In addition to the various classes of natural products described above, there are some hybrid natural products that span two classes. Cannabinoids for example are derived from the terpenoid pathway and the polyketide pathway . Prenyl-flavonoids and prenylstilbenoids would also fall under this category. Cannabinoids are a very interesting class of natural products, with some pretty remarkable medicinal properties. For the past 83 years cannabis or marijuana was considered a schedule one drug by the United States, which stunted the research into the molecules that cannabis makes. Although it is still classified as a schedule one drug by the federal government, the perception of the plant and the molecules it makes is starting to change. Thirty-three states now permit the use of medical marijuana and 10 states and Washington DC have legalized it for recreational use. The shift in perception may be due to recent clinical studies. A non-psychoactive component of cannabis, cannabidiol or CBD was FDA approved to treat severe childhood epilepsy, and has been suggested as a possible treatment for the spasticity associated with diseases like Parkinson’s Disease and Multiple Sclerosis . THC or D9 -tetrahydrocannabinol, the psychoactive component of marijuana is FDA approved as an appetite stimulant for patients going through chemotherapy.54 In addition to the conditions these cannabinoids are already approved for, both THC and CBD are cited to have antiemetic, antianxiety, anti-inflammatory, antidepressant and anticonvulsant properties. While THC and CBD are produced abundantly by cannabis plants, several groups have attempted to engineer microbes to produce cannabinoids and their intermediates in order to create a more sustainable source of these pharmaceutical compounds. As shown in Figure 1-7 the biosynthesis of cannabinoids can be broken down into three parts, cannabis grow room the polyketide biosynthesis , the terpenoid biosynthesis and the cannabinoid biosynthesis .

Gagne et al were the first group to transform the olivetolic acid biosynthetic pathway into yeast. Previous work by Taura et al had identified an olivetol synthase, capable of making the decarboxylated form of OA. Gagne et al posited that there was an additional enzyme in the biosynthetic pathway that would cyclize the polyketide chain to form OA instead of olivetol. In this work they identified olivetolic acid cyclase, which cyclizes tetraketide intermediate to form OA instead of olivetol. Gagne et al transformed the genes necessary to produce OA in S. cerevisiae, and reported a titer of 0.5 mg/L; however this titer is not optimized. Six years later Tan et al engineered the same pathway into E. coli. They added additional enzymes to increase the malonyl-CoA and hexanoyl-CoA precursors, and were able to obtain 80 mg/L of OA. At the other end of the biosynthesis , Zirpel et al have engineered several yeast strains to catalyze the final two steps in the biosynthetic pathway. First in 2017, Zirpel engineered P. pastoris to express a promiscuous prenyltransferase, NphB, and THCA synthase to convert OA into THC. While the yeast expressed both enzymes, they did not detect any THC until they lysed the cells and supplemented with two precursors, OA and geranyl pyrophosphate . Even then, the production of THC was slow at ~ 10 nmol/L/hr, which was attributed to an inefficient prenylation step, and low levels of the cannabinoid synthase. In later studies, Zirpel et al focused on engineering the yeast strains to improve the expression of the cannabinoid synthases THCAS and CBDAS. They identified five proteins that aid in the expression of the cannabinoid synthases. When they supplemented the yeast over expressing the cannabinoid synthases with CBGA, they were able to produce over 3 g/L of THCA and 400 mg/L of CBDA, a significant improvement to their first study.60 Only one group has reported the entire biosynthesis of cannabinoids in yeast, however the titers are low.

They produced the precursor cannabigerolic acid at 8 mg/L, THCA at 1.1 mg/L and CBDA at 4.2 µg/L, when the cultures were supplemented with 1 mM hexanoic acid. These low titers indicate the difficulty associated with engineering a microbe to produce a complex natural product. When the authors supplemented cultures with 1 mM of the precursor OA, they were able to produce higher levels of the CBGA intermediate 200 mg/L a 55% conversion of the OA added. This may indicate that OA biosynthesis is limiting in their strain, probably due to the low malonyl-CoA concentrations. Similarly to other engineered microbes for opioid, paclitaxel and artemisinic acid production, the cannabinoid pathway is very complex. The root of the pathway is the essential precursor acetyl-CoA, which is required to make both the aromatic polyketide component, and the isoprenoid component of the cannabinoid. Additionally there is always the possibility of product or intermediate toxicity, which can limit titers. Further, expression of the cannabinoid synthases can be challenging. Zirpel et al observed low titers of THC until they co-expressed the synthase with several chaperones, and an enzyme to improve cofactor biosynthesis. Listed above are 15 examples of microbes that have been engineered to produce various natural products , and it’s evident that microbial engineering allows for the production of vastly different natural products. Microbial fermentation is an important tool for the production of natural products, and is used in several industrial processes. However, the cases presented above demonstrate the highly variable nature of titers, ranging from µg/L to g/L, and its apparent that some studies are more successful than others. The studies that are more successful tend to be shorter metabolic pathways, and they do more than simply add an exogenous bio-synthetic pathway, they try to drive carbon flux into the target pathway. This highlights one inherent problem with engineering microbes.

