Production of FAME biodiesel in E. coli by direct methylation with an insect enzyme

Most biodiesel currently in use consists of fatty acid methyl esters (FAMEs) produced by transesterification of plant oils with methanol. To reduce competition with food supplies, it would be desirable to directly produce biodiesel in microorganisms. To date, the most effective pathway for the production of biodiesel in bacteria yields fatty acid ethyl esters (FAEEs) at up to ~1.5 g/L. A much simpler route to biodiesel produces FAMEs by direct S-adenosyl-L-methionine (SAM) dependent methylation of free fatty acids, but FAME production by this route has been limited to only ~16 mg/L. Here we employ an alternative, broad spectrum methyltransferase, Drosophila melanogaster Juvenile Hormone Acid O-Methyltransferase (DmJHAMT). By introducing DmJHAMT in E. coli engineered to produce medium chain fatty acids and overproduce SAM, we obtain medium chain FAMEs at titers of 0.56 g/L, a 35-fold increase over titers previously achieved. Although considerable improvements will be needed for viable bacterial production of FAMEs and FAEEs for biofuels, it may be easier to optimize and transport the FAME production pathway to other microorganisms because it involves fewer enzymes.

Increasing energy consumption and the detrimental environmental impact of fossil fuels has led to increased interest in developing sustainable and renewable sources of energy. The utilization of engineered microorganisms to produce chemicals from renewable biomass is a promising alternative to petroleum-derived fuels and chemicals. Fatty acid derived compounds are particularly promising because fatty acid derivatives are highly reduced, aliphatic compounds with high energy density that are not miscible with water 1 . Notably, their similarity to diesel fuels makes them compatible with existing infrastructure. As a result, many strategies have been developed to overproduce microbial fatty acids and then further convert the fatty acids into biofuels such as alkanes, fatty alcohols, and fatty acid methyl or ethyl esters [2][3][4][5][6][7][8] .
Microbial production of fatty acid methyl or ethyl esters (FAME, FAEE respectively) is of particular interest because FAME and FAEE are the main component of biodiesel currently in use. Typically, biodiesel is made by transesterification of triacylglyceride oils extracted from renewable biomass with short chain alcohols (e.g methanol or ethanol) using an alkaline catalyst 9 . However, the use of feedstock oils needed for biodiesel production is a major obstacle for the broader use of biodiesel due to lack of arable land and competition with the food supply. Therefore, a possible alternative to plant and animal oil-based biodiesel is the direct biosynthetic production of biodiesel in metabolically engineered microorganisms (reviewed in 10 ).
Steinbuchel and co-workers were the first to develop a pathway for the production of FAEE biodiesel in E. coli, and their approach was further developed by the Keasling group for increased yields of FAEE and fatty alcohols 2,4,7 . To produce FAEEs, two orthogonal pathways were introduced that simultaneously generated ethanol and fatty acyl-CoA. In the last step, ethanol and fatty acyl-CoA were then condensed to the FAEE using a wax ester synthase ( Fig. 1) 2,7 . After optimization, titers as high as 1.5 g of long chain FAEEs per liter of culture were obtained.
As a more straightforward approach to produce biodiesel in microorganisms, the Lykidis group attempted to produce FAMEs in E. coli through direct methylation of fatty acids by the action of an S-adenosyl-L-methionine (SAM) dependent bacterial methyltransferase from M. marinum 3 . The Lykidis pathway has the advantage of being much simpler than the FAEE production pathway by using endogenous compounds (SAM and fatty acids) produced in E. coli. Nevertheless, the FAME titers obtained were nearly two orders of magnitude lower than the FAEE titers (16 mg/L). The low level of FAME production is likely due to the high specificity of the methyltransferase employed, which prefers rare fatty acids containing a 3-hydroxy group 3 .
