Sensing of carboxylate drugs in urine by a supramolecular sensor array

Article abstract of DOI:10.1021/ja4015748

A supramolecular sensor array consisting of eight chemosensors embedded in a hydrogel matrix was used to sense carboxylate drugs. The discriminatory power of the array has been evaluated using principal component analysis and linear discriminant analysis. The eight-member sensor array has been shown to accurately identify 14 carboxylates in water with 100% classification accuracy. To demonstrate the potential for practical utility in the physiological environment, analysis of carboxylate drugs in human urine was also performed achieving 100% correct classification. In addition, the array performance in semiquantitative identification of nonsteroidal anti-inflammatory drugs has been investigated, and the results show that the sensor array is able to differentiate six typical nonsteroidal anti-inflammatory drugs at concentrations of 0.5-100 ppm. This illustrates the potential utility of the designed sensor array for diagnostic and environmental monitoring applications.

Full text of DOI:10.1021/ja4015748

Subscriber access provided by Otterbein University
Article
Sensing of Carboxylate Drugs in Urine by a Supramolecular Sensor Array
Yuanli Liu, Tsuyoshi Minami, Ryuhei Nishiyabu, Zhuo Wang, and Pavel Anzenbacher
J. Am. Chem. Soc., Just Accepted Manuscript • DOI: 10.1021/ja4015748 • Publication Date (Web): 23 Apr 2013
Downloaded from http://pubs.acs.org on April 27, 2013
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of the American Chemical Society is published by the American Chemical
Society. 1155 Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 8
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
Sensing of Carboxylate Drugs in Urine by a Supramolecular
Sensor Array
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Yuanli Liu,† Tsuyoshi Minami,† Ryuhei Nishiyabu,‡ Zhuo Wang,† and Pavel Anzenbacher, Jr.†,*
†Department of Chemistry and Center for Photochemical Sciences, Bowling Green State University, Bowling Green, Ohio, 43403,
USA.
‡Department of Applied Chemistry, Graduate School of Urban Environmental Sciences, Tokyo Metropolitan University, Tokyo 192-
0397, Japan.
ABSTRACT: A supramolecular sensor array consisting of eight chemosensors embedded in a hydrogel matrix was used to sense carboxylate
drugs. Discriminatory power of the array has been evaluated using the principal component analysis and linear discriminant analysis. The
eight-member sensor array has shown to accurately identify fourteen carboxylates in water with 100% classification accuracy. To demonstrate
the potential for practical utility in physiological environment, analysis of carboxylate drugs in human urine was also performed achieving
100% correct classification. In addition, the array performance in semi-quantitative identification of non-steroidal anti-inflammatory drugs
(NSAIDs) has been investigated, and the results show that the sensors array is able to differentiate six typical non-steroidal anti-inflammatory
drugs at concentrations 0.5 - 100 ppm. This illustrates the potential utility of the designed sensor array for diagnostic and environmental mon-
itoring applications.
INTRODUCTOIN
Carboxylates are important anions frequently encountered in Na-
ture as well as in a number of biological processes. Their utility in
chemical, pharmaceutical, food and beverages industry is wide-
spread.1 A number of drugs contain carboxylate function —notably
non-steroidal anti-inflammatory drugs (NSAIDs).2 Due to their
extensive use, these drugs also present a significant environmental
burden.3 For this study, we selected a group of carboxylates includ-
ing antimalarial artesunate,4 and well-known NSAIDs (ibuprofen,
naproxen, diclofenac, flurbiprofen, ketoprofen, mefenamic acid)
commonly used to relieve pain, inflammation, and fever.2 Also in-
cluded were aminoacids (alanine, tyrosine, sarcosine) and small-
molecule carboxylates (mevalonate, thyroxine) known to play an
important role in human metabolism. This is because sarcosine is a
potential biomarker for human prostate cancer,5 while mevalonic
acid is an intermediate in steroids biosynthesis.6 Tyrosine7 is a pre-
cursor of catecholamine neurotransmitters and hormone thyroxine.
Current detection methods for carboxylates generally utilize
solid-phase extraction (SPE) pre-concentration while the analysis
of the concentrated sample is typically performed by liquid chro-
matography–mass spectrometry,8 or gas chromatography–mass
Figure 1. Molecular structures of target carboxylates D1-D14 (Top)
and S1-S8 used for the sensor array (Bottom).
spectrometry.9 These methods, however, are not easily amenable to
determination of carboxylates in biological fluids.
Previously, we demonstrated that calixpyrrole sensors doped
into polyurethane films could be used for sensing of aqueous car-
Recently, chemosensors possessing binding sites for carboxylates
have been investigated10 including sensors based on cross-reactive
boxylates. The preliminary experiments showed colorimetric sens-
arrays inspired by the mammalian olfactory system.11 The increased
ing of three carboxylates of medical interest.14 Inspired by this work
popularity of array-based sensors is largely due to their capability to
we developed a fluorescence-based sensor array capable of differen-
recognize a number of analytes with high classification accuracy.12
tiation of fourteen pharmaceutically and biologically important
However, few sensors array exists for carboxylate anions that func-
carboxylates (Figure 1, Top) in both water and urine with high
tion in aqueous solution.13 To the best of our knowledge, sensor
classification accuracy. The array is prepared by casting a solution
arrays capable of sensing carboxylate anions in complex biological
of polyurethane (PU) and octamethylcalix[4]pyrrole sensor (Fig-
milieu, such as human urine, have not yet been developed.
ure 1, Bottom) into a multi-well microtiter plate.
