How two different LC–MS/MS methods with ESI and APCI were used for low-level analysis of 72 pesticides
Two different liquid chromatography–tandem mass spectrometry (LC–MS/MS) methods with electrospray ionization (ESI) and atmospheric pressure chemical ionization (APCI) were used for low-level analysis of 72 pesticides (including the very hydrophobic and chlorinated pesticides analyzed by gas chromatography [GC]–MS) in cannabis. The ability to screen and quantitate all 72 pesticides, including the very hydrophobic and chlorinated GC–MS amenable pesticides, in cannabis with only LC–MS/MS with ESI and APCI makes this approach a novel way of screening and quantitation of pesticides in cannabis and different matrices with a single instrument.
More than half of the United States has legalized the use of medical cannabis because of its therapeutic benefits for ailments such as cancer, multiple sclerosis, and amyotrophic lateral sclerosis (ALS or Lou Gehrig’s disease) (1–3). Like traditional agriculture crops, pesticides are sometimes used in cannabis cultivation to protect plants from pests and improve growth yield. Chronic exposure to pesticides can pose serious health risks; therefore, pesticide analysis in cannabis is an important consumer safety topic. Recent news has reported an alarming percentage of cannabis products to be tainted by high levels of pesticide residue, prompting recalls and public-safety alerts. Banned pesticides such as myclobutanil, imidacloprid, abamectin, etoxazole, and spiromesifen have been detected as residues on cannabis flowers and concentrated further in extracts and edibles. In Colorado, 20,000 packages of cannabis flowers in October 2015 were recalled because of pesticide contamination, and in November 2016, Oregon officials issued a health alert for specific batches of cannabis. Moreover, many of today’s cannabis products are inhaled after combustion, so there is growing concern among consumers and regulators because of the unknown effects of pesticide compounds when they are inhaled (4,5). The growing conditions for cannabis are also conducive to the growth of molds and fungi, which can produce carcinogenic mycotoxins including ochratoxin A and aflatoxins. As a result, testing for the levels of pesticide and mycotoxins in cannabis is important to ensure consumer safety and quality control.
High performance liquid chromatography–tandem mass spectrometry (HPLC–MS/MS) has emerged as the method of choice for pesticide and mycotoxin analysis because it offers superior selectivity, sensitivity, ruggedness, and does not require extensive sample preparation before analysis. Although gas chromatography–tandem mass spectrometry (GC–MS/MS) methods have been developed for pesticide analysis in cannabis samples, they are only applicable to a smaller subset number of analytes. Compounds such as daminozide, a highly polar compound, and abamectin, a high-molecular-weight compound, are not amenable to analysis by GC–MS/MS because they are heat labile and degrade in either the GC injection port or the column at high temperature. GC–MS/MS methods are not as robust as LC–MS/MS methods for pesticide analysis in complex matrices because they require extensive sample preparation to prevent GC injection port contamination from complex matrices (6,7).
Because there is no federal guidance for the analysis of pesticides in cannabis samples, different states in the United States have developed their own testing guidelines. Oregon was the first state to come up with comprehensive guidelines for pesticide residue analysis in cannabis (8) and set regulatory limits for 59 pesticides in cannabis. However, California has issued more stringent action limits for 66 pesticides (including all but one of those found on the Oregon state list, and eight more) and five mycotoxins residues in cannabis flower and edibles (9). Numerous reports for pesticide analysis in cannabis have been published but these studies have certain deficiencies (10–12). Most of these studies either do not achieve detection limits to meet the California state action limits or use time-consuming sample preparation methods (for example, QuEChERS [quick, easy, cheap, effective, rugged, and safe] with dispersive solid-phase extraction [dSPE]) with poor recoveries for some of the pesticides, which require use of both LC–MS/MS- and GC–MS/MS-based instruments for analysis of all the pesticides. This requirement increases cost, complexity, and turnaround time of analysis substantially. In this work, 66 pesticides (including very hydrophobic and chlorinated pesticides typically analyzed by GC–MS/MS) and five mycotoxins spiked in cannabis flower extracts were analyzed at levels well below the action limits specified by California. An LC–MS/MS instrument was used with electrospray ionization (ESI) and atmospheric pressure chemical ionization (APCI) sources and a simple solvent extraction method with excellent recoveries for all analytes in acceptable range of 70–120%.
