Ultrafast ¹H MAS NMR Crystallography For Natural Abundance Pharmaceutical Compounds

Abstract

Magic angle spinning (MAS) NMR is a powerful method for the study of pharmaceutical compounds, and probes with spinning frequencies above 100 kHz enable atomic-resolution analysis of sub-micromole quantities of fully protonated solids. Here we present an ultrafast NMR crystallography approach for structural characterization of organic solids at MAS frequencies of 100-111 kHz. We assess the efficiency of ¹H-detected experiments in the solid state and demonstrate the utility of 2D and 3D homo- and heteronuclear correlation spectra for resonance assignments. These experiments are demonstrated for an amino acid, U-¹³C,¹⁵N histidine and also for the significantly larger, natural product Posaconazole, an anti-fungal compound investigated at natural abundance. Our results illustrate the power for characterizing organic molecules, enabled by exploiting the increased ¹H resolution and sensitivity at MAS frequencies above 100 kHz.

Graphical Abstract

INTRODUCTION

Solid state nuclear magnetic resonance (SSNMR) spectroscopy is ubiquitously employed for the characterization of solid-phase pharmaceuticals, both pure molecules and complex formulations^1–3. The span of SSNMR applications of pharmaceuticals varies from the identification of polymorphs and phase transitions to the assessment of formulation stability and homogeneity^4–7. One of the greatest strengths of SSNMR for the analysis of pharmaceutical products is that long-range order, such as present in crystals, is not required.

Most commonly, solid-state NMR experiments are performed with magic angle spinning (MAS), to achieve narrow lines and the accompanying increased sensitivity^8–10. At MAS frequencies of ~ 60 kHz, attainable by a number of MAS NMR probes used by SSNMR practitioners today, heteronuclear dipolar and CSA interactions are efficiently averaged out. In contrast, homonuclear ¹H-¹H dipolar couplings are not averaged at these MAS frequencies, and highly resolved ¹H-detected spectra cannot be generally acquired in fully protonated samples. This impediment can be alleviated by using extensively deuterated samples, which yield narrow ¹H lines^11–13. Unfortunately, deuteration is not a practical approach for many systems, including pharmaceutical formulations, since it involves additional synthesis.

The recent development of probes that can deliver MAS frequencies above 100 kHz has revolutionized the field of MAS NMR^14–16. At these MAS frequencies, significant gains in both resolution and sensitivity are observed because ¹H-¹H dipolar interaction can be efficiently averaged out by a combination of spinning and homonuclear decoupling using radiofrequency (rf) pulses^17–19. This regime, called “ultrafast MAS”, enabled the implementation of a broad range of ¹H-detected experiments for resonance assignments and structural characterization of fully protonated systems, including ¹H-¹H multidimensional correlation experiments²⁰. Such experiments are powerful for analyzing organic solids since ¹H chemical shifts provide valuable information on the molecular and supramolecular structure, not easily available from the chemical shifts of heteronuclei, such as ¹³C and ¹⁵N²¹. Furthermore, when combined with quantum chemical calculations of NMR parameters, they permit to extract information about supramolecular organization, such as the crystal lattice, in an approach termed “NMR crystallography”^22–25.

Here we use the NMR crystallography approach for the structural characterization of two molecules at an MAS frequency of 111 kHz. We examined the performance of several 2D and 3D experiments for crystalline U-¹³C,¹⁵N histidine. Specifically, we assessed different double quantum excitation schemes, including constant time variants²⁶, in homonuclear chemical shift correlation experiments and determined the optimum conditions for scalar coupling-based ¹³C-¹H correlation experiments. In addition, we also applied this approach to Posaconazole, a 700 Da antifungal agent²⁷. Using both examples, the optimum set of experiments for homonuclear (¹H-¹H) and heteronuclear (¹³C/¹⁵N-¹H) ¹H-detected correlation spectroscopy is discussed. In addition, we demonstrate that quantum chemical calculations of ¹H, ¹³C, and ¹⁵N chemical shifts provide invaluable information, including identification of intra- and intermolecular correlations, and permit spectral assignments.

MATERIALS AND METHODS

Samples

U-¹³C,¹⁵N histidine was purchased from Cambridge Isotope Laboratories and used without further purification. Posaconazole was kindly provided by Dr. Yongchau Su (Merck). Samples were packed into 0.7 mm rotors. Each rotor contained less than 1 mg of sample in 0.5 μL available sample volume.

