Pharmaceutical Analysis by NMR Can Accommodate Strict Impurity Thresholds: The Case of Choline

Abstract

The ICH guidelines recommend reporting thresholds for regular impurities in drug substances at the level of 0.05% or 0.03% (w/w) depending on the maximum daily intake. Therefore, any instrumental method of analysis applicable to the impurity analysis should be able to detect and quantify the analytes at those levels. This investigation was designed to verify the suitability of ¹H NMR spectroscopy for the detection of impurities, as a first step in the process before attempting quantification. In order to minimize demand on equipment, this study employed a 400 MHz instrument for structural confirmation and signal assignments of choline (1) and O-(2-hydroxyethyl)choline (2), a known impurity. The limit of detection (LOD) of 2 in 10 mg of 1 was established as 0.01% on a 400 MHz instrument and 2% on a 60 MHz (benchtop) NMR spectrometer, respectively. Thus, impurities for which quantification is required are readily detected at 400 MHz or above. These results are in contrast to the widespread belief that ¹H NMR sensitivity is insufficient for pharmaceutical impurity analysis. Further, our experiments revealed that as low as. The choice of solvent was recognized as a critical parameter for ¹H NMR LOD analysis. Furthermore, publicly available NMR raw data (HMDB) proved to be valuable for unveiling the otherwise cryptic information hidden in complex signal patterns via ¹H NMR iterative Full Spin Analysis. Finally, the study uncovered the less noticed, yet characteristic, ¹⁴N-¹H coupling in the -N⁺(CH₃)₃ groups, adding further strong arguments for the Raw NMR Data Initiative. Collectively, the data prove that the analytical capabilities of high-field NMR easily fulfill the ICH requirements for detection of impurity analysis. This study explores the potential of ¹H NMR spectroscopy to detect an impurity in the presence of an actual substance of interest which makes it a step closer to achieving regulatory standards.

1. Introduction

1.1. Choline Health Products

Choline is an essential nutrient for humans [1]. Consequently, choline in the form of various salts such as chloride, bitartrate, and citrate are widely used as food additives and health supplements [2]. On the other hand, derivatives such as succinylcholine, methacholine, succinylmonocholine, acetylcholine are employed as pharmacological agents (Chapter 1 in [3]. Reflecting their vast significance for human health, interest in novel methodologies for the pharmaceutical analysis of choline is growing, and just recently a quantification method by ¹⁴N NMR was introduced [4]. Choline used in commercial supplements is typically produced synthetically by reaction of trimethyl amine with ethylene oxide. O-(2-hydroxyethoxyethyl)trimethylammonium (2), is a by-product impurity occurring in choline obtained through this well-established synthetic process and may be formed by nucleophilic substitution of ethylene oxide to choline. Structurally, 2 is the O-2-hydroxyethyl ether of choline (Fig. 1) and represents Residual Complexity (go.uic.edu/residualcomplexity) of the choline synthetic process. The analytical determination of 2 as impurity of choline derivatives has wide application in quality control of foods, dietary supplements, and pharmaceuticals.

Fig. 1.

Structures of choline chloride (1), impurity (2) iodide, and choline bitartrate (3) along with their chemical formulas and molecular weights.

1.2. Analytical Detection and the Role of NMR Spectroscopy

Liquid chromatography (LC) with UV/Vis or mass spectrometric (MS) detection is the most widely used analytical procedure to detect and quantify impurities in drug substances. NMR spectroscopy, on the other hand, offers additional information to accurately identify and quantify organic compounds as it gives the comprehensive profile of all molecules (based on the nuclei-specific analysis) present along with the compound of interest (e.g., an Active Pharmaceutical Ingredient [API]).

The Limit of Detection (LOD) is the lowest amount of any substance that can be reliably observed by a given analytical method; it is one of the crucial parameters established in method validation. The LOD of an analyte depends on multiple factors that include, among others, its chemical nature, analytical detection methodology and the sample matrix. Numerically, the LOD can be determined for the relevant signal as the intensity for which the signal to noise (S/N) equals 3 [5]. The LOD may also be calculated based on the standard deviation of the response and the slope of a calibration curve. In quantitative ¹H NMR (qHNMR) spectroscopy, calibration curves are constructed based on the integration of specific resonance signals.

The Signal to Noise (S/N) in NMR spectroscopy depends on a number of factors (see overview in page S4): From a practical perspective, the SNR for an FT ¹HNMR spectrum is proportional to the square root of the number of scans (transients) during acquisition. The SNR of a particular signal in the spectrum is equal to 2.5 times the height of the signal divided by the baseline noise (where a signal is not present). The concentration at which the signal of the analyte being measured is observable may be considered the LOD in NMR. For example, the LOD of ephedrine-HCl was determined to be 0.025 mg on a 300-MHz spectrometer [6]. Cholesterol has been detected and quantified by NMR in HSV529, a vaccine candidate. The LOD was determined as 0.1 μg/mL on a 700-MHz spectrometer [7].

