GC-TOF Discovery-Based Profiling of CBD Oil Pet Supplements

December 18, 2019
Matthew Curtis

Sue D’Antonio

is an Application Scientists with Agilent Technologies in Wilmington, Delaware.

Anthony Macherone

is a Senior Scientist at Agilent Technologies and a Visiting Professor at The Johns Hopkins University School of Medicine in Baltimore, Maryland.

Cannabis Science and Technology, November/December 2019, Volume 2, Issue 6

In this study, six different brands of CBD oil pet supplements were obtained and untargeted analyses using gas phase-time of flight mass spectrometry were performed.

Cannabidiol (CBD) oils manufactured specifically for pets is a rapidly growing market. Many pet owners are supplementing their pets’ diet with products derived legally from hemp. In this study, we obtained five lots of six different brands of CBD oil pet supplements and performed untargeted analyses using gas phase-time of flight mass spectrometry. Chemical features of the raw data were extracted and annotated, and differential statistical analysis using commercially available bioinformatics software was used to examine each product.

With the recent phenomenon of legalization of cannabis, hemp, and cannabinoid products for both medicinal and adult recreational use in many U.S. states and Canada, many pet owners are feeding cannabidiol (CBD) oils derived from hemp to their pets. CBD is a nonpsychoactive, natural phytocannabinoid native to Cannabis spp. and is commercially available. A quick search of the web yields hundreds of products specifically made for pets. In this study, we used a nontargeted chemical analysis approach to evaluate the chemical profiles of commercially available CBD oils manufactured for pet consumption. We purchased five bottles of each product to investigate batch diversity and different lots of each product. It’s important to note that this workflow can be equally applied to plant material or other cannabinoid products for large scale cannabinomics, terpenomics, metabolomics, or other “big-data” studies.

Materials and Methods

Chemicals and Standards
Five bottles of six commercially available CBD oil pet supplements (n = 30) were purchased online. High purity dichloromethane (DCM) was used to solubilize and dilute the oil samples. A C5–C30 alkane mixture was used to define retention indices, and an isotopically labeled polyaromatic hydrocarbon (PAH) mixture containing naphthalene-d8, acenaphthalene-d10, chrysene-d12, and perylene-d12 was used as internal standards.

Instrument Hardware and Software
An Agilent 7890/7250 gas chromatography-quadrupole time-of-flight (GC-QTOF) system was used for all analyses. MassHunter software was used for acquisition and qualitative analysis. Mass Hunter Unknowns Analysis software was used for feature finding. Mass Profiler Professional software was used for differential statistical analysis. See Tables I and II for GC and QTOF parameters, respectively.

Sample Preparation
A 10 µL aliquot of each CBD oil sample (n = 30) was transferred to 5 mL volumetric flasks and 5 mL DCM was added (500-fold dilution). The flasks were capped and briefly vortexed. A 50 µL aliquot of each solubilized sample was added to 950 µL DCM (10,000-fold final dilution factor) in 2-mL autosampler vials and the isotopically labeled PAH mixture was added to each vial at 400 pg/µL. All 30 CBD oils were mixed equally to create a pooled quality control (QC) which was prepared as defined above. The alkane mixture was used as is. Pure DCM was used as solvent blanks.

Data Acquisition
A randomized batch sequence was created in MS-Excel. Five replicates of each CBD sample were collected, n = 180. Throughout the data collection sequence, DCM blanks were run 21 times, the alkane mixture was run four times, and the pooled QC was run four times. Data was collected in traditional 70 eV mode and in low energy mode at 14 eV.

Data AnalysisFeature Finding
To identify real spectral features in the raw data, Mass Hunter Unknowns Analysis software used the SureMass signal processing algorithm and Library Search to identify and annotate chemical entities in the raw data. Commercially available spectral libraries included a proprietary high-resolution accurate mass (HRAM) pesticides personal compound database-library (PCDL), and the unit mass Fiehn (1) and NIST14 (2) spectral libraries.

Differential Statistical Analysis
Mass Profiler Professional software (MPP) was used to determine relationships among the sample groups and variables. In this study, the sample group was named “Product” and the variables were Sample 1, Sample 2, Sample 3, Sample 4, Sample 5, Sample 6, Alkanes, Pooled, and DCM Blank. The statistical model was ANOVA.

Results and Discussion

Quality Control
The chemical composition of CBD oils extracted from hemp was quite complex. In the 70 eV experiments, 840 entities were annotated, and 74 entities were identified in the low energy (14 eV) experiments. We identified fatty acids, fatty acid esters, di- and triglycerides, terpenoids, cannabinoids, tocopherols, steroids and sterols, essential oils, fragrances, alkanes, alkenes, and alcohols in the CBD oil samples. In Figure 1 chromatograms of the DCM blank, the alkane mixture, the pooled reference QC, and an unknown CBD product are shown.

