SENSING OF ENANTIOMERIC EXCESS IN CHIRAL CARBOXYLATES
Ali Akdeniz
A Dissertation
Submitted to the Graduate College of Bowling Green State University in partial fulfillment of the requirements for the degree of
DOCTOR OF PHILOSOPHY
August 2016
Committee:
Pavel Anzenbacher Jr., Advisor
Carol Heckman Graduate Faculty Representative
H. Peter Lu
Andrew T. Torelli
ii
ABSTRACT
Pavel Anzenbacher Jr., Advisor
Chirality of organic compounds plays a crucial role in human and veterinary medicine.
More specifically, chiral carboxylates are extensively utilized in drug development processes,
and a number of them are commercialized as drugs. Due to distinct pharmacological properties of enantiomers of chiral drugs, asymmetric synthesis became an essential tool to synthesize enantiopure compounds. While combinatorial libraries enable to screen catalysts, auxiliaries, and conditions rapidly, the enormous amount of resulting samples cannot be tested simultaneously.
In this dissertation, we have reported two optical methods to determine the ee values of the chiral carboxylates in a high-throughput fashion.
In the first method, we have employed cinchona alkaloids, known as Corey-Lygo catalysts, as chemosensors for ee analysis of chiral carboxylates. N-alkylated 9- chloromethylanthracenyl cinchona alkaloids form complexes with carboxylates by means of electrostatic interactions and hydrogen bonds. Using characteristic changes in fluorescence intensities of the chemosensors upon addition of analytes, we were able to perform the successful qualitative and quantitative ee analysis of enantiomers of chiral carboxylates in aqueous media.
Using the sensor array comprising four N-alkylated cinchona alkaloids, we were able to achieve
100 % correct classification of chiral carboxylates, including several enantiomers of non-
steroidal anti-inflammatory drugs (NSAIDs) successfully. Furthermore, quantitative ee analysis of (S)-Ibuprofen, (S)-Naproxen, and (S)-Ketoprofen shows prediction errors as low as 3%.
The second method involves fluorescent macrocyclic sensors containing the chiral
BINOL moiety, and H-bond donors in the pocket to form complexes with carboxylates. The iii
intrinsic structure of the pocket of macrocycles displays an enantioselective behavior for chiral
carboxylates, while the substituents at the 3,3’-positions of the BINOL moiety allow for tuning
the shape and size of the cavity to enhance the recognition performance of the macrocycles. The
resulting macrocycles show distinct responses for a number of carboxylates, including the enantiomers of ibuprofen, ketoprofen, 2-phenylpropanoate, mandelate, and phenylalanine. The fingerprint-like responses of the macrocycles, for structurally similar analytes, enabled to perform a qualitative analysis of 12 carboxylates with 100 % correct classification. Quantitative ee analysis of ibuprofen, ketoprofen, and phenylalanine shows that the sensors correctly identify the ee values of the unknown samples with an error of prediction < 3% even in the presence of impurities.
iv
To my beloved wife, Fatmanur, and son, Salih
v
ACKNOWLEDGMENTS
First of all, I owe my sincere gratitude to my advisor, Dr. Pavel Anzenbacher, Jr., for his continuous support, guidance, and patience, which allowed me to grow as a scientist. I am very
grateful for his help, professionalism, and most importantly, friendship throughout my graduate
studies.
I would like to thank all my committee members, Dr. Carol Heckman, Dr. H. Peter Lu,
and Dr. Andrew T. Torelli, for their service on the committee.
I would like to thank my colleagues, members of our group, and collaborators from
different universities. Particularly, I want to thank Drs. Mosca, Minami, Caglayan for all the
enlightening discussions, support, and guidance. Without their great help, this dissertation would
be very difficult to finalize. In addition, I am grateful to Prof. Tadashi Ema for his collaboration
and to Vincent M. Lynch from the University of Texas for X-Ray measurements.
I would like to acknowledge all faculty and staff at the Center for Photochemical
Sciences. Especially, I would like to thank Dr. Stephania Messersmith for her guidance, help,
and friendship. I need to express my gratitude and appreciation to Ms. Hilda E. Miranda, Mr.
Doug Martin, Mr. Jadrez Romanowicz, Ms. Alita Frater, and Ms. Nora Cassidy for their help
during these years.
Finally, and most importantly, I would love to express my heartfelt gratitude to my
wife, Fatmanur, for her continuous encouragement and patience throughout our odyssey. Her
love was the biggest support in my life. At the beginning of my Ph.D. adventure, we were
blessed with an adorable son, Salih, who has been my source of motivation and energy. This
accomplishment would not have achieved without them. vi
Financial support from the McMaster Endowment and Center for Photochemical Science is greatly appreciated.
vii
TABLE OF CONTENTS
Page
CHAPTER I. INTRODUCTION ...... 1
1.1. Background ...... 1
1.2. Sensors ...... 2
1.3. Cross-Reactivity of A Sensor...... 3
1.4. Sensitivity and Selectivity of the Sensors ...... 4
1.5. High-throughput Screening (HTS) Methods and Instrumentations ...... 6
1.6. Microplate Arrays ...... 7
1.7. Polymer Chips ...... 9
1.8. Data Analysis ...... 12
1.8.1. Linear Discriminant Analysis (LDA) ...... 12
1.8.2. Support Vector Machine (SVM) ...... 14
1.9. Enantiomeric Excess ...... 15
1.10. Combinatorial Chemistry ...... 16
1.10.1. Split and Pool Method ...... 17
1.10.2. Parallel Synthesis ...... 17
1.10.3. Detection of Enantiomeric Excess (ee) ...... 18
1.10.3.1. HPLC ...... 18
1.10.3.2. HPLC-CD ...... 19
1.10.3.3. Gas Chromatography ...... 20
1.10.3.4. Capillary Electrophoresis ...... 20
1.10.4. Nuclear Magnetic Resonance(NMR) Spectroscopy ...... 21 viii
1.11. References ...... 21
CHAPTER II. SENSING OF ENANTIOMERIC EXCESS IN CHIRAL CARBOXYLIC
ACIDS BY CINCHONA ALKALOIDS ...... 28
2.1. Introduction ...... 28
2.2. Results and Discussions ...... 32
2.3. Conclusion ...... 45
2.4. Experimental Section ...... 46
2.4.1. General ...... 46
2.4.2. Preparation of the Polymer Chips ...... 47
2.4.3. Synthesis ...... 48
2.4.4. Preparation of TBA Salts of Guests ...... 51
2.4.5. Qualitative Analysis ...... 52
2.4.6. Semi-quantitative and Quantitative Analysis of Naproxen ...... 53
2.4.7. Semi-quantitative Analysis of Ibuprofen ...... 55
2.4.8. Semi-quantitative Analysis of Ketoprofen...... 57
2.5. References ...... 58
CHAPTER III. DETERMINATION OF ENANTIOMERIC EXCESS OF
CARBOXYLATES BY FLUORESCENT MACROCYCLIC SENSORS ...... 61
3.1. Introduction ...... 61
3.2. Results and Discussion ...... 63
3.3. Conclusion ...... 78
3.4. Experimental Section ...... 80
3.4.1. Synthesis ...... 82 ix
3.4.1. Qualitative Linear Discriminant Analysis ...... 87
3.4.2. Semi-quantitative Analysis of Phenylalanine ...... 88
3.4.3. Semi-quantitative Analysis of Ibuprofen ...... 89
3.4.4. Semi-quantitative Analysis of Ibuprofen with 5% Acetate ...... 90
3.4.5. Semi-quantitative Analysis of Ibuprofen with 5% Pyrophosphate ...... 91
3.4.6. Semi-quantitative Analysis of Ibuprofen with 0 to 10 % Acetate ...... 92
3.4.7. Semi-quantitative Analysis of Ibuprofen with 0 to 10 % Acetate ...... 93
3.4.8. Semi-quantitative Analysis of Ketoprofen...... 94
3.5. References ...... 95
CHAPTER IV. TRI-SERINE TRI-LACTONE SCAFFOLD FOR QUANTIFICATION OF
CITRATE IN URINE ...... 99
4.1. Introduction ...... 99
4.2. Results and Discussion ...... 101
4.3. Conclusion ...... 114
4.4. Experimental Section ...... 116
4.4.1. Preparation of Polymer Chips ...... 116
4.4.2. Synthesis ...... 117
4.4.3. Job’s plot Experiments ...... 121
4.4.4. Qualitative Analysis ...... 122
4.4.5. Semi-quantitative and Quantitative Analysis-in Water ...... 123
4.4.6. Semi-quantitative Analysis-in Buffer Solution ...... 125
4.4.7. Semi-quantitative Analysis-in Urine ...... 126
4.5. References ...... 127 x
APPENDIX A. LIST OF ABBREVIATIONS, ACRONYMS, AND SYMBOLS ...... 129
APPENDIX B. SUPPLEMENTARY DATA FOR CHAPTER II ...... 132
APPENDIX C. SUPPLEMENTARY DATA FOR CHAPTER III ...... 152
APPENDIX D. SUPPLEMENTARY DATA FOR CHAPTER IV ...... 184
xi
LIST OF FIGURES
Figure Page
1.1. Representation of a chemosensor that utilizes reporter and receptor moieties ...... 3
1.2. A) Representation of a cross-reactive sensor array. B) Response of array upon addition of the analyte of interest. C) Fingerprint-like response of the analyte after subtraction of A. ... 4
1. 3. Relative selectivity of the cross-reactive sensor is the ratio of the association constants of corresponding analytes...... 5
1. 4. A: The sensor has a higher selectivity towards to the analyte A, whereas the response to analyte A is weaker than for analyte B. B: Upon addition of the A and B analytes mixture to the sensor solution, a weak response is observed due to the stronger binding of analyte A ...... 6
1. 5. Example of a microplate (A), BioRAPTR microfluidic dispenser (B), manual syringe dispenser (Hamilton) (C), and CLARIOstar microplate reader (D) ...... 8
1. 6. Example of a custom-made glass micro-well slide...... 10
1. 7. Fabrication process of polymer chips ...... 11
1. 8. Schematic representation of the response matrix ...... 12
1. 9. A) Distribution of two groups of two variables, X1, and X2. B) One-dimensional linear discriminant function Y ...... 13
1. 10. Illustration of mapping the input into a feature space ...... 14
1. 11. Structures of (S)-thalidomide (left) and (R)-thalidomide (right)...... 16
1. 12. HPLC with parallel columns...... 19
2. 1. Structure of cinchona alkaloids...... 28
2. 2. Four conformations of cinchonine (adapted from Ref. 1)...... 29
2. 3. Structural features of N-alkylated cinchona alkaloids...... 30 xii
2. 4. Structures of the carboxylate guests used in this study...... 31
2. 5. (A) Mass spectrum of the S2-Ibuprofen complex. Inset: Calculated isotopic pattern for
+ C47H50N2O3 . (B) Mass spectrum of the S2-Naproxen complex. Inset: Calculated isotopic pattern
+. for C48H46N2O4 (C) Mass spectrum of the S3-Ketoprofen complex. Inset: Calculated isotopic
+ pattern for C50H46N2O4 ...... 33
2.6. X-ray structure of the S3 complex with (S)-Ibuprofen (left) and (R)-Ibuprofen (right).34
2.7. (Left) Fluorescence titration of S1 by (S)-Ibuprofen. (Right) Fluorescence titration of S4 by (S)-Naproxen...... 35
