Complete chromatographic separation was achieved for almost all 23 analytes. An oven temperature of 45 °C allowed for optimal separation of these selected analytes (Supplementary Fig. 3). The target scan time was set to 1 s and provided a sufficient number of data points for quantification. Increasing the flow rate did not significantly alter the resolution between peaks (Supplementary Fig. 4). Trans-citral (2) and carvacrol (3), as well as CBD (12) and CBG (11), overlapped; however, they were distinguishable by mass in MRM. Figure 1 provides a chromatographic overview of the total ion chromatogram (TIC). Varying intensities were observed for all terpenes and cannabinoids, although being present at the same concentration. Two diastereomers, namely menthone (4, 6) and citral (1, 3), were chromatographically separable. Diastereomer ratios can provide valuable information for authenticity and quality control of natural products due to stereoselective synthesis of terpenes in plants [24]. Figure 2 provides an overview of the extracted ion chromatogram (XIC) of all analytes. The chromatographic analysis included the acid forms of THC and CBD, as their total content is relevant to legal limits, and the current chemotype classification is based on their ratio. Cannabinoid acids can be detected using APCI and the same mass transitions as their neutral forms due to reproducible decarboxylation in the ionization source. Decarboxylation of the cannabinoid acids in the APCI source was further observed during compound tuning and has been described in the literature [30, 31]. In positive mode APCI ionization, cannabinoid acids undergo a partial and consistent decarboxylation, as evidenced by the lower peak areas compared to their neutral counterparts. Although APCI in negative mode offers improved ionization efficiency by generating [M-H]⁺ parent ions, the simultaneous analysis of both cannabinoids and terpenes requires a compromise [32]. Therefore, the use of neutral transitions for the cannabinoid acids in positive mode APCI allows adequate ionization and quantification of all analytes. Since the acid forms are primarily found in the plant but not in processed cannabis products such as resins and oils, the focus of this study was also on CBDV, CBN, and CBG. Nonetheless, to further prove the broad applicability and flexibility of the method, a randomly selected authentic sample was screened for CBDVA, CBNA, and CBGA using transitions from the neutral forms. All three analytes were found to be present and were baseline-separated from their neutral form (Supplementary Fig. 5). From the calibration, we were able to deduce that the areas for THCA and CBDA at the same concentration were, on average, 11 and 10 times lower than their neutral counterparts, respectively. These consistent ratios between the areas/concentrations of the acid and neutral forms could be used as a correction factor for quantification using the neutral cannabinoids as a calibration standard. However, these correction factors are dependent on ionization source and MS settings and require individual assessment. This approach would require further investigation, and could, however, be interesting from an economic point of view, as the neutral forms are more cost-effective.

In order to compensate for possible variations and measurement inaccuracies, we opted to select an internal standard (IS) as a correction factor. Cis-3-hexen-1-ol, propylparaben, and Δ⁹-THC-D3 were considered as potential IS candidates. In our in-house GC-FID method, cis-3-hexen-1-ol has successfully been used as an IS. However, during MRM tuning, it became apparent that very small and, therefore, unspecific fragments (< 50 Da) were selected for cis-3-hexen-1-ol. Thus, it was decided to continue with propylparaben and Δ⁹-THC-D3. Propylparaben proved suitable as an IS for cis- and trans-menthone, as well as citronellol. Δ⁹-THC-D3 served as an IS for all other analytes. In order to control for carry-over during method validation, a blank EtOH sample was injected after every five measurements. No carry-over was observed. Robustness was ensured by testing different oven temperatures and flow rates on critical pairs such as α-humulene and β-caryophyllene. Ultimately, two calibration ranges were chosen because the calibration models for the acids THCA and CBDA did not fully overlap in these ranges, and CBG showed saturation at higher concentrations. Additionally, a different weighting factor for CBD appeared better suited for both the high and low concentration ranges. Two dilution steps were necessary to cover both cannabinoids and terpenes fully. Major cannabinoids required a 100-fold dilution step as they oversaturated the detector when measured undiluted. In return, terpenes were only sufficiently detectable in undiluted samples. To compensate for this discrepancy, various injection volumes and dilutions of authentic samples were evaluated to ensure optimal coverage of all analytes. An injection volume of 5 μL provided symmetric peak shapes, and selecting two different sample preparations allowed an optimal coverage of all analytes. Unlike cannabinoids, terpenes did not ionize as efficiently, which increased their limit of detection. We further investigated whether it was possible to extend the minor calibration range to 60 μg/mL for CBD, CBDV, CBN, and Δ⁹-THC, thus eliminating the need for a second calibration. While feasible, this approach appeared impractical for our research question, as we aimed to retain the possibility of extending this method to resins or oils. In an industrial application, additional considerations would have to be made as to what type of cannabis products are to be analyzed and what the working calibration range is. Method validation took place on 3 separate days and resulted in a bias below 15% for all analytes in the minor and major calibration range. Inter- (RSD_R) and intraday (RSD_T) variability yielded similar results with, the exception of CBDA as a major component at the QC_low level, which slightly exceeded 15%. If a value did not deviate by more than 20%, which was the case for all observed analytes, it was considered acceptable. The validation results are detailed in Table 4 for the minor components and Table 5 for the major components. Further, calibration models and details can be viewed in the supplementary materials (Tables 2–3 and Figs. 1–2). Considering that terpenes are traditionally analyzed by GC, which has the advantage of high chromatographic resolution and separation performance, guaranteeing selectivity and specificity becomes even more important. Therefore, three criteria were selected to guarantee identification and quantification. To ensure analyte identity and, therefore, specificity, the relative retention time between the analyte and IS had to be within ± 0.02, both qualifier and quantifier ions had to be present, and the ion ratio had to closely match the expected ion ratio. Differences in the ion ratio < 20% were considered acceptable, < 40% marginal, and anything above 40% unacceptable. During validation all analytes met these criteria. Due to the minimal mass differences and structural similarities among terpenes, the resulting fragments alone were unreliable for identification. To exclude further interferences with other cannabinoids and their acids, a mixture consisting of 10 µg/mL THCV, THCVA, CBG, CBGA, CBDV, CBDVA, CBC, CBCA, CBNA, and Δ⁸-THC in acetonitrile was analyzed using the same method. No interferences were observed, and all cannabinoids, including their acids, were baseline separated (Supplementary Figs. 6 and 7). If required, the method presented here can be expanded to contain these 8 cannabinoids including their acids, without the need to acquire additional mass transitions. This further emphasizes the broad applicability of the method. To assess potential matrix effects such as ion enhancement and suppression, ethanolic H. lupulus extracts were analyzed blank and spiked at three different concentrations in the high, medium, and low range of the minor calibration, and the RE was determined. As the H. lupulus extract is an analyte-free matrix for cannabinoids, it was possible to observe matrix effects in great detail. The results presented in Table 4 were determined for the minor components only, as the matrix is diluted 100-fold with EtOH for the major component determination. It was therefore assumed to have a minimal effect. The RE was in an acceptable range between 80 and 120% for all analytes. Minimal to no matrix effects were observed, which is consistent with the observation, that APCI is less affected by matrix effects compared to electrospray ionization (ESI) [33].

