**1. Introduction**

Monoterpenes are volatile, lipophilic compounds in the essential oils of plants, which often find application as flavors and fragrances in food, cosmetics, and household chemicals. Limonene is the predominant monoterpene in the essential oils of citrus fruits and can be found in oaks, pines, and spearmint as well. Recently, limonene has been investigated as a promising alternative or additive for solvents [1] and jet fuels [2–4]. Limonene also shows antimicrobial properties [5], can be easily functionalized because of its two double bonds [6], and thus finds application as a building block for several commodity chemicals and pharmaceuticals. The oxygenated derivatives of limonene show potent pharmaceutical activities. As an example, perillyl alcohol, which can be obtained by the regiospecific oxygenation of limonene via whole-cell biotransformation [7,8], has proven anti-cancer properties [9]. The application of monoterpenes as starting materials for industrially or pharmaceutically relevant compounds requires efficient synthesis routes [10]. Nowadays, limonene is mainly produced as a by-product of orange juice production. However, the establishment of new applications will lead to a rapidly growing global market. The low concentrations of monoterpenes in natural sources make their isolation often economically unfeasible. Chemical synthesis might offer alternative production strategies. However, the chemical synthesis of these complex and often

chiral molecules is typically difficult, involves many synthesis steps, and suffers from low yields. In order to ensure a stable and sustainable limonene supply, the development of a biotechnological process for limonene synthesis complements the traditional production route. Moreover, such a process could serve as a basis for the production of other monoterpenes of interest and subsequent selective functionalization.

During recent years, recombinant microbial strains have been engineered for limonene synthesis [11]. The production of isoprenoids with bacterial hosts was challenged by the low supply of the common precursors isopentenyl pyrophosphate (IPP) and dimethylallyl pyrophosphate (DMAPP) via the native 2-C-methyl-d-erythritol 4-phosphate (MEP) pathway. Higher precursor availability was realized by the introduction of a heterologous mevalonate (MVA) pathway from *Saccharomyces cerevisiae* in *Escherichia coli*, and isoprenoid titers above 100 mg·L−<sup>1</sup> were achieved for the first time. A nine-enzyme pathway was constructed on three plasmids to produce amorpha-4,11-diene, which is the sesquiterpene precursor to artemisinin, an antimalarial drug [12]. Based on this study, an equivalent set of plasmids was designed to produce limonene with recombinant *E. coli* [8]. The pathway was optimized by balancing the involved enzymes in several iterative steps, and the number of plasmids was reduced to a single plasmid (Figure 1). Cultivations of the engineered *E. coli* strain in shake flasks using glucose as carbon source in a complex medium resulted in limonene titers of up to 400 mg·L<sup>−</sup>1.

**Figure 1.** The heterologous mevalonate (MVA) pathway and limonene synthase introduced into *Escherichia coli* for the production of (S)-limonene. Acetoacetyl-CoA synthase from *E. coli* (atoB), HMG-CoA (hydroxymethylglutaryl-CoA) synthase from *Saccharomyces cerevisiae*(HMGS), an N-terminal truncated version of HMG-CoA reductase from *S. cerevisiae* (HMGR), mevalonate kinase (MK), phosphomevalonate kinase (PMK), phosphomevalonate decarboxylase from *S. cerevisiae* (PMD), isopentenyl diphosphate isomerase from *E. coli* (idi), a truncated and codon-optimized version of geranyl pyrophosphate synthase from *Abies grandis* (trGPPS), and a truncated and codon-optimized version of limonene synthase from *Mentha spicata* without the plastidial targeting sequence (LS).

Willrodt et al. constructed another *E. coli* strain harboring a two-plasmid system (pBAD:LS, pET24:AGPPS2) and operated a two-liquid phase fed-batch setup with a minimal medium in a stirred-tank bioreactor [13]. In this study, the addition of an inert organic phase was used to prevent product inhibition, toxicity effects, and the evaporative loss of limonene. Diisononyl phthalate (DINP) was selected as a biocompatible organic carrier solvent because of its favorable partition coefficient and lack of detectable impact on the growth of *E. coli* [14]. Final limonene concentrations of 1350 mg·L−<sup>1</sup> were reached with glycerol as the sole carbon source, which was an almost 4-fold increase in limonene formation compared to that from glucose fermentations using the same strain. The use of glycerol resulted in a prolonged growth and production phase, leading to a more stable process with a maximum space-time yield of about 40 mg·L−1·h−<sup>1</sup> for carbon-limited cultivation [13].

Rational strain optimization, as well as reaction engineering, demonstrated the potential of the biotechnological production of monoterpenes. Nevertheless, space-time yields and product titers are still not applicable for industrial production. Additionally, data obtained at the bioreactor scale are rare. This study aims at the development of a feasible bioreactor scale process for monoterpene production with a recombinant *E. coli* strain that is genetically optimized for limonene synthesis.

## **2. Results**

#### *2.1. Influence of Inducer Concentration on Limonene Yields*

Previous studies with a single plasmid strain (*E. coli* DH1 pJBEI-6409) cultivated in complex medium elucidated that low inducer concentrations (0.025 mM isopropyl <sup>β</sup>-d−1-thiogalactopyranoside (IPTG)) resulted in the highest limonene titers [8]. It was hypothesized that the amount of LacI produced by the single copy of *lacI* in the vector might not be enough to fully repress all three promoters in pJBEI-6409. The fully expressed MVA and limonene pathway at high IPTG levels could be too stressful for efficient limonene production. In the present study, different IPTG levels (0.025, 0.05, 0.1, 0.2, 0.5, and 1 mM) were tested for the optimal expression of heterologous genes. In comparison to the mentioned study, a different single plasmid strain was used (*E. coli* BL21 (DE3) pJBEI-6410), which carries a version of pJBEI-6409 harboring ampicillin resistance instead of chloramphenicol resistance. Furthermore, fermentations were carried out in M9 minimal medium instead of a complex medium. It turned out that the highest biomass specific yields could be obtained with IPTG concentrations of 0.05 mM and 0.1 mM (Figure 2). These values are high compared to the reported inducer concentrations for the producer strain *E. coli* DH1 pJBEI-6409 [8]. Following the hypothesis of Alonso-Gutierrez et al., the higher optimal inducer levels could be explained by a higher *lacI* expression level [8]. In contrast to *E. coli* DH1, the host strain *E. coli* BL21 (DE3) carries a Lac regulatory construct in its genome [15]. This operon includes *lacI*q, which is a mutant of *lacI* with a 10-fold higher expression level that leads to a lower basal expression of T7 RNA and therefore to a more tightly controlled expression [16]. For the following experiments, the inducer concentration of 0.1 mM IPTG was chosen to ensure sufficient induction during bioreactor experiments.

**Figure 2.** Biomass specific yields for different concentrations of the inducer isopropyl <sup>β</sup>-d−1-thiogalactopyranoside (IPTG) after 12 h of cultivation. Two-liquid phase shake flask fermentations with *E. coli* BL21 (DE3) pJBEI-6410 in M9 minimal medium with 0.5% *w*/*v* glucose as the sole carbon source. The error bars relate to biological duplicates.
