Fast ¹⁹F Magic-Angle-Spinning NMR Crystallography for Structural Characterization of Fluorine-Containing Pharmaceutical Compounds

Abstract

Fluorine-containing compounds comprise twenty to thirty percent of all commercial drugs, and the proportion of fluorinated pharmaceuticals is rapidly growing. While magic angle spinning (MAS) NMR spectroscopy is a popular technique for analysis of solid pharmaceutical compounds, fluorine has been underutilized as a structural probe so far. Here we report a fast (40 kHz) MAS ¹⁹F NMR approach for structural characterization of fluorine-containing crystalline pharmaceutical compounds at natural abundance, using the antimalarial fluorine-containing drug mefloquine as an example. We demonstrate the utility of 2D ¹⁹F-¹³C and ¹⁹F-¹⁹F dipolar-coupling-based correlation experiments for ¹⁹F and ¹³C resonance frequency assignment, which permit identification of crystallographically inequivalent sites. The efficiency of ¹⁹F-¹³C cross-polarization (CP) as well as the effect of ¹H and ¹⁹F decoupling on spectral resolution and sensitivity were evaluated in a broad range of experimental conditions. We further demonstrate a protocol for measuring accurate interfluorine distances based on 1D DANTE-RFDR experiments combined with multi-spin numerical simulations.

Abstract Graphic:

INTRODUCTION

Fluorinated compounds are becoming increasingly dominant in medicinal chemistry¹. Fluorine-containing molecules comprise 20% of all pharmaceuticals and 30% of newly approved drugs^2–4, and according to current estimates, over 150 commercial drugs contain fluorine atoms or fluoroalkyl groups, including several of the most-prescribed and/or most-profitable on the market^2–4. Structural characterization of pharmaceuticals relies on crystallography and NMR spectroscopy; fluorine NMR is becoming increasingly important in this regard. ¹⁹F is a 100% naturally abundant spin 1/2 nucleus and possesses a high gyromagnetic ratio, making it a very attractive probe for NMR applications. ¹⁹F resonance frequencies are also exquisitely sensitive to the local structural and electronic environment around the fluorine nuclei, exhibiting a chemical shift range of over 300 ppm^5,6.

While ¹⁹F NMR has been extensively applied in the analysis of fluorinated organic and biological molecules in soIution^7–13, MAS NMR studies remain relatively scarce^14–26. This, in part, is due to i) strong dipolar couplings that result in broad lines, requiring high-power ¹H and ¹⁹F decoupling with specialized hardware, and ii) relatively low sensitivity in heteronuclear polarization transfer experiments in the common regime of magic angle spinning (MAS) frequencies, below 30 kHz. Increasing the MAS frequencies to 40 kHz and above results in significant improvements in sensitivity and resoIution^24–26, yielding substantially narrowed ¹⁹F lines, even in the absence of ¹H decoupling^19,20,25. Furthermore, homonuclear ¹⁹F-¹⁹F recoupling experiments, such as 2D radio frequency driven recoupling (RFDR) and delays alternating with nutation for tailored excitation (DANTE)-RFDR, are effective and can be used for interfluorine distance measurements, including in multi-spin systems^15,18,19,24. In particular, long-range ¹H-¹⁹F and ¹⁹F-¹⁹F interatomic distances²⁷ together with ¹⁹F-¹³C contacts,^13,21,28,29 are indispensable parameters³⁰ in structure elucidations.

Here we report on a fast MAS ¹⁹F NMR crystallography approach for structure determination of crystalline fluorine-containing pharmaceutical compounds at natural abundance. This integrated approach was developed by us for structural analysis of active pharmaceutical ingredients and provides detailed information on the 3D structure. Our methodology was established using the antimalarial drug mefloquine³¹ (Fig. 1) whose X-ray structure is known. Combining 2D homonuclear ¹⁹F-¹⁹F and heteronuclear ¹⁹F-¹³C dipole-coupling-based experiments with Density Functional Theory (DFT) calculations, complete assignments of ¹⁹F and ¹³C chemical shifts were obtained, revealing three crystallographically inequivalent fluorine sites in the unit cell of mefloquine. We show that at MAS frequencies of 40 kHz and above, low-power double-quantum cross polarization (CP) is optimal for achieving high ¹⁹F-¹³C polarization transfer efficiencies and for recording short- and long-range ¹⁹F-¹³C correlations. The effects and the efficiency of ¹H and ¹⁹F decoupling on ¹⁹F-¹³C heteronuclear correlation (HETCOR) spectra were evaluated and a comprehensive analysis of multi-spin effects was performed. Magnetization exchange curves recorded in DANTE-RFDR experiments allowed us to extract accurate intra- and intermolecular fluorine-fluorine distances.

Figure 1.

a) Chemical and 3D structure of mefloquine. b) Arrangement of mefloquine molecules in the crystal. The three types of inequivalent molecules are colored in purple, orange and cyan. The fluorine atoms are shown as spheres. The 8CF₃ and 2CF₃ groups are colored in light purple and light cyan, respectively.

It is well known that polymorphs and solvates can impact the therapeutic performance for certain drugs,³² and patenting new polymorphs after the original drug is a common strategy for extending drug commercial lifecycle.³³ The fast MAS ¹⁹F NMR crystallography approach, exemplarily demonstrated herein for crystalline mefloquine, allows for quick and efficient characterization of polymorphs. We envision that it will be broadly adopted for analysis of active pharmaceutical ingredients in pharmaceutical formulations of unknown structures.