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 2 of 8
Table 1 Binding constants (M-1) derived from titrations using sensors S1-S8 and anions.
1
2
Anions
Sensors
3
4
5
6
7
8
9
S1
S2
S3
S4
S5
S6
S7
S8
ND
ND
F-
Cl-
>1.0 × 108
3.5 × 105
2.6 × 106
2.1 × 106
ND
2.1 × 106
1.0 × 105
4.2 × 105
6.5 × 104
3.0 × 105
1.4 × 105
2.5 × 106
1.2 × 105
4.1 × 105
8.1 × 104
ND
1.0 × 107
2.9 × 105
5.8 × 105
5.8 × 104
4.8 × 105
3.1 × 105
1.0 × 107
2.9 × 105
1.1 × 106
6.2 × 104
ND
1.6 × 106
1.1 × 105
2.8 × 105
4.1 × 104
2.5 × 105
1.0 × 105
1.6 × 105
1.8 × 104
9.9 × 104
3.6 × 104
5.4 × 104
1.8 × 105
AcO-
1.6 × 104
4.3 × 105
9.8 × 104
ND
-
H2PO4
HPPi3-
BzO-
3.0 × 105
1.4 × 105
4.4 × 105
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Binding constants were determined in MeCN (S1-S7) and DMSO (S8), respectively using anions in the form of their tetra-n-butylammonium salts.
ND means not determined due to low affinity or biphasic nature of the isotherm. All errors are < 15%.
Additionally, this new chemosensor design utilizes both colori-
metric and fluorimetric responses, which yields information-rich
output useful for discrimination of analytes. The utility of this dual
signal transduction scheme is demonstrated in semi-quantitative
identification of NSAIDs over a wide range of concentration.
nonradiative dissipation of the excited state energy thereby increas-
ing the fluorescence (turn-on signal). This, together with IPCT
effect results in an information rich output by S8. Also, S8-anion
complexes are likely to display C3 symmetry, which is more com-
plementary to phosphate anion. As a result, S8 displays preference
for phosphate.
RESULTS AND DISCUSSION
The binding affinities of S1-S8 for anions are shown in Table 1,
which shows relatively high binding affinities of S1-S7 toward hal-
ide and acetate anions over benzoate. The selection of the test
group of anions was made using anions known to be bound by
calix[4]pyrrole receptors.15,17 S8, which does not show a significant
response to halides, was expected to be an important factor in the
analyte recognition by the array. This hypothesis was confirmed by
the analysis of the contribution of the individual sensors.
The sensors S1-S7 utilize the octamethylcalix[4]pyrrole (OMCP)15
receptor for anions including carboxylates, which is an excellent
structural platform for preparation of anion sensors. The infor-
mation-rich signal output generated by S1-S7 arises from attaching
different chromophores to OMCP. One of the OMCP pyrroles
communicates with an electron-withdrawing residue attached
through a vinyl bridge.
To visualize the fingerprint-like response pattern of the S1-S8 ar-
ray to individual analytes we show an image corresponding to a
preliminary experiment comprising fluorescence recorded using
three channels (blue, green, and red). Even by naked eye one can
see that the addition of aqueous solutions of carboxylates (D1-
D14) resulted in fingerprint-like fluorescence response pattern
(Figure 3). As expected, each sensor in the array generated a dis-
tinct change in the fluorescence imparted by the different carboxy-
lates. This response pattern can be analyzed using methods of pat-
tern recognition to achieve analytes discrimination.
EWG
Figure 2. Intramolecular partial charge transfer (IPCT) in the sensors
S1-S7 results in anion-induced in changes in fluorescence and color.
Control
D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D11 D12 D13 D14
This structural arrangement establishes an intramolecular partial
charge transfer (IPCT) cascade (Figure 2): As the anion attracts
the proton involved in hydrogen bonding, the electronic density of
the H-N bond shifts toward the pyrrole nitrogen and causes further
polarization of the pyrrole electronic cloud. An acceptor (electron-
withdrawing moiety, EWG) attached through a conjugated bridge
accommodates the excess partial charge, thus completing the par-
tial charge transfer (IPCT) cascade.10d,f, 15b The IPCT results in ani-
on binding-induced change in fluorescence or color. 15b,c, 16 Sensor
S8, a tripodal turn-on fluorescent sensor prepared from (2,4,6-
triethyl-1,3,5-trimethylamino)benzene,12e shows selectivity for
aliphatic carboxylates and phosphates (Table 1), and was included
in the array to increase its signal variability. The sensor S8 is flexible
in the resting state but forms a stable bowl-shape complex with the
anions. The increased rigidity of the complex results in limited
S1
S2
S3
S4
S5
S6
S7
S8
Figure 3. Fluorescence responses of the S1-S8 sensor array to the pres-
ence of carboxylates D1-D14 (1 mM in water at pH 8.5, 200 nL). The
color representation was generated by superimposing equal weighed
images corresponding to RGB channels.
ACS Paragon Plus Environment

Page 3 of 8
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Figure 4. (Top) Pattern generated by S1-S8 sensors array using fifteen fluorescence and colorimetric channels in response to carboxylates D1-D14
in water (500 µM, 200 nL, pH 8.5). (Bottom) Response profile of S1-S8 sensor array upon addition of ibuprofen, D4, (500 µM, 200 nL, pH 8.5).
