In this work, several steps were developed to improve the workflow in the analysis of pesticide residues from cannabis plant material by liquid chromatography tandem mass spectrometry (LC–MS/MS) and gas chromatography (GC)–MS/MS. The goals of the study were to optimize the extraction and analysis process to reduce sample background, improve pesticide recoveries, and allow direct injection of QuEChERS (quick, easy, cheap, effective, rugged, and safe) extracts, without dilution, for high performance liquid chromatography (HPLC) analysis. Reduced background and better pesticide recoveries were achieved by the use of a QuEChERS cleanup with zirconia/C18 functionalized silica, primary secondary amine (PSA), and a low-surface-area carbon blend. HPLC analysis was optimized to provide late elution of coextracted cannabinoids relative to the targeted pesticides, thus making it possible for a dump step to be used to prevent them from entering the MS detector. The HPLC column configuration used improved the peak shapes for the earliest eluted hydrophilic pesticides when injecting extracts in 100% acetonitrile. This then allowed a single set of extracts to be analyzed by both GC–MS/MS and LC–MS/MS.
Consumption of cannabis or cannabis-based products is currently legal in some form in 29 states in the United States plus the District of Columbia. Testing of the plant materials and products is required by many of these states; however, the specific test methods and target compound lists are not mandated in all cases. In October 2016, the state of Oregon took a major step forward by requiring that all laboratories testing cannabis be accredited by the Oregon Environmental Laboratory Accreditation Program (ORLEP) and licensed by the Oregon Liquor Control Commission (OLCC) (1). Consequently, Oregon Administrative Rules (OAR) list specific contaminants to be tested in marijuana samples, along with action levels (2). The pesticides on this list include carbamate, organophosphorus, macrocyclic lactone, neonicotinoid, pyrethroid, and triazole fungicides as well as others. Action levels per OAR vary from 0.2 to 1 µg/g, depending on the specific pesticide. In addition, the state of California, which legalized recreational cannabis in 2016, recently released a proposed list of pesticides that includes all but one of those found on the OAR list, plus eight more (3).
Because of its ease of use and applicability to a wide range of pesticides, the QuEChERS (quick, easy, cheap, effective, rugged, and safe) approach has been adopted by many testing laboratories for use on cannabis. After extraction, incorporation of a cleanup step is important for removing pigments as well as other contaminants. A QuEChERS cleanup using a mixture of primary secondary amine (PSA), C18, and graphitized carbon black (GCB) is often chosen for this purpose. PSA removes acidic interferences, C18 removes hydrophobic interferences, and GCB retains some pigments-specifically the green color imparted by chlorophyll. This mixture of sorbents thus retains a wide range of contaminants; however, it also has potential to reduce recoveries of target pesticides that are susceptible to hydrophobic retention on C18, or planar enough in structure to be strongly retained by GCB. In previous work by Stenerson and colleagues with cannabis, an alternative sorbent mix was evaluated for cleanup in the analysis of various pesticides, and found to offer a better balance than a traditional mixture of PSA/C18/GCB with regards to removal of pigmentation and pesticide recovery (4). This sorbent mix contained PSA, a zirconia-coated C18 functionalized silica, and a specialized carbon. The zirconia retains by Lewis acid-base interactions, and has been found to reduce the background of certain fatty compounds as well as some pigments. The carbon is a low surface area graphitized variety that has been shown to have weaker retention of small, planar molecules such as certain pesticides. In this work, the pesticide list tested previously was expanded to include many of those on the OAR list described above. The alternative sorbent mix was compared directly to a conventional blend of PSA/C18/GCB for cleanup and analysis of spiked replicates of cannabis plant material analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) and gas chromatography-tandem mass spectrometry (GC-MS/MS). For the LC-MS/MS portion of the analysis, the column and mobile phase selection was optimized to improve overall workflow, which is described here.
Dried cannabis was pulverized using an IKA T10 Ultra Turrax mixer. Next, 1.9 g was weighed into a 50-mL centrifuge tube and spiked with pesticides at 50 ng/g. After a 10-min equilibration time, the sample was mixed with 10 mL of deionized water and allowed to sit for 30 min. Then 10 mL of acetonitrile was added, and the sample was shaken at 2500 rpm for 30 min. A mixture of 4 g of magnesium sulfate, 1 g of sodium chloride, 0.5 g of sodium citrate dibasic sesquihydrate, and 1 g of sodium citrate tribasic dihydrate was added, and the sample was shaken for 1 min. The sample was then centrifuged at 5000 rpm for 5 min, and the supernatant was removed for cleanup.
