion on the structure–antimicrobial activity relation based on spec-
tral data of water solution samples. Moreover, spectra assigned and
discussed before [20,21] by Lewandowski, were registered with an
older type of apparatus. To achieve the goals of this paper and to
improve spectra quality (e.g. bands wavenumber or under band
area), we registered spectra with use of FT-IR technique.
18.2 MX/cm). The samples were incubated at 35 °C for bacteria
and at 25 °C for yeast. The increase in the number of colonies
was estimated after 24 h of incubation. The results were presented
as the percentage of growth inhibition compared to the control
sample (the optimal conditions for growth of micro-organisms).
% of growth inhibition ¼ 100% ꢀ ½ðA1 ꢁ 100%Þ=A0ꢂ
A1, the value of optical density in the sample with the tested com-
pound; A0, the value of optical density in the control sample.
Statistical calculations were done using STATGRAPHICS Plus 2.1
package and GRAMS/AI software with PLS/IQ extension running on
Windows XP platform. Cross validation diagnostic in PLS-1 calibra-
tion was used to create the robust statistical model. Maximal num-
ber of factors was set to 3. One file was left out while building the
model with cross validation diagnostic. Two spectral regions 1571–
1528 and 1415–1366 cmꢀ1 (94 points altogether) out of several
tested regions were set, to achieve the best correlation with data
on antimicrobial activity. No one of studied compounds was
outlier.
2. Materials and methods
The benzoic acid was Sigma analytical reagent. The complexes
of Ca were prepared by dissolving benzoic acid in aqueous solution
of metal hydroxide (Merck analytical reagent) in a stoichiometric
ratio. Solid benzoates of the studied metals were obtained by mix-
ing aqueous solutions of soluble salts of the metals with sodium
benzoate. For instance, zinc benzoate was synthesized by mixing
equal volumes of solutions of ZnCl2 (0.25 mol dmꢀ3
) with
C6H5COONa (0.5 mol ꢁ dmꢀ3).The precipitates were recrystallized,
filtered off, washed with water, and dried first at room tempera-
ture and then at 95 °C. Benzoates obtained under the above condi-
tions are anhydrous. Analytical data for the compounds were
complete in accordance with the expected metal-to-benzoate stoi-
chiometries, i.e., 1:2 for Mg, Ca, Cu and Zn.
Log1/T values against wavenumbers of Na, Mg, Ca, Cu and Zn
benzoates were measured. Correlation coefficients (R2) and
log(PRESS) (Prediction Residual Error Sum of Squares) values were
used to evaluate and select best models.
KBr pellets were used for transmittance of IR measurements of
studied benzoates within the range of MIR (400–4000 cmꢀ1) with a
Perkin Elmer System 2000 FT-IR spectrophotometer. The resolu-
tion was 2 cmꢀ1. For each studied compound, five separated pellets
were prepared to average influence of pellet preparation (grinding
and pressing with KBr) on the wavenumbers of spectral bands –
possibility of such influence was discussed in our previous paper
[7]. Moreover, for magnesium benzoate HATR technique, in which
pure sample (without KBr matrix) is used, was applied to confirm
wavenumbers of spectral bands discussed in this paper. Pellets
were prepared by mixing KBr powder (Sigma reagent) with pow-
dered benzoate, dried during 10 h in the temperature of 90 °C, in
ratio of 300 mg:1 mg. Powder mixture was grounded in the ball
mill for 2 min. Then powder was pressed in the hydraulic press
3. Results and discussion
3.1. Spectral data
Wavenumbers of selected spectral bands of solid state samples
gathered in Table 1, were assigned based on our previous experi-
ence [24] and data found in literature [25]. Presence of some char-
acteristic bands (i.e.
other bands (i.e. (O–H), b(O–H) and c(O–H)) in spectra of benzo-
m
(COOꢀ), b(COOꢀ)), or disappearance of some
m
ates in comparison to spectrum of benzoic acid proves that given
metal replaces the carboxylic group hydrogen.
