Available online at www.sciencedirect.

com

Procedia Engineering 47 (2012) 522 – 525

Proc. Eurosensors XXVI, September 9-12, 2012, Kraków, Poland

The application of an array of sensors based on boronic acid

derivative for the quantitative analysis of amino acids
A. Kutyáa-Olesiuka, M. JaĔczyka, X. Cetób, M. del Valleb,
P. Cioseka, W. Wróblewskiaa*
a
Department of Microbioanalytics, Faculty of Chemistry, Warsaw University of Technology, Noakowskiego 3, 00-664 Warsaw,
Poland
b
Grup de Sensors i Biosensors, Dept. Química, Universitat Autònoma de Barcelona, Edifici Cn, 08193 Bellaterra, Barcelona, Spain

Abstract

The objective of this work was to perform the quantitative analysis of 4 amino acids: phenylalanine (Phe), tyrosine
(Tyr), ornithine (Ort) and glutamic acid (Glu) in mixtures with the use of potentiometric sensor array (electronic
tongue) and Sequential Injection Analysis (SIA) measurement system. The flow-through sensor array was composed
of 6 miniaturized classical ion-selective electrodes based on polymeric membranes (plasticized PVC) containing
phenylboronic acid ionophore and/or an ion exchanger. The PLS analysis of the sensor array responses, measured in
amino acids mixtures prepared by the SIA system (in two buffer solutions), permitted a correct quantitative analysis
of only Glu and Ort. Further chemometric treatment, involving the extraction of dynamic components of the transient
response employing the Fourier transform plus training of an Artificial Neural Network (ANN) with its coefficients
led to the simultaneous determination of 3 amino acids, with a separate semiquantitative estimation of Tyr.

©2012
© 2012ThePublished by Elsevier
Authors. Published Ltd. Ltd. Selection and/or peer-review under responsibility of the Symposium Cracoviense
by Elsevier
Sp. z.o.o. Open access under CC BY-NC-ND license.
Keywords: amino acids analysis; boronic acids; ion-selective electrodes; automated electronic tongue; sequential injection analysis.

1. Introduction

Specific inherited inborn errors in amino acid biosynthesis have been recognized as causal in many
metabolism disorder. Standard metabolic diseases are detected through genetic testing which involve the
analysis of mutations in the genes coding for enzymes production or by quantitative analysis of amino

* Corresponding author. Tel.: +48 22 234 5631; fax: +48 22 234 5631.
E-mail address: wuwu@ch.pw.edu.pl

1877-7058 © 2012 The Authors. Published by Elsevier Ltd. Selection and/or peer-review under responsibility of the Symposium Cracoviense
Sp. z.o.o. Open access under CC BY-NC-ND license.
doi:10.1016/j.proeng.2012.09.199
 A. Kutyła-Olesiuk et al. / Procedia Engineering 47 (2012) 522 – 525 523

acids or their metabolites in physiological fluids. Therefore the determination of physiological amino
acids using HPLC, GC/MS, HPLC/MS or tandem mass spectrometry, provides rapid diagnosis of
metabolic diseases and the introduction of proper treatment [1]. The concentration of amino acids is
examined mainly in plasma and in a healthy person is on the level of 10-4 M (in blood plasma) [2].
An alternative, simple method of the amino acids determination involving an array of ion-selective
electrodes based on phenylboronic acid ionophore and a chemometric data processing tool was reported in
this work. Boronic acids derivatives are widely used as receptors of different bioanalytes i.e. diols, anions,
neurotransmitters and also amino acids [3]. Taking into account the metabolic pathways as well as the
relationship between the level of amino acids in blood and several diseases, the quantitative analysis of 4
amino acids: phenylalanine (Phe), tyrosine (Tyr), ornithine (Ort) and glutamic acid (Glu) was performed.

2. Experimental

The flow-through sensor array embraced: 4 miniaturized electrodes with PVC membranes containing
4-octyloxyphenylboronic acid and a lipophilic salt TDMAC or KTFPB and 2 electrodes containing only
an ion-exchanger in membrane. The method of the membranes preparation and electrodes conditioning
were the same as for the standard ISEs. The membranes contained the ionophore (2 wt%), 10 mol%
versus ionophore lipophilic salt, 65-66 wt% plasticizer, and 31-33 wt% high-molecular-weight PVC (the
membranes based on an ion-exchanger contained 3 wt% of TDMAC or 1 wt% KTFPB). The membrane
components (200 mg in total) were dissolved in 2ml of THF. A detailed architecture of the miniaturized
ion-selective electrodes compatible with a flow-through module was presented in [4]. The constructed
ISEs were preconditioned in a dilute solution of internal electrolyte (0,01M NaCl) for 24 hours.
The flow-through sensor array was connected to the SIA system providing the automated operation
and generation of amino acid sample mixtures, thanks to the precise dosing and mixing of volumes of
stock solutions. An 8-channel signal conditioning circuit connected to the National Instruments
Multifunction DAQ analogue inputs (Model NI6221, USA) was used for the EMF measurements. The
whole system was controlled by a PC using a virtual instrument developed in Labview.
All measurements were carried out in cells of the following type: Ag, AgCl; KCl 3MūCH3COOLi
1Mūsample solution̴membrane̴internal filling solution; AgCl, Ag.
Data analysis was performed in MatLab 7.1 (The MathWorks, Inc., Natick, USA) with specific
routines written by the authors using its Neural Network Toolbox (v4.0.6) and Origin (Microcal Software,
Inc, Northampton, USA) software. Chemical images of samples were processed using Partial Least
Squares – Discriminant Analysis (PLS-DA). For the dynamic treatment, the transient response of each
sensor was first compressed employing Fast Fourier Transform (FFT), and then extracted coefficients
were used as inputs of the ANN model which carried out the quantification of the amino acids [5].

3. Results and discussion

The electrodes constructed by incorporating the phenylboronic acid within the PVC/DOS membrane
containing anionic additives (KTFPB) were found to respond potentiometrically (Fig 1a). The highest
sensitivity was recorded for Phe and Glu, whereas the values of the response slope for Ort and Tyr did not
exceed 10 mV/decade. Moreover, the slope values depended on the pH of the sample solution – the acidic
conditions promoted higher electrode sensitivity of the electrodes based on KTFPB additive. Surprisingly,
the electrodes formulated with 4-octyloxyphenylboronic acid and cationic additive (TDMAC) displayed
completely different performances (Fig 1b). Depending on the experimental pH conditions, anionic
(pH=4) or cationic (pH=9) responses with slopes <10 mV/decade in quite a narrow linear range were
determined towards amino acids (except phenylalanine responses in pH=4).
524 A. Kutyła-Olesiuk et al. / Procedia Engineering 47 (2012) 522 – 525

200 200
a) b)
Phe pH=4 KTFPB

150
150

Glu
EMF [mV]

EMF [mV]
pH=9 KTFPB
100
100

Ort 50
Tyr pH=9 TDMAC
50
pH=4 TDMAC
0
-6 -5 -4 -3 -2 -1 -6 -5 -4 -3 -2 -1

log c AA log a Phe

Fig. 1. (a) potentiometric responses of ISEs based on PVC/DOS membranes with 4-octyloxyphenylboronic acid (10% mol of
KTFPB) towards given amino acids at pH=4; (b) potentiometric responses of ISEs based on PVC/DOS membranes with 4-
octyloxyphenylboronic acid (10% mol of TDMAC or KTFPB) towards phenylalanine at various pH.

