Genetic studies support PhaG has acyl-ACP:CoA transacylase activity
Homologs of PpuPhaG have been used extensively as a means of enhancing the production of mcl PHA in bacteria19,27,28,29,30,31. That said, the specific activities catalyzed by PpuPhaG are debated. In vitro studies have confirmed the ability of PhaG to generate 3-hydroxyacyl-ACP from the corresponding CoA species and holo-ACP20,21. This is the reverse reaction of the one desired for oleochemical production studies and no in vitro data on acyl-ACPs, the substrate in the forward direction is available. From 1998 to 2012, PhaG was generically called a 3-hydroxyacyl ACP:CoA transacylase, based on in vitro data. In 2012, Nomura and co-workers challenged the name and the ability of the enzyme to perform the transferase reaction30. In this study, E. coli BL21 cells harboring a plasmid for expressing an mcl-PHA polymerase (PhaC1) were transformed with plasmids expressing PhaG and/or a CoA ligase from P. putida. Cells expressing both enzymes produced more than ten times the amount of PHA than those lacking the CoA ligase. The conclusion drawn from this study is that the ligase is needed for high-flux PHA generation and PhaG acts primarily as a thioesterase. Subsequent papers have used the name thioesterase, but have not provided further evidence to support the presence of thioesterase activity. In contrast, prior studies demonstrated that co-expression of TesB, a promiscuous CoA thioesterase enhanced the production of 3-hydroxy-fatty acids in both E. coli and P. putida32,33. TesB was similarly used to produce 3-hydroxy fatty acids as precursors to methyl esters34.
The complicated history motivated us to confirm that PhaG had substantial transferase activity. Therefore, we compared the metabolic product profiles (looking for production of methyl ketones or 3-hydroxy fatty acids) of specifically engineered strains of E. coli (MG1655 ΔfadA, ΔfadI, ΔfadD, ΔfadR, pTRC99a-PpuphaG–EcfadM) to determine if heterologously expressed PhaG demonstrated more thioesterase or transacylase activity. Strains were designed to create a 3-hydroxy fatty acid product sink to indicate potential PhaG thioesterase activity and a methyl ketone product sink for PhaG transacylase activity (Fig. 2A). Deletion of fadD removes the dominant acyl-CoA synthetase activity and prevents the reactivation of free fatty acids generated by either FadM or PhaG. Deletion of fadA and fadI removes known thiolase activities from E. coli and blocks β-oxidation from catabolizing any acyl-CoAs produced in vivo. Deletion of fadR removes repression of fadB expression and thereby upregulates a bi-functional enoyl-CoA hydratase/3-hydroxyacyl-CoA dehydrogenase responsible for converting 3-hydroxyacyl-CoA thioesters to 3-ketoacyl-CoA thioesters. FadM4,35 is overexpressed to provide 3-ketoacyl-CoA thioesterase activity, resulting in the conversion of any 3-hydroxyacyl-CoAs generated by PhaG to the corresponding methyl ketones. Cultures of E. coli RADI strain harboring pTRC99a-PpuphaG–EcfadM were grown in Clomburg media at 30 °C for 48 h. Culture samples were extracted and derivatized for GC/FID and GC/MS analysis. The samples contained a total of 170 mg/L C7–C13 methyl ketones but no detectable 3-hydroxy methyl esters (Fig. 2B) indicating that PpuPhaG functions primarily as a 3-hydroxyacyl ACP:CoA transacylase. Strains expressing FadD (E. coli RAI strain pTRC99a-PpuphaG–EcfadM) produced equivalent amounts of methyl ketones, indicating that the carbon flux for methyl ketone synthesis is not enhanced by FadD-catalyzed free fatty acid reactivation. Strains lacking PhaG overexpression (E. coli RADI strain pTRC99a-EcfadM) produced small amounts (<1 mg/L) of methyl ketones (Fig. 2B) that have been previously observed in strains expressing FadM4,35. Strains lacking EcFadM overexpression (E. coli RADI strain pTRC99a-PpuphaG) contained a total of 55 mg/L C8–C12 3-hydroxy methyl esters (Fig. 2C) consistent with prior studies32,33. Together, these data suggest that 3-hydroxy fatty acid production observed in past studies likely comes from thioesterase activities encoded by native enzymes (e.g., YciA, FadM, TesB can potentially catalyze the cleavage of (R)-3-hydroxyacyl-CoA to 3-hydroxy fatty acids) that are outcompeted by the methyl ketone synthesis pathway we introduced.
