Determination of cannabinoids in Cannabis sativa L. samples for recreational, medical, and forensic purposes by reversed-phase liquid chromatography-ultraviolet detection

Currently, an increasing demand of cannabis-derived products for recreational and medical use is observed. Therefore, the reliable and fast quantification of cannabinoids in hemp samples is essential for the control of product from Cannabis sativa, L. strains. In general, gas chromatography (GC) is the method of choice for the quantification of cannabinoids whereas this method is time consuming and the detection of acidic precursor is not feasible without derivatization. We report the successful development and validation of an accurate and broadly applicable reversed-phase high-performance liquid chromatography (RP-HPLC) method coupled to an ultra violet (UV) detector including an optimized extraction procedure for the separation and quantification of eight different cannabinoids. The optimized method is able to separate cannabidivarin, cannabidiolic acid, cannabigerolic acid, cannabigerol, cannabidiol, cannabinol, Δ9-tetrahydrocannabinol, and tetrahydrocannabinolic acid within 10 min. For all target analytes, the %-Bias at the lower and upper calibration range varied from − 1.3 to 10.3% and from − 3.9 to 8.6%, respectively. The most suitable agent for extracting cannabis plant samples was evaluated to be a mixture of acetonitrile and water in a ratio 1:1. The extraction efficiency was more than 95% for all analytes in recreational hemp samples. Stability studies on acidic cannabinoids showed a high likeliness of decarboxylation at 100 °C and aromatization after exposure to UV light, respectively. A modified loss on drying method revealed a moisture content between 4 and 10%. The developed method was successfully applied to measure the cannabinoid content in recreational and forensic hemp samples representing broad range of cannabinoid amounts and patterns. The present work proposes validated methods for the determination of cannabinoids in cannabis samples. The use of RP-HPLC-UV renders this method broadly applicable and allows the detection of acidic precursor in even less time compared to GC-based methods.


