Quantitative ³¹P-NMR for Purity Determination of Sofosbuvir and Method Validation

Abstract

Quantitative ¹H-NMR (¹H-qNMR) is useful for determining the absolute purity of organic molecules; however, it is sometimes difficult to identify the target signal(s) for quantitation because of their overlap and complexity. Therefore, we focused on the ³¹P nucleus because of the simplicity of its signals and previously reported ³¹P-qNMR in D₂O. Here we report ³¹P-qNMR of an organophosphorus compound, sofosbuvir (SOF), which is soluble in organic solvents. Phosphonoacetic acid (PAA) and 1,4-bis(trimethylsilyl)benzene-d₄ (1,4-BTMSB-d₄) were used as reference standards for ³¹P-qNMR and ¹H-qNMR, respectively, in methanol-d₄. The purity of SOF determined by ³¹P-qNMR was 100.63 ± 0.95%, whereas that determined by ¹H-qNMR was 99.07 ± 0.50%. The average half bandwidths of the ³¹P signal of PAA and SOF were 3.38 ± 2.39 and 2.22 ± 0.19 Hz, respectively, suggesting that the T₂ relaxation time of the PAA signal was shorter than that of SOF and varied among test laboratories. This difference most likely arose from the instability in the chemical shift due to the deuterium exchange of the acidic protons of PAA, which decreased the integrated intensity of the PAA signal. Next, an aprotic solvent, dimethyl sulfoxide-d₆ (DMSO-d₆), was used as the dissolving solvent with PAA and sodium 4,4-dimethyl-4-silapentanesulfonate-d₆ (DSS-d₆) as reference standards for ³¹P-qNMR and ¹H-qNMR, respectively. SOF purities determined by ³¹P-qNMR and ¹H-qNMR were 99.10 ± 0.30 and 99.44 ± 0.29%, respectively. SOF purities determined by ³¹P-qNMR agreed with the established ¹H-qNMR values, suggesting that an aprotic solvent is preferable for ³¹P-qNMR because it is unnecessary to consider the effect of deuterium exchange.

Introduction

Currently, quantitative NMR (qNMR) is used in many fields as an absolute quantitative method to provide accurate quantitation values of analytes without additional reference standards (RSs). In addition, qNMR with International System of Units (SI) traceability can be performed using appropriate protocols. Therefore, qNMR, particularly ¹H-qNMR, is a suitable method to evaluate RSs. Previously, our group conducted a study on accurate qNMR with an internal reference substance (AQARI) to determine the purity of reagents used as RSs for HPLC assays for crude drugs and Kampo formulations in the Japanese Pharmacopoeia (JP).¹⁾ Validation studies showed that AQARI can be used to determine purity with an accuracy of approximately two significant figures for target reagents with a molecular weight of approximately 300 and mass of approximately 10 mg in 1 mL of deuterated solvent.^2,3) Furthermore, to perform ¹H-NMR measurements with the intended accuracy, we found that the coupling of the target compounds’ signals and the chemical shifts of RSs are important factors in addition to the handling of impurity signals from target compounds and RSs.^4,5) Based on these studies, AQARI has been adopted as a method for determining the purity of four reagents that are used as RSs for quantitative HPLC assays of the crude drug section of the 16th Edition Supplement 2 of the JP.⁶⁾ Moreover, we found that humidity affects the purity of the reagents. Using thermogravimetric analysis, we confirmed that the hygroscopicity of compounds could alter their purity by increasing the water content. Therefore, humidity control prior to and during weighing is essential for reproducible preparation, and the absolute amount, which is not affected by water content, is essential for hygroscopic compounds measured by ¹H-qNMR, rather than purity.^7,8) We also established an optimized and reproducible ¹H-qNMR method for sample preparation of hygroscopic marker compounds for crude drugs, such as ginsenoside Rb₁,⁸⁾ and hygroscopic chemical drugs, such as indocyanine green.⁹⁾ Through our series of studies, 19 reagents evaluated by ¹H-qNMR are listed as RSs for HPLC assays of 37 crude drugs and Kampo formula extract monographs up to the publication of JP 18th Edition.¹⁾

We established an optimized and reproducible ¹H-qNMR sample preparation method suitable for the various properties of the target compounds. However, depending on the purity and structure of the compound, it is sometimes difficult to select target signals for quantitation because of their overlap and complexity. To overcome this problem, we have considered qNMR determination with other nuclei because the target signals are simpler. First, we tried ¹³C-qNMR, but its lower sensitivity than ¹H-NMR make it difficult to determine quantities accurately. Then, we examined ³¹P-qNMR because the sensitivity of ³¹P is relatively high among NMR-active nuclei and there is a relatively large number of organophosphorus pharmaceutical compounds. Our first ³¹P-qNMR study with an organophosphorus anticancer drug, cyclophosphamide hydrate, revealed that the purity values determined by ³¹P-qNMR agreed well with the values determined by the established ¹H-qNMR method in a validation study in multiple laboratories.¹⁰⁾ In a previous study, D₂O was used as the dissolving solvent, and KH₂PO₄ or O-phosphorylethanolamine were used as RSs for ³¹P-qNMR because of the chemical shift and solubility of cyclophosphamide hydrate. This demonstrates that selecting appropriate RSs and solvents is important for ³¹P-qNMR measurements.¹⁰⁾

In this study, we selected sofosbuvir (SOF), which is an organophosphorus anti-hepatitis C virus drug that is soluble in organic solvents, to examine the appropriate RS and dissolving solvent for ³¹P-qNMR measurements. Multi-laboratory ³¹P-qNMR validation studies of SOF were conducted by comparison with ¹H-qNMR.

