Engineered Rhodobacter capsulatus as a Phototrophic Platform Organism for the Synthesis of Plant Sesquiterpenoids

Sesquiterpenoids are a large class of natural compounds offering manifold properties valuable for food, cosmetics, agriculture, and pharma industry. Production in microorganisms is a sustainable approach to provide sesquiterpenoids for research and industrial use independent of their natural sources. This requires the functional transfer of the respective biocatalytic pathways in an adequate host microorganism offering a sufficient supply of precursors that is ideally adjusted to the individual demand of the recombinant biosynthesis route. The phototrophic purple bacterium Rhodobacter capsulatus offers unique physiological properties that are favorable for biosynthesis of hydrophobic terpenes. Under phototrophic conditions, it develops a large intracytoplasmic membrane suitable for hosting membrane-bound enzymes and metabolites of respective biosynthetic pathways. In addition, Rhodobacter harbors an intrinsic carotenoid biosynthesis that can be engineered toward the production of foreign terpenes. Here, we evaluate R. capsulatus as host for the production of plant sesquiterpenoids under phototrophic conditions using patchoulol and valencene as a proof of concept. The heterologous expression of patchoulol synthase PcPS from Pogostemon cablin as well as the valencene synthases CsVS from Citrus sinensis and CnVS from Callitropsis nootkatensis led to the production of the respective sesquiterpenoids in R. capsulatus. To analyze, if gradually adjustable formation of the key precursor farnesylpyrophosphate (FPP) is beneficial for sesquiterpene synthesis under phototrophic conditions, the intrinsic 1-deoxy-D-xylulose 5-phosphate (DXP) pathway genes as well as the heterologous mevalonate pathway genes were modularly expressed in various combinations. To this end, different plasmids and chromosomally integrated expression tools were developed harboring the strong and tightly controlled Pnif promoter for heterologous gene expression. Notably, comparative studies identified a distinct combination of precursor biosynthetic genes as best-performing setup for each of the tested sesquiterpene synthases. In summary, we could demonstrate that R. capsulatus is a promising alternative platform organism that is suited for sustainable sesquiterpenoid formation under phototrophic cultivation conditions. A modular engineering of R. capsulatus strains via tailored co-expression of FPP biosynthetic genes further allowed adaptation of sesquiterpene precursor formation to its catalytic conversion by different plant terpene synthases.


Supplementary Figure 1. Construction scheme of pRhon5Hi-2-based expression vectors for sesquiterpenoid production in R. capsulatus. (A)
The nifHDK operon of R. capsulatus encodes the molybdenum dependent nitrogenase complex and comprises the structural genes nifH (dinitrogenase reductase), nifD and nifK (subunits of dinitrogenase). To exploit the phototrophic physiology of R. capsulatus for sesquiterpene production, we constructed the new expression plasmid pRhon5Hi-2 that carries the nifHDK-promoter region (P nif ): Dashed lines illustrate the genomic region which was  inserted as an NheI/XbaI fragment (NCBI Genbank Accession MG208548) into the respective sites of vector pRhotHi-2 (Katzke et al., 2010;doi:10.1016/j.pep.2009.08.008). As previously described, the broad-host range expression vector harbors two antibiotic resistance genes (chloramphenicol and kanamycin), an origin of replication (REP) and an origin of transfer (MOB). Target genes can be integrated into the multiple cloning site and are thereby placed under control of the Pnif promoter. fdxD: Fe 2 S 2 ferredoxin. (B) The scheme illustrates cloning steps for the example of patchoulol synthase PcPS from Pogostemon cablin. The synthase encoding gene was obtained as synthetic DNA fragment, flanked by recognition sequences for NdeI/HindIII, on a vector from Eurofins Genomics (pEX-K4-PcPS). The PcPS gene was isolated from the vector by use of NdeI/HindIII and ligated into likewise hydrolyzed expression vector pRhon5Hi-2 to yield pRhon5Hi-2-PcPS (a). Vectors pRhon5Hi-2-PcPS-ispA, pRhon5Hi-2-PcPS-dxs, pRhon5Hi-2-PcPS-idi and pRhon5Hi-2-PcPS-MVA were constructed by PCR-amplification of the respective genes with added recognition sequences for HindIII and XhoI in the primers, hydrolysis with these enzymes, and ligation of the fragments into likewise hydrolyzed pRhon5Hi-2-PcPS (b either with ammonium (NH 4 + ; P nif repressing conditions), or serine, or dinitrogen (Ser, N 2 ; P nif derepressing conditions), as sole nitrogen sources in the medium (inset in upper panel). Figure S3. Oxygen-dependent control of intrinsic terpene formation in R. capsulatus. Carotenoid formation was used as a measure to inspect the oxygen-sensitive intrinsic isoprenoid metabolism of R. capsulatus under different growth conditions. Control cultures were cultivated under standard photoheterotrophic anaerobic conditions. (A) R. capsulatus wildtype SB1003 was pre-cultivated in RCV medium with 0.1% ammonium, before test cultures were started with an OD 660nm of 0.05 under aerobic conditions in 100 mL unbaffeled shake flasks (+O 2 ) with different filling volumes of RCV medium supplemented with 0.1% serine. After 48 h, cells equivalent to OD 660nm = 1 were harvested and extracted with ethanol. Cell debris was pelleted and the supernatant containing cellular pigments was used for recording absorbance spectra from 300 to 900 nm in 1 nm intervals. Samples of phototrophically cultivated cells (-O 2 ) exhibited typical absorption maxima of bacteriochlorophyll a (BChla; 368, 600, and 770 nm), as well as characteristic maxima of the carotenoid spheroidene (Sph; 430, 456, and 487 nm). Under (micro)aerobic conditions, the same bacteriochlorophyll a-related absorption was detected (at lower levels), and the carotenoid-specific absorption of spheroidenone (Sph-OH; max. ~480 nm) was found, as expected. (B) Plotting the absorption at the wavelength of 478 nm in all samples showed that the pigment formation increased in +O 2 cultures with higher filling volumes up to 60 mL. Under this condition, about 65% of carotenoid-dependent absorption was reached compared to the photoheterotrophic anaerobic control. Data represent mean values and respective standard deviations from three independent cultivations.  into a Trace GC Ultra gas chromatograph coupled to ITQ 900 mass spectrometer (Thermo Scientific). Separation was achieved in a 30 m × 0.25 mm diameter capillary, with a 0.25 µm film of FS-5 supreme (CS Chromatographie Service). Split mode with a split ratio of 10 was used for the injector with the inlet temperature set to 250 °C. The oven was programmed to start at 100 °C and a 1 min hold, after which temperature increased to 300 °C at a rate of 5 °C/min. Helium was used as carrier gas and was adjusted to a flow rate of 1 mL/min. MS data were collected from 50 to 300 m/z during the temperature ramp.