RESULTS

¹⁹F MAS NMR Spectra: MAS Frequency and Decoupling

¹⁹F MAS spectra of mefloquine acquired at MAS frequencies of 10, 40, and 60 kHz are shown in Fig. 2. Very strong signals emerge in only a single scan, and the signal-to-noise ratios (SNR) for spectra acquired with 32 scans at MAS frequencies of 10, 40, and 60 kHz, without and with ¹H decoupling, are remarkably high: 307/561, 1215/1655, 1535/2120, respectively. At a MAS frequency of 10 kHz, a manifold of 8 spinning sidebands was observed for each of the two broad peaks with isotropic chemical shifts of 16.2 and 8.8 ppm. The corresponding line widths for the center bands are 922 and 745 Hz in the absence of ¹H decoupling. When ¹H Spinal-64 decoupling at an RF B₁ field of 90 kHz was applied, the lines became somewhat narrower, 843 and 317 Hz, respectively. The upfield resonance exhibited asymmetric lineshape but remained unresolved. The ¹⁹F T₁ are 1.4 s and 2.5 s for the downfield and upfield peaks, respectively, while ¹H T₁ are 4.7–5.0 s.

Figure 2.

¹⁹F MAS spectra of mefloquine acquired at MAS frequencies of 60 kHz (top two traces), 40 kHz (middle two traces), and 10 kHz (bottom two traces), with or without SW_f-TPPM ¹H decoupling at the RF field strength, as indicated next to each spectrum. The blue traces are expansions around the isotropic peaks (marked with asterisks). The signal-to-noise ratios (SNR) are indicated next to each trace. The spectra were acquired at 11.7 T using a 1.3 mm HFX MAS probe, averaging 32 scans.

The spectra at 40 kHz and 60 kHz provided in Fig. 2 reveal 5 distinct resonances, with isotropic chemical shifts of 16.2 ppm, 15.9 ppm (a shoulder), 15.6 ppm, 8.8 ppm, and 8.2 ppm. Dramatic improvements in resolution are seen when increasing the MAS frequencies to 40 and 60 kHz. At 40 kHz, the line widths of the downfield and upfield resonances are 440 and 230 Hz without decoupling; with ¹H SW_f-TPPM decoupling at the RF field of 10 kHz, those are 378 and 143 Hz, respectively. At 60 kHz without ¹H decoupling, the line widths are 322 and 171 Hz; those in the presence of 15 kHz ¹H SWf-TPPM are 300 and 117 Hz, respectively. Taken together, these results unequivocally demonstrate that fast MAS frequencies are critical from both sensitivity and resolution, and that, even at 60 kHz, ¹H decoupling is necessary. Conversely, even with high-power (90 kHz) ¹H decoupling, the resolution is very poor at 10 kHz MAS and the SNR is 4-fold lower than at 60 kHz MAS.

¹⁹F-¹³C Cross-Polarization: Optimal Conditions for Fast MAS Experiments

¹H-¹³C and ¹⁹F-¹³C CPMAS spectra of mefloquine are shown in Fig. 3. All carbon resonances are detected in the ¹H-¹³C CPMAS spectrum (Fig. 3a) as well as the non-¹H decoupled ¹⁹F-¹³C spectrum (Fig. 3b), except for C14, which is farthest away from the two CF₃ groups. In the ¹⁹F-¹³C CPMAS experiments, the polarization transfer efficiency between fluorine and the aromatic carbons is similar to that in the ¹H-¹³C CPMAS experiment, with a 0.6 signal-to-noise ratio (SNR) per square root of scans.

Figure 3.

a) ¹H-¹³C and ¹⁹F-¹³C CPMAS spectra of mefloquine. The spectra were acquired at 20.0 T using a 1.9 mm HX MAS probe, averaging 512 scans and 1920 scans and CP contact time of 1.4 and 6 ms, respectively. The ¹⁹F-¹³C spectrum was acquired without decoupling. b) ¹⁹F-¹³C CPMAS spectra of mefloquine without decoupling acquired (top trace), with ¹⁹F decoupling (middle trace), and with ¹H and ¹⁹F decoupling (bottom trace). The spectra were acquired at 16.4 T using a 1.6 mm HFXY MAS probe, averaging 512 scans; the CP contact time was 7 ms. The MAS frequency was 40 kHz in every case. Assignments for individual carbon signals are shown in the different spectra.

Effects and efficiency of ¹H and ¹⁹F decoupling was evaluated for different decoupling sequences (Fig. 3b–d). For fluorine decoupling, high-power π-pulses were introduced every rotor period³⁴ and proton decoupling employed the low power XiX scheme³⁵, applied simultaneously with fluorine decoupling. Fluorine decoupling (Fig. 3c) increased the signal intensity and narrowed the CF₃ carbon resonances, as expected. At the same time, with only fluorine decoupling, non-CF₃ carbon signal intensities are reduced by different degrees relative to the ¹H-decoupled spectra. When proton and fluorine decoupling is applied simultaneously (Fig. 3d), the spectrum exhibits improved resolution throughout.

