Ketosynthase mutants enable short-chain fatty acid biosynthesis in E. coli

Abstract

Cells build fatty acids in tightly regulated assembly lines, or fatty acid synthases (FASs), in which β-ketoacyl-acyl carrier protein (ACP) synthases (KSs) catalyze sequential carbon-carbon bond forming reactions that generate acyl-ACPs of varying lengths—precursors for a diverse set of lipids and oleochemicals. To date, most efforts to control fatty acid synthesis in engineered microbes have focused on modifying termination enzymes such as acyl-ACP thioesterases, which release free fatty acids from acyl-ACPs. Changes to the substrate specificity of KSs provide an alternative—and, perhaps, more generalizable—approach that focuses on controlling the acyl-ACPs available for downstream products. This study combines mutants of FabF and FabB, the two elongating KSs of the E. coli FAS, with in vitro and in vivo analyses to explore the use of KS mutants to control fatty acid synthesis. In vitro, single amino acid substitutions in the gating loop and acyl binding pocket of FabF shifted the product profiles of reconstituted FASs toward short chains and showed that KS mutants, alone, can cause large shifts in average length (i.e., 6.5–13.5). FabB, which is essential for unsaturated fatty acid synthesis, blunted this effect in vivo, but exogenously added cis-vaccenic acid (C18:1) enabled sufficient transcriptional repression of FabB to restore it. Strikingly, a single mutant of FabB afforded titers of octanoic acid as high as those generated by an engineered thioesterase. Findings indicate that fatty acid synthesis must be decoupled from microbial growth to resolve the influence of KS mutants on fatty acid profiles but show that these mutants offer a versatile approach for tuning FAS outputs.

1. Introduction

Fatty acid biosynthesis is a centrally important metabolic process that generates lipophilic precursors for membrane assembly, cofactor production, energy storage, and signaling (Bogdanov et al., 2008; Guo et al., 2020; Ingólfsson et al., 2014; Muro et al., 2014; Sztalryd and Kimmel, 2014; Wakil and Abu-Elheiga, 2009). In both prokaryotic and eukaryotic cells, multi-enzyme assembly lines called fatty acid synthases (FASs) coordinate the iterative condensation and reduction of two-carbon extender units to generate acyl-acyl carrier proteins (ACPs) of varying lengths—the building blocks of membrane lipids and oleochemicals (Liu et al., 2021; Schweizer and Hofmann, 2004; Smith and Tsai, 2007; White et al., 2005). Ketosynthases (KSs) catalyze the carbon-carbon bond forming reaction at the core of this process; by controlling the profile of available acyl-ACPs, they regulate flux to different products (Chen et al., 2011; Gajewski et al., 2017b; Torella et al., 2013; Zhu et al., 2017, 2020). A detailed understanding of the molecular determinants of substrate specificity in KSs, and the influence of that specificity on FAS product profiles, could inform the design of microbes that produce valuable oleochemicals and other KS-dependent products (e.g., polyketides, an important class of secondary metabolites; (Gajewski et al., 2017a; Klaus et al., 2020)).

KSs can be classified loosely into two groups—initiating or elongating—and five families, depending on sequence and tertiary structure (Chen et al., 2011). Initiating KSs catalyze the first condensation reaction, which condenses acetyl-CoA (or propionyl-CoA) with malonyl-ACP to form short-chain β-ketoacyl-ACPs. Elongating KSs extend acyl-ACPs by condensing them with malonyl-ACP, which adds two carbons. The FAS of E. coli is a well-studied and industrially relevant model system comprised of discrete monofunctional enzymes (i.e., a type II system); this FAS has one initiating KS, denoted FabH, and two elongating KSs, FabB and FabF (Edwards et al., 1997; Jeffrey L. Garwin et al., 1980; J.L. Garwin et al., 1980; Lai and Cronan, 2003; Tsay et al., 1992). The elongating KSs have similar substrate specificities but differ in two ways: (i) FabB can elongate cis-dec-3-enoyl-ACP, which initiates unsaturated fatty acid synthesis, and (ii) FabF is more active than FabB on C16:1-ACP (Feng and Cronan, 2009; J L Garwin et al., 1980). In vitro and in vivo studies indicate that both enzymes can also initiate fatty acid synthesis without FabH—a commonly ignored secondary activity (Alberts et al., 1972; Borgaro et al., 2011; Mains et al., 2022; McGuire et al., 2001). It is the unique contribution of FabB and FabF to elongation, however, that makes them attractive targets for tuning the product profiles of FAS-based pathways.