A Metabolic pathways used to test for the presence of PhaG-dependent thioesterase and/or transferase activity. B In cells lacking FadR, FadA, FadI, and FadD (E. coli RAID harboring pTRC99a-PpuphaG), PhaG expression leads to the production of 50 mg/L medium-chain 3-hydroxy fatty acids (n = 3 biologically independent samples). These products could be generated by either direct PhaG thioesterase activity on medium-chain acyl-ACPs or by CoA thioesterase activity on stranded pools of 3-hydoxyacyl-CoAs made via PhaG transferase activity. ***P = 0.0005 was analyzed based on student two-tailed t test assuming unequal variances. C Co-expression of PhaG, and FadM in E. coli RAI (MG1655 ΔfadR ΔfadA ΔfadI harboring pTRC99a-PpuphaG–EcfadM) results in the production of 150 mg/L of medium-chain methyl ketones (n = 3 biologically independent samples). FadD expression did not impact methyl ketone production indicating that free fatty acid activation is not required for PhaG-dependent methyl ketone production. D A two-dimensional cluster map created with the Enzyme Similarity Tool37 displays the sequence similarity of PhaG variants tested in bioprospecting studies. Quantitative pairwise percent amino-acid identity of each homolog can be found in Supplementary Figure 2. Colored boxes and dots are used to indicate the sequences tested. Mean methyl ketone titers for constructs of PhaG homologs using E. coli RAI harbor pTRC99a-‘PpuphaG’-EcfadM (n = 3 biologically independent samples). All cultures were grown in Clomburg medium containing 20 g/L glycerol at 30 °C and shaking at 250 r.p.m. **P = 0.009 was analyzed based on student two-tailed t test assuming unequal variances. All data represent the mean ± s.d. of biological triplicates. Source data underlying B–D are provided as a Source Data file.
Bioprospecting identifies active PhaG variants
Next, we sought to identify higher activity variants through bioprospecting. We conducted a homology search based on the Pseudomonas putida KT2440 PhaG (PpuPhaG) sequence using the Basic Local Alignment Search Tool (BLAST)36 to identify candidate protein sequences. BLAST hits were sorted using the Enzyme Similarity Tool37. Among the homologous sequences that have similarity >45%, we found >95% of sequences belong to Pseudomonas species, indicating that PhaG provides activity unique from other PHA-producing bacteria38. We selected 13 PhaG homologs, which had a protein sequence similarity range of 24–88% based on a pairwise comparison, shown in Fig. 2D and Supplementary Figure 2. We were particularly interested in homologs from Mycobacteria and Corynebacteria species because of their potential to interface with substrates linked to the ACP domain of type I fatty acid synthase (FAS) found in these species39. The activity of PhaG homologs was assayed in vivo by monitoring methyl ketone production using E. coli RADI harboring a pTRC99a-’PkphaG’-EcfadM plasmid. Most PhaG variants generated similar methyl ketone titers to PpuPhaG albeit with reduced levels of 2-heptanone. The PhaG variants from Mycobacteria and Corynebacteria failed to produce methyl ketones. We found that the P. koreensis PkPhaG showed the highest production at 350 mg/L total methyl ketone, 1.6 times higher than the PpuPhaG, shown in Fig. 2D. The methyl ketone profile included 34 mg/L 2-heptanone, 85 mg/L 2-nonanone, 85 mg/L 2-undecanone, and 145 mg/L 2-tridecanone indicating a broad activity against medium-chain 3-hydroxyacyl-ACPs. The product distribution did not vary significantly across the tested variants. Therefore, substrate preference will need to be addressed with protein engineering efforts analogous to those targeted to acyl-ACP thioesterases15.
Metabolic engineering to enhance methyl ketone production
A central tenet of metabolic engineering states that enzyme activity must be balanced across a metabolic pathway to minimize unwanted accumulation of intermediates and maximize pathway flux. In order to assess the relative activity of PhaG to pathway flux, we varied co-overexpression of PpuPhaG, the more-active PkPhaG, and enzymes that convert 3-hydroxyacyl-CoAs to 3-ketoacyl-CoAs. In particular, we were concerned about the relative activity of ketoreductases on the two 3-hydroxyacyl-CoA stereoisomers. PhaG generates (R)-3-hydroxyacyl-CoA for direct polymerization into PHA. In contrast, β-oxidation passes through (S)-3-hydroxyacyl-CoAs. FadB, the dual-function enoyl-CoA hydratase, and dehydrogenase, can isomerize (R)-3-hydroxyacyl-CoA via the corresponding enoyl-acyl-CoA, likely at a reduced rate relative to its regular substrate isomer40,41,42,43. PHA-producing bacteria solve this problem by expressing an alternative enoyl-CoA hydratase (PhaJ) that can generate the preferred R-isomer (Fig. 3A) from β-oxidation intermediates. We selected EcFadB and EcFadJ, an anaerobically expressed FadB homolog, and four previously studied PhaJ variants from P. aeruginosa44. These enzymes were important optimization points because deletion of fadB and fadJ eliminated PhaG-dependent methyl ketone production (Fig. 3B).
A Detailed metabolic pathway is used to enable PhaG-dependent production of methyl ketones. B Evaluation of alternative routes of isomerizing 3-hydroxyacyl-CoAs with respect to PhaG activity. A series of mutations were made to the base strain, E. coli RAI (MG1655, ΔaraBAD ΔfadR ΔfadA ΔfadI) (n = 3 biologically independent samples). Cultures of each strain harboring pTRC99a-PpuphaG–EcfadM or pTRC99a-PkphaG–EcfadM were grown for 48 hrs in Clomburg medium containing 20 g/L glycerol at 30 °C and shaking at 250 r.p.m. All data represent the mean ± s.d. of biological triplicates. ***P = 0.0001, **P = 0.003, *P = 0.02 were analyzed based on student two-tailed t test assuming unequal variances. Source data underlying B are provided as a Source Data file.
A combination of EcFadB, EcFadB/EcFadJ, and four PpuPhaJ homologs was cloned into an operon linked to a PTRC promoter on a pACYC vector. Each of these vectors was co-expressed with pTRC99a-PpuphaG–EcfadM or pTRC99a-PkphaG–EcfadM in E. coli RAI. Cultures of each strain were grown at 30 °C for 48 h. Methyl ketones were extracted from culture samples and quantified by GC/FID (Fig. 3). Strains expressing PpuPhaG all produced ~160 mg/L of methyl ketones with similar distributions to prior experiments. In contrast, when PkPhaG was expressed, methyl ketone titers increased two- threefold relative to the corresponding PpuPhaG strains. In this series, co-expression of PhaJ1 and PhaJ3 had the biggest impact on methyl ketone titer, surpassing 0.5 g/L in cultures of the best strains. These experiments indicate that 3-hydroxyacyl-CoA isomerization can be a limiting step when PhaG transacylase activity is increased. Further, these experiments suggested that titers could be improved with more-active PhaG enzymes.
