Gram‐scale production of the sesquiterpene α‐humulene with Cupriavidus necator

Terpenoids have an impressive structural diversity and provide valuable substances for a variety of industrial applications. Among terpenes, the sesquiterpenes (C15) are the largest subclass with bioactivities ranging from aroma to health promotion. In this article, we show a gram‐scale production of the sesquiterpene α‐humulene in final aqueous concentrations of 2 g L−1 with the recombinant strain Cupriavidus necator pKR‐hum in a fed‐batch mode on fructose as carbon source and n‐dodecane as an extracting organic phase for in situ product removal. Since C. necator is capable of both heterotrophic and autotrophic growth, we additionally modeled the theoretically possible yields of a heterotrophic versus an autotrophic process on CO2 in industrially relevant quantities. We compared the cost‐effectiveness of both processes based on a production of 10 t α‐humulene per year, with both processes performing equally with similar costs and gains. Furthermore, the expression and activity of 3‐hydroxymethylglutaryl‐CoA reductase (hmgR) from Myxococcus xanthus was identified as the main limitation of our constructed C. necator pKR‐hum strain. Thus, we outlined possible solutions for further improvement of our production strain, for example, the replacement of the hmgR from M. xanthus by a plant‐based variant to increase α‐humulene production titers in the future.

necator (Eisenreich et al., 2004). However, there are examples of the expression of the MVA pathway in C. necator (Hyeok-Won et al., 2019;Krieg et al., 2018). C. necator can grow on heterotrophic carbon sources, for example, fructose or glycerol, as well as litoautotrophically on CO 2 and H 2 . Two hydrogenases catalyze the oxidation of H 2one membrane-bound hydrogenase transfers electrons into the electron transport chain, and one cytosolic hydrogenase generates reducing power by NADH regeneration for CO 2 fixation. The growth on CO 2 is mediated by the Calvin-Benson-Bassham cycle (Bowien & Schlegel, 1981). Under nutrient, for example, nitrogen limitation, C. necator directs most part of its carbon flux to the synthesis of the storage compound polyhydroxybutyrate (PHB), a biopolymer which is stored in granules in the cell (Reinecke & Steinbüchel, 2009). C. necator is able to store up to 70% of its own weight in PHB (Ishizaki et al., 2001) while growing autotrophically. However, even values up to 90% were reported (Aragno et al., 1977). The natural pathway of PHB synthesis in C. necator has acetoacetyl-CoA as a key intermediate. A PHB-deficient C.necator strain (Raberg et al., 2014) was utilized to use the flux toward acetoacetyl-CoA and redirect it to the recombinant MVA pathway combined with the α-humulene synthase from Zingiber zerumbet for the production of α-humulene. The original pathway was expressed in Methylobacterium extorquens (Sonntag et al., 2015). Due to a similar codon usage, the heterologous pathway for M. extorquens could be used directly in C.
necator. Therefore, the genes coding for the mevalonate pathway, the farnesyl pyrophosphate synthase (erg20), the IPP isomerase (fni), and the a-humulene synthase (zssI) were amplified, and introduced into a previously constructed and characterized vector for expression in C. necator pKRrha (Sydow, Pannek, et al., 2017) to form the plasmid pKR-hum. The resulting strain C. necator pKR-hum was already shown to be able to produce α-humulene under heterotrophic and autotrophic conditions as well as in an electrochemical system (Krieg et al., 2018).
In this article, we compare the theoretically possible industrialscale production of α-humulene with the C. necator pKR-hum strain under autotrophic and heterotrophic conditions in terms of costefficiency and investigate the actual production limits in a heterotrophic fermentation. In addition, we identify the limitations of the system and propose an optimization strategy to increase the αhumulene production with C. necator to industrially relevant production titers.

| Modeling of heterotrophic versus autotrophic production of α-humulene
The economic efficiency of the heterotrophic and autotrophic production processes of α-humulene with C. necator was calculated based on simulations with literature data. For the modeling of the different production systems and their respective unit operations, SuperPro Designer 11 (Intelligen, Inc.) was applied. The exact description and the applied parameters can be found in the supplementary information.