We will briefly detail the enzymology of FabB and FabF before zooming out to focus on their metabolic influence. Recent biostructural studies have elucidated the molecular mechanisms of FabB- and FabF-catalyzed elongation reactions in exceptional detail (SI Note 1; (Mindrebo et al., 2021, 2020a, 2020b)). Both enzymes use a ping-pong mechanism comprised of two major steps. First, they bind to acyl-ACP, which transfers an acyl-chain from its phosphopantetheine arm to a cysteine to form an acyl-enzyme intermediate and holo-ACP (Fig. S1). Second, they bind to malonyl-ACP, which displaces holo-ACP and undergoes a decarboxylative Claisen-like condensation with the acyl-enzyme to generate β-ketoacyl-ACP. X-ray crystallography, molecular dynamics (MD) simulations, and mutational analyses indicate that two loops, denoted “loop 1” and “loop 2”, maintain reaction order. Loop 1 has a gating residue (F400) that blocks the active site of the apo enzyme; when acyl-ACP binds, this loop opens to yield a transient pocket for acyl chain delivery to the active site. Closure of loop 1 over the bound substrate directs the acyl chain to an acyl binding pocket and restores an oxyanion hole, which helps stabilize tetrahedral intermediates formed during the acylation and condensation half reactions. Loop 2 interacts with both loop 1 and bound ACP, and may serve as an allosteric relay system that coordinates the related processes of gate opening and substrate binding—though additional studies are necessary to resolve this allosteric functionality in detail (Mindrebo et al., 2020b). Both loops have residues that are conserved between or within KS families (loops 1 and 2, respectively), an indication of their broad importance for the catalytic activity—and, perhaps substrate specificity—of KS enzymes (Mindrebo et al., 2021).

In this study, we used recent crystallographic insights to guide the design of FabF and FabB mutants with altered substrate specificities, and we used these mutants to examine the influence of KS mutations on FAS activity. Our analyses allowed us to test recently refined theories of KS enzymology and probe the metabolic adjustments that buffer cells against changes in acyl-ACP elongation, an essential and highly regulated biochemical process. Findings provide important lessons for using protein engineering to control microbial fatty acid synthesis.

2. Materials and methods

2.1. Model predictions

We modeled fatty acid synthesis by using a previously developed kinetic model that captures the activities of nine enzymes necessary to convert malonyl-CoA to free fatty acids. The model includes eight enzymes from the E. coli FAS, the acyl carrier protein (ACP), a cytosolic variant of the periplasmic thioesterase TesA, acyl-ACP intermediates, pathway substrates (acetyl-CoA and malonyl-CoA), and free fatty acids (Ruppe et al., 2020). The model uses rate equations based on detailed reaction mechanisms supported by experimental measurements of isolated enzymes and avoids a priori equilibrium assumptions by using separate association and dissociation steps for all heteromeric complexes. To frame our modeling analyses, we defined a reference condition consisting of 1 μM of each Fab enzyme, 10 μM TesA, 10 μM ACP, 500 μM malonyl-CoA, 500 μM acetyl-CoA, 1 mM NADPH, and 1 mM NADH. The relevance of these concentration ratios to in vivo fatty acid production is supported by measurements of intracellular concentrations of FAS components, and by the consistency of prior in vitro and in vivo studies of component-specific changes in concentration on FAS activity (Mains et al., 2022; Ruppe et al., 2020; Xiao et al., 2013; Yu et al., 2011). We solved the kinetic model at different FabF and FabB concentrations with MATLAB solver ode15s with relative and absolute error tolerances of 10⁻⁶ and a vectorization step to reduce solve time.

2.2. Materials and reagents

We purchased the substrates malonyl CoA (malonyl coenzyme A lithium salt) and acetyl CoA (acetyl coenzyme A sodium salt) from Millipore Sigma, and NADPH (Nicotinamide adenine dinucleotide phosphate) from Cayman Chemical. We obtained inducers isopropyl β-d-1-thiogalactopyranoside (IPTG) and anhydrotetracycline (aTC) from Thermo Fisher Scientific, and arabinose from Acros Organics. We bought fatty acid standards from Acros Organics (decanoic acid, myristic acid and pentadecanoic acid), Alfa Aesar (dodecanoic acid), and Millipore Sigma (all others); antibiotics from Thermo Fisher Scientific (carbenicillin, kanamycin sulfate and tetracycline) and Millipore Sigma (spectinomycin and chloramphenicol); and media components from Thermo Fisher Scientific (tryptone, yeast extract, and S.O.C media). For cloning, we purchased all necessary enzymes from New England Biolabs. We used NEB^® Turbo Electrocompetent and BL21 (DE3) E. coli cells (New England Biolabs) for cloning and protein expression, respectively. We purchased all chromatography columns from GE Healthcare.