MAS NMR Spectroscopy

MAS NMR experiments were collected on a 16.4 T Bruker Avance III standard bore spectrometer, outfitted with a Bruker 0.7 mm HCN MAS probe. The MAS frequency was 111 kHz, controlled to within ± 25 Hz with a Bruker MAS III controller. Heteronuclear multiple quantum coherence (HMQC) 2D correlation spectra and double quantum (DQ) 1D spectra on histidine were also acquired on a 11.7 T Bruker Avance NEO wide bore spectrometer outfitted with a Bruker 0.7 mm HCN MAS probe. Typical 90° pulse lengths were 0.7 μs (¹H), 1.5 μs (¹³C), and 2.4 μs (¹⁵N).

All experiments on U-¹³C,¹⁵N histidine used a pulse delay of 6 s and 10 kHz WALTZ16 ¹³C and ¹⁵N decoupling during ¹H acquisition. ¹H 1Ds were acquired with a Hahn echo sequence to avoid dead time and ¹H background.

To evaluate the performance of DQ excitation under ultrafast MAS, a number of conditions were tested for three different DQ excitation schemes. SPC-5₃^{28,29 1}H DQ measurements used 368 kHz rf field and optimized excitation and reconversion times of 86.49 μs and 108.11 μs, respectively. An average single value of 97 μs was chosen for the excitation and reconversion periods in the 2D experiments. R14₄⁵-symmetry based³⁰ DQ measurements require a 194 kHz B₁-field for excitation and reconversion at 111 kHz MAS, with the optimum time being 102.96 μs for both excitation and reconversion. BABA^{31 1}H DQ experiment used 357 kHz excitation pulses of 0.7 μs duration and a 0.5 μs z-filter, with 4 rotor periods for BABA DQ excitation (36.036 μs DQ evolution time) to attain optimal performance. 128 points were acquired in the t₁ dimension (11.5 ms) and for each t₁ point 16 transients were averaged.

¹H-¹H fpRFDR 2D spectra were acquired over a range of mixing times from 67.5 μs to 1 ms. 512 points were acquired in the t₁ dimension (13.8 ms); 16 transients were averaged. The ¹H-¹H RFDR DQ-SQ 3D experiment was performed with a 36 μs DQ excitation time and 125 μs RFDR mixing. 32 points were acquired in each indirect dimension (0.6 ms); 16 transients were averaged. The 2D ¹³C-¹H HETCOR spectrum was acquired with random exponentially biased (2.5 ms) non-uniform sampling (NUS) to 25%; the sampling schedule was generated in Topspin using the automatic sampling scheduling tool. For cross polarization (CP), the rf fields were 148.75 kHz and 27.75 kHz on ¹H and ¹³C, respectively and a 10% linear ramp was applied on ¹H The contact times were 1 ms and 0.8 ms for out and back transfers, respectively. 128 points were acquired in the t₁ dimension (2.3ms); 16 transients were averaged. The 2D ¹⁵N-¹H HETCOR utilized 77 kHz on ¹H with a 30% ramp, and 34 kHz on ¹⁵N for CP. Contact times were 2 ms out and 0.4 ms back. 236 points were acquired in the t₁ dimension (4.2 ms); 4 transients were averaged.

To evaluate the efficiency of ¹³C-¹H heteronuclear multiple quantum coherence (HMQC) and heteronuclear single quantum coherence (HSCQ) scalar transfers, several evolution times were tested: 1.43 ms to 4.167 ms (120 Hz to 350 Hz J coupling) for HMQC, and 0.714 ms to 1.667 ms (150 Hz to 350 Hz J coupling) for HSQC. The ¹³C-¹H HMQC 2D used 2.5 ms (200 Hz) J evolution. 128 points were acquired in the t₁ dimension (1.2 ms); 64 transients were added up for each FID. The ¹⁵N-¹H HSQC 2D also used 2.5 ms (200 Hz) J evolution. 256 points were acquired in the t₁ dimension (4.6 ms); 16 transients were averaged.

All experiments on Posaconazole used a recycle delay of 10 s (considering an average ¹H T₁ of 5s, Figure S1). For the ¹³C-¹H HETCOR experiment, the rf fields during the CP were 148.75 kHz on ¹H with a 10% ramp, and 27.75 kHz on ¹³C. The contact times were 1 ms and 0.8 ms for out and back CP transfers, respectively. 256 points were acquired in the t₁ dimension (2.3 ms); 64 transients were accumulated, for a total experiment time of 45 hours. The ¹³C-¹H HMQC 2D used 2 ms (250 Hz) J evolution. 128 points were acquired in the t₁ dimension (1.2 ms); 128 transients were averaged, for a total experiment time of 45 hours.