| Cultivation media
Lysogeny broth (LB) was used as a complex medium for precultures, if not stated otherwise. LB was composed of (in g L −1 ): tryptone/peptone 10, yeast extract 5, and NaCl 10. Minimal medium (MM) for C. necator was composed of (in g L −1 ): were supplemented with 15 µg mL −1 tetracycline hydrochloride (Tc) for recombinant strains. L-rhamnose was used as inducer at a final concentration of 11 mM (0.2%) using a 20% (w/v) stock solution for αhumulene production. All basic media components were purchased from Sigma-Aldrich and Carl Roth.

| Plasmid construction and transformation of plasmids into E. coli and C. necator
The construction of the pKRha, the pKR-hum, and the pKR-hum ΔzssI was described previously (Krieg et al., 2018). The construction of the pKR-hum ΔMVA plasmid is described in detail in the supplementary information.
Escherichia coli strains were made chemically competent for plasmid uptake by a standard protocol modified after Hanahan (1983). For details see the supplementary information methods section.
Plasmids were transferred to the recipient C. necator H16 PHB − 4 (Raberg et al., 2014) by conjugation. The details are given in the supplementary information.

| Heterotrophic cultivation in shake flasks
Strains were cultivated at 30°C and 37°C for C. necator and E. coli, respectively. Liquid cultures were shaken at 180 rpm in the incubation shakers Minitron or Ecotron (Infors AG) with a deflection of 25 mm. LB was used as a complex medium for all strains, if not stated otherwise, and supplemented with 15 µg mL −1 tetracycline for both bacteria, according to the applied plasmids.

| Autotrophic cultivation in injection flasks
Autotrophic cultivation was performed in 200 mL septum flasks containing 20 mL minimal medium without fructose and 5 mL n-dodecane as an organic phase for in situ product removal. A heterotrophic preculture was used as inoculum. The defined gas atmosphere was composed of H 2 /CO 2 /O 2 (64:16:20). Cultures were grown at 30°C and 180 rpm (Minitron or Ecotron, Infors AG) with a deflection of 25 mm and induced at exponential growth phase with 11 mM L-rhamnose. Samples of aqueous and organic phases were taken frequently to follow growth and α-humulene production (via LC-MS/MS), respectively.  2.8 | Heterotrophic fed-batch fermentation in DASGIP® parallel fermentation system C. necator (pKR-hum) was cultivated in a SR1000ODLL bioreactor system connected to the sensor module pH4pO4 for pH and dissolved oxygen (DO) monitoring and control, TC4SC4 module for agitation and temperature control, exhaust gas analyzer GA4, multipump module MP8, and compressed air supplier MX4/4 (DASGIP).

| Quantification of fructose in the supernatant with HPLC
For pH control, the pH was maintained at 6.8 with a 4 M KOH solution. The applied feeding solution composition was as follows: 446.4 g L −1 fructose, 111.6 g L −1 (NH 4 ) 2 SO 4 , and one-time addition of trace elements as in the regular minimal medium. After 7.5 days of cultivation, the feeding solution was replaced by one without (NH 4 ) 2 SO 4 . The reactors were placed into a Bioblock to control the temperature at 30°C. The regulation of the process was performed with DASGIP control, v4.5. The initial settings for gassing were set to six standard liters per hour (sL h −1 ) and the initial agitation speed was set to 400 rpm. The agitation and gassing were regulated to keep the DO signal at 20%. Initial liquid volume was set to 0.3 L. A heterotrophic preculture grown overnight in MM was centrifuged, resuspended in the bioreactor medium and the reactor was inoculated to an optical density (OD) of 0.1.
Cultures were induced at 0.84 ± 0.03 gCDW L −1 after 13-14 h with 0.2% L-rhamnose. At induction time, 20% n-dodecane (75 ml) was added to the culture through a port to remove α-humulene in situ. The feed was started after 14 h with the initial flow rate of 0.75 ml h −1 and was increased throughout the fermentation until a maximal feed of 1.10 ml h −1 . Increase in biomass, product formation, and concentration of carbon and nitrogen source was followed over time. After the bioreactors had been stopped, the total amount of The oxygen transfer rate (OTR) and carbon dioxide transfer rate (CTR) were calculated via the gas balancing method (supplementary information).

