From these initial observations we constructed a focused library with all but one of the constructs in the second library exhibiting activity at least 100-fold higher than WT NphB in an endpoint assay . The best two constructs, M23 and M31, exhibited dramatically improved activity and specificity. Both had kcat values 1000-fold higher than WT NphB and both produce only the correct prenylated isomer, CBGA. As shown in Figure 2-3C, WT NphB produces CBGA, but the dominant product is a prenylated side-product, 2-O-geranyl olivetolate, whereas M23 makes CBGA almost exclusively. Overall, the designed enzyme is a much more active and specific CBGA synthase than WT NphB, and is easier to work with than the natural cannabis prenyltransferase, which is an integral membrane protein. Our soluble, CBGA synthase could potentially be applied in both cell-free and in vivo systems to improve cannabinoid production. Production of cannabinoid precursors With our designed CBGA synthase in hand , we tested the ability to produce CBGA directly from glucose and OA using the full synthetic biochemistry system, including the PDH bypass . The initial productivity of the system using M23 was 67 mg L-1 hr-1 with a final titer of 744 ± 34 mg L-1 CBGA—100-fold faster and 21-fold higher titer than CBGA production using WT NphB . We noted that with the mutant NphB enzyme, maximum titers were reached within 24 hours, after which production spontaneously stopped. In contrast, the system with the wild-type enzyme ran continuously for up to 4 days, suggesting enzymes and cofactors remain active and viable for longer periods of time, air racking consistent with prior work. So what is stopping the reaction at the higher titers? We observed that reactions turned cloudy once ~500mg L-1 CBGA was produced.
We collected the precipitate and identified a mix of enzymes in the precipitate by SDS-PAGE analysis , indicating high-levels of CBGA in solution may be causing enzymes to precipitate. We therefore sought to continually remove the product in situ during the reaction . Initially a fixed volume nonane overlay was used for each reaction to extract CBGA. Unfortunately, CBGA is more soluble in water than nonane, limiting the amount of CBGA that can be extracted with a simple overlay. We therefore designed a flow system that would capture CBGA from the nonane layer and trap it in a separate buffered reservoir . By implementing this flow system we hoped to maintain a lower concentration of CBGA in the reaction vessel to mitigate enzyme precipitation. The flow system indeed improved the final titers to 1.25 ± 0.07 g/L, however enzyme precipitation still occurred at about 24 hours. We next evaluated the system flexibility by replacing OA with divarinic acid to produce the precursor of many rare cannabinoids, CBGVA. We first tested whether our designed enzymes would be active with DA as the substrate. Kinetic analysis indicated that M31 effectively prenylates DA, with catalytic efficiencies 15-fold higher than M23 and 650-fold higher than WT NphB. We therefore utilized M31 to produce CBGVA from glucose and DA. As shown in Figure 2-4A, CBGVA was produced at a maximum productivity of ~107 mg L-1 hr-1 , and reached a final titer of 1.74 ± 0.09 g L-1 , converting 92% of the added DA to CBGVA. The nonane flow system was not needed for the production of CBGVA because CBGVA was less potent in precipitating enzymes. To illustrate the production of other cannabinoids from the central cannabinoids CBGA and CBGVA, we employed CBDA synthase to convert CBGA into CBDA and CBGVA into CBDVA.
Conversion of CBGA into CBDA has been demonstrated by several groups. In our case, we simply transferred the nonane overlay containing CBGA to an aqueous solution containing CBDA synthase, and indeed we were able to convert CBGA into CBDA at a constant rate of 14.4 ± 0.8 mg L-1 hr -1 mg total protein-1 over the course of 4 days converting 25% of the CBGA added to CBDA . To our knowledge it is not known whether CBGVA can be converted into the rare cannabinoid CBDVA using the CBDA synthase. So, we added CBGVA, extracted from the cell-free system, to a reaction containing CBDA synthase. CBDVA was produced by CBDA synthase at a rate of 7.1 ± 0.1 mg L-1 hr -1 mg total protein-1 for 24 hours. We note that the cannabinoid acids can undergo spontaneous decarboxylation or heat induced decarboxylation to ultimately form additional bioactive cannabinoids cannabidiol and cannabidivarin . Thus, our system provides opportunities for ultimately producing a wide-variety of cannabinoids.Our results demonstrate the power and flexibility of a cell-free approach, not only for the production of pure, therapeutically relevant cannabinoids and other prenylated natural products, but for bio-derived chemicals in general. Freedom from worries about cell viability allowed us to focus on pathway optimization rather than minimizing GPP toxicity, while the lack of a cell membrane barrier freed us to design a system with added aromatic molecules, which would not be possible in cells. Moreover, we could flexibly change the input from OA to DA to target rare cannabinoids without redesigning an entire pathway. Finally, it was straightforward to identify and focus our efforts on fixing the bottleneck steps. When we started this project we were only able to produce 9 mg/L of CBGA using the monoterpene pathway developed by Korman et al.
By introducing the PDH bypass and optimizing for co-factors, enzymes and environmental factors we were able to increase those titers to 132 mg/L. To improve titers further we engineered the NphB prenyltransferase, which further increased titers to 600 mg/L of CBGA. The final bottleneck was enzyme stability in the presence of CBGA, so by limiting the CBGA in the reaction vessel, we increased the titer to 1.25 g/L of CBGA, nearly a 140-fold improvement. Solutions were quickly implemented due to speedy design-build-test cycles, rapidly yielding results that far exceed published results using living cells. Like all new technology, the current system will need additional technical developments to become commercially viable, but our results suggest that synthetic biochemistry can become a realistic option for producing bio-based chemicals.The PyOx/PTA reactions were assembled in two pieces. First the co-factors and substrates were combined in one tube, and the enzymes were combined in another. The amount of enzyme added to each reaction is detailed in Table 2-1. The co-factors and enzymes were mixed to initiate the reaction, and a 500 µL nonane overlay was added to the top. The reactions were incubated at room temperature shaking gently on a gel shaker. When the aromatic substrate was the varied component 0.5 to 5 mM of the aromatic substrate was added to the reaction, drying weed and the reactions were quenched at 24 hours. When time was the varied component, 5 mM of 1,6 DHN was added, and separate reactions were quenched at ~12, 24, 48 and 72 hours. Conditions for the olivetolate and divarinic acid reactions were altered slightly. Optimization of the cannabinoid pathway showed that the same titers could be achieved with less glucose, so we reduced the glucose concentration to 150 mM . Additionally, increasing the NADP+ concentration to 6 mM and decreasing the ATP concentration to 1 mM led to higher titers of CBGA. The olivetolate concentration was set at 5 mM. The amount of NphB added to the reaction was variable. The data shown in Figure 2-2C utilized 1.5 mg/mL NphB, and the reactions were quenched at ~4, 8, 14, 24, 48, 72 and 96 hours. The data shown in Figure 2-4A was achieved with 0.5 mg/mL of WT NphB and M23 and M31 , and reactions were quenched at ~ 6, 9, 12, 24, 48, 72 and 96 hours. The conditions were identical to the method above with the following exceptions, the final concentration of the aromatic substrates was 1 mM and the initial glucose concentration was 150 mM. Additionally, the final concentration of the prenyl-transferase was 1 mg/mL, and we tested AtaPT, NovQ and NphB with apigenin, daidzein, genistein, naringenin and resveratrol. We also tested NphB with olivetol, olivetolate and 1,6 DHN. The reactions were quenched at 24 hours. To quench the reactions, the aqueous and organic layer were transferred to a 1.5 mL microcentrifuge tube. The reaction vial was washed with 200 µL of ethyl acetate, which was then pooled with the reaction in the microcentrifuge tube. The samples were vortexed for 5-10 seconds and then centrifuged for 3 minutes at 16,060 x g. The organic layer was removed, and the remaining aqueous layer was extracted 2 additional times with 200 µL of ethyl acetate. For each sample the organic extract was pooled, and then evaporated using a vacuum centrifuge. The samples were re-dissolved in methanol for HPLC analysis. For olivetolate / CBGA Due to the observed protein precipitation, the CBGA reactions shown in Figure 2-4A were extracted in the presence of 0.12 g of urea , to facilitate the extraction of CBGA. This was unnecessary for the WT NphB CBGA data in Figure 2-2C because the proteins did not precipitate. The reactions were fractionated by reverse phase chromatography on a C18 column using a Thermo Ultimate 3000 HPLC. The column compartment temperature was set to 40 ºC, and the flow rate was 1 mL/min.