Hardware and Software
Chromatographic separation was conducted on a PerkinElmer LC–MS/MS LX50 ultrahigh-pressure liquid chromatography (UHPLC) system, and detection was achieved using a PerkinElmer Q-Sight 220 MS/MS detector with a dual ionization ESI and APCI source; ESI and APCI operate independently with two separate inlets. Instrument control, data acquisition, and data processing were performed using Simplicity 3Q software (PerkinElmer).
Sample Preparation Method
Below is the step-by-step sample preparation procedure with 10-fold dilution:
âLC Method and MS Source Conditions
The LC method and MS source parameters are shown in Table I. (See upper right for Table I, click to enlarge. Table I: LC method and MS source conditions.)
Analytical Challenges for Testing Pesticide Residues in Cannabis Samples
Since the pesticides tested in this study included both polar and nonpolar compounds, 100% acetonitrile was used to extract all the analytes from the samples. Because of the cannabis matrix’s hydrophobicity, further dilution of the extract was performed with the aqueous mobile phase to make it compatible with the reversed phase column. This protocol resulted in lower recoveries of some of the pesticides because of precipitation. To achieve a higher performing method, cannabis extracts were diluted with acetonitrile by an overall factor of 10 to achieve high recovery of pesticides and reduce matrix effects. However, the reversed-phase LC method used aqueous mobile phase at the beginning of the LC run to help better retain the polar compounds on the column. Injecting an organic solvent, such as an acetonitrile sample, on the LC leads to poor chromatographic peaks for early eluted polar compounds. To overcome this problem, a small sample injection volume of 3 µL was used in this study.
Pesticide analysis in cannabis is very challenging since its matrix composition is very complex and contains compounds from different classes such as cannabinoids, terpenes, hydrocarbons, sugars, fatty acids, flavonoids, and others. Sample matrix effect remains the main concern for LC–MS/MS, and leads to variable signal ion suppression and matrix interference. Moreover, quantification of pesticide residues in cannabis is a difficult task because of the great disparity in high concentration levels of naturally occurring cannabinoids as well as high terpene content. In this work, we used a generic extraction method with dilution, selected the best multiple reaction monitoring (MRM) transitions and optimized the LC gradient to allow low-level analysis of pesticides with good recovery in a complex cannabis matrix.
Normally, the analysis of pesticides in cannabis and other food matrices is done by both GC–MS/MS and LC–MS/MS since some nonpolar and chlorinated pesticides are difficult to ionize with an electrospray ion source (13,14). To demonstrate the convenience of a single method, an LC–MS/MS method was developed using both APCI and ESI techniques to analyze all the pesticides on the California regulated pesticide list with the additional benefits of improved throughput, reduced complexity, and lower cost of analysis. Typically, the dirty matrix found with cannabis samples would quickly foul the conventional GC–MS/MS and LC–MS/MS systems and this contamination would increase the maintenance costs and downtime resulting in a loss of productivity. We showed that the LC–MS/MS method developed in this work would be more immune to contamination from the dirty cannabis matrix.
Detectability and Reproducibility
Figure 1 shows MRM chromatograms with excellent signal-to-noise ratios (S/N) for a representative set of pesticides spiked at a level of 0.01 µg/g in the cannabis flower. (See upper right for Figure 1, click to enlarge. Figure 1: MRM chromatogram of representative set of pesticides: (a) oxamyl, (b) metalaxyl, (c) fenpyroximate, (d) mycyclobutanil, (e) etofenprox, and (f) azoxystrobin spiked at a level of 0.01 µg/g in cannabis matrix.) The limits of quantification (LOQs) and response reproducibility at the LOQ level for each of the pesticides (category II and I) and mycotoxins in cannabis extract are summarized in Tables II–IV. (See upper right for Tables II-IV, click to enlarge. Table II: LOQs for California category II pesticides with LC–MS/MS in cannabis. Red/Green: pesticides typically analyzed by GC–MS/MS; red/black: pesticides analyzed by LC–MS/MS with ESI; green: pesticides analyzed LC–MS/MS with APCI. Table III: LOQs for California category II mycotoxins with LC–MS/MS in cannabis. Table IV: LOQs for California category I pesticides with LC–MS/MS in cannabis. Red/green: pesticides typically analyzed by GC–MS/MS; red/black: pesticides analyzed by LC–MS/MS with ESI; green: pesticides analyzed by LC–MS/MS with APCI.) The LOQs were determined by considering both the signals of the quantifier and qualifier ions (S/N > 10 for both) and ensuring that the product ion ratios were within the 20% tolerance windows of the expected ratio. As demonstrated in Tables II and III, the LOQs determined in this study are well below the California action limit by a factor of 2 to 600 for all category II pesticides and mycotoxins listed. The response relative standard deviation (RSD) for each pesticide and mycotoxin at its LOQ level in the cannabis matrix was less than 20%. The retention time for each analyte was reproducible within ±0.1 min over a 24-h period. This demonstrates that the method is more than adequately sensitive and reproducible for pesticides and mycotoxins analysis in cannabis at the regulatory limit specified by the state of California.
