David A.
Hirsh
a,
Aaron J.
Rossini
*bc,
Lyndon
Emsley
d and
Robert W.
Schurko
*a
aDepartment of Chemistry and Biochemistry, University of Windsor, Windsor, ON N9B 3P4, Canada. E-mail: rschurko@uwindsor.ca; Tel: +1-519-253-3000 ext. 3548
bDepartment of Chemistry, Iowa State University, Ames, IA 50011, USA. E-mail: arossini@iastate.edu; Tel: +1-515-294-8952
cUS DOE Ames Laboratory, Ames, Iowa 50011, USA
dInstitut des Sciences et Ingénierie Chimiques, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, CH-1015, Switzerland
First published on 24th August 2016
In this work, we show how to obtain efficient dynamic nuclear polarization (DNP) enhanced 35Cl solid-state NMR (SSNMR) spectra at 9.4 T and demonstrate how they can be used to characterize the molecular-level structure of hydrochloride salts of active pharmaceutical ingredients (APIs) in both bulk and low wt% API dosage forms. 35Cl SSNMR central-transition powder patterns of chloride ions are typically tens to hundreds of kHz in breadth, and most cannot be excited uniformly with high-power rectangular pulses or acquired under conditions of magic-angle spinning (MAS). Herein, we demonstrate the combination of DNP and 1H–35Cl broadband adiabatic inversion cross polarization (BRAIN-CP) experiments for the acquisition of high quality wideline spectra of APIs under static sample conditions, and obtain signals up to 50 times greater than in spectra acquired without the use of DNP at 100 K. We report a new protocol, called spinning-on spinning-off (SOSO) acquisition, where MAS is applied during part of the polarization delay to increase the DNP enhancements and then the MAS rotation is stopped so that a wideline 35Cl NMR powder pattern free from the effects of spinning sidebands can be acquired under static conditions. This method provides an additional two-fold signal enhancement compared to DNP-enhanced SSNMR spectra acquired under purely static conditions. DNP-enhanced 35Cl experiments are used to characterize APIs in bulk and dosage forms with Cl contents as low as 0.45 wt%. These results are compared to DNP-enhanced 1H–13C CP/MAS spectra of APIs in dosage forms, which are often hindered by interfering signals arising from the binders, fillers and other excipient materials.
Solid APIs are commonly characterized using X-ray diffraction (powder or single-crystal), 1H and 13C solid-state NMR (SSNMR), thermogravimetric methods, and other spectroscopic techniques.12–18 In many cases, these techniques provide adequate characterization of the bulk forms of APIs; however, they are often of limited use for dosage forms (especially those with low weight percentages, wt%, of the API). In particular, both pXRD patterns and 13C SSNMR spectra of dosage forms often display interfering signals from the excipient (e.g., binding ingredients and fillers), which obscure signals arising from the API.
Prior studies by our group19–21 and others22,23 have demonstrated that 35Cl SSNMR is a valuable tool for characterizing the bulk and dosage forms of APIs that have been synthesized as HCl salts (excipients do not contain chloride ions, so there are no interfering signals in 35Cl SSNMR spectra of dosage forms). Since 35Cl is a quadrupolar nucleus (I = 3/2), its spectra are influenced by a combination of anisotropic chemical shift and quadrupolar interactions. The latter are particularly sensitive to small structural differences in the local Cl− anion environments arising from variations in local hydrogen bonding.19,20,24–27 As a result, each solid phase of an API produces a distinct 35Cl SSNMR spectral fingerprint. Given the importance of identifying low concentrations of API phases within dosage forms (including impurities), it is crucial to improve the lower detection limit (LDL) of 35Cl SSNMR experiments.
Our research group has developed pulse sequences that enable the rapid acquisition of broad 35Cl SSNMR patterns (hundreds of kHz or more) with high S/N, even at moderate field strengths (e.g., 9.4 T). Unlike older methods,28–30 these pulse sequences rely on phase-modulated frequency-swept WURST (wideband uniform-rate smooth truncation) pulses31,32 for broadband excitation and polarization transfer. The WURST-CPMG (WCPMG)33,34 and broadband adiabatic inversion cross-polarization (BRAIN-CP)35 pulse sequences are used for direct (35Cl) and indirect (1H–35Cl) broadband excitation of 35Cl SSNMR spectra, respectively. The BRAIN-CP-WCPMG sequence (BCP for short) uses BRAIN-CP to transfer spin polarization from abundant nuclides (e.g., 1H) to dilute nuclides (e.g., 35Cl), and a subsequent WURST-CPMG pulse and windowed acquisition train for further signal enhancement (Fig. S1, ESI†).
Over the past few years, high-field dynamic nuclear polarization (DNP) has become a prominent method for achieving high gains in S/N for SSNMR experiments.36–39 Recent developments in DNP NMR instrumentation (e.g., high-frequency gyrotrons,40,41 low-temperature MAS probes42,43), optimized radical polarizing agents,44–48 and the availability of commercial DNP NMR spectrometers49,50 have enabled signal increases in excess of 300, representing potential time savings by a factor of 90000. DNP has enabled SSNMR experiments that were previously considered challenging/impossible, allowing the detailed the study of materials that were previously inaccessible to SSNMR.51–60 Most materials are prepared for DNP experiments by using a simple incipient wetness impregnation procedure to coat the surface of the particles, or fill the porous volume with a radical polarizing agent solution.47,61 Saturation of the EPR transitions of the radicals with microwaves results in enhanced polarization of the nuclei (most often protons) that are in close proximity to the polarizing agents.46 In the case of micro-particulate organic solids (e.g., APIs), DNP-enhanced polarization can be relayed from the surface of the particles into the interiors of the solids by 1H–1H spin diffusion without perturbing their macroscopic structure.62,63 With this technique, DNP-enhanced solid-state NMR can be applied to organic solids,63–66 pure APIs,63,66–69 and low API wt% dosage forms.69 These developments have enabled natural isotopic abundance 13C–13C, 1H–15N and 13C–15N correlation SSNMR experiments, which would be challenging or impossible without DNP.
To date, most DNP SSNMR studies have been limited to the characterization of nuclei with fairly narrow powder patterns (i.e., breadths on the order of tens of kHz or less). These reports include the spectral acquisition of a variety of spin-1/2 nuclei as well as quadrupolar nuclei in highly symmetric environments (e.g., 2H, 14N overtone, 17O, 27Al, and 51V).55,56,64,66,70–75 However, the enhancements afforded by DNP also make it an attractive technique for the acquisition of NMR spectra of broader patterns (e.g., 35Cl SSNMR) due to their inherently low S/N that largely results from the distribution of NMR signal across a wide frequency range. During the final preparation of the manuscript describing our work, Kobayashi et al. published one such example, in a report on DNP-enhanced ultra-wideline 195Pt SSNMR.76 One factor that hinders the acquisition of DNP-enhanced wideline spectra from stationary samples is that moderate- and fast-magic angle spinning (MAS) results in substantially higher DNP and sensitivity enhancements.47,50,77–79 MAS rates of ca. 6 to 40 kHz provide DNP enhancements (ε) that are ca. 3 to 5 times higher than those obtained from static (i.e., stationary sample) experiments.47,50,77,78 Unfortunately, MAS experiments are not generally useful for patterns with breadths on the order of 100 kHz or more, especially for quadrupolar nuclides. First, MAS only results in partial averaging of the effects of the second-order quadrupolar interaction, and second, even very fast MAS may not result in the separation of the spinning sidebands from the isotropic centerband.80 Matters are further complicated by the effects of first-order quadrupolar interactions and/or large chemical shift anisotropies. In many cases, MAS spectra of broad patterns are distorted and have low S/N, which prevents the accurate determination of quadrupolar and chemical shift tensor parameters.
Herein, we show that DNP can be used to enhance S/N in static wideline 35Cl SSNMR patterns of APIs acquired with BCP methods. We detail a new protocol, called spinning-on spinning-off (SOSO) acquisition, for enhancing the DNP polarization under MAS and subsequently acquiring a wideline 35Cl pattern under static conditions. These techniques result in DNP enhancements in 35Cl SSNMR spectra of up to a factor of 110 at 100 K and B0 = 9.4 T. We demonstrate the application of 35Cl DNP SSNMR for the characterization of APIs in their bulk forms, as well as in dosage forms with Cl contents of as low as 0.45 wt%. The application of these techniques for polymorph differentiation, impurity identification, and the discovery of new solid phases are also demonstrated.
The full experimental parameters used for the 13C and 35Cl experiments are given in the ESI† (Tables S3–S6). NMR-13C: all 1H–13C CP/MAS experiments used a CP contact time of 2 ms, and a constant 13C spin lock rf field of ca. 74 kHz. An MAS frequency of 8 kHz (or 9 kHz for isox) was used. The 1H spin lock amplitude was linearly ramped86–88 from ca. 75 kHz to 83 kHz. Proton decoupling was applied for each acquisition using an rf field of 100 kHz and the SPINAL-64 pulse sequence.89 NMR-35Cl: 1H–35Cl CP-CPMG and CP-echo spectra of hist were acquired using conventional rectangular pulses with ca. 70 kHz rf on both channels. The 1H–35Cl BRAIN-CP-WURST-CPMG (BCP) pulse sequence35 was used to acquire the spectra of all of the other samples. The sweep direction of the WURST pulse applied during the BCP contact period can result in lopsided powder patterns; to minimize these effects on the 35Cl powder patterns of isox and diph, two sub-spectra were acquired with opposite sweep directions for the BCP contact pulses. These sub-spectra were then co-added together to form the final spectrum (Fig. S11, ESI†). With the exception of the spectra of hist, all of the 35Cl spectra were acquired using WURST-CPMG (WCPMG) refocusing pulses.33,34 To process these spectra, the echoes in each FID were co-added to form a single echo, which was then Fourier transformed and magnitude processed.
Because of the breadth of the 35Cl powder pattern of ceti (vide infra), its 1H–35Cl BCP spectrum was acquired using frequency-stepped acquisition.28,29,90 Four pieces were acquired with the transmitter frequencies separated by 50 kHz increments. These pieces were then co-added together to produce the final spectrum.
The dipolar hetero-nuclear multiple-quantum correlation rotatory-resonance recoupling (D-HMQC-R3) pulse sequence91,92 was used to obtain a 2D dipolar 13C–35Cl correlation spectrum in the hist sample. See Table S4 for the experimental parameters (ESI†).
The 1H–35Cl CP-CPMG NMR spectra of hist acquired with and without DNP are shown in Fig. 2b. Even under completely static sample conditions, a large DNP enhancement is observed when the microwaves are turned on (εCP(35Cl) = 50) (Fig. 2b). However, when the sample is rotated during part of the 30 s recycle/polarization delay, an additional 2.2-fold signal gain over the static DNP experiment is observed (Fig. 2b). We call this technique spinning-on spinning-off (SOSO). To acquire the spectrum with SOSO, we manually controlled the sample spinning at ca. 200 to 2000 Hz during the recycle delay and stopped the sample spinning several seconds before collecting a spectrum under static conditions (see Fig. S1 for a schematic diagram of the pulse sequence timings, ESI†). This procedure was repeated for each of the 4 to 8 scans in the experiment. SOSO allows for a larger 1H polarization build-up due to improved DNP while the sample is spinning and acquisition of a distortion-free wideline spectrum under static conditions. The improved DNP enhancement with spinning is consistent with previous results that show increased DNP enhancements in 1H–13C CP/MAS NMR spectra with increasing sample spinning speeds up to ca. 10–15 kHz.47,50,77,78
In order to examine the effects of slow MAS on DNP efficiency (without the application of SOSO), a 35Cl SSNMR spectrum was acquired with uninterrupted slow MAS at 250 Hz and the CP-echo pulse sequence (Fig. 3a). Slow MAS yields a central transition powder pattern that is very similar in appearance to that obtained from a corresponding experiment on a static sample (Fig. 3a and Fig. S2, ESI†). The signal in the spectrum acquired with continuous slow MAS is ca. 3.1 times greater than that acquired under static conditions (Fig. 3a). Thus, the DNP enhancement under continuous MAS is slightly greater than that obtained with SOSO (2.2 times, vide supra). The 35Cl DNP enhancements in these spectra are lower than the enhancements seen in the 13C NMR spectra, due in part to the slower MAS frequencies used in both the SOSO and slow MAS 35Cl SSNMR experiments.
Since hist has a small quadrupolar coupling constant, a high-quality DNP enhanced CP-echo spectrum was also acquired at a faster MAS frequency of 8 kHz (Fig. S3, ESI†), resulting in a powder pattern free of overlapping spinning sidebands. The DNP enhancement of this MAS spectrum (εCP(35Cl) = 230) is comparable to the enhancements seen in the 13C NMR spectra (εCP(13C) = 260, vide supra). While slow or moderate MAS experiments may be useful for acquiring spectra of narrow central transition patterns that can be partially averaged (i.e., static pattern breadths <50 kHz, Fig. S3, ESI†), overlap of the MAS powder patterns with spinning sidebands is problematic for spectra with broader central transition patterns. Such overlap yields spectra without clearly defined discontinuities, which are difficult to analyze (Fig. S4, ESI† and vide infra). Given that the 35Cl SSNMR spectra of anionic chlorides in hydrochloride salts of APIs typically have powder patterns with breadths spanning 100–300 kHz at moderate field strengths,19–21 MAS 35Cl experiments with spinning rates between 5 and 15 kHz are not suitable for the characterization of most APIs. Furthermore, the polarization transfer from 1H to quadrupolar nuclei is often challenging to optimize and inefficient under MAS conditions.95
All of the 35Cl SSNMR spectra of hist can be simulated with the same quadrupolar tensor parameters: CQ = 1.8(1) MHz, ηQ = 0.72(2), δiso = 16(5) ppm (spectral simulations of these and all other 35Cl powder patterns in this work can be found in the ESI:† Fig. S5–S9 and Table S7). While these parameters agree with those reported in a recent study by Pandey et al. (CQ = 1.8 MHz and ηQ = 0.66),22 they do not match those reported by Bryce et al. for the room temperature spectrum of hist (CQ = 4.59 MHz, ηQ = 0.46, δiso = 93 ppm).24 It is possible that the study by Bryce et al. involved a different polymorph of hist.
DNP enhancement also provides access to two-dimensional experiments that would otherwise be challenging or impossible. One such example is a 2D heteronuclear dipolar correlation spectrum of proximate 13C and 35Cl nuclei. As a proof of concept, we obtained a 13C–35Cl dipolar heteronuclear multiple-quantum correlation spectrum (Fig. 3b) with rotatory-resonance recoupling (D-HMQC-R3).91,92 This 2D spectrum shows correlations between the 13C and 35Cl nuclei that are close to each other. Such results may provide valuable distance constraints on the structure of the molecule that may be useful for NMR crystallography.96 With DNP at higher magnetic fields and/or with faster sample spinning rates, it should be possible to acquire 13C–35Cl correlation spectra for APIs with larger values of CQ. 2D 13C–35Cl correlation NMR spectra could enable overlapping 35Cl powder patterns in APIs with multiple 35Cl sites to be resolved by correlation to high resolution 13C resonances.
As with hist, DNP provides a considerable signal enhancement (εCP(35Cl) = 15) in the 1H–35Cl BCP NMR spectra acquired under static conditions (Fig. 4b). If the SOSO procedure is used (i.e., where the sample is spun slowly during most of the recycle delay and then is stationary during the pulse and acquisition periods), the DNP enhancement is further increased by a factor 2 and εCP(35Cl) = 30 is observed (Fig. 4b). The combination of DNP and BCP produces a high S/N spectrum, spanning roughly 200 kHz, (ca. 10 times broader than that of hist).
Given the breadth of the 35Cl powder pattern of ambr at 9.4 T, it is not possible to use conventional MAS (i.e., constant spinning throughout the experiment). At the spinning speeds typically used for DNP-enhanced 13C NMR experiments (i.e., between 8 and 15 kHz), the presence of spinning sidebands distorts the 35Cl pattern and makes analysis of the powder pattern challenging (Fig. S4, ESI†). These issues result because (i) slower MAS rates do not average the effects of the second-order quadrupolar interaction, leading to patterns with many overlapping sidebands that are difficult to simulate, and (ii) spinning speeds exceeding 40 to 50 kHz are necessary to separate the spinning sidebands from the isotropic centerband for typical values of 35Cl CQ at 9.4 T (or hundreds of kHz for other quadrupolar nuclides).80 However, DNP MAS probes with faster spinning rates have recently become available,79 and may enable acquisition of spectra exhibiting undistorted isotropic centerbands for 35Cl sites with larger values of CQ.
There is a slight distortion in the low-frequency shoulder (at ca. −70 kHz) of the pattern acquired with the SOSO method. This distortion may arise from an improperly refocused CPMG echo train, which results from the sample not coming to a complete stop after spinning during the recycle delay – even very slow spinning can be disastrous for these experiments (Fig. S10, ESI†). The starting and stopping of the sample spinning was performed manually for this preliminary set of experiments. In the future, these issues could be addressed with the addition of specialized hardware to precisely control the spinning rate and stop/start timings of the sample. Alternatively, longer recycle delays could be employed, at the expense of a slight reduction in sensitivity.
Fig. 5 SSNMR spectra of the bulk and dosage forms of isox impregnated with a 15 mM TEKPol/TCE solution acquired at 100 K and B0 = 9.4 T. The left column has 1H–13C CP/MAS spectra of (a) the bulk API and (b) the dosage samples acquired with microwaves on and off. The bottom inset shows the full spectra of the dosage form without vertical clipping. The right column has 1H–35Cl BCP spectra of (c) the bulk API with and without microwaves and (d) the bulk API and dosage samples with DNP enhancement. The lineshapes in (c) are lopsided to the high frequency side because only one sweep direction of the BCP contact pulse was used (see Fig. S10 for more details, ESI†). |
Another complication is that the DNP enhancements of signals arising from the API, solvent, and excipient molecules are not the same, as was observed in a prior study of several cetirizine dosage forms using DNP-enhanced 13C SSNMR.69 In the case of isox (Fig. 5b, inset), the strongest signal is observed for a feature at 74.8(2) ppm, which corresponds to the solvent, TCE (εCP(13C) = 104). This intense feature dominates the spectrum of the dosage form and obscures several peaks from the API. The DNP enhancement of the unobscured API signal (e.g., peaks at ca. 110–135 ppm), εCP(13C) = 32, is less than that of the solvent. Finally, there are features that correspond to various types of excipient molecules, including polysaccharides (20–40 ppm), synthetic polymers (60–75 ppm) and stearates (100–110 ppm),69,97,98 which have enhancements ranging from 12 to 50. Overlapping signals from the API and excipient make it challenging to determine the phase of an API in 13C NMR spectra even without the use of DNP;21 however, the differences in the DNP-enhancement such as those observed in the spectra of isox can further complicate the analysis. While the signal from the solvent can be decreased by adding a spin-echo to the pulse sequence99 doing so only marginally improves the resolution of the features from isox (Fig. S14, ESI†).
35Cl SSNMR can selectively probe the API in the dosage form without interfering signals from the excipients, since chloride ions are only found in the API and the 35Cl signal from covalently-bound chlorines would be extremely broad and of too low intensity to be detected.100–102 A comparison of the DNP-enhanced 1H–35Cl BCP NMR spectra of bulk and dosage forms of isox is shown in Fig. 5d. The DNP enhancement observed for isox under static conditions (Fig. 5c, εCP(35Cl) = 12) is comparable to that of ambr (cf.Fig. 4b). Both the 35Cl SSNMR spectra of the bulk and dosage forms of isox are consistent with spectra acquired without DNP19 (see Fig. S7 for the spectral simulation and associated quadrupolar parameters, ESI†). The fact that the powder pattern of the tablet matches that of the bulk compound confirms that both samples contain the same polymorph of isox. There are additional features in the centers of both patterns (with centers of gravity at ca. 0 ppm). While the origin of this feature is still under study, it may result from Cl− anions coordinated to H2O (e.g., as a result of disproportionation of the HCl salt), or some other chemical or physical alteration of the sample (see VT-pXRD patterns in Fig. S12, ESI†).
A primary advantage of DNP experiments is that dosage forms of APIs can be studied even if the wt% of the API is very low, as is the case for isox (4.95 wt% API, 0.52 wt% Cl). Here, the combined acquisition time of the 35Cl spectra of the pure and dosage forms of isox with DNP was just over 10 hours, roughly half the experiment time necessary to acquire a comparable set of spectra at room temperature.21 Such time savings are critical in the high-throughput screening of dosage forms with low wt% APIs. Given the long acquisition times required to obtain sufficient 35Cl signal without microwaves, we have not attempted to acquire these spectra, and cannot report an enhancement in the spectrum of dosage isox (or the other dosage forms, vide infra).
SOSO experiments were not conducted on isox because of its short T1(1H) of 15 s. The short T1(1H) limits the amount of time available for 1H polarization build up before the polarization is lost to longitudinal relaxation. Roughly 25 s were required to start the sample spinning and completely stop rotation, which was not fast enough to fit within the optimal polarization time (20 s) for the experiments with isox. Of course, a longer polarization time could be applied to perform the SOSO method; however, the gains from increased DNP enhancement would be partially offset from reduced sensitivity arising from use of a recycle delay longer than 1.3 × T1. As with ambr (vide supra), the 35Cl powder pattern is too broad for conventional MAS to be used.
Fig. 6 13C and 35Cl SSNMR spectra of the bulk and dosage forms of diph impregnated with a 15 mM TEKPol/1,3-dibromobutane solution acquired at 100 K and B0 = 9.4 T. The left column has 1H–13C CP/MAS spectra of (a) the bulk API and (b) the dosage samples acquired with microwaves on and off. Asterisks denote spinning sidebands. The right column has 1H–35Cl BCP spectra of (c) the bulk API with and without microwaves and (d) the bulk API and dosage samples with DNP enhancement. The lineshape in (c) is lopsided to the high frequency side because only one sweep direction of the BCP contact pulse was used (see Fig. S10 for more details, ESI†). |
The DNP enhancement observed in these spectra is not as high as those observed in the spectra of the other samples (Fig. 2, 4, and 5). One contributing factor is that diph, like isox, has a relatively short T1(1H) (ca. 18 s), which limits the build-up of enhanced polarization in the microcrystalline solid. The choice of solvent for the radical also plays an important role, as prior studies have reported decreased enhancements when using DBB rather than TCE, however, DBB was required for solubility reasons.47,48,103 Another disadvantage of the use of DBB is that it produces a broad solvent peak in the 13C SSNMR spectra, which can obscure features from the API and excipient and make confirmation of the phase and purity even more difficult. Clearly, the identification of other compatible solvents is an important future step for the further optimization of DNP experiments on APIs.
Unlike the 13C spectra, the 1H–35Cl BCP NMR spectra of diph (Fig. 6d) are free from signal interference from the excipient and solvent. The breadths of the powder patterns and locations of the discontinuities are identical, which confirm that the same phase of diph is present in both the bulk and dosage forms. These features are also consistent with previous work done at room temperature.21 As with the spectra of isox, there is an additional low-intensity feature at ca. −4(2) ppm that may result from disproportionation of the API. It is difficult to measure the DNP enhancement of these spectra due to the low S/N in the spectrum acquired without microwaves (Fig. 6c). The estimated minimum value of εCP(35Cl) = 7 was obtained by comparing the spectra after applying a Fourier transform directly to the CPMG echo train (see Fig. S15, ESI†).
Prior 35Cl SSNMR studies of this compound have relied on direct excitation experiments (e.g., WCPMG)21 because attempts to use CP at room temperature were unsuccessful due to poor CP efficiency. While CP experiments are still a challenge at low temperature, they are possible with the use of DNP. Unfortunately, the variation of CP efficiency with temperature when using the BCP pulse sequence is not well understood. DNP experiments, such as those reported here, could provide opportunities for further understanding the CP dynamics, and lead to improvements to the BCP sequence and related experiments under DNP conditions.
The 1H–13C CP/MAS NMR spectra of the bulk and dosage samples of ceti are shown in Fig. 7a and b, respectively, and are consistent with previous studies of this compound using DNP-enhanced 13C SSNMR.69 Several distinct features from the API are apparent in the spectrum of bulk ceti (e.g., peaks at 60–70 ppm and 130–200 ppm). However, only the strongest signal from the API (at ca. 130 ppm) can be distinguished in the spectrum of the dosage form due to interference from the excipients. In both spectra of ceti, there is less interference from the solvent than was observed in the spectra of diph (cf.Fig. 6a and b), because the DNP enhancement of the solvent peak is not as large (ε(DBB, ceti) = 24, ε(DBB, diph) = 50). The enhancement of the solvent varies from sample to sample, which may depend on the types and concentrations of different excipients, the quality of the glass formation when the radical-containing solution freezes, the amount of dissolved O2(g), the dielectric properties of the sample that affect the microwave field,106 or the other factors discussed above.
The enhancement observed in the spectrum of bulk ceti (εCP(13C) = 20) is close to what was previously reported for this compound (εCP(13C) = 31).69 However, for ceti in the dosage form we measured εCP(13C) = 8, which is lower than the 55-fold enhancement previously reported. We attribute these decreased enhancements to the use of DBB as the radical solvent. In the previous DNP SSNMR study, ceti was found to be sparingly soluble in TCE.69 We chose to use DBB (in which ceti is insoluble) to maximize the amount of solid sample in the rotor and to better maintain the structure of the API in the dosage form. Given that the discrepancy in enhancements is particularly apparent in the spectra of the dosage form, it is possible that DBB does not penetrate the excipients as well as TCE. Further optimization of the sample preparation could likely yield improved enhancements.
Ceti is a dihydrochloride, and its 35Cl SSNMR spectra should show two distinct 35Cl patterns corresponding to structurally unique Cl− anion sites. With the use of DNP, it is possible to identify two overlapping powder patterns in the 35Cl spectrum of the bulk form in just 5 minutes (εCP(35Cl) = 5.8, Fig. 7c). Due to hardware limitations, it is challenging to uniformly excite the entire breadth of the two 35Cl patterns of ceti, even with BCP. As such, we used frequency-stepped acquisition28,29 and combined 4 sub-spectra at evenly spaced transmitter frequencies to obtain the full pattern (which is ca. 250 kHz broad, Fig. S9, ESI†). The two overlapping 35Cl powder patterns with distinct quadrupolar parameters can be readily distinguished using analytical simulations. Nonetheless, the most important discontinuities in the two patterns can be observed in the central sub-spectrum (as in Fig. 7c and d).
Acquiring a 35Cl SSNMR spectrum of the dosage form with similar signal-to-noise ratio takes more than 11 hours, due to the extremely low wt% Cl in this sample (5.78 wt% API, 0.45 wt% Cl), which is the lowest discussed herein. Comparison of the 35Cl spectra of the bulk and dosage forms (Fig. 7d) shows that the dosage form likely contains two 35Cl environments that are similar to those of the bulk form. However, the poorer resolution of the discontinuities in the 35Cl spectrum of the dosage form (most evident at the center of the pattern) is consistent with the lower crystallinity of the API within the formulation. This API was previously confirmed to exist as an amorphous form in all formulations using DNP-enhanced 13C and 15N SSNMR in the previous study.69
The techniques we have reported in this study will help expand the use of DNP to the study of other wideline and ultra-wideline (breadths > 250 kHz) powder patterns. While the spectra reported herein are dominated by the second-order quadrupolar interaction, the techniques described should would work equally well for patterns that are broadened by the first-order quadrupolar interaction, chemical shift anisotropy, or combinations thereof. These developments make DNP useful for the study of a wide range of materials whose NMR spectra suffer from inherently low S/N largely due to the wide breadth of the signal.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6cp04353d |
This journal is © the Owner Societies 2016 |