¹H-¹H fpRFDR 2D spectra were acquired at three mixing times: 90 μs, 200μs, and 1 ms. 512 points were acquired in the t₁ dimension (13.8 ms); 16 transients were averaged, for a total experiment time of 23 hours per spectrum. The ¹H-¹H RFDR DQ-SQ 3D experiments were performed using a 27 μs DQ excitation time and 150 μs RFDR mixing, with {XY4}₄¹ phase cycling. 48 points were acquired in each indirect dimension (0.8 ms); 16 scans were averaged, for a total experiment time of 102 hours.

All spectra were processed in Topspin without apodization for 1Ds and 30° or 60° sine squared or Lorentzian-to-Gaussian apodization for 2Ds. The processed data were analyzed in Sparky³² or CCPNMR^33,34.

Density Functional Theory (DFT) Calculations

All DFT calculations were performed using the Gaussian09³⁵ programs. All calculations were carried out for a four molecule cluster of Posaconazole in the crystal (CCDC database identifier YIMVUO³⁶). The all-atom geometry optimization was carried out using Gaussian09, at the B3LYP/6-31G level. Magnetic shielding tensor calculations were performed with the gauge-including atomic orbitals (GIAO) formalism^37–40, as implemented in Gaussian09. The NMR chemical shift ab-initio calculation was carried out at the B3LYP/cc-pVTZ level. Experimental ¹³C and ¹H isotropic chemical shifts were referenced to TMS and ¹⁵N isotropic chemical shifts were referenced by converting isotropic magnetic shieldings σ_iso into chemical shifts using the relation δ_iso = σ_ref − σ_iso, with the value of σ_ref determined by linear regression between calculated and experimental shifts.

RESULTS AND DISCUSSION

¹H-¹H 2D Correlation Experiments

To test the performance of various ¹H-based experiments, a series of 2D and 3D through-space and through-bond correlation spectra were acquired on U-¹³C,¹⁵N histidine. We first evaluated the efficiency of ¹H-¹H homonuclear double quantum (DQ) experiments. Multiple quantum correlations can provide valuable information through enhanced resolution or spectral editing⁴¹. The SPC-5₃, R14₄⁵, and BABA DQ excitation schemes tested are schematically illustrated in Figure 1A. The required ¹H field strengths for these measurements range from 194 kHz to 390 kHz at the MAS frequency of 111 kHz, which can be readily achieved with a 0.7 mm probe.

Figure 1.

(A) General pulse sequence for double quantum (DQ) correlation experiments: DQ excitation, followed by evolution in the indirect dimension, then reconversion. The delay t_d is an optional dephasing delay for the removal of any residual transverse magnetization. Elements for different 3 DQ recoupling sequences are also shown. (B) Comparison of resonance intensities for the various ¹H homonuclear double quantum recoupling schemes for U-¹³C,¹⁵N histidine, relative to a Hahn echo spectrum (black): SPC-5₃ (purple) 368 kHz recoupling, R14₄⁵ (blue) 194 kHz recoupling, BABA (pink) 357kHz recoupling. All spectra were acquired as 1-dimensional experiments at a MAS frequency of 111 kHz.

For the SPC-5₃ ¹H DQ experiment^28,29, the optimized DQ excitation and reconversion times were 86.49 μs and 108.11 μs, with an average and maximum DQ yield of 38% and 43%, respectively (Figure 1B, magenta). For the triple-quantum (TQ) SPC-5₃ experiment, the average TQ yield was 11%, using a single optimized excitation and reconversion time of 184 μs (Figure S2). The single optimized DQ excitation and reconversion time for the R14₄⁵ DQ experiment³⁰ was 102.96 μs, with an average and maximum DQ yield of 38% and 41%, respectively (Figure 1B, blue). Back-to-back (BABA) homonuclear DQ recoupling³¹ (Figure 1B, pink) resulted in average and maximum DQ yields of 28%, and 32.5%, respectively and had poor performance for sites with short T₂s despite a short, 36.04us recoupling duration (Hahn echo T₂ decays shown in Figure S3).

