Production of fatty acids in Ralstonia eutropha H16 by engineering β-oxidation and carbon storage

Bacterial growth and induction

Strains and plasmids used in this study are listed in Table S1. E. coli strains were grown at 37 °C in lysogeny broth (LB) supplemented with 50 µg/ml kanamycin as appropriate. R. eutropha strains were grown at 30 °C in rich medium (16 g/L nutrient broth, 10 g/L yeast extract, 5 g/L (NH₄)₂SO₄) + 10 µg/ml gentamicin, and supplemented with 200 µg/ml kanamycin as appropriate. Rich medium plates were prepared from the same media plus 1.5% agar (BD). Lauric acid minimal media plates were produced with 1.5% agar and the following salts: 4.0 g/L NaH₂PO₄, 4.6 g/L Na₂HPO₄, 0.45 g/L K₂SO₄, 0.39 g/L MgSO₄, 0.062 g/L CaCl₂, 0.05% (w/v) NH₄Cl, 0.1% lauric acid, 1% NP40 and 1 ml/L of a trace metal solution (15 g/L FeSO₄ ⋅ 7H₂O, 2.4 g/L MnSO₄ ⋅ H₂O, 2.4 g/L ZnSO₄ ⋅ 7H₂O, and 0.48 g/L CuSO₄ ⋅ 5H₂O in 0.1 M HCl). For lauric acid minimal media plate growth experiments, R. eutropha strains were streaked out onto rich medium plates for 2 days at 30 °C, and then a single colony was transferred onto minimal media plates containing 0.1% laurate and grown at 30 °C for 5 days before imaging with a Canon EOS Rebel T3i Digital SLR camera.

To generate liquid cultures of E. coli or R. eutropha, a single freshly-grown colony was inoculated into 10 mL of culture medium in a 14 mL BD Falcon disposable polypropylene culture tube and shaken at 220 rpm overnight. R. eutropha free fatty acid (FFA) production experiments were conducted in 2 mL 96-well plates (Thermo Scientific, Waltham, MA, USA) or 30 mL glass culture tubes. In both cases, individual R. eutropha colonies were first inoculated in 10 mL rich medium with appropriate antibiotics and grown at 220 rpm at 30 °C for 20–22 h. For FFA production in 96-well plates, R. eutropha overnight cultures were diluted 1:20 into 1 mL fresh rich medium plus necessary antibiotics and shaken at 30 °C, 1,200 rpm on a Titramax 1,000 platform shaker (Heidolph, Schwabach, Germany) for 4–5 h before induction with 0.5% L-arabinose. For FFA production in glass culture tubes, R. eutropha overnight cultures were diluted 1:20 into 1 ml fresh rich medium supplemented with required antibiotics and shaken at 180 RPM at 30 °C. After 20 h, these cultures were induced with 0.5% L-arabinose and a 20% (v/v) dodecane overlay added. OD₆₀₀ was measured using an Ultrospec 10 spectrophotometer (Amersham, Little Chalfont, UK).

Construction of plasmids

All L-arabinose-inducible plasmids were generated by conventional restriction cloning. L-arabinose-inducible mCherry was digested from pJT390 (constructed by Joseph P. Torella) with SpeI and NaeI and ligated into SpeI and EcoRV sites in pBBR1MCS2 to generate pJC007. C-terminally his-tagged UcFatB2 (codon-optimized for expression in Synechococcus elongatus) was PCR amplified from pJT204 (constructed by Joseph P. Torella) using forward primers JC016–JC019 and 5′-phosphorylated reverse primer JC023 to generate four RBS variants driving UcFatB2. The resulting PCR products were digested with XbaI and inserted into XbaI and AleI sites in pJC007 to construct plasmids pJC022–pJC025. All markerless deletion plasmids were generated by conventional restriction cloning or Gibson assembly. A gene deletion fragment consisting of 240 bp stretches directly upstream and downstream the A0285 open reading frame was linked by a PacI restriction site, and the 498 bp product with flanking BamHI sites was synthesized directly by Integrated DNA Technologies (IDT). The synthesized insert and backbone vector pCB96 was digested with the double-cutter BamHI, and ligated to generate the A0285 markerless deletion plasmid, pJC043. The B1148 markerless deletion plasmid was generated as described by Lu et al. (2012), using primer pairs pJC078/JC079 and pJC080/JC081 for overlap extension PCR. The resulting PCR product and parent plasmid pCB96 were digested with XbaI and SacI and ligated to generate pJC046. The gene deletion fragment for A2794 was synthesized by IDT as described above, except the fragment was flanked upstream by XbaI and downstream by SacI, with ∼30 bp overlap regions to pCB96 on both ends. The fragment was assembled to the pCB96 backbone (linearlized by XbaI/SacI) via Gibson Assembly to generate the A2794 markerless deletion plasmid, pJC045. All newly-ligated plasmids and Gibson assembly reactions were transformed into TSS-competent E. coli TURBO or Mach1 cells, isolated and re-transformed into E. coli S17-1 cells for conjugative transfer of plasmid DNA into R. eutropha. Following conjugation, UcFatB2 thioesterase plasmids (pJC022–pJC025) were extracted from R. eutropha and frozen for subsequent electroporation. Primers used in plasmid construction are listed in Table S2.

Genomic modifications

Genetic knockouts were constructed as described previously in Brigham et al. (2010). R. eutropha was conjugated with a donor E. coli strain S17-1 harboring a markerless gene deletion plasmid containing homology to the 250 bp regions upstream and downstream of the gene of interest, and containing a kanamycin resistance cassette and sacB gene for sucrose-based counter-selection. To perform the mating procedure, 10 ml of R. eutropha and E. coli were grown separately in rich medium at 30 °C and LB at 37 °C, respectively. Each culture was harvested after 20 h of shaking, washed twice and resuspended in 1 ml 0.85% NaCl. Next, 30 µl of resuspended R. eutropha was mixed with 20 µl of resuspended E. coli and the 50 µl mixture plated onto an LB plate without selection for 24 h at 30 °C. Following conjugation, a cell scraper was used to remove half the cells from the plate, which were then transferred to an eppendorf tube and resuspended in 300 µl 0.85% NaCl. To select for R. eutropha with successful plasmid integration, either 100 µL or 10 µL of the cell suspension was plated onto rich medium agar plates with 10 µg/ml gentamicin and 200 µg/ml kanamycin and grown for 48 h at 30 °C. R. eutropha colonies were picked and restreaked onto rich medium plates supplemented with gentamicin and kanamycin and grown for 24 h at 30 °C. To allow the second recombination event to occur, the restreaked colonies were grown in 10 mL rich medium with 10 µg/ml gentamicin at 30 °C. After 20 h, 1 ml from the overnight culture was harvested, washed twice with 0.85% NaCl and resuspended in 1 ml 0.85% NaCl. 10 µl of resuspended cells were then plated onto rich medium plates containing 10 µg/ml gentamicin and 5% sucrose. The plates were incubated at 30 °C for 36–48 h, and colonies screened with colony PCR for successful recombination-based loss of the gene of interest. Primers used in strain verification are listed in Table S2. Note that although the B1148 markerless deletion plasmid was constructed, we were unable to knock out the B1148 gene in parallel with other fadD homologues identified by RNA-Seq. The lack of viable colonies from multiple attempts to knock out B1148 in WT and ΔfadD3 background strains may reflect that the gene is essential, or be a consequence of the locally high GC content of the genome.

Generation of R. eutropha MCFA production strains

Freshly-prepared electrocompetent R. eutropha strains were generated for all FFA production experiments using UcFatB2 thioesterase plasmids (pJC022–pJC025). Single R. eutropha colonies were grown overnight in rich medium, diluted 1:10 in 10 ml rich medium in 15 ml falcon tubes, and harvested at OD₆₀₀ = 0.8. Cultures were submerged in ice for 10 min, pelleted and washed twice with 1 ml ice-cold 10% sucrose. The cell pellet was resuspended in 1 ml 10% sucrose, and 50 ul electrocompetent cells were combined with 1–2 µg of plasmid DNA. The cell suspension was transferred to a pre-chilled 0.1 cm gap cuvette and electroporated using the Biorad electroporator manually set at 1.15 kV. Immediately following electroporation, 250 µl rich medium was added and the cells were transferred to an eppendorf tube. Electroporated cells were recovered at 30 °C for 2 h before being plated on selective medium and grown for 2–3 days at 30 °C.

Enzymatic assays for FFA measurements

Fatty acids were measured using the enzymatic Free Fatty Acid Half Micro Test (Roche, Basel, Switzerland) as described by Torella et al. (2013). For FFA measurements with 20% dodecane, 5 µl of the overlay was diluted into 20 ul dodecane before assaying with the same sample volume as previously discussed. Because acyl-CoAs are not cell permeable, detection of these compounds by the enzymatic assay is not possible in the cell supernatant or dodecane overlay.

GCMS identification of lauric acid

Fatty acids were extracted from 400 µl of acidified culture with ethyl acetate and esterified with ethanol. Fatty ethyl esters were extracted with hexane and run on an Agilent GC-MS 5975/7890 (Agilent Technologies, Santa Clara, CA, USA) using an HP-5MS (length: 30 m; diameter: either 0.25 or 0.50 mm; film: 0.25 µm) column, as described by Torella et al. (2013). This method is unable to detect acyl-CoAs, because they cannot be extracted into the solvent layer due to their hydrophilic properties.

