Improved 19 F { 1 H } Saturation Transfer Difference Experiments for Sensitive Detection to Fluorinated Compound Bound to Proteins

The 19F{1H} saturation transfer difference (STD) method was improved for sensitive 19F detection using a human serum albumin-diflunisal complex. Because NMR (nuclear magnetic resonance) experiments with 19F detection are feasible for the selective detection of fluorinated compounds, more sensitive NMR methods are required to be developed for purposes of practicality. The present research focused on the investigations of 19F{1H} STD pulse techniques and experimental parameters, leading to the development of detection methods with higher sensitivity.


Introduction
NMR (nuclear magnetic resonance) spectroscopy can be a useful method for analyzing protein-ligand interactions in the solution state.Moreover, various NMR-based screening methods to observe 1 H ligand signals have been proposed, such as NOE-pumping [1], saturation transfer difference (STD) [2], water-ligand observed via gradient spectroscopy (WaterLOGSY) [3,4], and reverse NOE-pumping [5] experiments.NMR-based screening methods have also been applied to 19 F detection [6][7][8][9].The characteristic properties of 19 F are a 100% natural abundance and high sensitivity for observation.The inclusion of fluorine atoms in a drug molecule generally alters its chemical properties and biological activities, and can influence the interaction with its target [10].The fluorine substitution in ligand molecules is expected to affect binding owing to the change of intra-and intermolecular forces [11].For the purpose of developing more practical NMR-based screening methods with 19 F detection, 19 F{ 1 H} saturation transfer difference (STD) experiments with higher sensitivity were optimized using the complex of diflunisal (Figure 1) and human serum albumin (HSA).Diflunisal contains two fluorine atoms per molecule, and is a nonsteroidal anti-inflammatory drug.Since the X-ray crystal structure of a diflunisal-HSA complex is available (pdb: 2BXE), this complex could be a suitable model system for studying the molecular interactions of 1 H and 19 F using NMR spectroscopy.

Introduction
NMR (nuclear magnetic resonance) spectroscopy can be a useful method for analyzing proteinligand interactions in the solution state.Moreover, various NMR-based screening methods to observe 1 H ligand signals have been proposed, such as NOE-pumping [1], saturation transfer difference (STD) [2], water-ligand observed via gradient spectroscopy (WaterLOGSY) [3,4], and reverse NOEpumping [5] experiments.NMR-based screening methods have also been applied to 19 F detection [6][7][8][9].The characteristic properties of 19 F are a 100% natural abundance and high sensitivity for observation.The inclusion of fluorine atoms in a drug molecule generally alters its chemical properties and biological activities, and can influence the interaction with its target [10].The fluorine substitution in ligand molecules is expected to affect binding owing to the change of intra-and intermolecular forces [11].For the purpose of developing more practical NMR-based screening methods with 19 F detection, 19 F{ 1 H} saturation transfer difference (STD) experiments with higher sensitivity were optimized using the complex of diflunisal (Figure 1) and human serum albumin (HSA).Diflunisal contains two fluorine atoms per molecule, and is a nonsteroidal anti-inflammatory drug.Since the Xray crystal structure of a diflunisal-HSA complex is available (pdb: 2BXE), this complex could be a suitable model system for studying the molecular interactions of 1 H and 19 F using NMR spectroscopy.

Instrumentation and Chemicals
All of the NMR spectra were recorded at 20 • C on a Varian 600 MHz NMR system equipped with an HFX probe (Varian, Palo Alto, CA, USA).Diflunisal and HSA were purchased from Sigma-Aldrich (Tokyo, Japan).A 600-µL volume of solution containing 0.05 mM HSA and 2.5 mM diflunisal was prepared in 100% 2 H 2 O.

NMR Spectroscopy
The experimental parameters of the 19 F{ 1 H} STD experiment were as follows: data points = 16,384, spectral width of 19 F = 6818 Hz, number of scans = 2048 or 4096.The saturation times for the selective excitation of protein were arrayed in the range of 0.5-3.0s.The on-resonance frequency of 1 H was 0.6 ppm, and the off-resonance frequencies were arrayed at 6000 Hz increments in the range of −50 to 50 ppm for comparison.The shaped pulses of I-BURP-1, I-BURP-2, Q 3 , G3, and Gaussian were made using VnmrJ software (Agilent Technologies, Santa Clara, CA, USA).The pulse widths (H 1 /2) were 4.67 ms (715 Hz), 4.87 ms (1133 Hz), 3.77 ms (900 Hz), 3.86 ms (637 Hz) and 0.866 ms (1426 Hz), respectively.

Results and Discussion
An optimization of the 19 F{ 1 H} STD experiments was carried out based on the 19 F{ 1 H} STD pulse sequence (Figure 2a) used in our past study [12].Regarding the selective excitation of protein signals, several pulse schemes, such as a single rectangular pulse and various shaped pulses, were evaluated using the pulse sequences shown in Figure 2. The saturation time was set to 1.5 s in each experiment for comparison.The repetitive time n shown in Figure 2b was adjusted to set the total saturation time.

Instrumentation and Chemicals
All of the NMR spectra were recorded at 20 °C on a Varian 600 MHz NMR system equipped with an HFX probe (Varian, Palo Alto, CA, USA).Diflunisal and HSA were purchased from Sigma-Aldrich (Tokyo, Japan).A 600-μL volume of solution containing 0.05 mM HSA and 2.5 mM diflunisal was prepared in 100% 2 H2O.

NMR Spectroscopy
The experimental parameters of the 19 F{ 1 H} STD experiment were as follows: data points = 16384, spectral width of 19 F = 6818 Hz, number of scans = 2048 or 4096.The saturation times for the selective excitation of protein were arrayed in the range of 0.5-3.0s.The on-resonance frequency of 1 H was 0.6 ppm, and the off-resonance frequencies were arrayed at 6000 Hz increments in the range of −50 to 50 ppm for comparison.The shaped pulses of I-BURP-1, I-BURP-2, Q 3 , G3, and Gaussian were made using VnmrJ software (Agilent Technologies).The pulse widths (H1/2) were 4.67 ms (715 Hz), 4.87 ms (1133 Hz), 3.77 ms (900 Hz), 3.86 ms (637 Hz) and 0.866 ms (1426 Hz), respectively.

Results and Discussions
An optimization of the 19 F{ 1 H} STD experiments was carried out based on the 19 F{ 1 H} STD pulse sequence (Figure 2a) used in our past study [12].Regarding the selective excitation of protein signals, several pulse schemes, such as a single rectangular pulse and various shaped pulses, were evaluated using the pulse sequences shown in Figure 2. The saturation time was set to 1.5 s in each experiment for comparison.The repetitive time n shown in Figure 2b was adjusted to set the total saturation time.The S/N ratios calculated using the VnmrJ software (Agilent Technologies) are shown in each spectrum (Figure 3).The most sensitive spectrum was acquired by an application of a Gaussian shaped pulse (Figure 3f).The sensitivities were almost equal among the rectangular pulse, I-BURP-2, and Q 3 shaped pulses.In setting the rectangular pulse, a higher power was expected to result in a higher sensitivity; however, it could also lead to serious damage of the probe.Considering the sensitivity and maintenance of the probe, the Gaussian shaped pulse was expected to be the most feasible pulse scheme in the 19 F{ 1 H} STD experiments.
The S/N ratios calculated using the VnmrJ software (Agilent Technologies) are shown in each spectrum (Figure 3).The most sensitive spectrum was acquired by an application of a Gaussian shaped pulse (Figure 3f).The sensitivities were almost equal among the rectangular pulse, I-BURP-2, and Q 3 shaped pulses.In setting the rectangular pulse, a higher power was expected to result in a higher sensitivity; however, it could also lead to serious damage of the probe.Considering the sensitivity and maintenance of the probe, the Gaussian shaped pulse was expected to be the most feasible pulse scheme in the 19 F{ 1 H} STD experiments.In the STD experiments, to observe ligand signals bound to proteins, on-resonance frequency was set on the protein (e.g., methyl region) and off-resonance frequency was set outside the signal region for reference (e.g., −20 ppm).Although the 1 H off-resonance frequency was generally set without careful consideration, it was arrayed in the range of −50 to 50 ppm to observe the impact on the sensitivity (Figure 4).The rectangular and Gaussian shaped pulses, corresponding to the previous [12] and currently optimized pulse schemes, were employed for comparison.The 1 H off-resonance frequency was optimized to be δ 1 H = −30 ppm, indicating that its careful setting was crucial for obtaining the sensitive spectra.The expanded 19 F{ 1 H} STD spectra, acquired using the past parameters [12] and the optimized parameters, are shown in Figure 4c,d.
In the STD experiments, to observe ligand signals bound to proteins, on-resonance frequency was set on the protein (e.g.methyl region) and off-resonance frequency was set outside the signal region for reference (e.g.−20 ppm).Although the 1 H off-resonance frequency was generally set without careful consideration, it was arrayed in the range of −50 to 50 ppm to observe the impact on the sensitivity (Figure 4).The rectangular and Gaussian shaped pulses, corresponding to the previous [12] and currently optimized pulse schemes, were employed for comparison.The 1 H off-resonance frequency was optimized to be δ 1 H = −30 ppm, indicating that its careful setting was crucial for obtaining the sensitive spectra.The expanded 19 F{ 1 H} STD spectra, acquired using the past parameters [12] and the optimized parameters, are shown in Figures 4c and Figures    In the 1 H{ 1 H} STD experiments, the STD build-up curves could be obtained at various saturation times [17].The slope of the STD build-up curve at a saturation time of 0 s was obtained by fitting to the monoexponential equation: STD = STDmax(1 − e (−ksat × t) ), where STD stands for the STD signal intensity at saturation time t; STDmax is the maximal STD intensity at long saturation times; and ksat stands for the observed saturation constant.The values of ksat × STDmax are generally used for epitope mapping [17].The 19 F{ 1 H} STD build-up curves were obtained using a rectangular pulse and Gaussian shaped pulses for 1 H saturation (Figure 5).The saturation times were arrayed in the range of 0.5-3.0s, and the values of the STD effect were normalized by referencing the signal of F2' with the largest STD effect as 100%.The normalized values of F4' using a rectangular pulse and Gaussian shaped pulses were 21.7 and 28.7%, respectively, indicating that F2' made a more significant contribution to binding with HSA (human serum albumin).In the 1 H{ 1 H} STD experiments, the STD build-up curves could be obtained at various saturation times [17].The slope of the STD build-up curve at a saturation time of 0 s was obtained by fitting to the monoexponential equation: STD = STD max (1 − e (−ksat × t) ), where STD stands for the STD signal intensity at saturation time t; STD max is the maximal STD intensity at long saturation times; and k sat stands for the observed saturation constant.The values of k sat × STD max are generally used for epitope mapping [17].The 19 F{ 1 H} STD build-up curves were obtained using a rectangular pulse and Gaussian shaped pulses for 1 H saturation (Figure 5).The saturation times were arrayed in the range of 0.5-3.0s, and the values of the STD effect were normalized by referencing the signal of F2' with the largest STD effect as 100%.The normalized values of F4' using a rectangular pulse and Gaussian shaped pulses were 21.7 and 28.7%, respectively, indicating that F2' made a more significant contribution to binding with HSA (human serum albumin).

Conclusions
The pulse schemes for 1 H saturation were evaluated, and the Gaussian shaped pulse was considered to be most feasible for the purposes of sensitivity and maintaining the instruments.The

Conclusions
The pulse schemes for 1 H saturation were evaluated, and the Gaussian shaped pulse was considered to be most feasible for the purposes of sensitivity and maintaining the instruments.The 1 H off-resonance frequency affected sensitivity, suggesting the importance of its optimization.Repetition of the shaped pulses could lead to some artifacts or phase distortion, which would be the reason why subtle differences in the 1 H off-resonance frequency affected sensitivity, besides off-resonance effects.The proposed technique with 19 F detection is expected to be a useful NMR screening method for fluorinated compounds.

Figure
Figure 4.The 19 F{ 1 H} STD spectra acquired using (a) a rectangular pulse and (b) Gaussian shaped pulses for 1 H saturation.The 1 H on-resonance frequency was set to 0.6 ppm, and the 1 H off-resonance frequencies were arrayed in the range of −50 to 50 ppm.The expanded spectra acquired using (c) a rectangular and (d) Gaussian shaped pulses.The 1 H off-resonance frequencies were (c) −20 ppm and (d) −30 ppm.

4 .
Figure 4.The 19 F{ 1 H} STD spectra acquired using (a) a rectangular pulse and (b) Gaussian shaped pulses for 1 H saturation.The 1 H on-resonance frequency was set to 0.6 ppm, and the 1 H off-resonance frequencies were arrayed in the range of −50 to 50 ppm.The expanded spectra acquired using (c) a rectangular and (d) Gaussian shaped pulses.The 1 H off-resonance frequencies were (c) −20 ppm and (d) −30 ppm.

Figure 4 .
Figure 4.The 19 F{ 1 H} STD spectra acquired using (a) a rectangular pulse and (b) Gaussian shaped pulses for 1 H saturation.The 1 H on-resonance frequency was set to 0.6 ppm, and the 1 H off-resonance frequencies were arrayed in the range of −50 to 50 ppm.The expanded spectra acquired using (c) a rectangular and (d) Gaussian shaped pulses.The 1 H off-resonance frequencies were (c) −20 ppm and (d) −30 ppm.

Figure 5 .
Figure 5.The 19 F{ 1 H} STD spectra acquired with various saturation times using (a) a rectangular pulse and (b) Gaussian shaped pulses for 1 H saturation.