While there is little overall difference in the performance of SPC-5₃ vs R14₄⁵, R14₄⁵ has lower rf field requirements, while the SPC-5₃ sequence appears to have a slightly better excitation bandwidth, as suggested by the slightly higher intensity of the downfield ¹H resonances. Therefore, we used the SPC-5₃ sequence in multidimensional experiments. While it is generally anticipated that a constant time (CT) variation during the DQ evolution period²⁶ improves ¹H resolution, CT requires a long and fairly uniform transverse DQ spin relaxation, a condition hard to meet with only few exceptions. In the case of histidine, the short T₂ relaxation times (Figure S3) caused loss of several DQ-SQ correlations (Figure S4). Therefore, the CT version of the experiment was not used for our studies of Posaconazole.

To establish through-space ¹H-¹H connectivities, a 2D DQ-SQ ¹H-¹H spectrum was acquired with SPC-5₃ excitation using 97 μs excitation and reconversion. As shown in Figure 2A (pink spectrum), many well-resolved resonances are obtained and can be readily assigned.

Figure 2.

(A) Through-space and through-bond ¹H-detected correlation experiments for U-¹³C,¹⁵N histidine at 111 kHz MAS. Full resonance assignments could be completed with this set of ¹³C-¹H, ¹⁵N-¹H, and ¹H-¹H spectra. The RFDR mixing time was 500 μs and contact times were 1 ms out, 0.8 ms back for ¹H-¹³C CP, and 2 ms out, 0.4 ms back for ¹H-¹⁵N CP. Both ¹³C-¹H HMQC and ¹⁵N-¹H HSQC used 200 Hz J evolution. Other experimental parameters are provided in the Materials and Methods section. (B) Efficiency of ¹³C-¹H HSQC and HMQC correlations at selected evolution times. (C) 3D structure of histidine. Carbon atoms are colored gray, nitrogen atoms blue, oxygen atoms red, and hydrogen atoms white.

We also acquired a series of ZQ ¹H-¹H correlation spectra with fpRFDR mixing⁴², using the {XY4}₄¹ phase cycling scheme⁴³ for detection of through-space ¹H-¹H connectivities. For the small molecule histidine, outstanding resolution and efficient ¹H-¹H mixing are obtained at a mixing time of 500 μs, as shown in Figure 2A (grey spectrum). For the larger molecule Posaconazole, we found that it is helpful for assignments to acquire spectra at several mixing times, as discussed below.

¹H-Detected 2D Heteronuclear Correlation Experiments

¹⁵N-¹H and ¹³C-¹H 2D correlation spectra for U-¹³C,¹⁵N histidine employ scalar-based and dipolar-based transfers in HMQC/HSQC and HETCOR experiments, respectively. As illustrated in Figure 2A, these experiments yield complementary information: the cross peaks in ¹⁵N-¹H HSQC and ¹³C-¹H HMQC spectra arise from single-bond correlations, while those in the HETCOR spectra are also associated with long-range through-space correlations. Therefore, a full set of experiments is required for resonance assignments and extracting distance information. Indeed, we were able to assign all resonances using a combination of 2D ¹³C-¹H and ¹⁵N-¹H HMQC/HSCQ spectra, and long-range correlations are extracted from the HETCOR spectra. It is worth noting that for histidine and generally small molecules, long-range correlations are present with using CP contact times for out/back transfers of 1/0.8 ms (¹H-¹³C) and 2/0.4 ms (¹H-¹⁵N). Remarkably, the dipolar-based ¹H-¹H, ¹³C-¹H, and ¹⁵N-¹H spectra of histidine also contain cross-peaks corresponding to correlations between histidine signals and those of a crystallographic water molecule. Observation of these cross peaks was possible due to the rigid nature of this water molecule.

Peak intensities for the Cε1-Hε1 cross peak, extracted from the 1D traces through the direct dimensions of the HMQC and HSQC spectra for different evolution times are shown in Figure 2B. As can be appreciated, the efficiency of magnetization transfers is substantially higher at longer J-evolution time in the HMQC compared to the HSQC transfer. Furthermore, the ¹H linewidths in the HMQC experiment are independent of the scalar evolution time: the linewidths are 279 and 276 Hz at 1.43 ms (350 Hz) and 3.33 ms (150 Hz), respectively. In contrast, in the HSQC experiment, broader lines are observed at longer scalar evolution times: 243 and 269 Hz at 0.714 ms (350 Hz) and 1.67 ms (150 Hz), respectively. Thus, the HMQC sequence is superior for small organic solids and pharmaceutical compounds in powder formulations.

