ABSTRACT
Pseudomonas putida is a saprophytic bacterium with robust carbon metabolisms and strong solvent tolerance making it an attractive host for metabolic engineering and bioremediation. Due to its diverse carbon metabolisms, its genome encodes an array of proteins and enzymes that can be readily applied to produce valuable products. In this work we sought to identify design principles and bottlenecks in the production of type III polyketide synthase (T3PKS)-derived compounds in P. putida. T3PKS products are widely used as nutraceuticals and medicines and often require aromatic starter units, such as coumaroyl-CoA, which is also an intermediate in the native coumarate catabolic pathway of P. putida. Using a randomly barcoded transposon mutant (RB-TnSeq) library, we assayed gene functions for a large portion of aromatic catabolism, confirmed known pathways, and proposed new annotations for two aromatic transporters. The tetrahydroxynapthalene synthase of Streptomyces coelicolor (RppA), a microbial T3PKS, was then used to rapidly assay growth conditions for increased T3PKS product accumulation. The feruloyl/coumaroyl CoA synthetase (Fcs) of P. putida was used to supply coumaroyl-CoA for the curcuminoid synthase (CUS) of Oryza sativa, a plant T3PKS. We identified that accumulation of coumaroyl-CoA in this pathway results in extended growth lag times in P. putida. Deletion of the second step in coumarate catabolism, the enoyl-CoA hydratase lyase (Ech), resulted in ‘drop-in’ production of the type III polyketide bisdemethoxycurcumin.
1 INTRODUCTION
Secondary metabolites of fungi, plants, and bacteria have been used as medicines and supplements for millennia [1]. Compounds such as naringenin, raspberry ketone, resveratrol, and curcumin are widely used as nutraceuticals and are biosynthesized through similar pathways [2,3]. Commercially, these chemicals are either extracted directly from plants or produced synthetically, as in the case of raspberry ketone [4]. Renewable microbial production of these compounds will decrease reliance on agriculture and fossil fuel-derived chemical synthesis. The biosynthesis of these compounds (naringenin, raspberry ketone, resveratrol, and curcumin) relies on a class of enzymes called type III polyketide synthases (T3PKSs). T3PKSs carry out iterative Claisen condensation reactions typically with coenzyme A (CoA)-based starter and extender units, both of which vary widely among these enzymes [5]. In the case of the tetrahydroxynapthalene synthase of Streptomyces coelicolor (RppA) the starter and extender units are simply malonyl-CoA, while in many plant T3PKSs the starter unit is a phenylpropenoyl-CoA thioester, usually derived from ferulate, coumarate, or cinnamate [6,7].
Coumarate and ferulate are components of lignin found in lignocellulosic hydrolysate (LH), which has been proposed for use as a renewable feedstock for biocatalysis [8,9]. However, the limited capabilities of commonly used model organisms, such as Escherichia coli, have hindered progress toward making LH a viable feedstock [10]. Characteristics such as high solvent tolerance and diverse carbon metabolisms are lacking in many E. coli strains, but are essential for a microbe to be used in LH valorization. With this in mind, it has become apparent that progress needs to be made in developing microbes with these robust traits. Among these emergent organisms is the saprophytic bacterium Pseudomonas putida. P. putida has been studied for its ability to catabolize phenolics and withstand solvents, making it an attractive host for LH valorization [11–13].
In this work we sought to leverage the native catabolism of P. putida KT2440 for use in the biosynthesis of a plant T3PKS product from coumarate. RppA of S. coelicolor was first expressed to determine optimal production conditions for T3PKS product accumulation. Next we sought to identify alternative catabolic pathways that could result in reduced titers. Using a randomly barcoded transposon mutant (RB-TnSeq) library of P. putida, we assayed for genes involved in the catabolism of coumarate and seven other related aromatic compounds. The native feruloyl/coumaroyl-CoA synthetase (Fcs) of P. putida was then used to produce the coumaroyl-CoA starter unit to the curcuminoid synthase (CUS) of Oryza sativa. In the process, we found that accumulation of coumaroyl-CoA was toxic to P. putida. Finally, relying on the native expression levels of fcs from the chromosome and plasmid-based expression of CUS resulted in drop-in production of bisdemethoxycurcumin (BDC) from coumarate (Figure 1).