2.3. Protein expression and purification

We overexpressed all FAS enzymes (FabA, FabB, FabD, FabF, FabG, FabH, FabI, FabZ), ACP, and ‘TesA (i.e., a leaderless, cytosolic variant of TesA), which we denote simply as “TesA” in this study, in E. coli as described previously (Mains et al., 2022). We began by adding FAS genes with N-terminal polyhistidine tags to pET vectors. Using these plasmids, we transformed chemically competent BL21 (DE3) cells, recovered the cells in S.O.C. (Super Optimal broth with Catabolite repression) media (2% tryptone, 0.5% yeast extract, 10 mM NaCl, 2.5 mM KCl, 10 mM MgCl₂, 10 mM MgSO₄, and 20 mM glucose) for an hour, plated them on Luria–Bertani (LB) agar plates supplemented with antibiotics (10 g/L tryptone, 5 g/L yeast extract, 10 g/L sodium chloride; 50 μg/mL carbenicillin or kanamycin), and grew the cells at 37 °C overnight. Using individual colonies, we inoculated 20 mL cultures of LB media containing antibiotics (50 μg/mL carbenicillin or kanamycin) and incubated them at 37 °C and 225 rpm until the cultures become cloudy (~5 h). We diluted the 20-mL cloudy cultures into 1 L of rich induction media supplemented with antibiotics (20 g tryptone, 10 g yeast extract, 5 g sodium chloride, 0.4% glucose, 72 mL of 5X M9 solution; 50 μg/mL carbenicillin or kanamycin) and grew the 1-L cultures at 37 °C and 225 rpm until the OD₆₀₀ reached 0.4–0.8 (with a path length of 1 cm). At this OD, which is indicative of mid-exponential growth, we added 0.5 mM IPTG and transferred the cultures to 22 °C and 225 rpm for another 14–18 h. We harvested cells by centrifuging them at 5000 rpm for 15 min, and we froze cell pellets at −80 °C for future use.

We purified proteins using a fast protein liquid chromatography (FPLC) instrument (AKTA Pure): (i) We lysed cell pellets with standard lysis buffer (per gram cell pellet: 4 mL B-PER (Thermo Fisher Scientific), 2 mg TAME, 2 mg magnesium sulfate heptahydrate, 3.5 mg TCEP, and 3.75 μl PMSF, 0.5 mg lysozyme, and 200 units DNase I (New England Biolabs)), (ii) isolated protein with a nickel-affinity column (HisTrap HP column with 50 mM Tris-HCl, pH 7.5, 0.5 mM TCEP, 300 mM NaCl, and 0–500 mM imidazole), (iii) buffer exchanged the isolated protein with a desalting column (a HiPrep 26/10 column with 50 mM Tris-HCl, pH 7.5 or 8.5, and 0.5 mM TCEP), (iv) purified the isolated protein further with an anion-exchange column (HiPrep Q HP 16/10 column with 50 mM Tris-HCl, pH 7.5 or 8.5, 0.5 mM TCEP, and 0–1 M NaCl), and (v) carried out a final buffer exchange prior to freezing (HiPrep 26/10 as before). We concentrated purified proteins with a spin concentrator (5 or 10 kDa VivaSpin; Sartorius), measured final concentrations with a Bradford assay using bovine serum albumin (BSA) as a standard, and stored them in 20% glycerol at −80 °C. We note: For FabZ, we used only nickel-affinity chromatography, which was sufficient to achieve single-band purity with SDS-PAGE.

2.4. In vitro analysis of fatty acid synthases

We studied the influence of mutations on FabF activity by replacing wild-type FabF in reconstituted FASs. To focus on the contribution of FabF, we used a FabB-less reference system (1 μM of a FabF variant, 0 μM FabB, 1 μM all other Fab enzymes, 10 μM TesA, 10 μM holo-ACP, 10 mM HEPES, pH 7.4, 150 mM NaCl, 0.5 mM TCEP). For initial rates, we used 100 μL reactions; for product profiles, we scaled them up to 400 μL. We initiated reactions by adding a mixture of substrates and cofactors (initial rates: 1.3 mM NADPH, 0.5 mM malonyl-CoA, and 100 μM acetyl-CoA; product profiles: 2.6 mM NADPH, 1.5 mM malonyl-CoA, and 300 μM acetyl-CoA) and let them run at room temperature (~21 °C) for either 2.5 min (initial rates) or 2 h (product profiles). For initial rates, we monitored the conversion of NADPH to NADP⁺. We measured absorbance (340 nm) in a SpectraMax iD3 multi-mode plate reader, as described in prior work (Xiao et al., 2013), converted the absorbance values to NADPH via a standard curve (Fig. S13), and used the ratio 14 μM NADPH:1 μM palmitic acid to calculate initial rates over the 2.5-min period in “palmitic acid equivalents per minute”. We examined each reaction with at least three independently reconstituted mixtures, and we used negative controls to correct for any background signal. To examine product profile, we used the GC-MS method described below.