Sample Matrix-Matched Calibration Standards
Matrix-matched calibration is the preferred analytical procedure for quantitation because it compensates for matrix effects that are prevalent in cannabis samples. The decrease or increase in response is attributed to ion suppression of the analytes during ionization by the presence of coeluted matrix compounds. Because of sample matrix effects, a matrix matched calibration curve was used for quantitation and generated by injecting blank cannabis flower extracts and blank cannabis flower extract samples spiked with varying concentrations of pesticides and mycotoxins over a range of 0.1–1000 ng/mL. The calibration curves for all pesticides and mycotoxins were linear with a calibration fit R2 of greater than 0.99 for all the analytes.
Recovery Studies with Solvent Extraction
The QuEChERS extraction technique is a common approach for the extraction of low levels of contaminants such as pesticides from fruit and vegetable matrices with higher water content (15). The method includes extraction of a broad range of pesticides and removal of sugars, organic acids, and other compounds commonly found in fruits and vegetables (16–20). It is not a suitable method for very polar pesticides, such as daminozide, which are included in both the California and other states regulatory framework. Daminozide is too polar to be extracted efficiently with QuEChERS; it remains in the aqueous phase and does not partition into the organic solvent during the salting out step. The recovery of daminozide from a cannabis matrix with QuEChERS extraction has been reported to be less than 10% (10). Moreover, a typical cannabis matrix contains mostly hydrophobic compounds such as cannabinoids and terpenes, and therefore the QuEChERS extraction method does not remove the matrix interfering compounds during the salting out step. Different groups have tried to develop an advanced QuEChERS method with a dSPE step that utilizes primary secondary amine (PSA) and other adsorbents to remove matrix from cannabis extract. But the addition of the dSPE step to the QuEChERS method not only makes this method more laborious and expensive, but also leads to low recoveries of compounds such as spinosad, spirotetramat, spioroxamine, ochratoxin A, and a few others (11,12). These low recoveries are a result of these compounds binding to the PSA adsorbent in the dSPE step.
In light of the above mentioned shortcomings of the QuEChERS method for the extraction of pesticides from cannabis matrix, the application team used a simple acetonitrile-based solvent extraction method for extraction. To confirm this method, fortified cannabis flower samples were used to determine pesticides and mycotoxin recovery. The cannabis flower samples were tested to confirm the absence of pesticides before they were spiked. Five cannabis flower samples were spiked at two levels (low and high) of all pesticide (0.1 and 1 µg/g) and mycotoxin (0.02 and 0.1 µg/g) standards. These two levels were chosen based on regulatory limits for pesticides and mycotoxins in cannabis from California and other states. Tables V–VII show that absolute recoveries of 66 pesticides and five mycotoxins at two different levels were within the acceptable range of 70–120% with RSD less than 20% for five cannabis flower samples. (See upper right for Tables V-VII, click to enlarge. Table V: Recovery of category II pesticides at two different levels from cannabis with acetonitrile solvent extraction method. Table VI: Recovery of category II mycotoxins at two different levels from cannabis with acetonitrile solvent extraction method. Table VII: Recovery of category I pesticides at two different levels from cannabis with acetonitrile solvent extraction method.) For two pesticides, the recovery values were not reported at the low spiked value since it was below their LOQ value.
LC–MS/MS Method with Optimum MRM Transitions for Challenging Analytes in Cannabis Matrices
As stated, cannabis is a challenging matrix to test, and this is compounded by the low concentration level of the pesticides. To ensure the highest analytical confidence, multiple MRM transitions for a number of pesticides with minimal matrix interference in the cannabis matrix were determined for low-level detection. For example, acequinocyl is an insecticide and can be ionized easily as a protonated molecular ion in a standard, but the MRM transitions, based on protonated molecular ion in the cannabis matrix, showed a poor LOQ of 0.5–1 µg/g, about 5–10 times higher than its action limit for the state of California. Therefore, MRM transitions based on alternative modes of ionization, such as adduct formation, were determined to reduce matrix interference and achieve an LOQ of 0.025 µg/g (four times below action limits) for acequinocyl in the cannabis matrix. Figure 2 shows the signal overlay of blank cannabis matrix and acequinocyl spiked at level of 0.1 µg/g in cannabis with MRM transitions based on protonated molecular ion and adduct ion of acequinocyl. (See upper right for Figure 2, click to enlarge. Figure 2: (a) Overlay of response of cannabis matrix (red) and acequinocyl (green) spiked at 0.1 µg/g in cannabis matrix with MRM transition based on protonated molecular ion. (b) Overlay of response of cannabis matrix (red) and acequinocyl (green) spiked at 0.1 µg/g in cannabis matrix with MRM transition based on adduct ion.) This figure displays that optimum acequinocyl MRM transitions helped in achieving lower detection limits because of minimal matrix interference.
