Rapid Measurement of Heteronuclear Coupling Constants in Complex NMR Spectra

The NMR analysis of fluorine-containing molecules, increasingly widespread due to their importance in pharmaceuticals and biochemistry, poses significant challenges. Severe peak overlap in the proton spectrum often hinders the extraction of critical structural information in the form of heteronuclear scalar coupling constants, which are crucial for determining pharmaceutical properties and biological activity. Here, a new method, IPAP-FESTA, is reported that drastically simplifies measurements of the signs and magnitudes of proton–fluorine couplings. Its usefulness is demonstrated for the structural study of the steroidal drug fluticasone propionate extracted from a commercial formulation and for assessing solvent effects on the conformational equilibrium in a physically inseparable fluorohydrin mixture.


IPAP-FESTA pulse sequence
The detailed pulse sequence for Bruker implementation is shown in Figure S1. Narrow and wide filled rectangles represent hard 90° and 180° pulses, respectively. The durations of the hard 90° pulses (p1) for each sample are given in Section 2. The two selective 180° iBURP shaped pulses (with duration p14) and the 180° rSNOB shaped pulse (with duration p12) were used to select the specified 19 F and 1 H coupled spins, respectively. These pulses were generated using Wavemaker (see Section 1e). If only one 19 F resonance is present, the selective 19 F 180° pulses can be replaced by hard 19 F 180° pulses, as heteronuclear selectivity is not required.
TOCSY transfer was achieved using the DIPSI-2 isotropic mixing scheme with a mixing time (d9) of 100-150 ms dependent on the spin system being analysed. The trapezoids on either side of the DIPSI-2 element indicate low-power 180° chirp pulses used to suppress the effects of zero quantum coherences. Their durations were set to 10 and 30 ms, respectively. The field gradient pulse labelled G1 in the pulse sequence is used to select the desired coherence transfer pathway, with an amplitude of 23 % of the maximum gradient strength and a duration of 1 ms. Gradient pulse G3 is a homospoil gradient with a duration of 1 ms and amplitude of 79 %, to dephase any transverse magnetization. G1 and G3 have a smoothed square shape (gpnam1 = gpnam2 = SMSQ10.100). Gradient pulses G1 and G3 are followed by recovery delays (d16 and d18) of 200 µs. Gradient pulses G0 and G10 are applied during chirp pulses to suppress the effects of zero-quantum coherences and are followed by recovery delays of 100 µs. The delays Δ1 (equal to half of the delay Δ detailed in the main manuscript) are set to 1/(4JHF) for AnX and AnX3 spin systems, and 1/(8JHF) for AnX2 spin systems, where n is the number of equivalent protons, and JHF is the heteronuclear coupling constant between the specified proton and fluorine. The recommended minimum phase cycle is 16 steps; the full phase cycle is given in Table S1. Table S1: Phase cycle for IPAP-FESTA described in Figure S1.

Using Wavemaker to generate shaped pulses
Wavemaker is a pulse shaping function within TopSpin® that allows shaped pulses to be generated without extensive manual set-up by the user. Wavemaker reads the relevant lines within the pulse program, and creates the shaped pulses required from the specified parameters. In this pulse sequence, there are 3 shaped pulses designed by Wavemaker: -An rSNOB shape (sp12, channel F1) to be used as the 1 H 180° refocusing pulse. -Two inversion iBURP pulses (sp14 and sp15, channel F2) to invert the specified 19 F resonance.
Wavemaker is automatically downloaded as part of the current standard TopSpin® installation. Any Wavemaker updates are available in the Bruker User Library. Alternatively, each shaped pulse can be set up manually. a. Set ZGOPTNS in the Acquisition window ('ACQPARS') to '-DIPAP' and the acquisition dimension to '2D'. To acquire only the IP or AP MODO-FESTA spectrum, set ZGOPTNS to '-DIP' or '-DAP' respectively, and the acquisition dimension to '1D'. b. Choose an appropriate DIPSI-2 duration (30 -200 ms) for the TOCSY transfer to observe the full spin system under study. c. Set up the zero quantum filter (ZQF) chirp pulses. 2 4. Acquire the 2D IPAP-MODO-FESTA experiment to obtain IP and AP spectra of the 1 H NMR spectra which only contain protons that belong to the chosen 19 F-1 H spin system.