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Encouraged by the apparent information-rich response pattern,
we attempted to discriminate the fourteen carboxylates (D1-D14)
in aqueous solutions by the S1-S8 sensor array. We have recorded
the information from both the fluorimetric (eleven channels) and
colorimetric (four channels) images. Figure 4 (top panel) shows
the overall response pattern of the sensor array in the presence of
carboxylates. As hypothesized, each carboxylate induced a distinc-
tive change in the individual sensor. Figure 4 (bottom) shows a
response of sensors S1-S8 to ibuprofen (D4) with characteristic
signal increases and decreases (relative to the control). This illus-
trates the assertion that a multidimensional response pattern matrix
is created by compiling the (D1-D14) × (S1-S8) response data,
which are then utilized in a pattern recognition investigation.
The sensor array consisting of eight sensors generates a signal
output in the form of a multidimensional response (120 dimen-
sions = 15 channels × 8 sensors). The signal output comprises both
fluorimetric and colorimetric data.18 The array response was evalu-
ated by utilizing the statistical multivariate analysis method - prin-
cipal component analysis (PCA).19 Here, the PCA of the dataset
(10 repetitions for each carboxylate) acquired from the eight-
member sensor array requires 15 dimensions (PCs) out of 120 to
describe 95% of the discriminatory range (12% of all PCs). This
attests to an extraordinarily high degree of dispersion of the data
obtained by the S1-S8 sensor array. This discrimination capacity is
unusually high in comparison with that reported for a sensor array
generally displaying 95% of discrimination in the first two PCs.11a,c
Here, the PCA score plot (Figure 5) shows clear clustering of the
data using only the first three PCs (representing 69.4% of variance).
The high dispersion level of the data shown by the PCA score plot
L-Tyrosine
Sarcosine
8
Ibuprofen
Ketoprofen
Ritalinic Acid
4
L-Alanine
Mevalonic Acid
Mefenamic Acid
0
Flurbiprofen
Salicylic Acid
Diclofenac
-4
Naproxen
Artesunate
L-Thyroxine
-8
Control
-12
-6
12
0
6
6
0
-6
12
-12
Figure 5. PCA score plot of the first three principal components of sta-
tistical significance for 150 samples (200 nL, 500 µM in water, pH 8.5,
14 carboxylates plus a control, 10 trials each) produced by the S1-S8
sensors array.
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 4 of 8
1
2
3
4
5
6
7
8
L-Tyrosine
Sarcosine
Ibuprofen
PCA
PC1
PC2
PC3
50
S1 S2 S3 S4
Eigenvalue
Proportion
46.09
38.40%
24.77
20.70%
9.58%
11.04%
13.26%
9.72%
12.43
10.40%
30.18%
12.76%
6.65%
Ketoprofen
Ritalinic Acid
L-Alanine
Mevalonic Acid
Mefenamic Acid
Flurbiprofen
Salicylic Acid
Diclofenac
Naproxen
S5 S6 S7 S8
S1
S2
S3
S4
S5
S6
S7
S8
8.42%
12.91%
15.62%
15.86%
16.12%
8.07%
0
6.42%
Artesunate
L-Thyroxine
Control
9.81%
10.04%
17.21%
6.75%
Excluded S2
S3 S4
12.75%
14.35%
19.49%
9
14.82%
8.18%
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
-260
80
-480
0
100
9.99%
F1(64.2%)
PCA
PC1
PC2
PC3
Eigenvalue
Proportion
16.06
35.70%
28.91%
10.62
23.6%
30.98%
5.72
12.70%
48.83%
S1
S2
S3
S4
S5
S6
S7
S8
60
0
S1 S5 S8
-
-
-
-
-
-
-
-
-
46.90%
29.06%
16.71%
-20
-40
-
-
-
-
-
-
Excluded
S1
24.19%
39.96%
34.46%
-200
0
50
F1(75.8%)
PCA
PC1
PC2
PC3
40
Eigenvalue
Proportion
11.99
40.00%
-
-
-
-
7.69
25.70%
-
-
-
-
3.15
10.50%
-
-
-
-
S5 S8
S1
S2
S3
S4
S5
S6
S7
S8
0
LDA Jackknified Classification
66.26%
26.63%
44.02%
Number of sample
Correct Number
150
150
-
-
-
-
-
-
-20
Correct Proportion 100%
30.74%
60.67%
55.98%
-40
-20
0
160
F1(80.3%)
Figure 6. Schematic representation of the investigation of the most important contributors to the overall statistical significance: (Top) PCA for the
complete set of sensors (S1-S8) shows that the main contributors to the cluster dispersion are S1, S5, and S8. (Center) Sensors S2, S3, S4, S6 and S7
were excluded from the data set and the remaining data were analysed again with PCA. PCA shows that the main contributors were S5 and S8. (Bot-
tom) S1 was excluded and LDA was carried out using the remaining data set. Cross-validated LDA shows 100% accurate classification for all three
arrays.
can be attributed to the differences in response of the IPCT-based
sensors to the carboxylates. The receptors also display significant
cross-reactivity to the carboxylates, which enables the sensors to
respond to a wide variety of carboxylate analytes. Generally speak-
ing, it is the synergistic effect of the selective yet cross-reactive fea-
ture of the chemosensors that provides the good resolution (sepa-
ration) of the clusters (carboxylates) in the PCA score plot.