Isolation of RNA and RNAseq analysis

For RNA-Seq experiments, individual colonies of R. eutropha strain Re2061 (ΔphaCAB) were inoculated in triplicate into rich medium, grown overnight, then diluted 1:100 into 30 mL minimal medium (Brigham et al., 2010) in a 125 mL flask, supplemented with 2.0% Nondiet-P40 (v/v), 0.54% NH₄Cl, 10 µg/ml gentamicin, and either 0.4% fructose (w/v) or 0.1% lauric acid (w/v). Cells were grown to stationary phase, diluted 1:100 again into the same medium, and each culture monitored at 2 h intervals for cell density by OD₆₀₀. Once a given culture entered mid-log phase (OD₆₀₀ = 0.5–0.6), 1 mL was immediately taken and stabilized with RNA Protect (Qiagen, Hilden, Germany), followed by purification with an RNEasy Mini Kit (Qiagen, Hilden, Germany). The purified RNA was then analyzed by Genewiz on an Illumina HiSeq-2500, and the resulting data analyzed with Rockhopper (McClure et al., 2013) for alignment to the R. eutropha H16 genome and differential gene expression analysis.

Expression of the medium-chain thioesterase UcFatB2 enables MCFA production in R. eutropha

To produce the MCFA lauric acid, we cloned UcFatB2, a thioesterase selective for 12-carbon acyl-ACP substrates from the plant U. californica (Voelker & Davies, 1994), into an arabinose-inducible, R. eutropha-compatible replicative vector (pJC007; Table S1). Ribosome Binding Sites (RBS) with a range of predicted expression strengths (Translational Initiation Region (TIR) values from 14,000 to 84,000 in R. eutropha) were designed using the RBS Calculator (Salis, Mirsky & Voigt, 2009) and cloned upstream of UcFatB2 (pJC022–pJC025; Table S1). One variant (pJC024, TIR = 66,000) resulted in 22 mg/L total free fatty acid (FFA) production within 48 h (Fig. 2A). The absence of detectable fatty acid production in the variant with the highest predicted TIR (84,000, pJC025) may reflect toxicity or inclusion body formation resulting from high expression, or insufficient protein production (e.g., in the case that the actual TIR is much lower than predicted).

GCMS analysis of the ReH16-pJC024 culture confirmed production of lauric acid at 14.3 ± 2.6 mg/L after 24 h of induction; laurate production did not occur in wild-type ReH16 (Fig. 2B). Compared to lauric acid, significantly smaller amounts of myristic (C_14:0) and palmitic (C_16:0) acids were detected from both wild-type and UcFatB2-expressing strains (Fig. 2B, ∼2 mg/L), while the concentration of shorter and unsaturated products was negligible.

Although FFA production was significant within the first ∼24 h following arabinose induction, by 50 h post-induction FFAs were undetectable in the culture (Fig. 3A). This behavior is consistent with the expectation that fatty acids should be re-consumed by R. eutropha through β-oxidation, something previously observed in other microbes engineered to produce fatty acids (Lu, Vora & Khosla, 2008; Runguphan & Keasling, 2014).

Fatty acid production in R. eutropha is limited by β-oxidation and PHB synthesis

To evaluate the role of β-oxidation in FFA reconsumption, we expressed UcFatB2 in R. eutropha strain Re2303, which contains knockouts in two putative β-oxidation operons and is incapable of growth on oleic acid (Brigham et al., 2010). In contrast to the wild-type, Re2303 (Δβ-oxidation) produced FFA concentrations up to 20 mg/L without reconsumption over a 150 h time course, indicating that β-oxidation in the wild-type contributes to the decline in FFA yields (Fig. 3A).

Although Re2303 lacks two β-oxidation operons, the acyl-CoA ligases that catalyze entry of fatty acids into the β-oxidation pathway are not located in these operons. We therefore hypothesized that FFA yield might be limited by uptake of fatty acids and conversion to acyl-CoAs and other β-oxidation intermediates, which are not detected in either of our fatty acid assays (Methods). We assayed FFA production in Re2312, which contains a deletion in a known acyl-CoA ligase in R. eutropha, fadD3 (Brigham et al., 2010), and observed a burst of 30 mg/L FFAs within 4h post-induction along with a steady increase up to 54 mg/L FFA over the following week. By 150 h post-induction, Re2312 (ΔfadD3) led to a 2.8-fold improvement over the Re2303 (Δβ-oxidation) mutant strain and an 8.5-fold increase over wild-type parent strain (Figs. 3A and 3B), indicating that disruption of the first step in β-oxidation enhanced the production of FFAs beyond what is achieved by blocking downstream steps in β-oxidation (e.g., in Re2303).

