Background
Cannabis sativa (cannabis) is a recognisable medicinal plant containing cannabinoids which have therapeutic effects in the treatment of inflammation, psychiatric syndromes, epilepsy, muscle spasticity in multiple sclerosis, and other human health conditions. With the acceleration of cannabis liberalisation worldwide, the global cannabis market and demand for cannabinoid products have rapidly expanded in recent years.
A challenge for the cannabis industry is to develop a high throughput, efficient and sustainable extraction process to meet this growing demand. Cannabinoids are conventionally extracted from cannabis using maceration in which the plant materials are soaked in hexane, chloroform or other similar volatile organic compounds (VOCs) for cannabinoid extraction, and then the solvent removed by filtration and evaporation. This technology is easy in operation but is not environmentally friendly and has low selectivity [1]. Other cannabis extraction technologies using pure ethanol, supercritical CO2, deep eutectic solvents and ionic liquids have also been extensively investigated in recent years; however, high capital investment, viscosity issues and safety concerns (flammable solvents, high operating pressure and temperature) are some of the limitations restricting their applications in large-scale manufacturing [2-3]. In comparison to other technologies, liquid-liquid extraction (LLE) is an outstanding technology that can provide a low cost, easy operation, small footprint (e.g., solvent extraction column) and environmentally friendly (e.g., using bio-degradable solvents) approach for natural compound extraction. LLE has been widely applied in the poppy industry [4], which makes it a promising extraction technique for cannabis.
Another challenge in conducting cannabis/cannabinoid research and development activities are the regulatory barriers, and this challenge is even more significant when considering pilot-scale studies. Therefore, it is critical to identify compounds that can mimic the behaviors of cannabinoids in the extraction process.
Method
In this study, a list of mimic candidates for psychoactive neutral cannabinoids (e.g., tetrahydrocannabinol acid (THC), cannabidiol (CBD), and cannabinol (CBN)) was proposed. The mimic selection was based on an evaluation matrix that included the following selection criteria: market price, toxicity, physiochemical properties (solubility, pKa) and the partition coefficients of mimic candidates in different aqueous/organic liquid-liquid systems (Aqueous: ethanol/water, alkaline solution; VOCs: toluene, xylene, heptane; Green organic solvents: limonene, pinene, cymene).
The physiochemical properties and partition coefficients of cannabinoids and mimic candidates in different solvent systems were predicted by the COnductor-like Screening MOdel (COSMO) within the Amsterdam Modeling Suite program (developed by Software for Chemistry & Materials, Amsterdam, The Netherlands [5]). The leaching, extraction and stripping performance of the selected cannabinoid mimic using the above liquid-liquid systems was also determined experimentally. In addition, leaching studies were conducted on high-THC cannabis materials to validate the developed process.
Results and Conclusions
Based on the evaluation matrix, 4-tert-amylphenol (4-TAP) was selected as the mimic of choice for neutral cannabinoids. The preferred aqueous solutions for 4-TAP extraction were 40–60 v/v% ethanol/water, while higher ethanol content in aqueous phase were favored for 4-TAP stripping. Alkaline solution at pH > 10 could also be used for mimic extraction. For organic solvents, toluene and xylene showed the highest 4-TAP extractability compared to green solvents and heptane. However, by having high partition coefficients of the cannabinoid mimic (i.e., log P4-TAP Organic/Aqueous » 14–47 [6]), green solvents remain a competitive option to replace VOCs in cannabinoid extraction. Moreover, the trends predicted by the COSMO model and those observed in experimental studies on mimic and cannabis materials were comparable, indicating that the use of model and mimic is an effective approach in developing LLE process for cannabis and other medicinal plants.
Acknowledgements
The authors would like to acknowledge the funding provided by the Australian Research Council to the Industrial Transformation Research Hub for Medicinal Agriculture (ARC MedAg Hub, IH180100006), a Linkage Program grant (LP160101317), and the Australian Department of Education Regional Research Collaboration Program – Next Generation Protected Cropping in a Regional Manufacturing Facility. We would also like thank Cann Group Ltd. (Victoria) for providing the cannabis material for this project.