Simultaneous identification and quantification of synthetic cannabinoids (cannabimimetics) in serum, hair, and urine by rapid and sensitive HPLC tandem mass spectrometry screenings: overview and experience from routine testing

Specimens

Serum, hair, and/or urine specimens were sent from other laboratories, hospitals, or withdrawal clinics. Control material was provided by volunteers, including the authors, from the laboratory.

Chemicals, analytical standards, and reagents

Standard substances used for complete method validation in serum and hair, that is, JWH-018, JWH-019, JWH-073, JWH-081, JWH-122, JWH-200, JWH-210, JWH-250, WIN48,098, WIN55,212-2, AM-694, and CP47,497 (Figure 1) were purchased from THC Pharm (Frankfurt, Germany). The deuterated internal standard JWH-018-d₁₁ was from Chiron AS (Trondheim, Norway), whereas THC-d₃ was from Radian International (Austin, TX, USA) and supplied by LGC Standards (Wesel, Germany). Further standard substances used for semi-quantitative reports due to incomplete validation, that is, UR-144, AM-1220, JWH-398, AB-001, AM-2201, RCS-04, RCS-08, JWH-307, MAM-2201, AM-1248, and AKB-48, as well as all hydroxylated urine metabolites (Table 1) were from Cayman Chemical Company (Ann Arbor, MI, USA) and purchased from LGC Standards. Acetic acid (anhydrous for analysis), methanol (for analysis), sodium carbonate, and sodium hydroxide were supplied by Merck (Darmstadt, Germany), ethyl acetate and n-hexane by Sigma-Aldrich (Taufkirchen, Germany), and LC-MS grade methanol by VWR International (Darmstadt, Germany).

Preparation of standard solutions and calibrators

As-received stock solutions of JWH-018, JWH-019, JWH-073, JWH-081, JWH-122, JWH-200, JWH-210, JWH-250, WIN48,098, WIN55,212-2, AM-649, and CP47,497 and further compounds (usually c=1 mg/mL methanol) were diluted in methanol (1:10). Then, 10 μL of each of these 12 solutions was further added to 1.88 mL methanol to give a standard working solution with a concentration of 500 ng/mL for each analyte. From this working solution (which was also produced if compounds were provided as dry substance), two more working solutions were prepared with final concentrations of 50 and 5 ng/mL, respectively. Similarly, JWH-018-d₁₁ and THC-d₃ were used as internal standards by further diluting the respective stock solutions to a final concentration of 100 ng/mL methanol.

Sample preparation: serum

For calibration, standard solutions (six points ranging from 50 pg to 2 ng absolute amount yielding final concentrations of 0.25, 0.5, 1.0, 2.5, 5.0, and 10 ng/mL, plus zero value) together with 20 μL of internal standard mixture were transferred into 2-mL polypropylene vials and gently evaporated to dryness. Analyte-free serum (200 μL) and 100 μL of saturated sodium bicarbonate solution were added. Similarly, 200 μL of unknown specimen plus 100 μL sodium bicarbonate were added to evaporated internal standard. Liquid-liquid extraction was then performed twice with 900 μL hexane/ethyl acetate (9+1 by volume) by rigorous shaking for 3 min. After centrifugation and evaporation of the organic phase, samples were reconstituted in 100 μL of methanol/ammonium acetate buffer (5 mmol/L, pH 3.9) 4+1 by volume and transferred to autosampler vials, from which 10 μL was injected into the LC-MS/MS.