Random mutagenesis improves PhaG activity
We initiated a protein engineering study to seek variants with enhanced activity altered and if possible altered product specificity. We constructed an error-prone PCR library of PkphaG ORFs and screened for the ability to complement a lipoic acid auxotrophy; this approach was previously used to isolate C8-specific thioesterases with enhanced Vmax14. E. coli strains lacking LipB require supplementation of lipoic acid or octanoic acid in the media to enable the formation of active pyruvate dehydrogenase and aerobic growth on glucose. We constructed a lipoic acid auxotrophic strain, E. coli CM23-ΔlipB, that also lacks β-oxidation genes and many fermentation pathways. To link the octanoic acid selection to PhaG activity, we added heterologous enzymes to convert the 3-hydroxyoctanoyl-CoA to octanoic acid (Fig. 4A). These enzymes include PkPhaG, PaPhaJ3, a Treponema denticola trans-enoyl-CoA reductase (TdTER)8 and a Mycobacterium sp. acyl-CoA thioesterase MaTesB A197D (referred as to MaTesB*)45 (Fig. 4A). Purified Mycobacterium avium MaTesB* has been shown to hydrolyze octanoyl-CoA and generate octanoic acid in vitro45. We neglected the octanoic acid synthesis activities by endogenous E. coli acyl-CoA thioesterase because EcTesB generally has activities toward longer chain acyl-CoA (>C10)41 and EcYciA has activities toward shorter chain acyl-CoA (<C8)46,47. The base strain, expressing the wild-type PkphaG, produced ~20 mg/L octanoic acids after 48 h, whereas the corresponding strain without PkPhaG produced <1 mg/L octanoic acid. In order to reduce the baseline octanoic acid titer, we subcloned PkphaG onto a low-copy vector (pBTRCK)8 and MatesB* onto a high copy number vector pTRC99a. The latter was performed to ensure that octanoic acid production would be limited solely by PhaG activity. After tuning the copy number of PkphaG and MatesB*, E. coli CM23-ΔlipB harboring pTRC99a-MatesB*–TdTER + pACYC-PaphaJ3 + pBTRCK-’PkphaG’ plasmids produced ~7 mg/L octanoic acids after 48 h. This strain was used to perform selections of the error-prone PhaG library.
A Metabolic pathways involved in PhaG complementation of a ΔlipB driven lipoic acid auxotrophy in E. coli CM23 ΔlipB. EcFadD deletion blocks the reactivation of fatty acids. EcFadBJ deletion blocks hydration and dehydrogenation of octenoyl-CoA to β-ketooctanoyl-CoA. EcFadAI deletion blocks thiolase-driven elongation or reduction of acyl-CoA chains. LipB deletion blocks activation of apo pyruvate dehydrogenase complex (PDC) E2 domain to octanoylated PDC E2 domain. (R)-3-hydroxyoctanoyl-CoA is dehydrated and hydrogenated by PaPhaJ3 and TdTER to generate octanoyl-CoA. Octanoyl-CoA is hydrolyzed by MaTesB* to release octanoic acid and CoA. Octanoic acid is ligated to Apo PDC E2 domain by LplA. LipA creates the functional lipoylated E2 domain restoring PDH activity. A functional selection was performed by introducing PhaG variants into the selection strain and plating cells on minimal MOPS-glucose agar. In the selection strain, MaTesB* and TdTER were expressed from pTRC99a, PaPhaJ3 was expressed from pACYC, PkPhaG, and other variants, were expressed from a low-copy vector, pBTRCK. B FAME analysis of cultures harboring PhaG variants containing combinations of point mutations identified in selection experiments (n = 3 biologically independent samples). E. coli CM23 harboring pBTRCK-’PkphaG’ + pACYC-PaphaJ3 + pTRC99a-TdTER–MatesB* were cultured in test tubes containing 5 mL Clomburg 20 g/L glycerol and 1 mM IPTG at 30 °C for 48 hrs. All data represent the mean ± s.d. of biological triplicates. ***P = 0.0001 (Q45R Y138F G142V vs. wild-type), ***P = 0.0006 (PkPhaG Q45R G142V vs. PkPhaG), **P = 0.004 (PkPhaG Y138F vs. PkPhaG), *P = 0.02 (PkPhaG Q45R vs. PkPhaG) and *P = 0.03 (PkPhaG G142V vs. PkPhaG) were analyzed based on two-tailed student t test assuming unequal variances. Source data underlying B are provided as a Source Data file.