2 RESULTS
2.1 RppA as a screen for optimal T3PKS production conditions
The tetrahydroxynaphthalene synthase of Streptoyces coelicolor was recently applied as a biosensor for malonyl-CoA concentrations in several hosts including P. putida [14]. We sought to utilize this system to rapidly assay growth conditions for increased T3PKS product accumulation. RppA was initially codon optimized for expression in P. putida and cloned into the construct pBADT-rppA-OW. We then constructed two more vectors, one with the complete rppA cloned from the S. coelicolor genome (pBADT-rppA-NW) and another with a 5’ truncation of 75 base pairs (pBADT-rppA-NT) as this had been described previously to increase enzymatic activity [6].
Since the product of RppA, 1,3,6,8-tetrahydroxynaphthalene, spontaneously oxidizes to the red pigment flaviolin, we used a colorimetric assay to determine how glucose concentrations affect the production of T3PKS products (Figure 2a). Concentrations of glucose ranging from 0 - 100 mM supplemented into LB medium were tested, and the production of flaviolin was measured by absorbance at 340 nm, as previously reported [14]. Unexpectedly, the rppA expression vectors using the native S. coelicolor codons produced more flaviolin than the P. putida codon optimized variant (Figure 2b). In all three constructs tested, there was an increase in the accumulation of flaviolin in cultures containing greater than 25 mM glucose (Figure 2b). Given this result, production experiments for bisdemethoxycurcumin were conducted in media supplemented with 100 mM glucose.
2.2 Functional genomics to validate aromatic catabolisms of P. putida
While aromatic catabolism has been extensively studied in P. putida, genes implicated in these pathways are still being discovered [15]. The genes involved in the first steps of coumarate catabolism reside in an operon with a putative acyl-CoA dehydrogenase (PP_3354) and a putative beta-ketothiolase (PP_3355), which have been proposed to be involved in an alternative catabolic pathway [16]. As functional redundancy in coumarate catabolism could result in loss of the type III polyketide precursor, coumaroyl-CoA, we sought to identify any pathways that could potentially impact product titers. To assay for genes involved in coumarate and related metabolisms, we grew a randomly barcoded transposon mutant (RB-TnSeq) library of P. putida KT2440 in minimal medium with a variety of different aromatic compounds often found in LH (p-coumarate, ferulate, benzoate, p-hydroxybenzoate, protocatechuate, vanillin, vanillate, phenylacetate) and glucose as sole carbon sources. The fitness of each gene was calculated by comparing the abundance of barcodes before versus after growth selection, using barcode sequencing (BarSeq) [17,18]. Negative values indicate that the gene was important for growth in that condition.
The results of the RB-TnSeq assay indicate that the primary route for ferulate and coumarate catabolism is through the feruloyl/coumaroyl-CoA synthetase (Fcs) and the enoyl-CoA hydratase lyase (Ech) (Figure 3 and Figure S1). The genes in the proposed secondary pathway of coumarate and ferulate catabolism, PP_3354 and PP_3355, had no significant fitness phenotype, indicating that these genes are not necessary for coumarate or ferulate catabolism [19].
Of all the known reactions depicted in the map of aromatic catabolism (Figure 3), we observed significant negative fitness values for all but three of their corresponding genes: catA-1, catA-2, and pcaC. In the case of catA-1 and catA-2 we believe the cause for no fitness phenotype is the functional redundancy of these two genes, i.e. a cell carrying a transposon insertion in one will still have the other gene present. In the case of pcaC, we noticed that there was a strong phenotype (-log2 < −2.0), but the significance fell below our cutoff in all conditions tested (|tscore| < 4.0). This is likely due to the low frequency of transposon insertions into this gene in the library (n=4). While our results heavily support the current models of aromatic metabolism in P. putida, our data also indicated that some gene annotations should be revised. PP_3272 is currently annotated as encoding an acetate permease but given this data and its homology to other systems [20], it should be reannotated as the phenylacetate transporter (phaJ). The PP_1376 gene is annotated as encoding a 4-hydroxybenzoate transporter; however, we only observed a fitness detriment for this gene with protocatechuate as the sole carbon source.