As PHB synthesis is a sink both for acetyl-CoA (Cook & Schlegel, 1978) and β-oxidation intermediates (Shimizu et al., 2013) in R. eutropha, we hypothesized that it may also limit FFA yield. Expression of UcFatB2 in a phaCAB deletion strain (Re2061) resulted in a considerable 5-fold increase in FFA yield over wild-type by 150 h. Unlike the initial burst of FFAs observed in Re2312 (ΔfadD3), the increase in FFA production by Re2061 (ΔphaCAB) began 40 h after arabinose induction, suggesting a different mechanism for increasing FFA yield. Indeed, PHB synthesis is induced in response to nutrient limitation (Verlinden et al., 2007), which may explain the increase in FFA synthesis at a late time-point in this strain (Figs. 3A and 3B). It is worth noting that simultaneously knocking out PHB synthesis and beta-oxidation genes was unable to enhance FFA production over that achieved by ΔfadD3 alone (Fig. S1), suggesting that beta oxidation is the limiting factor for FFA yield. Our study therefore focused on manipulating the acyl-CoA ligases responsible for lauric acid consumption to enhance fatty acid yields in R. eutropha.

Consistent with the hypothesis that Re2312 (ΔfadD3) had increased FFA production due to a lower rate of FFA consumption, we found that it had a slower rate of exogenous FFA uptake (Fig. 3C). Wild-type R. eutropha, Re2303, Re2061 and Re2312 (none expressing UcFatB2) were incubated with 120 mg/L laurate and the levels of extracellular FFA monitored over a 24 h time period. The wild-type, Re2303 (Δβ-oxidation) and Re2061 (ΔphaCAB) strains had similar consumption rates with laurate no longer detectable in the media by 12 h. Re2312 (ΔfadD3) cells exhibited a slower rate of consumption relative to the other strains, with laurate becoming undetectable by 24 h (Fig. 3C). Laurate consumption was not blocked entirely, suggesting that other acyl-CoA ligases may still be active in Re2312. Indeed, we found that Re2312 was still capable of growth on laurate minimal medium (Fig. S2).

Identification of three additional acyl-CoA ligases that limit FFA yield through RNA-Seq

Deletion of the primary acyl-CoA ligase in R. eutropha, fadD3, results in slower uptake of fatty acids (Fig. 3C) but does not block their utilization as a sole carbon source (Brigham et al., 2010). This is consistent with computational predictions that the R. eutropha genome harbors an apparently redundant β-oxidation system (Pohlmann et al., 2006), estimated to contain 51 homologues of the acyl-CoA ligase alone (Shimizu et al., 2013) (Fig. 4A).

To narrow down the list of acyl-CoA ligases likely to limit the production of fatty acids, a preliminary RNA-Seq experiment was performed to identify acyl-CoA ligases that are up-regulated during growth on lauric acid. Three independent cultures of Re2061 were grown from an OD₆₀₀ of 0.02 in minimal medium containing fructose or lauric acid as a sole carbon source (Fig. 4B). During mid-exponential phase (OD 0.3–0.6), a sample of each culture was taken for analysis by RNA-Seq (Methods). Successfully aligned transcript frequency was low (∼40–50% of reads), likely due to poor sample quality, and we therefore elected not to analyze the resulting data in full here (a future manuscript is forthcoming). Nevertheless, it provided enough information on acyl-CoA ligase expression patterns to engineer improved FFA production in R. eutropha.

Read counts normalized to upper-quartile gene expression (McClure et al., 2013; Bullard et al., 2010) were used to identify four genes annotated as acyl-CoA ligases in the R. eutropha genome that were upregulated strongly (≥5-fold) and significantly (q-value < 0.001) during growth on laurate as compared to fructose: fadD3, A2794, A0285, and B1148 (Fig. 4C). We generated mutants of each gene alone and in combination with fadD3 with the exception of B1148, which we were unable to knock out after several attempts (Methods). FFA production and OD₆₀₀ were measured for each strain 96 h after induction; a large increase in FFA production was associated with knocking out fadD3 (Re2312), A0285 (JC357) and A2794 (JT338), with A2794 providing the greatest FFA yield at 62 ± 2 mg/L (Fig. 4D). The ΔA0285ΔfadD3 double knockout (JC358) failed to increase FFA yield significantly above what was achieved for either knockout alone, while the ΔA2794ΔfadD3 double knockout (JT339) actually decreased FFA yield. As the latter was the only strain tested with impaired biomass yield (Fig. 4D, bottom panel), this reduced mean yield may reflect a synthetic loss of fitness associated with the ΔA2794ΔfadD3 double knockout.