Continuous-Flow Synthesis of Δ9-Tetrahydrocannabinol and Δ8-Tetrahydrocannabinol from Cannabidiol

A challenging step in the preparation of tetrahydrocannabinol analogs is an acid-catalyzed intramolecular cyclization of the cannabidiol precursor. This step typically affords a mixture of products, which requires extensive purification to obtain any pure products. We report the development of two continuous-flow protocols for the preparation of (−)-trans-Δ9-tetrahydrocannabinol and (−)-trans-Δ8-tetrahydrocannabinol.


GC-FID Analysis
GC-FID analysis was performed on a Shimadzu GC FID 230 with a flame ionization detector, using an RTX-5MS column (30 m × 0.25 mm ID × 0.25 μm) and helium as carrier gas (40 cm/sec linear velocity).
The injector temperature was set to 280 °C. After 1 min at 50 °C, the temperature was increased by 25 °C/min to 300 °C and kept constant at 300 °C for 4 min. The detector gases used for flame ionization were hydrogen and synthetic air (5.0 quality). A representative GC chromatogram is provided below, see Section 1.5. Note: In the case of GC and GC/MS care should be taken to ensure that the analysis is representative of the reaction and that decomposition does not occur due to the analytical method.

GC-MS Analysis
In the study by Tsujikawa et al., they investigated the thermal decomposition of CBD to Δ 9 -THC by GCMS. 1 They identified that decomposition was not observed when using split mode, under any condition. We used split mode and we also confirmed that no decomposition occurs by injecting samples of standards for CBD, Δ 9 -THC and Δ 8 -THC under our GC-MS and GC-FID and we did not see any decomposition products.
THC and CBD exist as tetrahydrocannabinol acid (THCA) and cannabidiol acid (CBDA) in their natural form within the plant. The main advantage of HPLC is that it can identify the acidic components, THCA and CBDA, before conversion to their corresponding free forms of THC and CBD. 2 Figure S1. Representative GC-FID chromatogram of the acid-catalyzed cyclization of CBD (1), example 1: full chromatogram (above), zoomed-in (below).

Acid Screening
We investigated the acid-catalyzed cyclization of CBD (1) by studying the influence of different Lewis and Brønsted acids on the reaction performance (Scheme S1). A time profile was collected for each experiment. Initially we screened each acid reagent at −10 °C and then operate at higher temperatures when necessary to promote the reaction or increase the reaction rate.
Scheme S1. General reaction scheme for the screening of Brønsted and Lewis acids via batch profiling. The general conditions are stated within the scheme and the experimental procedure is described below.
An exemplary batch procedure is described below. If not otherwise noted, batch experiments were executed using this procedure.
Substrate 1 (0.157 g, 0.500 mmol) was dissolved in anhydrous dichloromethane (5 mL, 0.1 M). The reaction was maintained at the desired temperature and the solution was stirred (minimum 500 rpm) using a magnetic stirrer. Then the acid (1.2 eq) was added and the reaction was profiled over time.
Aliquots were taken from the reaction and added to a saturated solution of NaHCO3 to quench the reaction. Then an aliquot was taken for analysis: 25 μL reaction mixture was diluted in 975 μL dichloromethane and analyzed by GC analysis. Table S1. Selected results for the measured responses from the Lewis acid screening. Conditions reported in Scheme S1 were used unless otherwise stated.

Lewis Acids
a Determined by GC-FID peak area percent. Percent of product with respect to all peaks except the substrate. b Reaction was performed in PhMe as solvent. RT = room temperature. Figure S2. Reaction profiles and kinetic fitting for BF3 . OEt2 and In(OTf)3. See also Fig. 1 Table S2, entries 4-6). After the reaction, the microwave vial was cooled into an ice/NaCl bath, and a saturated solution of NaHCO3 was added to quench the reaction. An aliquot of the reaction was taken for analysis: 25 μL reaction mixture was diluted in 975 μL dichloromethane and analyzed by GC analysis. CBD (1) (0.157 g, 0.5 mmol) was dissolved in anhydrous dichloromethane (5 mL, 0.1 M). The reaction flask was immersed in an ice/NaCl bath to keep the temperature at −10 ° C. Subsequently, the supported acid was added in the desired amount (see below). The reaction was stirred (minimum 500 rpm) using a magnetic stirrer and controlled to the desired temperature. An aliquot of the reaction was taken for analysis: 25 μL reaction mixture was diluted in 975 μL dichloromethane and analyzed by GC analysis.  Figure S3. Reaction profiles and kinetic fitting for MK10 and Si-BF3. Figure S4. Reaction profiles and model fitting for PVP-BF3. S11  Table S5. Results for the measured responses for the batch optimization using AlCl3.

AlCl3 Further Batch Optimization
General conditions: All experiments were performed using 0.5 mmol of 1 in 5 mL of solvent. a Determined by GC-FID peak area percent. Percent of product with respect to all peaks except the substrate.