¹⁹F-¹³C polarization transfer efficiencies were assessed by recording CPMAS spectra for different Hartmann-Hann matching conditions with radiofrequency (rf) fields varied from 5 kHz to 120 kHz. Both double-quantum (DQ) and zero-quantum (ZQ) conditions were examined. We found that several DQ and ZQ conditions led to efficient CP transfers for 15 kHz or 35 kHz ¹⁹F radio frequency (rf) fields. DQ transfers appear to be more efficient than their ZQ counterparts, with the low-power CP transfers considerably more efficient than the high-power conditions (Fig. S1 and Table S1 of the Supporting Information). The highest transfer efficiency was achieved for first-order DQ CP with 15 kHz ¹⁹F and 25 kHz ¹³C rf fields (Fig. S1, Fig. S2a–b).

In addition, ¹⁹F-¹³C CPMAS spectra were recorded with different contact times, to obtain magnetization buildup profiles for different resonances. Contact times of 6 ms or higher are necessary to cross-polarize the aliphatic carbons while intensities for the aromatic carbons reach their maximum at 7 ms and decrease thereafter (Fig. S2c–g and Fig. S3). Overall, our results indicate that it is advantageous to record ¹⁹F-¹³C CPMAS spectra with contact times of both 7 and 10 ms. Setting the contact time to 10 ms in the 2D HETCOR experiments permits detection of cross-peaks corresponding to long-range correlations, as discussed below.

¹H-¹³C and ¹⁹F-¹³C HETCOR: Resonance Assignments and Long-Range Correlations

¹³C resonances were assigned using 2D ¹H-¹³C and ¹⁹F-¹³C HETCOR spectra and DFT calculations (Fig. S4 of the Supporting Information); all chemical shifts are summarized in Table 2. While most of the ¹³C signals could be assigned unambiguously based on the calculated frequencies, remaining ambiguities were resolved by analyzing the 2D ¹H-¹³C HETCOR spectra recorded with CP contact times of 0.5 ms and 1.1 ms. Strong peaks present in the 2D HETCOR spectrum recorded with a contact time of 0.5 ms correspond to correlations between carbons and their directly bonded protons, while correlations between carbons and protons separated by two or three bonds appear in the spectrum with 1.1 ms contact time (Fig. 4a). The resonances of the 2CF₃ and 8CF₃ groups give rise to intense signals and could be assigned with confidence on the basis of the 2D ¹⁹F-¹³C HETCOR spectrum (contact time of 1 ms), despite their partial overlap with those from C8 and C4a (Fig. 4b). All carbon resonance assignments were validated from the contact-time dependencies of 1D ¹H-¹³C and ¹⁹F-¹³C CPMAS spectra (Fig. S3a, Fig. S4b). For instance, carbon resonances of CF₃ groups as well as C2, C8a and C7, which are all within three bonds of the ¹⁹F atoms, are clearly observed in spectra with 1 ms contact time. In contrast, signals corresponding to carbons distant from the fluorine atoms, are relatively weak (C5 and C6) or missing (C4). Similarly, resonances of aromatic carbons, such as C8a, C8 and C4a, have low intensities in the 1 ms ¹H-¹³C CPMAS spectra. Assignments of C15, C16 and C17 resonances are tentative as the corresponding signals partly overlap. To the best of our knowledge, the ¹³C and ¹⁹F chemical shifts of mefloquine have not been reported to date. For several carbon atoms, including C2, C3, C4, C6 and the carbon of 8CF₃ group, several sets of distinct peaks are observed, which correspond to the crystallographically inequivalent positions of these atoms in the unit cell (Fig. 4c). We also note that the application of ¹H and ¹⁹F decoupling results in significantly improved spectral resolution, consistent with our observations in the 1D ¹⁹F-¹³C and ¹H-¹³C CPMAS experiments (Fig. 4b–d and Fig. S5 of the Supporting Information).

Table 2.

MAS NMR Experimental and DFT Calculated ¹³C Isotropic Chemical Shifts and Interatomic Distances in Mefloquine

Carbon atom	δ13C (ppm) MAS NMR		δ13C (ppm) DFT	Distance to 19F (Å)
	a	b		2CF3	8CF3
C4	147.5	148	157.2	4.5*	5.2
C2	144.3	145	150.9	2.3	4.1
C8a	143.6	143.2	146.7	4.1	3
C7	129.7		134.3	6.3	3.4
C5	126.8		132.4	6.2	4.9
C6	125.2	124.7	128.7	6.8	4.6
8CF3	123.0	122.6	126.2	5	1.3
C8	122.2#		126.6	5	2.4
C4a	120.8#		125.1	4.9	4.3
2CF3	120.1		122.4	1.3	4.7
C3	114.3	114.1	114.2	3.4	5.2
C11	67.7		71.7	5.8	6.7
C12	59.3	58.8	58.4	6.3	7
C14	47		44.9	8.5	9
C15	23.1, 22.4	(22.3–23.6)	24.8	8.1	9.1
C17	21.7, 21.4	(21.3–22.0)	21.4	5.7	7
C16	21.0, 20.1	(19.8–20.9)	20.3	6.8	7.7

^*Resonances were assigned on the basis of ¹⁹F-¹³C CPMAS, ¹H-¹³C CPMAS and ¹⁹F-¹³C HETCOR spectra. Correlations detected in 2D FC-HETCOR with the 7.0 ms CP contact time are highlighted in green and those only detected with the 1ms CP contact in light green.