2.5. Plasmid construction

We received pCDM4 (Addgene #49796) from Mattheos Koffas; pBad18-TesA-R3M4 (TesA_C8) from Brian Pfleger; pAM053 from Shelley Copley; and ptsss9-gRNA and pSS9 (Addgene #71655) from Ryan T. Gill. For transcriptional repression of FabB, we used CRISPR interference (CRISPRi): using pFD152-FabB, we expressed a catalytically inactive Cas9 (dCas9) and a guide RNA that targets the promoter region of fabB in the E. coli genome (Mains et al., 2022). We employed Gibson Assembly to build all new plasmids listed in Table S1 by using the primers listed in Table S2. We used a gblock from integrated DNA technologies as the template for the tacO1 promoter in pCDM4-tacO1-FabF. (Browning et al., 2019).

2.6. Strain construction

We received the strain RL08ara, a derivative of K-12 MG1655 lacking genes fadD and araBAD, as a gift from Brian Pfleger (Lennen et al., 2010). Removal of fadD enhances the production of medium-chain fatty acids, and the removal of araBAD enables arabinose titration of genes under control of the pBAD promoter. We used this strain as the chassis for our in vivo studies.

For genomic modification of the fabF gene, we paired lambda red recombination with a Cas9-gRNA system, as described by Bassalo et al. (2016). To inactivate fabF, we used a separate recombination template that replaces fabF with TetA, a gene for tetracycline resistance. We began template construction by building a plasmid with fabF homology arms (pKMM-fabF HA); briefly, we isolated the genomic DNA of RL08ara using the DNeasy Blood and Tissue kit (Qiagen), amplified the fabF gene with primers that target 300 bp up- and downstream of fabF, and integrated the cloned gene into pBad18 with Gibson assembly. Next, we constructed a plasmid in which TetA is flanked by homology arms for the region of the genome that flanks fabF (pKMM-fabF HA-Tet-SPM-p189). Here, we amplified TetA from pSS9 (Addgene #71655) and used Gibson assembly to swap it for fabF in pKMM-fabF HA, taking care to add a synonymous PAM mutation (SPM) to prevent Cas9 from cutting after recombination. We generated our target dsDNA template, in turn, by (i) amplifying TetA along with the 300 bp genomic overlaps from pKMM-fabF HA-Tet-SPM-p189, (ii) purifying this amplicon using DNA gel electrophoresis, and (iii) extracting the DNA using an Omega E.Z.N. A.^® Gel Extraction Kit. Finally, we constructed our gRNA system by using Gibson assembly to build ptsss9-gRNA-FabF-p189, a plasmid that constitutively expresses a guide RNA (gRNA) targeting the promoter region of fabF. To mutate—rather than inactivate—the fabF gene (i.e., I108F and G198F), we constructed a single plasmid with both template and gRNA segments (Garst et al., 2017). Here, we used Gibson assembly to build CREATE plasmids, which contain both (i) gRNAs that target fabF immediately upstream of the desired point mutation and (ii) a 120-bp fragment of fabF overlap with the desired point mutation accompanied by an SPM.

We constructed our modified strains with a standard CRISPR/Cas9 mediated lambda red recombination protocol. In brief, we transformed electrocompetent RL08ara cells with pAM053 (which contains Cas9 and lambda red proteins), recovered these cells in S.O.C. media, plated them on LB plates supplemented with antibiotics (34 μg/mL chloramphenicol), and grew them overnight at 30 °C. We inoculated 5 mL LB media (34 μg/mL chloramphenicol) with a single colony and grew the resulting culture overnight at 30 °C and 225 rpm. We used 500 μL of this overnight culture to inoculate 50 mL of LB (34 μg/mL chloramphenicol), which we grew at 30 °C and 225 rpm to an OD₆₀₀ of 0.4–0.6. We induced expression of lambda red machinery by shaking the 50-mL culture at 42 °C for 15 min and 0 °C (i.e., on ice) for 5 min. Immediately after induction, we made the cells electrocompetent by carrying out three steps: (i) We centrifuged the cells (4000g for 10 min at 4 °C), removed the supernatant, and resuspended them in 25 mL water (ice cold). (ii) We repeated step i). (iii) We repeated step i) but resuspended cells in 1 mL water (ice cold).

We transformed 100 μL aliquots of electrocompetent cells with either both (i) ptsss9-gRNA-FabF p189 (Table S1) and (ii) the gel-purified tetracycline resistance cassette with genomic overlap, or (iii) a fabF CREATE plasmid (I108F or G198F). We plated the recovered cells on LB agar (carbenicillin (50 μg/mL), chloramphenicol (34 μg/mL) and tetracycline (fabF inactivation only; 10 μg/mL)), grew them overnight at 30 °C to select for positive knockout strains, and used individual colonies to inoculate 5 mL of LB media and grew the cultures at 37 °C to cure the plasmids. We extracted genomic DNA from the selected strains with the DNeasy Blood and Tissue kit (Qiagen) and verified the presence of the tetracycline resistance gene or desired point mutation via PCR amplification and Sanger sequencing (Fig. S14). We stored verified knockout strains as glycerol stocks (20% glycerol).

Generating a FabF H268A mutant strain with the above method was unsuccessful, so we used an alternative approach. Briefly, after replacing the fabF gene with tetA, we repeated the recombineering process with ptsss9-gRNA-ACP 227, which expresses a guide RNA targeting the end of the ACP gene (which is directly upstream of the fabF gene in the RL08ara genome), and a dsDNA template (amplified from pKMM-fabF HA-SPM ACP 227-FabF mutant) containing a corresponding SPM, the fabF mutant gene, and ~300 bp genomic overlap on each side of the mutant. This alternative approach for modifying the genome worked quickly, so we used it for all other mutant strains generated in this study.

2.7. Fatty acid biosynthesis

We carried out fatty acid production with RL08ara and its engineered derivatives cultured in 15-mL culture tubes (5 mL working volume) as described previously (Mains et al., 2022). We transformed each electrocompetent strain with the plasmids of interest, grew them on LB agar (100 μg/mL spectinomycin, 50 μg/mL carbenicillin, 50 μg/mL kanamycin, and/or 10 μg/mL tetracycline) at 37 °C overnight. We inoculated 5 mL LB media (100 μg/mL spectinomycin, 50 μg/mL carbenicillin, 50 μg/mL kanamycin, and/or 10 μg/mL tetracycline) with individual colonies and grew the resulting 5-mL cultures overnight at 37 °C and 225 rpm. We diluted these overnight cultures 1:200 (v/v) into 5 mL fresh LB (100 μg/mL spectinomycin, 50 μg/mL carbenicillin, 50 μg/mL kanamycin, and/or 10 μg/mL tetracycline, 0.5% glycerol). For cultures requiring cis-vaccenic acid (C18:1), we diluted our C18:1 stock solution (20 mg/mL C18:1 in ethanol) in LB media to a final concentration of 80 μg/mL. We grew these “production” cultures at 37 °C and 225 rpm to an OD₆₀₀ of 0.4–0.6 and induced fatty acid production by adding 0–1000 μM IPTG, 0–0.1% (w/v) arabinose and/or 0–0.25 μg/mL aTC. Immediately after induction, we moved the culture to 30 °C for 24 h. We harvested cells for fatty acid measurements at 24 h. We performed each cell-based measurement at least three times.

2.8. Extraction, esterification, and quantification of fatty acids

We prepared extracted fatty acids for GC/MS analysis by using the procedure described in prior work (Mains et al., 2022). We quantified fatty acids present in unknown samples with standard curves for each fatty acid. We calculated (i) total fatty acid production (μM) by summing the concentrations of all measured fatty acids, (ii) average chain length (C_avg) by multiplying the fraction of each chain length by the number of carbons in that chain and summing the resulting value for all chain lengths reported, and (iii) standard deviation of fatty

$$\text{SD}=\sqrt{\frac{\sum \left\{{\left(\text{n}-{\text{C}}_{\text{avg}}\right)}^2*\left[{\text{C}}_{\text{n}}\right]\right\}}{\left(\sum \left[{\text{C}}_{\text{n}}\right]-1\right)}}$$ (1)

acid profiles using Eq. (1), where n is the number of carbons in each fatty acid (n = 4 + 2i for i = 1 to 7), and [C_n] is the concentration (μM) of each length.