1.3. Established Quality Control Guidelines vs. NMR Approaches

For drugs, the reporting, identification, and qualification thresholds of impurities depend on dosage. As per International Council for Harmonisation (ICH) guidelines for quantification of impurities in new drug substances (Q3A(R2)) [8], the reporting threshold is 0.03 % (w/w) if the maximum daily dose of a particular drug is greater than 2 g. The reporting threshold, which is the amount beyond which an impurity has to be reported albeit without further characterization, is 0.05% when the maximum daily dose of the drug is less than 2 g. The identification thresholds for new drug substances, being the amount beyond which the impurity also has to be identified/characterized, are roughly two times the reporting thresholds.

Applying the ICH guidance to choline, the recommended adequate intake for adult (19 years and older) men and women is 550 mg and 425 mg, respectively. The maximum daily allowance of choline for adults is 3500 mg per day [1]. Therefore, an analytical procedure must be able to quantitate an impurity in choline materials at the level of 0.03%. Before quantitation could be performed, it is a prerequisite that the instrument is able to detect the impurity. While it is generally known that NMR spectrometers, esp. high field, can produce analyzable signals for low concentrations, as per our literature survey, systematic study of the limits of detection (LOD) of an impurity in the presence of a drug substance (or in this case a dietary ingredient) are sparse: while the use of ¹³C satellites for LOD validation has been recognized for long [9,10], early NMR-based impurity profiling method were typically limited to levels >1–3%, one notable exception being the qNMR method described for 0.1% dihydrolovastatin in lovastatin [11]. Therefore, this inquiry also sought to provide support for the ability of NMR spectroscopy, despite frequently being termed a relatively insensitive technique, to detect impurities at levels recommended by ICH. Reflecting economic factors and considering recent trends in benchtop NMR instrumentation, the study compared the suitability of both high- and low-field ¹H NMR for the detection and identification of 2 in the presence of 1. Furthermore, in order to demonstrate the feasibility of employing low-field benchtop NMR instruments for choline analysis, the study also performed quantum mechanics-based ¹H NMR full spin analysis (HiFSA) [12]. This not only contributed to unequivocal identification, but also made the presented data universal by allowing for seamless adjustments to instruments with various magnetic field strengths. Further, the calculated spectrum could be used to subtract the main component from the experimental spectrum to expose impurities.

2. Materials and Methods

2.1. Instrumentation

High-field NMR spectra were acquired on a JEOL 400 YH ECZ400 NMR spectrometer (Tokyo, Japan) at 25.5°C with a triple resonance broadband cryoprobe. Hydrogen resonated at 399.78 MHz. Low-field ¹H NMR spectra were acquired on a Magritek Spinsolve 60 Carbon Ultra (Malvern, PA, USA) spectrometer at 25°C. Hydrogen resonated at 60.81 MHz. Norell (Morganton, NC, USA) sample tubes (XR-55–7) with 5 mm o.d. were used. Spectra processing was performed with MestreNova v 14.1.2 (Mestrelab Research S.L., Santiago de Compostela, Spain). The spectra and raw data are available as Supplementary Materials.

2.2. Chemicals

Choline chloride (1), O-(2-hydroxyethoxyethyl)trimethylammonium iodide (choline RCA, 2) and choline bitartrate were received from the US Pharmacopeial Convention (Rockville, MD, USA). NMR solvents and chemicals: D₂O (99.9%) was purchased from Sigma Aldrich, (St. Louis, MO, USA) and DMSO-d₆ (99.9%) from Cambridge Isotope Laboratories Inc, (Tewksbury, MA, USA); 3-(trimethylsilyl)-1-propanesulfonic acid-d₆ (DSS-d₆, 92.3%) was a gift from FUJIFILM Wako Pure Chemical Corporation (Osaka, Japan).

2.3. NMR sample preparation

Samples were weighed to five decimal places on XS-105 balance (Mettler Toledo, OH, USA) and the solutions were pipetted using calibrated micropipettes (Drummond Scientific Co., PA, USA). Chemical shift referencing was done with the residual solvent signal DMSO-d₅ (2.5000 ppm) when samples were measured in DMSO-d₆. Because the residual HOD signal is broad and changes its chemical shift values depending on the water (moisture) content in the sample, an internal reference standard, DSS-d₆, was added to D₂O to calibrate the ppm scale. D₂O was spiked with 1.5 % (w/v) of DSS-d₆ and the trimethylsilyl signal of DSS-d₆ was referenced to 0.0000 ppm when samples were dissolved in D₂O.

2.4. Structure confirmation

Impurity 2, ~6 mg (equivalent to ~3.23 mg of (2)) and ~4 mg (equivalent to ~2.15 mg of (2)) dissolved in 0.5 mL of DMSO-d₆ and D₂O respectively, were subjected to 1D (¹H, ¹³C), and 2D (COSY, HSQC and HMBC) NMR spectral analyses to assign the ¹H and ¹³C NMR signals.

