Terpenes are natural products with diverse functions in different organisms. These compounds have benefited human society since antiquity, applied as materials , traditional medicines , pharmaceuticals , and cosmetics . The versatile functions and applications of terpenes owe to their diverse chemical structures, with more than 55,000 molecules known . Nature employs a concise paradigm to build such structures, in which isopentenyl pyrophosphate and dimethylallyl pyrophosphate , fivecarbon units derived from the 2-C-methyl-D-erythritol 4-phosphate pathway or mevalonate pathway , are first polymerized head-to-tail into prenyl pyrophosphates, including geranyl , farnesyl , and geranylgeranyl pyrophosphates. Then, the prenyl pyrophosphates are processed, mostly cyclized, by terpene synthases to produce terpene scaffolds . Lastly, the optional tailoring proteins, including P450s and glycosyltransferases , modify the scaffolds to afford the final products. Through this paradigm, structural diversities of the final products are introduced from different numbers of five-carbon units polymerized, varied cyclizations , and optional post modifications.Terpenes with non-multiple of five carbons are rare in nature and challenging to sample, even though this chemical space is valuable to explore for novel applications. We reported heterologous expression of the lepidopteran mevalonate pathway, a propionyl-CoA ligase, and terpene cyclases in E. coli to produce several novel sesquiterpene analogs containing 16 carbons . The LMVA pathway produces 6-carbon analogs of IPP and DMAPP, homoIPP and homo-DMAPP , in a reaction sequence highly similar to the canonical MVA pathway . The difference is that the thiolase condenses a propionyl-CoA and an acetyl-CoA making 3-ketovaleryl-CoA in the LMVA pathway instead of the condensing two acetyl-CoA to produce acetoacetyl-CoA in the canonical MVA pathway . The “extra” carbon from the propionyl-CoA is transformed into HIPP/HDMAPP,grow trays combined with two 5-carbon isoprenoid precursors to afford the sixteen-carbon products.
With the LMVA pathway expressed in E. coli, we established a biosynthesis platform for novel homoterpenes deviating from the “multiple of five carbons” rule. While the previous study successfully produced the final homosesquiterpenes, the details of the LMVA pathway, especially the production of HIPP and HDMAPP in the E. coli host was not assayed. Moreover, the production needs optimization because low C16 terpene titers have hindered us from accumulating and purifying these products for characterization. Here, we investigate the LMVA pathway by introducing a promiscuous phosphatase, NudB , to hydrolyze the terpene precursors to their corresponding alcohols. These alcohols are readily detected by gas chromatography . Using the alcohols as the final product, we were able to engineer and optimize the LMVA pathway to increase the production of HIPP, the direct product of LMVA and the starting substrate for homoterpenes synthesis. Also, the higher 3-methyl-3-buten-1-ol analogs have excellent fuel properties, making them candidates for next-generation biofuels.The atoB knockout and atoB yqeF double knockout strains were generated using the λ red recombinase protocol previously described . A kanamycin resistance cassette flanked by FLP recognition target sites was amplified from the plasmid pKD13 with primers KO_kan_F and KO_kan_R. The 1000 base pair homology regions upstream and downstream of atoB and yqeF were amplified using the primer pairs atoB_US_F/atoB_US_R and atoB_DS_F/atoB_DS_R, and yqeF_US_F/yqeF_US_R and yqeF_DS_F/yqeF_DS_R using BW25133 genomic DNA as template, respectively. The homology regions were designed to knock out the entire gene except for the starting ATG and the last 21 base pairs to avoid disrupting potential downstream ribosomal binding sites. The upstream and downstream homology sequences were combined with the kanamycin cassette by using NEBuilder HiFi DNA Assembly , and a second PCR was done to amplify the three-part fusion DNA sequences using primer pairs atoB_US_F/atoB_DS_R or yqeF_US_F/yqeF_DS_R using the corresponding assembly product. To generate marker-free atoB knockout, E. coli 6C01, the pKD46 plasmid with the λ red recombination genes and a temperature-sensitive replicon was transformed into chemically competent E. coli BL21 cells and selected on carbenicillin LB agar plates at 30oC. Plasmid-containing cells were made electrocompetent, and expression of recombination genes was induced with 0.1% arabinose. ~600 ng of the three-part PCR product was introduced into the cells by electroporation, and positive colonies were selected for on LB kanamycin plates grown at 37oC overnight.