High molecular weight compounds such as abamectin, and some early eluting polar compounds, such as daminozide, are difficult to measure at low levels using GC–MS/MS since they decompose either in a high-temperature GC injector or a GC oven. Although high-molecular-weight compounds such as abamectin and polar compounds such as daminozide can be ionized with the ESI source, they are also prone to decomposition at high temperatures. Figure 3 shows abamectin response as a function of hot-surface induced desolvation (HSID) or heated interface temperature and source temperature. (See upper right for Figure 3, click to enlarge. Figure 3: Abamectin signal as a function of (a) ESI source and (b) HSID temperature.) Based on these results, the optimum temperature values for the ESI source and HSID temperature were set to maximize signals for high-molecular-weight and polar pesticides. Abamectin is also prone to sodium and potassium adduct formation from the sodium and potassium ions leached into mobile phase from glassware. Because it is difficult to control the amount of sodium and potassium ions leached from glassware, the use of the sodium adduct for abamectin as Q1 (parent ion) mass for analysis would lead to response variation. To reduce sodium or potassium adduct formation, a controlled amount of ammonium salt was added to the mobile phase. The combination of ammonium salt in the mobile phase and optimum temperature conditions resulted in good and reproducible signals for abamectin.
Analysis of Pesticides Typically Analyzed by GC–MS/MS by LC–MS/MS
A number of pesticides in cannabis, regulated by California and other states, are traditionally analyzed using GC–MS/MS with an electron ionization source since these pesticides have low proton affinity, which results in low ionization efficiency with the ESI source. Some examples of these pesticides analyzed normally analyzed with GC–MS are cypermethrin, cyfluthrin, captan, naled, permethrin, and pyrethrins. To achieve the required sensitivity, the selected MRMs were optimized with a heated electrospray source. LOQ for these analytes were in the range of 0.01–0.25 µg/g, which was well below the California action limits.
Analysis of Pyrethrins Isomers in Cannabis
The pyrethrins are a class of organic compounds normally derived from Chrysanthemum cinerariifolium that have potent insecticidal activity by targeting the nervous systems of insects. Pyrethrins are a group of six isomers and their structures are displayed in Figure 4. (See upper right for Figure 4, click to enlarge. Figure 4: Structure of six isomers of pyrethrins.) The naturally occurring pyrethrins, extracted from chrysanthemum flowers, are esters of chrysanthemic acid (pyrethrin I, cinerin I, and jasmolin I) and esters of pyrethric acid (pyrethrin II, cinerin II, and jasmolin II). In the United States, the pyrethrum extract is standardized as 45–55% w/w total pyrethrins and in a commercially available pyrethrin standard, the percentage of pyrethrin I, pyrethrin II, cinerin I, cinerin II, jasmolin I, and jasmolin II is about 56.1%, 27.8%, 5.7%, 3.8%, 4%, and 2.6%, respectively. A number of compounds in cannabis mimic the structure of pyrethrins, and therefore the analysis of pyrethrins in cannabis is very difficult because of matrix interference. The optimum MRM transitions and LC gradient were developed to analyze the six pyrethrins at low levels in the cannabis matrix with minimal matrix interference. The LOQs obtained with the LC–MS/MS method using optimum MRM transitions and LC gradient for six pyrethrins (pyrethrin I, pyrethrin II, cinerin I, cinerin II, jasmolin 1, and jasmolin II) were 0.1, 0.1, 0.01, 0.03, 0.025, and 0.01 µg/g, respectively, in cannabis flowers.