2.5. ¹H NMR acquisition conditions

¹H NMR (¹³C decoupled using the GARP composite decoupling scheme; see refs. [13,14] and references therein) spectra were acquired under two different conditions: 1. For mixtures with 10 mg of choline chloride and 5.4 – 0.027% of 2: (a) At 400 MHz: 8, 128, and 256 scans (~15 min, 1.25 h, 2.5 h respectively of total experimental time); 90° pulse angle, 30 s relaxation delay; 128 k data points, equivalent to an acquisition time of: 8.74 s; 20 ppm spectral width. (b) At 60 MHz: 64 and 1024 scans; 90° pulse angle, AQ+d1: 30 s; 7.6 k data points; 80 ppm spectral width. 2. For mixtures with 100 mg of choline chloride and 0.012 – 0.05% of 2: at 400 MHz: 1024 scans (~ 20 min total experimental time); 1 s relaxation delay; 128 k data points; 20 ppm spectral width. No apodization was used to process the acquired spectra.

2.6. Preparation of impurity (2) solutions for the LOD experiments

Aliquots of 500 μL of each of the following dilutions were transferred to NMR tubes. Impurity 2 (2.09 mg) was dissolved in 10.00 mL of D₂O to create a 200 μg/mL stock solution equivalent to 108 μg/mL of 2. The stock solution (250 μL) was diluted to 500 μL with D₂O to create a 27 μg sample. The stock solution (100 μL) was diluted to 500 μL with D₂O to create a 11 μg sample. The stock solution (100 μL) was diluted to 1.00 mL with D₂O to create a 5.4 μg sample. The stock solution (25 μL) was diluted to 500 μL with D₂O to create a 2.7 μg sample. The stock solution (25 μL) was diluted to 1.00 mL with D₂O to create a 1.3 μg sample. The stock solution (25 μL) was diluted to 2.00 mL with D₂O to create a 0.67 μg sample.

2.7. Stock solutions

Two samples of choline chloride (49.72 and 50.12 mg) were dissolved in 2.00 mL of DMSO-d₆ and D₂O respectively to create ~ 25 mg/mL solutions. Samples of 2 (9.92 and 9.98 mg, equivalent to 5.34 and 5.38 mg) were dissolved in 0.5 mL of DMSO-d₆ and D₂O respectively to create ~ 20 mg/mL solutions.

2.8. Stock working solutions of impurity (2) for spiking (A-D)

The impurity 2 stock solution was diluted accordingly as follows: For 2 mg, 100 μL of the stock solution of 2 was used. For 0.4 mg, 100 μL of stock solution was diluted to 500 μL with deuterated solvent (working solution A). For 0.1 mg, 100 μL of working solution A was diluted to 400 μL (working solution B). For 0.02 mg, 100 μL of working solution B was diluted to 500 μL (working solution C). For 0.01 mg, 100 μL of working solution C was diluted to 200 μL (working solution D).

2.9. Mixtures of choline (1) and impurity (2)

Stock solution of choline chloride (400 μL) and stock and working solutions for spiking (A-D) of 2 (100 μL; after diluting accordingly as mentioned above to obtain target concentrations) were added into each NMR tube in order to maintain 500 μL of the 1 and 2 mixture in all five samples.

2.10. Choline derivatives

Each choline derivative (4–7 mg, exact amounts given in Table S4) were dissolved separately in 500 μL of DMSO-d₆, and D₂O.

2.11. Analysis of impurity (2) in 100 mg choline (1)

Impurity 2 stock solution was prepared by dissolving 2.55 mg in 1.00 mL of D₂O. The stock solution (100 μL) was diluted to 500 μL with D₂O to create a 50 μg/100 μL solution A. Solution A (100 μL) was diluted to 200 μL with D₂O to create a 25 μg/100 μL solution B. Solution B (100 μL) was diluted to 200 μL with D₂O to create a 12.5 μg/100 μL solution C. 100 mg samples of Choline chloride were dissolved in 400 μL of D₂O. Aliquots of 100 μL of Solutions A, B and C were added to choline chloride solutions to obtain 500 μL of solutions containing 100 mg choline chloride spiked with 100, 25 and 12.5 μg of impurity 2 respectively.

2.12. HiFSA

HiFSA of choline was performed on the ¹H NMR spectrum available in the raw data section of HMDB (HMDB0000097_62–49-7P.fid) using the pre-release version of ‘Cosmic Truth’ (CT; ct.nmrsolutions.io), a web-based, client/server software for automatic and semi-automatic spectrum analysis from NMR Solutions Ltd. (Kuopio, Finland), as described earlier [12]. The original FID from HMDB was subjected to a strong resolution enhancement to reveal the signal splittings from the ²J and ³J couplings between N⁺ and the neighboring hydrogens.

3. Results

3.1. Comparison of impurity (2) with choline derivatives

Impurity 2 could be the byproduct during the synthesis of choline chloride/bitartrate. Therefore, the ¹H NMR spectrum of 2 was compared with those of choline chloride and choline bitartrate, by stacking, acquired in DMSO-d₆ at 400 MHz (Figs. 2A and S1). This would reveal a possible overlap of signals thus helping in determining the suitability of solvent and method. While the N,N,N-trimethyl (NTM) signals of choline and 2 appear in the narrow window of 3.19 – 3.21 ppm, the triplet at 4.658 ppm of 2 hydroxy does not coincide with any other signals. However, when the counter-ion for choline is an organic acid (e.g., bitartrate, citrate etc), there is a possibility of hydrogen (of hydroxy) exchange with the carboxyl hydrogens in the acid. In which case, the hydroxy triplet of choline would be missing. Some signals in choline chloride and choline bitartrate (and likely other organic acid counter-ions) overlap with methylene hydrogens (H-2) of 2 signal at 3.836 ppm.