To remove the kanamycin resistance cassette, pCP20 was transformed into the selected kanamycin resistance cells using electroporation and selected on carbenicillin and kanamycin LB agar plates at 30oC. The resulting colonies were streaked on an LB plate without antibiotic and grew overnight at 42oC to remove the cassette. The resulting colonies were streaked on an LB plate without antibiotic, a carbenicillin LB agar plate, and a kanamycin LB agar plate, respectively, to confirm the loss of the kanamycin cassette and pCP20.Two plasmids bearing the pathway genes were co-transformed into E. coli expression hosts using a room temperature electroporation method, in which the electrocompetent bacterial cells were prepared at room temperature freshly .The gene sequences were codon-optimized for E. coli and cloned into the pJL02 and pJL01 plasmid series . With the constructs in hand, we tested C6-isoprenol production in E. coli. First, we screened the CoA ligases, catalyzing the first step in beta-oxidation, with the well-characterized MeD as the acyl-CoA dehydrogenase. After production, we used GC-FID to quantify C6-isoprenol in the broths. The C6-isoprenol production varied substantially with different CoA ligases, suggesting this is a key step in the whole pathway. Among the genes we tested, PcCL gave rise to the highest production at 58.6 mg/L , a result consistent with the reported kinetics data . Also, we noticed that the combination of CsCL with MeD had a decreased production of 6.9 mg/L, down from the 27.3 mg/L produced by CsCL/MlD, suggesting that MlD is a better homolog for the second step . This hypothesis was supported by the screening results of the second step catalyzed by acyl-CoA dehydrogenase. The PcCL and MlD pair gave the highest C6-isoprenol production at 110.3 mg/L, so we used this pair of genes for the first two steps of the beta-oxidation pathway in the following experiments. Endpoint optical density analysis was performed to evaluate the impact of the expression of different beta-oxidation genes on cell growth. The results indicated the expression of the CoA ligases and acyl-CoA dehydrogenase barely impacts the growth. The slight growth inhibition effect in the high C6- isoprenol production, e.g., the PcCL/MlD,drying marijuana may result from the toxicity of C6-isoprenol. After optimizing the pathway genes to increase the production efficiency, we turned to the host genes that may degrade the key intermediate, 3-ketovaleryl-CoA.
As mentioned before, this intermediate can be degraded by a thiolase into acetyl-CoA and propionyl-CoA. Therefore we focused on two E. coli chromosomal thiolase genes, atoB and yqeF, which have substrate preference for short-chain betaketoacyl-CoA and were proposed to degrade 3-ketovalerylCoA. We conducted in-frame single-gene knockouts to delete these two genes in E. coli BL21 in sequence, and PCR confirmed the genotypes of the knockout mutants . The intermediate strain E. coli BL21 ΔatoBand the final double knockout strain E. coli BL21 ΔatoB ΔyqeF were then used for C6-isoprenol production using the plasmid combination of pJL01-MlD/pJL02-PcCL. The production of C6-isoprenol and the consumption of valeric acid were quantified. The results showed that while the knockout of atoB did not impact the C6-isoprenol titer, the double knockout strain, 6C02, increased the C6-isoprenol production to 390 mg/L, and the C6- isoprenol yield from valeric acid doubled to 44 mol% over the wild-type strain . We noticed even after knocking out the thiolase genes, only around half of the valeric acid is transformed into C6-isoprenol. The loss of valeric acid may come from the evaporation and the flux into side directions in the metabolic network, such as the degradation of 3-ketovaleryl-CoA by other thiolases in E. coli .In the C6-isoprenol runs, we noticed that the levels of isoprenol were generally negatively correlated to the levels of C6-isoprenol. For the sources of isoprenol, we reason that in addition to the E. coli native MEP pathway, the LMVA pathway may contribute the major portion via acetoacetylCoA, instead of 3-ketovaleryl-CoA, to C6-isoprenol. Through the LMVA pathway, the productions of C6-isoprenol and isoprenol use the same precursor, acetyl-CoA, resulting in the negatively correlated levels of C6-isoprenol and isoprenol. Also, the knockout of the short-chain acyl-CoA thiolase increased the supply of acetoacetyl-CoA and the production of IPP, resulting in an increased isoprenol titer of 14.4 mg/L compared to 2.1 mg/L for the wild-type strain. Hence, we proposed that acetoacetyl-CoA is readily accepted by the LMVA pathway, and it’s betaoxidation precursor, butyric acid, might be a substate of the beta-oxidation LMVA pathway. To test this hypothesis, we fed 1 g/L butyric acid instead of valeric acid to strain 6C02 containing pJL01-MlD and pJL02-PcCL.
After production, GC-FID and GC-MS confirmed the isoprenol production, quantified at 301.8 mg/L . This result validates butyric acid is a good substrate for the beta-oxidation LMVA pathway for IPP/isoprenol production. The successful transformation of butyric acid into C5 alcohols by C4A suggests that the beta-oxidation LMVA pathway has substrate promiscuity. To explore the substrate spaces of this pathway, we tested other fatty acids as substrates. Without supplementing hexanoic acid, E. coli 6C02 with the beta-oxidation LMVA pathway also produces a small amount of C7-isoprenol, confirmed by the synthetic standard using GC-FID and GC-MS . Supplementing hexanoic acid increased the production of C7-isoprenol significantly . Therefore, the betaoxidation LMVA pathway can activate hexanoic acid and transform it to C7-isoprenol, albeit with low efficiency. C7-isoprenol production without hexanoic acid supplementation is likely from endogenous hexanoyl-CoA in E. coli. We also tried fatty acid analogs with functional group substitutes, including 5- chloro-valeric acid, 4-pentenoic acid, 4-amino-butyric acid, 5,5,5-trifluorovaleric acid, and 4- bromobutyric acid. We conducted comparative GC-FID/GC-MS analysis and fragment search in GC-MS for lack of standards of the expected alcohol products. However, none of the expected substituted alcohols were detected, suggesting these fatty acid analogs are poor substrates for the beta-oxidation LMVA pathway . The substrate promiscuity assays suggest the beta-oxidation LMVA pathway can produce IPP and C7-IPP, expanding the product chemical space of the homoterpene biosynthesis platform. The isopentenols, including isoprenol and prenol, are drop-in alcohol biofuels and have versatile potential fuel applications: prenol is one of the top 10 Department of Energy Co-Optimization of Fuels & Engines gasoline blendstocks and has synergistic blending effects for research octane number , and isoprenol is the precursor of 1,4- dimethylcyclooctane , a high-performance jet fuel blendstock . Previous studies have shown that for alcohol biofuels, molecules with longer chain lengths have a better blend stability with conventional fuels, and are less hygroscopic than their shorter chain congeners . Our pathway to C6-isoprenol and C7-isoprenol from biomass-derived fatty acids makes it possible to produce these chain-extended isoprenol analogs sustainably. With the synthesized C6-isoprenol and C7-isoprenol, we were able to test some important fuel properties of these novel isoprenol analogs. We first estimated their water solubility based on their logP values. High water solubility contributes to the high hygroscopic nature of alcohol fuels, increasing the possibility of phase separation when blended with conventional hydrocarbon fuel. Also, for microbial bio-fuel production process, molecules with high logP and low water solubility will partition into the organic phase in a two-phase extractive fermentation, resulting in low product toxicity to the producing microbes . Increasing carbon chain length leads to decreasing polarity of alcohols, resulting in lower water solubilities. The logP of isoprenol analogs were determined using an HPLC method, with C4-C8 strain chain alcohols as references. The result revealed that the one-carbon increase of the chain length of isoprenol decreases the water solubility to 22.12 g/L from 65.36 g/L. Moreover, the addition of two carbons to isoprenol further decreases the water solubility to 6.6 g/L . The decreasing trend of the water solubility was expected, and these data reflect the trend quantitatively. The energy density of the isoprenol analogs was determined by testing their gross heats of combustion using a standard method . The energy density tests revealed C6-isoprenol has a higher heating value of 35.524 MJ/kg, while C7-isoprenol had an HHV of 39.468 MJ/kg . These numbers are similar or higher to the HHVs of isopentenols tested in the same batch. We also determined the RONs of the 10% alcohol RBOB gasoline blends using the recently published AFIDA method .