Figure 4.

a) 2D ¹H-¹³C HETCOR spectra acquired with CP contact times of 0.5 ms (yellow) and 1.1 ms (blue). b)-d) 2D ¹⁹F-¹³C HETCOR spectra acquired with CP contact times of 1ms and 7 ms (b), 7 ms and 10 ms with ¹⁹F decoupling (c), and with a CP contact time of 7 ms with ¹H and ¹⁹F decoupling (d). The rf fields for ¹⁹F and ¹³C were 15 kHz and 25 kHz (b) or 50 kHz (c), respectively. The spectrum shown in d) was acquired at 16.4 T with 8 scans. All other spectra were acquired at 20.0 T with 256 scans. The MAS frequency was 40 kHz. Carbon assignments are indicated. e) Short- and long-range intramolecular ¹⁹F-¹³C distances in the mefloquine crystal structure consistent with correlations in the HETCOR spectra for 1ms and 7 ms contact times. f) Intermolecular ¹⁹F-¹³C distances < 6 Å in the crystal lattice.

The ¹⁹F-¹³C multiple-bond correlations were extracted from a series of ¹⁹F-¹³C HETCOR spectra acquired with different contact times. Resonance assignments labeled in Fig. 4b–d were made based on the following considerations: correlations to C3, C4 and the aliphatic carbons are only observed for the ¹⁹F signal at 8.8 ppm; therefore, this signal is assigned to the 2CF₃ group, which is close to the aliphatic carbons. By exclusion, the ¹⁹F signal at 15.9 ppm is therefore assigned to the 8CF₃ group. This assignment is consistent with the DFT calculations, which indicate that the 2CF₃ group is downfield-shifted and that the ¹⁹F chemical shift difference between the 2CF₃ and 8CF₃ groups is 8.1 ppm.

To unequivocally distinguish between short- and long-range ¹⁹F-¹³C correlations, we carefully examined the 2D HETCOR spectra acquired with different CP contact times. As can be noted, ¹⁹F-¹³C correlations corresponding to atom pairs separated by 3 or more bonds are only detected when the CP contact time is 7 ms or longer (Fig. 4c and Fig. S5 of the Supporting Information), while ¹⁹F-¹³C correlations involving directly bonded carbons and carbons within 3 bonds of the fluorine are observed with contact time as short as 1 ms. To observe fluorine correlations with the piperidine ring carbons, such as C12 and C15–17, contact times of 9 ms or longer are necessary. Overall, ¹⁹F-¹³C correlations, corresponding to intra- and intermolecular distances of as long as 6.8 Å were observed. All experimental correlations described above are consistent with the structure of mefloquine and map unambiguously to specific distances (Fig. 4e,f).

¹⁹F-¹⁹F RFDR: Assignment of Resonances from Crystallographically Inequivalent Sites

At room temperature, the fluorine atoms of the CF₃ groups are expected to give rise to a single ¹⁹F resonance due to motional averaging^36,37. However, multiple ¹⁹F chemical shifts were observed for carbon and fluorine atoms of the 2CF₃ and 8CF₃ moieties in the ¹⁹F-¹³C HETCOR spectra (Fig. 4b–d), also consistent with the 1D ¹⁹F MAS spectra (Fig. 2). The presence of several distinct fluorine signals for each trifluoromethyl group can be explained by the crystal structure of mefloquine: there are three molecules in the unit cell, which result in inequivalent atom positions for the CF₃ groups. To confirm this interpretation, we performed 1D ¹⁹F DANTE excitation experiments^38,39, where selective DANTE irradiation was applied at different ¹⁹F frequencies (Fig. 5a–d). When DANTE pulses were applied on the 8CF₃ resonance at 15.9 ppm (Fig. 5c) and 2CF₃ at 8.8 ppm (Fig. 5d), multiple resolved resonances can be seen at each of the two positions, with resolution increasing with increased DANTE interpulse delay from 2 to 4 rotor cycles. With DANTE selective excitation, at least two signals for the 2CF₃ moiety and three for the 8CF₃ moiety were observed. The line widths of the three 8CF₃ resonances are 0.42 ppm, 0.35 ppm and 0.67 ppm, and the overall width of the 8CF₃ peak in the absence of the DANTE excitation is 1.13 ppm.

Figure 5.

a)-d) Pulse sequence (a), 1D ¹⁹F MAS NMR spectrum (b) with ¹⁹F DANTE selective excitation of the 8CF₃ resonance at 15.9 ppm (c) and 2CF₃ resonance at 8.8 ppm (d). Spectra were acquired at 20.0 T, with a MAS frequency of 40 kHz, averaging 16 scans and a DANTE interpulse delay of 4 rotor periods. e) 2D ¹⁹F-¹⁹F RFDR spectra with increased mixing times from 1.6 ms to 30.4 ms. Intramolecular and intermolecular correlations are shown in black and orange, respectively. f) The intra- and intermolecular interfluorine distances in crystal structure of mefloquine. The notation for each ¹⁹F-¹⁹F distance is identical to the corresponding correlation 2D ¹⁹F-¹⁹F RFDR spectra. The fluorine atoms in CF₃ groups are shown in light cyan. g) Left panels: 2D ¹⁹F-¹⁹F RFDR spectra acquired at the MAS frequencies of 40 kHz (left), 50 kHz (middle), and 60 kHz (right). The RFDR mixing times were 3.2 ms (gray) and 8.0 ms (dark purple). Right panel: the 1D traces of the 2D spectra with 8 ms RFDR mixing extracted at 8.61 ppm for MAS frequencies of 40 kHz (cyan), 50 kHz (black) and 60 kHz (orange). The peak widths are indicated in the slices.

In addition, 2D ¹⁹F-¹⁹F RFDR spectra were acquired with RFDR mixing times ranging from 1.6 ms to 30.4 ms, resulting in multiple resolved ¹⁹F signals for each CF₃ moiety and multiple sets of cross peaks between 2CF₃ and 8CF₃ groups (Fig. 5e). Interestingly, the RFDR buildup profiles for the cross peaks are very different. For example, cross peaks labeled “1” and “2” in the spectra were detected with mixing times as short as 1.6 ms, whereas other signals only appear at longer mixing times. To corroborate the presence of these multiple ¹⁹F signals in the RFDR spectra, the experiments were performed at MAS frequencies of 40 kHz, 50 kHz and 60 kHz with the RFDR mixing times set to 3.2 ms and 8.0 ms. As shown in Fig. 5g, the spectral resolution increases considerably with the MAS frequency, and the individual peaks become well resolved at 60 kHz. The line widths for the diagonal peaks are 0.35, 0.30, and 0.27 ppm at 40, 50, and 60 kHz, respectively. Furthermore, as a result of enhanced transfer efficiencies at higher spinning frequencies, the peak intensities increase by approximately 25% with every 10 kHz increase in the MAS frequency.

To assign the individual ¹⁹F resonances to inequivalent CF₃ groups in the unit cell, DFT calculations were performed for the cluster of 8 molecules shown in Fig. 1b. The resulting calculated shifts for molecules a, b and c are distinct (Table 1), consistent with the experimental findings. The assignments of ¹⁹F chemical shifts belonging to inequivalent groups and the corresponding intra-and intermolecular ¹⁹F-¹⁹F correlations are consistent with the observed relative cross peak intensities in the RFDR spectra, acquired with different mixing times (Fig. 5e) and the buildup profiles for the different cross peak intensities (Fig. 5f). Specifically, intramolecular correlations between the 2CF₃ and 8CF₃ moieties are detected at a mixing time as low as 1.6 ms, whereas correlations for medium- and long-range intermolecular contacts appear only at mixing times of 4.0 ms or longer. Intermolecular 8CF₃-8CF₃ correlations between inequivalent CF₃ groups are strong at the 1.6 ms mixing time, consistent with the short 3.5 Å distances.

Table 1.

MAS NMR Experimental and DFT Calculated ¹⁹F Isotropic Chemical Shifts for Mefloquine

Crystallographically inequivalent molecules	Functional group	δiso (ppm) MAS NMR	δiso (ppm) DFT
a	2CF3	8.8	7.4
a	8CF3	15.6	15.5
b	2CF3	8.2	1.5
b	8CF3	16.1	16.1
c	2CF3	8.6	1.6
c	8CF3	15.8	14.9

Taken together, these above findings indicate that ¹⁹F chemical shifts are very sensitive to the slight differences in local environments, allowing for the observation and assignment of inequivalent fluorine positions.

Measurement of Accurate Interfluorine Distances

We determined ¹⁹F-¹⁹F distances using ¹⁹F DANTE-RFDR magnetization exchange profiles. The original version of the experiment by McDermott and coworkers¹⁵ was modified such that a DANTE excitation pulse train was applied on either 2CF₃ or 8CF₃. followed by a non-selective homonuclear mixing using RFDR (Fig. 6a). The resulting DANTE-RFDR spectra are shown in Fig. 6b,c and Fig. S8 of the Supporting Information. Increased intensity for the CF₃ signals is clearly observed for increasing RFDR mixing times from 0.8 ms to 16 ms, while without RFDR mixing, no intensity buildup is observed.

Figure 6.

a) Pulse sequence for the 1D RFDR experiment with ¹⁹F DANTE-excitation. b),c) 1D ¹⁹F-¹⁹F DANTE-RFDR spectra with DANTE 90° selective excitation applied to the ¹⁹F resonances of 2CF₃ (b) and 8CF₃ (c), respectively. Spectra acquired with RFDR mixing times of 1.6 ms and 8.0 ms are shown in black and blue, respectively. The position of the DANTE excitation is shown with arrows and the resonances to which the magnetization was transferred by asterisks. d),e) Left: Experimental and simulated ¹⁹F-¹⁹F DANTE-RFDR magnetization exchange curves for the 8CF₃ (d) and 2CF₃ (e) resonances. The experimental data points are shown as black circles, the simulated curves, as dashed lines. In d), the 2CF₃ spins were excited by DANTE pulses and magnetization was transferred to the 8CF₃ spins during RFDR mixing period. In e), DANTE excitation was applied to the 8CF₃ resonances and magnetization was transferred to the 2CF₃ groups. Errors in the data points as defined by the standard deviation of the noise in a region of over 10 ppm are smaller than the size of the circles. The RMSDs of the simulated DANTE-RFDR magnetization exchange curves (dashed lines) are 0.008 (orange, the 2CF₃ — 2CF₃ distance is 7.2 Å) and 0.014 (black, the 2CF₃ — 2CF₃ distance is 7.4 Å) for 8CF₃ (d) and 0.014 for 2CF₃ (e). Right: Sets of interfluorine distances used in the 5-spin simulations, see also Table 3.

¹⁹F-¹⁹F DANTE-RFDR magnetization exchange profiles (Fig. 6d,e) are clearly dominated by multi-spin effects, similar to our recent findings for crystalline difluorobenzoic acids²⁴. Numerical simulations were performed to extract interfluorine distances. In order to account for multi-spin effects, we constructed a large number of 2-spin, 3-spin, 4-spin, 5-spin, and 6-spin systems using combinations of the interfluorine distances from the crystal structure as a set of starting values, which were then varied to assess the resulting error (Fig. S9–10 of the Supporting Information). Clearly, 5 spins are necessary and sufficient to reproduce the experimental magnetization exchange curves (Fig. S9). Analysis of the simulated magnetization exchange profiles for different distance combinations showed that the interfluorine distance between the spin excited by the DANTE sequence and the spin to which the magnetization is transferred dominate the exchange profiles. This is clearly seen from the comparison of the simulation of a 2-spin system with a 2CF₃-8CF₃ distance of 5.4 Å and that of a 5-spin system with four different 2CF₃-8CF₃ distances (with the shortest one being 5.4 Å) and one additional 2CF₃-2CF₃ distance (Fig. S10b). The shortest distance determines the initial buildup rate while the addition of other longer distances influences the detailed oscillation profile after the first intensity maximum. Taking longer distances into account distinctly improves the agreement of the fit with the experimental curve. The overall accuracy of the interfluorine distances extracted from DANTE-RFDR experiments is 0.1–0.2 Å for distances shorter than 7.0 Å and 0.2–0.4 Å for distances in the 7.0 – 11.0 Å range, similar to our recent findings for difluorobenzoic acids²⁴.

We also carried out simulations using several 5-spin systems in which the intramolecular distance between 2CF₃ and 8CF₃ groups as well as several intermolecular distances were included. Excellent agreement is observed between the simulated and experimental curves if intra- and intermolecular ¹⁹F-¹⁹F distances shorter than 11 Å are taken into account (Fig. 6d–e). The best agreement for the 2CF₃ exchange curve is observed for a 2CF₃-8CF₃ intramolecular distance of 5.5 6, and intermolecular distances of 8.0 Å, 9.9 Å and 10.5 Å, with an added 8CF₃-8CF₃ distance of 7.0 Å included as well. Likewise, the best agreement for 8CF₃ group was obtained with identical 2CF₃-8CF₃ distances and two additional intermolecular 2CF₃-2CF₃ distances of 7.4 Å and 10.6 Å. Remarkably, the 2CF₃-2CF₃ distance determined by the simulation is 7.2 6, only 0.2 6 different from the actual distance of 7.4 Å in the X-ray crystal structure. Overall, all distances determined by the simulations are consistent with ¹9F-¹9F distances within the cluster of CF₃ groups in the crystal lattice.

DISCUSSION

While DANTE-RFDR protocols for accurate interfluorine distance measurements have been applied here for mefloquine, whose X-ray structure is known, it is important to point out that the present ¹⁹F fast MAS NMR crystallography approach is applicable for analysis of any crystalline fluorinated solid of unknown structure. By integrating the information from ¹⁹F-¹⁹F DANTE-RFDR magnetization exchange curves with ¹⁹F, ¹³C, and ¹H chemical shifts and various internuclear correlations, structures can be determined without the need for single crystal diffraction data, as has been shown for other systems⁴⁰. In our earlier study, we have demonstrated that an unbiased grid search could be successfully applied to derive distance distributions from the experimental ¹⁹F-¹⁹F DANTE-RFDR data in 2,5-difIuorobenzoic acid, and that five-spin systems are sufficient to reproduce the magnetization exchange profiles for many organic crystals with extensive coupling networks²⁴. In principle, the experiments discussed in this work can also be applied to amorphous systems, albeit with likely lower information content due to their associated structural heterogeneity and broader line widths.

Magnetic field strength is an important consideration for the ¹⁹F fast MAS NMR experiments described here, given that the ¹⁹F chemical shift range and chemical shift anisotropy (CSA) are proportional to the field strength. We see clear benefits of higher magnetic fields for fast MAS ¹⁹F spectroscopy: going from 11.7 T to 16.4 T to 19.9 T in experiments on mefloquine results in increased sensitivity and resolution. Naturally, higher fields are generally advantageous when multiple fluorine sites with similar isotropic chemical shifts are present in the system. This is certainly seen for mefloquine, investigated here, but applies also to fluorinated tryptophans,²⁵ difluorobenzoic acids²⁴ as well as HIV-1 CA protein assemblies labeled with 5F-Trp residues,²⁶ which we reported on previously. At this juncture it is important to note is that for many fluorinated organic moieties the ¹⁹F CSA lie in the 40–80 ppm range, which can be efficiently averaged out upon spinning at MAS frequencies exceeding 40 kHz, even at high magnetic fields (19.9 T).

As to the choice of MAS frequency, as observed here for mefloquine and in previous studies,^24–26 spinning frequencies of at least 40 kHz are required to obtain well-resolved spectra in the absence of decoupling, with further gains in sensitivity and resolution seen at 60 kHz MAS. As shown here, spinning frequencies of 40 kHz and higher were critical for acquiring high-resolution/high-sensitivity ¹⁹F-¹³C/¹³C-¹⁹F and ¹⁹F-¹H/¹H-¹⁹F HETCOR spectra of mefloquine. Even further gains in sensitivity and resolution are anticipated at ultrafast MAS frequencies of 111 kHz and above.

CONCLUSIONS

A fast MAS ¹⁹F NMR approach for the structural characterization of fluorine-containing natural abundance pharmaceutical compounds is presented. ¹⁹F-¹³C HETCOR and ¹⁹F-¹⁹F RFDR experiments together with DFT calculations readily permit assignments and identification of inequivalent sites in the crystal. Accurate interfluorine distances are obtained from DANTE-RFDR magnetization exchange profiles and multi-spin numerical simulations. The NMR crystallography approach presented here can be extended to pharmaceuticals of unknown structures and is broadly applicable to organic and biological molecules, including crystalline organic compounds, peptides and proteins as well as protein assemblies possessing long-range order, such as assemblies of virus proteins and amyloid fibrils.

MATERIALS AND METHODS

Chemicals

Natural abundance mefloquine hydrochloride was purchased from Acros Organics and used without further recrystallization. For MAS NMR experiments, sample amounts were as follows: 3 mg (1.3 mm rotor for measurements at 11.7 T), 3.7 mg (1.3 mm rotor for measurements at 14.1 T), 9.5 mg (1.6 mm rotor for measurements at 16.4 T), and 13.5 (1.9 mm rotor for measurements at 19.9 T).

MAS NMR spectroscopy

¹⁹F and ¹³C-detected experiments were performed on a 20.0 T narrow bore Bruker AVANCE III spectrometer outfitted with a 1.9 mm HX MAS probe. The Larmor frequencies were 850.4 MHz for ¹H, 800.1 MHz for ¹⁹F and 213.8 MHz for ¹³C. For all ¹⁹F-detected experiments, the ¹H channel was tuned to ¹⁹F. All MAS NMR spectra were acquired at a MAS frequency of 40 kHz maintained within ± 10 Hz by Bruker MAS III controller. The sample temperature was calibrated with KBr as an external temperature sensor and was maintained at 12.0±0.3 °C by a Bruker variable temperature controller. Typical 90° pulse lengths were 1.5 μs for ¹H, 1.1 is for ¹⁹F and 3.0 μs for ¹³C. ¹⁹F chemical shifts were referenced externally with respect to those of trifluoroacetic acid (100 μM solution in 25 mM sodium phosphate buffer, pH 6.5) used as an external reference (0 ppm), which relates to other commonly used reference standards as: neat trifluoroacetic acid (−2.8 ppm), trichloro-fluoro-methane (73.55 ppm), Teflon (−48.45 ppm). ¹³C chemical shifts were referenced externally to adamantane.

The ¹⁹F-¹³C cross-polarization was performed with a linear amplitude ramp of 70–100 % on ¹³C and the center of ramp was Hartmann-Hahn matched at the first or second spinning sideband; the carrier frequency on ¹³C was set to 100 ppm. For optimization of ¹⁹F-¹³C CP, ¹⁹F rf fields of 15, 25, 30, 35, 45 and 55 kHz were applied, and Hartmann-Hahn matched at 1~3 times of the spinning frequency (v_r). Zero quantum (ZQ) or double quantum (DQ) CP was matched with the ¹⁹F rf field fixed while the ¹³C rf field was systematically varied over a range of 0 kHz to 75 kHz. ¹⁹F-¹³C CPMAS spectra were acquired with 512 scans and CP contact times varied systematically from 1.0 ms to 10.0 ms; the rf fields were 15 kHz for ¹⁹F and 25 kHz (DQ-CP) or 55 kHz (ZQ-CP) for ¹³C. For 2D ¹⁹F-¹³C HETCOR experiments, the CP contact times were 1.0, 7.0 and 10.0 ms; both DQ-CP and ZQ-CP conditions were used; 38 complex points were acquired in t₂ dimension. The carrier frequencies in ¹³C were set to 100.0 ppm. In several experiments, π-pulse ¹⁹F decoupling at RF field of 208 kHz was applied during evolution in ¹³C dimension. A recycle delay of 6.0 s was used for all experiments.

For ¹H-¹³C CPMAS experiments, the ¹H-¹³C cross polarization was performed with a linear ramp; the ¹H and ¹³C RF fields were at 13 kHz and 28 kHz, respectively; the typical CP contact times were 0.5–1.4 ms. 2D ¹H-¹³C HETCOR spectra with 0.5 ms and 1.1 ms CP contact were acquired with 448 and 384 transients, respectively; 80 complex points were collected in the indirect dimension.

2D ¹⁹F-¹⁹F RFDR spectra were acquired without decoupling with RFDR mixing times of 1.6 ms, 4 ms, 8 ms, 12 ms, 20 ms and 30.4 ms. The typical length of the RFDR π pulse was 8.3 μs and a XY-16 phase cycle was used during the RFDR mixing. For each ¹⁹F-¹⁹F spectrum, the data were collected with 120 complex points in t₂ dimension using States-TPPI phase sensitive detection; 16 transients were averaged for each FID. The 1D ¹⁹F DANTE spectra were acquired with 16 scans; 22 0.1-μs DANTE pulses were applied at 8.8 ppm and 15.9 ppm for selective irradiation of 2CF₃ and 8CF₃ signals, respectively. The DANTE interpulse delay was set to 4 rotor cycles. The recycle delay was 5.0 s.