Sample preparation: hair

Hair was washed by sonicating in acetone for 10 min, in distilled water for 5 min, again in acetone for 5 min, and finally dried at 50°C. All extracts were stored separately. Aliquots of fine-cut hair (∼50 mg) were weighed into 10-mL tubes. Known analyte-free hair was spiked with the respective standard solutions (six points ranging from 250 pg to 1.5 ng absolute amount yielding final concentrations of 5, 10, 15, 20, 25, and 30 pg/mg, plus zero value), and 10 μL of internal standard solution was added to all samples. Then, 1 mL sodium hydroxide (1 mol/L) was added, and the solution was heated under reflux at 96°C for 10 min. The fluid phase was extracted with 5 mL hexane/ethyl acetate (9+1 by volume) by mixing for 2 min. After centrifugation and evaporation of the organic phase under nitrogen, samples were reconstituted in 100 μL methanol/ammonium acetate buffer and transferred to autosampler vials, from which 10 μL was injected into the LC-MS/MS.

Sample preparation: urine

To 1.5 mL of urine, 50 μL β-glucuronidase (type H2 from Helix pomatia, Sigma-Aldrich) in phosphate buffer (pH 6.0) together with 25 ng of JWH-073 N-(3-hydroxybutyl)-d5 as the internal standard (Cayman Chemical Company, supplied by LGC Standards) were added. Hydrolysis of conjugates was achieved by incubation at 56°C for 1 h. After cooling, samples were adjusted to pH 9.0 and extracted with diethyl ether. The organic phase was transferred to a glass vial and processed as described above. All biological specimens were handled in a separate room with no contamination by stock standard solutions.

Apparatus and conditions

LC-MS/MS analyses were performed using an API 4000 tandem mass spectrometer (Applied Biosystems/Sciex, Darmstadt, Germany) with coupled turbo ion spray interface in both positive and negative MRM mode. The HPLC system consisted of a 1100 series binary pump (Agilent, Waldbronn, Germany) and a HTC-PAL autosampler (CTC-Analytics, Zwingen, Switzerland). Chromatographic separation was accomplished on an Onyx Monolithic C18, 50 mm×4.6 mm column (Phenomenex, Aschaffenburg, Germany). Data acquisition and evaluation were done using Analyst V.1.4 software (Applied Biosystems/Sciex).

Eluents were composed of methanol/1% acetic acid with a phase composition of 80+20 (by volume) for positive, and 90+10 (by volume) for negative MRM mode, respectively. General adjustments for MS were as follows: curtain gas, 30 units; nitrogen gas 1, 40 units; nitrogen gas 2, 50 units; and turbo spray temperature, 450°C. In positive and negative (for CP47,497 solely) MRM modes, CAD gas (nitrogen) was set to 4 and 6 units, ionization voltage was 4600 and –4500 V, entrance potential was 10 and –10 V, and dwell time was 20 and 50 ms, respectively. The optimized parameters for each (validated and non-validated) analyte, together with the MRM transitions chosen for the qualifier and quantifier transitions, are listed in Table 2. For routine analyses in serum and hair, triple injections were necessary: the first one for the 11 validated compounds in the positive MRM mode, the second one for the 11 non-validated compounds in the positive MRM mode (in order to achieve sufficient registration points per chromatographic signal), and the third one for CP-47,497 in the negative MRM mode. Overall run times are 5.0 min for serum and hair extracts (last eluting substance at 2.4 min) and 8.5 min for urine extracts (last eluting substance at 5.8 min).

Method validation

Chromatographic separation and proof of negative specimens

Separation of synthetic cannabinoids was performed by chromatography and mass transitions. When compared to data for serum (Figure 2) and hair samples (Figure 3) spiked with 11 compounds plus internal standard (respective panels A), peak areas in blank samples spiked with internal standard only (respective panels B) were all below the detection limit. Irrespective of the short retention times, an unequivocal assignment of signals was always possible (Figures 2 and 3, Table 2).

Figure 2

Extracted ion chromatograms of 23 positive MRM pairs (quantifier ions, qualifier ions, and internal standard JWH-018-d₁₁) in pool serum spiked with 11 synthetic cannabinoids, 5.0 ng/mL each (A), and in pool serum blank with internal standard JWH-018-d₁₁ (B).

For chromatographic conditions and structures of the synthetic cannabinoids, see text, Figure 1, and Table 2.