Because of this, PP_1376 should be reannotated as a protocatechuate transporter.
2.3 Accumulation of coumaroyl-CoA is toxic to P. putida
The first step in the biosynthetic pathway for bisdemethoxycurcumin is the activation of coumarate with coenzyme A (CoASH) (Figure 1). Because Pseudomonas putida KT2440 natively produces coumaroyl-CoA during coumarate catabolism, we knocked out the subsequent gene in the native catabolic pathway, ech, to prevent P. putida from consuming this necessary precursor (Figure 1) [20]. Initial production experiments in ∆ech strains overexpressing fcs and CUS from a synthetic operon resulted in an extended lag phase (data not shown). To determine the cause, we overexpressed fcs alone under control of the arabinose-inducible araBAD promoter (PBAD) in the presence and absence of coumarate (Figure 4A). Increasing the inducer concentration resulted in increased lag times only when coumarate was present in the medium (Figure 4B). This suggested that the coumaroyl-CoA intermediate is toxic to P. putida. We therefore sought to minimize the burden of this intermediate by relying on the native chromosomal expression of fcs in future experiments.
2.4 Production of bisdemethoxycurcumin
In order to produce bisdemethoxycurcumin in P. putida, we employed the biochemical pathway outlined in Figure 1. Exogenously added coumarate is activated by the native feruloyl/coumaroyl-CoA synthetase of P. putida (Fcs), then two resultant coumaroyl-CoA molecules are condensed with malonyl-CoA by the curcuminoid synthase of O. sativa (CUS) to yield bisdemethoxycurcumin [21]. The CoA synthetase, fcs, was expressed from its native chromosomal locus, while CUS was expressed from the pBADT plasmid and induced with L-arabinose [22].
Our initial production strategy was to induce CUS until the cultures reached stationary phase. Then the cultures were pelleted and resuspended in fresh LB medium supplemented with 5 mM coumarate. These samples were then incubated for another 72 hours. A similar approach had been used successfully in E. coli [23]; however, our titers were less than 0.5 mg/L (Figure S2). Given that bisdemethoxycurcumin is insoluble in water [24], a 10 % v/v oleyl alcohol overlay was used to extract the product as the fermentation progressed. Production levels were low (approximately 0.1 mg/L) in P. putida ∆ech when the medium was supplemented with 10 mM coumarate, likely due to the toxicity of coumaroyl-CoA (Figure 5). We therefore lowered the concentration of externally-supplied coumarate in an attempt to decrease precursor toxicity. Supplementation with 5 mM coumarate resulted in a ~5-fold increase in bisdemethoxycurcumin titers in the P. putida ∆ech strain relative to wild-type (Figure 4A).
3. DISCUSSION
Pseudomonas putida is among the most well studied saprophytic bacteria. Its diverse metabolisms enable it to catabolize a wide variety of complex carbon sources, including lignocellulosic hydrolysate [25]. The robust catabolic pathways of P. putida, while useful for producing valuable molecules from diverse carbon sources, can also serve as an obstacle to achieving high product titers as it can often metabolize the desired products [26]. Given recent advances in gene editing techniques [27–31] and our ability to rapidly assay for gene function with transposon site sequencing [17,18,32], engineering non-model hosts like P. putida for industrial applications has become less challenging.
Using RB-TnSeq mutant fitness assays we were able to rapidly confirm entire pathways of aromatic catabolism (Figure 3 and S1). The genes downstream of fcs, previously described as a possible alternative route for coumarate/ferulate catabolism [16,20], showed no significant fitness detriment on any of the carbon sources tested. These genes may be structural remnants of a β-oxidation pathway that eventually evolved into the coumarate/ferulate pathway requiring fcs, ech, and vdh [33]. We then revised the annotations of two genes required for transport of the aromatic compounds, phenylacetate and protocatechuate, an important plant hormone and lignin metabolite respectively [34]. There is a large amount of information in these data about regulatory and structural genetic elements that could be useful to engineers and biologists.