3. Results

3.1. Mutation-derived changes in FabF substrate specificity alter FAS product profiles in vitro

Our study begins with FabF. Unlike FabB, which is necessary for unsaturated fatty acid biosynthesis, FabF is not essential for E. coli growth (i.e., its deletion is not lethal). It is a promising starting point for protein engineering (Fig. 1A). Guided by crystal structures of FabF bound to dodecanoic acid (C₁₂) and ACP_C16, we sought to change the substrate specificity of this enzyme by mutating residues in loop 1, loop 2, and the acyl binding pocket (Fig. 1A–1B; SI Note 2).

Fig. 1. Mutants of FabF alter the product profiles of reconstituted FASs.

(A) A depiction of the fatty acid synthase (FAS) of E. coli with FabF highlighted in gray. (B) Crystal structures of FabF bound to C16 crypto-ACP (PDB: 6OKG; left) or C12 (PDB: 2GFY; right). Highlights: FabF (gray), ACP (teal), the acyl chain (blue), the catalytic residues (dark blue), and residues selected for site-directed mutagenesis in loop 1 (yellow), loop 2 (orange), and the acyl binding pocket (red). Bottom: Two-dimensional illustrations depict conformational changes in the active site associated with acyl chain (left) delivery and (right) transfer. Closed circles depict the approximate position of mutations. (C) The product profiles of FASs reconstituted in vitro from purified enzymes (2 h). Each reaction consists of 2.6 mM NADPH, 1.5 mM malonyl-CoA, 0.3 mM acetyl-CoA, 10 μM ACP, 10 μM TesA, 1 μM of each Fab, and 1 μM of a FabF variant: wild-type (WT) or mutant. The reference system includes 1 μM FabB; all other systems have no FabB. Colors denote the position of mutations in accordance with B. Each plot depicts the mean, standard error (SE), and individual measurements for n ≥ 3 biological replicates.

To characterize each mutant, we examined the product profiles of experimentally reconstituted FASs containing a single FabF variant (and no FabB). To our satisfaction, these in vitro systems produced fatty acids with average chain lengths of 6–12 carbons (Fig. 1C). The two mutations in loop 1 were the least influential; they reduced the production of C14 fatty acids but otherwise maintained broad profiles. Mutations in loop 2 and the acyl binding pocket yielded more distinct changes, though all, with the exception of H268S, increased the production of C8. Broadly, our in vitro analyses show that mutation-derived changes in the activity of FabF can alter the product profiles of FAS systems.

Our prior work indicates that changes in the ratio of FabF to TesA can shift average chain length by altering relative rates of acyl-ACP elongation and termination. Mutations that reduce the activity of FabF, in turn, could plausibly mimic low FabF concentrations. We explored this potential mutational effect by measuring initial rates of fatty acid synthesis (Fig. S3A). In general, most FASs with altered product profiles exhibited initial rates that were similar to—that is, at least 60% of—the wild-type rate, though two mutants yielded slightly lower rates: T403Y and G198M (43% and 57% of wild-type). Indeed, these rates were similar to the rate of a FAS with ten-fold less FabF (46% of wild-type); however, the product profiles afforded by the two mutants were quite different (Fig. S3B; Mains et al., 2022). In general, we observed no correlation between initial rates and product profiles, an indication that mutations in FabF exert their effect by altering its substrate specificity, rather than its catalytic activity.

The product profiles of influential mutants are consistent with recent crystal structures. Mutants A193M and I108F, which were reported in prior work (Val et al., 2000), and A193V, L111M, and I108M, which are new to this study, yielded narrow profiles centered at C8, while G198M and G198F generated similar amounts of C4-C8 (Fig. 1C). The influence of these mutations, which restrict elongation beyond C8, suggests that C6 sits just above the acyl binding pocket—a position consistent with the crystal structure of the FabF-C12 complex. Mutations in loop 2, by contrast, enhanced the production of short-chain fatty acids but yielded broader product profiles. Notably, H268A and T270F produced more C12 and C4, respectively, than other mutants. When loop 2 is closed, residue H268 participates in a main-chain hydrogen bond with S271 and a side-chain bond with T270. When the loop is open, the bonding pattern flips from H268_(H)N-S271(O)C to H268_(O)C-S271_(N)H, and T270 breaks its hydrogen bond with H268 to interact with D35 on ACP. The H268A mutation may destabilize the closed conformation of loop 2, while T270F might destabilize both. Both effects appear to disfavor activity on long-chain acyl-ACPs. H268S, in turn, may retain the side-chain hydrogen bond with T270 in the closed conformation, allowing FabF to elongate medium- and long-chain acyl-ACPs.

3.2. FabB limits the influence of genomically integrated FabF mutants

In the complete FAS, FabB could obscure the influence of FabF mutants by elongating short-chain acyl-ACPs. We observed this “buffering” phenomenon in our study of ratiometric tuning; FabB compensated for a reduction in FabF concentration (Mains et al., 2022). We used our in vitro system to reexamine this behavior with FabF mutants. As expected, the addition of FabB to mutant-containing FASs increased average chain length, even for a selective mutant like I108F, and high concentrations of I108F did not restore its influence (Fig. S4). Next, we used our detailed kinetic model to examine different concentrations of FabB. Indeed, we found that even low concentrations of FabB were sufficient to blunt the influence of FabF mutants (Figs. S5 and S9A). Taken together, the results of our in vitro and in silico analyses indicate that strong transcriptional repression of fabB is required to amplify shifts in product profiles afforded by mutants of FabF in cells.

We evaluated the influence of FabF mutants in vivo by integrating them into the E. coli genome. We focused initially on I108F and G198F, two mutants with extreme product profiles: narrow production centered at C8, and fairly even across C4-C8, respectively. We swapped each mutant for the native fabF gene and transformed the modified strains with plasmid-borne control systems for TesA and FabB (Fig. 2A): The TesA system enabled arabinose-inducible transcriptional activation of tesA; the FabB plasmid, anhydrotetracycline-inducible repression of fabB. Without fabB repression, the product profiles afforded by I108F, G198F, and wild-type FabF were nearly indistinguishable (Fig. 2B). Notably, FabF mutants failed to produce cis-vaccenic acid (C18:1), an indication of successful genomic integration, which we also confirmed with Sanger sequence verification (Fig. S14); FabB cannot generate this product (Garwin et al., 1980). Nonetheless, the distribution of other products remained unchanged. Transcriptional repression of fabB, in turn, enhanced the production of C8 fatty acids for all strains but still failed to yield detectable differences in their product profiles. All three produced C12 as the dominant product (Fig. 2B–C). We used proteomics to confirm the functional integrity of our FabB control system. Indeed, transcriptional repression reduced the intracellular concentration of FabB by five-fold (Fig. S6A). Perhaps most importantly, it caused overall titers to drop for all strains (23–34%, p < 0.05)—an effect consistent with an observed reduction in growth rate (Fig. S6B) and the established essentiality of fabB (Cronan et al., 1969). This effect indicates that further repression of fabB is unlikely to improve the titers of short-chain fatty acids while this enzyme remains essential for cellular growth.

Fig. 2. Genomic integration of FabF mutants has little influence on product profiles.

(A) We swapped two FabF mutants—G198F and I108F—in place of fabF in the genome of E. coli RL08ara. Plasmid-borne control systems for TesA and FabB enabled inducible tesA activation and fabB repression (pBad18-TesA and pFD152-FabB, respectively). Genes: β-ketoacyl-ACP synthase II (fabF), thioesterase 1 (tesA), and β-ketoacyl-ACP synthase I (fabB) (B) Repression of fabB (i.e., 0.25 μg/ml anhydrotetracycline, or aTC) increased the mole fraction of short chains and decreased titers for both wild-type and mutant strains (RL08ara with pBad18-TesA and pFD152-FabB, 0.1% arabinose, and 0 or 0.25 μg/mL aTC). The mutant strains do not produce C18:1—an indication of successful genomic integration of the mutant genes—but fail to shift overall product profiles (Garwin et al., 1980). (C) The product profiles associated with the 0.25 μg/mL aTC case from B. In B-C, plots of fatty acid production depict the mean, SE, and individual measurements for n ≥ 3 biological replicates. Plots of length depict the mean and SE for n ≥ 3 biological replicates. We measured fatty acids at 24 h after induction (30 °C).

3.3. FabB limits the influence of plasmid-borne FabF mutants

We speculated that genomically integrated mutants of FabF (i.e., a single genomic copy) might express poorly, so we explored plasmid-borne control systems. We removed fabF from the genome (i.e., we replaced it with tetA, a gene for tetracycline resistance) and transformed the ΔfabF strain with plasmids for modulating the expression of TesA, FabB, and FabF. The FabF control system enabled IPTG-inducible expression of FabF variants. Intriguingly, the removal of fabF from the genome caused a nearly imperceptible increase in average chain length (7%, p < 0.05; Fig. 3B) and a large increase in unsaturated fraction (46%, p < 0.05; Fig. 3C), suggesting that FabB is overexpressed. Proteomics results confirm this interpretation; the intracellular concentration of FabB was approximately three-fold higher in the ΔfabF strain, even under transcriptional repression of fabB (Fig. S8). This increase highlights the importance of native transcriptional circuits in controlling FabB levels.

Fig. 3. Expression of FabF mutants from plasmids has little influence of product profiles.

