The necessity for biological processes utilizing renewable resources becomes critical in part due to global warming and depletion of petroleum. Most biological engineering efforts have been focused on developing microbial systems that can be genetically altered to produce high-value chemicals and petroleum substitutes. Although our understanding of tuning microbes into efficient workhorses for chemical production is still limited, technological developments in synthetic biology promise to make biological engineering more predictable, faster, and less expensive . All major classes of secondary metabolites including phenylpropanoids, terpenoids, and polyketides have been produced in microbes [7, 8, 13, 21–25]. Yet most of the natural products are still supplied primarily through fungal fermentation and plant cultivation accompanied by subtle chemical modifications. In this respect, the yeast engineered to produce artemisinic acid is not only a step towards a potential economically viable method for producing artemisinin at a commercial scale but also a research model to study the microbial manufacture of specialty and commodity chemicals.
In this study, we demonstrated that the production titer of artemisinic acid can be increased by altering the plasmid selection marker and culture medium. Without any optimization, the artemisinic acid yields from both shake-flask and bio-reactor cultivation of the engineered yeast were 10-fold higher than those yields reported previously . Further optimization of gene expression and cultivation conditions in order to take advantage of the full metabolic capacity of these microbes should improve artemisinic acid yields even more.
Although all of the amorphadiene produced was not completely oxidized to artemisinic acid, the conversion rate is estimated to be at least 32% by simply comparing the yields of amorphadiene and artemisinic acid (Figure 1B and Figure 2B). However, this value may be underestimated because the stability of pESC-Leu2d::T is two-fold lower than that of pESC-Leu2d::ADS. Considering the lower plasmid stability in artemisinic acid-producing yeast strain, the conversion rate from amorphadiene to artemisinic acid could be as high as 64%. We attempted to measure the amorphadiene that was not oxidized in the artemisinic acid-producing yeast by overlaying dodecane in the culture; however, dodecane sequesters amorphadiene efficiently and hence the total artemisinic acid production was severely impaired by dodecane (data not shown). Although almost immiscible, terpenes still have a low, but measurable water solubility. For example, a range of 5–30 mg per liter solubility in water was reported for monoterpenes such as limonene, myrcene, and pinene [26, 27]. Thus, an excess of unoxidized amorphadiene is likely to remain in the medium and diffuse through the yeast membrane before being oxidized by AMO/CPR or vaporized into the atmosphere. The sequestration of amorphadiene by dodecane in cultures of artemisinic acid-producing yeast suggests that AMO/CPR inside the cell could not instantly oxidize endogenously produced amorphadiene. It is difficult to accurately measure the "in-cell conversion rate", but apparently the AMO/CPR-oxidation capability needs to be improved to fully convert the amorphadiene to artemisinic acid. Improvement of the kinetic properties of P450s is likely to be challenging, though such improvements have been demonstrated in a microbial system . Nonetheless, genes/enzymes in plant secondary metabolism are evolutionarily recent, and hence there may be room for further improvement of their catalytic efficiencies by in vitro evolution or site-directed mutagenesis.
It was unexpected that the plasmid stability of the yeast strain expressing ADS in pESC-Leu2d reached 84% in non-selective rich medium, which was the highest of all conditions tested (Figure 1D). The plasmid stability in non-selective medium was even higher than that of yeast strain harboring the same plasmid in the selective medium. FPP and other mevalonate pathway intermediates have been reported to be cytotoxic in E. coli [21, 29]. It is likely that yeast is also under similar stresses when FPP and other precursors accumulate inside the cells. In the yeast strain expressing ADS on pESC-Leu2d in rich medium, the expression level of ADS per cell could be elevated by placing it on a Leu2d-based high-copy plasmid and by supplying a high-level of nutrients in rich medium. Hence, potentially cytotoxic FPP could be rapidly converted to amorphadiene, which is then transported or diffuses across the plasma membrane and cell wall and evaporates from the culture medium. The reduced intracellular concentration of FPP in cells harboring a high-copy ADS and grown in rich medium might stabilize the plasmid harboring ADS. However, it should be noted that converting too much FPP to amorphadiene could also be toxic to cells as it would deplete precursors needed for ergosterol production.
The mechanism of olefin secretion from plant cells remains unknown, but hydrophobic olefins such as amorphadiene might be diffusible through the cell membrane without involving transporters. During cultivation (and production) at 30°C, amorphadiene is likely to diffuse through the membrane, and the volatile amorphadiene is readily vaporized. On the other hand, artemisinic acid produced endogenously should be ionized under neutral cytosolic conditions. Since ionized artemisinic acid requires transporters to cross the yeast plasma membrane, engineered yeast might utilize its own non-specific transporters to secrete artemisinic acid into the medium. As artemisinic acid is likely to be produced faster than it can be transported from the cell, the engineered yeast harboring the plasmid carrying AMO is at a selective disadvantage relative to the same yeast with no plasmid, which results in the low plasmid stabilities observed in non-selective medium (Figure 2D). This suggests that artemisinic acid might have a certain level of cytotoxicity in yeast.