SI). Here, three factors describe 89.4% of the total information
(variance) contained in the dataset. This graphical representation
shows clusters of similar data and demonstrates the quality and
predictability of the output provided by the sensor array. The cross-
validation routine shows 100% accuracy for the classification of all
carboxylates (see SI).
Part of the motivation of this work was to establish the correla-
tion between the structural features of sensors and the discrimina-
tory power of the sensors array, an effort that could provide impor-
tant information for developing an effective analytical device for
carboxylate anions in physiological milieu. Toward this end, we
determined which sensors contribute most to the discriminatory
capacity. The screening of sensors was accomplished by excluding
certain sensors utilizing a sequence of PCA followed by LDA.12h
In addition to the PCA, the multidimensional response pattern
was further evaluated by linear discriminant analysis (LDA)19 to
explore the discriminatory power of the sensors array. LDA is a
statistical approach widely used for classification using a cross-
validation (leave-one-out) routine to assess the overall ability to
correctly classify the observations. The LDA graphical output
shows canonical score plots for the first three canonical factors (see
ACS Paragon Plus Environment

Page 5 of 8
Journal of the American Chemical Society
This way the sensors contributing most to the discriminatory
power (as judged by PCA) remain in the array while the less con-
1
2
3
4
5
6
7
8
L-Tyrosine
Sarcosine
Ibuprofen
tributing sensors are excluded (Figure 6). An ideal array should
comprise the lowest number of sensors that allow maintaining the
100% correct classification accuracy. Those should be the sensors
contributing most to the individual principal component with sta-
tistical significance,20 as estimated from the sensor contribution to
the principal component. This contribution can be evaluated from
the factor loadings which correlate to the cosine of the angle be-
tween the original variable and principal component axis.
Ketoprofen
Ritalinic Acid
L-Alanine
Mevalonic Acid
Mefenamic Acid
Flurbiprofen
Salicylic Acid
Diclofenac
Naproxen
40
0
Artesunate
L-Thyroxine
Control
-40
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
120
80
Figure 6 shows the screening process for sensors. First, PCA is
used to identify the three sensors (S1, S5, S8) with the highest con-
tribution to statistically significant principal components (PC1,
PC2, PC3). The selection of S1, S5, and S8 makes sense from the
supramolecular chemistry perspective as well because S1 and S2
display the highest affinity for acetate (as a model for aliphatic car-
boxylates, Ka~106 M-1, Table 1) while S8 displays the overall high-
est selectivity (binds acetate but not benzoate, which is probably
too large to fit well in the binding cavity of the tripodal receptor of
S8). It is quite likely that the sensors with the highest affinity and
selectivity would have the highest impact on the variance within the
response data. We believe that it is the complementarity between
S5 and S8 that increases the discriminatory capacity. We find it
reassuring that our understanding of supramolecular chemistry
principles and the pattern recognition methods arrive at the same
conclusion. The LDA confirmed that the selection of the sensors
S1, S5, and S8 did not result in compromised classification accuracy.
60
40
0
0
-40
Figure 7. LDA canonical score plots for the response of S1-S8 sensors
-60
array to fourteen carboxylates in urine. The first 3 factors were used in
order to describe 80% of the total variance. The cross-validation rou-
tine shows 100 % correct classification.
accomplish the anion sensing in such a complex medium requires
highly responsive sensing elements capable of distinguishing
among a large number of similar analytes.
To obtain maximum discriminatory information, we also re-
corded the response from both the fluorimetric and colorimetric
channels (combined 15 channels were recorded). Once again, mul-
tivariate statistical methods PCA and LDA were used to evaluate
the response pattern. The PCA of the dataset requires 20 dimen-
sions (PCs) out of 120 to describe 95% of the discriminatory range
(which corresponds to mere 17% of all PCs). The PCA score plot
(see SI) shows clear clustering of the data with the first three PCs
(representing 65.8% of variance). The high dispersion level of the
data shown by the PCA score plot reflects the fact that the eight-
member sensor array possesses high discriminatory capacity to the
carboxylic drugs in urine. Similarly, the LDA canonical score plot
(Figure 7) shows clear clustering of the data with the first three
factors describing 82.0% of the total information (variance) con-
tained in the dataset. LDA cross-validation routine shows 100%
classification accuracy for the fourteen carboxylates D1-D14 in
urine. This confirms that the S1-S8 sensor array possesses very high
discriminatory capacity even in highly competitive milieu.
Further exploring reveals that S5 and S8 are the most important
contributors to discriminatory power of the array. LDA cross-
validation routine demonstrated that the two sensors still provide
100% correct classification even though the response space is not
large as with the eight sensors (Figure 6).
While from the array fabrication perspective it is convenient to
decrease the number of the sensors in the array, from the view of
need to record and work with a large amount of data from multiple
channels one may be interested in learning about the lowest num-
ber of channels required for 100% correct classification of the four-
teen analytes (D1-D14). Toward this end we performed a similar
analysis as above aimed at evaluation of detection channels. We
learned that when we use all eight sensors, we actually need only
four channels to achieve 100% correct classification (for the LDA
results, see SI).