Next, 1 mL of extract was added to a 2-mL tube containing the mixture of cleanup sorbents. Two different sorbent mixtures were used:
Samples were shaken for 1 min, and then centrifuged at 5000 rpm for 3 min. The supernatant was removed for analysis.
A single set of extracts was analyzed directly after cleanup, and without dilution, on both an Agilent 6460 LC-MS/MS system and an Agilent 7000C GC-MS/MS system. Pesticides that did not yield response by LC-MS/MS were analyzed by GC-MS/MS. The HPLC column used was a 10 cm x 2.1 mm, 3.0-µm Ascentis RP-Amide column, equipped with a 3 cm x 2.1 mm, 5-µm RP-Amide guard column. A gradient of 5 mM ammonium formate, 0.1% formic acid in 5:95 water-acetonitrile and 5 mM ammonium formate, 0.1% formic acid in 95:5 water-acetonitrile was used at a flow rate of 0.4 mL/min. The gradient started at 10% organic, held for 1 min at 10%, ramped to 100% in 13 min, and finally held at 100% for 6 min. The sample injection volume was 5 µL, and the column temperature was maintained at 30 °C.
GC analysis was performed on a 20 m x 0.18 mm, 0.18-µm film thickness SLB-5ms column, with an oven program starting at 50 °C, with a 2-min hold, and ramping at 8 °C/min to 325 °C with a 10-min hold. Helium carrier gas was used at a constant flow rate of 1.2 mL/min. The injection port temperature was 250 °C, and a 1-µL injection was performed in pulsed splitless mode with a 50-psi pulse held until 0.75 min. The liner used was a 4-mm i.d. tapered FocusLiner containing quartz wool.
Quantitation was done against a five-point matrix-matched calibration curve prepared in unspiked cannabis extract. Separate curves were prepared for each cleanup. No internal standards were used, thus all recovery values reported are absolute.
HPLC Column and Mobile-Phase Selection
Typical cannabis samples analyzed by testing laboratories contain high levels of cannabinoids, often in the range of 20-25% by weight. These compounds will coextract with the pesticides during the QuEChERS process, and are only partially removed during cleanup. In the case of LC-MS/MS analysis, these remaining cannabinoids can build up on the detector, resulting in the need for more frequent system maintenance. In this study, column and mobile-phase selection were based on conditions that would force elution of the cannabinoids as late as possible in the run, ideally after the pesticides. Under these conditions, the diverter valve on the LC-MS/MS system could be set to flow to waste after elution of the last pesticide. This setting will then prevent a majority of the cannabinoids from entering the detector.
To facilitate the appropriate high performance liquid chromatography (HPLC) conditions, a screening experiment was designed to study elution of the major cannabinoids relative to the targeted pesticides on several different column chemistries, and using both acetonitrile and methanol based gradients. The column chemistries screened were as follows:
All columns were 10 cm x 2.1 mm columns with 2.7-µm fused-core particle packings. The fused-core particle architecture version of these chemistries was initially chosen for both efficiency and speed. The HPLC conditions were similar to those listed previously in the experimental section, with ultraviolet (UV) used for detection, and ammonium formate omitted from the mobile phase. Samples were injected in 100% acetonitrile to emulate samples resulting from the QuEChERS extraction. As expected, on all five columns this injection solvent resulted in poor peak shapes of the earliest eluted pesticides. Focusing on overall retention profiles, the RP-Amide column yielded the least amount of overlap between the pesticide and cannabinoid elution ranges. Overlap was notably less than obtained using the C18 column (Figure 1), which is a commonly used chemistry for this application. (See upper right for Figure 1, click to enlarge.) In addition, comparing acetonitrile to methanol in the mobile phase, using the former in the gradient eluted the pesticides faster, further decreasing overlap with the cannabinoids.
To simplify the method as much as possible, the same QuEChERS extract was analyzed by both HPLC and GC. However, as indicated previously in the column screening experiment, injection of 100% acetonitrile into the high aqueous starting conditions of the gradient produced poor peak shapes for the early eluted pesticides. To improve the peak shapes of these compounds, a 3-µm fully porous RP-Amide column was substituted for the 2.7-µm fused-core RP-Amide column. The installation of a guard column further improved peak shape (Figure 2), most likely because of increased retention and improved mixing of the sample with the mobile phase. (See upper right for Figure 2, click to enlarge.) In addition, when working with high background samples, the use of a guard column is highly recommended to extend the life of the analytical column.