Figure 3

Extracted ion chromatograms of 23 positive MRM pairs (quantifier ions, qualifier ions, and internal standard JWH-018-d₁₁) in hair sample spiked with 11 synthetic cannabinoids, 15 pg/mg each (A), and in hair sample blank with internal standard JWH-018-d₁₁ (B).

For chromatographic conditions and structures of the synthetic cannabinoids, see text, Figure 1, and Table 2.

Thorough analyses were performed to exclude any possibility of false-positive results. Following conditions set forth by the German Society for Toxicological and Forensic Chemistry (GTFCh), especially spatial and personal separation of sample and standard handling, no positive results in control serum, hair, or urine specimens were detected. Several hair coloring/dying techniques, based on anamnestic exploration of volunteers, did not produce interfering signals. By nature, generation of false-negative results due to hair coloring cannot be excluded.

Methodological aspects: limits, imprecisions, linearities, and matrix effects for serum and hair matrix

LOD and LOQ values as calculated from signal-to-noise ratio and B.E.N. software are shown in Table 3. According to DIN32645, LOD and LOQ ranged from 0.018 to 0.192 ng/mL (0.084 if JWH-210 is considered as outlier) and from 0.062 to 0.708 ng/mL serum (0.304 if JWH-210 is considered as outlier). In tendency, LOD values calculated on the basis of the signal-to-noise ratio were slightly lower in most cases. In hair, variability between substances was similar with LOD and LOQ ranging from 0.140 to 0.820 pg/mg and from 0.760 to 2.870 pg/mg hair, respectively. With certain analytes, values for LOQ were lower than the actual values for LOD; in those cases, LOD and LOQ could be assigned to the same (higher) value. As a consequence, preliminary cut-off values (decision levels) were set at 0.1 ng/mL serum and 1 pg/mg hair. Results of the intra-day imprecision measurements (expressed as coefficient of variation=% relative standard deviation) are listed in Table 4. With the only exception of WIN48,098, all results met the acceptance criterion of <15%. Similarly, inter-assay imprecisions (CV) with serum ranged from 7.7% to 14.5% at 0.5 ng/mL, from 5.4% to 14.4% at 2.0 ng/mL, and from 5.7% to 14.6% at 10 ng/mL for tested cannabinoids, respectively, whereas inaccuracies varied from –1% to +24% (over all tested concentrations) with no systematic dependence on target concentrations, the only exception being again JWH-210 where inaccuracy ranged from 33% to 36%. At present, no explanation for this behavior is obvious; a possible contribution of interfering substances could be excluded by selectivity experiments. Values for hair matrix were in the same order of magnitude with inaccuracies between –27% and +3% (details not shown).

As compared to data from the literature reported to date, imprecision measures reported herein are grossly similar to those obtained by Teske et al. and Kacinko et al. [26, 28] at least for JWH-018. Using a similar methodological approach, LOD values reported by Neukamm et al. [29] were, on average, by a factor 3 lower for hair matrix but by a factor of 10 higher for serum matrix as related to the method described here. It appears that – without any intention – the method developed in our laboratory, using the API4000 MS/MS instrument, favors analytical sensitivity for the serum rather than for the hair matrix, which appears useful in view of the concentrations reached in real samples (see below). In this context, it is worth mentioning that the technical approach developed herein allows fast runs (retention times up to around 2 min for extracts of serum and hair, and up to 6 min for urine metabolites) on the API4000 system. Previously documented methods are working with retention times ranging from 3 to 8 min [26, 28, 29, 32].

Testing for linearity yielded correlation coefficients of r>0.99 for both concentration ranges I and II in both serum and hair matrix (details not shown).

The absolute matrix effect for serum at low concentration (2 ng/mL) is, at least with the present approach, characterized by moderate ion suppression with matrix effect ranging from 67.5% to 87.8% for all analytes. At higher concentration (10 ng/mL), absolute matrix effects increase slightly, ranging from 100.7% to 118.2% indicating a slight decrease of ion suppression. Whereas the ion suppression effect with the method described herein for JWH-018 in serum is significantly lower than reported previously [26], the analyte concentration-dependent change in matrix effect has also been observed in previous studies [26, 28], In hair matrix, however, only ion suppression was found. Even at a low concentration (5 pg/mg) absolute matrix values range from 15.4% to 71.5%. Whether this is indeed a systematic effect needs to be clarified; nevertheless, deviations between the five different specimens used for each matrix were characterized by a relative standard deviation of <16%. Recoveries (i.e., extraction efficiencies) ranged from 33.7% to 93.7% for serum and from 61.6% to 104.9% for hair (values combined from two concentrations each for both matrices).

Results from real serum, hair, and urine specimens

Within the time period of October 2011 to April 2013, a total of 359 routine serum samples were analyzed. Of these, only 31 tested positive (values above the decision limit defined above) with a total of 52 positive results for individual cannabimimetics. The continuous evaluation of serial analyses reveals that most of the serum samples tested positive contain only one of the compounds considered, but around one-third contain two or more cannabimimetics (Figure 4). This finding may either result from consumption of a single product containing different active ingredients or from consumption of different products within a very short period. With this or similar statistics it has, however, to be realized that the spectrum of compounds detected by analytical procedures becomes continuously broader possibly leading to a “dilution” of detection probabilities for individual compounds. By contrast, analytical progress, legislative initiatives, and press or internet releases may influence consumers’ behavior and substance availability on the market. Respective changes, such as the disappearance of JWH-210 in samples by May 2012, or the appearance and disappearance of AB-001 between May and August 2012, can be followed by serial follow-up of results with the limitation that client populations are not randomly distributed over time (details not shown).

Figure 4

Number of different cannabimimetics in serum samples tested positive in consecutive time order within the period of October 2011 to April 2013. Around 9% of all routine specimens received were positive.

In serum samples, 11 different cannabimimetics have been found to date (Figure 5). Concentrations span a very wide range up to around 40 ng/mL. This value represents a dynamic range of a factor 400 relative to the decision level. Although such high levels are rare, they should be expected and must be covered by the analytical approach chosen. Including all substances detected to date, the median concentration amounts to around 0.5 ng/mL; this value is by a factor 5 higher than the decision level and by a factor 75 lower than peak concentrations.

Figure 5

Concentrations of cannabimimetics in positive serum samples (same data pool as for Figure 4). Owing to the wide range, a logarithmic scale was applied for ordinate values.

In a parallel manner, a total of 84 hair samples have been tested between August 2011 and April 2013, of which 45 were positive with a total of 154 positive results for individual cannabimimetics. The finding that up to 13 different compounds can be detected in one individual sample attracts special attention (Figure 6). This observation allows the conclusion that consumers may change products rather frequently, either dependent on availability in headshops or electronic shops or “recommendations” in internet-based communities or personally by other users. From a clinical point of view, this finding draws attention to a high-risk behavior as far as consumers obviously change products rapidly and are presumably not aware of the highly different receptor affinities of the psychoactive components (although consumption behaviors would probably not change if consumers were aware of this risk). A limitation in the interpretation of positive hair samples results from the fact that, at least at present, only parent compounds are detected with the consequence that external airborne contamination cannot be excluded. This is in contrast to other drugs such as ethanol or cocaine, where a hepatic first-pass effect or a passage through the organism can be proven by the presence of the metabolites ethylglucuronide or benzoylecgonin (even more specific: norcocaine) in hair, respectively.

Figure 6

Number of different cannabimimetics in hair samples tested positive in consecutive time order within the period August 2011 to April 2013. Around 54% of all routine specimens received were positive.

Sixteen different cannabimimetics have been detected over time in hair specimens (Figure 7). Extremely high values are likewise rare, but it has to be realized that a dynamic range of a factor 3000 is obviously needed. Including all substances detected to date, the median concentration amounts to around 9 pg/mg; this value is by a factor 9 higher than the decision level and by a factor 300 lower than peak concentrations.