2.8. Fluorescence titrations of S1 (red) and S3 (blue) with the incremental amounts of (S)-
Ketoprofen, and (R)-Ketoprofen...... 36
2.9. Images of the sensors upon addition of the corresponding analytes ...... 38
2.10. Linear discriminant analysis of ten carboxylates ...... 39
2.11. Linear discriminant analysis plot of ketoprofen. Ten different enantiomeric compositions of ketoprofen were discriminated with 100% correct classification...... 40
2.12. Results of linear regression using support vector machine (SVM) afford quantitative analysis of enantiomeric excess of ketoprofen...... 41
2.13. Results of linear regression using support vector machine (SVM) afford quantitative analysis of enantiomeric excess of ibuprofen (left) and naproxen (right)...... 42
2.14. Linear discriminant analysis (LDA) of ibuprofen mixtures ...... 44
2.15. The canonical scores plot of qualitative analysis of 9 analytes and a control by using S1-
S4 in hydrogel matrix...... 52
2.16. Linear discriminant analysis (LDA) of enantiomeric excess of (S)-Naproxen ...... 53
2.17. The result of the linear regression using support vector machine (SVM) ...... 53 xiii
2.18. The canonical scores plot of Semi-quantitative analysis of (S)-Naproxen ...... 54
2.19. Linear discriminant analysis (LDA) of enantiomeric excess of (S)-Ibuprofen ...... 55
2.20. The result of the linear regression using support vector machine (SVM) ...... 55
2.21. The canonical scores plot of Semi-quantitative analysis of (S)-Ibuprofen ...... 56
2.22. The canonical scores plot of Semi-quantitative analysis of (S)-Ketoprofen ...... 57
3.1. Structures of macrocyclic chemosensors (S1–S4)...... 62
3.2. Synthetic route of macrocyclic chemosensors (S1–S4)...... 64
3.3. A) X-ray crystal structure of macrocyclic chemosensor (S2) with a water molecule in the cavity. B) DFT-optimized structure of S2 at the B3LYP/6-31G* level ...... 65
3.4. UV-VIS absorption and fluorescence spectra of S1 (A), S2 (B), S3 (C), and S4 (D) 66
3.5. ESI-MS spectra of S1+Ibuprofen (A), S2+ Ketoprofen (B), and S3+Mandelate (C) with
the calculated spectra of the complexes...... 67
3.6. Digital photographs of S3 (left) and S4 (right) upon addition of enantiomers of ibuprofen
and a control (in the middle of both panels) ...... 68
3.7. Structures of the carboxylates used in this study...... 70
3.8. Corresponding isotherms of the fluorescence titrations of (A) S1 with PPA, (B) S2 with
IBP, (C) S3 with KTP, and (D) S4 with Phe ...... 71
3.9. Linear discriminant analysis (LDA) plot of twelve analytes results in 100 % correct classification using S1–S4 sensor array ...... 73
3.10. A) Semi-quantitative analysis of ibuprofen. B) Quantitative analysis of ibuprofen using
SVM. C) Semi-quantitative analysis of ketoprofen. D) Quantitative analysis of ketoprofen using
SVM. The analysis was achieved by the S1–S4 array ...... 74 xiv
3.11. A) Linear discriminant analysis (LDA) of the semi-quantitative assay of enantiomeric
composition of ibuprofen in the presence of 5 % acetate. B) Quantitative analysis of the
enantiomeric composition of ibuprofen in the presence of 5 % acetate...... 76
3.12. A) Linear discriminant analysis (LDA) of semi-quantitative assay of enantiomeric
composition of ibuprofen in the presence of increasing composition of acetate from 0–10 % B)
Quantitative analysis of enantiomeric composition of ibuprofen in the presence of increasing
composition of acetate from 0–10%...... 77
3.13. A) Linear discriminant analysis (LDA) of the semi-quantitative assay of enantiomeric
composition of phenylalanine by employing only the S4 array. B) Quantitative analysis of the
enantiomeric composition of phenylalanine ...... 78
3.14. The canonical scores plot of qualitative assay...... 87
3.15. The canonical scores plot of qualitative assay of phenylalanine...... 88
3.16. The canonical scores plot of semi-quantitative assay of ibuprofen...... 89
3.17. The canonical scores plot of semi-quantitative assay of ibuprofen in the presence of 5 % acetate…………...... 90
3.18. The canonical scores plot of semi-quantitative assay of ibuprofen in the presence of 5 % pyrophosphate ...... 91
3.19. The canonical scores plot of semi-quantitative assay of ibuprofen in the presence of 0 to
10 % increasing concentration of acetate...... 92
3.20. The canonical scores plot of semi-quantitative assay of ibuprofen in the presence of 0 to
10 % increasing concentration of pyrophosphate ...... 93
3.21. The canonical scores plot of semi-quantitative assay of ketoprofen ...... 94
4.1. Structure of enterobactin produced by Escherichia coli ...... 100 xv
4.2. Structures of the sensors (S1-S2) based on tri-serine tri-lactone scaffold ...... 101
4.3. Synthetic route of the sensors (S1 and S2)...... 102
4.4. (A) ESI mass spectrum of the complex of [S1+Citrate+Na+MeCN]-. Inset: Calculated
isotope pattern for [S1+Citrate+Na+MeCN]- (B) ESI mass spectrum of the complex of
S2+Citrate+TBA. Inset: Calculated isotope pattern for S2+Citrate+TBA ...... 103
4.5. 1H NMR (500 MHz) titration of S1 (top) and S2 (bottom) upon the addition of citrate as
tri-tetrabutylammonium salts ...... 104
4.6. Left: Fluorescence titration of S1 Right: Fluorescence titrations of S2. Insets:
Corresponding isotherms ...... 106
4.7. Structures of the analytes used in this study ...... 107
4.8. A) Digital fluorescence images of the S1 (top) and S2 (bottom) in the polyurethane
microchips. B) Qualitative analysis of eight carboxylates, including mono, di, and tri
carboxylates ...... 110
4.9. Linear discriminant analysis of semi-quantitative analysis of citrate in buffer ...... 111
4.10. Linear discriminant analysis plot of semi-quantitative analysis of citrate in urine .. 113
4.11. Job’s plot experiments of S1 (left) and S2 (right) with citrate...... 121
4.12. Job’s plot experiments of S1 (left) and S2 (right) with acetate...... 121
4.13. The canonical scores plot of qualitative analysis of 8 analytes and a control ...... 122
4.14. Linear discriminant analysis (LDA) of Citrate-Tri-Na in hydrogel matrix...... 123
4.15. The result of the linear regression using support vector machine (SVM) ...... 123
4.16. The canonical scores plot of Semi-quantitative analysis of citrate tri-sodium salt by using
S1 and S2 in hydrogel matrix...... 124 xvi
4.17. The canonical scores plot of semi-quantitative analysis of citrate tri-sodium salt in buffer solution ...... 125
4.18. The canonical scores plot of semi-quantitative analysis of citrate tri-sodium salt in urine by using S1 and S2 in hydrogel matrix...... 126
xvii LIST OF TABLES
Table Page
2.1. Photochemical Properties of Cinchona Alkaloids ...... 32
2.2. Affinity Constants Obtained by Fluorescence Titration ...... 37
2.3. ANN Analysis of Unknown Ibuprofen Samples ...... 44
2.4. The Jackknifed Classification Matrix of Qualitative Analysis of 9 Analytes and A Control
by Using S1-S4 in Hydrogel Matrix...... 52
2.5. The Jackknifed Classification Matrix of Semi-Quantitative Analysis of (S)-Naproxen by
Using S1-S4 in Hydrogel Matrix...... 54
2.6. The Jackknifed Classification Matrix of Semi-Quantitative Analysis of (S)-Ibuprofen by
Using S1-S4 in Hydrogel Matrix ...... 56
2.7. The Jackknifed Classification Matrix of Semi-Quantitative Analysis of (S)-Ketoprofen
by Using S1-S4 in Hydrogel Matrix...... 57
3.1. Enantiomeric Fluorescence Difference Ratios ...... 69
3.2. Association Constants (Ka, M–1) Determined by Fluorescence Titrations. A...... 72
3.3. The Jackknifed Classification Matrix of Qualitative Assay Using S1-S4...... 87
3.4. The Jackknifed Classification Matrix of Semi-Qualitative Assay of Phenylalanine. 88
3.5. The Jackknifed Classification Matrix of Semi-Qualitative Assay of Ibuprofen...... 89
3.6. The Jackknifed Classification Matrix of Semi-Qualitative Assay in The Presence of 5 %
Acetate Using S1-S4...... 90
3.7. The Jackknifed Classification Matrix of Semi-Qualitative Assay in The Presence of 5 %
Pyrophosphate Using S1-S4...... 91 xviii
3.8. The Jackknifed Classification Matrix of Semi-Qualitative Assay in The Presence of 0 To
10 % Increasing Concentration of Acetate Using S1-S4...... 92
3.9. The Jackknifed Classification Matrix of Semi-Qualitative Assay in The Presence of 0 To
10 % Increasing Concentration of Pyrophosphate Using S1-S4...... 93
3.10. The Jackknifed Classification Matrix of Semi-Qualitative Assay of Ketoprofen .... 94
4.1. Association Constants (Ka, M-1) [A] Of Analytes Used in This Study ...... 108
4.2. The Jackknifed Classification Matrix of Qualitative Analysis of 8 Analytes and A
Control by S1 and S2 in Hydrogel Matrix...... 122
4.3. The Jackknifed Classification Matrix of Semi-Quantitative Analysis of Citrate Tri-
Sodium Salt by Using S1 and S2 in Hydrogel Matrix...... 124
4.4. The Jackknifed Classification Matrix of Semi-Quantitative Analysis of Citrate Tri-
Sodium Salt in Buffer Solution by Using S1 and S2 in Hydrogel Matrix...... 125
4.5. The Jackknifed Classification Matrix of Semi-Quantitative Analysis of Citrate Tri-
Sodium Salt in Urine by Using S1 and S2 in Hydrogel Matrix...... 126
xix
LIST OF EQUATIONS
Equation Page
1.1. Binding constants were calculated by nonlinear least-square methods using the equation
for a 1:1 binding model...... 13
1.2. Equation is used to calculate the enantiomeric excess (ee) of solutions...... 16
2.1. Binding constants were calculated by nonlinear least-square methods using the equation
for a 1:1 binding model...... 47
2.2. The final responses (R) were evaluated as indicated in this equation...... 48
2.3. General route for the synthesis of all the sensors ...... 48
3.1. Binding constants were calculated by nonlinear least-square methods using the equation
for a 1:1 binding model...... 81
3.2. Route for the synthesis of all the sensors ...... 82
4.1. The final responses (R) were evaluated as indicated in this equation...... 117
4.2. The first step of synthesis of the tri-serine tri-lactone scaffold...... 117
4.3. Formation of quaternary ammonium salts of the tri-serine tri-lactone...... 118
4.4. Substitution of reporter moieties of S1 to tri-serine tri-lactone scaffold...... 119
4.5. Substitution of reporter moieties of S2 to tri-serine tri-lactone scaffold...... 120 1
CHAPTER I. INTRODUCTION
1.1. Background
Anions are ubiquitous species that play important roles in many natural processes.