| Measurement of transcription levels of the pKR-hum plasmid with RT-qPCR
For analyzing the expression of the pKR-hum genes using RT-qPCR, C. necator cells were harvested 1.5 h after induction (OD600 of~0.8) from shake flasks cultivations in triplicates in minimal medium. As a control, uninduced cells were cultivated in parallel and harvested at the same time. The cells were pelleted by centrifugation at 10,000 × g at 4°C for 5 min and total RNA was isolated immediately with the RNeasy Mini Kit (Qiagen) and stored at −20°C until quantitative reverse-transcription polymerase chain reaction (RT-qPCR) analysis. The primers and details of the RT-qPCR are listed in Table S6.
2.10 | Statistical analysis for the autotrophic α-humulene production tests in the injection flasks An unpaired two-tailed t-test was used to test the null hypothesis that there was no difference between the two conditions. A 99% or 95% confidence interval was used to reject the null hypothesis and thus determine the significance of single variables (p values < 0.01 or < 0.05, indicated in the results part).

| Cloning of M. xanthus genes into E. coli expression vectors
The Myxococcus xanthus genes encoding the MVA-pathway enzymes hmgS, hmgR, mvaK, mvaD, and mvaK2 were amplified by PCR from the plasmid pKR-hum and cloned (ligation independent cloning) into a proprietary backbone vector (pGT1960) containing a medium copy origin of replication (p15A-ori; 15-20 plasmid copies/cell), an ampicillin resistance gene as selection marker and a constitutive promoter driving recombinant gene expression. Details are described in the supplementary information.
2.12 | Testing of MVA genes in lycopene-producing E. coli cells To test the activity of the MVA enzymes, two previously constructed lycopene producing E. coli strains were used: strain BL21.G45 with a lycopene biosynthetic pathway integrated into the genome and BL21.G171 carrying in the genome the lycopene pathway as well as a functional MVA lower portion pathway previously constructed using plant genes.
For testing the hmgS-hmgR enzymes, the plasmids pGT2148 and pGT2184 were transformed into BL21.G171 and for testing the mvaK-mvaD-mvaK2 the plasmids pGT2071 and pGT2225 were transferred into BL21.G45. As a negative control, an empty plasmid without MVA genes was used.
The cells expressing the MVA enzymes were grown in 2 mL FIT-B media (using glucose as a carbon source and supplemented with 100 mg L −1 ampicillin, media composition in Table S8). The cultivation was performed at 28°C and 350 rpm for 48 h, using a 24-Deepwell microtiter plate (Duetz System). The 2 mM mevalonate was added to the media to test the activity of the mvaK-mvaD-mvaK2 enzymes. After the incubation period, the cultures were centrifuged at 14,000 rpm for 2 min and the lycopene was extracted from the cell pellet with acetone. The concentration of lycopene in the acetone extracts was determined by measuring the extinction at 502 nm using a spectrophotometer (Evolution™ 201/220 UV-Vis-Spektrophotometer, Thermo Fisher Scientific).

| RESULTS AND DISCUSSION
3.1 | Theoretical comparison of autotrophic and heterotrophic production of α-humulene in C. necator C. necator can grow on different substrates (e.g., fructose or mixtures of H 2 /O 2 /CO 2 ). To assess which of the systems is currently economically feasible, a model-based comparison was carried out.
For the theoretical evaluation of the autotrophic versus the heterotrophic production of α-humulene with C. necator the target production was set to 10 t a −1 for both production types. As the maximal yield of α-humulene, the maximal respective literature value for P(3)HB production of 2.42 g L −1 h −1 was considered, which when accounted for the same carbon flux, corresponds to a α-humulene production of 1.5 g L −1 h −1 (7.3 mM h −1 ) (Kim et al., 1994). For the autotrophic production, lower optical densities  and therefore lower P(3)HB titers are described. The highest literature value for P(3)HB is 1.55 g L −1 h −1 (Tanaka et al., 1995), which corresponds to the productivity of 1.0 g L −1 h −1 (4.89 mM h −1 ) α-humulene.
For the heterotrophic production, literature values for medium components and setpoints are applied (Table S1). The process is performed in a stirred-tank reactor in a batch mode with 80% of the maximal working volume. For the autotrophic production, an airlift reactor was chosen with 80% of the maximal working volume. Apart from the same process parameters, the use of gasses as carbon and energy sources requires an atmosphere of 16% CO 2 , 64% H 2 , and 20% O 2 and an initial pressure of 1.5 bar (Krieg et al., 2018).
In the simulation of the heterotrophic production with SuperPro For both production modes setting up the next preculture in the preculture reactor directly after inoculation of the main reactor shortens the operation time by 218.5 h, leading to 35 production runs in 1 year and 11 t of α-humulene annually. The main production parameters of the process were calculated and can be found in Table S2.
The heterotrophic process has higher total fixed costs than the autotrophic process due to the use of a stirred-tank reactor which is more complicated in its construction and maintenance than an airlift reactor, which does not contain any stirrer as the stirring is performed with air. Therefore, the payback period is 1.6 times longer for the heterotrophic process (2.13 a) compared with the autotrophic process (1.33 a, Table S2). However, after the payback period, the net gain of the heterotrophic process is higher than the net gain of the autotrophic process due to slightly higher productivity and makes it therefore more profitable in the long run. Nevertheless, both processes are comparable and can be considered industrially relevant. We would like to emphasize that this is only a rough estimation and that more detailed modeling is needed, once the production strain has been optimized.