Bands present in the spectra originate from both carboxylic an-
ion vibrations and ring vibrations. Bands originating from carbox-

286
M.H. Borawska et al. / Journal of Molecular Structure 919 (2009) 284–289
Table 1
Wavenumbers, intensities and assignments of selected bands occurring in the IR spectra of Ca(II), Mg(II) Cu(II) and Zn(II) benzoates
Mg benzoate
Wavenumber
Ca benzoate
Cu benzoate
Wavenumber
Zn benzoate
Assignment
Intensity
Wavenumber
Intensity
Intensity
Wavenumber
Intensity
. . .
(C C)ar
1631
1598
1553
1492
1423
1069
1029
930
vs
vs
vs
m
vs
w
w
w
m
w
w
1630
1595
1558
1498
1444
1071
1025
935
vs
vs
vs
m
vs
m
m
w
m
w
w
1633
1597
1555
1492
1412
1069
1027
931
vs
vs
vs
m
vs
w
m
w
m
w
w
1639
1611
1575
1493
1382
1071
1026
941
vs
vs
vs
m
vs
m
m
w
m
m
w
m
m
m
m
m
•
. . .
asym(COOꢀ)
(C C)ar
•
. . .
(C C)ar
•
sym(COOꢀ)
b(C–H)
b(C–H)
b(C–H)
846
767
524
846
767
521
846
767
520
841
772
572
bsym(COOꢀ)
c
sym(COOꢀ)
basym(COOꢀ)
vs, very strong; s, strong; m, medium; w, weak; vw, very weak; sh, shoulder.
ylate anion vibrations are: (1) carboxylic anion asymmetric
841 for Mg, Ca, Cu and Zn benzoates, respectively. Thus,
stretches
m
asym(COOꢀ), located at 1553, 1558, 1555 and
bsym(COOꢀ) bands located at around 846 cmꢀ1 each, do not really
1557 cmꢀ1 for Mg, Ca, Cu and Zn benzoates, respectively; (2) car-
distinguish studied complexes, while
D separation does. Regular
boxylic anion symmetric stretches
m
sym(COOꢀ) at 1423, 1444,
increase in the value in the following order: Mg, Ca, Cu, Zn indi-
D
1412, and 1382 cmꢀ1 for Mg, Ca, Cu and Zn benzoates, respec-
tively; (3) in plane symmetric deformations bsym(COOꢀ) at 846,
846, 846 and 841 cmꢀ1 for Mg, Ca, Cu and Zn benzoates, respec-
tively; (4) in plane asymmetric deformations basym(COOꢀ) at 524,
521, 520 and 575 cmꢀ1 for Mg, Ca, Cu and Zn benzoates, respec-
cates chelation structure in case of Mg benzoate, bridging structure
in case of both Ca and Cu benzoates and monodentate structure in
case of Zn carboxylates, which suggests coordination number of
central ion (Zn2+) equal to 2, similarly as in ordinary zinc salts.
To recapitulate, the bonds within carboxylic anion of studied com-
plexes are of different polarity. Consequently, structure of carbox-
ylate anion significantly varies along with studied bivalent cation
change.
Moreover, a different influence of a given metal cation on vibra-
tions of aromatic ring is observed. The metal–aromatic ring inter-
action is transferred by carboxylate anion (M–O, C–O and C–C
bonds). Consequently, bands originating from aromatic ring vibra-
tions are located at different wavenumbers, depending on coordi-
nated metal. In the high frequency spectral region, above
1600 cmꢀ1, two bands occur (see the first and second rows in Table
1) and both of them slightly distinguish studied complexes. Alike,
other aromatic ring bands listed in rows seven and eight in Table 1,
distinguish studied compounds as well. Wavenumbers of the
bands originating from the same ring vibrations of studied
compounds vary, which indicates that the strength and type of
metal–ring interactions, transmitted through the differently struc-
tured carboxylic anions (differences between structures of carbox-
ylic anions of the studied complexes were proved above) depend
on the carboxylic anion structure.
tively. (5) out of plane symmetric deformations
c
sym(COOꢀ) at
767, 767, 767 and 772 cmꢀ1 for Mg, Ca, Cu and Zn benzoates,
respectively. The remaining bands present in spectra originate
from aromatic ring vibrations, the bands originating from metal–
oxygen vibrations are below spectral range involved in present
measurement.