The target pathway engineered into the microbe is always competing with native metabolic pathways for cofactors and metabolites. Therefore, essential pathways will always deplete resources from the target pathway. Additionally, metabolites from the background metabolism have the potential to inhibit the target pathway. These interactions are extremely challenging if not impossible to identify in vivo, and can limit the overall titers of any pathway. The final product can also decrease cell viability, like in the case of monoterpenes and alcohols, and limit product titers. Finally, while engineering microbes is fairly simple and efficient compared to other organisms, it still takes a significant amount of time to develop strains, approximately “150 person-years” of work were required to engineer yeast to produce artemisinic acid at 25 g/L.62 There are two alternatives to metabolic engineering, cell-free biosynthesis and synthetic biochemistry. They are built on the same principles, using enzymes to produce a target molecule, but instead of engineering a microbe, the biocatalysts are reconstituted either in a cell lysate or in vitro. The difference between cell-free and synthetic biochemistry is small, with cell-free relying on the cellular machinery in a lysate to recycle essential co-factors, whereas synthetic biochemistry relies on pathways with the ability to recycle the necessary co-factors. Cell-free systems are most commonly used to power protein expression, however there are some cases where lysates have powered the production of a biosynthetic pathway. There’s. lot of flexibility within this set up. Kay et al produced 2,3 butanediol at 84 g/L using a lysate derived from an E. coli strain expressing all three enzymes in the pathway, but Dudley et al expressed the desired proteins individually, lysed the cells and mixed the various lysates together to produce mevalonate at 17.5 g/L. Interestingly, when Dudley applied the same concept to the production of limonene , the titers were significantly lower at approximately 100 mg/L.This type of system shares some of the challenges with metabolic engineering. Even though the cells are lysed, metabolic enzymes still remain active; therefore its possible for carbon flux to be diverted from the target pathway. Dudley et al cite this as a major challenge for the cell-free production of limonene. The native E. coli farnesyl pyrophosphate synthase remains active, grow trays and siphons flux away from the monoterpene pathway to produce a sesquiterpene alcohol. Additionally, metabolites present in the lysate can inhibit the pathway of interest and limit titers. However, cell viability is no longer limiting, products can be extracted in real time limiting product inhibition and the target pathway protein levels are easily manipulated.Synthetic biochemistry provides several advantages over metabolic engineering. Many of the successful studies mentioned above required altering central metabolism to account for the target pathway, but in every instance it was a balancing act between essential pathways and the target pathway. A synthetic biochemistry approach also allows for rapid design, build, test cycles, which makes it easier and faster to identify pathway bottlenecks. Once bottlenecks are identified, it is very easy to tailor the enzyme activity to alleviate the problem, something that is very challenging to do in vivo. Additionally, because enzymes can be expressed and purified from a range of expression platforms , it increases the pool of potential enzymes that can be used.

Finally, like the cell-free approach the products can be easily extracted, limiting problems associated with product toxicity. However there is a challenge with this approach. Without the cell, the in vitro pathways need to incorporate components that will balance and regenerate co-factors. In their simplest form, in vitro enzymatic systems can be broken down into two modules, the sugar breakdown module and the build module. The sugar breakdown module catabolizes the sugar into 2-3 carbon building blocks and generates high energy co-factors like ATP and reducing equivalents in the process. Those components are then assembled in the build phase to generate the final product. The challenge is the high energy cofactors produced in the sugar breakdown module are not always balanced with the build module. This is the case for the in vitro biosynthesis of monoterpenes and isobutanol. Korman et al engineered a system with 27 enzymes to biosynthesize monoterpenes. To produce a monoterpene, the feedstock, glucose, was converted into pyruvate via glycolysis, pyruvate was converted into acetyl-CoA with pyruvate dehydrogenase , acetyl-CoA fed into the mevalonate pathway to yield geranyl pyrophosphate , which was converted into a monoterpene via a monoterpene synthase. The pathway originally was not stoichiometrically balanced for the reducing equivalents, NADH. The sugar breakdown module originally produced 6 moles of NADH, and the build module required 2 moles of NADPH. Not only were the reducing equivalents stoichiometrically unbalanced, they were not the type the build module needed. To counteract this imbalance Korman et al employed a molecular purge valve. The purge valve consists of three components. First, an NAD+ specific glyceraldehyde-3-phosphate dehydrogenase is responsible for maintaining carbon flux through glycolysis. An NADH oxidase was used to burn excess reducing equivalents, and an NADP+ specific GAPDH was used to generate the reducing equivalents required for the mevalonate pathway. All three components were required for the in vitro system to run efficiently. Using the NADP+ specific GAPDH alone leads to a buildup of NADPH, which prevents the conversion of glyceradehyde-3-phosphate into 1,3-bisphosphoglycerate . This would eliminate ATP regeneration via the glycolysis pathway and eventually stop the system. The pathway yielded 12.5 g/L of limonene and 14.9 g/L of pinene, which is significantly higher than titers achieved in microbes. Opgenorth et al engineered an in vitro enzymatic system to produce isobutanol. As previously mentioned alcohols are fairly toxic to microorganisms. Therefore, an in vitro, cell-free approach might lead to higher titers. For isobutanol, the sugar breakdown module is glycolysis which nets 2 reducing equivalents and 2 ATP per cycle. The isobutanol build module requires 2 moles of NADPH, but 0 ATP. Therefore, in order to balance the system Opgenorth et al utilized GapN, an enzyme that converts G3P and NADP+ into 3-phosphoglycerate and NADPH, thereby eliminating a step that regenerates an ATP. This leads to a system that is stoichiometrically balanced, however, this may not be optimal in a cell-free system. Opgenorth et al found that ATPase activity from purified enzymes or spontaneous ATP hydrolysis can deplete ATP stores and limit product titers. So, they engineered a component to generate excess ATP under high phosphate conditions , called the molecular rheostat. The rheostat is a branchpoint in glycolysis providing two paths that convert G3P into 3PG, with one path produces ATP and the other does not. The GapN path directly converts G3P into 3PG yielding no ATP. The phosphorylation path converts G3P into BPG with an NADP+ specific GAPDH, and BPG is converted into 3PG using phosphoglycerate kinase, which regenerates an ADP into ATP. Because the activity of the GAPDH enzyme is dependent on the phosphate concentration, the rheostat ATP production is also dependent on the phosphate concentration.