LC is not the method of choice for the analysis of volatile compounds such as terpenes. This becomes even clearer when considering the structural similarities of terpenes and their widespread occurrence. Cannabis, in particular, is a multi-component mixture consisting of numerous cannabinoids, terpenes, and flavonoids. This special feature makes cannabis an interesting medicinal plant but also makes its analytical analysis a challenge. This multitude of chemical diverse classes often requires a combination of different analytical techniques to gain a complete understanding of the composition of cannabis, which can be both time-consuming and expensive [23, 34–36]. A single method that is able to capture as much compositional variability with minimum sample preparation in flowers of C. sativa would be desirable. Our study showed that terpenes can be efficiently ionized by APCI and that volatility is not a deal-breaker for the analysis with LC. Furthermore, our chromatographic method was able to separate 16 terpenes and at least 15 cannabinoids within 25 min, which is significantly shorter than most GC methods. In addition, the derivatization reactions, a process that can have side reactions and be incomplete, can be omitted. To test the reliability of the method, we compared our LC-APCI-MS/MS method with a state-of-the-art GC-FID method by randomly selecting and analyzing 15 authentic samples as a subset from the 55 authentic samples for their terpene composition. The same analytes were quantified in the samples by GC-FID, and the results of the two methods were compared using Bland–Altman plots. For reliable results, we only selected analytes that could be quantified with both methods above the LoQ in at least 10 of the samples. Samples in which the analytes were detected but were not within the calibration range were excluded from the method comparison. The calibration models, weights, and range for the analysis using GC-FID can be studied in the supplementary material Table 5. These criteria were only fulfilled by α- and β-pinene, myrcene, linalool, limonene, α-humulene, and β-caryophyllene, terpenes which are known to be major and ubiquitous components in cannabis. With the exception of limonene, all analytes were in good agreement between both methods. Figure 3 depicts the results from the method comparison using a Bland–Altman analysis. Results were considered acceptable if the percentage difference between the two methods was not greater or smaller than 20%.

Limonene concentrations determined by LC-APCI-MS/MS differed by more than double from those determined by GC-FID. However, the limonene ratio between samples appeared consistent in both the GC-FID and LC-APCI-MS/MS method. As it was initially thought that another substance was co-eluting with limonene, samples were also investigated by GC-FID. Especially α- and γ-terpinene were suspected as co-eluting analytes since both analytes are isomeric forms of limonene, sharing identical masses and fragments. However, neither were detected in any of the samples, and co-elution of these two structural isomers of limonene could be excluded. In general, the concentrations determined by LC were slightly higher than those measured by GC, which may be related to the detection method and split-injection mode. Ultimately, the co-elution of isomeric analytes is far more likely using LC than GC due to lower resolution power. For the other terpenes studied, this seemed to not be a problem. In particular, co-elutions of terpenes occurring at extremely low concentrations do not appear to have an influence on the quantification of major terpenes. This method allows faster and more reliable detection of the main constituents of C. sativa than traditional methods. It also has the advantage that the acid forms of the cannabinoids can be analyzed without derivatization. For the observation of extreme minor terpenes, it is still recommended to rely on the high separation power of GC. Since a minimum concentration of 0.05% v/w is required for terpenes to have a pharmacological effect, it is actually sufficient to consider and quantify only the major terpenes when cannabis is used for medicinal purposes [37]. We propose that the analysis of minor terpenes can be of particular interest when, for example, the origin or environment of the plant is of interest.