The measurements of the sensor array signals were conducted in 80 sample mixtures containing 4
amino acids in the concentration range of 0,1 – 3 mM (except tyrosine: 0,05 – 1,5 mM) prepared by the
Sequential Injection Analysis system in 2 carrier solutions of different pH controlled by 1 mM MES
(pH=5) and TRIS (pH=9) buffers. Data matrices involving the signals of the sensors were formed – every
sample was characterized by 12 variables (6 sensors signals obtained in 2 carrier solutions). The 80x12
data matrix was divided into the train (64x12) and the test (16x12) sets, serving as inputs for Partial Least
Squares (PLS) models.
The model performance was characterized after linear fitting of the real data (real concentrations of
every amino acid) to the predicted data (PLS predicted concentrations of every amino acid). Values of
slope (“a”), intercept (“b”) and determination coefficient (“R2”), were calculated for the test samples and
train samples. The correct quantitative analysis (for train and test sets) was achieved only for Glu and Ort,
(Fig 2), whereas for Phe and Tyr, the values of a, b, and R2 were not acceptable.

3 3
Train Ɣ Train Ɣ
y = 0,9656x + 2·10-5 ; R2 = 0,9656 y = 0,967x + 4·10-5 ; R2 = 0,967
2,5 Test ż 2,5 Test ż
y = 0,7855x + 3·10-4 ; R2 = 0,8984 y = 0,9789x + 2·10-5 ; R2 = 0,9569
2 2
PRED C/mM

PRED C/mM

1,5 1,5

1 1

0,5 0,5

Ort Glu
0 0
0 0,5 1 1,5 2 2,5 3 0 0,5 1 1,5 2 2,5 3
REAL C/mM REAL C/mM

Fig. 2. Model performance characterized after linear fitting of the real data (real concentrations of amino acid) to the predicted data
(PLS predicted concentrations of aminoacid). Training subset (Ɣ, solid line), testing subset (ż, dotted line).
 A. Kutyła-Olesiuk et al. / Procedia Engineering 47 (2012) 522 – 525 525

Since the application of the steady-state response of each sensor was not sufficient to build a model
capable of carrying out the quantification of the four amino acids, a more complete data treatment
involving the use of the dynamic component of the signal was attempted. For this, the transient response
of each sensor was first compressed employing the Fast Fourier Transform, which allowed the extraction
of some coefficients which were used as inputs to the ANN model. As before, the same data division was
performed for the train and test samples, evaluating the model performance after linear fitting of the real
concentrations vs. the predicted ones by the ANN model, both for the train and test samples (Fig 3). In
this way, the introduction of the dynamic component provided the correct quantitative analysis of the
three amino acids (i.e. Glu, Ort and Phe), while the concentration of tyrosine could be predicted
semiquantitatively in an another step.

Fig. 3. Model performance characterized after linear fitting of the real concentrations of the three amino acids to the predicted data
for the FFT-ANN treatment. Training subset (Ɣ, solid line), testing subset (ż, dotted line) and theoretical diagonal line (dashed line).

Acknowledgements

This work has been financially supported by the project LIDER/17/202/L-1/09/NCBiR/2010 and by
the project CTQ2010-17099. X. Cetó thanks the support of Dept. d'Innovació, Universitats i Empresa de
la Generalitat de Catalunya for the predoctoral grant.

References

[1] Murray RK, Granner DK, Mayes PA, Rodwell VW. Harper’s Illustrated Biochemistry. 26th ed. McGraw-Hill, 2003.
[2] Abdulrazzaq YM, Ibrahim A. Determination of aminoacids by ion-exchange chromatography on filter paper blood samples
stored at different temperatures and for different periods: comparison with capillary and venous blood. Clin Biochem
2001;34:399-406
[3] Mohler LK. Czarnik AW. D-amino acid chelative complexation by an arylboronic acid. J Am Chem Soc 1993;15:7037-8.
[4] Witkowska E, Buczkowska A, Zamojska A, Szewczyk KW, Ciosek P. Monitoring of periodic anaerobic digestion with
flow-through array of miniaturized ion-selective electrodes. Bioelectrochemistry 2010;80:87-93.
[5] Calvo D, Durán A, del Valle M. Use of pulse transient response as input information for an automated SIA electronic
tongue. Sensor Actuat B-Chem 2005;131:77-84.