Experimental

The list of equipment and materials used for sample preparation in each laboratory is provided in Supplementary Table S1 (in methanol-d₄) and Table 1 (in dimethyl sulfoxide-d₆ (DMSO-d₆)), and those used for ¹H-qNMR and ³¹P-qNMR measurements in each laboratory are summarized in Supplementary Tables S2 and S3 (in methanol-d₄) and Tables 2, 3 (in DMSO-d₆).

Table 1. Instruments and Parameters Used for Sample Preparation in DMSO-d₆ at Different Laboratories

Laboratory	A	B	C	D	E	F	G	H	I	J
Sample preparation
Reference standard for 1H-qNMR	DSS-d6 (FUJIFILM Wako)
Reference standard for 31P-qNMR	Phosphonoacetic acid (Sigma-Aldrich)
Analyte	Sofosbuvir (AchemBlock)
Solvent	DMSO-d6	DMSO-d6	DMSO-d6	DMSO-d6	DMSO-d6	DMSO-d6	DMSO-d6	DMSO-d6	DMSO-d6	DMSO-d6
	>99.95 atom%D (acros)	99.96 atom%D (Merck)	99.96 atom%D (Aldrich)	99.96 atom%D (Aldrich)	99.9 atom%D (Kanto)	99.9 atom%D (ISOTEC), 99.9% (High purity) (FUJIFILM Wako)	>99.95 atom%D (MERCK)	99.9 atom%D (Cambridge Isotope Laboratories)	99.9 atom%D (ISOTEC)	99.9 atom%D (FUJIFILM Wako), 99.9% (High purity) (FUJIFILM Wako)
Balance	Ultramicro balance	Ultramicro balance	Micro balance	Micro balance	Micro balance	Ultramicro balance	Ultramicro balance	Ultramicro balance	Ultramicro balance	Ultramicro balance
Minimum indicated value (mg)	0.0001	0.0001	0.001	0.001	0.001	0.0001	0.0001	0.0001	0.0001	0.0001
Condition
Reference standard for 1H-qNMR: DSS-d6 (mg)	1.0327	1.0982	2.011	2.013	1.574	1.0537	1.0675	1.0152	1.0022	1.0065
Reference standard for 31P-qNMR: PAA (mg)	6.0156	6.1474	12.452	11.952	8.914	5.8293	6.0119	5.9655	6.0196	5.9310
Analyte volume: SOF (mg)	20.1469	19.8674	41.993	40.190	29.586	19.6370	20.1830	18.8965	20.0513	20.8355
Solvent volume (mL)	1	1	2	2	1.5	1	1	1	1	1
Equilibration period before weighing (h)*	1	1	1	1	1	1	1	3	2	1

* Keep PAA at 50% relative humidity (RH) or less.

Table 2. Instruments and Parameters Used for ¹H-qNMR Measurements in DMSO-d₆ at Different Laboratories

Laboratory	A	B	C	D	E	F	G	H	I	J
Observation nuclear	1H	1H	1H	1H	1H	1H	1H	1H	1H	1H
Spectrometer frequency	600 MHz	600 MHz	500 MHz	400 MHz	500 MHz	700 MHz	400 MHz	400 MHz	500 MHz	600 MHz
Probe type	Cryogenic (SuperCOOL)	Cryogenic (SuperCOOL)	Normal	Normal	Normal	Cryogenic	Normal	Normal	Normal	Normal
Spectral width	20 ppm	20 ppm	20 ppm	20 ppm	20 ppm	20 ppm	20 ppm	20 ppm	20 ppm	20 ppm
Pulse offset	5 ppm	5 ppm	5 ppm	5 ppm	5 ppm	5 ppm	5 ppm	5 ppm	5 ppm	5 ppm
Spinning	No	No	No	No	No	No	No	No	No	No
Digital filter	ON	ON	ON	ON	ON	ON	ON	ON	ON	ON
Pulse angle	90°	90°	90°	90°	90°	90°	90°	90°	90°	90°
Digital resolution	0.25 Hz	0.25 Hz	0.25 Hz	0.25 Hz	0.25 Hz	0.25 Hz	0.25 Hz	0.25 Hz	0.25 Hz	0.25 Hz
Relaxation deley time	60 s	60 s	60 s	60 s	60 s	60 s	60 s	60 s	60 s	60 s
Acquisition time	4 s	4 s	4 s	4 s	4 s	4 s	4 s	4 s	4 s	4 s
Measurement temperature	25 °C	25 °C	25 °C	25 °C	24 °C	30 °C	25 °C	25 °C	25 °C	25 °C
13C decoupling	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Decoupling sequence	MPF8	MPF8	MPF9	MPF9	MPF8	MPF9	MPF8	MPF9	MPF8	MPF8
Scan times	8	8	8	8	16	8	16	8	8	8
Dummy scan times	2	2	2	2	2	4	2	2	2	2
S/N (4.9 ppm)	1263	1390	796	267	1176	3382	954	660	1094	595

Table 3. Instruments and Parameters Used for ³¹P-qNMR Measurements in DMSO-d₆ at Different Laboratories

Laboratory	A	B	C	D	E	F	G	H	I	J
Observation nuclear	31P	31P	31P	31P	31P	31P	31P	31P	31P	31P
Spectrometer frequency	243 MHz (1H: 600 MHz)	243 MHz (1H: 600 MHz)	202 MHz (1H: 500 MHz)	162 MHz (1H: 400 MHz)	202.5 MHz (1H: 500 MHz)	162 MHz (1H: 400 MHz)	162 MHz (1H: 400 MHz)	162 MHz (1H: 400 MHz)	202 MHz (1H: 500 MHz)	243 MHz (1H: 600 MHz)
Probe type	Cryogenic (SuperCOOL)	Cryogenic (SuperCOOL)	Normal	Normal	Normal	Normal	Normal	Normal	Normal	Normal
Spectral width	50 ppm	50 ppm	50 ppm	50 ppm	50 ppm	50 ppm	50 ppm	50 ppm	50 ppm	50 ppm
Pulse offset	10 ppm	10 ppm	10 ppm	10 ppm	10 ppm	10 ppm	10 ppm	10 ppm	10 ppm	10 ppm
Spinning	No	No	No	No	No	No	No	No	No	No
Digital filter	ON	ON	ON	ON	ON	ON	ON	ON	ON	ON
Pulse angle	90°	90°	90°	90°	90°	90°	90°	90°	90°	90°
Acquisition time	2.3 s	2.3 s	3.3 s	2.3 s	2.6 s	2.5 s	2.4 s	8 s	4.0 s	2.7 s
Relaxation deley time	30 s	30 s	30 s	30 s	30 s	30 s	30 s	30 s	30 s	30 s
Digital resolution	0.43 Hz	0.43 Hz	0.3 Hz	0.43 Hz	0.39 Hz	0.4 Hz	0.42 Hz	0.125 Hz	0.25 Hz	0.37 Hz
Measurement temperature	25 °C	25 °C	25 °C	25 °C	24 °C	30 °C	25 °C	25 °C	25 °C	25 °C
1H decoupling	Inverse gated decoupling (No-NOE)	Inverse gated decoupling (No-NOE)	Inverse gated decoupling (No-NOE)	Inverse gated decoupling (No-NOE)	Inverse gated decoupling (No-NOE)	Inverse gated decoupling (No-NOE)	Inverse gated decoupling (No-NOE)	Inverse gated decoupling (No-NOE)	Inverse gated decoupling (No-NOE)	Inverse gated decoupling (No-NOE)
Decoupling sequence	Waltz	Waltz	Waltz16	Waltz16	Waltz	Waltz	Waltz	Waltz	WALTZ	Waltz
Scan times	32	32	64	32	64	64	64	64	32	64
Dummy scan times	2	2	2	2	2	2	2	2	4	2
S/N (−11.1 ppm: SOF)	429	420	479	274	244	739	290	340	155	266

Facilities

Ten investigators from 10 laboratories (A–J) performed separate experiments.

Reagents, RSs for ¹H-qNMR and ³¹P-qNMR, and Solvents

SOF was purchased from Advanced ChemBlocks (Hayward, CA, U.S.A.). Sodium 4,4-dimethyl-4-silapentanesulfonate-d₆ (DSS-d₆; IUPAC name: sodium 3-(trimethylsilyl)propane-1-sulfonate-1,1,2,2,3,3-d₆) (MW = 224.36), 1,4-bis(trimethylsilyl)benzene-d₄ (1,4-BTMSB-d₄) (MW = 226.50), and certified reference materials (CRMs) traceable to the National Metrology Institute of Japan/National Institute of Advanced Industrial Science and Technology were purchased from FUJIFILM Wako Pure Chemical Corporation (Osaka, Japan) and used as the RSs for ¹H-qNMR. Phosphonoacetic acid (PAA) [MW = 140.03 (C₂H₅O₅P)], a qNMR CRM with SI traceability (TraceCERT^®; purity: 99.23%), was purchased from Sigma-Aldrich (MO, U.S.A.) and used as an RS for ³¹P-qNMR. Methanol-d₄ (>99.8% atom% D) and DMSO-d₆ (>99.9% atom% D) were used as deuterated solvents for qNMR (Supplementary Table S1 and Table 1, respectively). The structures of SOF and qNMR CRMs (DSS-d₆, 1,4-BTMSB-d₄, and PAA) are shown in Fig. 1.

Fig. 1. Structures of Sofosbuvir (SOF), and qNMR CRMs Traceable to the SI Reference Standards of ¹H- and ³¹P-NMR [DSS-d₆, 1,4-BTMSB-d₄, and Phosphonoacetic Acid (PAA)]

*¹ IUPAC name: sodium 3-(trimethylsilyl)propane-1-sulfonate-1,1,2,2,3,3-d₆.

Instruments and Equipment

An ultra-microbalance and a microbalance with readability of 0.0001 and 0.001 mg, respectively, were used (Supplementary Table S1 in methanol-d₄ and Table 1 in DMSO-d₆). A 700 MHz NMR spectrometer equipped with a cryogenic probe, two 600 MHz NMR spectrometers equipped with a supercool (cryogenic) probe, and one 600 MHz, three 500 MHz, and three 400 MHz NMR spectrometers equipped with normal (room temperature) probes were used for ¹H-qNMR measurements (Supplementary Table S2 in methanol-d₄ and Table 2 in DMSO-d₆). For ³¹P-qNMR measurements, two 243 MHz NMR spectrometers equipped with a supercool (cryogenic) probe and one 243 MHz, three 202 MHz, and four 162 MHz NMR spectrometers equipped with normal probes were used (Supplementary Table S3 in methanol-d₄ and Table 3 in DMSO-d₆).

Preparation of Sample Solutions

NMR Validation Test

Approximately 20–40 mg of SOF and 1–2 mg of the RS for ¹H-qNMR (1,4-BTMSB-d₄ or DSS-d₆) or 6–12 mg of the RS for ³¹P-qNMR (PAA) were precisely weighed, placed in the same vial for each tare, and dissolved in the NMR solvent (methanol-d₄ or DMSO-d₆; 1–2 mL). The sample solution (0.6 mL) was added to an NMR tube and sealed (Supplementary Table S1 in methanol-d₄ and Table 1 in DMSO-d₆).

Moisture Adsorption and Desorption Analysis

A vapor sorption analyzer (Dynamic Vapor Sorption Intrinsic Plus, Surface Measurement Systems Ltd., U.K.) was used under the following conditions: purge gas, nitrogen; feed rate in a humidity chamber, 200 mL/min; sample weight, 5–8 mg; sample pan, stainless steel; temperature in weight equilibration determination, 25 °C; relative humidity, increased stepwise from 0 to 80% for two cycles.

Humidity Conditions

First study (methanol-d₄): the humidity recorded in nine laboratories was 30%–48% (data not shown). Second study (DMSO-d₆): the observed humidity in 10 laboratories was 11%–35% (data not shown). SOF, 1,4-BTMSB-d₄, DSS-d₆, and PAA were equilibrated for 1–3 h under each set of conditions before weighing (Supplementary Tables S1, Table 1).

Conditions for qNMR

The instruments and parameters employed for sample preparation in each laboratory are summarized in Supplementary Table S1 (methanol-d₄) and Table 1 (DMSO-d₆). Supplementary Table S2 (in methanol-d₄) and Table 2 (in DMSO-d₆) show the instruments and parameters used for ¹H-qNMR measurements in each laboratory. The RSs for ¹H-qNMR (1,4-BTMSB-d₄, DSS-d₆) were also used as the chemical shift reference signal (0 ppm). The δ values are expressed in ppm. The observed spectral width was 20–25 ppm, and a digital filter was used. The center of the spectrum was set to 5 ppm, and the pulse width was set to the time at which a 90° pulse was obtained. The acquisition time was 4 s, the digital resolution was 0.25 Hz, and the delay time was 60 s. An auto field gradient shim or TopShim was used for shimming. The determination temperatures were 24, 25, and 30 °C. ¹³C decoupling was performed using MPF8 or MPF9. Scans were performed 8 or 16 times, and a dummy scan was performed 2 or 4 times.

Supplementary Table S3 (in methanol-d₄) and Table 3 (in DMSO-d₆) summarize the instruments and parameters employed for ³¹P-qNMR measurements in each laboratory. The RS for ³¹P-qNMR (PAA) was used as the chemical shift reference signal (0 ppm). The observed spectral widths were 50, 51, or 86 ppm, and a digital filter was used. The center of the spectrum was set at 10, 11, 11.23, or 11.32 ppm, and the pulse width was set to the time at which a 90° pulse was obtained. The acquisition time was 2.2–8 s, the digital resolution was 0.125–0.46 Hz, and the delay time was 30 s. An auto field gradient shim or TopShim was used for shimming. The determination temperatures were 24, 25, and 30 °C. ¹H decoupling with inverse-gated decoupling (No-NOE) was conducted. Scans were performed 32 or 64 times, and dummy scans were performed two or four times.

In principle, the measurements were performed three times for each sample following the internal standard method (AQARI) to ensure that the signal-to-noise (S/N) ratio of the quantitative signal was 100 or higher (Tables 2, 3, Supplementary Tables S2, S3). ALICE 2 for qNMR (JEOL Tokyo, Japan), Purity Pro (JEOL), Delta (JEOL), TopSpin (Bruker MA, U.S.A.), and Mnova (Mestrelab Research, Santiago de Compostela, Spain) software were used for NMR data processing.

The trimethylsilyl signal of the RSs for ¹H-qNMR (1,4-BTMSB-d₄ and DSS-d₆) and the phosphorus signal of the RS for ³¹P-qNMR (PAA) were set as 0 ppm. Phase correction, baseline correction, and determination of the signal integration ranges were performed manually or automatically. All the integrated values in this study are expressed in terms of purity (%). The purity of the reagents was calculated using the equation described in a previous study.¹⁰⁾

The following values were used for the calculations: number of methyl group protons in 1,4-BTMSB-d₄ (RS for ¹H-qNMR) of 18 (CH₃ × 6); molecular weight of 1,4-BTMSB-d₄ of 226.50; number of methyl group protons in DSS-d₆ (RS for ¹H-qNMR) of 9 (CH₃ × 3); molecular weight of DSS-d₆ of 224.36; molecular weight of SOF of 529.45 (C₂₂H₂₉FN₃O₉P); molecular weight of PAA (RS for ³¹P-qNMR) of 140.03 (C₂H₅O₅P).

Results and Discussion

Solvent and RSs for ¹H-qNMR and ³¹P-qNMR

A protic solvent, methanol-d₄, was selected for ¹H- and ³¹P-qNMR analyses of SOF because of the solubility properties of the molecule. In addition, because of its solubility, 1,4-BTMSB-d₄, an SI-traceable CRM, was selected as the ¹H-qNMR RS. PAA can be used as a commercially available CRM for ³¹P-qNMR.¹¹⁾ Thus, we chose PAA (δP: approximately 18.3 ppm) because of its solubility characteristics, and the ³¹P phosphorus chemical shift was referenced to PAA at 0 ppm. The chemical shift of SOF was as follows: δP of approximately 4.4 ppm (−13.9 ppm in this study) in methanol-d₄.

In the second experiment, DMSO-d₆ was selected as the aprotic solvent. Because DSS-d₆ (a CRM) was chosen as the ¹H-qNMR RS, PAA was also used as the RS for ³¹P-qNMR (δP: approximately 15.6 ppm in DMSO-d₆) due to its solubility and the chemical shift of SOF in DMSO-d₆ [δP: approximately 4.4 ppm (−11.1 ppm in this study)].

SOF NMR Assignment

There are no reports describing NMR assignment of SOF; therefore, the ¹H- and ¹³C-NMR assignments of SOF in DMSO-d₆ are given in Supplementary Table S4, and the two-dimensional correlations of SOF are shown in Supplementary Fig. S1.

Relaxation Delay Time for ³¹P-qNMR Measurements

For quantitative NMR experiments, a relaxation delay time exceeding seven times that of T₁ is required for stabilization.¹²⁾ The T₁ values of the ³¹P signals of SOF and PAA in methanol-d₄ were 1.6 and 1.8 s and those in DMSO-d₆ were 0.5 and 1.5 s, respectively. Therefore, the relaxation delay time for ³¹P-qNMR spectroscopy was set to at least 30 s.

Hygroscopicity of SOF and PAA

The hygroscopicity of SOF and PAA were investigated using moisture adsorption and desorption analyses prior to quantitative NMR measurements. Under conditions of 0–80% relative humidity, the weight of SOF did not change (0.1% or less in 1 h) (data not shown). In contrast, the weight of PAA increased drastically and it deliquesced at a relative humidity of ˃50%. Therefore, weighing and sample preparation were performed at a relative humidity of < 50%.

¹H- and ³¹P-qNMR Measurements in a Protic Solvent (Methanol-d₄)

In the first study, the conditions summarized in Table 4 were used to determine the purity of SOF by ¹H- and ³¹P-qNMR measurements. A solution of SOF, 1,4-BTMSB-d₄, and PAA in methanol-d₄ was prepared. The ¹H- and ³¹P-qNMR spectra of SOF in methanol-d₄ are shown in Figs. 2A–C, and the quantitative data are given in Tables 5–7.

Table 4. Purity (%) of SOF Determined by ¹H- and ³¹P-qNMR during the First and Second Study in Eight to 10 Laboratories

Study	Reference standard for qNMR*		Analyte	Solvent	1H-qNMR Purity ± S.D. (%) (n = 3, eight or 10 Labs)	31P-qNMR Purity ± S.D. (%) (n = 3, eight or nine Labs)
	1H	31P
1st	1,4-BTMSB-d4	Phosphoacetic acid (PAA)	SOF	Methanol-d4	Signal H-3 99.07 ± 0.50	100.63 ± 0.95
		PAA (Purity: determined by 1,4-BTMSB-d4)			—	100.36 ± 0.53 (PAA: 99.22 ± 0.31)
2nd	DSS-d6	PAA	SOF	DMSO-d6	Signal H-28 99.44 ± 0.29	99.10 ± 0.30
		PAA (Purity: determined by DSS-d6)			—	99.34 ± 0.55 (PAA: 99.46 ± 0.47)

* Purity of 1,4-BTMSB-d₄: 100.0% (CRM); purity of DSS-d₆: 92.4% (CRM); purity of PAA: 99.23% (CRM); S.D.: Standard deviation

Fig. 2. Typical ¹H-qNMR (A and B) and ³¹P-qNMR (C) Spectra of SOF with PAA and 1,4-BTMSB-d₄ in Methanol-d₄ Obtained during the First Study

(B) shows the enlarged spectra of each signal in (A).

Table 5. Purities (%) of SOF Determined by ¹H-qNMR in Methanol-d₄ at Nine Laboratories

Position		Laboratory									Average (%) of nine labs	S.D. (%) of nine labs
		A (2)	B	C	D	E	G	H	I	J
5.4 ppm	Average (%)	98.94	98.87	98.49	98.26	99.02	99.82	99.23	99.50	99.52	99.07	0.50
	S.D. (%)	0.04	0.66	0.17	0.07	0.41	0.19	0.13	0.22	0.23

S.D.: Standard deviation

Table 6. Purities (%) of SOF Determined by ³¹P-qNMR in Methanol-d₄ at Nine Laboratories

Position		Laboratory									Average (%) of nine labs	S.D. (%) of nine labs
		A (2)	B	C	D	E	G	H	I	J
−13.9 ppm	Average (%)	100.94	99.77	100.81	99.91	100.19	100.56	101.15	99.65	102.73	100.63	0.95
	S.D. (%)	0.97	0.53	0.16	0.02	0.86	0.13	0.36	0.28	0.23

S.D.: Standard deviation

Table 7. Purities (%) of PAA as a Reference Standard for ³¹P-qNMR Determined by ¹H-qNMR in Methanol-d₄ at Eight Laboratories

Position		Laboratory								Average (%) of eight labs	S.D. (%) of eight labs
		A (2)	B	C	D	E	G	H	I
2.6 ppm	Average (%)	99.18	99.25	98.92	99.70	98.74	99.47	99.09	99.43	99.22	0.31
	S.D. (%)	0.04	0.13	0.03	0.17	0.27	0.03	0.11	0.03

S.D.: Standard deviation

In the ¹H-qNMR spectrum of the mixture of SOF and PAA with 1,4-BTMSB-d₄ in a protic solvent (methanol-d₄), the doublet signal of H-3 (δ_H 5.4 ppm) of SOF was selected as the target signal for quantitation considering its multiplicity, sufficient separation, and height (Fig. 2A). The purity value of SOF determined by ¹H-qNMR in nine laboratories using 1,4-BTMSB-d₄ as the RS was 99.07 ± 0.50% (Table 5).

In the ¹H-NMR spectrum of the mixture, the methylene signal of PAA was observed as a doublet at δ_H 2.64 ppm (J = 21.5 Hz) due to PO(OD)(OH)-CH₂-COOH because of the coupling with the neighboring ³¹P nucleus. Because PAA has acidic deuterium-exchangeable protons, namely, phosphoric and carboxylic protons, new doublet signals were gradually observed at δ_H 2.69 ppm (J = 21.4 Hz) and δ_H 2.62 ppm (J = 21.4 Hz) due to the PO(OD)(OH)-CH₂-COOD and PO(OD)₂-CH₂-COOH methylenes, respectively. The former doublet is shown on the right side in Fig. 2B, and the latter doublet, which is not visible in the figure, appeared as a small shoulder signal 9 h after sampling. The integrated region of the methylene signal included all these doublet signals, which means that the deuterium exchange of the acidic protons did not affect the integrated values. However, this is the active methylene, and the active methylene protons themselves are deuterium exchangeable. In fact, 23 h after sampling at 25 °C, a decrease of approximately 1% of the total intensity of the signals was observed. Therefore, before the validation studies, the time-course for the decrease was examined, and the results suggested that the decrease in intensity was negligible when the study was performed within 4 h. Therefore, the validation study followed this rule, and the average purity of PAA determined by ¹H-qNMR in nine laboratories was 99.22 ± 0.31%, which is the same as the certified value of 99.23% (Table 7).

We then performed a ³¹P-qNMR analysis of the mixture. The ³¹P-spectrum is shown in Fig. 2C. The enlarged spectra in Fig. 2C show the results of a ³¹P-qNMR time course study of PAA. The sideband signal was assigned as ¹³C coupling and the gradually increasing small signal at δ_P −0.65 ppm was assigned as PO(OD)(OH)-CH₂-COOD because the long-range correlation between the signal and the proton doublet at δ_H 2.69 ppm was observed by ¹H-³¹P HMBC (data not shown). Unfortunately, the signal corresponding to PO(OD)₂-CH₂-COOH was not observed, and we believe it overlapped with the main PAA signal. Therefore, we decided to include the carbon satellite and the deuterium shifted signal at δ_P −0.65 ppm in the integrated region of PAA. For the SOF signal, we observed a small shoulder signal, which did not increase after 17 h. Given that the estimated coupling constant (around 10 Hz) was within the literature values¹³⁾ and its signal height was around 1.3% of the main signal, we deduced that this signal was the overlapping counterpart signals caused by three ²J_31P-13C couplings and that the other counterpart signals thought to be buried in the main signal. Therefore, we decided to integrate the signal including this small shoulder as the SOF signal.

The determination of SOF by ³¹P-qNMR was conducted at nine laboratories and resulted in a mean purity value of 100.63 ± 0.95% (Table 6), which was greater than the value determined by ¹H-qNMR (99.07 ± 0.50%; Table 5). Considering that the purity value of PAA determined by ¹H-qNMR coincided with the certified value, we decided to investigate the cause of the 1.6% increase in purity in the ³¹P-qNMR experiments because the integrated value of the shoulder impurity signal in SOF was negligible (Fig. 2C). It is well-known that the half bandwidth of signal is inversely correlated with its T₂ relaxation time. Therefore, we checked the half bandwidth of the ³¹P signal of PAA, and compared it with that of SOF. The obtained values for PAA and SOF from nine laboratories ranged from 1.42 to 9.18 Hz and from 1.96 to 2.48 Hz, and their averages were 3.38 ± 2.39 and 2.22 ± 0.19 Hz, respectively. This suggests that the T₂ relaxation time of the PAA signal was shorter than that of SOF and varied significantly among the test laboratories. The acidic protons, including the active methylene of PAA, are gradually deuterated, causing instability in the PAA chemical shift. The required range for integration of the signal for sufficient quantification depends on the half bandwidth of the signal, and ¹H-qNMR for the range of 99.9% determination is 630 times the half bandwidth.¹⁴⁾ Therefore, the unstable and relatively small T₂ relaxation time of PAA due to the exchange of acidic proton(s) to deuterated one(s) appeared to result in the smaller integrated intensity of the PAA signal than the certified purity of 99.23% of the CRM, and this in turn appeared to cause the higher purity value of SOF (100.63 ± 0.95%; Table 6). It should be noted that the deuterium exchange appearing in a protic solvent such as methanol-d₄ does not cause a conceivable problem for ¹H-qNMR, as in most cases for quantification there are some selectable signals that remain unaffected by the exchange.

¹H- and ³¹P-qNMR Measurements in an Aprotic Solvent, DMSO-d₆

In the second experiment, an aprotic solvent, DMSO-d₆, was used with DSS-d₆ and PAA as RSs for ¹H and ³¹P nuclei, respectively (Table 4). The ¹H-qNMR and ³¹P-qNMR spectra of the mixture are shown in Fig. 3. In the ¹H-qNMR spectrum, we observed a broad impurity signal for water at approximately δ_H 6 ppm (Fig. 3A). Considering the impurity, the simplicity of coupling, and signal overlap in the 400 MHz NMR, we selected the septet signal at δH 4.9 ppm from the C28 methine proton. Even after the last (third) measurement, we did not observe a new small signal derived from deuterium exchange (Figs. 3A, 3B). In the ³¹P-qNMR spectrum, we observed a shoulder signal in the PAA signal with ¹³C satellite signals, which did not increase after 24 h (Fig. 3C). The signal height was around 0.5% of the main signal and the estimated coupling constant was around 10 Hz. Therefore, we assigned it as ²J_31P-13C coupling by the carboxyl carbon and decided to integrate it with PAA. The purity values of SOF as determined by ¹H-qNMR and ³¹P-qNMR in DMSO-d₆, which were obtained in 10 and nine laboratories, were 99.44 ± 0.29 and 99.10 ± 0.30%, respectively (Tables 8, 9). The ranges of the half bandwidth of PAA and SOF were from 0.34 to 0.98 Hz and from 1.43 to 3.35 Hz, and their averages were 0.74 ± 0.22 and 2.22 ± 0.74 Hz, respectively. Considering the integrated ranges, the averages of both half bandwidths were sufficiently small, and they did not vary significantly among the 10 laboratories. Notably, the average purity of PAA determined by ¹H-qNMR in 10 laboratories was 99.46 ± 0.47%, which was the same as the certified value of 99.23% (Table 4). Because the SOF purities measured by ³¹P-qNMR agreed with the established ¹H-qNMR values, we suggest that an aprotic solvent is preferable for ³¹P-qNMR, which uses PAA as an RS for organic compounds because it is not necessary to consider the effect of deuterium exchange.

Fig. 3. Typical ¹H-qNMR (A and B) and ³¹P-qNMR (C) Spectra of SOF with PAA and DSS-d₆ in DMSO-d₆ Obtained during the Second Study

(B) shows the enlarged spectra of each signal in (A).

Table 8. Purities (%) of SOF Determined by ¹H-qNMR in DMSO-d₆ at 10 Laboratories

Position		Laboratory										Average (%) of 10 labs	S.D. (%) of 10 labs
		A	B	C	D	E	F	G	H	I	J
4.9 ppm	Average (%)	99.17	98.84	99.87	99.69	99.48	99.39	99.33	99.64	99.42	99.59	99.44	0.29
	S.D. (%)	0.28	0.20	0.11	0.12	0.34	0.03	0.09	0.11	0.09	0.05

S.D.: Standard deviation

Table 9. Purities (%) of SOF Determined by ³¹P-qNMR in DMSO-d₆ at Nine Laboratories

Position		Laboratory									Average (%) of nine labs	S.D. (%) of nine labs
		A	B	C	D	E	F	G	H	I
−11.1 ppm	Average (%)	99.88	98.71	99.01	98.89	99.47	98.98	99.31	99.62	99.02	99.10	0.30
	S.D. (%)	0.15	0.16	0.53	0.58	0.13	0.17	0.13	0.06	0.29

S.D.: Standard deviation

³¹P-qNMR Measurements in a Protic Solvent (Methanol-d₄) Using a High-Sensitivity Probe

Deuterium exchange occurs dynamically; therefore, faster measurement reduces the inconvenience of using methanol-d₄ as a solvent. Accordingly, we investigated the use of a cryogenic probe (Supercool probe), which has a high sensitivity for ³¹P nuclei, in Laboratory A [A(1), Supplementary Tables S1–S3]. The ³¹P-qNMR measurement time of SOF was shorter (90 min) than that with the normal room-temperature probe (180 min), and a similar S/N ratio was obtained. The mean SOF purity values determined by ¹H- and ³¹P-qNMR in methanol-d₄ following three measurements were 99.26 ± 0.29 and 99.54 ± 0.26%, respectively (data not shown), which were similar. Therefore, we believe that experimental time is essential for precisely quantifying ³¹P-qNMR, especially when using a protic solvent with PAA as the RS.

Conclusion

In this study, SOF, an organophosphorus compound soluble in organic solvents, was selected as the target compound, and the appropriateness of a different RS and dissolving solvents was investigated to establish a ³¹P-qNMR absolute determination method. PAA was used as the RS for ³¹P-qNMR in a protic solvent, methanol-d₄. The SOF purity determined by ³¹P-qNMR (100.63 ± 0.95%) was 1.6% higher than that determined by ¹H-qNMR (99.07 ± 0.50%). This suggests that using a protic solvent, such as methanol-d₄, is inappropriate for ³¹P-qNMR because of the deuterium exchange with PAA, resulting in a small integrated intensity. Consequently, an aprotic solvent, DMSO-d₆, was used to determine the purity of SOF. The data revealed that the SOF purities determined by ³¹P-qNMR (99.10 ± 0.30%) agreed well with the established ¹H-qNMR values (99.44 ± 0.29%) and suggested that the absolute quantitation of SOF using both ³¹P-qNMR and ¹H-qNMR is possible in DMSO-d₆. When a target organophosphorus compound or RS with exchangeable protons in the molecule is used for purity determination by ³¹P-qNMR, an aprotic organic solvent, such as DMSO-d₆, avoids deuterium exchange affecting the results. The ³¹P-qNMR spectrum is simpler than the ¹H-qNMR spectrum. Typically, the ³¹P-qNMR spectrum only contains two signals; one from the RS and the other from the target compound. Therefore, it is an attractive technique for determining the purity of complicated organic compounds with ³¹P nuclei. However, it is sometimes difficult to select an appropriate RS, considering the solubility, chemical shift, acid or base stability, and deuterium exchange ability of both the target compound and RS. Therefore, it is necessary to develop new RSs for ³¹P-qNMR and a higher-sensitivity probe for the ³¹P nuclei.

Acknowledgments

This research was supported by the Japan Agency for Medical Research and Development (AMED) (Grant Nos. JP21mk0101129 and JP22mk0101220).

Conflict of Interest

The authors declare no conflict of interest.

Supplementary Materials

This article contains supplementary materials.

References

• 1) The Ministry of Health, Labour and Welfare, Japan. “The Japanese Pharmacopoeia Eighteen Edition (2021).”: ‹https://www.mhlw.go.jp/stf/seisakunitsuite/bunya/0000066597.html›, cited 16 July, 2022.
• 2) Hosoe J., Sugimoto N., Goda Y., Pharmaceutical and Medical Device Regulatory Science, 41, 960–970 (2010).
• 3) Hosoe J., Sugimoto N., Suematsu T., Yamada Y., Hayakawa M., Katsuhara T., Nishimura H., Goda Y., Pharmaceutical and Medical Device Regulatory Science, 43, 182–193 (2012).
• 4) Hosoe J., Sugimoto N., Suematsu T., Yamada Y., Miura T., Hayakawa M., Suzuki H., Katsuhara T., Nishimura H., Kikuchi Y., Yamashita T., Goda Y., Pharmaceutical and Medical Device Regulatory Science, 45, 243–250 (2014).
• 5) Goda Y., Pharmaceutical and Medical Device Regulatory Science, 48, 615–619 (2017).
• 6) The Ministry of Health, Labour and Welfare, Japan. “The Japanese Pharmacopoeia Sixteen Edition Supplement II (2014)”: ‹https://jpdb.nihs.go.jp/kyokuhou/files/000152770.pdf›, cited 12 August, 2022.
• 7) Uchiyama N., Hosoe J., Miura T., Sugimoto N., Ishizuki K., Yamada Y., Iwamoto Y., Suematsu T., Komatsu T., Maruyama T., Igarashi Y., Higano T., Shimada N., Goda Y., Yakugaku Zasshi, 140, 1063–1069 (2020).
• 8) Uchiyama N., Hosoe J., Miura T., Sugimoto N., Ishizuki K., Yamada Y., Iwamoto Y., Suematsu T., Komatsu T., Maruyama T., Igarashi Y., Higano T., Shimada N., Goda Y., Chem. Pharm. Bull., 69, 26–31 (2021).
• 9) Uchiyama N., Hosoe J., Sugimoto N., et al., Chem. Pharm. Bull., 69, 118–123 (2021).
• 10) Uchiyama N., Hosoe J., Sugimoto N., et al., Chem. Pharm. Bull., 69, 630–638 (2021).
• 11) Rigger R., Rück A., Hellriegel C., Sauermoser R., Morf F., Breitruck K. B., Obkircher M., J. AOAC Int., 100, 1365–1375 (2017).
• 12) qNMR primary guide. “A guide to quantitative analysis for beginners-from basics to practice,” qNMR primary guide working group, Kyoritsu publication, Tokyo, 2015, pp. 96–97.
• 13) Pretsch E., Bühlmann P., Badertscher M., “Structure Determination of Organic Compounds. Table of Spectral Data,” 5th ed. Springer, Berlin, 2020.
• 14) The European Network of Forensic Science Institutes (ENFSI). “Guideline for qNMR analysis, Ver. 001.”: ‹https://enfsi.eu/wp-content/uploads/2017/06/qNMR-Guideline_version001.pdf›, cited 12 August, 2022.