Analysis of n-dodecane-mediated sesquiterpenoid extraction from phototrophically grown R. capsulatus.
Usually, sesquiterpenoids are extracted from microbial cell cultures via an n-dodecane layer (1/30 of the culture volume) which is added prior to cultivation and acts as an organic solvent phase (Rodriguez et al., 2014;doi:10.1038/nprot.2014). In the here presented work, we used the photosynthetic bacterium R. capsulatus as alternative sesquiterpenoid production host. After phototrophic cultivation, the sealed Hungate tubes were shaken horizontally at 30 °C and 130 rpm overnight in the dark in a Multitron Standard incubation shaker (Infors HT) to facilitate sesquiterpenoid extraction into the organic phase. Subsequently, 100 µL n-dodecane samples were subjected to gas chromatographic (GC) analysis as described in the Materials and Methods section.
To quantify the final product titers, calibration curves with authentic references (-)-patchoulol and (+)-valencene were used. However, the mere correlation of signals from n-dodecane extracted samples with the reference signals does not take into account extraction efficiencies of individual sesquiterpenoids when using n-dodecane as organic solvent. It can be assumed that, in dependence of their specific properties, different sesquiterpenoids only diffuse to a certain extent into the n-dodecane layer. In addition, terpenes that are produced in the cytoplasm of R. capsulatus can additionally be retained by the intracytoplasmic membrane system thereby further affecting the transfer into the organic phase. Therefore, we first determined the transfer efficiency of valencene and patchoulol from cultivation medium into n-dodecane in the presence of intact and disrupted Rhodobacter cells. For this purpose, the respective reference compounds were first mixed with 14 mL of phototrophically grown R. capsulatus SB1003 cells (cultivation parameters: anaerobic growth, 30 °C, approx. up to OD 660nm = 2.5) in appropriate amounts (giving signal intensities comparable to samples from R. capsulatus production cultures; patchoulol: 15 mg/L; valencene: 5.71 mg/L). For this purpose, 130.4 µL valencene, which is an oil, was added as a 10-fold dilution in diethyl ether, while the solid patchoulol had to be solved in diethyl ether prior to use gaining a 2 mg/mL stock solution of which 110.5 µL was added. After addition, the cultures were sealed and vortexed for 1 min. Subsequently, reference substances were extracted using n-dodecane as described above. The transfer efficiency was determined via GC analysis by comparing peak areas of the specific signals from appropriately diluted solutions to samples that had undergone extraction (Supplementary Figure S5). Figure  S5: Transfer efficiency of the patchoulol (black bar) and valencene (grey bar) reference compounds from cultivation medium into the n-dodecane phase in the presence of intact R. capsulatus cells. For extraction, 15 mg/L patchoulol or 5.71 mg/L valencene were added to 14 mL cell cultures (OD 660nm = 2.5). For details, see text above. Data represent means and standard deviations of three independent measurements (n = 3).