General Flow Configurations
For pumping feed solutions, syringe pumps (Syrris Asia) equipped with syringes appropriate for the desired flow rate were used. All of the pumps were used with check valves (Upchurch, CV-3321) and internal pressure sensors. The pressure limit of the pumps was set to 20 bar. The pumps would turn-off automatically for safety reasons above this pressure. The syringe pumps were calibrated by pumping for a specified time and checking the mass balance. All pumps were found to dose within ± 2%. Standard PFA tubing (0.8 mm or 1.6 mm i.d.), PTFE or PEEK fittings and T-pieces were used in the flow setups.
The reactor coil was cut to length depending on the desired volume required. The reactor coil was kept at the desired temperature (−20 to 40 °C) by placing it inside the heating fluid which was maintained at a constant temperature using a thermostat (Huber Ministat 230).
We selected five acids from the screening to continue our investigations in: TMSOTf, BF3·Et2O, AlCl3, PVP-BF3 and Montmorillonite K10. We investigated the influence of changing substrate concentration, acid equivalents, residence time and temperature.

Montmorillonite K10 / PVP-BF3
Montmorillonite K10 or PVP-BF3 was packed into a column (Omnifit) and then placed on the heater module (Syrris Asia) to perform the reaction at the desired temperature.

AlCl3
For pumping light suspensions, i.e., in the case of AlCl3, a peristaltic pump (Vapourtec SF-10) was used.
Above this pressure and the pump would turn-off automatically for safety reasons. S14  Table S8. Measured responses for a flow experiment using PVP-BF3.

PVP-BF3
Experiment was performed using 0.1 M solution of CBD (1). a Determined by GC-FID peak area percent. Figure S5.

AlCl3
Table S10 . Results for the measured responses for the flow optimization using AlCl3.

Long Run to Prepare Δ 9 -THC
We performed a long run for the preparation of Δ 9 -THC (2) using the optimized conditions shown in Table S10, entry 7, to assess the robustness of our protocol.

Before/after experiment:
Before running the reactions, the system was flushed with technical grade dichloromethane for 10 min.
After the experiments, the setup was rinsed with dichloromethane and then the system was stored under isopropanol.

Flow reaction protocol:
The system was operated for a total of 268 min (from start-up to collecting the final fraction), which corresponded to the processing 4.  (Table S11). 1,2-dichloro-4-nitrobenzene was used as internal standard to measure the NMR assay yield, which was determined to be 90% (value based on the average of three separate NMR samples). The collected fractionated vials were combined, filtered and then solvent was removed under reduced pressure (the rotary evaporator bath was maintained at 35 °C) to obtain 2 (4.07 g, 12.9 mmol, 97% yield) as a yellow oil (Fig. S6). The yellow oil was characterized by 1 H NMR, 13 C{ 1 H} NMR and GC-MS (Section 5). S19 Figure S6. Δ 9 -THC (2).

Preparative-scale Experiment to Prepare Δ 8 -THC
We performed an experiment for the preparation of Δ 8 -THC (3) using the optimized conditions shown in Table S6 Entry 14, to assess the robustness of our protocol.

Before/after experiment:
Before running the reactions, the system was flushed with technical grade dichloromethane for 10 min.
After the experiments, the setup was rinsed with dichloromethane and then the system was stored under isopropanol.

Feed preparation:
Feed solutions were prepared in volumetric flasks. Substrate feed preparation: Cannabidiol (1)

Flow reaction protocol:
The system was operated for a total of 24 min (from start-up to collecting the final fraction). The two feed solutions were introduced at the same flow rate using syringe pumps (Syrris Asia). The combined flow rate was 2.25 mL/min. A simple T-piece was used to mix the two feeds prior to the reactor. The reactor coil (4.5 mL internal volume) was submerged in the thermostat heating solution (EtOH) to control the temperature, which was set at 25°C. The reaction outlet was collected in 10 mL vials containing a quench (NaHCO3 in CH2Cl2) and a stirring bar, for the first 3 min (1 vial every 1 min), then in 10 mL vials for the central 16 min (1 vial every 2 min), and again in a 4 mL vial for the last three min. Fourteen fractions were collected and analyzed on the GC-FID (Table S12). 1,2-dichloro-4-nitrobenzene was used as internal standard to measure the NMR assay yield, which was determined to be 87% (value based on the average of three separate NMR samples). The collected fractionated vials were combined, filtered and then solvent was removed under reduced pressure (the rotary evaporator bath was maintained at 35 °C) to obtain 3 (1.48 g, 4.71 mmol, 98% yield) as a red oil (Fig. S10). The red oil was characterized by  Figure S9. Concentration of the different components over the duration of the experiment for the preparation of Δ 8 -THC (3). If a rate constant displayed no sensitivity to the fit then it was removed. The model structure which was fitted to the reaction profiles is shown in Scheme S2. The four rate limiting steps were simultaneously fitted for the reaction profile. The reactions were fitted as first order. In the case of the batch reaction with MK10 and Si-BF3 an additional rate-limiting step, k5, Δ 9 -THC (2) to CBN (S1) was also fitted (Fig.   S3).

Scheme S2.
Model structure for the fitting of the rate constants.