Identification and quantification of anions are desired in a number of fields, such as the
pharmaceutical industry, forensic sciences, pollutant monitoring, and for diagnostic purposes.1
More importantly, reliable analytical methods for determination of anions that are toxic to human
and other species would allow us to be aware of possible complications and find proper solutions
for the resulting consequences.2
A number of instruments and protocols for the detection and quantification of anionic species have been developed over the last century.3 Analytical methods based on the intrinsic
properties of analytes, such as gas chromatography–mass spectrometry (GC-MS) and high-
performance liquid chromatography (HPLC), are intensively used. The majority of such
methods, however, is labor-intensive, requires expensive instrumentation, or may involve pre-
treatment and purification of the product.4 Furthermore, few methods can be utilized for high- throughput analysis and the range of analytes that can be tested is limited. Recently, chemical sensors have attracted great interest due to their remarkable properties, such as being amenable
to high-throughput analysis and real-time acquisition.5 Unlike conventional analytical methods,
chemosensors do not mostly require any sample pre-treatment and are relatively inexpensive.
In the field of supramolecular analytical chemistry, chemical sensors are defined as
synthetic chemical structures that create assemblies which result in signal modulation upon the
addition of analytes.6 A typical qualitative or quantitative analysis of an analyte using
chemosensors usually requires three steps: generation of a signal, detection of the generated 2
signal, and processing of the data collected by a detector. The first step, signal modulation,
usually depends on the chemical interaction and equilibrium of the sensor with the analyte of
interest. Generated signals can be in the form of change in the thermal7, electrical8, mechanical9,
or optical properties10 of the sensors. The second step of the workflow of analysis is signal
detection. Generated signals are collected by a detector, which converts the signals to datasets.
The third step of the analysis, processing, employs chemometric methods to convert output data to final results. For example, in the detection of mercury by an organic field effect transistor, the current (signal) is generated by the interaction of mercury with the transistor and then data are collected by source meter.11 The resulting data matrix is then treated using chemometric
methods. In this dissertation, the term “chemosensors” refers to the fluorescent hosts that display
detectable changes in their fluorescence intensity upon interaction with the analytes of interest
(guests).
1.2. Sensors
Chemosensors, are array based sensors, are designed to utilize the two functional parts:
the receptor and the reporter. It is expected that receptor moieties form a complex with the
analyte of interest, whereas the reporter’s signals are modulated upon complex formation. The conventional chemosensor design mostly relies on the “lock and key” paradigm of molecular recognition proposed by Emil Fischer in 1894 (Figure 1.1).12 This model of recognition was inspired by the enzyme-substrate interaction. The substrate-specific geometry of the active site of an enzyme shows high selectivity for the particular substrate and a few structurally similar compounds. From the lock-and-key model, chemosensors should be selective and show affinity toward a particular analyte. However, the design and synthesis of selective chemosensors with a shape and desired geometry complementary to the analyte can be a difficult task. In addition to 3
geometrical considerations, selective chemosensors come with multiple drawbacks.5 First,
analytes with small structural alterations may result in false-positive signals. Second, the
unknown structure of the sample components will not allow to design the selective chemosensors
in advance. As a final point, selective chemosensors will not be able to analyze complex
mixtures that are recognized by their components.13
Analyte
Reporter: Off Reporter: On Analyte
Receptor Receptor
Figure 1. 1. Representation of a chemosensor that utilizes reporter and receptor moieties. Chemosensors demonstrate a turn-on response upon complex formation with the analyte.
1.3. Cross-Reactivity of A Sensor
As the number of the samples increase, an emerging approach, cross-reactive sensor arrays, are coming to the fore.14 In the mammalian olfaction and gustation systems, cross-
reactive receptors are capable of differentiating thousands of odors and flavors by generating
“fingerprint” response patterns of the analytes.15 Unlike the lock-and-key model of sensing,
recognition is achieved by the differential interaction of cross-reactive receptors with odorants
and flavor molecules. The resulting fingerprint-like responses are then stored in the brain for
future recalls. As given in an earlier example, cross-reactive sensors are not selective; that is,
they display an affinity to multiple analytes to a different extent. Hence, cross-reactive sensor
arrays do not necessarily require high specificity to particular analytes. The resulting fingerprint
responses of the analytes can be evaluated using pattern recognition protocols (Figure.1.2).16 4
A) B) C) Cross-reactive sensors with Finger-print response Cross-reactive sensors Analyte of interest of the analyte
Figure 1. 2. A) Representation of a cross-reactive sensor array. B) Response of the array upon addition of the analyte of interest. C) Fingerprint-like response of the analyte after subtraction of A.
One of the most significant advantages of the cross-reactive sensor array is the ability to
perform qualitative and quantitative analysis of multi-analyte mixtures.17 Following the
introduction of optical cross-reactive sensors by Walt (1996)18 and Suslick (2000)19, sensor
arrays for organic anions20, proteins21, nucleic acids22, metal ions23, and beverages24 have been
reported. Non-optical signal transduction mechanisms, such as cross-reactive sensors based on
radiofrequency25, potentiometry26, conductivity27, and others28, were also studied. Insufficient
discrimination power of chemosensors may require increasing the sensing elements in the
method. For example, piezoelectric sensors, or electronic noses with polymer-coated quartz crystals contain hundreds of sensing elements to perform satisfactory classification. 29
1.4. Sensitivity and Selectivity of the Sensors
A typical chemosensor (selective or cross-reactive) forms a complex with the analyte of
interest and displays signal modulation. In the case of supramolecular chemosensors, the driving
forces of these complex formations are noncovalent interactions, such as hydrogen bonding, π–π stacking, salt bridges, hydrophobic interactions, etc.30 Even though covalent bonds are stronger, 5
noncovalent interactions are preferred for two reasons. First, noncovalent interactions are
reversible processes. Covalent bonds can also be reversible to some extent; however, they mostly
require catalytic support to run reactions faster. Second, noncovalent interactions reach
equilibrium quickly and component exchanges are possible. As mentioned earlier, cross-reactive
chemosensors show a different affinity for each analyte and their different signal modulations.
This property is controlled by supramolecular interactions and depends on two intrinsic
properties of the chemosensors: sensitivity and selectivity.
Sensitivity can be described as the slope of the calibration curve and is the response of
the sensor as a function of analyte concentration.31 The limit of detection (LOD) strongly depends on the slope; thus, the magnitude of change in the response of the sensor. The selectivity of the sensor mainly depends on the receptor moiety. The geometry of the receptor imposes the selectivity among the same type of analytes. The selectivity of a chemosensor depends on the
32 affinity to the analyte of interest, which can be assessed by its association constant (Ka).
Relative selectivity of the sensors can be determined by the ratio of the association constants
(Figure 1.3):
[ ] + = [ ] [ 𝑒𝑒𝑒𝑒] 𝐴𝐴𝐴𝐴 𝑆𝑆𝑆𝑆 = 𝑆𝑆 𝐴𝐴 ⇆ 𝑆𝑆𝑆𝑆 𝐾𝐾 𝑒𝑒𝑒𝑒 𝑒𝑒𝑒𝑒 𝑆𝑆 𝐴𝐴 𝐾𝐾𝑆𝑆𝑆𝑆 [ ] 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑆𝑆𝑆𝑆 + = 𝐾𝐾 [ ] [ ] 𝑆𝑆𝑆𝑆 𝑒𝑒𝑒𝑒 𝑆𝑆 𝐵𝐵 ⇆ 𝑆𝑆𝑆𝑆 𝐾𝐾𝐴𝐴𝐴𝐴 𝑆𝑆 𝑒𝑒𝑒𝑒 𝐵𝐵 𝑒𝑒𝑒𝑒 Figure 1. 3. Relative selectivity of the cross-reactive sensor is the ratio of the association constants of corresponding analytes. 6
The combination of selectivity and sensitivity of a sensor enable to perform qualitative
and quantitative analyses of samples successfully. To illustrate the significance of these
properties, an example of a hypothetical cross-reactive chemosensor and two analytes is given.
A B
𝐾𝐾𝐴𝐴 ≫ 𝐾𝐾𝐵𝐵
Figure 1. 4. A: The sensor has a higher selectivity towards to the analyte A, whereas the response to analyte A is weaker than for analyte B. B: Upon addition of the A and B analytes mixture to the sensor solution, a weak response is observed due to the stronger binding of analyte A.
As is illustrated in Figure 1.4.A, the fluorescence modulation of analyte B is stronger than for analyte A. However, it is assumed that the chemosensor shows a higher selectivity toward analyte A. In the case of the addition of an analyte mixture to the sensor solution (Figure
1.4.B), weak signal modulation similar to the analyte A response is observed. This is due to cross-reactive behavior of the chemosensor that prefers to bind analyte A even in the presence of
B; hence, mainly the signal modulation of analyte A is observed.
1.5. High-throughput Screening (HTS) Methods and Instrumentations
Comprehensive analysis techniques, such as methods based on chromatography or NMR, suffer from a chronic weakness: they test one sample at a time. This is a time-consuming 7
approach and in the most cases require trained personnel. High-throughput screening (HTS)
methods enable thousands of chemical and pharmacological tests to be performed rapidly.
Instead of analyzing each individual sample, high-throughput methods are capable of performing
up to 100,000 assays/day.33 High-throughput screening was first introduced in 1970 in the field
of materials research to analyze multi-component systems that are synthesized separately.34
Today, most pharmaceutical research and drug development facilities use HTS systems, and
most of the drug discoveries are at least in part due to advancements in these systems. In addition
to the pharmaceutical industry, there is increasing demand for high-throughput assays in
endeavors such as clinical chemistry, proteomics, biochemistry, and genetics.35
Optical methods have attracted considerable attention due to their easy implementation and potential applications in high-throughput analysis and, including, differential sensing.36
Fluorescent chemosensors based on differential sensing are especially appealing due to their high
sensitivity and the number of possible modes, such as photon counting, polarization, and
lifetime. It is worth to note that even though colorimetric methods (absorbance as a read-out) can
be inexpensive and instrument-free (detected with a naked eye), fluorimetric methods hold their
ground due to their high sensitivity and low signal-to-noise ratio. Thus, fluorimetric methods
tend to provide lower LOD in many cases. In this dissertation, two different platforms for analyte
exposition are used. In the following sections, the details and the procedures of microplates9c and
polymer chips9b,9d will be explained.
1.6. Microplate Arrays
Microplate arrays are static systems that allow sensors and analyte molecules to interact
in miniaturized multi-well plates.37 The introduction of analytes into a chemosensor solution 8
results in the signal modulation of the chemosensor, sometimes with an incubation time. It is
worth noting that microplates and robotic dispensers are required to be compatible with the
solvent used, otherwise expensive chemical-resistant microplates are needed (Figures 1.5 B and
C).
A C
B D
Figure 1. 5. Example of a microplate (A), BioRAPTR microfluidic dispenser (B), manual syringe dispenser (Hamilton) (C), and CLARIOstar microplate reader (D).
Regardless of the way of preparation, multimode plate readers are used to acquire a
protocol based on fluorescence intensity or other modes, such as polarization, lifetime,
absorbance, etc.38 The output signals are recorded in the response matrix and cannot be
interpreted visually. Hence, chemometric methods are mostly used to decrease the
dimensionality of the response matrix, which usually has more than 10 channels (10
dimensions).
In this dissertation, microplate assays were designed in propionitrile solutions due to the
insoluble nature of chemosensors in aqueous media. Solutions of chemosensors and analytes 9
were introduced manually to chemical-resistant multi-well plates (polypropylene) as illustrated in Figure 1.5.A. Resulting signals from the chemosensors were recorded using a CLARIOstar microplate reader equipped with excitation and emission monochromators that allow the
selection of excitation and emission filter bands (Figure 1.5.D). Multiple acquisition channels
were set to record signals as fluorescence photon counts. The resulting datasets were reported as
a matrix in the form of analytes (X) × variables (Y). Variables represent the fluorescence
intensity counts, which are collected using the different sets of filters. To avoid instrumentation
and human errors, ten repetitions of each analyte were pipetted. Evaluation of the response
matrix was then performed by pattern recognition protocols.
1.7. Polymer Chips
The multi-well plate method is advantageous due to the possible modes that can be
acquired.39 However, acquisition time builds up as the number of the channels and wells
increase. Furthermore, evaporation of the samples becomes a concern as the time extends. Unlike
platforms based on multi-well plates, the polymer chip imaging method uses sensitive CDD or
CMOS cameras that can detect several multiple features of the whole platform in significantly shorter time. Although the sensors are insoluble in water, they are supported in a hydrogel matrix by the polymer to enable analysis in aqueous media. Our previous reported work on the behavior of poly (ether-urethane) polymers suggests that the polyethylene oxide part of the PUs absorbs water, thereby preventing hydration of the sensors and analytes, which in turn compete with the binding.40 The custom-made glass slides for the polymer chips were fabricated using ultrasonic
drilling. The typical multi-well slide has 10 × 21 wells with a diameter of 1000 ± 10 μm and a
depth of 250 ± 10 μm (Figure 1.6). 10
Figure 1. 6. Example of a custom-made glass micro-well slide. Each well has a diameter of 1 mm and depth of 0.25 mm. Most of the glass slides used in this dissertation were 10 × 21 micro- wells.
The polymer chip preparation scheme is illustrated in Figure 1.7. First, sensors were dissolved in polymer solution (4% poly(ether-urethane in THF). Next, 200 nL sensor solutions were dispensed into each well in the micro-well slide using a Hamilton dispenser (Figure 1.5.C).
The well bottom was covered with 5 μm of polymer film that converted into a hydrogel upon addition of water (400 nL). Whereas the amount of water absorbed will differ depending on the polymer, Liu et al. characterized the poly(ether-urethane) polymers available to our group.38
Poly (ether-urethane) polymers are designed with three fundamental parts. While the
polyethylene oxide component is hydrophilic, the polybutylene oxide and urethane connector
show hydrophobic properties. The optimum polymer that provides the best condition for our
applications was found to be polyurethane (polyethylene oxide (PEO)/polybutylene oxide (PBO)
= 4.8) polymer, which absorbs up to 60% of water (w/w). 11
Figure 1. 7. Fabrication process of polymer chips.
The controlled water uptake plays an essential role in the recognition process by excluding excessive amount of water, which may inhibit the complex formation.41 Third, 400 nL
of water was pipetted to form a hydrogel matrix after each well was covered with a polymer
solution and fully dried. Finally, 200 nL of aqueous analyte solution was manually cast on each
well and scanned using a 4000 mm KODAK scanner (Figure 1.7) after evaporation of the
excessive water at room temperature. 12
1.8. Data Analysis
Variables
Samples
Figure 1. 8. Schematic representation of the response matrix. Samples are in the rows and features, such as fluorescence intensity, polarization, lifetime, absorbance, etc., are in the columns.
In both the microplate arrays and polymer chips, fluorescence intensity changes were
acquired in the form of a multidimensional response matrix. Figure 1.8 shows a typical response
matrix, which displays the number of samples in rows and a number of variables in columns. As
the number of the channels increase, the response output matrix grows dramatically. Analysis of
such multidimensional data matrix was performed by statistical analysis tools.17 In this
dissertation, linear discriminant analysis (LDA) and support vector machine (SVM) regression analysis were mainly used. 42
1.8.1. Linear Discriminant Analysis (LDA)
Linear discriminant analysis is used to generate a classification model based on known
groups, also called learning or training objects, and to place unknown samples into appropriate
groups in the model. The classification model is obtained by forming linear discriminant function
(LDF) Y, which is a linear combination of all the recorded channels during the experiment.43 13
= + + +
𝑌𝑌 𝑎𝑎1𝑋𝑋1 𝑎𝑎2𝑋𝑋2 ⋯ 𝑎𝑎𝑛𝑛𝑋𝑋𝑛𝑛 Equation 1.1. Linear discriminant function.
As mentioned earlier, n measurements are performed for each sample. Using LDF, n- dimensional response data are reduced to a one-dimensional function of Y. Coefficients of the
LDF are chosen to maximize the separation of groups from each other and minimize the
separation between samples that are in the same group.
Figure 1. 9. A) Distribution of two groups of two variables, X1, and X2. B) One-dimensional linear discriminant function Y results in 100% correct classification into one of two groups (adapted from ref.42).
To illustrate this concept, a sample of spectral data comprises two variables and two groups
are given. Figure 1.9.A shows the distribution of each group on the two-dimensional graph. As
observed, neither variable, X1 nor X2, is enough to separate the clusters completely. However, a
linear combination of variables with coefficient values that maximize the distance between the groups and minimize within-group distances can result in a correct classification (Figure 1.9.B). 14
The performance of the LDF can be tested by two methods.44 The first method, the simplest
method, classifies each sample within the group and reports the percent of the samples correctly
allocated. We did not rely on this method due to over-optimistic results. This is due to the fact that discriminant function Y is formed using all the data in the matrix, including the samples to
be tested. The second method, cross-validation, is always performed in this dissertation to test
the performance of our analysis. Cross-validation, also called the leave-one-out method, forms
LDF by excluding one data point and allocating the excluded data point in the appropriate group.
This procedure is repeated for each data point in the data matrix and performance is evaluated.
The resulting data are reported as a jackknife classification matrix. It is expected that all data
points can be 100% correctly allocated in their groups using both methods for a robust and
reliable analysis.
1.8.2. Support Vector Machine (SVM)
Figure 1. 10. Illustration of mapping the input into a feature space that can be linearly separated. The support vector machine uses kernel functions to form linear support samples while maximizing the distance between support vectors.
The support vector machine is an approach for multivariate clustering and data mining methods. It was first proposed in the 1990s and attracted great attention due to its distinct
properties.42 Unlike LDA, SVM finds the boundaries between classes without employing all the 15 available recorded data. Figure 1.10 illustrates the boundary samples; that is, support vectors of imaginary two classes. Boundary samples were chosen to maximize the space between the groups. Then, unknown samples were classified according to the side of the boundary to which they are allocated. However, multivariable data collected from optical sensors show nonlinear responses and simple linear support vectors cannot be drawn. Here, the way of the boundaries formed needs to be established differently. The SVM handles this problem easily using kernel functions that can be in linear, polynomial, or radial form as the basis function to map nonlinear data into a feature space, where linear support vectors can be drawn. SVM tries to maximize the distance between two classes while the linear trend is being regenerated. Although further explanation goes beyond the scope of this dissertation, in the studies we have performed, SVMs have provided excellent performance in quantitative analysis.
1.9. Enantiomeric Excess
Drug stereochemistry plays a crucial role in pharmacology. Altered positions in space of the four different substituents on a carbon atom result in enantiomers that are non- superimposable mirror images of one another.45 Enantiomers of the same chiral compound may display different pharmacological properties; that is, whereas one enantiomer of a compound may be an active drug, another enantiomer may not have any pharmacological properties or may even be toxic. One example is a drug, thalidomide, was prescribed widely in Europe for the treatment of morning sickness in pregnant women in the 1960s (Figure 1.11).46 Thalidomide is a chiral molecule and only the (R)-thalidomide was produced as a drug. However, studies showed that (R)-thalidomide racemizes in vivo and (S)-thalidomide, which has a teratogenic effect, is formed. Ten thousand babies suffered severe birth defects as a result of their mothers’ use of this drug. 16
O O O O NH NH N O N O
O O
Figure 1. 11. Structures of (S)-thalidomide (left) and (R)-thalidomide (right).
Due to the different pharmacological activities of enantiomers, the US Food and Drug
Administration (FDA) requires separate testing of the toxicological parameters of both
enantiomers of chiral drugs.47 In the case of the high toxicity of the enantiomeric pair, chiral
drugs need to be sold in their enantiopure forms.
Asymmetric synthesis is a cost-efficient method to obtain enantiopure compounds using a chiral catalyst and auxiliaries.48 To find optimum conditions that lead to an enantiopure product,
screening and testing of a range of catalysts, auxiliaries, and conditions are required. The
efficiency of the asymmetric synthesis is quantified by enantiomeric excess (ee) calculations.
[ ] [ ] = × 100 [ ] [ ] 𝑅𝑅 − 𝑆𝑆 𝑒𝑒𝑒𝑒 � 𝑅𝑅 + 𝑆𝑆 � Equation 1.2. Equation is used to calculate the enantiomeric excess (ee) of solutions.
1.10. Combinatorial Chemistry
Discovery of procedures that yield enantiopure compounds are usually conducted using rational design of the experiments which carry out the experiments one at a time. However, the activity of the catalyst is possibly affected by many factors, including temperature, solvent, substrate, reaction time, etc. Although rational methods are at the forefront of asymmetric synthesis, the testing of many catalysts, auxiliaries, and conditions is time-consuming. In the
1990s, a new method, combinatorial chemistry, was introduced as a tool for the testing catalysts 17
and auxiliaries of a number of conditions and libraries of compounds in a high-throughput
fashion. With advancements in reaction discovery, determination of enantiopurity of each
condition/catalysis becomes a major bottleneck to combinatorial synthesis. Here, two
combinatorial methods are explained briefly as further explanation goes beyond the scope of this
dissertation.
1.10.1. Split and Pool Method
Split and pool techniques are time-efficient methods to synthesize a large number of
compounds in a short time.49 In this method, reagents (A, B, and C) are coupled to polymer resin
beads. Coupled polymer resin beads are combined and then split into n number of groups to react
with n number of reagents. This procedure is repeated with different reagents m times to produce
n number of libraries with nm products. The screening of these products is carried out to reveal
the catalyst/axillaries that yield the desired product. Identification of the product is performed
either by further deconvolution or encoding techniques.50 The split and pool method enables a large number of reactions in a short time.
1.10.2. Parallel Synthesis
Parallel synthesis includes all the advantages of both combinatorial techniques and rational synthetic methods. A smaller number of libraries are tested in a parallel fashion, whereas
each reaction is tested individually. Due to advancements in high-throughput instrumentation,
such as robotic liquid dispensers and micro-well plate readers, the synthesis and screening of a large number of chiral compounds become much easier and faster. In addition to speed and
accuracy, high-throughput instrumentation allows reactions to be performed in a wide range of
conditions (temperature, pH, solvent, etc.).51 Miniaturized reaction settings also reduce the cost
and waste production. 18
Combinatorial synthesis and parallel synthesis are time-efficient approaches to developing catalysts that yields highly enantiopure products.52 However, in addition to their
advantages, these methods are, in part, limited by a lack of methods to screen for ee of the products. Due to the inability to test a large number of products, researchers have to reduce the number of the catalyst/auxiliary to reduce the number of the products that need to be tested.