We hypothesized that if we could find a broad range fatty acid methyl transferase, perhaps we could improve upon the Lykidis approach for FAME production. Here we show that Drosophila melanogaster Juvenile Hormone Acid O-Methyltransferase (DmJHAMT) has broad specificity for medium chain free fatty acids and can be used to produce FAMEs in E. coli. By introducing DmJHAMT to engineered E. coli strains tolerant to high levels of endogenously produced medium chain fatty acids, we observed in vivo FAME production 11 . Enriching the endogenous SAM pool further increased FAME production with final titers showing a 35-fold increase from titers previously reported 3 .

Results and Discussion
DmJHAMT is robustly expressed and broadly active on medium chain fatty acids. Several SAM-dependent juvenile hormone acid methyltransferases have been previously found to methylate insect sesquiterpenoid hormones that play central roles in the development and growth of these organisms [12][13][14] . D. melanogaster Juvenile hormone acid O-methyltransferases (DmJHAMT) appeared to be a promising enzyme for FAME production because it showed some activity with unbranched saturated medium and long-chain fatty acids such as lauric and palmitic acids, and could be expressed in E. coli 13 .
We expressed the DmJHAMT protein recombinantly in E. coli to investigate its substrate specificity. DmJHAMT expression was robust in E. coli (up to 200 mg of protein per liter of culture) with no apparent effect on cell growth. As shown in Fig. 2a, DmJHAMT is active on fatty acids ranging in size from C12 to C16. We saw no activity with shorter chain, C8:0 and C10:0, fatty acids, however. DmJHAMT is most active on medium chain fatty acids, showing the highest activity with lauric acid (C12:0) among the substrates tested. The kinetic parameters with lauric acid were K m = 59 μM and k cat = 0.15 min −1 (Fig. 2b). Although the low k cat indicates that the enzyme is not very efficient with these non-natural substrates, the high expression and broad specificity of DmJHAMT suggested that it might be effective at producing FAMEs, particularly medium chain FAMEs, in E. coli. DmJHAMT produces FAME biodiesel in E. coli. E. coli has been utilized as a host for over-production of free fatty acids (FFAs) of various lengths and properties 2,5,15-24 . Introduction of bacterial and plant acyl-ACP thioesterases in a ΔfadD mutant E. coli strain defective in fatty acid degradation allows overproduction of free fatty acids by liberating fatty acids attached to acyl-carrying proteins (ACPs), while simultaneously removing acyl-ACP mediated regulation of the fatty acid biosynthesis pathway, effectively redirecting lipid biosynthesis into free fatty acid production 25 . Since DmJHAMT is most active with medium chain FFAs, we opted to utilize the acyl-ACP thioesterase from Umbellularia californica (BTE), which has a preference for medium chain fatty acids and leads to accumulation of lauric acid when expressed in an E. coli ∆fadD strain 5,26 .
We first prepared strain SS3B (∆fadD DmJHAMT/BTE) bearing a ∆fadD mutation and expressing DmJHAMT and BTE from plasmids (Table 1). In strain SS3B we observed relatively high production of medium-chain fatty acid methyl esters (Fig. 3). The initial titer of FAMEs was 240 ± 15 mg/L of culture, already a dramatic improvement over prior results 3 . Since medium-chain FAMEs are somewhat volatile, we added dodecane as an organic overlay at the stationary phase to trap the FAMEs, which further increased the titer of FAMEs to 312 mg/L of culture 2 . The majority of FAMEs contained 12-carbon acyl chains (73%), mostly unsaturated C12 methyl laurate (Fig. 3).
While a high level of FAMEs were produced, we were surprised to find that strain SS3B (∆fadD DmJHAMT/ BTE) still produced a considerable amount of free fatty acids (FFA) that were not methylated (860 ± 20 mg/L of culture). Indeed a majority of the FFAs generated in strain SS3B were not converted into FAMEs. We therefore sought to increase the conversion of the excess FFAs to FAMEs.

Increasing SAM levels.