Due to the significance of NSAIDs we decided to demonstrate
the potential for quantitative sensing using the present method.
Ibuprofen / H2O
0.15
Diclofenac / H2O
Finally, we also attempted the reduction of both the number of
sensors in the array as well as the number of channels recorded.
100% correct classification was achieved with four sensors utilizing
output in four channels. This corresponds to reduction of the orig-
inal dataset by 87% (only 13% of the original dataset was used to
recognize the fourteen carboxylates D1-D14). This illustrates ex-
cellent recognition capabilities of the sensors and suggests an ap-
proach one may use to reduce this method to practice.
0.10
0.12
0.08
0.05
0.04
The detection of carboxylates in biological fluids such as human
urine is of clinical importance since the levels of drugs reflect the
state of human health.21 However, to detect carboxylates in urine is
a challenging problem as urine is a highly competitive medium that
contains high concentrations of electrolytes such as chloride, phos-
phates, and carboxylates as well as a large number of proteins. To
0.00
0.00
0
1
2
3
4
0
10
20
30
40
50
Concentration (ppm)
Figure 8. Concentration analyses of ibuprofen and diclofenac. The
inset in the graph shows the linear region in the concentration range of
0- 4 ppm. Analysis shows a limit of detection LOD ~ 0.1 ppm.
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 6 of 8
Two analyses were performed: First, a quantitative analysis of ibu- the same point (0 analyte concentration, control). This is also a
1
2
3
4
5
6
7
8
profen and diclofenac was performed using the PU : S3 film (Figure
8). Second, to illustrate the ability of the sensor array (S1-S8) to
recognize multiple NSAID carboxylates, a semi-quantitative analy-
sis was applied to six carboxylates: Diclofenac, flurbiprofen, ibupro-
fen, salicylic acid, ketoprofen, and naproxen.
reason why at low concentrations the data points appear to be close.
This is reasonable considering the isotherms shown in Figure 8.
Importantly, however, all concentrations for all six NSAIDs are
resolved (separated). It could be argued that one may not need an
array sensor for six different analytes; indeed, six NSAIDs in a semi-
quantitative analysis was, in fact, a stress test rather than a realistic
assignment. The significance of this result is in the proof of simul-
taneous quantitative sensing of anionic analytes utilizing an array-
based approach at low concentration, which is a rare achievement.
Figure 8 shows quantitative analyses of ibuprofen and diclofenac.
The overall response isotherm displays saturation behavior and a
linear portion at low analyte concentrations. The inset shows the
linear response to ibuprofen and diclofenac in the concentration
range of 0-4 ppm. Ibuprofen and diclofenac analysis suggests a limit
of detection LOD ~0.1 ppm. The fact that the PU : S3 shows a
reasonably low LOD attests to the potential applicability of this
approach.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
CONCLUSIONS
This study demonstrates that the eight hydrogen-bonding based
chemosensors embedded in hydrophilic polyurethane films can be
used in the sensors array for the detection of carboxylates in water
or in urine. The seven sensors (S1-S7) utilize a common receptor,
calix[4]pyrrole, attached to a chromophore via a conjugated moiety
thereby establishing an intramolecular partial charge transfer
(IPCT) chromophore that yields a strong fluorescence and color
change in the presence of anion. Another tripodal sensor (S8) was
included in the array to increase the selectivity of the array and
variance in the output datasets. Pattern recognition methods (PCA,
LDA) were employed to evaluate the sensing performance of the
array. The fourteen carboxylates were detected in water with 100%
classification accuracy. To demonstrate the practical utility, the
100% correct classification of carboxylates was performed in hu-
man urine. Finally, the simultaneous semi-quantitative analysis for
six NSAIDs was performed in the wide range concentrations (0.5-
100 ppm). It was demonstrated that the array responds to the pres-
ence of each NSAID analyte by an analyte-unique response which
can be transformed in an isotherm-like function. Our preliminary
results suggest that these dependences may be used as calibration
curves for rigorous regression treatments. We believe that these
results open up an avenue for development of future array sensors
for detection of carboxylates in biological and health-related appli-
cations.
Encouraged by this result, analysis of six NSAID-related car-
boxylates was performed using the present sensors array (Figure 9).
NSAIDs diclofenac, flurbiprofen, ibuprofen, salicylic acid, ketopro-
fen, and naproxen were tested in concentrations ranging from 0 to
100 ppm. First, PCA was carried out to reveal the clustering and the
trends in the response data. The PCA was used to plot the concen-
tration response function in the score plot space. Figure 9 shows
response functions composed of the average scores for each con-
centration of
a given carboxylate. From the concentration-
dependent response functions it can be seen that the sensor array
was able to discriminate between six different carboxylates in a
wide range concentration from 0.5 to 100 ppm, which covers the
typical NSAIDs urinary concentration.8b It should be noted that the
response functions are not the same as isotherms. The response
functions are the result of dimensionality reduction to mere three
dimensions. Thus, for each analyte, data from 150 variables are
reduced to a response function, which is then projected into three-
dimensional space (PC1×PC2×PC3).