The wavenumbers of asymmetric and symmetric stretches
(both of very strong intensity) are highly sensitive to the electronic
charge distribution, and therefore to the structure of carboxylate
group. Consequently, the separation of those bands i.e.
D
=
m
asym(COOꢀ) ꢀ
m
sym(COOꢀ) is also strongly indicative of the
is the parameter well known
structure of a given carboxylate.
D
in literature [26,27]. A general decreasing trend of the following or-
der: uncoordinated acid > unidentate coordination > bidentate
(bridging > chelating) coordination > free carboxylate anion is ob-
served for both stretching frequencies and/or band separation
values in carboxylate metal complexes. Also the IR spectra of me-
tal–acetate complexes are often described in terms of the value of
wavenumber separation between the asymmetric and symmetric
stretches
to Deacon [28] and Nakamoto [29], very low
100 cmꢀ1) indicate chelation or a combination of chelation and
bridging while larger
values (around 150 cmꢀ1) are indicative
D
rather than absolute wavenumbers [28,29]. According
Data of water solutions of studied compounds in concentration
0.01% are gathered in Table 2. Less number of bands occurs in spec-
tra of water solution samples compared to spectra of solid state
samples. Bands originating from carboxylic anion vibrations are
of small intensity (contrary to the same bands in spectra of solid
state samples), likely due to intermolecular H2O–COOꢀ interaction.
On the other hand, bands originating from ring vibrations are of
strong intensity and wide shape. That was similar.
Spectra of water solution of different metals are analogous, thus
depicted by the similar number of related intensity bands located
in comparable spectral range. Very intense and wide band of 8a
or 8b vibrations is located in the 1810–1430 cmꢀ1 region. In the
right shoulder of this intensive signal (around 1540 cmꢀ1), charac-
teristic band of weak intensity, originating from asymmetric car-
boxylic anion stretches, appears. Then around 1380 cmꢀ1 a band
of medium intensity, originating from symmetric stretches of
COOꢀ group, can be found. Within the spectral region of 1100–
400 cmꢀ1 two bands of very strong intensity are located around
770 and 470 cmꢀ1, respectively. They are present in spectra of all
studied compounds. Those bands originate from ring vibrations
D
values (around
D
of chelating or bridging acetate coordination, while substantially
greater value (around 200 cmꢀ1) are indicative of monodentate
acetate coordination. The
D values calculated from theoretical
wavenumbers [30] support the empirical data. Large
D values
around 241 cmꢀ1 are obtained from monodentate complexes,
while significantly smaller values (around 127 cmꢀ1) are calcu-
lated for bidentate and bridging structures. Although
D is com-
monly used and verifiable parameter, in order to distinguish
between different carboxylic anion configurations additional infor-
mation about the system under study is sometimes required (e.g.
location of bsym(COOꢀ) band) [25]. Shift of the wavenumber of
bsym(COOꢀ) band toward the lower wavenumbers is recognized
as identification of more ionic structure of carboxylate anion.
Following data in Table 1, for studied here complexes
D values
are as follows: 130, 114, 143, 193 for Mg, Ca, Cu and Zn benzoates,
respectively, while wavenumbers of bsym(COOꢀ) are 846, 846, 846,

M.H. Borawska et al. / Journal of Molecular Structure 919 (2009) 284–289
287
Table 2
Wavenumbers, intensities and assignments of selected bands occurring in the IR spectra of water solution of Ca(II), Mg(II) Cu(II) and Zn(II) benzoates
Mg benzoate
Wavenumber
Ca benzoate
Cu benzoate
Wavenumber
Zn benzoate
Assignment
Intensity
Wavenumber
Intensity
Intensity
Wavenumber
Intensity
. . .