¹H-Detected 3D Homonuclear Correlation Experiments

While for small molecules, such as histidine, a combination of 2D heteronuclear correlation experiments is sufficient for complete resonance assignments, this is not the case for larger systems with potential resonance overlap. For more complex molecules, additional frequency dimensions are required. We therefore examined the performance of a 3D ¹H-¹H-¹H correlation experiment, where the RFDR recoupling is combined with SPC5₃ double-quantum-to-single-quantum transfer (DQSQ)^14,20,28,29. The pulse sequence for a RFDR-DQSQ correlation spectrum and representative 2D planes extracted from the 3D spectrum recorded for U-¹³C,¹⁵N-histidine are displayed in Figure 3A and 3B, respectively. Numerous well-resolved cross peaks are observed in the spectrum and were readily assigned.

Figure 3.

(A) Pulse sequence of and (B) 2D planes extracted from the 3D ¹H-¹H-¹H RFDR-DQSQ (SPC5₃) experiment of U-¹³C,¹⁵N-histidine. The chemical shift in the indirect ¹H DQ dimension is indicated in the lower right corner of each panel.

Applications to Natural Abundance Pharmaceuticals: Posaconazole

We also applied ¹H-detected 2D and 3D homo- and heteronuclear experiments to the anti-fungal pharmaceutical compound Posaconazole. This natural product serves as an example for a larger, complex pharmaceutical. For resonance assignments of Posaconazole a series of 2D and 3D ¹H-¹H and ¹³C-¹H correlation spectra were acquired (Figure 4). Since the sample is not isotopically labeled, inherently low sensitivity is observed in the ¹³C-¹H HMQC and HETCOR spectra, and longer range ¹³C-¹H correlations are not as readily observed. Therefore, fewer connectivities are present compared to the spectrum of U-¹³C,¹⁵N histidine acquired with the same CP contact time.

Figure 4.

(A) Blue (top): through-bond (HMQC) and through-space (HETCOR) ¹³C-¹H correlation spectra of Posaconazole acquired at 111 kHz MAS. The contact times were 1 ms out and 0.8 ms back for ¹H-¹³C CP. 250 Hz J evolution was used for the ¹³C-¹H HMQC. Gray (bottom) ¹H-¹H 2D fpRFDR correlation spectra for three different mixing times: 90 μs (black), 200 μs (medium gray), and 1 ms (light gray) and slices from a 3D RFDR-DQSQ (SPC-5₃) correlation experiment. The chemical shift of the DQ dimension is indicated in the lower left of each panel. The apparent lower resolution of the HMQC relative to the HETCOR is due to the lower sensitivity of the HMQC spectrum, requiring both an increase in the number of scans (128 vs. 64) and shorter sampling time in the indirect dimension (1.2 vs. 2.3 ms) for the same total experiment time. (B) Molecular structure of Posaconazole.

To assist with the resonance assignments of Posaconazole, NMR parameters were calculated using DFT for a four-molecule cluster extracted from a recent X-ray crystal structure of Posaconazole³⁶ (Figure 5A). By combining the DFT calculations with our experimental data, most resonances were readily assigned. The calculated ¹³C chemical shifts are in excellent agreement with the solution NMR values reported previously⁴⁴ (Figure S5). The correlation between calculated and experimental ¹³C chemical shifts is shown in Figure 5B. While overall agreement between experimental and DFT-calculated ¹H chemical shifts is good, several averaged DFT ¹H shifts exhibited deviations of up to several ppm from the experimentally determined values (Figure 5C), although the calculated ¹H chemical shifts agree well with another recent DFT study of Posaconazole⁴⁵. ¹H shifts that exhibit significant differences between averaged DFT ¹H shifts and experimental ¹H shifts include the triazolone group (H44). Inspection of the 4-molecule cluster used for DFT and resulting chemical shifts (Table S2) indicate good agreement with experimental ¹H shifts when an intermolecular hydrogen bond is formed with a neighboring molecule of Posaconazole (Figure 5D). In addition, notable shift differences between the experimental solution ¹H chemical shifts⁴⁴ and the solid state isotropic shifts determined here are also present (Figure 5C; bottom).

Figure 5.

(A) DFT optimized cluster of four molecules of crystalline Posaconazole (CCDC database identifier YIMVUO). (B) (top) Correlation between DFT-calculated and experimental solid state ¹³C chemical shifts; (bottom) correlation between solution and experimental solid state ¹³C isotropic chemical shifts for Posaconazole. (C) (top) Correlation between DFT-calculated and experimental solid state ¹H chemical shifts; (bottom) correlation between experimental solution and solid-state NMR ¹H isotropic chemical shifts of Posaconazole. Best fits are shown as dotted lines and y = x as a solid black line. All calculations were performed with the Gaussian suite of programs. (D) Illustration of the intermolecular hydrogen bond between H44 of molecule 4 and O41 of molecule 3 in the crystal. The experimental and calculated ¹H chemical shifts for H44 are listed at the bottom. Due to the small size of the DFT cluster, only molecule 4 in the cluster exhibits this hydrogen bond, resulting in better agreement with solid state NMR shift than the average over the 4 molecules.