Diclofenac
5
Flurbiprofen
Ibuprofen
EXPERIMENTAL PROCEDURES
Salicylic acid
Sensors S1,12b S2,15f and S812e were synthesized previously. The
synthesis and characterisation of sensors S3, S4, S5, S6, and S7
were described in the supporting information (SI). The multi-well
15×8 (sub-microliter) array chips were fabricated by ultrasonic
drilling of microscope slides (well diameter: 1000±10 μm, depth:
250±10 μm). The sensor solutions (500 μM) were prepared by
dissolving S1-S8 in a polyurethane hydrogel TecophilicTM THF
solution (4 wt%). In a typical array, the sensor solution (200 nL)
was spotted into the wells of the multi-well chip and dried to form a
5 µm thick polymer film in each well. The aqueous solutions of
carboxylic drugs (200 nL, 500 μM or 1 mM), whose pH values
adjusted at pH 8.5, were then added to each well containing the
sensor.
0
Ketoprofen
Naproxen
-5
30
-20
20
%)
PC1 (52.1
-10
10
0
0
%)
PC2 (32.6
Figure 9. PCA score plots describe the response of the S1-S8 array to
six different NSAID carboxylates at a concentration range between 0.5
and 100 ppm. The response functions are derived from calculating the
average of the scores for each concentration of a given drug. The limit
of detection for six carboxylates is ~0.1 ppm
Human urine used in the experiments displayed following charac-
teristics: pH 7.1, sodium ion (69 mEq/L), potassium (22.1
mEq/L), chloride (70 mEq/L), phosphate (34.8 mg/dL),
creatinine (24.1 mg/dL), and µ-albumine (10 mg/L).
Images from the sensor array were recorded using a Kodak Image
Station 440CF (for preliminary experiments) and a Kodak Image
As expected, all six response functions in Figure 9 originate from
ACS Paragon Plus Environment

Page 7 of 8
Journal of the American Chemical Society
Hewavitharan, A. K.; Chen, D. D.Y. J. Chromatogr. A 2012, 1267, 162–
169.
(6) Miziorko, H. M. Arch. Biochem. Biophys. 2011, 505, 131–143.
(7) Fernstrom, J. D.; Fernstrom, M. H. J. Nutr. 2007, 137, 1539S–1547S.
(8) (a) Hirai, T.; Matsumoto, S.; Kishi, I. J. Chromatogr. B 1997, 692,
375–388. (b) Aresta, A.; Palmisano, F.; Zambonin, C. G.; J. Pharm. Bio-
med. Anal. 2005, 39, 643–647.
Station 4000MM PRO (for qualitative and semi-quantitative ex-
periments). The images recorded using the Image Station
4000MM PRO reflect both fluorescence (11 channels) and color
intensity (four channels), respectively. In fluorescence detection,
1
2
3
4
5
6
7
8
the combinations of channels (excitation (λex) and emission (λem
)
filters) are: λex(430 nm)-λem(480 nm), λex(430 nm)-λem(535 nm),
λex(430 nm)-λem(600 nm), λex(470 nm)-λem(535 nm), λex(470 nm)-
λem(600 nm), λex(500 nm)-λem(535 nm), and λex(500 nm)- λem(600
nm). No excitation filters were used when using broadband UV as
excitation light, and the combinations of channels are: λex(UV)-
λem(440 nm), λex(UV)-λem(535 nm), λex(UV)-λem(480 nm), and
λex(UV)-λem(600 nm). In colorimetric detection using the
4000MM PRO, the filters for color-intensity measurement are 440,
480, 535, and 600 nm, respectively. After acquiring the images, the
integrated (non-zero) grey pixel value (n) is calculated for each
well in each channel. Images of the sensor chip were recorded be-
fore (b) and after (a) the addition of an analyte. The final responses
(R) were evaluated as indicated in the following equation (1):
(9) Maurer, H. H.; Tauvel, F. X.; Kraemer, T. J. Anal. Toxicol. 2001, 25,
237–244.
(10) (a) Sessler, J. L.; Gale, P.A.; Cho, W.-S. Anion Receptor Chemistry.
Monographs in supramolecular Chemistry; Royal Society of Chemistry:
Cambridge, 2006. (b) Bayly, S. R.; Beer, P. D. Struct. Bond. 2008, 129, 45–
94. (c) Valeria, A.; Fabbrizzi, L.; Mosca, L. Chem. Soc. Rev. 2010, 39,
3889–3915. (d) Duke, R. M.; Veale, E. B.; Pfeffer, F. M.; Kruger, P. E.;
Gunnlaugsson, T. Chem. Soc. Rev. 2010, 39, 3936–3953. (e) Li, A.-F.;
Wang, J.-H.; Wang, F.; Jiang, Y.-B. Chem. Soc. Rev. 2010, 39, 3729–3745.
(f) Wenzel, M.; Hiscock, J. R.; Gale, P. A. Chem. Soc. Rev. 2012, 41, 480–
520. (g) Ngo, H. T.; Liu, X.; Jolliffe, K. A. Chem. Soc. Rev. 2012, 41, 4928–
4965. (h) Pal, R.; Beeby, A.; Parker, D. J. Pharm. Biomed. Anal. 2011, 56,
352–358. (i) Butler, S. J.; Parker, D. Chem. Soc. Rev. 2013, 42, 1652–1666.
(11) (a) Gardner, J. W.; Bartlett, P. N. Electronic Noses: Principles and
Applications; Oxford University Press: New York, 1999. (b) Rakow, N. A.;
Suslick, K. S. Nature 2000, 406, 710–714. (c) Albert, K. J.; Lewis, N. S.;
Schauer, C. L.; Sotzing, G. A.; Stitzel, S. E.; Vaid, T. P.; Walt, D. R. Chem.