Supplementary
By using intact R. capsulatus cells, a transfer efficiency of 54% (patchoulol) and 73% (valencene) could be determined. Thus, it could be shown that there are some methodological losses, which have to be considered for product quantification. To moreover analyze if putative interaction of intracellularly produced sesquiterpenoids with the Rhodobacter ICM can decrease product transfer, the experiment was repeated using disrupted cells. For this, equally cultivated R. capsulatus wildtype cells (OD 660nm = 2.5) were disrupted using a ball mill (3 x 10 min, 30 Hz, Mixer Mill MM 400, Retsch GmbH, Germany) and subsequently mixed with the same amount of reference compound as described previously. Extraction and quantification was performed as described for intact cell samples and signals were subsequently compared to those of the non-extracted reference compounds (Supplementary Figure S6). Figure  S6: Transfer efficiency of the patchoulol (black bar) and valencene (grey bar) reference compounds from cultivation medium into the n-dodecane phase in the presence of disrupted R. capsulatus cells. For extraction, 15 mg/L patchoulol or 5.71 mg/L valencene were added to 14 mL cell lysate (OD 660nm = 2.5). For further details, see text above. Data represent means and standard deviations of three independent measurements (n = 3).

Supplementary
For patchoulol, no significant decrease of the transfer efficiency could be observed for lysed cells (55%) in comparison to the previous measurement using intact cells (54%). In contrast, a strong decrease of transfer efficiency could be detected for valencene (only 40%, in comparison to 73% when intact cells were used), suggesting that this more hydrophobic terpenoid (logP = 5.86 in comparison to patchoulol with a logP of 4.19; values were calculated using the ALOGPS2.1 online tool described by Tetko et al. 2005;doi:10.1007/s10822-005-8694-y) can be retained more efficiently by the intracytoplasmic membrane system. Hence, for calculating the final production titers, individual transfer efficiencies for disrupted cell cultures (here termed c t 'transfer efficiency coefficient'; patchoulol: 1.4521; valencene: 1.6) were taken into account.
Besides the above described negative effect of cellular components on the extraction efficiency, we further analyzed, if repeated n-dodecane-dependent sesquiterpenoid extraction should be considered for an optimal estimation of the overall production titers. Thus, an experiment with repeated sesquiterpenoid extractions from disrupted wildtype cultures that were mixed with reference compounds as described above was performed over four days (Supplementary Figure S7). For quantitative analysis of sesquiterpenoids, calibration curves with the authentic references of (-)-patchoulol and (+)-valencene ranging from 0.25 to 2 mg/500 µL n-dodecane, were used (slope: 380.37 and 475.37, respectively; see also depicted below in Supplementary Figure S9). Figure S7: Extraction efficiency of the patchoulol (black bars) and valencene (grey bars) reference compounds from cultivation medium in the presence of disrupted R. capsulatus cells by repeatedly using n-dodecane as organic solvent over four days. See text above for details. For repeated extraction, 15 mg/L patchoulol or 5.71 mg/L valencene were added to 14 mL cell lysate (OD 660nm = 2.5).

Supplementary
Subsequently, 500 µL n-dodecane was used for 24 h over a time period of four days. Single extraction procedures were repeated four times and the sesquiterpenoid concentration of each fraction was analyzed via GC. Data represent means and standard deviations of three independent measurements (n = 3).
For patchoulol, the overall extraction efficiency was increased up to 32% by repeated extraction in comparison to the amount determined after the first extraction. For valencene, extraction efficiency increased even more (up to 157%). Therefore, we also took into account these factors by which the quantification on day 1 underestimates product titers (here termed c ex 'coefficient for repeated extraction'; patchoulol: 1.3165; valencene: 2.5732 in order to calculate the final product titers. Finally, we analyzed if the presence of an n-dodecane layer can positively or negatively affect sesquiterpene formation in R. capsulatus cells during cultivation. Therefore, an experiment with repeated sesquiterpene extraction out of production cultures that had been cultivated (5 days) with and without an n-dodecane layer before extraction was performed over four days (Supplementary Figure S8). For the analysis of sesquiterpene producing R. capsulatus cultures grown without a

8) (S3) Repeated DC extraction +Ref PAPER
patch valen solvent layer, equal amounts of n-dodecane were added after cultivation and prior to the extraction procedure.
Supplementary Figure S8: Comparison of relative patchoulol and valencene formation in R. capsulatus production strains cultivated with (black bars) and without an n-dodecane layer (grey bars). Data was normalized to the amount of sesquiterpene extracted from cultures with n-dodecane. See text above for details. For R. capsulatus cultures containing the n-dodecane layer, 500 µL of the solvent was added before cultivation. In contrast, the same amount of solvent was added to cultures without n-dodecane after the cultivation. Compounds were extracted and analyzed as described above. Data represent means and standard deviations of three independent measurements (n = 3).
Almost no changes of sesquiterpene formation could be observed in the absence of the n-dodecane layer. Remarkably, production titers for valencene even increased slightly without using the organic solvent. Hence, the n-dodecane layer can alternatively be added after cultivation of the Rhodobacter production strains prior to the extraction procedure without any product losses.
In summary, product titers of R. capsulatus sesquiterpenoid production cultures were determined by analysis of n-dodecane extraction samples from disrupted cells. To this end, R. capsulatus strains were cultivated without the solvent, disrupted and then extracted one time with n-dodecane. Using the calibration curves obtained with reference compounds (Supplementary Figure S9) and taking into account above described results on losses of this procedure (Supplementary Figure S6 and S7), we used the following equation for calculating the final patchoulol and valencene titers: