Strains and plasmids used in this study are listed in Table 1. Escherichia coli was used as cloning host and was cultivated in lysogeny broth (LB (Bertani, 1951)) at 37°C and 180 rpm. C. glutamicum was used for production experiments. Precultures were cultivated in 100 mL baffled shake flasks containing LB supplemented with 10 g L^-1 glucose at 30°C and 120 rpm. The pre-cultivated cells were used to inoculate the main culture to an initial optical density (wavelength: 600 nm, OD₆₀₀) of 1 determined by using the V-1200 spectrophotometer (VWR, Radnor, PA, United States). All trans-nerolidol producing experiments, including response surface methodology (see Section 2.3), were performed in 48-well FlowerPlates (Beckman Coulter, Brea, CA, United States) with a filling volume of 1 mL per well at 1,100 rpm and 30°C in the BioLector XT microcultivation system (Beckman Coulter, Brea, CA, United States) for 24 h. For the transfer of the refined trace elements to patchoulol and (+)-valencene, 100 mL baffled shake flasks with a filling volume of 10 mL and additional 10% (v v⁻¹) dodecane as organic overlay were used instead of the BioLector. Cultivation lasted 48 h at 30°C and 120 rpm. 40 g L^-1 glucose was added to the CGXII medium as carbon and energy source (Keilhauer et al., 1993) for all main cultures. If appropriate, the medium was supplemented with kanamycin (25 μg mL^-1), tetracycline (5 μg mL^-1), chloramphenicol (7.5 μg mL^-1 for C. glutamicum or 30 μg mL^-1 for E. coli). 1 mM isopropyl-β-D-1-thiogalactopyranoside (IPTG) was added at the start of the main cultivation to induce gene expression.

Primers for DNA amplification and sequencing (Supplementary Table S1) were purchased from Sigma-Aldrich (St. Louis, MO, United States). Plasmid DNA isolation (QIAwave Plasmid Miniprep, Qiagen, Venlo, Netherlands) and PCR clean-up (NucleoSpin^® Gel and PCR Clean-up, Macherey-Nagel, Düren, Germany) were performed according to the manufacturer’s instructions. Nerolidol synthase (NS) gene from Tripterygium wilfordii (GenBank: KU588405) (Su et al., 2017) was codon optimized (Codon Optimization Tool, 2025) and synthesized by Twist Bioscience (San Francisco, CA, United States). Gene fragments were amplified using the Allin™ HiFi DNA polymerase (highQu GmbH, Kraichtal, Germany) and cloned into BamHI-digested (Thermo Fisher Scientific, Waltham, MA, United States) plasmids using Gibson Assembly (Gibson et al., 2009). CaCl₂ chemocompetent E. coli DH5α were transformed with the Gibson Assembly reaction mix by heat shock at 42°C (Sambrook and Russell, 2001). Cloned DNA fragments were verified by DNA sequencing. Trans-Nerolidol producing C. glutamicum strains were created by transforming electrocompetent cells via electroporation and subsequent heat shock at 46°C with the respective plasmids (Eggeling and Bott, 2005).

Cultivation was performed as described in Section 2.1. Due to the total amount of 1 mL per well, the trace elements were prepared separately as 100x stock solutions to facilitate pipetting. Design and analysis of the experiment was performed by using R 4.4.1 (R Studio Team 2024) and the rsm package version 2.10.5 (Lenth, 2009). Within the experiment, the effect of the trace elements MgSO₄, CaCl₂, FeSO₄, MnSO₄, and ZnSO₄ on the trans-nerolidol titer was investigated. The central composite design consisted of a 2⁵ full factorial design as cube portion, ten face-centered star points and six center points. The concentration of each trace element at the center point was regarded as 100%. 50% and 150% of the center point concentration were chosen for the cube portion of the model. For the star points, only one out of the five trace elements was set to either 0% and 200%, respectively, while the others remained at 100%. The overall workflow was: first CCD with standard CGXII trace elements as center points, followed by a steepest ascent experiment. The best trace elements composition of the steepest ascent experiment was used as the center point for the second CCD. All combinations and volumes can be found in the Supplementary Material. The other media components remained constant as described in Section 2.1.

For the extraction of trans-nerolidol, 1 mL of hexane was added to 330 µL of cultivation broth and incubated at 50°C and 1,000 rpm for 30 min (ThermoMixer C, Eppendorf, Hamburg, Germany). The extraction mixture was centrifuged, the organic phase was removed, dried with sodium sulphate and transferred to GC-MS analysis. For patchoulol and (+)-valencene, the dodecane overlay was separated from the aqueous phase by centrifugation and was directly used for analysis. All terpenes were quantified using a TRACE GC Ultra gas chromatograph and a ISQ single quadrupole mass spectrometer equipped with an AS 3000 autosampler and a TraceGOLD™ TG-5 MS column (30 m × 0.25 mm x 0.25 µm) (Thermo Fisher Scientific, Waltham, MA, United States). 1 μL was injected in splitless mode. Helium was used as carrier gas at a constant flow of 1 mL min^-1. Temperatures were set as the following: injector (250°C), interface (250°C), and ion source (220°C). The oven profile was set as the following: 80°C for 1 min, increased to 120°C at a rate of 10°C min^-1, followed by 3°C min^-1–160°C, and a further increase at 10°C min^-1–270°C, which was held for 2 min. Mass spectra were recorded after a solvent cutoff at 14 min using a scanning range of 50–750 m/z at 20 scans s^-1. Chromatograms were evaluated using Xcalibur 2.1.0 (Thermo Fisher Scientific, Waltham, MA, United States). Extracted-ion chromatograms at m/z = 93 were used for quantification of trans-nerolidol, m/z = 138 and m/z = 220 for patchoulol, and m/z = 161 for (+)-valencene. Analytical standards for trans-nerolidol, patchoulol, and (+)-valencene were purchased from Extrasynthese (Lyon, France), Biosynth (Staad, Switzerland), and Sigma-Aldrich (St. Louis, MO, United States), respectively.

Fed-batch fermentation using strain NERO5 was performed in a bioreactor with a total volume of 3.7 L (KLF, Bioengineering AG, Wald, Switzerland). The aspect ratio of the reactor was 2.6:1.0. Three six-bladed Rushton turbines were placed along the stirrer axis at 6, 12, and 18 cm from the bottom of the reactor with a stirrer to reactor diameter ratio of 0.39. 1 L of the high cell density medium based on Knoll et al. (2007) was used with some alterations: all trace elements were replaced by the refined trace elements in a 20x excess compared to the regular cultivation in CGXII (see Supplementary Table S2). Antibiotics and IPTG were added accordingly and the medium was inoculated to an OD₆₀₀ of 1. 500 mL of a 600 g L^-1 glucose solution was used as feed medium. A headspace overpressure of 0.5 bar was applied. The temperature was kept at 30°C during fermentation. The pH of 7 was automatically maintained by the addition of 10% (v v⁻¹) H₃PO₄ and 25% (v v⁻¹) NH₃. An initial airflow of 0.25 NL min^-1 was provided from the bottom through a ring sparger, which was manually increased during the fermentation to 0.75 NL min^-1 when oxygen supply became limiting. The stirrer speed increased automatically in a stepwise manner from 400 rpm to 1,500 rpm, every time the relative dissolved oxygen saturation (rDOS) fell below 30%. The feed pump was primed when the rDOS fell below 60% for the first time. Subsequently, the feed pump was activated every time the rDOS exceeded 50% and stopped as soon as the rDOS fell below 50% thereby preventing overfeeding. To control foam formation, 0.6 mL L^-1 of antifoam 204 was already added to the batch medium. If required, an antifoam probe controlled the supply of antifoam 204 during the process. Samples were automatically taken over the course of the fermentation and stored at 4°C until further use.

The genome of C. glutamicum codes for two geranylgeranyl pyrophosphate (GGPP) synthases (crtE and idsA), but neither of them synthesizes significant amounts of the sesquiterpene precursor farnesyl pyrophosphate (FPP) from the end products of the MEP pathway isopentenyl pyrophosphate (IPP) and dimethylallyl pyrophosphate (DMAPP) (Frohwitter et al., 2014). Trans-Nerolidol production in C. glutamicum wild type was enabled by overexpressing FPP synthase ispA from E. coli together with a codon-optimized nerolidol synthase from T. wilfordii from the plasmid pECXC99E (NERO1, see Figures 1, 2). To prevent further conversion of FPP to GGPP and subsequently to carotenoids, a metabolically engineered wild type derivative lacking both carotenoid operons as well as the GGPP synthase idsA (Henke et al., 2018b) was transformed with the plasmid pECXC99E-ispA _Ec -NS _Tw (NERO2). This strain did not synthesize carotenoids, thus increasing the trans-nerolidol titer 5.9-fold from 1.9 ± 0.1 to 11.2 ± 0.8 mg L^-1. The engineered C. glutamicum strain NERO2 was subsequently used to study the effects of the trace element composition.

Although some media formulations for C. glutamicum contain additional trace elements like Na₂MoO₄ and H₃BO₃ (Weuster-Botz et al., 1997), the classical CGXII medium contains the seven trace elements MgSO₄, CaCl₂, FeSO₄, MnSO₄, ZnSO₄, CuSO₄, and NiCl₂ (Keilhauer et al., 1993). To reduce the number of experiments, it was decided to neglect the two trace elements with the lowest concentration, being CuSO₄ and NiCl₂. This is in accordance with the fact that out of the seven trace elements in CGXII, copper and nickel ions are the metal ions used least as cofactors in enzymes (Waldron et al., 2009). Therefore, it was hypothesized that variation of the copper and nickel ion concentrations might influence terpene biosynthesis the least. A CCD approach, with the standard CGXII trace elements concentrations (see Supplementary Table S3) as center points, was performed in the BioLector microcultivation system using strain NERO2 (see Section 2.3).

The analysis of the trans-nerolidol titer showed no significant two-factor interactions between the trace elements (data not shown). This is why the data was subsequently evaluated considering only first order and quadratic effects. Significant first-order effects of MgSO₄ as well as quadratic effects of MgSO₄ and FeSO₄ were observed (see Supplementary Table S4). Although the model possessed no significant lack of fit, the quadratic effects might be just artificial as the star points containing no MgSO₄ or FeSO₄ did not grow and therefore did not produce any trans-nerolidol (see boxplots in Supplementary Figure S1 and Supplementary Table S3). This is why only first order effects were considered, showing strong positive effects of MgSO₄ (Table 2).

By applying only a first order model, no stationary point with a predicted optimum could be obtained. Furthermore, it was expected to gain more insight into the interactions between the different trace elements and their effects on trans-nerolidol production. A reasonable next step therefore is to follow the direction along where the response increases the fastest to end up at trace element concentrations that result in higher titers than the starting condition. This position would be the new center point used for a second round of CCD (Roberts et al., 2020). The so-called path of steepest ascent was calculated by the results of the first order model out of the CCD, yielding a set of trace element compositions and predicted titers (see Supplementary Table S5). The results of the steepest ascent analysis are shown in Figure 3. Although fluctuating, a trend of increasing production along the path of steepest ascent was observed. At a distance of 4.5, the highest titer (17.1 ± 0.7 mg L^-1) was observed, being significantly higher than the control. This trace element composition was subsequently used as a new center point for another CCD, allowing to investigate concentration ranges that might be closer to an optimum (see Supplementary Table S6). After having performed the second CCD experiment, neither two-factor interactions, quadratic effects or first order models could be fitted significantly (data not shown).

The second round of CCD did not result in a significant result. However, some of the tested trace element combinations within this new range of concentration showed a clearly increased titer compared to the control. To validate these observations, the best composition from the second CCD (see Supplementary Table S6, run 45) was compared to the standard trace elements and to the best combination of the steepest ascent experiment, which was used as the center point of the second CCD. As MgSO₄ showed a strong positive effect in the first CCD (Table 2), the influence of a 4x increased MgSO₄ concentration was investigated as well. The tested trace element compositions are listed in Supplementary Table S2 and the results are shown in Figure 4. All three trace element compositions resulted in significantly increased titers. The highest titer of 14.3 ± 0.4 mg L^-1, corresponding to an improvement of 34%, was achieved using the trace element composition found within the second CCD. No statistical significance was found when comparing the high MgSO₄ composition with the one from the second CCD. However, as the overall amount of trace elements changed the least when compared to the other compositions (see Supplementary Table S2), the combination obtained within the second CCD was now chosen to be the refined trace element combination.

Trace element refinement improved trans-nerolidol production of the base strain NERO2. To answer the question, if the refined trace elements would also support higher trans-nerolidol production, strain NERO2 was further metabolically engineered (Figure 1) before standard and refined media were compared.

To improve the terminal biosynthesis, the synthetic operon of FPP synthase and trans-nerolidol synthase genes was cloned into the pECXT-P_syn vector containing the strong constitutive P_syn promoter (Henke et al., 2021) yielding strain NERO3. Using the standard trace elements, the stronger expression increased the titer significantly to 14.5 ± 0.3 mg L^-1 (compare strains NERO2 and NERO3 in Figure 5). The addition of the empty vector pVWEx1 to NERO3 (NERO4; empty vector control) slightly decreased the titer. To improve the entry reaction into the MEP pathway as well as the interconversion of IPP and DMAPP, dxs and idi were overexpressed as a synthetic operon (NERO5) resulting in a titer of 25.1 ± 0.7 mg L^-1, which was 2.4 times higher compared to strain NERO2 (Figure 5). Next, it was tested, if the refined traces would have a positive impact on trans-nerolidol production by strain NERO5. Indeed, the refined trace elements further significantly improved trans-nerolidol production to the titer to 28.1 ± 1.0 mg L^-1 (+12%).

Since the combination of trace element refinement and metabolic engineering was successful in a microcultivation system, it was evaluated whether the trans-nerolidol production using C. glutamicum strain NERO5 can be stably transferred to a 1.5 L fed-batch process, corresponding to a 1,500-fold volume increase. The high cell density medium based on Knoll et al. (2007) was used for fermentation. However, the trace element composition of this medium corresponds neither to the standard CGXII nor to the refined trace elements. Hence, the refined trace element composition was used instead, while all other media components remained unchanged. To account for the higher cell density, the concentration of refined trace elements was increased proportionally 20-fold relative to standard CGXII cultivations in BioLector or shake flasks (see Supplementary Table S2). The glucose feed was consumed after 74 h resulting in a biomass titer of 57.5 g_CDW L^-1. The trans-nerolidol titer reached 0.41 g L^-1, corresponding to a volumetric productivity of 5.6 mg L^-1 h^-1 (Figure 6).

To test if the refined medium can be transferred as a general strategy for the production of other sesquiterpenes, it was used to cultivate previously established strains overproducing patchoulol (Henke et al., 2018b) and (+)-valencene (Binder et al., 2016). The refined trace elements increased the (+)-valencene titer from 6.0 ± 0.3 mg L^-1 using standard CGXII trace elements to 10.3 ± 1.7 mg L^-1, corresponding to a 72% improvement. Also, the patchoulol titer was increased, i.e., from 10.4 ± 0.4 mg L^-1 to 11.9 ± 0.2 mg L^-1, corresponding to a 15% improvement (Figure 7). Thus, the medium with refined trace elements developed for trans-nerolidol production also significantly improved production of two other sesquiterpenes.