The chromatographic gradient was optimized to resolve all 30 mono- and sesquiterpenoids of interest using the selected ion monitoring (SIM) mode (Table S1; Figures 1 and 2). Using this thermal program, the stable working conditions for the system were optimized by establishing analyte recoveries and response repeatability. This was evaluated using a range of stock solvents, headspace carrier liquids, and vial incubation and loop/sample path temperatures (with 10–12 replicate injections under all conditions evaluated).

The first headspace liquid examined was brine, as this is an inexpensive and easily prepared solution that should increase the equilibrium vapor pressure of terpenes. Terpenoid standards were prepared in hexanes at 1 μg/mL and fortified to 5 ng/mL in 5 mL of brine. These samples were analyzed with a 5:1 split ratio and vial incubation and sample loop temperatures of 85 °C. The intention with this approach was to leverage the ability of hexane to quantitatively extract terpenes from plant materials with the reduction of coextraction interferences that headspace analysis would provide. Analyte responses showed low calculated relative standard deviations (RSD; ≤ 5%) for the low-boiling analytes (mostly nonoxygenated monoterpenes, compounds 1–13, Figure 1), and larger variations (up to 20% RSD) for the higher boiling components (oxygenated monoterpenoids and sesquiterpenoids, compounds 14–30, Figure 1). To improve method precision, different sample path temperatures (95–125 °C) were evaluated, as the higher sample path temperatures might prevent condensation of analytes. In addition, higher septum purge flows (10 and 20 mL/min) were also evaluated for possible flow dynamic effects within the high-temperature headspace sampler. Results showed that an offset of 10–20 °C between vial incubation and sample path temperatures and a purge flow of 20 mL/min improved the precision of instrumental replicates (n = 10). At 5 mL/min purge flow, a 100 °C sample path temperature produced all RSD values below 20%, with all but three below 10%. When higher septum purge flows were applied, all RSDs were less than 10% at 100 and 125 °C sample path temperatures (Table 1). Neither increasing septum purge flow beyond 20 mL/min nor changing needle purge flow levels improved the RSDs further. A substantial gap between vial incubation and sample path temperatures (85 and 190 °C) did not improve repeatability (Table 1).

Additional headspace liquids were explored, including dimethyl sulfoxide (DMSO), dimethylformamide (DMF), and glycerol. It was hypothesized that their higher boiling points compared to water might improve the equilibrium vapor pressure of terpenes with lower inherent volatility (i.e., higher boiling terpenes). Also, the ability to use high temperatures with glycerol would facilitate investigations into profiling terpenoids across a wider range of incubation temperatures (vide infra). Analysis of terpenes and terpenoids fortified in DMSO and DMF at various incubation temperatures up to 175 and 140 °C, respectively, revealed low signal-to-noise ratios for most analytes, and nondetects for many others (data not shown). For this reason, further evaluations with these solvents were not explored. Conversely, when glycerol was used, the most abundant raw responses for higher boiling terpenes were observed (relative to brine), likely due to the utilized vial incubation temperature of 180 °C and a sample path temperature of 200 °C. Of the headspace liquids evaluated, glycerol was unique in permitting the exploration of very high incubation temperatures due to its boiling point of 290 °C, versus 189 and 153 °C for DMSO and DMF, respectively. This facilitated the better transfer of analytes to the gas phase, and improved precision.

It was interesting to note that analyte recoveries for calibrators and spike-recovery samples using glycerol as the headspace solvent were dependent on the solvent used to prepare the terpenoid stock solutions. The low-boiling solvent hexanes (68 °C) and the high-boiling solvent tetradecane (254 °C) were used to investigate this effect. Analyses of samples fortified with terpenoids in the absence of plant material (i.e., a calibration sample) in hexane-based stocks showed that glycerol samples displayed higher absolute responses for low-boiling terpenoids (compounds 1–13, Figure 1), but lower absolute responses (as low as 10–20%) for high-boiling terpenoids (compounds 14–30, Figure 1) compared to brine samples. When the analogous tetradecane-based terpenoid stocks were used, most analytes displayed higher responses with glycerol compared to brine (Figure S1). This suggested that the solvents used to prepare stock solutions increased the partial pressure in the headspace such that the amount of analyte in the headspace decreased when the low-boiling hexane was used. It is postulated that the higher boiling point of tetradecane mitigated this effect. When the plant material was also present in the headspace vials, higher absolute responses relative to the data obtained in brine were observed for the majority of terpenoids when glycerol was used regardless of terpenoid stock solvents (hexanes or tetradecane). The reason for this effect is currently being explored.

Based on these investigations, terpenoid stock solutions for calibration samples and spike-recovery tests were prepared in tetradecane and glycerol was selected as the optimal headspace carrier liquid.

The developed method was evaluated for the dynamic range, repeatability (n = 5), intermediate precision (n = 17, over three days), carry-over, method detection limit (MDL), method reporting limit (MRL), and spike recovery. Calibration functions for all 30 terpenes and terpenoids were determined from 25 to 4000 ng/mL (equivalent to 2.5–400 ppm in plant material; dynamic range = 2.2). The split ratio can be adjusted to accommodate different expected terpenoid concentrations or to permit the evaluation of different quantities of plant material (i.e., not 50 mg). However, the issue of suitable sample quantity to overcome potential homogeneity issues needs to be considered; erring on the side of more material sampled for a given analysis is generally the prudent approach, with the authors suggesting 50 mg as a minimum quantity. Across a range of 25–4000 ng/mL, all calibration functions had correlation coefficients >0.99 and the accuracy of all calibration concentrations were ± 20% (data not shown). The calibration functions were consistent across three days of method validation and subsequent sample analyses. Additionally, all analytes presented with <0.1% carry-over when a method blank was analyzed immediately after a high concentration calibration sample.

Stinging nettle was selected as the matrix for method development, and validation as the dry tissues of this plant contain nominal incurred terpenoids, which made it an ideal blank matrix with which spike-recovery studies could be conducted (Figure S2). This was necessary for the method development reported herein, as the intended matrices for this method (i.e., cannabis and industrial hemp) have high concentrations of many terpenoids, such that performing spike-recovery studies would be problematic due to limitations in the dynamic range. Using dry, ground stinging nettle tissue, terpenoid spike recoveries of at least 70% were achieved, with 27 out of 30 terpenes and terpenoids reaching 80–108% at 10 ppm (Table 2). At the evaluated mid and high spike-recovery concentrations (125 and 400 ppm, respectively), the spike-recovery data ranged from 82 to 104% for all analytes, except for (−)-caryophyllene oxide, which has a high concentration spike recovery of 129% (Table 2). MDLs for all analytes were less than 4 ppm in the plant (40 ppm in the vial), and MRLs were verified at 10 ppm in the plant (100 pm in the vial) for 50 mg of plant material (Table 2). Lower MRLs could be achieved for many analytes, but given the expected concentrations in cannabis and industrial hemp, it was not necessary to fully optimize this statistically verified reporting limit. All of the method validation data reported herein demonstrate the suitability of the method for the intended goal of quantitating terpenoids.

As a relatively new market segment, there is currently a dearth of proficiency testing and/or certified reference materials for cannabis and industrial hemp. This is especially true for terpenoids that, unlike cannabinoids or pesticides, are not faced with any regulatory restrictions regarding their absolute concentrations. The absence of these materials makes it challenging to demonstrate the accuracy of an analytical method for a broad range of terpenoids—a new program exists in the United States for hemp flower, but the number of terpene values provided is limited to ten. To provide a better assessment of method performance for all terpenes and terpenoids evaluated herein, a cross-validation approach was utilized to verify method accuracy. Since the presence of incurred terpenoids was not a concern for the cross-validation, hops (Humulus lupulus L. flower), a close phylogenetic relative of the family Cannabaceae (i.e., cannabis), was used.

The method used for the cross-validation is an ISO/IEC 17025 accredited method for quantitating terpenes in plant material based on a hexane extraction, followed by direct analysis of the organic extract using GC–tandem mass spectrometry (method validation parameters were consistent with those reported herein for the high-temperature headspace method; Table S5). In addition to being accredited, this method has also passed several rounds of proficiency testing for terpenes in hemp oil. While not an exact matrix match, these results do provide strong support for the accuracy of the method as a benchmark. When comparing analytes above the respective MRLs for each method, a correlation matrix was obtained that demonstrated the accuracy of the high-temperature headspace method developed herein (Figure 3A).

The original focus of this method development project was on the quantitation of terpenes in plant material. Paralleling our method development efforts in plant material were reports coming out about the safety of e-cigarettes and related vaping products intended to be inhaled after high-temperature vaporization.^18,20,31 E-juices are generally formulated in a mixture of glycerol and propylene glycol, with other common ingredients, including nicotine, caffeine, and flavoring agents (i.e., terpenes).³² Given our finding that glycerol was a suitable matrix for the high-temperature headspace quantitation of terpenes, it was decided that the scope of this method should be extended to include the quantitation of terpenes in e-juices and related products. Understanding the terpene concentrations is likely to become a critical regulatory issue, as it has been documented that flavoring agents can decompose during consumer use of e-cigarettes to produce potentially harmful breakdown products.^18,33−35 Since nicotine is frequently included in e-juice formulations, it was also included in the expansion of the analytical scope of the method reported herein (Figure 4).

A mixture of glycerol and propylene glycol (1:1 ratio) was used as a surrogate for typical e-juice matrices to develop the method for terpene/terpenoid and nicotine quantitation. Using this mixture, terpenoid spike recoveries of at least 93% were obtained (Table 3). At the evaluated mid and high spike-recovery concentrations in e-juice (75 and 250 ppm for terpenoids, and 7.5 and 25 mg/mL for nicotine), the spike-recovery data ranged from 83 to 104% for all analytes. MDLs for all terpenoids and nicotine were less than 3 ppm for terpenoids and 0.1 mg/mL for nicotine, and MRLs were verified at 10 ppm for terpenoids and 0.5 mg/mL for nicotine (Table 3). These method performance parameters are all suitable for the intended scope, with typical nicotine concentrations in e-juices ranging mostly from 1 to 30 mg/mL.^36,37

Using this method, the terpenoids of interest were quantitated in commercially available terpenoid distillates and nicotine-containing e-juices. Owing, again, to the lack of suitable proficiency testing and/or certified reference materials, a cross-validation approach (using the same method as reported for cross-validation of terpene quantitation in plant material) was utilized to evaluate method accuracy for the terpenoid distillates. The data produced were in excellent agreement (Figure 3B), demonstrating the applicability of this method for the quantitation of terpenoids in e-juices and other glycerol-based vaporization products.

In addition to terpenoids, the method reported here is suitable for the analysis of nicotine (Figure 4, Table 3)—a bioactive compound for which concerns about the combined neurological impact with cannabis^38,39 has been raised. In 2018, the Canadian Task Force on Cannabis Legalization and Regulation also recommended that mixed products, including those containing both cannabis and nicotine and/or caffeine, should be prohibited.⁴⁰ In this context, this method could be used as a regulatory tool to screen for nicotine, while at the same time quantitating (or profiling) the terpenoids present.