Fig. 2.

Sections of ¹H NMR spectra of various cholines: A: DMSO-d₆, B: D₂O (+DSS-d₆). A, B: From top to bottom: 1 in 60 MHz (violet), Choline bitartrate in 60 MHz (dark blue), 2 in 60 MHz (blue-green), 1 in 400 MHz (green), Choline bitartrate in 400 MHz (olive green), 2 in 400 MHz (maroon). For full spectra, refer to SI: Figs. S1–S4. For sample weights, see Table S4.

Likewise, the 400 MHz ¹H NMR spectrum of 2 was compared with those of choline chloride, and choline bitartrate acquired in D₂O (Figs. 2B and S2). In this solvent, the signals in 2 do not show any interferences with those of the choline salts. Hence, D₂O was identified as the most suitable solvent for the detection of 2 in the tested choline salts.

All samples were also analyzed at 60 MHz. In DMSO-d₆, the most distinctive signal for 2 was that of the hydroxy hydrogen at 4.658 ppm (Figs. 2A and S3): it was diagnostic for the presence of 2 in samples of choline chloride and choline bitartrate, as shown by stacking of the spectra. Generally, the methylene resonances in 2 show partial overlap with the resonance of the methylene groups of the same two samples.

The H-2 methylene signal of 2 spectrum acquired in D₂O at 60 MHz appears at 3.979 ppm. However, it overlaps considerably with the signal at 4.055 ppm in both choline chloride and choline bitartrate (and likely other organic acid counter-ions) (Figs. 2B and S4). In order to identify 2 in cholines (chloride and bitartrate), its signal at 3.65 ppm was rendered to be most useful.

Summarizing these initial studies, the stacked ¹H NMR spectra (Fig. 2) show that 2 may be detected and identified in the presence of choline and/or any of the other selected choline derivatives at both 400 and 60 MHz with either DMSO-d₆ or D₂O as a solvent. However, because choline has two signals that are close to 2 signals, artificial mixtures of 1 and 2 were analyzed in this investigation. This was done to support the hypothesis that, if 2 can be detected in the presence of choline, as shown in the stacked ¹H NMR spectra (Fig. 2), it can be detected in the presence of other choline salts and derivatives as well.

3.2. LOD of impurity (2) iodide in D₂O at 400 MHz

To evaluate the threshold at which sensitivity limits the capability of NMR to recognize an impurity, and before performing experiments in mixtures, a series of experiments was undertaken to determine the LOD of neat 2 in D₂O at 400 MHz. A series of 6 tubes were created with quantities of 2 ranging from 50 μg to 1.25 μg. The spectra were acquired using a set of conditions described as (2) in section 2.5. As can be observed in Fig. 3, all the signals in the NMR spectra were visualized from 50 μg down to 2.5 μg. The SNR for H-1’ (3.662 ppm) for the 50 μg to 2.5 μg samples was in the range of 74 to 5.8 (Table S6). In the 1.25 μg sample, there is one visible signal, of the NTM group, at 3.195 ppm (which would not be useful because it was found to overlap with NTM of choline salts) along with two spurious signals, likely from the sample preparation and/or solvent, a triplet at 3.599 ppm (overlapping with the methylene H-1) and a singlet at 3.340 ppm. The lowest limit at which 2 (2.5 μg) was detected is equivalent to 1.35 μg of 2 cation. Thus, the 400 MHz instrument equipped with cryoprobe was clearly capable of detecting and identifying amounts of 2 consistent with ICH requirements. However, to verify if the lower limit of detection remains the same in the presence of actual substance, several artificial mixtures of 1 and 2, including the concentrations at which ICH requires impurity to be quantitated, were prepared and analyzed.

Fig. 3.

¹H NMR spectra (stacked, 400 MHz, D₂O) of 2 at various concentrations to determine the limit of detection. *At 1.25 μg, signals at 3.46 and 3.35 ppm, belonging to unknown component(s) of the sample of impurity 2, or likely related to solvent or sample preparation were observed.

3.3. ¹H NMR assignments of impurity (2)

This study also achieved the full ¹H NMR assignments of 2 in both DMSO-d₆ and D₂O, generated by the analysis of its 400 MHz 1D and 2D NMR spectra (Tables S1 and S2; Figs. S12–S21). Interestingly, the choice of solvent affects not only the chemical shifts, but also the order of the methylene signals of 2. In DMSO-d₆, the order of the methylene signals is H₂-1’, H₂-2’, H₂-1, and H₂-2 from low to high frequency, with H₂-2’ and H₂-1 being highly overlapped (Fig. S22). In D₂O, the order of methylene signals becomes H₂-1, H₂-1’, H₂-2’, and H₂-2, showing well-resolved signals (Fig. S23). The subsequent ¹H NMR iterative Full Spin Analysis (HiFSA) of 1 and 2 revealed the complete spin-spin couplings of these two molecules (Tables S12–S18).