Since ¹H chemical shifts are very sensitive to intermolecular interactions⁴⁶, they are significantly influenced by hydrogen bonding in the crystal. Indeed, in recent room-temperature MAS NMR⁴⁷ and dynamic nuclear polarization (DNP) studies⁴⁸ of amorphous solid dispersions (ASD) of Posaconazole, different ¹H shifts from those observed here for the pure molecule are seen, underscoring the high sensitivity of ¹H chemical shifts to critical structural characteristics of molecular state and/or formulation. All calculated and solid state NMR chemical shifts are listed in Tables S1 (¹³C, ¹⁵N) and S2 (¹H), along with the previously reported solution NMR chemical shifts⁴⁴. In total, from the spectra presented above we assigned 27 of 28 ¹H and 27 out of 37 ¹³C resonances. Unassigned carbon resonances are those without attached hydrogens.

Additional structural information can be extracted from a series of ¹H-¹H 2D fpRFDR spectra at different mixing times. At short mixing times, only spatially close atoms will yield cross peaks, while at longer mixing times longer range correlations are seen. At the shortest ¹H-¹H mixing time of 90 μs, only correlations for distances of ~ 2-4 Å are observed, while at the longest mixing time of 1 ms correlations corresponding to up to distances of ~ 4-6 Å are present. Several intra- and intermolecular correlations and the corresponding distances in the crystal structure for H50 are illustrated in Figure 6. These correlations are in excellent agreement with distances in the recent crystal structure of Posaconazole³⁶. As an aside, we note that, the 2D RFDR spectra of Posaconazole exhibit limited resolution and complete assignment, solely based on these spectra, was not possible. Therefore, the 3D ¹H-¹H-¹H DQ-SQ-RFDR experiment and DFT NMR calculation were particularly helpful for reliable ¹H chemical shift assignment. A major advantage of the ¹H-¹H-¹H 3D experiment over spectra that contain a ¹³C dimension (e.g. ¹³C-¹H-¹H) is the significant time saving for natural abundance materials. The 2D ¹³C-¹H HETCOR experiment of Posaconazole required 45 hours of signal averaging (Figure 4), and the 3D ¹H-¹H-¹H spectrum was recorded in 102 hours. A 3D experiment with a ¹³C dimension (e.g. ¹³C-¹H-¹H) could not be acquired in a reasonable amount of time without implementing time-saving methods such as non-uniform sampling. Taken together, our results also underscore the benefits of combining several 2D and 3D ¹H-detected experiments for in-depth structural characterization of pharmaceutical compounds at natural abundance. Double-quantum based spectra provide additional information complementary to the single-quantum data sets. While the optimal set of measurements is determined by the system of interest, the general approach presented here can be adapted to a broad range of pharmaceutical solids.

Figure 6.

A) Selected region of the 2D ¹H-¹H RFDR correlation spectra for two different mixing times. Several intra- and intermolecular correlations for H50 labeled (only a single ¹H resonance is observed for the 3 protons of this methyl group). Correlations at 90 μs (dark gray) correspond to distances of ~ 2-4 Å. At 1 ms (light gray), additional correlations, corresponding to distances of up to ~ 4-6 Å are observed. B) Intermolecular distances involving H50 in the crystal structure of Posaconazole that yielded correlations in the 2D ¹H-¹H RFDR correlation spectra (CCDC database identifier YIMVUO, ³⁶). Intermolecular distances are shown by pink dotted lines, while intramolecular distances are shown by black dotted lines.

CONCLUSIONS

Here we reported on applications of ultrafast ¹H MAS NMR to small molecules and in particular to the anti-fungal drug Posaconazole. Our work demonstrates that ultrafast MAS NMR spectroscopy combined with DFT calculations in ‘NMR crystallography’ is a powerful approach for structural characterization of pharmaceutical compounds at natural abundance with micromole quantities. We envision that applications of ultrafast MAS methods will extend to a wide range of pharmaceutical formulations, for probing atomic-level details of structure, polymorphism, and dynamics.