Pesticides That Don’t Ionize Effectively with ESI Analyzed with APCI
Hydrophobic and halogenated pesticides (for example, pentachloronitrobenzene and chlordane) are traditionally analyzed by GC–MS/MS since they do not ionize effectively by LC–MS/MS with an ESI source. For reference, the structure of the chlorinated pesticides is shown in Figure 5. (See upper right for Figure 5, click to enlarge. Figure 5: Structure of (a) pentachloronitrobenzene and (b) chlordane.) Since pentachloronitrobenzene (PCNB) does not contain either hydrogen atoms, for loss of protons, or functional groups with either high proton affinity or that can form ammonia or sodium adducts, it cannot be ionized with the ESI source. Similarly, chlordane is highly chlorinated and has very low proton affinity and is therefore difficult to ionize efficiently with an ESI source. Because an APCI ion source is better suited for ionization of very hydrophobic and nonpolar analytes, APCI was used to determine the detection limits of pentachloronitrobenzene and chlordane in cannabis. Also, the APCI ion source was used for low-level analysis of chlorfenapyr in cannabis, since limits of detection for chlorfenapyr were improved by a factor of two with the APCI source in comparison to ESI source because of less ion suppression. Figure 6 shows an excellent signal-to-noise ratio (S/N â©¾100) for PCNB spiked at level of 0.1 µg/g in the cannabis matrix using an LC–MS/MS system with an APCI source. (See upper right for Figure 6, click to enlarge. Figure 6: Sample chromatogram of pentachloronitrobenzene (PCNB) spiked at level of 0.1 µg/g in a cannabis matrix using LC–MS/MS system with APCI source.) Using a fast 6-min LC–MS/MS method with a short LC gradient and APCI source, the LOQs of pentachloronitrobenzene, chlordane, and chlorfenapyr in cannabis were 0.01, 0.05, and 0.05 µg/g, respectively.
Long-Term Stability Data with Self-Cleaning Source in LC–MS/MS
Long-term stability data for pesticide and mycotoxin analysis in cannabis samples was collected using an LC–MS/MS system, fitted with ESI and APCI sources, and combined with a heated and self-cleaning source with a laminar flow interface. Figure 7 shows long-term response and stability of the method for 100 ng/mL of diazinon spiked in cannabis extract over one week. (See upper right for Figure 7, click to enlarge. Figure 7: Long term stability data over one week of injections of diazinon at a level of 100 ng/mL spiked in cannabis flower matrix extract.) Long-term stability data for pesticide analysis in cannabis showed that response RSD over one week for most of the pesticides and mycotoxins was 1.5–20%. These results demonstrated that the heated self-cleaning source in the LC–MS/MS system would reduce maintenance needs that are usually prevalent with this matrix. Most of the LC–MS/MS methods published in the literature do not show long term stability data or state that they have to clean the electrospray source frequently to maintain the sensitivity of the mass spectrometer (21). Also, they divert the LC flow to waste for the first few minutes (and after the last peak is eluted) to reduce contamination from unretained and late eluted matrix compounds. In this study, excellent long-term stability data was obtained without diverting the LC flow from the MS in the first few minutes, at the end of run, and without periodical cleaning of ion sources.
This study demonstrates a unique, quantitative, rapid, and reliable LC–MS/MS method for analysis of different cannabis pesticide and mycotoxin residues in cannabis samples. The proposed solvent extraction method is suitable for laboratories wanting to comply with the state of California regulations, because the recovery of all pesticides and mycotoxins from a cannabis matrix was in the acceptable range of 70–120% with RSD less than 20%. This method allowed identification and quantification of 66 pesticides and five mycotoxins at low levels (0.005 to 0.25 µg/g), which is well below the actions limits set by the state of California with good precision. The ability to screen and quantitate all 66 pesticides, including the very hydrophobic and chlorinated compounds normally analyzed on a GC–MS/MS system, and the five mycotoxins, makes this method a novel way to screen and quantitate pesticides and mycotoxins in cannabis with a single instrument.
Avinash Dalmia and Jason P. Weisenseel are with PerkinElmer, Inc., in Shelton, Connecticut. Erasmus Cudjoe, Toby Astill, and Feng Qin are with PerkinElmer, Inc., in Woodbridge, Ontario, Canada. Jacob Jalali is with PerkinElmer, Inc., in San Jose, California. Molly Murphy is with SC Labs in Tigard, Oregon. Travis Ruthenberg is with SC Labs in Santa Cruz, California. Direct correspondence to: Toby.Astill@perkinelmer.com
A. Dalmia, E. Cudjoe, T. Astill, J. Jalali, J.P. Weisenseel, F. Qin, M. Murphy, and T. Ruthenberg, Cannabis Science and Technology1(3), 38-50 (2018).