We hypothesized that SAM levels may be a limiting factor in the conversion of FFAs to FAMEs. To test the possibility that low SAM levels during stationary phase contributed to low FAME production, we lysed strain SS3B after two days of growth and supplemented the lysate with exogenous SAM. We observed increases in all FAME species indicating that the DmJHAMT remained active but the SAM levels may be limiting (Fig. 4a). To increase SAM production, we introduced the methionine synthase protein from rat liver, Mat1A, into E. coli strain SS3B 27,28 . Mat1A was shown to dramatically increase the intracellular.
SAM pool in E. coli cells 28 . Mat1A expression from a plasmid in strain SS4B (∆fadD DmJHAMT/BTE/Mat1A) increased SAM levels 8.5-fold (from 73.3 to 636.8 nmoles per gram of cells) compared to the control strain SS3B (∆fadD DmJHAMT/BTE) after two days of growth. Nevertheless, we found that Mat1A overexpression actually decreased both FFA and FAME titers. Mat1A overexpression may have unexpected deleterious effect on FAME production such as toxicity, competition for expression with other proteins, high metabolic ATP demand for SAM production, or the complication of harboring three different plasmids, among other possibilities 29 .
To simplify the system and reduce the expression of Mat1A, we incorporated a single copy of the Mat1A gene into the E. coli genome under the control of a T7 promoter. When Mat1A was incorporated into the genome, we observed ~3-fold increase of SAM levels in strain SS33 (∆fadD::Mat1A BTE/DmJHAMT) compared to SS3B (∆fadD BTE/DmJHAMT) after two days of growth (192 ± 3 nmoles SAM per gram of cells compared to 71 ± 19 nmoles per gram of cells in control the strain, Fig. 4b). More importantly, we saw a 19% increase in FAME production, from 312 mg/L to 370 mg/L in cells carrying Mat1A in the genome. In addition, this strain had a higher ratio of SAM to S-adenosylhomocysteine (SAH), a by-product SAM-dependent methylation and a potent inhibitor of methyltransferases (Fig. 4c). While the levels of SAH were similar in these strains, the levels of SAM showed considerable increases in Mat1A-carrying strains after 48 hours of growth 30 . Overall, Mat1A expression improved the production of FAMEs.
Δaas further increases the FAMEs titers in E. coli. Short and medium chain FFAs are toxic to E. coli cells, most likely due to membrane stress 23,[31][32][33] . It is possible that the production of excess FFAs in our strains is deleterious to FAME production. We recently reported that the deletion of the aas gene can alleviate medium chain FFA toxicity 11 . The Aas protein acts in a FFA salvage pathway that can incorporate exogenous medium chain FFAs directly into the lipid bilayer with deleterious consequences. We therefore attempted to reduce the toxicity of the medium chain fatty acids by deleting the aas gene in strain SS3B to produce strain SS34 (Δaas ΔfadD::Mat1A BTE/DmJHAMT). Indeed strain SS34 showed an almost 50% increase in the FAME production (559 mg/L of culture) compared to the same strain with a wild type aas gene SS33 (ΔfadD::Mat1A BTE/DmJHAMT). Overall,

Conclusion
We have engineered a strain of E. coli that produces FAMEs at levels comparable to the best FAEE production strain and at levels that are more than an order of magnitude greater than FAME titers previously attained 2,3 . Essential developments were the identification of a FFA methyltransferase that has broad specificity for fatty acids and could be overproduced in E. coli and deletion of the aas gene to reduce incorporation of toxic medium chain-length FFAs into the bilayer. The fact that more than half of the FFAs generated (1.45 g of FFAs vs 0.559 g FAME) are not methylated in the highest producing strain (SS34) suggests that there is still considerable room for improvement. We do not know why FFAs are not fully converted to FAMEs, but presumably some portion of the FFAs is sequestered from DmJHAMT (e.g in the membrane) because there is still sufficient SAM (211 nmoles per g of cells) and active enzyme present after several days, yet FFAs remain. It is also possible that the FFAs that escape from the cell are not reabsorbed efficiently due to the ΔfadD mutation, the normal route for uptake of long-chain free fatty acids. Poor re-uptake may be particularly problematic for medium chain FFAs even with fadD intact 37 , so perhaps better results will be obtained with strains that can produce longer chain (C16 and C18) FFAs on which DmJHAMT is active. Screening of other methyltransferases or the engineering of methyltranferases for broader specificity should allow for still further improvements and diversification of the FAME products. While heat of combustion and cetane number, a measure of diesel ignition quality, are similar in these molecules, increasing the proportion of unsaturated acyl groups in this biofuel mix adds beneficial properties such as lower cloud point and lower freezing temperature 38,39 . Current studies are underway to increase branched and unsaturated fatty acid yields in E. coli that could potentially be used in our one-step biodiesel production method [40][41][42] .

Materials and Methods
Materials. T4
To knock-in genes into the E. coli genome, we generated a plasmid, called pCDF-Cat, that contains a chloramphenicol resistance gene (cat) flanked by the FLP recognition target (FRT) sites 45 . To do that, the cat gene cassette containing FRT sites was amplified from the pKD3 plasmid using pKD3-Cat-pCDF-Forw and pKD3-Cat-pCDF-Rev primers and the pCDF-1B plasmid was amplified using pCDF 385-Rev and pCDF 425-Forw primers. The resulting PCR fragments were ligated together using the AMM kit so that the cat gene was inserted into the 385-425-base pair region of the pCDF-1B plasmid 43 . The Mat1A gene was then cloned into pCDF-Cat the same way as Mat1A was inserted in pCDF-1B vector and the resulting Mat1A-pCDF-Cat plasmid was used as template to amplify the Mat1A-FRT-cat-FRT fragment that was inserted into the E. coli genome (see below). The primers used for the cloning are listed in Table 2. All cloned genes were verified by sequencing. E. coli strains construction. E. coli strains K-12 MG1655, JW 1794-1 (Δfad::kan) and JW2804-1 (Δaas::kan) were used as the starting point for strain construction 46 . The SS19 strain carrying a double Δfad Δaas deletion was generated as previously described 11 . A Mat1A knock-in PCR fragment was generated by using the primers FadD KO -pCDF1 P1-1 and FadD KO -CAT P2-1 for amplification on the Mat1A-pCDF-Cat plasmid and further extended in a second round of PCR using the primers FadD-P1-pKD4-Primer2 and FadD-P2-pKD4-Primer2. This PCR fragment was employed to insert Mat1A into the fadD gene region of the K-12 MG1655 and JW2804-1 strains. Subsequent cat gene removal was performed according to protocol from Datsenko and Wanner 45 . A λ DE3 prophage was integrated and BTE-pBAD/p15A and DmJHAMT-pET15b plasmids were transformed into each strain. The list of strains and their genotypes are in Table 1.
Protein Expression and Purification. DmJHAMT was expressed from DmJHAMT-pET-28a(+) plasmid in a BL21(DE3) strain and purified using Ni-NTA affinity chromatography. 2 mL of an overnight starter culture was transferred to 2 L of LB media containing 50 μg/ml kanamycin and incubated at 37 °C. When the OD 600 of

Name of Primer Sequence
XhoI-pBAD/p15A-BTE 5′-GGGTTTTCTCGAGGAGTGGAAGCCGAAGCCGAA-3′ FadD-P2-pKD4-Primer2 5′-GCGTCAAAAAAA ACGCCGGATTAACCGGCGTCTGACGACTG-3′ Enzyme Assays. DmJHAMT activity was measured using an enzyme-coupled colorimetric assay for SAM-dependent methyltransferases 49 . Enzyme assay solutions contained 20 μM LuxS, 10 μM MTAN, 500 μM SAM and various concentrations of fatty acid substrates in degassed 50 mM potassium phosphate [pH 8.0] at a final volume of 500 μL. 3 μM DmJHAMT was used for the k cat /K m calculations of lauric acid (Fig. 2b) and 10 μM DmJHAMT was employed for reaction rate calculation with other fatty acids (Fig. 2a). Fatty acids were added from stock solutions prepared at 1 mg/mL in 100% ethanol. C16 palmitic acid was insoluble at concentrations >50 μM so comparison of the reaction rates for different fatty acid substrates was performed at 40 μM fatty acid. 5-200 μM range of lauric acid was used to obtain k cat /K m values for this specific substrate 50 . All components of the assay except DmJHAMT were combined and mixed and the reaction was initiated by addition of DmJHAMT at 30 °C. 60 μL of the reaction mixture was taken out at various time points and quenched by adding 180 μL of 260 μM DTNB, 0.5 mM EDTA, 6 M GuHCl (room temperature) and the absorbance at 412 nm read after a 20 min incubation. A standard curve for SAH consumption by MTAN/LuxS was developed and used to quantify FAME production in the enzyme-coupled assays. All experiments were done in duplicate or triplicate and standard deviation from the mean value was used for error bars.
SAM/SAH Assay. The SAM/SAH measurement protocol from cultures was modified from 51 . E. coli cells were pelleted by centrifugation (6000 rpm, Eppendorf F45-30-11 rotor, 5 min, 4 °C) and the wet cell weight was measured for each sample. The cells were resuspended and lysed by vortexing in 5% trifluoroacetic acid at 4 °C for 2 min (4 ml/g of wet cell weight). The cell lysate was clarified by centrifugation (13000 rpm, Eppendorf F45-30-11 rotor, 5 min, 4 °C) and 120 μl of supernatant was analyzed by high performance liquid chromatography (HPLC) as described in 51 . The concentrations were calculated using SAM and SAH standards of known concentrations. All measurements were performed in triplicate.
Cell growth. Most of the strains did not reach saturation point in minimal media supplemented with either glycerol or glucose and terrific broth (TB) with 1.5% glycerol was used for cell growth and subsequent analysis. The media was supplemented with ampicillin (50 μg ml −1 ), chloramphenicol (34 μg ml −1 ) or kanamycin (50 μg ml −1 ) as appropriate. 5 mL of TB-glycerol were inoculated from a single colony and cultured overnight at 37 °C. The seed cultures were then used to inoculate 30 mL TB-glycerol medium with appropriate antibiotics in 150 mL culture tubes and cultivated at 25 °C in a rotary shaker (210 rpm). BTE and DmJHAMT expression was induced at an OD 600 of 0.1 with 50 μM Isopropyl-β -D-thio-galactoside and/or 0.002% L-arabinose. For samples with a dodecane overlay, 6 ml of dodecane were added after 24 hours of growth. Cultures were grown for an additional 1 day prior to FA/FAME analysis as described below.
Metabolite extraction and identification. FFAs and FAMEs were extracted by addition of 6 mL of a 2:1 chloroform/methanol mixture (spiked with 0.15 mg/L of either methyl tridecanoate or methyl heptadecanoate as an internal control) to 5 ml of culture. For consistency in data analysis, 1 mL of dodecane layer was similarly treated with 6 mL of a 2:1 chloroform/methanol mixture before gas chromatography (GC) analysis. Quantification of FAs/FAMEs was conducted by GC-FID using an HP 5890 Series II gas chromatograph equipped with an HP-Innowax Column (0.32 mm x 30 m x 0.25 μm, Agilent). All samples were analyzed using the following parameters: inject: 1 μl; inlet temperature 250 °C with split ratio 1:1; carrier gas: helium; flow: 5 ml/ min; oven temperature: initial temperature of 160 °C, hold 3 min; gradient to 255 °C at 5 °C/min; hold 3 min; inlet temp: 270 °C, detector temp: 330 °C. The amount of FAs/FAMEs was determined by comparison to a standard curve of various FAs and FAMEs and methyl tridecanoate or methyl heptadecanoate concentrations. To identify all FA/FAME products, GC/mass spectrometry analysis was additionally performed using an Agilent 6890-5975 equipped with HP-Innowax Column (0.32 mm x 30 m x 0.25 μm, Agilent). Peak identification was performed through comparison with GC retention time, known standards and mass spectra with the National Institute of Standards and Technology (NIST) database.