| Production in a heterotrophic fed-batch system
The α-humulene production process in the DASGIP parallel reactor system was divided in three different phases. In the first phase, the batch phase, the C. necator strain was grown until induction. The second phase was the fed-batch phase with different feed rates (f, Figure 1) of a mixed fructose/ammonium sulfate solution. In the third phase, this feed was replaced with only fructose feed.
The growth of C. necator did not show any lag phase (Figure 1a). The induction was performed 14 h after inoculation. The α-humulene production started directly after induction and the addition of 20% ndodecane as the second phase for in situ product removal (ISPR). There was a correlation between the production of α-humulene and the biomass formation of C. necator during the linear growth phase, which was regulated by the linear feed. After switching the feed, the cell growth did not stop due to the ammonia accumulation in the medium. At the end of the process, more than 1 g L NH 4 + was detected in the fermentation broth. Especially during the feeding with a fructose-only feed, fructose was nondetectable in the fermentation broth. Oxygen limitation did not occur at any time during the fermentation as shown by the OTR signal, since the dissolved oxygen (DO) signal was regulated at 20% by an increase in stirring and gassing (Figure 1b). The biological respiration was controlled by the fructose availability, indicated by an increase in OTR and CTR after an increase in feeding rate of the fructose stock solution.
After 144 h, a decrease in biological activity was detected, indicated by a slow but steady decrease in OTR and CTR. Simultaneously, an increase in foam production was observed, which led to the assumption that increased cell lysis of old cells was observed. In addition, we could observe a formation of interphase between the organic and the aqueous phase containing cell debris. This further confirmed our cell lysis hypothesis. By daily addition of antifoam, the amount of foam in the reactor was reduced to a minimum (spiking not shown). The antifoam had no obvious impact on the cell metabolism, as there was no response in CTR and OTR. The respiratory quotient (RQ) was between 0.9 and 1.0 during the whole process time (Figure 1b). The value means that the consumption of oxygen matched the production of carbon dioxide, which is typical for an aerobic fermentation process on sugars (Bendtsen, 2000). Final During the whole process time, 156 and 160 g fructose were consumed by C. necator (pKR-hum) in reactor 1 and 2, respectively. These amounts per reactor volume correspond to 7857 and 8017 mmol C L −1 .
By measuring the optical density of the culture, the α-humulene concentration, and the carbon dioxide concentration in the exhaust gas stream, a carbon balance was set up. In total, 8952 and 8374 mmol C L −1 were found in the products. The majority was found in carbon dioxide, followed by biomass with approximately 25% of the carbon recovered.
Despite the high titers of 10 mM, only 2% of the total carbon from the substrate source can be found in the target compound α-humulene (Table 1) A recent review compares the product titers of several production systems of different sesquiterpenes (Moser & Pichler, 2019). After E. coli and Saccharomyces cerevisiae, production processes with C. necator are now shown to be able to produce the highest concentrations of sesquiterpenes.
To the best of our knowledge, the achieved α-humulene titers are the highest reported to date. With future improvements of the limitations caused by the expression of the MVA pathway, the titers could be further improved towards an industrially relevant α-humulene production with C. necator. Therefore, we had a closer look at the current MVA performance in C. necator.
3.3 | Production of α-humulene in C. necator with the MVA pathway C. necator carrying the pKR-hum plasmid and variations thereof (Krieg et al., 2018) was cultivated autotrophically in injection flasks to see the differences in α-humulene production between the pKR-hum plasmids and the controls. The full plasmid pKR-hum was compared to the α-humulene synthase deletion mutant pKR-hum ΔzssI, the MVA pathway deletion mutant pKR-hum ΔMVA and an empty plasmid control pKRrha (Table 2).

| 2699
The uninduced vector pKR-hum, the vector without the αhumulene synthase pKR-hum Δ zssI as well as the empty vector pKRha did not produce α-humulene in significant amounts, as ex-  (Table S10). The 16S-RNA sequence (Lawrence et al., 2005) was used as the reference for the normalization of the mRNA signal in the RT-qPCR. The primer sequences can be found in Table S6.
As depicted in Figure 3 (3-hydroxymethylglutaryl-CoA reductase) and mvaD (phosphomevalonate reductase). All genes, however, already were expressed before induction due to a slight leakiness of the pBAD promotor . Therefore, the genes are likely expressed to a higher extend. Thus, the two identified genes hmgR and mvaD were identified as bottlenecks, one being part of the upper and the other part of the lower MVA pathway (the recombinant pathway is shown in Figure S2). If these bottlenecks were to be improved, the next limitation might be the final enzyme in the pathway, the α-humulene synthase.

| Towards optimization of the MVA pathway
In C. necator, the genes hmgR and mvaD from M. xanthus showed rather limited expression levels ( Figure 3) and are likely to limit performance of the implemented MVA pathway.
To check whether this limitation is host-dependent, a series of experiments were conducted in the host E. coli. Moreover, M. xanthus MVA enzymes were evaluated in comparison to alternative candidate enzymes from plants to check for optimization potential.
For this purpose, the M. xanthus genes from plasmid pKR-hum were cloned into a plasmid backbone suitable for expression in E. coli.
Two plasmids were generated, one with the so-called upper portion (UP) genes of the MVA pathway (pGT2148, hmgS, and hmgR) and a second one with the lower portion (LP) genes (pGT2225, mvaK, mvaD, and mvaK2). As α-humulene synthase is poorly active in E. coli ( Figure 2), the functionality of these constructs was tested in a system employing lycopene-producing E. coli-strains to provide an easy detectable terpene product build from the building blocks provided by the MVA pathway ( Figure 4).
For the testing of UP genes, the lycopene-producing E. coli strain BL21.G171 was used, which also carries plant lower portion genes (LP2) integrated into the E. coli genome. The results of this experiment are summarized in Figure 4a. Without expression of the MVA pathway, BL21.G171 was able to produce 1.3 mg gCDW −1 lycopene.
Feeding of this strain with external mevalonate results in an over 400% (to 6.6 mg gCDW −1 ) increase in lycopene due to the presence/ more lycopene than the control strain lacking LP genes or without the addition of mevalonate, whereas LP2 lead to 88% higher product yield upon mevalonate feeding. As proven by the results in Figure 4a, lycopene synthesis capacity/sink strength is not limiting in the chosen test system. Strikingly, the LP2 in this experimental setup showed significantly lower performance than the genome integrated copy of the same LP2 enzyme cassette used in Figure 4a. It has been concluded that the higher recombinant expression of the LP2 genes (due to the higher gene dosage caused by the plasmid system) negatively impacted LP activity. Achieving proper expression levels of these genes is key to gain maximum activity. This fact needs to be considered in future optimization work.

| CONCLUSION
In this study, we provided evidence that C. necator is a very promising host organism for heterotrophic and autotrophic production of terpenes at similar yearly production amounts and costs in silico. With 2 g L −1 in a heterotrophic fed-batch system, the produced α-humulene titers are the highest titers reported up to date. However, the full biosynthetic potential has not been gained yet, as has been shown by the modeling data and the expression data of the MVA pathway. The implementation of the MVA pathway to boost biosynthesis had limited success in improving the product yields. Experimental data show that this is most likely due to the poor expression and poor activity of the hmgR enzyme from M.
xanthus used in these experiments. However, a replacement of the hmgR enzyme would be only one step in the optimization work, that needs to be performed in the future, as other limitations within the upper portion of the MVA revealed. The correct balancing of enzyme expression would be an additional topic to improve the MVA activity. Furthermore, the tested alternative enzymes from plants showed much higher activity and can serve as a basis for future optimization work to unravel the full potential of heterotrophic terpene production in C. necator.

ACKNOWLEDGMENTS
We thank the BMBF for funding this study [grant numbers MES 031A226, BioFlex 031B0347A]. Open access funding enabled and organized by Projekt DEAL.

CONFLICT OF INTERESTS
The authors declare that there are no conflict of interests.