(C C)ar
1644
1559
1389
1318
–
1074
755
523
vs
sh
vs
m
1637
1548
1390
1307
1177
1029
776
vs
sh
m
vw
vw
w
s
1620
1538
1328
–
vs
sh
m
1645
1546
1392
1319
1143
1120
779
vs
sh
s
vw
m
v
m
•
m
asym(COOꢀ)
sym(COOꢀ
)
m
–
w
s
s
1069
829
576
m
s
s
b(C–H)
s
s
bsym(COOꢀ)
basym(COOꢀ)
487
s
487
vs, very strong; s, strong; m, medium; w, weak; vw, very weak; sh, shoulder.
and overlap all other bands which are expected in this region (e.g.
out of plane vibrations of carboxylic anion).
Resuming spectral data, it is to say that studied here salts of dif-
ferent metals differ significantly in both structure of carboxylate
anion and aromatic ring. Therefore, it is likely that their antimicro-
bial activity would differ.
evaluated with great care, as this is the average of five independent
measurements, performed in the same experimental conditions. In
order to compare inhibition strength of studied compounds, it is
necessary to take the confidence interval into account. For exam-
ple, stimulation effect was observed for Mg benzoate action against
E. coli. In this case, the result of each replicate was under zero, with
confidence interval 2.4 and average value of ꢀ11.2. Similar stim-
ulation was observed for Mg or Ca benzoates action against S. cere-
visiae. Action of Cu and Zn benzoates against E. coli and Mg
benzoate against B. subtilis can be evaluated as neutral as in some
experiments small inhibition while in others small stimulation was
observed. On the other hand, Mg benzoate inhibited P. anomala
growth in each replicate, Ca benzoate inhibited E. coli, B. subtilis,
P. anomala, in every experiment and Cu and Zn benzoate inhibited
B.subtilis, P. anomala and S. cerevisiae each time. In order to perform
statistical correlations within this paper, the averages were set as
the values of antimicrobial activity of studied compounds.
After 24 h of incubation the studied compounds influence the
growth of studied micro-organisms except P. anomala at a signifi-
cantly lower level than control sodium benzoate. Among studied
benzoates, there are statistically significant differences between
the level of activity, denoted in the Table 3 by ‘‘p” footnote, there-
fore it was possible to distinguish those compounds by the quan-
tity of their action against micro-organisms. The level of growth
inhibition increases in the following order Cu > Zn = Cu > Mg;
Cu = Ca = Zn > Mg; Cu = Ca = Zn = Mg; and Zn > Cu > Mg = Ca for E.
coli, B. subtilis, P. anomala and S. cerevisiae, respectively.
3.2. Biological data
Table 3 contains data on inhibition effect of the studied Mg, Ca,
Cu and Zn benzoates on the growth of E. coli, B. subtilis, P. anomala
and S. cerevisiae after 24 h of micro-organism incubation. Data
gathered in Table 3 are averages of five replications. Averages in
Table 3 are followed by confidence intervals. In addition to tests
samples, control samples in conditions optimal for micro-organ-
isms growth were incubated. The sodium benzoate, which is com-
monly used preservative agent, was used as a reference material,
as it was done in our previous papers [7,8]. Sodium benzoate con-
centration in current experiments was higher than in the previous
papers, as it had to correspond to the level of concentration of cur-
rently studied compounds. Although Na benzoate concentration
was 2-fold higher in current experiments, in some cases its activity
against the same micro-organisms studied currently and before,
was unexpectedly lower comparing to previous results. This phe-
nomenon might be explained by fact that some inhibitors demand
specific concentration to show maximum activity, and increased
concentration does not improve activity.
Some data presented in Table 3 are of relatively high confidence
intervals (e.g. ꢀ11.2 2.4 for Mg benzoate against E. coli). One
must keep in mind, that those are data obtained from biological
measurements, and although all experimental conditions were
kept unchanged for each sample, high variations might have oc-
curred. Activity of the studied compounds changes from growth
inhibition to growth stimulation which is denoted by minus signs
in Table 3. Stimulation effect, as well as inhibition effect, must be
To evaluate the effect of cation alone on E. coli and S. cerevisiae
survival, microbiological tests were done with water-soluble inor-
ganic salts of Mg2+–[Mg(NO3)2], Ca2+–(CaCl2), Cu2+–(CuSO4), and
Zn2+–(ZnCl2) in the concentration of given cation equal to the
concentration of this cation in case of benzoate samples (0.01%).
Results are presented in Table 4. Our assumption was that if the
antimicrobial activity depends only (mainly) on cation, the inhibi-
tion degree expressed by inorganic salt built of the same cation as
Table 3
The degree of growth inhibition of bacteria Escherichia coli and Bacillus subtilis and yeast Pichia anomala and Saccharomyces cerevisiae caused by Mg, Ca, Cu, Zn and Na benzoates
after 24 h of incubationa
Escherichia coli
Bacillus subtilis
Pichia anomala
Saccharomyces cerevisiae
1
2
3
Na benzoate
Mg benzoate
Ca benzoate
20.5 16.6
p1–2,3,4,5 < 0.05
ꢀ11.2 2.4
p2–3,4,5 < 0.05
2.7 0.4
22.1 1.3
p1–2,3,4,5 < 0.05
0.1 1.3
p2–3,4,5 < 0.05
6.0 1.7
23.4 8.7
16.9 6.6
p1–2,3,4,5 < 0.05
ꢀ2.4 1.8
p2-4,5 < 0.05
ꢀ5.3 3.0
p3–4,5 < 0.05
4.6 3.1
23.2 2.3
19.2 8.8
p3–4,5 < 0.05
ꢀ1.4 2.5
4
5
Cu benzoate
Zn benzoate
7.6 2.1
5.3 4.4
24.7 2.8
27.0 5.4
ꢀ0.6 2.6
7.6 3.9
a
The negative values stand for growth stimulation. Each value is the average of five replicates and is followed by its confidence interval. The concentration of each
compound was 0.01% expressed as a concentration of benzoic acid. The p value stands for statistically significant difference between samples denoted by numbers from 1 to 5
at the 95% confidential level.

288
M.H. Borawska et al. / Journal of Molecular Structure 919 (2009) 284–289
Table 4
Table 5
The degree of growth inhibition of bacteria Escherichia coli and yeast Saccharomyces
cerevisiae exhibited by Mg, Ca, Cu and Zn inorganic salts after 24 h of incubation
Statistical data obtained with PLS-1 calibration (cross validation diagnostic with one
file out)
Inorganic salts
S. cerevisiae
E. coli
Number of factors in model
R2
Log(PRESS)
24 h of incubation
PLS calibration
E. coli
B. subtilis
P. anomala
S. cerevisiae
Mg2+
Ca2+
Cu2+
Zn2+
ꢀ9.4
ꢀ12.0
ꢀ9.2
8.0
7
1
7
2
3.6
ꢀ2.1
3.8
6
7
1
2
1
1
3
2
0.81
0.70
0.76
0.55
3,23
2,39
2,10
3,46
7.4
Two spectral regions of total 94 points of five different samples were involved in
PLS-1 calibration.
in case of benzoate, will be the same as inhibition degree shown by
studied benzoate. In the case of E. coli there is a statistically signif-
icant difference between the degree of action of Zn, Ca, Cu and Mg
benzoates and their inorganic salts, while in the case of S. cerevisiae
the differences are statistically insignificant for Zn salts.
A conclusion can be drawn that in case of benzoates anion is
mainly responsible for antimicrobial action.
inhibition was found. For E. coli (24 h of incubation) correlation
coefficient equal to 0.81 and log(PRESS) = 3.23 was obtained with
one-factor model. One factor model was selected, as R2 for two-
and three-factor models were significantly lower, while log(PRESS)
significantly higher. For B. subtilis one-factor model produced cor-
relation coefficient of 0.70, with log(PRESS) = 2.39, and for P. ano-
mala three-factor model produced correlation coefficient of 0.76
with log(PRESS) = 2.10. Two- and three-factor models obtained
for B. subtilis and one and two factor models obtained for P. ano-
mala showed significantly lower R2 and significantly higher
log(PRESS). Fig. 1 presents plot of actual versus predicted antimi-
crobial activity of studied compounds against B. subtilis.
3.3. Structure–antimicrobial activity relationship
Chemometrics is widely used to establish the correlation be-
tween two sets of variables derived from the same samples. Within
multivariate models IR spectral data are related with data of con-
tent of proteins, sugars, fats and other components of foodstuff
[31–33] or with physical and chemical properties of gasoline
[34]. This technique enables to achieve results similar to those
obtainable with standard methods at shorter times, lower costs
and safer sample handling. Relating wavenumbers of selected IR
bands with antimicrobial activity is quite new idea. The model
once validated can be applied as a tool in routine procedure, e.g.
for preselecting particularly interesting compounds from large
group of compounds due to biological activity.
In our approach different data were processed using simple
regression followed by multivariate regression. Spectral parame-
ters e.g. various single band wavenumbers, band intensities, or dif-
ferences between selected band wavenumbers or non-spectral
parameters e.g. atomic mass, reversed atomic mass, electro nega-
tivity were initially analyzed due to their relation to microbiologi-
cal activity. Unfortunately, relation between those parameters and
biological activity was at statistically insignificant level.
For solid state samples statistically significant models were only
obtained when frequencies of various spectral bands were consid-
ered altogether. Data of solid state benzoate samples and data of
activity against E. coli were initially processed with PCA technique.
PCA however, produced principal components which did not really
distinguish studied compounds towards their spectral structure, as
their values were almost the same for every compound, which
eventually resulted in extremely high (enormous) correlation coef-
ficient. Contrary, PLS produced three-elemental model correlating
spectral data and the antimicrobial activity against E. coli with
Correlation coefficients found for S. cerevisiae with application
of spectral data of water solution samples and cross validation
diagnostic technique were statistically less significant (see
Table 5).
Statistically significant correlation coefficients obtained for
studied micro-organisms and compounds confirm the existence
of correlation between IR wavenumbers and biological activity.
Even small changes in chemical structure change biological prop-
erties, and all those changes are reflected in IR spectra. Spectral re-
gions involved in the model created contain stretches of symmetric
and asymmetric vibrations – of COOꢀ group as well as aromatic
ring vibrations. Wavenumbers of bands of anion or bands of ring
analyzed separately according to Lambert–Beer law did not pro-
duce good results. So that, whole molecule except cation is respon-
sible for biological activity of studied series of benzoates. Cation
insignificance is confirmed by lack of good correlation in case of so-
lid state samples.
t
relatively high correlation coefficient equal to 0.83 and
log(PRESS) = 3.12. Data of activity against remaining micro-organ-
isms did not produced statistically significant models.
Spectral data of water solution samples differ significantly from
data of solid samples (see Tables 1 and 2). Water solution data
seem to be more suitable for building a statistical model, as mi-
cro-organism incubation was conducted in water environment.
Various spectral regions were analyzed in order to find good corre-
lation, while applying two of them (1571–1528 and 1415–
1366 cmꢀ1) altogether within PLS-1 calibration produced high
correlation coefficients and relatively low PRESS values. Table 5
presents results obtained from cross validation diagnostics used
within PLS-1 calibration.
For three out of four studied micro-organisms statistically sig-
nificant correlation between actual and predicted degree of growth
Fig. 1. Predicted with PLS-1 model versus actual antimicrobial activity showed by
studied benzoates against B. subtilis after 24 h of incubation.

M.H. Borawska et al. / Journal of Molecular Structure 919 (2009) 284–289
289
PLS reduces the number of predictors to a set of uncorrelated
components and performs least squares regression on these com-
ponents. Band wavenumbers (predictors) used to construct the
PLS model were the same as for the models we have created before
for another set of compounds: asymmetric and symmetric
stretches of carboxylic anion, asymmetric deformations of carbox-
ylic anion and 8a and 19b ring vibrations [7]. The response was
antimicrobial activity. In Table 1, wavenumbers of 8a ring vibration
are in the first row, while wavenumbers of 19b ring vibration are in
the fourth row. This is of great importance that both previously and
in this work we are statistically analyzing wavenumbers of bands
assigned to the same vibrations. That confirms the proper selection
of bands originating from the same bonds, involved in the model.
Wavenumbers of those bands correlate with the antimicrobial
activity of compounds. Therefore, the claim of determining antimi-
crobial activity by molecular structure (characterized by spectral
parameters) is supported. However, as we have already stated,
wavenumbers of carboxylate anion or ring vibrations taken into
account separately did not produce satisfactory correlation. That
means structure of whole molecule, but not anion or ring alone
is of importance.
In the model proposed herein, validation with leave-one-out
procedure was performed. Obtained statistical model is valid for
current data, with correlation coefficient (R2) equal to 0.81, 0.70
and 0.76, for E. coli, B. subtilis and P. anomala, respectively. Compar-
ison of log(PRESS) values (see Table 5) allowed selection of stron-
gest model out of one-, two- and three-factor models. High
correlation coefficient confirm good correlation between spectral
and biological data (current data), while moderately strong predic-
tion strength is depicted by relatively high PRESS value. This weak-
ness of proposed models is probably a result of a small number of
data involved in the experiment, as well as error in relatively high
confidence intervals of biological data.
of factors involved in model calculated for different micro-organ-
isms suggests somewhat different mode of action against studied
micro-organisms. However, it is very likely that the most impor-
tant factor, as regards antimicrobial activity and mode of action,
is the ability of anion to coordinate hydrogen cations and there-
fore, at least partially, regulate pH value of the medium. In other
words, buffer capacity of the solution of studied benzoates,
which depends on both salt concentration and stability constants
of weak acid incorporated in the salt, which is in turn deter-
mined by electronic charge distribution, is the factor regulating
pH value of the medium. Therefore we suggest molecular struc-
ture, characterized by IR data, to be indirectly responsible for
antimicrobial properties of given compound.
Acknowledgement
This work was financially supported by Polish Ministry of Edu-
cation and Science (Grant No. 2PO5F 02428).
References
[1] P. Koczon´ , P. Mos´cibroda, K. Dobrosz-Teperek, W. Lewandowski, J. Mol. Struct.
565 (2001) 7–11.
[2] P. Koczon´ , W. Lewandowski, A.P. Mazurek, Vib. Spectrosc. 20 (1999) 143–149.
´
[3] W. Lewandowski, P. Koczon´ , K. Dobrosz-Teperek, R. Swisłocka, L. Fuks, W.
Priebe, A.P. Mazurek, Spectrochim. Acta A 59 (2003) 3411–3420.
´
[4] P. Koczon´ , T. Hrynaszkiewicz, R. Swisłocka, M. Samsonowicz, W. Lewandowski,
Vib. Spectrosc. 33 (2003) 215–222.
[5] P. Koczon´ , J.Cz. Dobrowolski, W. Lewandowski, A.P. Mazurek, J. Mol. Struct. 655
(2003) 89–95.
[6] P. Koczon´ , W. Lewandowski, E. Lipin´ ska, E. Sobczak, J. Agric. Food Chem. 49
(2001) 2982–2986.
´
[7] P. Koczon, J. Piekut, M. Borawska, W. Lewandowski, J. Mol. Struct. 651–653
(2003) 651–656.
´
´
[8] P. Koczon, J. Piekut, M. Borawska, R. Swisłocka, W. Lewandowski, Spectrochim.
Acta A 618 (2005) 1917–1922.
´
[9] P. Koczon´ , J. Piekut, M. Borawska, R. Swisłocka, W. Lewandowski, Anal. Bioanal.
Chem. 384 (2006) 302–308.
[10] P. Buning-Pfaue, Food Chem. 82 (2003) 107–115.
[11] E. Wüst, L. Rudzik, J. Mol. Struct. 661–662 (2003) 291–298.
[12] H. Büning-Pfaue, R. Hartmann, J. Harder, C. Urban, Fresenius J. Anal. Chem. 360
(1998) 832–835.
[13] C. Miralbes, Food Chem. 88 (2004) 621–628.
[14] R. Faggianhi, E. Howard-Lock, C.J.L. Lock, B. Lippert, B. Rosenberg, Can. J. Chem.
60 (1982) 529–534.
[15] C. Hansch, T. Fujita, J. Am. Chem. Soc. 86 (1964) 1616–1626.
[16] X. Cui, C. Joannou, M. Hughes, R. Cammack, FEMS Microbiol. Lett. 98 (1992)
67–70.
4. Conclusions
The structure of carboxylic anion of solid state samples of the
studied compounds is of different type: chelation structure in the
case of Mg benzoate, bridging structure in the case of both Ca
and Cu benzoates and monodentate structure in the case of Zn car-
boxylates which is proved by
D value obtained for every com-
pound. Electronic charge distribution within carboxylic anion is
therefore different in all studied molecules, and slight differences
occur in the aromatic ring.
[17] Y. Kima, S. Farrah, R.H. Baney, Int. J. Antimicrob. Agents 29 (2007) 217–222.
[18] J.L. Buchbinder, J. Baraniak, P.A. Frey, G.H. Reed, Biochemistry 51 (1993)
14111–14116.
[19] G. Kurisu, T. Kinoshita, A. Sugimoto, A. Nagara, Y. Kai, N. Kasai, S. Harado, J.
Biochem. 121 (1997) 304–308.
In the case of E. coli, for solid state samples statistical model cre-
ated with PLS technique produced good relation between spectral
data and antimicrobial activity. Comparable results had been
obtained in the past for another series of compounds and the same
micro-organism [14]. However, even though the present model fits
current data properly (R2 = 0.83), its predictive strength is only
moderately strong. Considering the lack of such relation for
remaining salts, we suggest that the defense system of E. coli might
be significantly different from other micro-organisms’ defense
systems, or the mode of action of the studied compounds against
E. coli involve different properties of the molecule.
In case of spectra of water solutions model with statistically
significant correlation coefficients and low log(PRESS) value
was obtained for three out of four studied micro-organisms.
Importance of both carboxylic anion and ring structure deter-
mining compound antimicrobial activity is confirmed by spectral
region selection. Bands generated from vibrations of both car-
boxylic anion and ring are located in the used spectral regions.
If other regions were analyzed alone or added to be analyzed
together with selected carboxylic anion and ring vibrations,
correlation coefficient dropped dramatically. Different number
´
[20] W. Lewandowski, H. Baranska, J. Raman Spectrosc. 17 (1991) 17–22.
[21] W. Lewandowski, M. Kalinowska, H. Lewandowska, J. Inorg. Biochem. 99
(2005) 1407–1423.
[22] N.J. Brownless, D.A. Edwards, M.F. Mahon, Inorg. Chim. Acta 28 (1999) 789–
794.
[23] A. Wilcks, N. Jayaswal, D. Lereclus, L. Andrup, Microbiology 144 (1998) 1263–
1270.
[24] K. Bajdor, P. Koczon´ , E. Wie˛ckowska, W. Lewandowski, Int. J. Quant. Chem. 4
(1997) 385–392.
[25] G. Varsanyi, Assignments for Vibrational Spectra of 700 Benzene Derivatives,
Akademiai Kiado, Budapest, 1973.
[26] T. Ishioka, Y. Shibata, M. Takahashi, I. Kanesaka, Y. Kitagawa, K.T. Nakamura,
Spectrochim. Acta A 54 (1998) 1827–1835.
[27] P. Persson, M. Karlsson, L. Ohman, Geochim. Cosmochim. Acta 62 (1999) 3657–
3668.
[28] G.B. Deacon, R.J. Phillips, Eur. J. Inorg. Chem. 11 (2002) 2965–2974.
[29] K. Nakamoto, Infrared and Raman Spectra of Inorganic and Coordination
Compounds, Wiley, New York, 1986.
[30] J.D. Kubicki, G.A. Blake, S.E. Apitz, Geochim. Cosmochim. Acta 60 (1996) 4897–
4911.
[31] R. Quilitzsch, M. Baranska, H. Schulz, E. Hoberg, J. App. Bot. Food Qual. 79
(2005) 163–167.
[32] E. Wust, L. Rudzik, J. Mol. Struct. 661–662 (2003) 291–298.
[33] S. Jacobsena, I. Søndergaardb, B. Møllera, T. Deslerc, L. Munck, J. Cer. Sci. 42
(2005) 281–299.
[34] B. Pavoni, N. Rado, R. Piazza, S. Frignani, Ann. Chim. 94 (2004) 521–553.