3.4. Comparison of choline (1) and impurity (2) ¹H NMR spectra

Comparative stacking of the ¹H NMR spectra of 1 and 2 acquired on a 400 MHz NMR spectrometer in DMSO-d₆ (shown in Fig. S22), reveals that three of the four methylene signals are distinguishable from the two choline methylene signals, with the H-2 methylene signals overlapping. The stacked 400 MHz ¹H NMR spectra of 1 and 2 in D₂O (Fig. S23) reveal that all four methylene signals of 2 are distinguishable from those of the methylene signals of 1, with the methylene signals of 2 resonating between those of the two methylene signals of 1. In D₂O, the hydroxy signals are invisible due to rapid H-exchange with the solvent. In both DMSO-d₆ and D₂O, the NTM “singlets” are prominent but slightly overlapped. For this series of experiments, the stacked spectra of 1 and 2 were acquired separately at concentrations of ~5 mg each (see Materials and Methods). However, the overlap of signals likely increases with the difference in concentration between the two compounds. When 2 occurs as impurity of 1, only signals H₂-1’ and H₂-2’ of 2 are suitably separated from choline signals.

3.5. LOD of impurity (2) in the presence of 10 mg choline (1)

For the next series of experiments, five artificial mixtures with concentrations of 2, from 2 mg (20%) to 10 μg (0.1%) in 10 mg of choline, were prepared in DMSO-d₆ and D₂O. The ¹H NMR spectra were acquired at both 400 and 60 MHz. The ¹H NMR signals in the spectra of the mixtures were identified based on the assignments shown in Tables S1 and S2.

The artificial mixtures of 1 and 2 in DMSO-d₆ were acquired at 400 MHz (Figs. 4A and S5): The NTM and H-2 methylene signals of 2 and choline were overlapped. However, the H₂-1, H₂-1’, and H₂-2’ methylene signals of 2 were readily distinguished from those of the methylene signals of 1. The triplet corresponding to the hydroxy group appeared at ~ 5.62 ppm for choline, whereas it was observed at ~ 4.85 ppm for 2. However, as discussed earlier, in cases of choline bitartrate or citrate, the hydroxy signal may not appear due to exchange and so is not a dependable signal to distinguish analytes. The signals of 2 were clearly observed in samples with concentrations of 2 mg (20%) down to 100 μg (1%) as evidenced by the SNRs of the OH signal from 521 to 3.86 for the signal at 4.856 ppm (Table 1, S7). The signals of 2 in the 20 μg (0.2%) and 10 μg (0.1%) samples could be observed when examining the baseline while increasing the y-scale. The comparison of determining an LOD in a pure 2 sample vs. 2 in the presence of a greater amount of 1 demonstrates that the LOD also depends on the relative amounts of (other) analytes in the sample. The 400 MHz instrument with cryoprobe could detect and identify as little as 10 μg (0.1%) of 2 in the presence of 10 mg of 1. However, it was not enough to meet the ICH requirement of a quantitation limit, 0.03% (30 ug) or less. To reach those levels, choline sample was increased to 100 mg, adding 12.5 ug of 2 to determine if 0.01% of 2 could be detected.

Fig. 4.

Sections of ¹H NMR spectra of 1–2 mixtures in different ratios in A: DMSO-d₆, 400 MHz. B: D₂O (+DSS-d₆), 400 MHz. C: DMSO-d₆, 60 MHz. D: D₂O (+DSS-d₆), 60 MHz. The pale-orange colored regions are signals of 2 and rest are contributed by 1. For full spectra, refer to SI: Figs. S5–S8. From top to bottom: Molar ratios of 1 to 2: 90:10 (violet), 98:2 (blue), 99.5:0.5 (green), 99.9:0.1 (olive green), 99.95:0.05 (maroon). For exact amounts in the NMR samples, see Table S5. As the relatively dense stacked plots result from the 200:1 dynamic range and need for uniform vertical scaling, the raw NMR data is made available for individual viewing and detailed expansions.

Table 1.

Consolidated table of the signal-to-noise ratios (SNRs) of the various ¹H NMR resonance patterns of 2 at different concentrations in 10.0 mg of choline chloride (NS 256 for 400 MHz, NS1024 for 60 MHz^*).

				SNR
Solvent	Instrument (MHz)	Hydrogen	ppm	2 mg (1.08 mg)^	400 μg (216 μg)^	100 μg (54 μg)^	20 μg (10.8 μg)^	10 μg (5.4 μg)^
DMSO-d5	400	OH	4.856	520.76	131.85	31.29	10.91	3.86
	60	OH	4.832	81.74	21.53	-	-	-
D2O	400	H-2’	3.743	361.17	89.95	24.54	21.31	7.27
		H-1’	3.650	387.66	93.41	26.67	25.71	10.18
	60	H-2’	3.706	90.61	55.28	-	-	-
		H-1’	3.688	100.79	6.50	-	-	-

^*The Signal to Noise ratios were calculated by using the automated script “SNR calculation” in MNova. S/N values could be obtained for the signals only which are recognized by the algorithm.

^{^}equivalent weights of 2 (to 2 iodide). Complete data is given in Tables S7–S10.

The same set of artificial mixtures of 2 with 10 mg choline chloride in DMSO-d₆ were analyzed on 60 MHz (Figs. 4C, S7 and S10) to examine the sensitivity of the benchtop NMR spectrometer. Except for the hydroxy signal, all the signals of 2 significantly overlapped with those of choline in the 60 MHz spectra. The triplet signal corresponding to the hydroxy at 4.832 pm of 2 was observed in mixtures containing 2 mg and 400 μg (4%) of 2. In the mixture containing 100 μg of 2, a slight bump was observed, but its triplet shape was not evident. Increasing the number of scans did not show any improvement in visualizing the signal in the mixtures containing lower amounts of 2 (20 μg and 10 μg), as can be seen from the comparison in Figs. S8 and S11. The SNR for a 2 mg sample with 64 scans was 25.7, compared to 81.7 with 1024 scans (Table 1, S8). In these measurements, the sensitivity of the 60 MHz instrument was four times less than the 400 MHz instrument with cryoprobe.

A similar set of artificial mixtures of 1 and 2 in D₂O were acquired at 400 MHz (Figs. 4B and S7). The four methylene signals of 2 were clearly distinguished even in the presence of 10 mg of choline when measured in D₂O. All the methylene signals of 2 in the prepared concentrations (2 mg to 10 μg) were observed with SNR in the range of 361 to 7.27 for the H-2’ resonance at 3.743 ppm (Table 1, S9). Similar to the series measured in DMSO-d₆, it was straightforward to locate the signals down to 100 μg (1%) of 2. However, for the 20 μg (0.02%) and 10 μg (0.01%) mixtures, the methylene signals were more prominent in D₂O than DMSO-d₆.

The set of artificial mixtures of 2 with 10 mg of 1 in D₂O were also analyzed at 60 MHz (Figs. 4D, S8 and S9). All the signals of 2 overlapped with those of choline in the 60 MHz spectra, except for a certain part of the resonances of the three methylene groups around at 3.69 ppm. These resonances were clearly observed in mixtures containing 2 mg and 400 μg of 2. In the mixture containing 100 μg of 2, the resonances were still evident, albeit with poor shape with less well-defined line shape. While assignment to one or two specific nuclei might not be achievable at 60 MHz, it might be feasible to develop an empirical integral ratio based assay for the threshold determination of the impurity 2 in the future. In the same manner as was observed with DMSO-d₆, increasing the number of scans further did not improve the visibility of the combined H-2’, H-1 and H-1’ methylene signals in the lower 2 concentrations (20 and 10 μg added), as can be seen by comparing (Figs. S8 and S9). This instrumental limitation was also reflected in the achievable SNRs of the 3.706 signal: being 71.3 for 64 transients, it only increased by ~1.3 fold when acquiring with 1024 scans, which theoretically should have led to a 2-fold increase (Table 1, S10). Therefore, the LOD of 2 in 1 in D₂O was similar to that of DMSO-d₆, only differing in the reference signal.

3.6. LOD of impurity (2) in the presence of 100 mg choline (1)

In another set of samples, choline chloride was increased 10 fold to 100 mg, thereby increasing the 1 to 2 ratio to more closely reflect samples containing minor amounts of impurity. Samples containing 100 (0.1%), 25 (0.02%) and 12.5 μg (0.01%) of 2 along with 100 mg of 1 in D₂O were analyzed at 400 MHz (Figs. 5 and S6). These experiments could only be performed in D₂O due to solubility limitations. The results show that two of the four methylene signals of 2 are visible on the tails of the very intense choline methylene signals, with SNRs from 42.5 to 21.6 for the H-2’ signals at 3.748 ppm (Table S11). Ultimately, the 400 MHz NMR spectrometer equipped with a N₂-cryoprobe was able to detect the impurity in the presence of the actual compound levels, at which ICH requires the impurity to be quantitated.

Fig. 5.

Selected region ¹H NMR spectra in D₂O (+DSS-d₆), 400 MHz of 100 mg Choline (purple); 100 mg Choline + 100 μg of 2 (blue); 100 mg Choline + 25 μg of 2 (green); 100 mg Choline + 12.5 μg of 2 (maroon). For full spectra refer to SI: Fig. S11.

4. Discussion

4.1. Comparing 400 and 60 MHz instruments

NMR measurements at both 400 and 60 MHz were used to determine if the ICH impurity threshold levels can be reached by NMR at accessible cost. The 400 MHz instrument used in this study represents cryogenic equipment that is widely available in laboratories of major manufacturers, university and major research institutions, while posing some accessibility limitations for smaller colleges and laboratories. On the other hand, the 60 MHz instrument employed here represents the kind of non-cryogenic benchtop instruments that are gaining popularity for their smaller initial investment and low maintenance requirements. These instruments are likely to become available with higher sensitivities, and increasingly accessible and/or affordable, to benefit those seeking to conduct impurity analysis of regulated compounds.

4.2. The role of the 400 MHz NMR spectrometer in this study

One practical goal of the present study was to determine what can be achieved with less than 3 hrs of acquisition when using a 400 MHz spectrometer. While the instrument used was equipped with a N₂-cryoprobe, it can still be considered representative of modern entry-level (“standard”) high-field NMR instruments. On a 400 MHz NMR spectrometer with a N₂-cryoprobe, as little as 2.5 μg of 2 was detected with a complete set of identifiable signals, in 2.5 hrs. Currently, many research labs are equipped with at least a 400 MHz NMR spectrometer, and many of them with a cryoprobe. Another consideration is the level of uncertainty required for the given analysis. As quantification of a minor impurity, such as the LOD determination of 2 (alone) in the present study, is acceptable at a higher uncertainty level (typically, a few percent), the use of not strictly quantitative conditions with much shorter pulse recycle times can be acceptable. Reducing the recycle time can shorten the measurement substantially (e.g., from 2.5 h to about 30 min when going from ~38 to ~7.5 s, while still retaining the adequate quantitative potential for LOD determination. Collectively, the methodology described in this study may be readily implemented and can be recommended as a prototype universal procedure for impurity analysis. In fact, a 400 MHz instrument can detect the targeted levels of 2 in a short time.

4.3. 60 MHz NMR spectrometers

The spectra acquired on 60 MHz NMR spectrometers represented in Fig. 2 shows expected increased overlap between signals of choline with those of 2 compared to 400 MHz. However, there were two distinct solvent-dependent features in the 60 MHz spectra that aided in the detection of 2 in the presence of 1. In DMSO-d₆, the hydroxy signals were separated by ~0.8 ppm with no interference from any other signals, making this a characteristic difference (Fig. 2A). With D₂O, the signal of one methylene group of 2 stands out for quick identification (Fig. 2B). The advantages of performing this analysis on a 60 MHz instrument are its speed, intuitive conceptual simplicity, and economic operation, which are poised to make the benchtop option an attractive choice for small analytical laboratories. Not only is the initial cost of the low-field non-cryogenic instrument much less than its high-field cryogenic counterpart, but operation expenses and cost-per-sample are much lower. However, the sensitivity limited the application to a relatively high content of the impurity

4.4. The optimization of spectra acquisition on the 60 MHz instrument

In order to probe another aspect of the practicability of low-field analysis, the study also assessed what can be achieved within ca. 30 min on a benchtop 60 MHz instrument when compared to an overnight measurement.

The optimization of spectral acquisition on the 60 MHz instrument relies on several factors. In this study, the 60 MHz spectra acquired in 35 min (64 scans with a 30 s relaxation delay) were compared to ones acquired in 9.6 h (1024 scans with a 30-s relaxation delay), which theoretically creates a 4-fold increase in signal to noise ratio (Figs. 4C (1024 scans) and S10 (64 scans); Figs. 4D (1024 scans) and S8 (64 scans)). Such an increase in the number of scans did not enhance the signal intensity as theory would predict, as the SNR was improved by a ~1.3 factor. From a general point of view, another potential means of gaining sensitivity, especially in low field NMR instruments, depends on the solubility of the analyte and the geometry of the probe: in cases where it is practical to reduce the solvent volume to e.g. to as low as 200 uL (matching a coil length of ~ 0.5 cm in the instrument used here), the increased sample concentration makes the sensitivity gain predictable.

4.5. ¹H iterative Full Spin Analysis (HiFSA) of choline (1) and impurity (2)

The semi-automated ¹H NMR spectral analysis of 1 and 2 (in D₂O) revealed the nature of their entire ¹H,¹H spin-spin coupling networks and resulted in digital fingerprints (Tables S12–S18) through which one can calculate the ¹H NMR spectra of these molecules. Both 1 and 2 have similar coupling patterns (Tables S13, S16). Spin-spin coupling between the nitrogen in the -N⁺(CH₃)₃ and the methylene groups (¹⁴N-¹H coupling), which was described earlier [15–17], was revealed during the full spin analysis. Despite the resolution enhancement of the ¹H NMR spectrum of 2 by Gaussian window functions, the NTM signal appeared as a “singlet” due to line broadening. However, the full spin analysis clearly reveals the coupling between ¹⁴N and the methyl groups, resulting in a triplet. Furthermore, the ¹⁴N coupling did not go beyond the ethereal oxygen in 2.

These HiFSA profiles are the basis of generating “calculated spectra” and can be calculated for any field strength. For example, to identify a substance unambiguously, a spectrum could be acquired on an 80 MHz NMR spectrometer, and compared with the spectrum calculated from its known HiFSA profile. HiFSA profiles being the actual representations of ¹H NMR spectrum in a numerical form, and independent of magnetic field, enable rapid identification of a given substance without a chemically identical reference standard, and this approach works at any practical field strength.

5. Conclusions

5.1. NMR: a useful tool for pharmaceutical analysis

NMR is nearly a universal detection technique for organic compounds as the only requirement is the presence of nuclei resonating within the range of measurement (presence of H in H-NMR), which represents an advantage over other techniques used in pharmaceutical impurity analysis that require specific structural characteristics to response. NMR spectroscopy provides for determination or confirmation of identity with a level of specificity that is much more difficult to achieve or even unattainable by other analytical methods. Furthermore, NMR measurements do not require physical reference standards chemically identical to the analytes, thereby departing from the conventional analytical paradigm relied upon same substance calibration-based methods. The presented 400 MHz NMR data and HiFSA analysis have unequivocally assigned chemical shifts and coupling constants of the analytes being examined. These assignments, while more challenging on low-field instrumentation, can nonetheless take advantage of the HiFSA data derived from high-field measurements scaled appropriately to any magnetic field. Availability of high quality raw FID data is a source of NMR spectrum for HiFSA and is a potential reference standard for rapid identification of organic compounds.

5.2. NMR as a quantification instrument

The LOD for direct measurement of 2 in the presence of 1 on a 400 MHz NMR spectrometer with cryoprobe is sufficient to satisfy the requirements for reporting thresholds in ICH, making possible the determination of LOQ as the next step. The use of low field instruments is at this time limited to the determination of impurities present at ~200 μg in the maximum amount of sample soluble in 500 μL. Depending on the sample size and solubility, concentrations as low as 0.2% can be detected directly. For high soluble substances such as choline salts in water used in this work, sample size in the range of 100 mg may be used to reach those detection levels. Detection at this level is still suitable for identified and known impurities with acknowledged low toxicity, where tolerances in the range of 0.2 to 2% may be appropriate for the maximum daily dose.

5.3. Fitness for practical implementation

This study showed that NMR analysis using high field instrumentation with cryoprobe is fit for detection of minor impurities at the 0.03% level required by ICH for drugs with maximum daily dose of more than 2 g/day. The proof of concept was established for choline, with the known impurity, 2. The limit of detection of the impurity was found to be 2.5 μg (1.34 μg of 2) in 500 μL D₂O, when analyzed alone and by acquiring 1024 scans on a 400 MHz NMR spectrometer equipped with a liquid-N₂ cryoprobe. Keeping in mind that LODs depend on the molecular weight and to some degree on the characteristics of the resonance that differ for each substance, an impurity with molecular weight similar to that of choline impurity 2 (275 amu) could be detected by NMR. The method was able to detect 10 μg of 2 in the presence of 10 mg of choline on the 400 MHz NMR spectrometer, in both DMSO-d₆ and D₂O (0.1%). Furthermore, the method was able to detect 12.5 μg of 2 was in the presence of 100 mg of 1 (0.0125%), suggesting applicability to the analysis of impurities of pharmaceutical APIs.

Using the 60 MHz benchtop NMR spectrometer, 400 μg of 2 (4%) was readily identified in the presence of 10 mg of choline i.e.: ~ 200 μg of 2 in 10 mg of 1 (2%). As NMR technology of the typical contemporary benchtop spectrometer (40–60 MHz) stands today, it would, thus, require a 200 ug/0.03% = 667 mg sample of choline to detect the impurity at the ICH threshold level of 0.03%. Apart from intrinsic solubility limitations, this theoretical “scale-up approach” is not feasible as such high sample amounts fall beyond the acceptable dynamic range for these instruments. While subtraction techniques might help bridge some of these challenges, it is more likely that a reduction of a matrix effect can be achieved through the development of sample preparation/extraction procedures. Also, advancements in instrument hardware may help future generations of low-field NMR instruments to bridge the sensitivity gap by scale up of sample amount and, thereby, achieve thresholds recommended in the ICH guidelines.

However, the detection levels reached with these instruments (0.2–2%) are still suitable for identified impurities with low toxicity where the tolerances are higher. In summary, this investigation confirmed the ability of ¹H NMR spectroscopy with entry-level high-field instruments to detect impurities in synthetic products. While it is common knowledge that a major analyte in a mixture can be identified and quantified by ¹H NMR, this work demonstrates successful detection and identification of minor impurities in the presence of closely structurally related major ingredient meeting the internationally acceptable identification thresholds, in other words, modeling the very common scenario of pharmaceutical manufacturing and analytical paradigms in which pharmaceutical and regulatory scientists routinely operate and where chromatographic separations are still the methodology of choice. In the present study, the LOD of 2, by itself and in the presence of 100 mg of choline chloride, was found to be significantly below the levels required by ICH (for quantitation) on 400 MHz NMR with cryoprobe.

This conclusion upends the prevalent notion that ¹H NMR sensitivity is incapable of meeting the analytical demands of contemporary pharmaceutical analysis. The findings confirm immediate and wide-ranging applicability of ¹H NMR spectroscopy to modern analytical tasks faced by everyday quality control personnel in every conceivable setting. Finally, archiving and sharing FIDs had immense value for other users [18] as reported in the present study, raw NMR data (from HMDB) of choline was useful for full spin analysis and identification. In other cases, it is possible that raw NMR data could contribute to enhanced the utility of qNMR assay, such as impurity threshold determination.

Fig. 6.

HiFSA profile of choline (1) in D₂O. The spectrum was obtained from HMDB and the calculations were performed on the webserver of Cosmic Truth software (ct.nmrsolutions.io). The results were exported as numeric spin parameters and JCAMP-DX files and are available as supplementary material.

HIGHLIGHTS.

• The LOD of choline RCA in choline is < 0.01% at 400 Hz, and 2% at 60 MHz

• Demonstrates feasibility of establishing choline purity and identity by (q)NMR

• Entry-level high- and low-field (400/60 MHz) instruments are suitable equipment

• ¹H NMR can detect choline impurity in the presence of choline at required LOD levels

• QM-based analysis reveals unknown ¹⁴N-¹H splitting in choline NMR resonance patterns