Heterologous expression of bacterial T3PKSs, including the tetrahydroxynapthalene synthase of S. coelicolor (RppA), has previously been demonstrated in P. putida KT2440 [14,35,36]. Using variants of rppA, we were able to rapidly screen for optimal T3 polyketide production conditions. Expressing the codon optimized variant we created in this study, rppA-OW, resulted in less flaviolin production than the native codon variants rppA-NW and rppA-NT. It is possible that there are some factors affecting heterologous protein expression that are not sufficiently accounted for in current codon optimization algorithms [37]. However, we demonstrated in all our constructs that increasing the glucose concentration had a considerable effect on the production of flaviolin (Figure 2). These T3PKS “sensors” have broad utility in both rapidly assaying culture conditions, as described here, and as high-throughput screens of genetic libraries for increased malonyl-CoA accumulation [14]. Future work could employ these sensors to screen for increased intracellular malonyl-CoA concentrations from a complex substrate like LH.
To provide the coumaroyl-CoA substrate for the bisdemethoxycurucumin T3PKS, CUS, we sought to use the native CoA synthetase (Fcs) of P. putida. Plasmid-based induction of fcs expression in the presence of coumarate, however, resulted in an increase in lag times due to the build up of the toxic coumaroyl-CoA intermediate. This toxicity has been observed in Acinetobacter baylyi following disruptions in the gene encoding its enoyl-CoA hydratase lyase and in E. coli expressing the A. baylyi fcs homolog in the presence of coumarate [38]. This toxicity could also have been limiting other systems using bacterial coumaroyl-CoA synthetases, but the defective growth phenotype may not have been observed due to differences in experimental design [39,40]. The exact cause for coumaroyl-CoA toxicity is unclear and will be the subject of future investigations, as it is likely to interfere with other biosynthetic pathways requiring this intermediate.
To engineer P. putida for bisdemethoxycurcumin production, we deleted the native enoyl-CoA hydratase lyase (ech) responsible for the conversion of coumaroyl-CoA to acetyl-CoA and p-hydroxybenzaldehyde. In order to relieve coumaroyl-CoA toxicity, we relied on the native genomic copy of fcs instead of a plasmid-based system. We demonstrated that native expression of fcs generates a sufficient coumaroyl-CoA pool for the curcuminoid synthase (CUS). Extraction of the product during growth using an oleyl alcohol overlay also significantly enhanced titers (Figure S2 and 5). In the final P. putida production strain, we achieved ‘drop-in’ production of bisdemethoxycurcumin at titers of 2.15 mg/L (Figure 5).
This work is a significant first step towards the production of plant T3PKS-derived compounds in P. putida, but several issues remain to be addressed. The use of an overlay and the exogenous addition of coumarate are not feasible for an industrial process. Recent in vitro and in vivo experiments suggest that malonyl-CoA is a limiting factor in plant T3PKS pathways [41,42]. Titers could be improved by modulating this substrate pool. This approach has been successfully used to optimize titers of the T3PKS product naringenin, resulting in titers of 191.9 mg/L [35]. Future work could employ a similar strategy in P. putida to achieve higher titer production of plant T3PKS products. As we continue to understand more about its metabolism and engineer it more effectively, P. putida will become a more attractive host for renewable chemical biosynthesis.
4. METHODS
4.1 Media, chemicals, and culture conditions
General E. coli cultures were grown in Luria-Bertani (LB) Miller medium (BD Biosciences, USA) at 37 °C, while P. putida was grown at 30 °C. MOPS minimal media was used where indicated and comprised of the following: 32.5 μM CaCl2, 0.29 mM K2SO4, 1.32 mM K2HPO4, 8 μM FeCl2, 40 mM MOPS, 4 mM tricine, 0.01 mM FeSO4, 9.52 mM NH4Cl, 0.52 mM MgCl2, 50 mM NaCl, 0.03 μM (NH4)6Mo7O24, 4 μM H3BO3, 0.3 μM CoCl2, 0.1 μM CuSO4, 0.8 μM MnCl2, and 0.1 μM ZnSO4 [43]. Cultures were supplemented with kanamycin (50 mg/L, Sigma Aldrich, USA) when indicated. Technical grade oleyl alcohol was acquired from Alfa Aesar (Alfa Aesar, Thermo Fisher Scientific). All other compounds were purchased through Sigma Aldrich (Sigma Aldrich, USA). Construction of P. putida deletion mutants was performed as described previously [44].