1.10.3. Detection of Enantiomeric Excess (ee)
As mentioned earlier, the development of practical techniques for ee determination is crucial in drug development processes. The most important methods for analysis of enantiomeric compositions are explained below. These methods include chiral high-performance liquid chromatography (HPLC), HPLC coupled with circular dichroism (HPLC-CD), chiral gas chromatography (GC), capillary electrophoresis (CE), and nuclear magnetic resonance spectroscopy (NMR).53 Other methods, such as liquid crystals54, infrared thermography55,
molecularly imprinted polymers56, and enzymatic and antibody methods57 were reported but not
presented in this dissertation.
1.10.3.1. HPLC
Chiral separation using chromatographic techniques can be classified into two groups: indirect and direct separation. Indirect separation requires to derivatize the enantiomers of the chiral compounds to form diastereomers which can be separated easily. In contradiction, direct chiral separation includes a chiral stationary phase or mobile phase, which forms diastereomers with enantiomers of the analytes that differ in their physical properties. A reversible complex formation is preferred to allow recovery and reuse of the column. 19
Figure 1. 12. HPLC with parallel columns.
Many types of chiral columns are commercially available.58 However, no chiral column
fits for the ee analysis of all analytes. Ideally, chiral columns should be identified for a particular analyte and the method optimized. In addition to the necessity for analyte-specific chiral columns, HPLC is not compatible with high-throughput analysis. Relatively rapid HPLC analysis takes ten minutes to perform, which is not fast enough to analyze the thousands of samples that can be analyzed using combinatorial techniques. Several approaches have been proposed to reduce the analysis time. For example, parallel columns with individual pumps allow multiple samples to be analyzed simultaneously. Nonetheless, conventional parallel chiral HPLC
can take up to 16 hours to scan 96 samples and uses more than a liter of solvent.59
1.10.3.2. HPLC-CD
HPLC coupled with circular dichroism (CD) is another method proposed for ee analysis.60 Unlike chiral HPLC columns, HPLC-CD involves circularly polarized light to create a chiral environment for detecting enantiomers. HPLC-CD requires a chromophore in intimate contact with the chiral center. However, most analytes do not have a chromophore or 20
chromophore with a strong CD signal. Derivatization can be performed for the analytes that do
not utilize a chromophore.61 Nevertheless, the time-frame of chromatographic methods does not
allow high-throughput analysis.62
1.10.3.3. Gas Chromatography
Gas chromatography uses gasses as a mobile phase. Reversible interaction with chiral
stationary phase (CSP) allows differentiating enantiomers by forming diastereomers.63 The
advantages include high efficiency, sensitivity, and robustness of the analysis while a limited number of analytes can be analyzed due to the thermal stability of the samples. Gas chromatography involves evaporation of the samples at high temperature, which usually results in racemization or degradation of the samples or the column itself.64 Therefore, this technique
can only be used for analytes that are highly volatile and thermally stable.
1.10.3.4. Capillary Electrophoresis
Capillary electrophoresis (CE) is an electrokinetic separation technique performed in a
capillary. It is based on the migration of the charged species through an electrolyte solution
under the influence of an electric field. Chiral resolution is achieved by the addition of chiral
selectors, which mostly consist of charged cyclodextrins and polysaccharides.65 Advantages of
CE include high resolution, minimal sample diffusion, small sample size, and, most importantly,
it can be employed for parallel screening in high-throughput analysis.66 However, analysis
requires analytes to be ionizable in the buffer solution and must involve a chromophore or be
derivatized to be detected.67 21
1.10.4. Nuclear Magnetic Resonance(NMR) Spectroscopy
NMR is a spectroscopic method that can be utilized for ee analysis.68 This method involves chiral lanthanide shift reagents (CLSRs)69, or chiral derivatizing reagents (CDAs)65d to form diastereomeric complexes with each enantiomer that may result in different resonance signals. Relative integrals of the corresponding peaks can be easily used to reveal the enantiomeric composition of the samples. CLSRs can coordinate with chiral molecules via their metal centers; however, line broadening is observed due to the paramagnetic nature of lanthanide complexes. On the other hand, compounds must have a basic Lewis lone-pair to coordinate, or further derivatization is needed. The advantages of CDAs over the CLRSs are a better resolution in resonance signals and the resulting diastereomers can be separated by traditional chromatographic methods. However, promising studies of ee analysis using NMR spectroscopy in combinatorial chemistry are limited due to lack of applicability to high-throughput analysis.
1.11. References
1 (a) Harsanyi, G. In Sensors in biomedical applications: fundamentals, technology & applications, Technomic: Lancaster, PA, 2000.
(b) Fraden, J. In Handbook of modern sensors: physics, designs, and applications; AIP
Press/Springer: New York, USA, 2004.
(c) Wilson, J. S. In Sensor technology handbook, Wilson, J. S., Ed.; Elsevier: Amsterdam, 2005.
2 Supramolecular Chemistry of Anions; Bianchi, A., Bowman-James, K., Garcia-Espana, E.;
Wiley-VCH: New York, 1997.
3 Szabadváry, F. History of Analytical Chemistry. International series of monographs in analytical chemistry, Vol. 26. Pergamon Press: Oxford, 1966. 22
4 Leung, D.; Kang, S. O.; Anslyn, E. V. Chem. Soc. Rev. 2012, 41, 448.
5 Bell, T. W.; Hext, N. M. Chem. Soc. Rev. 2004, 33, 589.
6 You, L.; Zha, D.; Anslyn, E. V. Chem. Rev. 2015, 115, 7840.
7 Grate, J. W. Chem. Rev. 2008, 108, 726.
8 Hatchett, D. W.; Josowicz, M. Chem. Rev. 2008, 108, 746.
9 Diamond, D.; Coyle, S.; Scarmagnani, S.; Hayes, J. Chem. Rev. 2008, 108, 652.
10 (a) Grate, J. W.; Egorov, O. B.; O'Hara, M. J.; DeVol, T. A. Chem. Rev. 2008, 108, 543.
(b) Akdeniz, A.; Mosca, L.; Minami, T.; Anzenbacher, Jr., P. Chem. Commun. 2015, 51, 5770.
(c) Akdeniz, A.; Minami, T.; Watanabe, S.; Yokoyama, M.; Ema, T.; Anzenbacher, Jr., P.
Chem. Sci. 2016, 7, 2016.
(d) Akdeniz, A.; Caglayan, M. G.; Anzenbacher, Jr., P. Chem. Commun. 2016, 52, 1827.
11 Minami, T.; Sasaki, Y.; Minamiki, T.; Koutnik, P.; Anzenbacher, Jr., P.; Tokito, S. Chem.
Commun. 2015, 51, 17666.
12 Behr, J.-P. The Lock-and-Key Principle, The State of the Art-100 Years On; John Wiley &
Sons: Chichester, U.K., 2008
13 Palacios, M. A.; Nishiyabu, R.; Marquez, M.; Anzenbacher, Jr., P. J. Am. Chem. Soc. 2007,
129, 7538.
14 Lavigne, J. J.; Anslyn, E. V. Angew. Chem. Int. Ed. 2001, 40, 3118.
15 (a) Doty, R. L.; Dekker, M. Handbook of Olfaction and Gustation, New York, 2nd ed., 2003.
(b) Stopfer, M.; Jayaraman, V.; Laurent, G. Neuron, 2003, 39, 991.
16 Bishop, C. M. Neural Networks for Pattern Recognition; Bishop, C. M, Oxford University
Press: Oxford, U.K., 1995. 23
17 Otto, M. In Chemometrics : statistics and computer application in analytical chemistry,
Wiley-VCH: Weinheim, New York, 1998.
18 Dickinson, T. A.; White, J.; Kauer, J. S.; Walt, D. R. Nature 1996, 382, 697.
19 Rakow, N. A.; Suslick, K. S. Nature 2000, 406, 710.
20 Lavigne, J. J.; Anslyn, E. V. Angew. Chem. Int. Ed. 2001, 40, 3118.
21 Wright, A. T.; Griffin, M. J.; Zhong, Z.; McCleskey, S. C.; Anslyn, E. V.; McDevitt, J. T.
Angew. Chem., Int. Ed. 2005, 44, 6375.
22 Matson, R. S.; Rampal, J. B. Genomic/Proteomic Tech. 2003, 3, 37.
23 (a) Szurdoki, F.; Ren, D.; Walt, D. R. Anal. Chem. 2000, 72, 5250.
(b) Mayr, T.; Liebsch, G.; Klimant, I.; Wolfbeis, O. S. Analyst 2002, 127, 201.
(c) Mayr, T.; Igel, C.; Liebsch, G.; Klimant, I.; Wolfbeis, O. S. Anal. Chem. 2003, 75, 4389.
24 (a) Zhang, C.; Bailey, D. P.; Suslick, K. S. J. Agric. Food Chem. 2006, 54, 4925.
(b) Zhang, C.; Suslick, K. S. J. Agric. Food Chem. 2007, 55, 237.
25 Potyrailo, R. A.; Surman, C.; Morris, W. G. J. Comb. Chem. 2009, 11, 598.
26 Correia, D. P. A.; Magalhaes, J. M. C. S.; Machado, A. A. S. C. Microchim. Acta. 2008, 163,
131.
27 Kukla, A. L.; Pavluchenko, A. S.; Shirshov, Y.M.; Konoshchuk, N.V.; Posudievsky, O.Y. Sens
Actuators B Chem. 2009, B135, 541.
28 (a) Morita, K.; Shimizu, Y., Anal.Chem. 1989, 61,159.
(b) Li, B.; Santhanam, S.; Schultz, L.; Jeffries-El, M.; Iovu, M. C.; Sauve, G.; Cooper, J.; Zhang,
R.; Revelli, J. C.; Kusne, A. G.; Snyder, J. L.; Kowalewski, T.; Weiss, L. E.; McCullough, R. D.;
Fedder, G. K.; Lambeth, D. N. Sens Actuators B Chem. 2007, B123, 651. 24
(c) Jahnke, H.-G.; Rothermel, A.; Sternberger, I.; Mack, T. G. A.; Kurz, R. G.; Paenke, O.;
Striggow, F. Robitzki, A. A., Lab. Chip. 2009, 9, 1422.
29 Gautschi, G. Piezoelectric sensorics Springer Berlin, Heidelberg, New York, 2002
30 Steed, J. W.; Atwood, J. L. In Supramolecular chemistry; Wiley: Chichester, New York,
2003.
31 Mermet, J.-M.; Kellner, R.; Otto, M.; Valcarcel, M.; Widmer, M. Analytical Chemistry: A
Modern Approach to Analytical Science. Wiley: New York, 2004.
32 Anslyn E. V.; Dennis A. Modern Physical Organic Chemistry.University Science Books,
2006.
33 (a) Hertzberg, R.P.; Pope, A.J. Curr. Opin. Chem. Biol. 2000, 4, 445.
(b) Schena, M. In Microarray Analysis; Wiley-Liss: Hoboken, NJ, 2003.
34 Hanak, J.J. J. Mater. Sci. 1970, 5, 964.
35 Charlton, J., J. Advances in High-Throughput Analysis: Automated Radiochemical
Separations and Nanopillar based Separations and Field Enhanced Spectroscopy, Ph.D.
Dissertation [Online], University of Tennessee – Knoxville, 2015.
36 (a) Leung, D.; Kang, S. O.; Anslyn, E. V. Chem. Soc. Rev. 2012, 41, 448.
(b) You, L.; Zha, D.; Anslyn, E. V. Chem. Rev. 2015, 115, 7840.
37 Armstrong, J.W. HTS and Robotics Implementation: a critique of robotics equipment and
strategies. 19-33. Published by HTS Consulting Ltd. 1997.