The cellular stresses evoked by artemisinic acid were reflected by the massive transcriptional induction of the genes coding for ABC transporters as determined by RT-PCR and q-PCR. The unbiased microarray analysis further confirmed that the key ABC transporter genes (SNQ2, YOR1, PDR15, and PDR5) were among the genes highly up-regulated in the artemisinic acid-producing yeast. Nonetheless, yeast appears to tolerate the high-level of artemisinic acid by operating its efficient efflux pumps, such as the ABC transporters. It is difficult at this stage to identify the specific transporters or a group of transporters responsible for the secretion of artemisinic acid from yeast. In the future, it would be feasible to identify responsible transporters by deleting transporter genes in the engineered yeast or using deletion yeast strains to evaluate their sensitivity to artemisinic acid as was previously demonstrated in the artesunate (anti-malarial drug)-sensitivity test in mutant yeasts . The nature and kinetic competency of yeast efflux pumps for artemisinic acid are unknown. Thus, the identification and proper expression of authentic plant artemisinic acid transporters from the plant Artemisia annua might enhance the efflux of artemisinic acid from engineered yeast. It would be interesting to reconstitute plant transporters in engineered yeast to see if plants have evolved an efficient transporter for a specific class of natural product such as sesquiterpenoids.
The global microarray analysis of transcriptomes of the yeast producing artemisinic acid relative to the yeast producing amorphadiene was used to identify the stress responses associated with artemisinic acid production. The microarray analysis confirmed that expression of genes encoding ABC-transporters were up-regulated in artemisinic acid-producing yeast. Additionally, microarray data indicated multiple genes involved in oxidative and osmotic stresses and other general stress-responses were significantly up-regulated. Despite the difficulties in distinguishing whether these responses were the direct effects of artemisinic acid-related toxicity or indirect effects from the initial stress responses, the T-profiler and YEASTRACT analyses revealed that at least three cis-elements in multiple genes were clearly responsive to artemisinic acid and/or its biosynthesis. As predicted, the Pdr3p and related transcription factors involved in multi-drug resistant signaling pathways were identified, and Msn2p/Msn4p- and Reb1p-binding motifs were also identified. The reason for the involvement of the Reb1p (RNA polymerase I enhancer binding protein) in the artemisinic acid stress is not clear, but Msn2p/Msn4p might be related to oxidative stress associated with artemisinic acid synthesis as these zinc-finger transcription factors are known to regulate superoxide dismutase and thiol peroxidase [31–33]. In the P450 catalytic cycle, reactive oxygen species (ROS) are formed when activation of molecular oxygen and substrate oxygenation are uncoupled . Cytochrome P450 is a heme-thiolate enzyme in which the distal axial ligand of the heme-moiety is occupied by conserved cysteine residue . The strong electron donation by anionic thiolate to the heme moiety facilitates the heterolysis of dioxygen occupying the position trans to the thiolate [36, 37]. When cysteine was altered to glycine in the C439G AMO mutant, the heterolytic cleavage of dioxygen cannot occur, and hence ROS species are not generated.
Our microarray data are in general consistent with other yeast mutant and microarray data describing the weak acid stress-responses in yeast. The screening of gene-deletion yeasts identified PDR5 and TPO1 as major resistant determinants against the anti-malarial drug artesunate (weak acid) which is remarkably consistent with our data (PDR5 and TPO1 showed 9- and 4- fold induction at the 24 h time point; Figure 5). A comparison of yeast genes induced by treatment with 2,4-dichlorophenoxyacetic acid (2,4-D) and those by artemisinic acid produced by our engineered yeast revealed that multi-drug resistance transporters belonging to ABC and MSF transporters families (e.g. PDR5, PDR15, YOR1 and TPO1) were common genes strongly activated under both conditions . Considering all data together, it is apparent that the engineered yeast is under the cellular stress exerted by a weak acid (i.e., artemisinic acid).
Finally, microarray data suggests that artemisinic acid itself may evoke osmotic stress resulting from lipophilic weak acid stress in the engineered yeast. Undissociated artemisinic acid could passively diffuse into the yeast cytosol where it dissociates into a proton and an artemisinic acid anion, which can be actively exported by the cells, resulting in an accumulation of protons in the yeast cytosol and its acidification. Alternatively, artemisinic acid could directly cause a destabilization (e.g. increased porosity) of the yeast cell wall, leading to decreased plasma membrane stability. The glycosylphosphatidylinositol (GPI)-anchored cell wall proteins, such as SPI1 and YGP1, are believed to strengthen cell walls to protect yeast against stress caused by lipophilic weak acids, such as the food preservative benzoic acid [39, 40]. Induction of these genes during artemisinic acid production (approximately 1.7-fold and 4 fold, respectively) together with several other genes inducible by osmotic stress indicates that yeast tries to compensate for the membrane destabilization effect of artemisinic acid.
The fact that yeast can tolerate multiple stresses and produce one gram of artemisinic acid per liter in a laboratory scale bio-reactor suggests that yeast may be an ideal platform to produce artemisinic acid or other related lipophilic weak acids. Although the biological significance of the pleiotropic drug response and oxidative and osmotic stresses needs to be further investigated in this context and the potential resistance determinants to artemisinic acid production are still not clearly determined, this study provides a basis for endogenous artemisinic acid production-induced stresses and will be useful in selecting future potential targets for metabolic engineering.