Figure 7

Concentrations of cannabimimetics in positive hair samples (same data pool as for Figure 6). Owing to the wide range, a logarithmic scale was applied for ordinate values.

With the exception of specimens sent for inter-laboratory comparison after positive immunoassay-based pretesting, routine urine samples received to date were negative with respect to the 14 compounds listed in Table 1 at a preliminary but not yet validated decision level of 0.5 ng/mL.

Metabolic aspects

Metabolism of cannabinoids is a matter of interest for at least two particular reasons. First, from a pharmacological point of view, knowledge of metabolic routes (qualitative aspect) and conversion rates (quantitative aspect) is essential in assessing duration and profile of a clinical effect. In contrast to general opinion, metabolites of CB receptor ligands are not necessarily inactive. Even in THC-COOH, one of the final end products of THC obtained after consecutive [41] cytochrome P450 (CYP)-catalyzed hydroxylations, analgesic properties remain whereas the psychoactive effect is lost [42]. The endocannabinoid, anandamide, is hydroxylated by CYP3A4 and CYP4F2 in human liver and brain. The CYP-catalyzed formation of epoyeicosatrienoic acid-ethanolamine from anandamide can be interpreted as a bioactivation mechanism or as a bioinactivation mechanism dependent on the specific target considered because this epoxide formation shifts receptor selectivity from CB1 to CB2 [8]. In parallel, there is increasing evidence that primary and secondary metabolites of synthetic cannabinoids express pharmacological activities. For instance, monohydroxylated metabolites of JWH-018 and JWH-073 bind agonistically to CB2 receptors but are more effective in G protein activation than the respective parent compounds [43], whereas certain monohydroxylated JWH-073 metabolites as ligands of the CB1 receptor exhibit neutral antagonist to partial agonist activity [44].

From an analytical point of view, metabolism of cannabimimetics is challenging because for one individual compound, there are multiple sites accessible for CYP-dependent hydroxylations. Owing to the structural similarity of many of these compounds as outlined in Figure 1 and Table 1, this has the consequence that it may become difficult or impossible to assign a mass fragment to a specific parent compound. For instance, members of the JWH-series can be hydroxylated at different positions of the aliphatic side chain, at the aromatic ring of the indol system, or at the naphthoyl ring system. Furthermore, N-dealkylation and hydroxyl group oxidation to carboxylic acid have been described at least for JWH-018 [45]. Similarly, eight possible hydroxylated and oxygenated CP-47,497 metabolites [46], and even 24 possible metabolites of HU-210 [47] have been identified. Those wide metabolite spectra may result from the combination of different metabolic reactions (hydroxylations, dealkylations) as shown for RCS-4 [36].

There are two general strategies to identify metabolic routes. The first one (which may be called the in vitro approach) is based on incubation of the substance in question with human liver microsomes [45–47] or specifically overexpressed CYP systems indicating that CYP2C9 and 1A2 predominantly catalyze JWH-018 and AM-2201 oxidations, whereas CYP 3A4, 2C19, and 2D6 are involved to a minor degree [48]. This approach leads to the identification of metabolites which can then be used as specific targets for urine analysis. The second strategy (which may be called the in vivo approach) is based on complete analysis of urine from customers using one defined substance by GC-MS (gas chromatography-mass spectrometry) or HPLC-MS/MS techniques. Again beginning with JWH-018 and JWH-073 [35], the technique has been further optimized to include metabolite spectra from seven or eight different parent synthetic cannabinoids [22, 34]. In turn, this approach may help to identify possible enzyme-catalyzed routes for cannabimimetic metabolism.

In conclusion, HPLC-MS/MS analysis, including the technique presented in this paper, of synthetic cannabinoids and their metabolites in different matrices is useful to further decipher biotransformation processes, clarify duration of drug action for prognostic purposes in practical medicine, and may eventually be applied to anti-doping programs as recently suggested [49].