Introduction
Since centuries, Cannabis sativa L. (C. sativa) is used for industrial purposes but it is better known as illegal drug possessing psychotropic properties. However, C. sativa is also a highly decorated medicinal plant for the use as anticancer agent, for neuroprotection and as bone marrow stimulants (Velasco et al. 2016;Machado Rocha et al. 2008). With the legalization of cannabis for therapeutic use, the demand for pure and characterized samples has grown significantly (Corroon and Phillips 2018). Therefore, currently new pharmacopeial monographs are in development for quality control of C. sativa-based medicinal products . Besides the medical use, there is an enormous interest from consumers/patients in the utilization of low Δ9-tetrahydrocannabinol (THC) hemp for recreational use. In recent years, a kind of gold-rush fever is observed in Europe and all over the world and many new suppliers entered the market (Pellechia 2018). Since there is a complicated and different legislation for C. sativa products all over Europe, caution for quality control has to be taken. Although there is no upper limit for the cannabidiol (CBD) or cannabidiolic acid (CBDA) content in most European countries, maximum limits of THCor Δ9-tetrahydrocannabinolic acid (THCA) contents vary between 0.1 and 1% within Europe.
Cannabinoids belong to terpenophenolic compounds and are the main constituents of the cannabis plant. Terpenoids and phenols were also identified in the cannabis plant but are of lower pharmacological importance . Cannabigerolic acid (CBGA) is the starting point in the biosynthetic pathway of all cannabinoids, which are synthesized in vivo in a carboxylated form (Fig. 1). In the plant, CBDA and THCA are synthesized by enzymatic catalyzed reactions. However, ex vivo stress conditions like heat and UV light decompose these precursors to their decarboxylated form: CBGA ➔ cannabigerol (CBG), CBDA➔ CBD and THCA ➔ THC, respectively (Citti et al. 2018a;Sirikantaramas and Taura 2017). Under UV light, Δ9-THC is further aromatized to cannabinol (CBN). THC and CBD are two main biomarkers in commercial available hemp samples. THC is mostly responsible for psychotropic activities whereas CBD is more anxiolytic and sleep inducing. In comparison to THC, CBD is not considered a controlled substance. Numerous reports have been published for the qualitative and quantitative analysis of cannabinoids in cannabis and its preparations. This study will therefore focus on those substances for possible therapeutic use such as CBD, Δ9-THC, CBG, CBN, cannabidivarin (CBDV), cannabichromene, and tetrahydrocannabivarin (Amada et al. 2013;Thomas et al. 2007). Several comprehensive reviews of the chemical analysis of cannabis plants, corresponding preparations, and forensic specimens were presented in the past (Citti et al. 2018b;Wang et al. 2017;Patel et al. 2017;ElSohly and Salem 2000). The most widespread techniques applied for separation were gas chromatography (GC) with and without derivatization, high-performance liquid chromatography (HPLC), and to a lesser extend supercritical fluid chromatography (Wang et al. 2016;U.N.O.o. Drugs, Crime 2013). The GC method is still officially employed by the authorities for the determination of cannabinoids. But it is obvious that acidic forms are not accessible without prior derivatization, and further conversion of THCA to THC is not quantitative at all (Dussy et al. 2005). Some researchers postulate that an accurate cannabinoid profile should be evaluated by determining the acid and neutral forms separately Citti et al. 2018b;Calvi et al. 2018;Ambach et al. 2014). On the other hand, LC-based procedures render the derivatization step superfluous and enable the detection of the heat-labile acid precursor in less time. However, determining chromatographic conditions is more challenging. Additionally, the pre-analytical phase has to be taken into account for method development and validation. Extraction, storage conditions, and stability determination play a pivotal role in the analysis of C. sativa L.-derived products (Dussy et al. 2005;Brighenti et al. 2017;Mudge et al. 2017).
The main scope of this study was the development and validation of a fast and convenient UV-detector-based RP-HPLC method for the fast quantification of cannabinoids in CBD samples and forensic cannabis samples. The present study examines further pre-analytical conditions and the analytical stability of cannabinoids under different stress conditions. Eight authentic CBD-hemp materials and 12 forensic cannabis samples offering a wide range of cannabinoid patterns were analyzed. Results of the overall THC-content of forensic samples were compared with gas chromatographic method (U.N.O.o. Drugs, Crime 2013), the formerly gold standard in cannabinoid analysis. Additionally, a modified loss on drying method was applied to determine the moisture content of all cannabis samples. Finally, the developed method was transferred easily to an ultra-high performance liquid chromatography (UHPLC) device using know metrics, thus further reducing analysis time from 10 to less than 5 min.

Materials and methods
Analytical standards were obtained from Lipomed (Reinach, Switzerlanf). Formic acid (FA), methanol (MeOH), ethanol (EtOH), and acetonitrile (ACN) were obtained from Merck (Darmstadt, Germany) and were of LCMS grade. Pure-water was generated from an in-house water purification system from Labtec (Villmergen, Switzerland). For all experiments, Gilson DIAMOND tips were used. Hop strobiles (Humulus lupulus L.) were obtained from local pharmacies. CBD-hemp tobacco samples were purchased from several licensed producers within Switzerland. The Forensic Institute Zurich (Zurich, Switzerland) provided 12 forensic cannabis samples.

Chromatographic analysis HPLC conditions
Reversed-phase chromatography was done using a LaChrom Elite System (Hitachi, Ltd., Tokio, Japan) HPLC system consisting of a LaChrom Elite L-2200 autosampler, a LaChrom Elite L-2130 pump, a LaChrom Elite L-2350 column oven, and a LaChrom Elite L-2420 UV-VIS detector. For peak integration, Agilent EZChrom Elite was used. The final liquid chromatography analysis was performed on a Phenomenex Kinetex XB-C18 column (150 × 4.6 mm, 2.6 μm) applying gradient elution, using pure-water (with 0.1% FA) and acetonitrile (with 0.1% FA) as the organic phase. The injection volume was 15 μL, and the dwell volume of the HPLC system was 1.8 mL. The column-oven temperature was set to 50°C, and the flow rate was 0.8 mL/min. Monitoring of all cannabinoids was done at λ = 220 nm.

UHPLC conditions
Reversed-phase chromatography was done using a HITACHI ChromasterUltra UHPLC system consisting of a 6270 autosampler, a 6310 column oven, a 6170 binary pump, and a 6430 Diode Array Detector. For peak integration, Agilent EZChrom Elite was used. The final liquid chromatography analysis was performed on a Phenomenex Kinetex XB-C18 column (150 × 2.1 mm, 1.7 μm) applying gradient elution, pure-water (with 0.1% formic acid), and acetonitrile (with 0.1% formic acid) as the organic phase. The injection volume was 5 μL, and the dwell volume of the UHPLC system was 0.7 mL. The column-oven temperature was set to 50°C, and the flow rate was 0.8 mL/min. Monitoring of all cannabinoids was done at λ = 220 nm.

Extraction
All preliminary extraction experiments were performed using sample (A). Twenty milligrams of sample was extracted with 2.5 mL solvent in a cooled ultrasonic bath. Afterwards, samples were centrifuged at 10°C for 15 min at 4000 rpm. Supernatant was filtered using a PFTE filter (0.45 μm, Machery Nagel) prior to analysis and tenfold diluted with solvent. Recovery effect (RE) was tested at QC low level using three independent spiked hop samples. Extraction efficiency (EE) was determined in triplicate extracting CBD and THC rich samples three times.

Validation
Commercially available 1 mg/mL methanolic solutions of all analytes were used as stock solutions for calibration and QC spiking solutions. Four different concentrations of the analytes in the range of 1 -100 μg/mL were chosen. Working solutions were prepared by serial dilution from each stock solution in methanol. QC low and QC high samples were analyzed in duplicate on each of 6 days. Accuracy was given in terms of bias as the percent deviation of the mean calculated concentration compared to the theoretical value. Intra-day and inter-day imprecision was calculated as relative standard deviation (RSD) according to Peters et al. (2009). Phenprocoumon was used as internal standard (IS) at a final concentration of 200 μg/mL.

Stability studies
Stability of cannabinoids was tested in an oven (VD20 Binder, Huber) at 100°C and under UV light (Honle, Sol 2, 350-700 nm). For the heat stability experiment, the sample was placed in weighing flasks. For the UV stability, one weighing flask was covered with aluminum foil and the other was exposed to UV light. For both stability experiments, 20 mg of sample was taken from each of the flasks at indicated time points and was analyzed. At indicated time points, 20 mg of sample was taken from each of the vials and analyzed. All stability experiments were performed in duplicate.

Quantification
Authentic samples were extracted and quantified applying the developed and validated method. All samples (20 mg) were analyzed in duplicate. The final amount of analyte [%] was calculated using the dilution factor given by the procedure and the weighed amount of plant sample. The determined concentration of the authentic forensic samples was compared to those obtained by established GC-FID-based method as described previously (U.N.O.o. Drugs, Crime 2013). (Details can be found in Additional file 1).

Loss on drying
Loss on drying experiment of hop was performed in an oven (VD20 Binder, Huber) 105°C for 2 h (Pharmacopoea europaea (Ph. Eur.) 2.2.32) (Ph. Eur., Loss on Drying (2.2.32) 2018). Cannabis samples were placed in weighing flasks and were dried to constant mass at 60°C.

Chromatographic analysis
Reversed phase chromatography RP-HPLC was chosen for the separation of eight cannabinoids. The focus was set on C18 columns, since these were the most commonly used in the literature. Several C18 columns with different eluent compositions, flow rates, and column temperatures were tested (Additional file 1: Table S1). A baseline separation of all analytes was finally achieved using the Kinetex XB-C18 HPLC column (2.6 μm, 150 × 4.6 mm,) with H 2 O/0.1% FA and ACN/0.1% FA as solvent. Flow rate and temperature was set to 0.8 mL/min and 50°C, respectively (Additional file 1: Table S2). Selected cannabinoids were separated within 10 min under HPLC conditions (Fig. 2). After cleaning and reequilibration, complete run time of this method was 20 min. The resolution of all analytes was at least R s > 1.7 and therefore in the acceptable range for quantification. The asymmetry factor of all peaks is between 1.2 and 1.5 (Table 1). Although for peaks 3 and 4, resolution and asymmetry factor were not in the optimal range, validation data in terms of bias and imprecision for CBGA(3) and CBG(4) were acceptable. The same column was used by Mudge et al. (2017)and De Backer et al. (2009) (Citti et al. 2018b) to separate the same number of cannabinoids, but with separation times of 14 min and 20 min, respectively. As internal standard (IS), phenprocoumon was used. Under selected chromatographic conditions, a clear separation between the IS and all cannabinoids was achieved. Finally, the developed HPLC method was transferred to an UHPLC system coupled to a diode array detector (DAD). The chemistry of the column (Kinetex C18, 1.7 μm, 150 × 2.1 mm) was similar to the HPLC column, and the same mobile phases were used. The injection volume was reduced to 5 μL. Target analytes were separated in less than 5 min (Additional file 1: Figure S5). The resolution of all peaks was above 1.5 and the asymmetry (10%) between 0.9 and 1.1.

Extraction
The most commonly used extracting agent for cannabinoids according to literature (Patel et al. 2017;U.N.O.o. Drugs, Crime 2013;De Backer et al. 2009;Zoller et al. 2000) is a mixture of methanol (MeOH) and chloroform (CHCl 3 ) in a ratio of 9:1. Considering the volatility and toxicity of chloroform and taking green chemistry guidelines into consideration, this halogenated solvent should not be the solvent of choice. Performance of other extracting solvents and solvent mixtures were compared to the standard MeOH/CHCl 3 (9:1) mix. One of these mixtures was MeOH/H 2 O (4:1) which is also used by Mudge and coworkers (Mudge et al. 2017). The other two were ACN/H 2 O in a ratio of 1:1 and ACN 99.9%. All extraction experiments were performed using CBD hemp sample (A). The mass to volume ratio was chosen as proposed previously (Mudge et al. 2017). As shown in Fig. 3a, ACN (Fig. 3b). Samples extracted without vortexing yielded the same amount of cannabinoids as the standard method. Vortexing alone or shortening ultra-sonication time to 5 min lowered the yield to around 50% for CBGA, THCA, CBGD, CBG, and THC and to about 70% for CBDA. These results clearly indicate the need for 15 min in the ultrasonic bath whereas the influence of vortexing was insignificant. Extraction efficiency (EE) was controlled by repeating extraction process three times on samples (F) and (P5) showing highest contents in CBD/CBDA and THC/THCA, respectively. In general, the EE shows the completeness of an extraction procedure of authentic samples. For CBD-rich cannabis samples with low amounts of THC/THCA, the EE for THC, CBDA, and CBD was greater than 94.4% after one extraction step. For all other analytes, an EE > 99% was observed in the first step. For THC-and THCA-rich samples, first extraction yielded in an EE of around 99% for THC/THCA whereas after the second extraction step, all target analytes were extracted completely (Fig. 4). Interestingly, in THC-/ THCA-rich samples, no CBD or CBDA was found. EE for THC and THCA was only around 90% for the first extraction step. However, after the second extraction, no more target analytes were found in the samples.

Method development and validation
Since there is no cannabinoid-free cannabis matrix, extracts from a closely related plant were used as surrogate matrix. The use of hop (Humulus lupulus L.) appeared to be most appropriate, because it belongs to the same family (Cannabacea) as cannabis (Nuutinen 2018). Therefore, it is assumed that the general and non-specific composition and structure of hop is similar to cannabis. Blank extract of hop samples were investigated in detail, but no co-eluting substances in relation to any target analytes of cannabis were detected. The recovery effect (RE) of an analyte is the ratio of the detector response obtained from an amount of the analyte added before and after the extraction process to blank matrix or its extract, respectively. To calculate the RE, blank material was spiked with reference standards and re-extracted as described above (N = 3). The extraction recovery effect was calculated using Eq.
1. Since there was a full recovery of all analytes observed (Table 2) and no co-eluting substance in the blank matrix was observed, a matrix-matched calibration was not necessary. As surrogate matrix for calibrants and quality control (QC) samples, the solvent mix ACN/H 2 0 (1:1) is used. Four different concentrations of the analytes were chosen for the calibrations: 1, 10, 50, and 100 μg/mL. The dilutions were made using ACN/H 2 O 1:1. For six following days, the analysis of four calibrants and two QC samples (QC Low and QC High ) were performed. QC samples were analyzed in duplicate. Validation data is shown in  Table 2. Bias for QC Low and QC High for all analytes varied between 1.3 and − 10.3% and 3.9 and 8.6%, respectively. Blank samples were injected after the highest calibrant, and carry-over was not observed. Across the calibration range, 1/x 2 was found to be the best fit for all cannabinoids, whereas a non-weighted calibration was used for THC ( Table 2). The decision on weighted calibration was made by comparing the deviations of the back-calculated concentrations from the respective nominal concentrations of the calibrators. Limit of quantification (LoQ) for all analytes was set to 1 μg/mL because of low bias and acceptable imprecision data. The limit of detection was not investigated systematically. Due to the early elution of the IS phenprocoumon, 60 different small molecules were screened for a second internal standard for late eluting peaks. Sixty different small molecules were screened for a second internal standard for use as a second internal standard for late eluting peaks. Under given chromatographic conditions, co-elution was observed with one of the target analytes. The list of selected compounds is given in the supporting information (Additional file 1: Table S3). Because of the acceptable bias and imprecision data, phenprocoumon was accepted as IS for all target analytes.
Stability CBD-hemp sample (A) was analyzed after indicated time points. All analyses were performed in duplicate. In Fig. 5a, a time-dependent decline of the concentrations is observed for all three acidic cannabinoids (CBDA, CBGA, and THCA) and at the same time the formation of their decarboxylated forms (CBD, CBG, and THC) is detected. The decarboxylation process of THCA occurred more rapidly than for the other two acidic precursors. The concentrations of CBD, CBG, and THC show a saturation at different time points (Fig. 5a). Afterwards, the curves start to decline, indicating further conversion of these compounds. For CBD and CBG, the decline of the analyzed concentrations appeared to start roughly after 30 h; however, for THC, at about 4 h. At the same time, an increase in CBN content could be observed. Starting from 0.14 μg/mL concentration of CBN increased to around 4.25 μg/mL after 72 h. Therefore CBN concentration may be useful as quality marker for storage conditions of cannabis samples and should be further take into account for estimation of initial THC content. For the UV stability experiment, sample (B) was placed in a weighing flask and exposed to UV light (λ = 350-750 nm, Fig. 5b). Since the temperature under the UV lamp can slightly increase (up to 36°C), a control, covered with aluminum foil, was also analyzed (Additional file 1: Figure S6A). The samples were quantified after indicated time points of exposure. Data in Fig. 5b showed a time-dependent decrease of all target analytes under UV-light exposure. A fivefold increase in CBN content could be detected, from approximately 0.2 to 1 μg/mL, after 8 h. Interestingly, CBGA and CBG were more susceptible to UV light compared to heat conditions. In the control, only a slight increase in CBN and decrease of THCA was observed.

Quantification
THC and THCA concentrations of the 12 forensic hemp samples were analyzed using the developed HPLC-UV method. Results were compared with results obtained by the forensic institute Zurich using standard GC-FID method. In general, correlation of both methods was high (R 2 = 0.956, Fig. 6). For all samples, the difference between the expected and the found THC+THCA content was between − 0.3 and 2.4% (Additional file 1: Table  S4). The slope of the regression line and y-intercept in both cases was 1.033 and 0.447, respectively. In general, THC/THCA were quantified higher by the HPLC-based method than the GC method. This could be explained by incomplete conversion of THCA to THC or conversion of THC to CBN at higher temperatures. Dussy et al. showed that at 150°C THC reaches its stability optimum for an accurate analysis, and above that temperature, the conversion to CBN takes place (Dussy et al. 2005). Interestingly, in forensic samples, only THC/THCA and CBD/CBDA were detected. Other target analytes were found only in traces. The cannabinoid content of eight different legally sold CBD-hemp tobacco samples obtained in Switzerland were analyzed. For only two out of eight samples (B, C), a certificate of analysis was available online. As indicated in the certificate, reported cannabinoid values were not batch specific. For nearly all samples, CBD content was far below the declared values. Only for sample (C), CBD content was higher than indicated. In general, it can be observed that the THC+THCA contents of all tobacco samples were under the maximal legally allowed limit of 1%. CBDV and CBN were below the detection limit. Nevertheless, our analysis also showed that samples (F) and (H) with a concentration of 0.91% and 0.90% respectively were very close to this threshold. Since CBD tobacco is sold in whole plant pieces, the content of the cannabinoids can vary depending on which part of the flower is ripped and ground and used for analysis, because different parts of the plant produce different amount of substances. To minimize such fluctuations, all batches should be homogenized before being analyzed and sold.

Loss on drying
The content of volatile compounds in crude drugs is an important factor since it influences the concentrations of the substances with pharmacological activity. Through manipulation, a lower content of target compounds could be reported. The current monograph on cannabis presented in the German Pharmacopoeia does not provide any information about loss on drying (BfArM 2017). The monograph of hop includes drying the sample at 105°C for 2 h. The loss on drying of hop was performed as described in the Ph. Eur. and additionally till constant mass was achieved after drying at 60°C. Both procedures resulted in loss of 3.8% (Table 3). Since the cannabinoids are heat instable, hemp samples were dried till constant mass at 60°C. The loss on drying of all tested CBD-hemp samples is shown in Table 3. The highest loss was determined in samples (E) and (G) and was higher than 8%. The lowest loss of 4.5% was determined in sample (B). In general, all hemp samples showed a higher loss on drying compared to hop samples.

Conclusion
We developed a fast and reproducible HPLC-UV method for the quantification of hemp samples. Method validation confirmed that the method produces repeatable and accurate results for eight different cannabinoids in less time. The use of a water to acetonitrile mixture (50%) for extraction instead of chlorinated organic solvent mixtures renders this method more ecologically friendly The presented procedure is universally applicable in a wide range of settings from pharmacopeial monographs, research, quality control, and regulatory evaluation of this emerging field of herbal industry. Additionally, the transfer to an UHPLC-DAD system reduced the analysis time to less than 5 min providing additional ecological benefits. An alternative loss on drying experiment was further described and showed similar results when applied on closely related hop samples. CBN content could be used as a marker for storage control of cannabis samples. The results obtained from the analysis of authentic samples highlight the need for accurate determination of the cannabinoids concentrations in regularly time intervals of different C. sativa L. strains to limit the risk of increased THC content in legalized products and to provide characterized products for therapeutic use.

Additional file
Additional file 1: Table S1. Use d RP-HPLC columns. Table S2. Gradient elution HPLC method. Table S3. Used compounds for use as internal standard. Table S4. Results obtained by GC-and LC-based methods. Figure S5. UHPLC-UV chromatogram of Cal 2 recorded at λ = 220 nm. Figure S6.