Figure 2 is a principle component analysis (PCA) that was performed as a quality control measure on all identified chemical entities, and demonstrated that each CBD oil sample was differentiated from one another and were unique compared to the pooled reference standard, DCM, and the alkane mixture. Figure 3 is a Box-Whisker plot illustrating the variance of the internal standards in each sample over the entire dataset.

Statistical Analysis
In the MPP environment, the chemical entities were filtered by those present or marginally present, then by the frequency that they appear across the entire dataset. This filtering removes chemical entities that have very low abundance and do not appear often in the data. This was followed by ANOVA with p < 0.05 (5% occurrence by chance) and a fold-change >2 (two-times more abundant than the median ion abundance over the entire data set). The result of these processes trimmed the number of chemical entities to 152 out of 843.

The purpose of PCA is to reduce a complex dataset to a smaller number of variables (principle components). It often provides visualization of the variation and interrelationships in a dataset. After performing the statistical analysis defined above, the three-dimensional (3D) PCA plot in Figure 4 demonstrated the uniqueness and similarities amongst the CBD oil products collected at 70 eV. Table III is a select list of the annotated compounds after statistical analysis. Interestingly, psychoactive ∆9-tetrahydrocannabinol (THC) was identified in all the samples. This is perhaps not unusual since hemp and products derived from hemp may legally contain up to 0.3% THC by weight. There is evidence of acid cataylzed conversion of CBD to THC (3), and in situ synthesis of THC from highly concentration CBD oil diluted in dichloromethane (unpublished data). However, in soon to be published work using liquid chromatography–mass spectrometry (LC–MS), these two products contain >0.3% THC by weight. Figure 5 is a normalized Box-Whisker plot of the THC ion abundance and a plot of the raw THC intensities.

Low Energy Mode
A series of low energy experiments were performed on a subset of the samples. In these experiments, all GC-QTOF parameters remained the same except for the electron voltage, emission current, and the source temperature which were lowered to 14 eV, 0.8 µA, and 200 °C, respectively. Low energy mode often increases the molecular ion response and improves elemental composition calculations. Figure 6 is the 70 eV and 14 eV chromatograms and spectra for an unknown compound. The 14 eV spectra better retained the molecular ion and therefore improved elemental analysis and compound identification from a known mass spectral library. The correct identification was hentriacontane (C31H64) with a mass accuracy of -0.27 ppm. Figure 7 is the HRAM extracted ion chromatograms (EIC) and mass spectra for CBD and cannabichromene (CBC). It should be noted that in the previously mentioned unpublished work, this sample contains 22% by weight total CBD and about 1% by weight CBC. In this qualitative study, the relative abundance of CBC is an order of magnitude lower than CBD which agrees with the quantitative work using LC–MS.


In this study, six commercially available CBD oil pet supplements were analyzed using GC-QTOF in discovery (untargeted) mode at both 70 eV and in low energy mode at 14 eV. The data was analyzed in a qualitative manner using chemical feature finding and differential statistical tools. The CBD oil products contained a multitude of chemotypes synthesized by the hemp plants. All products contained CBD in variable relative abundances. Compared to traditional analysis at 70 eV, low energy mode (14 eV) offers a second level of information from the samples. This workflow can be applied for multistrain analysis, cannabinomics, terpenomics, or plant metabolomics.


Agilent products and solutions are intended to be used for cannabis quality control and safety testing in laboratories where such use is permitted under state or country law.


  1. T. Kind, G.Wohlgemuth, D.Y. Lee, Y. Lu, M. Palazoglu, S. Shahbaz, and O. Fiehn, Anal. Chem.81(24),10038–10048. DOI: 10.1021/ac9019522 (2009).

  1. ©2014 copyright by the U.S. Secretary of Commerce on behalf of the United States of America.

  1. R. Andrews and S. Paterson, J. Analyt. Tox.36, 61–65 DOI:10.1093/jat/bkr017 (2012).

About the Authors

Matthew Curtis and Sue D’Antonio are with Agilent Technologies in Santa Clara, California. Anthony Macherone is with Agilent Technologies and The Johns Hopkins University School of Medicine in Baltimore, Maryland. Direct correspondence to: amacher1@jhmi.edu

How to Cite this Article

M. Curtis, S. D’Antonio, and A. Macherone, Cannabis Science and Technology2(6), 48-52 (2019).

download issueDownload Issue : November/December 2019