4.2 Strains and plasmids
Bacterial strains and plasmids used in this work are listed in Table 1. All strains and plasmids created in this work are available through the public instance of the JBEI registry https://public-registry.jbei.org. All plasmids were designed using Device Editor and Vector Editor software, while all primers used for the construction of plasmids were designed using j5 software [45–47]. Plasmids were assembled via Gibson Assembly using standard protocols [48], or Golden Gate Assembly using standard protocols [49]. Plasmids were routinely isolated using the Qiaprep Spin Miniprep kit (Qiagen, USA), and all primers were purchased from Integrated DNA Technologies (IDT, Coralville, IA).
4.3 Plate based growth assays
Growth studies of bacterial strains were conducted using microplate reader kinetic assays. Overnight cultures were inoculated into 10 mL of LB medium from single colonies, and grown at 30 °C. These cultures were then diluted 1:100 into 500 μL of LB medium with appropriate concentrations of arabinose and p-coumarate in 48-well plates (Falcon, 353072). Plates were sealed with a gas-permeable microplate adhesive film (VWR, USA), and then optical density and fluorescence were monitored for 48 hours in an Biotek Synergy 4 plate reader (BioTek, USA) at 30 °C with fast continuous shaking. Optical density was measured at 600 nm.
4.4 HPLC detection of bisdemethoxycurcumin
HPLC analysis was performed on an Agilent Technologies 1200 series liquid chromatography instrument coupled to a Diode Array Detector (Agilent Technologies, USA). Compounds were separated at a constant flow rate of 0.4 mL/min over a Kinetex C18 column (2.6 μm diameter, 100Åparticle size, dimensions 100 x 3.00 mm, Phenomenex, USA) held at 50 °C. The mobile phase consisted of H2O + 0.1% trifluoroacetic acid (A) and acetonitrile + 0.1% trifluoroacetic acid (B). Separation was performed using the following gradient method: 0-3 minutes 95% A, 3-15 minutes 95-5% A, 15-17 minutes 5% A, 17-17.5 minutes 5-95% A, 17.5-20 minutes 95% A. The presence of bisdemethoxycurcumin was monitored and quantified at 440 nm.
4.5 RB-TnSeq experiments and analysis
BarSeq-based experiments utilized the P. putida RB-TnSeq library, JBEI-1, which has been described previously [44]. An aliquot of JBEI-1 was thawed on ice, diluted into 25 mL of LB medium supplemented with kanamycin and grown to an OD600 of 0.5 at 30 °C. Three 1 mL aliquots of the library were pelleted and stored at −80 °C to later serve as the t0 of gene abundance. Libraries were then washed in MOPS minimal medium, and diluted 1:50 in MOPS minimal medium with 10 mM p-coumarate, ferulate, benzoate, p-hydroxybenzoate, protocatechuate, vanillin, vanillate, phenylacetate, or D-glucose. Cells were grown in 600 μL of medium in 96-well deep well plates (VWR). Plates were sealed with a gas-permeable microplate adhesive film (VWR, USA), and then grown at 30 °C in an INFORS HT Multitron (Infors USA Inc.), with shaking at 700 rpm. Two 600-μL samples were combined, pelleted, and stored at −80 °C until analysis by BarSeq, which was performed as previously described [17,18]. All fitness data is publically available at http://fit.genomics.lbl.gov.