The compounds were separated using a gradient elution with water + 0.1% TFA and acetonitrile + 0.1 % TFA as the mobile phase. Solvent B was held at 20% for the first minute. Then solvent B was increased to 95% B over 4 minutes, and 95% B was then held for 3 minutes. The column was then re-equilibrated to 20% B for three minutes, for a total run time of 11 minutes. The cannabinoids were quantified using an external calibration curve derived from an analytical standard purchased from Sigma Aldrich. The 5-p-1,6-DHN and CBGVA NMR samples were used to generate an external calibration curve because authentic standards were not available . A known concentration of the standard was dissolved in water, and then extracted using the method detailed above. Due to the lack of authentic standards for the prenyl-products prenyl-apigenin, prenyldaidzein, prenyl-naringenin, prenyl-genistein, prenyl-resveratrol and prenyl-olivetol, we quantified the prenyl-products based on substrate consumption. To generate a standard curve, serial dilutions of each aromatic substrate were subjected to the reaction mix, but to prevent product formation the prenyl-transferase was left out. ESI-TOF measurements were carried out on a Waters LCT-Premier XE Time of Flight Instrument controlled by MassLynx 4.1 software . The instrument was equipped with the Multi Mode Ionization source operated in the electrospray mode. A solution of Leucine Enkephalin was used in the Lock-Spray to obtain accurate mass measurements. Samples were infused using direct loop injection on a Waters Acquity UPLC system. Samples were separated on a Waters Acquity UPLC system using an Acquity BEH C18 1.7 µm column and were eluted with a gradient of 30 – 95% solvent B over 10 min. Mass spectra were recorded from a mass of 300 – 2000 daltons.NMR spectroscopy was used to identify prenyl-products, and quantify 5-p-1,6-DHN. The PyOx/PTA cell-free system was used to produce prenyl-DHN. 200 µL reactions were pooled, and extracted 3 times with an equivalent amount of nonane and then the nonane was evaporated. The product of the reactions was suspended in 500 µL of deuterated methanol , with 2 mM 1,3,5-trimethoxybenzene as an internal standard. Spectra were collected on an AV400 Bruker NMR spectrometer. The amount of the prenylated compound in the sample was determined with reference to the internal TMB standard. We compared the proton signal from TMB at 6.05 ppm with an aromatic proton corresponding to 5-p-1,6-DHN at 7.27 ppm. NMR was also used to identify the product of the enzymatic system with divarinic acid as the aromatic substrate. The PyOx/PTA system was set up as detailed above, and the reactions were quenched at 24 hours. The reactions were extracted as detailed above, and analyzed on the HPLC. There was a new major peak at 6.7 minutes that we predicted to be the prenylated divarinic acid. We HPLC purified the peak, removed the solvent, and re-dissolved the pure component in 600 µL of CD3OD. A proton spectrum collected with an AV500 Bruker NMR spectrometer was compared to a proton spectrum published by Shoyama et al for CBGVA to confirm that CBGVA was the main product. Based on the report by Shoyama et al the study by Bohlmann et al, we conclude that the prenylation of divarinic acid occurs at the C3 carbon of divarinic acid.