38 Clariostar, http://www.bmglabtech.com/en/products/clariostar (accessed April 4th, 2016)
39 Schena, M. Microarray Analysis, Wiley-Liss, Hoboken, NJ, 2003, 630.
40 Anzenbacher, P., Jr.; Liu, Y.; Palacios, M. A.; Minami, T.; Wang, Z.; Nishiyabu, R. Chem.
Eur. J. 2013, 19, 8497. 25
41 Anzenbacher Jr., P.; Liu, Y.; Palacios, M.; Minami, T.; Wang, Z.; Nishiyabu, R. Chem. Eur. J.
2013, 19, 8497 – 8506
42 Applied Chemometrics for Scientists, Brereton, R. G. Wiley: Chichester, U.K., 2007.
43 J. N. Miller, and J. C. Miller Statistics and Chemometrics for Analytical Chemistry, sixth ed.,
Pearson, UK, 2010
44 R. G. Brereton Applied Chemometrics for Scientists, J. Wiley and Sons, Chichester, 2007
45 S. H.; Eliel, E. L. Stereochemistry of Organic Compounds, Wilen, John Wiley Interscience, 1st
Ed.1994.
46 Mellin, G. W.; Katzenstein, M. N Engl J Med. 1962, 267, 1184.
47 Development of New Stereoisomeric Drugs, U.S. Food and Drug Administration, U.S., 1992.
48 (a) Lin, G.-Q.; Li, Y.-M.; Chan, A. S. C. Applications of Asymmetric Reactions in the
Synthesis of Natural Products, in Principles and Applications of Asymmetric Synthesis, John
Wiley & Sons, Inc., New York, USA, 2001.
(b) Christmann, M.; Braese, S. Asymmetric Synthesis: The Essentials, Wiley-VCH, Weinheim,
New York, 2007.
49 Pescarmona, P. P.; Van der Waal, J. C.; Maxwell, I. E.; Maschmeyer, T. Catal. Lett., 1999, 63,
1.
50 Toy, P. H.; Lam, Y. Solid-Phase Organic Synthesis: Concepts, Strategies, and Application,
John Wiley & Sons, Inc., Hoboken, New Jersey. 2011.
51 Burgess, K.; Lim, H.-J.; Porte, A. M.; Sulikowski, G. A. Angew. Chem., Int. Ed. Engl. 1996,
35, 220.
52 (a) Ding, K.; Du, H.; Yuan, Y.; Long, J. Chem.--Eur. J. 2004, 10, 2872.
(b) Ding, K. Pure Appl. Chem. 2005, 77, 1251. 26
53 (a) Gasparrini, F.; Misiti, D.; Villani, C. J. Chromatogr. A, 2001, 906, 35.
(b) Traverse, J. F.; Snapper, M. L. Drug Discov Today 2002, 7, 1002.
(c) Guebitz, G.; Schmid, M. G. Mol. Biotechnol. 2006, 32, 159.
54 Demus, D.; Goodby J. W. G.; Gray, G. W. Handbook of Liquid Crystals. Vol. 1,
Fundamentals, Wiley-VCH, Weinheim, New York, Chichester, 1998.
55 Taylor, S. J.; Morken, J. P. Science, 1998, 280, 267.
56 Sellergren, B. Molecularly Imprinted Polymers: Man- Made Mimics of Antibodies and Their
Applications in Analytical Chemistry, Techniques and Instrumentation in Analytical Chemistry,
Elsevier Science, United Kingdom, 2001.
57 (a) Badalassi, F.; Wahler, D.; Klein, G.; Crotti, P.; Reymond, J. L. Angew. Chem., Int. Ed.
2000, 39, 4067.
(b) Korbel, G.A.; Lalic, G.; Shair, M. D. J. Am. Chem. Soc. 2001, 123, 361.
58 Ahuja, S. LCGC North Am. 2007, 25, 1112.
59 Welch, C. J. "Microscale chiral HPLC in support of pharmaceutical process research."
Chirality 2008, 21, 114.
60 (a) Bobbitt, D. R.; Linder, S. W. Trends Anal. Chem. 2001, 20, 111.
(b) Roussel, C.; Del Rio, A.; Pierrot-Sanders, J.; Piras, P.; Vanthuyne, N. J. Chromatogr., A
2004, 1037, 311.
61 Wolf, C.; Bentley, K. W. Chem. Soc. Rev., 2013, 42, 5408-5424
62 Chen, L.; Zhao, Y.; Gao, F.; Garland, M. Appl. Spectrosc. 2003, 57, 797.
63 He, L.; Beesley, T. E. J. Liq. Chromatogr. Relat. Technol. 2005, 28, 1075.
64 Zhang, Y.; Wu, D.-R.; Wang-Iverson, D. B.; Tymiak, A. A. Drug Discov. Today 2005, 10, 571
65 (a) Nishi, H.; Terabe,S. J. Chromatogr., A 1995, 694, 245 27
(b) Gubitz, G.; Schmid, M. G. J. Chromatogr., A 1997, 792, 179
(c) Verleysen, K.; Sandra, P. Electrophoresis 1998, 19, 2798
(d) Van Eeckhaut, A.; Michotte, Y. Electrophoresis 2006, 27, 2880.
66 Fang, Q.; Xu, G.; Fang, Z. Anal. Chem., 2002, 74, 1223.
67 Reetz, M. T.; Kuhling, K. M.; Deege, A.; Hinrichs, H.; Belder, D. Angew. Chem., Int. Ed.
2000, 39, 3891.
68 (a) Parker, D. Chem. Rev. 1991, 91, 1441.
(b) Wenzel, T. J.; Wilcox, J. D. Chirality 2003, 15, 256.
(c) T. J. Wenzel, Discrimination of Chiral Compounds Using NMR Spectroscopy, Wiley,
Hoboken, NJ, 2007
69 Whitesides, G. M.; Lewis, D. W. J. Am. Chem Soc. 1970, 92, 6979. 28
CHAPTER II. SENSING OF ENANTIOMERIC EXCESS IN CHIRAL CARBOXYLIC
ACIDS BY CINCHONA ALKALOIDS
2.1. Introduction
Cinchona alkaloids, namely, quinine, quinidine, cinchonine, and cinchonidine, are organic molecules isolated from the bark of cinchona trees (Figure 2.1). Although medicinal use of cinchona trees dates back to the mid-seventeen century, active compounds were marketed as an antimalarial drug in 1820.1 Besides antimalarial activity, these compounds are also commercialized today as muscle relaxants, cardiac depressants, and food additives. Nowadays,
700 metric tons of cinchona alkaloids are produced annually, and nearly half of production is used as a “bitters” additive (for example, tonic water) in the food and beverage industry.
R 8 9 Quinine OMe R S Quinidine OMe S R Cinchonine H S R Cinchonidine H R S
Figure 2. 1. Structure of cinchona alkaloids.
Cinchona alkaloids were introduced to organic chemistry by Pasteur in 1853.2 Pasteur reported that tartaric acid can be separated from its enantiomers by crystallization of their diastereomeric salts with cinchona alkaloids.2 Currently, there is a number of cinchona alkaloids and their derivatives used as resolving agents.3 In addition to the utilization of cinchona alkaloids as resolving reagents, they are also used as enantioselective tools.1 For example, cinchona 29
alkaloids and their derivatives are extensively used as stereoselective catalysts and chiral
auxiliaries for many types of reactions in organic chemistry. 4,5
Figure 2. 2. Four conformations of cinchonine (adapted from Ref. 1).
From the perspective of their structures, cinchona alkaloids have an internal rotation
around the C8-C9 bond that results in four low-energy conformers: anti-closed, anti-open, syn-
closed, and syn-open (Figure 2.2). In addition to their flexible structures, they can be modified
easily by various types of reactions.6 Importantly, N-alkylated cinchona alkaloids display efficient asymmetric induction by chiral ion-pairing mechanism.7 In addition to electrostatic interactions of N-alkylated cinchona alkaloids, their structures feature hydroxyl moiety as hydrogen bond donor and aromatic substrates to comprise π -π interactions (Figure 2.3).
30
Figure 2. 3. Structural features of N-alkylated cinchona alkaloids.
In this chapter, N-alkylated cinchona alkaloids with 9-(chloromethyl)anthracene, known
as Corey-Lygo catalyst8, are used as chemosensors for enantiomeric excess analysis of chiral
carboxylates. Upon complex formation between negatively charged chiral carboxylates and
positively charged N-alkylated quinuclidine moiety of cinchona alkaloids these alkaloids display distinctive changes in their fluorescence intensities.9 In addition to electrostatic interaction, the secondary driving force of complex formation, hydrogen bonding and π -π interactions may also contribute to the recognition process as well. Here, we decided to utilize methyl anthracene as a reporter moiety, which also comprises rings that can form additional π-π interactions. We have chosen anthracene as a reporter for the chemosensors because of its known photophysical properties, low energy absorption (> 350 nm) and relatively high fluorescence quantum yield
(~0.25) in organic solvents. Introduction of the reporter moiety was performed by nucleophilic substitution of 9-(chloromethyl)anthracene to the nitrogen of the quinuclidine moiety (N1) (for the synthesis, see experimental section). The reaction results in chloride salts of the N-alkylated cinchona alkaloid chloride salts that did not form any complexes with carboxylates due to the 31
high charge density of the chloride anions. To overcome this complication, we performed counter anion exchange with tetrafluoroborate, which is less nucleophilic than halides. Thus, the
chemosensors (cations) obtained become more reactive toward carboxylates (anions). Once again, in addition to electrostatic interactions, the hydrogen bonding with the 9-hydroxy group and π-π interaction of aromatic substrates favor complex formation, and all these interactions are likely to have an impact on guest-host complexation. Upon complexation, presumably, the fluorescence intensity of the anthracene moiety will be changed and the magnitude of the signal change will be depended on the structural features of the resulting complexes and the possible types of the interactions between the chemosensors and the analytes.
Figure 2. 4. Structures of the carboxylate guests used in this study.
In this study, we selected a group of chiral and achiral carboxylate guests, including the
active and less active (or inactive) enantiomers of nonsteroidal anti-inflammatory drugs
(NSAIDs) commonly used for pain relief and fever reduction (Figure 2.4)10, separately. 32
Several fluorescent chemosensors for the detection of chiral carboxylates are reported, including chiral macrocycles, diacridylnaphthalene11, various BINOL derivatives12, cyclodextrins13,
boronic acids14, fluorescent oligomers15, and others16. Alas, only a few of these receptors and
chemosensors are amenable for high-throughput screening.
2.2. Results and Discussions
We synthesized four cinchona alkaloids alkylated at the quinuclidine nitrogen with
chloromethyl anthracene to obtain N-alkylated (methyl anthracene) cinchona alkaloids, namely,
N-(9-Anthracenylmethyl) quininium tetrafluoroborate (S1), N-(9-Anthracenylmethyl)
quinidinium tetrafluoroborate (S2), N-(9-Anthracenylmethyl) cinchoninium tetrafluoroborate
(S3), and N-(9-Anthracenylmethyl)cinchonidinium tetrafluoroborate (S4). Photochemical
properties, synthetic yields, and melting points of chemosensors S1-S4 are listed in Table 2.1.