As expected, the coextracted chlorophyll generated a QuEChERS extract with a deep green color. After cleanup, a majority of the green color was removed by both cleanups, with the final extracts appearing as a light yellow. Analysis by GC-MS in full scan mode (not shown) produced similar peak patterns between both cleanups and the uncleaned extract, but a difference in amplitude of background peaks (which consisted mainly of terpenes and cannabinoids). Overall reduction in background was compared by summation of total peak area for each chromatographic analysis. Compared to no cleanup, the alternative sorbent mix reduced background by 35% and PSA/C18/GCB by 31%.
Elution of Cannabinoids
Using the optimized HPLC conditions described previously for the final LC-MS/MS analysis of the cannabis extracts, minimal overlap was observed between two of the major cannabinoids present in the samples, and the later eluted pesticides. Figure 3 shows an extracted ion chromatogram (EIC) of m/z 314.5, taken from a full scan LC-MS analysis of a cannabis extract. (See upper right for Figure 3, click to enlarge.) This mass represents the molecular ion of the two major cannabinoids detected in the sample extract; tetrahydrocannabinol (THC) and cannabidiol (CBD). As indicated, the last pesticide analyzed, pyridaben, was eluted just before cannabidiol. The most abundant cannabinoid present, THC, was eluted well after. Column flow could be switched to waste after elution of pyridaben, preventing some of the CBD and all of the THC from entering the MS system. Other cannabinoids; specifically cannabigerol (CBG), cannabinol (CBN), cannabidiolic acid (CBDA), cannabichromene (CBC), cannabigerolic acid (CBGA), and tetrahydrocannabinolic acid (THCAA), are known to be eluted after CBD on the RP-Amide phase. Thus, if present in a cannabis sample, all of these could also be diverted to waste as well.
The pesticides included in this study represented a majority of those on the OAR list. Two pesticides from this list, avermectin B1a and naled were not analyzed because of lack of response. Avermectin is prone to sodium and potassium adduct formation. The presence of ammonium formate in the mobile phase should reduce this occurrence (because it is monitored as an ammonium adduct). However, even with these measures, others have also reported issues with low-level detection of this compound (5,6). Naled is susceptible to adsorption by PSA, and thus did not make it through the cleanup process with either sorbent mix.
Comparing spike data between the two cleanup methods, the alternative mix exhibited better overall performance than PSA/C18/GCB. Looking at data for the individual pesticides (Table I), bifenthrin, chlorantraniliprole, clofentezin, fenproximate, fludioxinil, and hexythiazox showed notably better recoveries using the alternative sorbent mix. (See upper right for Table I, click to enlarge.) Although none of these are completely planar in structure, it is possible that recovery was reduced using PSA/C18/GCB because of the hydrophobic retention on the GCB phase, which has a higher surface area than the carbon in the alternative mix.
Several pesticides exhibited poor recoveries after both cleanup techniques. Bifenazate recovery was around 50%. This compound is susceptible to oxidation to bifenazate-diazine, which may have occurred to some degree during the extraction and cleanup process (7). Daminozide is a carboxylic acid, and thus its recovery was affected in both cleanups by the presence of PSA. Matrix interference prevented the analysis of dichlorvos in the PSA/C18/GCB extracts. In the extracts cleaned with the alternative mix, the peak could be detected, but recoveries were low and variable. This low recovery is most likely because of the retention on the zirconia-based sorbent as a result of Lewis acid-base interaction between the zirconia and the phosphate group present in the structure of dichlorvos. This same behavior has been observed by the author and others for this compound when using zirconia sorbents (8-10). Etofenprox is a very hydrophobic pesticide (log p = 7.1) and may exhibit poor extraction efficiency or retention by the C18 and zirconia portions of the cleanup sorbents (although less so on the later). Imazalil is a relatively polar pesticide, which can be retained by PSA. Recovery issues have also been observed by others with this compound when using zirconia containing sorbent mixtures (10,11). Low recoveries of spinosyn A and D, which are large macrocyclic lactones, have been reported when using C18, carbon, and zirconia-containing sorbents (12). The use of citrate buffering in the QuEChERS extraction has been observed to increase recovery, possibly by displacement of the analytes from the zirconia (10). Low spiroxamine recovery could not be attributed solely to cleanup, and may be because of an issue with extraction efficiency.
To improve workflow in the analysis of a highly varied list of pesticides in cannabis, such as that required by Oregon, several recommendations can be made.
The authors would like to thank Dr. Hari H. Singh, the Program Director at the Chemistry & Physiological Systems Research Brand of the National Institute on Drug Abuse at the National Institute of Health for supplying the cannabis sample.
Katherine K. Stenerson and Gary Oden are with MilliporeSigma in Bellefonte, Pennsylvania. Direct correspondence to: Katherine.firstname.lastname@example.org
K. Stenerson and G. Oden, Cannabis Science and Technology1(1), 48-53 (2018).