Processing of IPAP-FESTA data using TopSpin
Below is the recommended procedure for processing IPAP-FESTA data within TopSpin version 3.1.6. Minor modifications may be required to process IPAP-FESTA data in other versions of TopSpin.
1. Extract the 1D IP and AP-FESTA spectra from the 2D data set. For example, use the command 'ft 1 1001' to extract the 1 st data set (IP-FESTA) to processing number 1001, and the command 'ft 2 1002' to extract the 2 nd data set (AP-FESTA) to processing number 1002. Ensure that the IP-and AP-FESTA spectra used identical processing parameters (for example, SI, weighting functions, etc.) 2. Overlay the extracted IP-and AP-FESTA spectra. Determine an appropriate scaling factor k for the AP spectrum to ensure minimum cross-talk artifacts in the resulting IPAP spectra. 3. For addition/subtraction processing, use the 'split' macro to obtain the IP+AP and IP-AP spectra.
a. The macro will prompt the user to input a scaling factor (k) to be applied to the AP-FESTA spectra. b. Note that the data may be processed differently depending on the system and 'split' macro version used. c. Measure the signs and magnitudes of JHF using the sign and direction of displacement of the 19 Fα and 19 Fβ coupled 1 H signals. Note that only the relative signs of JHF values can be obtained, not absolute signs. 2 JHF couplings are, however, usually large and positive and can be used as a reference.
1.5. Processing of IPAP-FESTA data using Mathematica A Mathematica® notebook is supplied (see Section 5) to extract the IP and AP-FESTA spectra, apply an optimised scaling factor k (between 0.8 and 1.2) to each specified JHF coupled 1 H signal, and accurately extract the signs and magnitudes of JHF couplings. The default scaling factor (k) range of 0.8 -1.2 is typically sufficient for correcting relaxation loss in small molecules, but can be modified if required (see below).
The Mathematica notebook requires 7-9 inputs from the user: -The scaling factor (k) range to be applied to the AP-MODO-FESTA spectra (0.8 < fratio < 1.2). This range is stated in two lines of code starting with 'res1' and 'res2'. -A comma-separated list of the centres in ppm of the regions to be fitted (shifts Running the notebook should extract all the JHF coupling constants from the raw data, as well as the corresponding scaling factors k and root mean squared (RMS) uncertainties.

Computational procedures
The conformational preferences for cis-and trans-3-fluoro-2-hydroxytetrahydropyran diastereomers (Scheme S2) were computationally evaluated by varying the orientation of the O-H group relative to the six-membered ring system. The potential energy curves were scanned at the B3LYP/cc-pVDZ level of theory, varying the C3-C2-O-H dihedral angle from 0 to 360° in steps of 10° for the four significant conformers ( Figure S2). For each of the conformers the C3-C2-O-H dihedral angle was fixed and the rest of the molecule was allowed to relax during the geometry optimization calculations.
The geometries for the minima in the curves were then fully reoptimized and the Gibbs energy differences for the stable conformers were calculated at the M062X/aug-cc-pVTZ level of theory, in isolated phase (xyz cartesian coordinates listed below), and including solvent effects (CHCl3, DMSO and H2O) using the SMD model. The Gibbs energies for the stable conformers are listed in Section 3c, while calculated values for the n JFH coupling constants for the most stable conformers were obtained at the B3LYP functional level employing the EPR-III basis set for the hydrogen, carbon, and fluorine atoms and cc-pVDZ basis set for oxygen. The EPR-III basis set was not used for oxygen to save computational time as no couplings involving 17 O were calculated or measured. All calculations were performed in the Gaussian 16 program. 3 H IPAP-FESTA spectra were recorded with 2048 transients in an experiment time of 4 h 23 min. The 1 H signal at 5.41 ppm was selected using an rSNOB shaped pulse (4.1 ms, 450 Hz effective bandwidth) and the 19 F signal at -187.4 ppm was selected by using iBURP shaped pulses (2.8 ms, 1650 Hz effective bandwidth). These shaped pulses were designed using Wavemaker. A 100 ms mixing period was used for the DIPSI-2 (d9) element. When generating IPAP-FESTA by addition/subtracting processing, a scaling factor (k) of 1.072 was applied to the AP spectra.

Sample preparation and experimental data
Scheme S1: Molecular structures of major excipient 2-phenylethanol (left) and fluticasone propionate (right).

Fluorohydrin
Scheme S2: Equilibrium between the cis-and trans-isomers of fluorohydrin, each with 2 main conformers. Fast (grey arrows) and slow (black arrows) relate to the conformer exchange rate on the chemical shift timescale.
The cis-and trans-3-fluoro-2-hydroxytetrahydropyran compounds were previously prepared 4 in our (CT) laboratory. According to the procedure described in the literature, 5 a solution containing nitromethane (50 mL), water (10 mL) and 3,4-dihydro-2H-pyran (1.83 mL, 0.02 mol) was cooled to 3 °C and selectfluor (10 g, 0.03 mol) was added. The reaction was stirred at 25 °C for 12 h. After that, the solution was refluxed (110 °C) for 1 h and concentrated under reduced pressure to remove the nitromethane. The crude residue was dissolved with dichloromethane (50 mL), followed by the addition of sodium bicarbonate (5%) solution. The organic layer was washed with brine, dried with magnesium sulfate, and concentrated under reduced pressure. The crude product was chromatographed over SiO2 (70-230 mesh) using 7:3 hexane/ethyl acetate as eluent, and a mixture of cis-and trans-3-fluoro-2-hydroxytetrahydropyran diastereomers was isolated. The three fluorohydrin-based mixtures used in this work contained 0.68 M of 3-fluoro-2-hydroxytetrahydropyran in DMSO-d6 (99.9 atom % D, Sigma-Aldrich), 0.58 M in D2O (99.9 atom % D, Sigma-Aldrich), and 0.74 M in CDCl3 (99.9 atom % D, Sigma-Aldrich). Fluorohydrin spectra were recorded at 298 K on a Bruker Avance III 500 MHz spectrometer (Bruker Biospin) with a 5 mm BBO probe with standard transmitter routing, equipped with a z-gradient coil with a maximum nominal gradient strength of 53.  H IPAP-FESTA spectra were recorded with 32 transients, except for trans-fluorohydrin in CDCl3, which was recorded with 16 transients. The chemical shift of the 1 H and 19 F signal selected in each experiment is shown in Table 3. 1 H signals ware selected using an rSNOB shaped pulse (4.65 ms, 450 Hz effective bandwidth) and 19 F signals using iBURP shaped pulses (32.5 ms, 1650 Hz effective bandwidth. These shaped pulses were designed using Wavemaker. A 150 ms mixing period was used for the DIPSI-2 (d9) element. When generating IPAP-FESTA by addition/subtracting processing, a scaling factor (k) was applied to the AP spectrum (Table S4). Prior to Fourier transformation of IPAP data, zero-filling to 131072 complex points was applied, and Gaussian weighting with LB of -0.01 Hz and GB of 0.013 was used.    Due to severe peak overlap, the JHF values for 4 J5aF and 3 J4aF were measured using HD-HAPPY-FESTA (data not shown). In proton labels, 'a' corresponds to axial and 'b' to equatorial. Extracted JHF values are shown in (e). The signal at 4.95 ppm in (e) was scaled by a factor of 0.5. The JHF value of +11.3 Hz corresponding to 3 J4aF was measured manually due to signal overlap. In proton labels, 'a' corresponds to axial and 'b' to equatorial.     Extracted JHF values are shown in (e). The JHF value for 4 J5bF was measured from the HD-HAPPY-FESTA spectrum due to severe peak overlap. In proton labels, 'a' corresponds to axial and 'b' to equatorial.
2.3. Comparison of JHF couplings measured using IPAP-FESTA and using 1 H and 1 H{ 19 F} NMR spectra To confirm the accuracy of the JHF values obtained from IPAP-FESTA NMR data, they were compared to the JHF values of all the well-resolved signals in the 1 H NMR spectra of the fluorohydrin and fluticasone samples. Excellent agreement is seen between JHF couplings measured using IPAP-FESTA and with 1 H and 1 H{ 19 F} NMR spectra, as shown in Table S5. 2.4. Cross-talk artifacts and the scaling factor k in IPAP data Small systematic differences in peak intensity between the IP and AP data sets result in unwanted cross-talk artifacts when generating IPAP spectra by addition/subtraction processing. It is recommended to scale one spectrum, here the AP spectrum, by a scaling factor k to minimise such artifacts and obtain cleaner spectra. To highlight the importance of the scaling factor k, the IP and AP FESTA 1 H NMR spectra of trans-fluorohydrin in DMSO-d6 were processed using the TopSpin AU macro 'split', using a k factor of 1 (no scaling) and 1.126 (optimized scaling factor). Results are shown in Figure S14, where cross-talk artifacts for H3 in the IP-AP spectrum are reduced from a signal intensity of 7.5 % to 0.8 % when using the optimized scaling factor. Cross-talk artifacts for H4b are also highlighted, but signal intensities are not reported due to a small spectral overlap with the wanted signal.

Fluticasone propionate JHF values
All JHF6 couplings for fluticasone propionate obtained using IPAP-FESTA data are shown in Table S6. The Mathematica notebook used to measure JHF values and the raw data are available for download at DOI: 10.48420/21975908.  Figure S4.

Fluorohydrin conformer populations determination from Gibbs free energy
The Excel document used to calculate the conformer ratios for Table S8-S16 is available for download at DOI: 10.48420/21975908. The Boltzmann factor is the ratio of the population of each conformer to the population of the lowest energy conformer. The Gibbs energy (Tables S8-S15) was used to weight the populations of the conformations because it includes the entropic contribution to the Boltzmann distribution. Contributions from translational, electronic, rotational, and vibrational motion were considered.. For a single molecule, one of the most essential approximations throughout this analysis is that all the equations assume non-interacting particles and apply only to an ideal gas. To compute all contributions, the partition function for the corresponding component of the total partition function is used. Since the vibrational partition function depends on the frequencies, you must use a structure that is either a minimum or a saddle point. Therefore, the thermal energy (E therm ), enthalpy (H=E therm +PV), entropy (S) and Gibbs free energy (G=H-TS) were computed at the default conditions of 298.15 K and 1 atmosphere, which can be obtained from frequency calculation. 6

JHF values from DFT calculations and values extracted from experimental IPAP-FESTA data
The computational uncertainty in both the populations and J-values calculated via DFT is not discussed here as it is outside of the scope of this work. To determine the computational uncertainty, calculations using different methods (functional and basis set) and the evaluation of different solvent effect models (implicit and explicit) must be performed, which suggest using molecular dynamics approaches. Some coupling constants are population dependent, such as the 3 JHF couplings, and deviation/variation between experimental and theoretical is more pronounced. This is not the case for 2 JHF since in both conformations coupled nuclei are in a similar orientation, i.e., bond angle and bond distance are similar.    S19: JHF values obtained from DFT calculations and extracted from experimental IPAP-FESTA data ( Figure S6) for cis-fluorohydrin in CDCl3. From DFT calculations, the ratio of ax-eq conformer: eq-ax conformer was 39:61.     Figure S10) for trans-fluorohydrin, in DMSO-d6. From DFT calculations, the ratio of ax-ax conformer: eq-eq conformer was 39    3.5. Repeatability of JHF couplings measured from experimental IPAP-FESTA data

Cis isomer and its conformations
To confirm the repeatability of IPAP-FESTA for measuring the magnitude and sign of JHF couplings, ten repetitions of IPAP-FESTA data acquisition and processing were performed for cis-fluorohydrin in DMSO-d6, for eight multiplets. The data were processed with two different time-domain weighting functions (no weighting, and 1 Hz worth of exponential weighting), using a modified version of the Mathematica notebook in Section 5. The results are summarised in in Tables  S25 and S26. The standard deviations of the 16 JHF coupling values ranged from 0.3 to 1.1 mHz. This confirms the excellent repeatability of the IPAP-FESTA data processing, potentially giving mHz accuracy in the weak coupling limit. Importantly, IPAP methods will show systematic errors in the presence of strong couplings; these have not been analysed and will be spin system dependent. The repeatability data and the Mathematica notebook used are available from the DOI: 10.48420/21975908.