*3.1. Glycerol is a Suitable Carbon Source for Heterologous Limonene Production in Escherichia coli*

The beneficial effect of glycerol as a sole or supplementary carbon source has been reported before for the fermentative production of carotenoids in MEP engineered *E. coli* strains [17–19], for sesquiterpenes [12] and for limonene [13] in MVA-engineered *E. coli* strains. The use of glycerol as a carbon source resulted in higher limonene formation rates, a prolonged growth phase, and increased stability compared to the same whole-cell biocatalyst growing on glucose [13].

In the present study, we were able to transfer this knowledge to a bioprocess with an optimized one plasmid strain and showed that, compared to glucose, glycerol is definitely the preferred carbon source for the production of limonene with *E. coli*. The transferability to the bioreactor scale was validated, and reaction engineering was performed to further increase the limonene titer. This led to the highest monoterpene concentration obtained with a microorganism to date. Next to availability and low cost, various advantages make glycerol an attractive carbon source for fermentation processes compared to glucose. Firstly, beneficial effects on the viability of cells and productivity of recombinant proteins were observed [20]. Secondly, glycerol does not show any catabolic repression in combination with lactose, which might be the preferred inducer for heterologous gene expression instead of IPTG due to a reduced stress level for the production host. Catabolite repression occurs when excess glucose is present and leads to reduced lactose uptake rates, which causes the decreased expression of recombinant proteins [21]. Finally, glycerol is a suitable carbon source for anaerobic fermentation with *E. coli* strains producing biofuels and highly reduced compounds. The high degree of reduction of carbon atoms in glycerol (κ = 4.67) provides a distinct advantage over glucose (κ = 4.00) in the absence of other electron acceptors [22]. *E. coli* strains are able to utilize glycerol in such conditions for cell growth and need a suitable sink for the excess reducing equivalents generated during the formation of biomass [23]. Therefore, the ability to form a highly reduced product is essential for the microorganism. Limonene has a high degree of reduction (κ = 5.60), so it would be a suitable product and sink for reducing equivalents in anaerobic glycerol fermentation. The anaerobic environment could have another beneficial effect regarding the toxicity of limonene. Whereas limonene itself has relatively little toxicity towards *E. coli* cells, the common oxidation product limonene hydroperoxide, which forms spontaneously in aerobic environments, shows highly antimicrobial effects [24]. In this study, the inhibitory effects of limonene hydroperoxide were not observed, due to efficient product extraction in the organic phase.

#### *3.2. Progess to an Economic Limonene Production Process*

The optimized bioreactor process described in this study resulted in limonene productivity exceeding the threshold for developing a profitable production process for fine chemicals (100 mg·L-1·h−1) [25] for the first time (Table 1). However, a techno-economic assessment stated that a biotechnological production process for limonene needs to have a space-time yield above 700 mg·L−1·h−<sup>1</sup> and a 45% carbon specific yield to be competitive with established processes [26]. While our process already shows more than a fifth of the space-time yield required, the conversion of the carbon source into the product is still low, at less than 1%.



STY: space-time yield; STR: stirred tank reactor.

Different approaches to improve the yield have already been described. Non-growing but metabolically active *E. coli* cells can boost the production of the desired product due to reduced energy and carbon loss to biomass formation [28]. A fourfold increase in specific limonene yields relative to biomass was accomplished with this strategy. Moreover, further pathway debottlenecking and the optimization of the involved enzymes could increase the economics of the process. For example, the exchange of the geranyl pyrophosphate synthase with a neryl pyrophosphate synthase from *Solanum lycopersicum* led to increased limonene production with *E. coli* [27]. Another strategy to increase the product titer and the specificity of monoterpenes in *E. coli* was described by Chacón et al. [29]. The monoterpene geraniol was converted to the monoterpenoid geranyl acetate with an in vivo esterification and extracted in situ to an organic phase. Toxicity issues and the synthesis of by-products could be circumvented, resulting in a monoterpenoid concentration of 4.8 g·L<sup>−</sup>1. Similar approaches to the coupled synthesis and functionalization of limonene are described, which produce the valuable monoterpenoid perillyl alcohol [8,30].

The highest yields with more than 95% and titers of 15 g·L−<sup>1</sup> were achieved with a cell-free system consisting of 27 purified enzymes, which convert glucose into monoterpenes [31]. Other systems which incorporate acetic acid as a starting building block for the cell-free synthesis of terpenes are described as well [32]. However, a major drawback is the need for purified enzymes, which are associated with additional costs and the further input of glucose needed to produce them. The need for enzyme purification can be avoided with the use of enzyme-enriched *E. coli* lysates, but this approach appeared to suffer from low product titers of 90 mg·L−<sup>1</sup> [33]. Moreover, the involved enzymes are considered to have low stabilities in the in vitro environment, and cofactor regeneration could be a limiting aspect in cell-free applications as well [32]. The expensive cofactors CoA and NADPH must be effectively recycled in such systems, while the use of whole cells circumvents these drawbacks as the cofactors are regenerated by the primary metabolism. Therefore, microorganisms are preferred as the biocatalyst for the larger biotechnological production of limonene. Microbial hosts other than *E. coli* were recently investigated as producer strains, such as the cyanobacterium *Synechocystis* sp. [34] or the oleaginous yeast *Yarrowia lipolytica* [35], which was able to produce limonene from waste cooking oil. However, product titers were orders of magnitude lower compared to the processes based on engineered *E. coli*.

Next to the selection and optimization of the production system, a feasible bioprocess with high limonene titers involves the integrated development of in situ product removal strategies. Due to the high volatility and inhibitory effects on cell growth, the capturing of limonene during fermentation is required. Various methods are available, with two-liquid phase and gas stripping systems being especially suitable at higher scales [36]. In particular, two-liquid phase systems have the advantage that the products are effectively removed from the fermentation broth [37]. The choice of capturing method is also dependent on the further application of the product. If limonene is subsequently used as a pure compound, solvent-free systems might be the better choice, whereas application as an additive for, e.g., solvents might allow the use of the same solvent for in situ extraction [38]. In the present study, the in situ product removal strategy in combination with an engineered *E. coli* strain and a glycerol-limited fed-batch fermentation enabled the synthesis of the highest limonene concentration reported to date. Steps towards an economic process were made, and the potential of integrating already generated knowledge with the biotechnological production of terpenes was demonstrated.

#### **4. Materials and Methods**

#### *4.1. Chemicals and Bacterial Strains*

All chemicals used in this work were purchased from Carl Roth GmbH & Co. KG (Karlsruhe, Germany) and Merck KGaA (Darmstadt, Germany).

*E. coli* BL21 (DE3) harboring the plasmid pJBEI-6410 was used for all experiments. pJBEI-6410 carries the genes for the MVA pathway, a geranyl pyrophosphate synthase, and the limonene synthase from *Mentha spicata* [8]. It was a gift from Taek Soon Lee (RRID:Addgene\_47049; http://n2t.net/addgene:47049).