In the first-round of mutagenesis, hundreds of colonies appeared three days after plating on MOPS-glucose minimal agar containing 20 μM isopropyl β-d-1-thiogalactopyranoside (IPTG). On day 4, we picked ~180 of the largest colonies and quantified the octanoic acid titer from individual liquid cultures grown in Clomburg liquid media containing 20 g/L glycerol. We found 17 PkPhaG variants increased octanoic acid titer (3.3–16.3-fold) and total fatty acid titer (1.8–8.3-fold) relative to the parent PkPhaG (Supplementary Figures 3–5 and Supplementary Method 2). The 17 improved PkPhaG variants contained a total of 28 point mutations that we recreated individually. We co-expressed each PkPhaG variant from the high copy number pTRC99a-’PkphaG’-TdTER vector with pACYC–PaphaJ3 + pBTRCK-MatesB* in E. coli CM23. Each strain was cultured for 48 hrs at 30 °C and samples were harvested for fatty-acid quantification. We found that variants containing 6 of the 28 single point mutations (e.g., Q45R, R66H, H76Y, Y138F, G142V, Q277X) increased octanoic acid titer more than twofold over strains expressing the parent PkPhaG (Supplementary Figure 6). Among these, three individual PkPhaG mutants Q45R, G142V and Y138F strains had 1.9-, 1.8-, and 1.5-fold higher total fatty acid titers than strains expressing wild-type PkPhaG.
We constructed all four possible combinations of Q45R, G142V, and Y138F mutations in a vector pTRC99a-’PkphaG’-TdTER and repeated analogous fatty-acid production experiments. The best variant, PkPhaG Q45R G142V, produced 1.1 g/L of C8–C14 free fatty acids, a 4.0-fold increase compared with the original PkPhaG (Fig. 4B). The fatty-acid pool contained 41% tetradecanoic acid, 22% dodecanoic acid, 11% decanoic acid, and 26% octanoic acid, similar to the original PkPhaG-expressing strain. This indicates that the increased production of octanoic acid was due to a general increase in activity, not selectivity.
We next repeated the experiments described in Fig. 3 to confirm that the more-active PkPhaG* (referred as to PkPhaG Q45R G142V) variant was not exceeding the downstream FadB and PhaJ activities. Strains expressing PaPhaJ3, which was present in all protein engineering experiments, and PaPhaJ1 produced the highest fatty acid titers (Supplementary Figure 7) and more than any other alternative dehydratase. Similar to the observation in Fig. 2C, deletion of FadD did not substantially affect methyl ketone titer (Fig. 5A) generated by the enhanced PkPhaG*, suggesting that the modified enzyme does not have enhanced thioesterase activity. Note, these experiments were performed with a FadM thioesterase from P. sneebia that we found had higher activity than E. coli FadM4.
A Methyl ketone production by E. coli RAI or RADI harboring pTRC99a-PkphaG*-PsfadM + pACYC-PaphaJ3 plasmids. B Fatty alcohol production by E. coli RADI harboring pTRC99a-PkphaG*-TdTER + pACYC-PaphaJ3 + pBTRCK-MaACR. The corresponding negative control harbored a pTRC99a empty vector in place of the PhaG* vector. All cells were grown in 50 mL of Clomburg media containing 20 g/L glycerol, 20% (v/v) dodecane, and 1 mM IPTG at 30 °C for 48 hrs (n = 3 biologically independent samples). The shadings and non-shadings in graphs represented products in the dodecane organic layer and aqueous phase, respectively. All data represent the mean ± s.d. of biological triplicates. Source data underlying A, B are provided as a Source Data file.
Benchmarking PhaG-driven oleochemical production
Next, we benchmarked PhaG-driven production of methyl ketones and fatty alcohols as model oleochemical products. To produce methyl ketones, we cultured E. coli RADI harboring pTRC99a-PkphaG*-PsfadM and pACYC-PaphaJ3. To produce fatty alcohols, we cultured strain E. coli RADI harboring pTRC99a-PkphaG*-TdTER, pACYC-PaphaJ3, and pBTRCK-MaACR. A bi-functional alcohol forming acyl-CoA reductase MaACR from M. aquaeolei VT8 was chosen because of its successful use in prior studies8. Each strain was cultured for 72 hr at 30 °C in shake flasks containing rich glycerol media. The cultures produced 1.5 g/L total methyl ketone and ~1.1 g/L total fatty alcohol from 20 g/L glycerol (Fig. 5). The chain length distribution of each product was different. Methyl ketones were evenly distributed across 8–16-carbon chain lengths, whereas, fatty alcohols were dominated by 14- and 16-carbon alcohols. The titer and yield of fatty acids, methyl ketones, and fatty alcohols from the PhaG-dependent strategies are comparable to values reported in the literature using thioesterase strategies or reverse β-oxidation strategies to produce products in shake flasks (Table 1).
To further improve methyl ketone titers, we performed discontinuous fed-batch cultivations in a stirred bioreactor by adding media pulses after cells reached high cell densities. Enhanced aeration in bioreactors can lead to loss of volatile oleochemical products in the off-gas. Therefore, we added a 20% dodecane overlay to the culture4 to trap products in the reactor and designed a gas trap to recover products from the off-gas. Details of off-gas methyl ketone capture and ASPEN analysis can be found in the Supplementary information file (Supplementary Figure 8 and Supplementary Method 3). After 96 hrs of cultivation in reactors lacking an off-gas scrubber, cells consumed 90.0 g/L of glycerol, reached an OD600 ~60, and produced 6.8 g/L total methyl ketone (Fig. 6A). In a separate experiment, we bubbled the bioreactor off-gas through a jacketed gas dryer filled with dodecane maintained at 5 °C. After 96 hrs of induced cultivation, cells reached a density of OD600 ~60, consumed 95.2 g/L glycerol, and produced 6.7 g/L total methyl ketone (Fig. 6B). At the endpoint of the cultivation, we observed 3.4 g/L methyl ketone contained in the condensed dodecane phase of the off-gas trap, corresponding to an additional 0.51 g/L (per L of culture volume) of total methyl ketone captured from the bioreactor (Fig. 6C).
Time course of glycerol consumption, OD600, and methyl ketone titer from fed-batch bioreactor cultures (n = 3 biologically independent samples). A and bioreactor coupling a condenser (n = 2 biologically independent samples) B using E. coli RADI harboring pTRC99a-PkphaG*-PsfadM and pACYC-PaphaJ3 plasmids. C Evaluation of methyl ketone concentration from samples in the aqueous phase (dark gray), dodecane layer (light gray), and condenser (shading) after 96 h fermentation. All data represent the mean ± s.d. of biological triplicates. Source data underlying A–C are provided as a Source Data file.
This study demonstrated that PhaG is capable of supporting high-flux to three medium-chain oleochemical products, fatty acids, fatty alcohols, and methyl ketones. Through bioprospecting, we identified a PhaG from P. koreensis that demonstrated higher activity when producing methyl ketones. Random mutagenesis of PkPhaG produced 17 enhanced PhaG variants that were isolated by complementing lipoic acid auxotrophy. The improved activity was demonstrated by 3–16-fold more octanoic acid being produced when the PhaG variants were co-expressed in a cell designed to direct flux to octanoic acid (co-expressing TesB, PhaJ3, and Ter). Reconstitution of the individual point mutations led to the creation of a double mutant, PkPhaG Q45R G142V, that showed 4.0-fold higher activity relative to the parent enzyme. Finally, we demonstrated the production of 1.1 g/L C8–C14 free fatty acids, 1.5 g/L C7–C15 methyl ketones, and 1.1 g/L C8–C16 fatty alcohols in shake flasks and 7.2 g/L of methyl ketones in a fed-batch. These results demonstrate that PhaG is a viable alternative strategy that should be considered for oleochemical production. The PhaG-dependent strategy has the potential to achieve higher theoretical yields compared with the well-established thioesterase route. However, this impact, saving one ATP per product, is not likely to be observed until cells approach theoretical limits. Given the current state of the field, additional work is needed on both fronts. Our work motivates continued strain development as well as additional protein engineering15,48,49 to narrow the PhaG product profile and further increase activity.