¹⁹F-detected single pulse excitation spectra were also acquired on a 11.7 T wide bore Bruker AVANCE III spectrometer outfitted with a 1.3 mm HFX MAS probe. The Larmor frequencies were 500.13 MHz for ¹H and 470.59 MHz for ¹⁹F. The MAS frequencies were 10, 40, and 60 kHz maintained within ±10 Hz by Bruker MAS III controller. Spinal-64⁴¹ (10 kHz MAS) or swept-frequency two-pulse phase modulation (SW_f-TPPM)⁴² heteronuclear decoupling sequences were applied with the ¹H RF field strengths of 90 kHz, 15 kHz, and 10 kHz for the MAS frequencies of 10 kHz, 40 kHz, and 60 kHz, respectively. ¹⁹F 90° pulse length was 2.45 μs. The recycle delay was 6.0 s.

Additional 2D ¹⁹F-¹⁹F RFDR spectra were recorded 14.1 T, on a Magnex narrow-bore magnet interfaced with a Bruker AVIII HD spectrometer, and outfitted with a 1.3 mm Bruker HCN MAS probe. The H channel was tuned to the ¹⁹F Larmor frequency of 564.35 MHz and the typical ¹⁹F 90° pulse length was 3.3 μs for. ¹⁹F-¹⁹F RFDR spectra were recorded for MAS frequencies of 40 kHz, 50 kHz and 60 kHz with RFDR mixing times of 3.2 ms and 8.0 ms, 16 transients were averaged and the recycle delay was 2.0 s. The pulse length for the DANTE selective excitation pulses was 0.1 μs. The interpulse delay was set to 2 rotor cycles. The DANTE-RFDR magnetization exchange curves were recorded with RFDR mixing times of 0.8, 1.6, 3.2, 4.0, 5.6, 8.0, 12.0, 16.0, 21.6, 26.4, and 30.4 ms; (XY8)¹₄ phase cycle⁴³ was applied during the RFDR mixing.

Supplemental ¹⁹F-¹³C CPMAS and HETCOR NMR spectra were acquired on a 16.4 T Bruker spectrometer equipped with a PhoenixNMR 1.6 mm HFX MAS probe at a MAS frequency of 40 kHz. The Larmor frequencies were 700.1 MHz for ¹H, 658.8 MHz for ¹⁹F and 176.0 MHz for ¹³C. The typical 90° pulse lengths were 2.5 μs for ¹H, 2.0 μs for ¹⁹F, and 1.97 μs for ¹³C. The ¹⁹F-¹³C CP contact time was 7.0 ms. ¹H and ¹⁹F decoupling was applied simultaneously during the t₂ evolution in the ¹³C dimension. ¹H decoupling used low-power XiX³⁵ with an rf field of 12.5 kHz. For ¹⁹F decoupling, a π-pulse with the rf field of 125 kHz was applied every rotor period. A spin echo ¹H π-pulse was applied in the center of the t₁ evolution in the ¹⁹F dimension to refocus the ¹H offset and heteronuclear coupling. The recycle delay was 3 s.

All spectra were processed in TopSpin 4.0 and analyzed with Sparky⁴⁴ and Mnova.

Numerical simulations

The DANTE-RFDR magnetization exchange curves were simulated using SIMPSON⁴⁵ (version 3.1.0). In the multi-spin simulation, the magnetization exchange was followed starting from the non-selectively irradiated spin and the evolution of the spin to which the magnetization is transferred was performed from l_2x to −l_1z. The experimental decay curves of the signals that are selectively excited by the DANTE pulse are scaled to 1 and the experimental buildup curves of the nonselective signals are scaled to 0. The simulated DANTE-RFDR exchange curves were rescaled to match the experimental intensities. The example simulation script is present in the Supporting Information.

DFT calculations

¹⁹F and ¹³C magnetic shielding tensor calculations were carried out in Gaussian 09 (Revision D.01)⁴⁶. Molecular clusters of mefloquine comprising 8 molecules were generated from the crystal structures by Pymol⁴⁷. All-atom geometry optimizations were performed using a M06 functional with the cc-pVTZ basis set and geometry-optimized models were used for magnetic shielding tensor calculations at the same level of theory. The chemical shifts were referenced by converting absolute magnetic shielding constants, σ, into absolute chemical shifts, using the relation δ_i = σ_ref – σ_i with the value of σref determined by linear regression between calculated and experimental shifts⁴⁸.

Table 3.

Sets of Interfluorine Distances for the 5-spin Simulations and in the X-ray Crystal Structure

Type of Interfluorine Distances	Simulation	X-ray
2CF3 – 8CF3	5.5 Å	5.5 Å
	8.0 Å	8.0 Å
	9.9 Å	9.9 Å
	10.5 Å	10.5 Å
2CF3 – 2CF3 (Simulation for 8CF3 resonance)	7.4 Å / 7.2 Å	7.4 Å
	10.6 Å
8CF3 – 8CF3 (Simulation for 2CF3 resonance)	7.0 Å	7.0 Å