Rev. 2000, 100, 2595–2626. (d) Lavigne, J. J.; Anslyn, E. V. Angew. Chem.
Int. Ed. 2001, 40, 3118–3130. (e) Anzenbacher, P., Jr.; Lubal, P.; Bucek, P.;
Palacios, M. A.; Kozelkova, M. E. Chem. Soc. Rev. 2010, 39, 3954–3979.
(f) Paolesse, R.; Monti, D.; Dini, F.; Di Natale, C. Top. Curr. Chem. 2011,
300, 139–174. (g) Stitzel, S. E.; Aernecke, M. J.; Walt, D. R. Annu. Rev.
Biomed. Eng. 2011. 13, 1–25.
(12) (a) Nelson, T. L.; O’Sullivan, C.; Greene, N. T.; Maynor, M.S.;
Lavigne, J. J. J. Am. Chem. Soc. 2006, 128, 5640-5641. (b)Palacios, M. A.;
Nishiyabu, R.; Marquez, M.; Anzenbacher, P., Jr. J. Am. Chem. Soc. 2007,
129, 7538-7544. (c) Lavigne, J. J. Nat. Mater. 2007, 6, 548–549. (d) Pala-
cios, M. A.; Wang, Z.; Montes, V. A.; Zyryanov, G.V.; Hausch, B. J.;
Jursıkova, K. Anzenbacher, P., Jr. Chem. Commun. 2007, 3708–3710. (e)
Zyryanov, G. V.; Palacios, M. A.; Anzenbacher, P., Jr. Angew. Chem. Int. Ed.
2007, 46, 7849–7852. (f) Wang, Z.; Palacios, M. A.; Zyryanov, G. V.;
Anzenbacher, P., Jr. Chem. –Eur. J. 2008, 14, 8540–8546. (g) Wang, Z.;
Palacios, M. A.; Anzenbacher, P., Jr. Anal. Chem. 2008, 80, 7451–7459. (h)
Palacios, M. A.; Wang, Z.; Montes, V. A.; Zyryanov, G. V.; Anzenbacher, P.,
Jr. J. Am. Chem. Soc. 2008, 130, 10307–10314. (i) Shabbir, S. H.; Joyce, L.
A.; da Cruz, G. M.; Lynch, V. M.; Sorey, S.; Anslyn, E. V. J. Am. Chem. Soc.
2009, 131, 13125–13131. (j) Suslick, B. A.; Feng, L.; Suslick, K. S.; Anal.
Chem. 2010, 82, 2067–2073. (k) Liu, Y.; Palacios, M. A.; Anzenbacher, P.,
Jr. Chem. Commun. 2010, 46, 1860–1862. (l) Feng, L.; Zhang, Y.; Wen, L.
Y.; Chen, L.; Shen, Z.; Guan, Y. F. Chem. –Eur. J. 2011, 17, 1101–1104.
(m) Davey, E. A.; Zucchero, A. J.; Trapp, O.; Bunz, U. H. F. J. Am. Chem.
Soc. 2011, 133, 7716–7718. (n) Lin, H.; Jang, M.; Suslick, K. S. J. Am.
Chem. Soc. 2011, 133, 16786–16789. (o) Zhang, X.; You, L.; Anslyn, E. V.;
Qian, X. Chem. –Eur. J. 2012, 18, 1102–1110. (p) Anzenbacher, P., Jr.; Li,
F.; Palacios, M. A. Angew. Chem. Int. Ed. 2012, 51, 2345–2348. (q) Lim, J.;
Nam, D.; Miljanić , O. Š.; Chem. Sci. 2012, 3, 559–563. (r) Minami, T.;
Esipenko, N. A.; Zhang, B.; Kozelkova, M. E.; Isaacs, L.; Nishiyabu, R.;
Kubo, Y.; Anzenbacher, P., Jr. J. Am. Chem. Soc. 2012, 134, 20021–20024.
(13) Kitamura, M.; Shabbir, S. H.; Anslyn, E. V. J. Org. Chem. 2009, 74,
4479–4489.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
an
(1)
R 
1
 b
n
n
ASSOCIATED CONTENT
Supporting Information
Synthesis and characterization of S3, S4, S5, S6, and S7, UV-vis and
fluorescence spectra, results of multivariate analysis. This material is
available free of charge via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
pavel@bgsu.edu
ACKNOWLEDGMENT
We gratefully acknowledge the financial support from the NSF (CHE-
0750303 and EXP-LA 0731153 to P.A.) and BGSU (TIE Grant).
REFERENCES
(1) (a) Kroschwitz, J. I.; Seidel, A. Kirk-Othmer Encyclopedia of Chemi-
cal Technology, 5th ed., Vol. 5; Wiley-Interscience: Hoboken, N. J.: New
York, 2007. (b) Ullmann, F.; Gerhartz, W.; Yamamoto, Y. S.; Campbell,
T.; Pfefferkorn, R.; Rounsaville, J. F. Ullmann's Encyclopedia of Industrial
Chemistry, 7th ed.; Wiley-VCH: Weinheim, 2011.