4.5 Curcuminoid production
For production of bisdemthoxycurcumin without an overlay, an overnight culture of P. putida KT2440 ∆ech + pBADT-CUS was diluted 1:100 into 5 mL of LB supplemented with 50 mg/L kanamycin, 1 % w/v L-arabinose, and 100 mM glucose. The culture was grown to stationary phase over 12 hours then pelleted in a centrifuge at 5000 xg for 5 minutes. The cell pellets were resuspended in 2 mL of fresh LB with 50 mg/L kanamycin, 0.5 % L-arabinose, 100 mM glucose, and 5 mM coumarate. The culture was allowed to proceed for 72 hours.
Overnights of P. putida harboring pBADT-CUS were diluted 1:100 into 25 mL of fresh LB supplemented with 50 mg/L kanamycin, and 100 mM glucose. Arabinose and coumarate were added at the beginning of the fermentation at concentrations indicated. A 2.5 mL overlay of oleyl alcohol was added to extract the bisdemethoxycurcumin during growth. The fermentation was allowed to proceed for 72 hours.
4.6 Curcuminoid extraction
For cultures lacking an overlay, 0.5 mL of culture was acidified to pH 3 with 3 N HCl. Bisdemethoxycurcumin was then extracted with an equal volume of ethyl acetate. 250 μL of the ethyl acetate layer was removed and the solvent was allowed to evaporate overnight. The dried samples were then resuspended in 50 μL of acetonitrile for analysis with HPLC-DAD.
For cultures with an overlay, the cultures were acidified to pH 3 with 3 N HCl. Acidified cultures were then pelleted in a centrifuge and the oleyl alcohol overlays were extracted. To quantify bisdemethoxycurcumin, 100 μL of the extracted overlays were added to a black, clear bottom 96-well plate and absorbance was measured at 425 nm in a Biotek Synergy 4 plate reader (BioTek, USA). A standard curve was made with bisdemethoxycurcumin standards dissolved in oleyl alcohol (Figure S3)
4.7 Flaviolin production
Colonies of P. putida KT2440 strains carrying pBADT-rppA-NW, pBADT-rppA-NT, and pBADT-rppA-OW were used to inoculate LB medium with 50 mg/L kanamycin and cultured overnight. The overnight culture was then diluted 1/100 into fresh LB with 0.2 % w/v L-arabinose, 50 mg/L kanamycin, and 100, 50, 25, 12.5, 6.25, 3.125, 1.5625 or 0 mM glucose. Cultures were conducted in 24-well deep-well plates, and allowed to proceed for 48 hours. Cultures were pelleted in a centrifuge at 5000 xg for 5 minutes. Supernatants were removed, aliquoted into a 96-well black clear bottom plate, and the absorbance was measured at 340 nm in a Biotek Synergy 4 plate reader (BioTek, USA). For cultures expressing rppA-NW and rppA-NT, the supernatants were too opaque with red product and needed to be diluted 1:2 in fresh LB for accurate absorbance measurements.
Contributions
Conceptualization, M.R.I., M.G.T., J.M.B; Methodology, M.R.I., M.G.T., J.M.B.; Investigation, M.R.I., M.G.T., J.M.B., M.S., A.N.P.; Writing – Original Draft, M.R.I.; Writing – Review and Editing, All authors.; Resources and supervision, A.M.D., J.D.K.
Competing Interests
J.D.K. has financial interests in Amyris, Lygos, Demetrix, Napigen and Maple Bio
Acknowledgements
We would like to thank Professor Mattheos Koffas for providing us with the plasmid pEMT6-4CL-CUS. We thank Jesus Barajas for his careful reading of this manuscript and helpful suggestions during preparation. We also thank Morgan Price for assistance in analyzing RB-TnSeq data.
This work was part of the DOE Joint BioEnergy Institute (https://www.jbei.org) supported by the U. S. Department of Energy, Office of Science, Office of Biological and Environmental Research, supported by the U.S. Department of Energy, Energy Efficiency and Renewable Energy, Bioenergy Technologies Office, through contract DE-AC02-05CH11231 between Lawrence Berkeley National Laboratory and the U.S. Department of Energy. The views and opinions of the authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, expressed or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan)