Intermolecular Interactions between Eosin Y and Caffeine Using 1 H-NMR Spectroscopy

DETECHIP has been used in testing analytes including caffeine, cocaine, and tetrahydrocannabinol (THC) from marijuana, as well as date rape and club drugs such as flunitrazepam, gamma-hydroxybutyric acid (GHB), and methamphetamine. This study investigates the intermolecular interaction between DETECHIP sensor eosin Y (DC1) and the analyte (caffeine) that is responsible for the fluorescence and color changes observed in the actual array. Using 1 H-NMR, 1 H-COSY, and 1 H-DOSY NMR methods, a proton exchange from C-8 of caffeine to eosin Y is proposed.


Introduction
DETECHIP, a highly selective and sensitive molecular sensor that produces color and fluorescence changes in the presence of analytes, was recently developed. A quick, sensitive, and selective detection system is required for many applications, such as alerting security officers to the presence of explosives or their precursors, preincident monitoring/ screening for homeland security purposes [1] such as weapons of mass destruction, and detection and quantification of doping compounds in competitive sports [2][3][4][5]. The method that is currently most widely used for the detection of such substances is gas chromatography-mass spectrometry (GC-MS) [6,7]. However this method requires a skilled operator and cannot be easily miniaturized. Current screening reagents for abused narcotics like flunitrazepam (often used for date rape, assault, or theft) [8], methylephedrine, caffeine, nicotine, and others include immunoassays [9], ion trap mobility spectrometry [9][10][11], wet colorimetric assays [12][13][14], spot tests such as Marquis [13], Scott Drug Testing Company drug tests (http://www.scottcompany.com), or the b-Glucuronidase Drug Analysis Bundle (Sigma-Aldrich) and Magnotech technology testing [15].
DETECHIP uses an array of sensors that can be used for identification of drugs and other molecules by fluorescence and color changes [16][17][18]. This method relies on molecular interactions between the analyte molecules and the DETECHIP sensors [18]. Unlike other color tests which provide a single "yes" or "no" response, DETECHIP gives multiple simultaneous responses in the form of color and fluorescent changes using two different buffers, allowing users to quickly characterize suspect materials. Figure 1 shows an example of a 96-well plate DETECHIP assay. Eight sensors (DC1-DC8) are added to the rows in the plate in two different buffers (A + B), and the analytes are tested alongside a control in the columns of the plate.
While DETECHIP has worked well for many analytes, little is known regarding the intermolecular interaction between the sensors and the analytes. As further development of DETECHIP is in progress, it is important to determine the reasons behind the color or fluorescence changes observed. Understanding the underlying mechanism of sensor interactions with analytes allows for the selection and preparation of better sensors for future DETECHIP prototypes. Color or fluorescence changes may be associated with structural changes of the sensor. Structural changes of the sensors may be due to tautomerism, conformational changes, or the formation of a more complex supramolecular structure.
Xanthene dyes, such as eosin Y, exhibit different tautomeric structures with different protolytic forms, either proton "on" or "off" depending on pH as illustrated in Figure 2 [19]. The color associated with the dye derives from the quinoid structure of the anionic form [20,21]. The pKa of 2.02 in Figure 2 is associated with the phenolic proton. The pKa values may vary as reported by Bartistela et al. [19]. At pH = 7, the most prevalent form of eosin Y is the diionic form at a pKa of 3.8 [19,20]. In DETECHIP assays, the pH is buffered at pH = 7, and therefore the color changes must be a function of the chemical environment due to the presence of the analyte. Caffeine exhibits strong color and fluorescence changes with eosin Y. Eosin Y is therefore a good candidate for the investigation of the sensor-analyte intermolecular interaction. Figure 3 shows the structures of eosin Y in its di-ionic form and Caffeine.
We used proton nuclear magnetic resonance ( 1 H-NMR) to investigate the intermolecular interaction between the sensor, eosin Y, and the analyte, Caffeine. NMR is one of the most effective tools for the structural elucidation of complex organic molecules [22,23]. Structure determination and intermolecular bonding interactions studies can be achieved by use of 1 H-NMR and two-dimensional proton correlation spectroscopy ( 1 H-COSY) NMR. A COSY pulse-sequence is usually used to identify spins of the same isotope that are coupled to each other. It consists of a single RF pulse (p1) followed by the specific evolution time (t1) followed by a second pulse (p2) followed by a measurement period (t). The twodimensional spectrum that results from the COSY experiment shows the frequencies for the isotope under study, most commonly hydrogen ( 1 H) along both axes. This method allows us to determine which protons on the analyte are correlated to the protons of the analyte.
Experiments determining the presence of multiple components in a solution have recently gained popularity by the determination of the diffusion coefficient (D) using twodimensional diffusion ordered spectroscopy (DOSY) NMR [24,25]. The measurement of diffusion is carried out by observing the attenuation of the NMR signals during a pulsedfield gradient experiment. The degree of attenuation is a function of the magnetic gradient pulse amplitude and occurs at a rate proportional to the diffusion coefficient (D) of the molecule. A "normal" NMR spectrum is obtained in the horizontal axis, while in the vertical axis, the diffusion coefficient is obtained. Since components of different sizes have unique diffusion coefficients, the vertical axis provides diffusion peaks which can help determine how many components are in a mixture. DOSY allows us to evaluate if more than two components are in solution when the analyte is added to the sensor, such as the presence of products that form as a result of a chemical reaction.

Reagents and Materials
The sensors and Caffeine were purchased from Sigma-Aldrich, St Louis, MO. Dibasic potassium phosphate (K 2 HPO 4 ) was purchased from VWR International, West Chester, PA. The NMR deuterated solvent (D 2 O) and standard sodium-3-trimethylsilylpropionate d 4 (TSP) were purchased from Cambridge Isotope Laboratories, Cambridge, MA. All reagents were used as received.

NMR Measurements
All measurements were performed on a 400 MHz Bruker Avance III spectrometer equipped with a 5 mm z-gradient probe at room temperature. Chemical shifts (δ) are reported relative to the methyl signal of TSP (0 ppm).
The 1D 1 H spectra were acquired with presaturation to reduce the residual water. COSY data was acquired using gradient pulses for selection, purge pulses, and presaturation to reduce the water signal. DOSY experiments were acquired with a stimulated echo sequence incorporating a longitudinal eddy delay (LED), bipolar gradient pulses, and two spoil gradients.

Sample Preparation
An NMR buffer solution was prepared at 400 mM of dibasic potassium phosphate (K 2 HPO 4 ) in deuterium oxide (D 2 O) at pH = 7 using deuterium chloride (DCl) solution. The NMR standard used was 0.1% TSP. Using the buffer solution, 12.5 mM Caffeine and 20 mM eosin Y solutions were prepared, from which the NMR samples were made. The eosin Y and Caffeine samples were each used without further dilution. The final concentrations in the eosin Y-Caffeine mixture sample were 6.25 mM Caffeine and 10 mM eosin Y.

1 H-NMR Spectra Predictions
1 H-NMR spectra were calculated using ChemBioDraw 12.0 software (CambridgeSoft, Cambridge, UK) and used in the peak assignments in the experimental spectra.
To determine the change in spectral intensity, the 1 H-NMR spectra were fitted using Peakfit program (Seasolve software, Framingham, MA).

Ultraviolet-Visual Spectroscopy Experiments
Absorbance measurements were carried out in a Varian Cary 50 ultraviolet-visual spectrometer (UV-Vis). Both samples were made using 400 mM dibasic potassium phosphate buffer solution at pH = 7. A 20 d835df07M solution sample of eosin Y was made by a 1000-fold dilution of 20 mM eosin Y. A 1 : 1 mixture of 20 mM eosin Y and 12.5 mM Caffeine was diluted 1000-fold to make the mixture sample.

Results and Discussion
1 H-NMR spectra were obtained for Caffeine, eosin Y, and their mixture. Figure 4 shows the 1 H-NMR spectrum of caffeine. The peaks corresponding to the N-methyl protons are labeled a, b, and c at 3.33, 3.50, and 3.94 ppm, respectively. The C8 proton is labeled d and shows a peak at 7.90 ppm. Changes in position and/or intensity of these peaks, would signify changes in the structure of Caffeine and were monitored in the Caffeine-eosin Y mixture sample. Similarly, changes were monitored on the peaks of the eosin Y.
The 1 H-NMR spectrum of eosin Y is given in Figure 5. The corresponding proton peaks are labeled a and b at 7.47 (s), c at 7.17 (d), d at 7.60 (t), e at 7.66 (t), and f at 7.80 (d) ppm.
The 1 H-NMR spectrum of the Caffeine-eosin Y mixture is shown in Figure 6. There is a slight downfield shift in the proton peaks for eosin Y: a + bfrom 7.49 to 7.47 ppm and c 7.30 to 7.19 ppm. Table 1 shows the chemical shifts for the eosin Y protons in the eosin Y sample and in the eosin Y-Caffeine mixture sample. The downfield shift in the eosin Y proton signals is due to a reduction in electron density, caused by the transfer of a proton from Caffeine. The peak for Caffeine d proton (C-8), 7.90 ppm, in the Caffeine spectrum is shifted to 7.67 ppm and is drastically reduced in intensity in the eosin Y-Caffeine mixture spectrum. The eosin Y-Caffeine mixture spectrum was fitted using Peakfit software to enable comparing the area ratio of the peak at 7.90 ppm to the one at 3.35 ppm in the Caffeine spectrum and the area ratio of the peak at 7.69 ppm to that at 3.39 ppm in the Caffeine-eosin Y mixture spectrum. A 5-fold reduction is observed, implying that there is loss of a C-8 proton from Caffeine.
The chemical shifts for the Caffeine protons in the two samples are given in Table 2 and can be seen in Figure 7, where the upfield shifts of N-methyl proton peaks for Caffeine in the eosin Y-Caffeine mixture are evident. This is characteristic of an increase in electron density around the protons due to the formation of a Caffeine anion.
A potential mechanism for the abstraction of the Caffeine C-8 proton to eosin Y is given in Figure 8. The negatively charged carboxylate deprotonates the C-8 hydrogen which results in the protonated eosin Y and the deprotonated Caffeine.
The COSY 1 H-NMR spectrum in Figure 9 shows off-diagonal peaks for the C-8 proton in Caffeine correlated with "b" proton peaks. While this shows that the C-8 protons are still attached to the Caffeine molecules, the reduction in intensity confirms that some of the protons have been transferred.
To test for the presence and identity of different components in the mixture such as new products formed due to a chemical reaction between eosin Y and Caffeine, or dimerization between eosin Y molecules [26], a DOSY NMR experiment was obtained. The horizontal axis shows the normal proton NMR spectrum. The vertical axis represents the diffusion constant: each component diffuses at a certain rate thus, the separation.
Results from the DOSY NMR, Figure 10, show that the mixture comprises 3 components: water, eosin Y, and Caffeine. This argues that a supramolecular complex with hydrogen bonding or dimerization may not have occurred under these experimental conditions.
Based on the COSY results in Figure 9 and the DOSY results in Figure 10, Caffeine and eosin Y are not reacting to form a new molecule. No new diffusion peak appeared in the DOSY spectrum which would indicate that a product was formed due to a chemical reaction between Caffeine and eosin Y. However, as seen in Figure 7, the proton labeled "d" on C-8 disappears when eosin Y is added to Caffeine. This disappearance of the peak at 7.9 ppm can be explained by the following: (1) the peak shifted and overlaps with eosin Y peaks and is not visible anymore, (2) some of the Caffeine molecules lost the proton which transferred to eosin Y, or (3) there is a hydrogen bond interaction between the Caffeine and eosin Y. At pH = 4, eosin Y exists as dianion in equilibrium with the monoanion protonated at the carboxylic group as shown in Figure 2 [19]. In order to investigate this proton transfer, 1 H-NMR spectra were obtained at lower pH values. The spectrum observed at pH = 4 shows a pentet-like signal at approximately 7.6 ppm that is similar to the one observed in the eosin Y-Caffeine mixture spectrum in Figure 6. This illustrates that in the mixture, eosin Y exists in the protonated form.
UV-Vis spectra were obtained at 0.75 d835df07M eosin Y and mixtures of 0.75 d835df07M eosin Y and Caffeine at various concentrations (from 0.42 mM to 11.25 mM) in 400 mM dibasic potassium phosphate solution are shown in Figure 11. In the presence of Caffeine, the λ max is red shifted by up to 8 nm depending on Caffeine concentration. This is consistent with the color change observed in the 96-well plate DETECHIP assay (Figure 1). This red shift increases with increasing Caffeine concentration, very low eosin Y concentrations (less than 10 mM). Dimerization of eosin Y is not expected [26] at these concentrations of eosin Y and a proton transfer mechanism between Caffeine and eosin Y is feasible. 1 H-NMR results indicated that the negatively charged oxygen on eosin Y may deprotonate C-8 of Caffeine. This potentially created a negatively charged Caffeine intermediate and the change of conjugation in eosin Y led to a visible color change. The literature reports pK a values for the proton on C-8 in Caffeine between 10.4 [27] and 14.0 [28]. Thus, this proton is very labile and can be abstracted easily at pH = 7. Similar hydrogen exchange reactions in nucleic acids and similar heterocycles have been observed in previous studies [29].

Conclusion
DETECHIP is based on fluorescence and color changes from which a unique identification has been developed for many drugs. The results in this study have shown that some of the color and fluorescence changes observed in DETECHIP assays may be a consequence of a proton transfer reaction between the sensors and the analyte molecules or donor-acceptor hydrogen bond kind of interaction. Further detailed studies are in progress for the interaction between other DETECHIP sensors and other analytes.  Protolytic equilibriums of eosin Y with the most representative tautomeric structure.      Mechanism for proton transfer from Caffeine to eosin Y.