(2) (a) Nonsteroidal Anti-Inflammatory Drugs: Mechanisms and
Clinical Uses, 2nd ed.; Lewis, A. J.; Furst, D. E., Eds. Marcel Dekker: New
York, 1994; (b) Lemke, T. L. Williams, D. A. Nonsteroidal anti-
inflammatory drugs Foye's Principles of Medicinal Chemistry, 6th ed.;
Lippincott Williams&Wilkins: Philadelphia, 2008. (c) Pharmacotherapy
Handbook, 8th ed.; Wells, B. G.; DiPiro, J. T.; Schwinghammer, T. L.;
DiPiro, C. Eds. McGraw Hill: New York, 2012.
(3) (a) Corcoran, J.; Winter, M. J.; Tyler, C. R. Crit. Rev. Toxicol.
2010, 40, 287–304. (b) http://www.epa.gov/esd/bios/daughton/book-
summary.htm
(14) Nishiyabu, R.; Anzenbacher, P., Jr. J. Am. Chem. Soc. 2005, 127,
8270–8271.
(4) Li, Q.; Weina, P. Pharmaceuticals 2010, 3, 2322–2332.
(15) Reviews see: (a) Gale, P. A.; Anzenbacher, P., Jr. Sessler, J. L. Coord.
Chem. Rev. 2001, 222, 57–102. (b) P. Anzenbacher Jr., Top. Heterocycl.
Chem. 2010, 24, 205–235. (c) Anzenbacher, P., Jr. Top. Heterocycl. Chem.
2010, 24, 237–265. Examples see: (d) Miyaji, H.; Sato, W.; Sessler, J. L.
Angew. Chem. Int. Ed. 2000, 39, 1777–1780. (e) Nielsen, K. A.; Jeppesen, J.
O.; Levillain, E.; Becher, J. Angew. Chem. Int. Ed. 2003, 42, 187–191. (f)
Nishiyabu, R.; Anzenbacher, P., Jr. Org. Lett. 2006, 8, 359–362. (g)
Sokkalingam, P.; Kim, D. S.; Hwang, H.; Sessler, J. L.; Lee, C.-H. Chem. Sci.
2012, 3, 1819–1824.
(5) (a) Sreekumar, A.; Poisson, L. M.; Rajendiran, T. M.; Khan, A. P.;
Cao, Q.; Yu, J.; Laxman, B.; Mehra, R.; Lonigro, R. J.; Li, Y.; Nyati, M. K.;
Ahsan, A.; Kalyana-Sundaram, S.; Han, B.; Cao, X.; Byun, J.; Omenn, G. S.;
Ghosh, D.; Pennathur, S.; Alexander, D. C.; Berger, A.; Shuster, J. R.; Wei,
J. T.; Varambally, S.; Beecher, C.; Chinnaiyan, A. M. Nature 2009, 457,
910–914. (b) Wu, H.; Liu, T.; Ma, C.; Xue, R.; Deng, C.; Zeng, H.; Shen,
X. Anal. Bioanal. Chem. 2011, 401, 635–646. (c) Solimana, L. C.; Hui, Y.;
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 8 of 8
(16) (a) Gunnlaugsson, T.; Kruger, P. E.; Jensen, P.; Tierney, J.; Ali, H. D.
(19) (a) Jambu, M. In Exploratory and Multivariate Data Analysis; Aca-
P.; Hussey, G. M. J. Org. Chem. 2005, 70, 10875–10878. (b) Quinlan, E.;
Matthews, S. E.; Gunnlaugsson, T. J. Org. Chem. 2007, 72, 7497–7503. (c)
Duke, R. M.; Gunnlaugsson, T. Tetrahedron Lett. 2011, 52, 1503-1505.
(17) The choice of DMSO and MeCN was made because the binding
behavior observed in these polar organic solvents is also observed in arrays
when the sensor elements are embedded in polyurethane matrices and the
anions are administered as strictly aqueous solutions. See: Palacios, M. A.;
Nishiyabu, R.; Marquez, M.; Anzenbacher, P., Jr. J. Am. Chem. Soc. 2007,
129, 7538–7544.
demic Press: San Diego, 1991. (b) Beebe, K. R.; Pell, R. J.; Seasholtz, M. B.
In Chemometrics: a practical guide; Wiley: New York, 1998. (c) Otto, M.
Chemometrics, Statistics and computer application in analytical chemistry;
Wiley-VCH: Chichester. 1999.
(20) (a) Carey, W.; Beebe, K.; Kowalski, B. Anal. Chem. 1986, 58, 149–
153. (b) Avila, F.; Myers, D.; Palmer, C. J. Chemometrics 1991, 5, 455–465.
(c) Green, E.; Olah, M. J.; Abramova, T.; Williams, L. R.; Stefanovic, D.;
Worgall, T.; Stojanovic, M. N. J. Am. Chem. Soc. 2006, 128, 15278–15282.
(21) (a) Turley, S. M. Understanding Pharmacology for Health Profes-
sionals, 3rd ed.; Prentice Hall: New Jersey, 2002; (b) Cohen, B. J. Medical
Terminology: An Illustrated Guide, 4th ed.; Lippincott Williams & Wil-
kins: Philadelphia, 2003.
1
2
3
4
5
6
7
8
(18) The 120 dimensions = 15 channels × 8 sensors are comprised of 88
dimensions (11 fluorescence channels × 8 sensors) and 32 dimensions (4
colorimetric channels × 8 sensors).
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment