2.2. NMR Methodology

All ¹H NMR measurements were performed using a Bruker Ascend 400 spectrometer (400 MHz) with a PA BBI 400S1 H-BB-D-05 probe head or a Bruker UltraShield 400 MHz with a 5mm PASE 1H/D-13C Z-GRD probe head, both equipped with an automatic sample changer (SampleXpress H15000-01) (all equipment from Bruker Biospin, Rheinstetten, Germany). All spectra were recorded at 300.0 K after 5 min for thermal equilibration. For sample preparation, approximately 100 ± 2 mg of the sample was dissolved in 0.6 mL solvent which contained 0.05% TMS. The resulting solution was placed for NMR measurement in a 5 mm NMR sample tube (Deutero, Kastellaun, Germany).

In order to identify an appropriate solvent for the cannabinoid quantification experiment, NMR spectra of pure cannabidiol and commercial CBD oils based on hemp seed oil and MCT oil were recorded. Prior to the NMR experiments, the samples were dissolved in either pure solvent (CD₃OD, CDCl₃) or solvent mixtures (CDCl₃/CD₃OD (3:2, 2:1, v:v) and CDCl₃/DMSO-d₆ (5:1, v:v)). These solvents and solvent mixtures were previously proposed in the literature for NMR measurement of hemp extracts or edible oils [9,19,20,22,23,24,25]. To assign the cannabinoid signals in the ¹H NMR spectrum of CBD oils, reference spectra of different cannabinoids were recorded in CDCl₃ and their compound-specific coupling constants, multiplicities, and chemical shift were recorded and verified with literature data. The reference substances were selected based on the main naturally occurring cannabinoids of hemp and the recommendation of the EU Commission (EU) 2016/2115 on the monitoring of ∆⁹-THC, its precursors and other cannabinoids in foods [26]. Therefore, CBD, ∆⁹-THC, ∆⁸-THC, CBN, CBG and THCV were measured, as well as the cannabinoid acids THCA and CBDA. For signal assignment of the individual cannabinoids, one-dimensional ¹H NMR spectra were recorded for various cannabinoids and spike solutions with and without oil matrices were performed (Table 1). In case of ∆⁹-THC, ∆⁸-THC and THCV, the nondeuterated methanol of the commercial products were removed prior to the NMR experiment: the solutions were evaporated to dryness for approximately 30–90 min at 45 °C via nitrogen evaporation depending on the quantities and redissolved with deuterated NMR solvent.

Acquisition parameters of the initial NMR experiments for method development.

The NMR spectra were recorded using an adapted composite experiment from Bruker consisting of two one-dimensional ¹H NMR experiments (Table 2) including a pulse sequence that suppresses selected intense lipid signals in the spectrum. The lipid signals are suppressed in the frequency ranges of 347–357 Hz (0.87–0.89 ppm), 504–524 Hz (1.26–1.31 ppm), 640–650 Hz (1.60–1.63 ppm), 803–823 Hz (2.01–2.06 ppm), 919–929 Hz (2.30–2.32 ppm), 1102–1112 Hz (2.75–2.78 ppm), and 2116–2156 Hz (5.29–5.39 ppm). The multiple suppression was implemented to increase the signal intensity of minor components, e.g., the cannabinoids. Detailed descriptions of multiple suppression for oil matrix are given by Ruiz-Aracama et al. [27] and Longobardi et al. [28]. An automatic estimation of the 90° pulse width (P1) for each sample was implemented, because the effectiveness of the 90° pulse varies from sample to sample depending on their physicochemical properties [29].

Acquisition parameters of the developed multiple suppression NMR experiment.

^a automatic pulse estimation.

In order to achieve correct signal intensities and to avoid systematic measurement errors in the NMR method, the acquisition parameters were carefully optimized during method development. The receiver gain (RG) was set to 16 after optimization via multiple suppression. The temperature effects were investigated with a CBD and ∆⁹-THC spiked hemp seed oil in CDCl₃ between 280 K and 320 K. In addition, complete relaxation of the analyte signals and thus coherent signal intensity are necessary to avoid underestimation of the analyte concentrations. To verify the commonly chosen relaxation delay of five times the longitudinal relaxation time (T1), T1 was determined for the selected cannabinoid signals by means of an inversion recovery experiment.

The raw spectra were processed using the software TopSpin 3.2 (Bruker Biospin, Rheinstetten, Germany). The time domain was set to 131,072 data points with a spectral width of 20.5617 ppm. The size of real spectrum (SI) was extended to 262,144 by zero filling. For further spectra processing, the free induction decay (FID) was multiplied with an exponential window function to achieve a line broadening of 0.30 Hz and the spectra were automatically phase- and baseline-corrected.

The quantification was performed using PULCON (PULse length based CONcentration determination) which is based on the use of an external standard [29,30]. The technique directly relates the measured signal intensity per proton of the analyte to the measured signal intensity per proton of a calibration standard of known concentration for determination of the absolute concentration [30].

The external standard (also called quantification reference, QR) was composed of tetrachloronitrobenzene (TCNB, 4662 mg/L) and ethylbenzene (EB, 3506 mg/L) dissolved in CDCl₃. Note that for correct signal intensities, the device-specific response (ERETIC factor, f_ERETIC) relies on the full relaxation with T1 as the minimum spectra acquisition time [30]. Due to the long T1 of TCNB and EB in CDCl₃ (10.9 s and 8.8 s, respectively), empirical spectroscopic correction factors (1.30 for TCNB; 1.22 for EB) were added to the ERETIC factor-defining equation to correct the signal intensity (Equation (S1)). This enables a reduction of the measurement time to 14 s by a factor of 4 compared to complete relaxation. Due to the suppression ranges for the fatty acids, the signals at 1.24 and 2.65 ppm cannot be used for the calculation of the ERETIC factor. The factor is therefore calculated exclusively from the two signals of the respective aromatic protons. The defined signal ranges of TCNB (7.64–7.83 ppm, 1 proton, singlet) and EB (7.0162–7.2341 ppm, 3 protons, multiplet) were integrated and the average ERETIC factor was used for further calculations.

Using this modified ERETIC factor, the PULCON relation (Equation (S2)) yields the mass concentration of the analyte in (mg/L sample solution) based on NMR signal areas. Note that all experiments were recorded with the same NMR spectrometer as this relationship is only valid if both the QR and sample solutions were recorded with the same acquisition parameters. To report the cannabinoid content in (mg/kg sample), the results obtained with PULCON equation were converted to (mg/kg sample) using a simplified conversion (see Equation (S3)): (1) volume contractions that may occur during mixing of sample solution are neglected, (2) the sample weight is set to 100 mg sample for all calculations, neglecting the actual uncertainties, (3) an experimental determined average CBD oil density of 0.89 g/mL is used, neglecting any influence of the actual density of the individual carrier oil.

At the end of each measurement series, a control sample (sample with cannabinoids contents previously determined) was also measured. The results of the sample should not deviate by more than two times the standard deviation from the mean value of the precision measurements. The concentration quantification (signal integration and calculation) of the processed spectra were performed using MatLab (version 2015b, The MathWorks, Natick, MA, USA). In addition, all obtained spectra and fit graphs were visually checked to control the correctness of the MatLab routine. Cannabinoid contents whose value are below the determined LOD (see Section 3.6.2) were automated assigned to zero in the processing routine. In order to avoid or diminish the influence of the sample matrix on individual signals, the signal areas of the signals CBD at 4.63 ppm (called CBD 3 in the further work), ∆⁹-THC (δ = 6.15 ppm) and ∆⁸-THC (δ = 6.12 ppm) were determined using a line fitting algorithm.