Table 2. 1. Photochemical Properties of Cinchona Alkaloids
Sensor R 8 9 Φ τ (ns) MP (ºC) Yield Quinine S1 OMe R S 0.21 3.61 186 0.12
Quinidine S2 OMe S R 0.28 6.59 170 0.68
Cinchonine S3 H S R 0.25 4.89 176 0.59
Cinchonidine S4 H R S 0.20 4.19 130 0.26
Initially, the complex formation ability of N-alkylated cinchona alkaloids S1-S4 with
carboxylates was tested using electrospray ionization mass spectrometry (ESI-MS). Figure 2.5
shows the ESI-mass spectra of the complexes of the S2+Ibuprofen (A), S2+Naproxen (B), and
S3+Ketoprofen (C), with corresponding calculated spectra (insets). The molecular ions and 33
isotopic patterns of the complexes agree with the calculated spectra and isotopic patterns of the complexes with a 1:1 (chemosensor:analyte) stoichiometry.
Figure 2. 5. (A) Mass spectrum of the S2-Ibuprofen complex. Inset: Calculated isotopic pattern + for C47H50N2O3 . (B) Mass spectrum of the S2-Naproxen complex. Inset: Calculated isotopic + pattern for C48H46N2O4 . (C) Mass spectrum of the S3-Ketoprofen complex. Inset: Calculated + isotopic pattern for C50H46N2O4 .
Furthermore, the crystal structures of the S3 with enantiomers of ibuprofen were obtained by the slow diffusion of hexanes into a propionitrile solution of complexes. Figure 2.6 shows the
X-ray crystal structures of S3 with (S)-Ibuprofen (left) and (R)-Ibuprofen (right). Crystal structures of the complexes feature a combination of hydrogen bonding via the hydroxyl group at
C9 position and electrostatic interactions between the positively charged nitrogen of quinuclidine and negatively charged carboxylates of ibuprofen as we expected. Here, one can clearly see that
the orientation of the bulky benzyl group of the carboxylates differs in space, which results in the
vibrational and rotational motions of the fluorophores; which explain the difference in the
response observed for the two ibuprofen enantiomers. It is worth noting that solid-state studies of 34
the complexes may not display all the interactions that contribute to complex formation in a solution environment.
Figure 2.6. X-ray structure of the S3 complex with (S)-Ibuprofen (left) and (R)-Ibuprofen (right).
Lastly, we performed UV-Vis and fluorescence titrations of the chemosensors with
selected carboxylates (see Appendix B). There was no change observed in the absorption spectra
of the chemosensors upon addition of the analytes. Hence, fluorescence titration experiments
were carried out at the excitation wavelength of the absorption maximum of the chemosensors.
Fluorescence intensities of chemosensors S1-S4 were amplified upon addition of the analytes in
most cases. Exceptionally, only naproxen quenched the fluorescence of chemosensors S1-S4,
which was presumably due to a photo-induced electron transfer (PET) mechanism. Notably,
similar to all other carboxylates, excitation wavelength of the sensors did not overlap with the
absorption of naproxen. As an example, Figure 2.7 displays the titration of S1 with (S)-Ibuprofen
(left) and S4 with (S)-Naproxen. Whereas fluorescence intensity of S1 showed 60%
enhancement upon addition of ibuprofen, (S)-Naproxen quenched up to 75% of fluorescence
intensity of S4. Insets of Figure 2.7 show the corresponding isotherms of the titrations, 35
respectively. The fluorescence enhancement of the chemosensors was most likely due to
increased rigidity of the complexes, which limits vibrational and rotational motions of the fluorophore that otherwise would result in non-radiative decay.
1.0 1.0 ) i ) -I 0 0 -I 1.5 i 0.5
)/(I 0.5 -I)/(I
0 0.9 0 (I (I-I
6 -1 6 -1 0.0 Ka = 1.54 x 10 M 0.0 Ka = 2.22 x 10 M -5 -5 -5 -5 0.0 1.0x10 2.0x10 1.0 0.0 1.0x10 2.0x10 0.6
o [(S)-Naproxen] (M) [(S)-Ibuprofen] (M) I/I
0.5 0.3
0.0 0.0 380 400 420 440 460 480 500 380 400 420 440 460 480 500 Wavelength (nm) Wavelength (nm)
Figure 2.7. (Left) Fluorescence titration of S1 by (S)-Ibuprofen. (Right) Fluorescence titration of S4 by (S)-Naproxen. Titrations were performed in propionitrile and analytes were added as tetrabutylammonium salts ([Sensors] = 10 µM).
The fluorescence titrations of chemosensors S1-S4 indicate their cross-reactivity as well
as their ability to respond to the presence of an analyte by a change in their fluorescent intensity.
More importantly, the chemosensors show different signal responses for the enantiomers of the
chiral analytes. For example, Figure 2.8 displays the fluorescence titrations of S1 (red lines) and
S3 (blue lines) with the enantiomers of ketoprofen. One clearly can see that fluorescence
responses of chemosensors S1 and S3 significantly differ for enantiomers of the same compound.
Enantioselective behaviors of the chemosensors were not identical. 36
0.5
0.4 S1 (S)-Ketoprofen 0 S- /I
0 0.3 S1 (R)-Ketoprofen R- I-I S3 (R)-Ketoprofen 0.2 R- S3 (S)-Ketoprofen 0.1 S- 0.0
0.0 1.0x10-5 2.0x10-5 Ketoprofen (M) Figure 2.8. Fluorescence titrations of S1 (red) and S3 (blue) with the incremental amounts of (S)-Ketoprofen, and (R)-Ketoprofen.
The association constants (Kas) of the analytes were calculated based on the change in
fluorescence intensity of the chemosensors at λEm = 424 nm. Table 3 shows the Kas for all carboxylates included in this study. The association constants of the analytes suggest that in the case where enantiomers of the same compound elicit comparable signal changes, differences in the Ka values enable a reliable analysis of corresponding analytes. The significance of the affinity
of the chemosensors was explained in Chapter I (see 1.3. Sensitivity and selectivity of the
sensors). Different binding affinities and response patterns of the chemosensors for each analyte
that we tested encouraged us to perform a qualitative analysis of some of the selected
carboxylates. As chemosensors S1-S4 are not water soluble, we utilized a polymer chip platform17a, which allows us to accommodate the chemosensor array in the aqueous media. In
addition, to provide the mechanical support to the chemosensors, polymer chips also contribute significantly to the recognition process. Details of polymer chip platforms were discussed in
Chapter I (see 1.6. Polymer chips). For the preparation of polymer chips, chemosensors were 37
dissolved in 5% polyurethane polymer solution and were manually cast into multi-well glass
slides.17
-1 [a] Table 2.2. Affinity Constants (Ka, M ) Obtained by Fluorescence Titration
S1 S2 S3 S4 Acetate 3.34 × 106 5.31 × 105 7.53 × 105 > 107
Benzoate 1.44 × 106 7.65 × 105 2.17 × 106 1.08 × 106
(S)-Mandelate 2.33 × 105 4.51 × 104 8.53 × 104 1.13 × 106
(R)-Mandelate 1.40 × 105 3.23 × 104 6.87 × 104 3.17 × 105
(S)-Ibuprofen 2.22 × 106 1.17 × 106 3.41 × 106 > 107
(R)-Ibuprofen 3.29 × 106 3.91 × 106 2.58 × 106 > 107
(S)-Ketoprofen 2.12 × 106 3.00 × 105 ND[b] ND[b]
(R)-Ketoprofen 1.37 × 106 1.42 × 105 2.92 × 105 2.08 × 105
(S)-Naproxen 3.00 × 106 3.85 × 106 2.31 × 106 1.54 × 106
(R)-Naproxen 9.10 × 106 3.41 × 106 2.70 × 106 1.67 × 106
[a] The titrations are recorded in propionitrile and Kas were calculated based on the change in fluorescence intensity change at λEm = 424 nm using 1:1 model. The errors of the curve fitting were < 20%. [b] Ka could not be calculated due to the low magnitude of response.
Then, an aqueous solution of analytes was cast into the polymer chips with embedded chemosensors. The fluorescence images of chemosensors S1-S4 were captured using a UV-Vis scanner. It is worth noting that the polymer matrix forms a semi-solid environment for the chemosensors. Upon addition of the aqueous solution of analytes, water molecules are stripped away from the anions by the hydrophilic part of the polymer enabling the recognition and binding the chemosensors. To eliminate instrumentation errors, ten repetitions of each sample were performed. Resulting multivariate data from the sensor chip array based on S1-S4 were 38 evaluated using linear discriminant analysis (LDA).18 The performance of the LDA was tested using a leave-one-out routine.
Figure 2.9. Images of the sensors upon addition of the corresponding analytes recorded by UV- Vis scanner.
Before performing the qualitative analysis, we carried out a preliminary fluorescence experiment of the enantiomers of ibuprofen, ketoprofen, naproxen, and mandelate. Acetate, benzoate, and water were also added for control purposes. Figure 2.9 shows the fluorescence images of the chemosensors upon addition of the aqueous solutions of analytes. The fluorescence intensities of the chemosensors S1-S4 induced upon addition of aqueous solutions of the analytes show the same behavior as was seen in the organic solvent. More importantly, enantiomers of the similar compounds show significant differences in the fluorescence intensity of the chemosensors. Therefore, the analyte-specific response patterns of chemosensors can be used for qualitative analysis of chiral carboxylates.
39
Control Naproxen Acetate Benzoate S-Mandelate Ibuprofen R-Mandelate 20 Mandelate S-Ibuprofen R-Ibuprofen
) 0 Ketoprofen S-Naproxen
5%
.
3
( R-Naproxen
3 -20 F 20 S-Ketoprofen R-Ketoprofen -40 10
-50 ) % -25 0 .4 6 ( 0 2 F1 F (8 -10 6. 25 2% ) 50 -20 75
Figure 2.10. Linear discriminant analysis of ten carboxylates. The analytes include enantiomers of ibuprofen, ketoprofen, naproxen, and mandelate. Acetate and benzoate are also added for control purposes.
Different responses of the chemosensors to the chiral carboxylates encouraged us to perform a qualitative analysis of selected carboxylates. Figure 2.10 shows the LDA plot of the 11 carboxylates, including enantiomers of ibuprofen, ketoprofen, naproxen, and mandelate. Acetate, benzoate, and water were also added for control purposes. All (100%) correctly classified clusters of chiral carboxylates were significantly separated from each other. More importantly, the enantiomers of the same compounds showed well-resolved clusters. For example, enantiomers of ketoprofen are significantly separated. The outstanding results of this qualitative analysis encouraged us to perform enantiomeric excess (ee) analysis of the chiral carboxylates.
40
-20 Racemic- Ketoprofen 10% ee (S)-Ketoprofen 30% ee (S)-Ketoprofen 0
%) 40% ee (S)-Ketoprofen
5
.
1
( 60% ee (S)-Ketoprofen
3
F 70% ee (S)-Ketoprofen 20 80% ee (S)-Ketoprofen 90% ee (S)-Ketoprofen 20 -30 F Pure (S)-Ketoprofen 2 ( 0 0 8. 2%) ) 30 9.1% 1 (8 -20 F
Figure 2.11. Linear discriminant analysis plot of ketoprofen. Ten different enantiomeric compositions of ketoprofen were discriminated with 100% correct classification.
At first, we performed an analysis of ketoprofen. Ketoprofen is a nonsteroidal anti- inflammatory drug (NSAID) with antipyretic and analgesic effects. It is sold as a racemic mixture in the U.S. and an enantiopure (S)-Ketoprofen, with the name dexketoprofen, in the U.K.
The pharmacological properties of ketoprofen enantiomers are found to be different and enantiopure (S)-Ketoprofen is recommended.19 In this experiment, we prepared an aqueous solution of (S)-Ketoprofen of variable enantiomeric composition (10, 30, 40, 60, 70, 80, 90, and
100% ee) and a racemic mixture of the ketoprofen. Sensor arrays based on the four chemosensors (S1-S4) were able to resolve each cluster of ketoprofen solutions with a wide range of ee values and 100% accuracy (Figure 2.11). The clusters of ketoprofen show a clear trend with increasing ee values of (S)-Ketoprofen. Observing a clear trend in the semi- quantitative analysis is crucial to construct a calibration model for the quantitative analysis. The performance of the chemosensor array was tested by carrying our an additional semi-quantitative ee analysis of analytes, such as ibuprofen and naproxen (see Appendix B). Clear trends in the 41
clusters of both experiments suggest that chemosensor arrays based on S1-S4 can perform
quantitative ee analysis of diverse chiral carboxylates accurately.
100 90 80 Calibration data set Validation data set 70
)-Ketoprofen 60 S (
50 40 30 RMSEC=2.80 20 RMSECV=6.21 10 RMSEP=2.66 Predicted % ee % Predicted 0 0 102030405060708090100 Actual % ee (S )-Ketoprofen
Figure 2.12. Results of linear regression using support vector machine (SVM) afford quantitative analysis of enantiomeric excess of ketoprofen. The plots of actual vs. predicted concentrations show high accuracy of prediction for two enantiomeric compositions of ketoprofen. The root-mean-square errors (RMSEs) of calibration (C), cross-validation (CV), and prediction (P) attest to the quality of the model and prediction. Two unknown samples (red circles ●) were subsequently correctly analyzed.
Positive outcomes from the qualitative and semi-quantitative analyses of chiral carboxylates encouraged us to develop a linear regression model using obtained data from the array and the datapoints corresponding to two different ee values were excluded and used later to validate the regression analysis. The quantitative ee analysis of unknown samples was performed by regression analysis utilizing the support vector machine (SVM) algorithm.20 In the
quantitative ee analysis, 80% of the dataset was used to create a calibration model using SVM
while the remaining 20% of data used to validate the model. The performance of the analysis is
reported as the root-mean-square errors (RMSEs) of calibration (C), cross-validation (CV), and
prediction (P). Figure 2.12 shows the linear regression of the enantiomeric composition of 42
ketoprofen. The chemosensor array based on S1-S4 yielded an accurate quantitative regression
analysis of the enantiomeric composition of ketoprofen. Including the whole range of ee values
of ketoprofen (0-100% ee (S)-Ketoprofen) in the calibration model, ee analyses of two random points (red circles, ●) show very high accuracy, with RMSEP as low as 2.66.
100 Calibration data set 100 Calibration data set 90 Validation data set 90 Validation data set 80 80 70
)-Naproxen 70
)-Ibuprofen S ( S 60 60 (
50 50 40 40 30 30 20 RMSEC=1.12 20 RMSEC=1.86 10 RMSECV=1.70 Predicted % ee % Predicted 10 RMSECV=2.76 Predicted % ee % Predicted RMSEP=1.71 0 0 RMSEP=2.53
0 102030405060708090100 0 102030405060708090100 Actual % ee (S)-Ibuprofen Actual % ee (S)-Naproxen
Figure 2.13. Results of linear regression using support vector machine (SVM) afford quantitative analysis of enantiomeric excess of ibuprofen (left) and naproxen (right). The plots of actual vs. predicted concentrations show high accuracy of prediction for two enantiomeric compositions of ibuprofen and naproxen and the low error of prediction (RMSEP).
In addition to ee analysis of ketoprofen, we performed analyses of ee in ibuprofen and
naproxen samples to test the performance of the chemosensor array for analytes with similar
structures to ketoprofen. It is worth noting that the arrays based on chemosensors S1-S4 were
capable of discriminating enantiomers of the chiral compounds as well as similar structures with
small structural differences. Figure 2.13 plots the regressions of ibuprofen (left) and naproxen
(right). Similar to the ee analysis of ketoprofen, the chemosensor array was able to validate
successfully the unknown ee values of ibuprofen and naproxen. The RMSEPs of ibuprofen and
naproxen were determined as 1.71 and 2.53, respectively. 43
Finally, we successfully performed qualitative and quantitative ee analyses of carboxylates, such as ketoprofen, naproxen, and ibuprofen. However, the enantiomeric excess analysis is not concentration-dependent and the simultaneous determinations of ee and concentration of each enantiomer may be needed.21 Neither the linear discriminant function nor support vector machine algorithm is capable of revealing multiple information simultaneously. A trending method of analysis, artificial neural network (ANN)22, was performed to show the enantiomeric composition of unknown samples along with the actual concentration of each enantiomer. To utilize the ANN, we selected three concentrations of ibuprofen (the total concentration of enantiomers), namely, 0.5, 1, and 2 mM, of varying enantiomeric composition.
To examine the concentration- and ee-dependent trends revealed by the array, we performed a linear discriminant analysis. Figure 2.14 is the LDA plot of various concentrations and ee (100 to
100%) values of ibuprofen. Here, the definite trend in the concentration of ibuprofen was observed along the F1 axis, whereas the ee-dependent trend was seen in the F2 axis. A 100% correctly classified semi-quantitative analysis of ibuprofen (Figure 2.14) suggests that the array is capable of performing the quantitative analysis and ee analysis of ibuprofen. Hence, a network function of ANN was created and trained using all the data (except three data points used for validation). Table 4 shows the results of the ANN analysis of the three unknown ibuprofen solutions excluded from the training stage of analysis. The results of ANN analysis of ibuprofen show high accuracy in their ee values and the concentrations (error < 6%).
44
100% Correct Classification in Water 0.5 mM Pure (R)-Ibuprofen 0.5 mM -80% (S)-Ibuprofen 0.5 mM -40% (S)-Ibuprofen 0.5 mM -20% (S)-Ibuprofen 0.5 mM 20% (S)-Ibuprofen -60 0.5 mM 40% (S)-Ibuprofen 0.5 mM 80% (S)-Ibuprofen -40 0.5 mM Pure (S)-Ibuprofen
-20 1 mM Pure (R)-Ibuprofen ) 1 mM -80% (S)-Ibuprofen 0 2% 1 mM -60% (S)-Ibuprofen (2. 1 mM -40% (S)-Ibuprofen 3 20 F 1 mM -20% (S)-Ibuprofen 1 mM 20% (S)-Ibuprofen 40 1 mM 40% (S)-Ibuprofen 60 1 mM 60% (S)-Ibuprofen 1 mM Pure (S)-Ibuprofen 50 2 mM Pure (R)-Ibuprofen
) 2 mM -80% (S)-Ibuprofen % 2 mM -60% (S)-Ibuprofen 9 . 0 0 2 mM -40% (S)-Ibuprofen 6 ( 2 mM -20% (S)-Ibuprofen -60 1 F -40 2 mM 20% (S)-Ibuprofen -20 2 mM 30% (S)-Ibuprofen F2 0 2 mM 60% (S)-Ibuprofen (20 20 -50 .7% 2 mM Pure (S)-Ibuprofen ) 40 Figure 2.14. Linear discriminant analysis (LDA) of ibuprofen mixtures of various concentrations and ee (-100 to 100%) show perfect separation.
Table 2.3. ANN Analysis of Unknown Ibuprofen Samples[a]
Concentration Enantiomeric Excess
Actual Predicted (ANN) Error Actual Predicted (ANN) Error
0.5 0.50 0.003 20 20.12 0.002
1 0.93 0.064 20 18.61 0.023
2 1.99 0.003 60 62.10 0.105
[a] The analysis is performed using the scores from LDA results. The training set comprised 23 samples (seven repetitions) and three samples (seven repetitions) were used for prediction.
45
2.3. Conclusion
In this chapter, we demonstrated that N- alkylated cinchona alkaloids, a class of phase transfer catalysts, can be used successfully as fluorescent chemosensors for the qualitative and quantitative ee analyses of carboxylates.. First, we synthesized N-alkylated cinchona alkaloids, namely, N-(9-Anthracenylmethyl) quininium tetrafluoroborate (S1), N-(9-Anthracenylmethyl) quinidinium tetrafluoroborate (S2), N-(9-Anthracenylmethyl) cinchoninium tetrafluoroborate
(S3), and N-(9-Anthracenylmethyl)cinchonidinium tetrafluoroborate (S4). The resulting counter anion of chloride was exchanged with a less nucleophilic anion to enhance the affinity of the positively charged chemosensors toward the carboxylates. As a result, the S1-S4 cinchona salts were found capable of binding the carboxylate anions with a strong change in signal in most cases. Fluorescence intensities of the chemosensors S1-S4 were enhanced upon complex formation with carboxylate anions. In designing the chemosensor array, we considered the chiral chemosensors to form diastereomers with chiral carboxylates and the structural features of complexes results in enantiomer specific responses. Due to the formation of the diastereomeric complexes, enantiomers of the same compounds were discriminated. The analyte-specific response patterns of arrays based on S1-S4 were capable of successfully distinguishing and quantifying structurally similar chiral carboxylates and their corresponding enantiomers.
Specifically, we have demonstrated qualitative analysis enantiomers of chiral carboxylates, including several non-steroidal anti-inflammatory drugs (NSAIDs), such as enantiomers of ibuprofen, ketoprofen, and naproxen. Sensor arrays based on four fluorescent sensors (S1-S4) were capable of differentiating between a range of enantiomeric excess analyses of ibuprofen, ketoprofen, and naproxen (RMSEP = 2.66, 1.71, and 2.53, respectively). In addition to ee analysis of chiral carboxylates, we performed ANN experiments to reveal ee 46
values of samples along with their concentrations. Simultaneous determination of ee values and
concentration of ibuprofen shows less than 6% error of prediction.
These results suggest that fluorescent cinchona organocatalysts are well-suited
chemosensors S1-S4 for high-throughput ee analysis of a number of drugs, or drug intermediates, and high-throughput screening probes for the enantiomeric purity of chiral drugs.
2.4. Experimental Section
2.4.1. General
Chemosensors S1-S4 were synthesized using standard laboratory techniques. All starting materials were purchased and used as received. Compounds S1-S4 were prepared according to the literature procedures.23 1H- and 13C-NMR (APT) spectra were recorded using a Bruker®
Avance IITM 500 MHz UltraShieldTM (Bruker Corporation, Mass., USA) spectrometer at 25°C.
Solutions for optical measurements were prepared using freshly distilled propionitrile. Optically
dilute solutions (0.1 A) were used for all photophysical experiments. Fluorescence emission
spectra were acquired using an Edinburgh single photon counting spectrofluorometer (FLSP
920). Fluorescence emission spectra were recorded between 380 nm and 500 nm. The band
passes of both excitation and emission monochromators were set to 1.0 nm. The emission from
probes was scanned in 2 nm steps. The dwell time was adjusted to 0.30 s. Scans were taken
under ambient room conditions. Guest titrations were performed in propionitrile. Titration
isotherms were constructed from changes in the fluorescence maximum at 424 nm. Binding
constants were calculated by nonlinear least-square methods using an equation for a 1:1 binding
model.24 All the calculations were performed using Origin software package.25
47