(A) We replaced fabF with tetA, a gene that confers resistance to tetracycline, in the genome of E. coli RL08ara. Plasmid-borne control systems for TesA, FabB, and FabF allowed us to adjust intracellular concentrations of these enzymes (pBad18-TesA, pFD152-FabB, and pCDM4-tacO1-FabF*, respectively). The FabF system, which is new in this figure, enables IPTG-inducible transcriptional activation of fabF genes (wild-type and mutant). Genes: β-ketoacyl-ACP synthase II (fabF), thioesterase 1 (tesA), and β-ketoacyl-ACP synthase I (fabB), tetracycline resistance protein (tetA). (B-C) In E. coli, the removal of fabF (i.e., ΔfabF) causes (B) a nearly imperceptible increase in average chain length (4–7%) and (C) a sub-stantial increase in unsaturated fraction (RL08ara and RL08ara ΔfabF with pBad18-TesA and pFD152-FabB, 0.1% arabinose, and 0 or 0.25 μg/mL aTC). (D) The plasmid-borne expression of wild-type and I108F variants of FabF in the ΔfabF strain has a marginal influence on product profile (RL08ara ΔfabF with pBad18-TesA, pFD152-FabB, and pCDM4-tacO1-FabF* grown with 0.1% arabinose, 0.25 μg/mL aTC, and 0 or 100 μM IPTG). In B-D, plots of production depict the mean, SE, and individual measurements for n ≥ 3 biological replicates. Plots of length depict the mean and SE for n ≥ 3 biological replicates. For all plots, we measured fatty acids at 24 h after induction (30 °C).

The overexpression of both wild-type and I108F variants of FabF from plasmids yielded minor shifts in product profiles (i.e., a nearly imperceptible decrease in average length, 3–6%; p < 0.05), even under transcriptional repression of fabB (Fig. 3D). Intrigued by these shifts, we used proteomics to check FabF expression. To our surprise, the intracellular concentrations of FabF afforded by the plasmid-borne control system did not exceed those of the wild-type strain and failed to reduce elevated levels of FabB caused by knocking out fabF (Fig. S8). To determine if further enhancements in FabF expression—and, perhaps, higher intracellular concentrations of FabF variants—might shift product profiles, we turned to our kinetic model. Intriguingly, when we modeled increasing concentrations of I108F in the presence of a low FabB background (0.25 μM), the I108F mutant of FabF had no perceptible influence on product profiles, even when present at concentrations much higher than FabB (i.e., a FabF:FabB ratio of 40; Fig. S9B). This insensitivity suggests that FabF overexpression cannot amplify mutational effects in the presence of the reduced—yet still sub-stantial—background concentrations of FabB necessary to sustain growth.

3.4. Exogenously added cis-vaccenic enhances short-chain fatty acid biosynthesis

When E. coli is grown in the presence of oleic acid, it no longer needs FabB (Nosho et al., 2018). Long-chain fatty acids can be taken up by FadL, converted to acyl-CoAs by FadD, and used by PlsB and PlsC to build membrane lipids (Fig. 4B; (Black et al., 1992; Nunn and Simons, 1978; Silbert et al., 1972)). Oleic acid is particularly interesting because it can downregulate FabB by binding to FabR, a transcriptional repressor of fabB (Feng and Cronan, 2011). We speculated that exogenously added cis-vaccenic acid, an isomer of oleic acid, might decouple growth from FabB concentration and improve transcriptional repression of FabB. We note: Cis-vaccenic (or octadec-11-enoic) acid, which is generated by E. coli, is an analogue of oleic (or octadec-9-enoic) acid, which is made by plants and animals. To test our strategy, we swapped I108F for fabF in the E. coli genome and transformed this strain with plasmid-borne control systems for TesA, FabB, and FadD (Fig. 4A). The FadD system enables constitutive expression of fadD, which was knocked out of our parent strain. Consistent with our prior work, transcriptional repression of fabB increased the fractions of C6 and C8 fatty acids, even alongside wild-type FabF (Mains et al., 2022). Our prior work indicates that this shift results from a reduction in the rate of acyl-ACP elongation (relative to acyl-ACP termination). By itself, exogenously added C18:1 had little effect on product profile; when paired with transcriptional repression of fabB, however, it reduced average chain length for both wild-type and I108F systems significantly (Fig. 4C). For wild-type FabF, the titer of C6 fatty acids increased by over 400%; for I108F, the C8 titer increased by ~300% (relative to the case with no fabB repression or C18:1 addition). We used proteomics to confirm the influence of C18:1 on FabB; indeed, when paired with transcriptional repression of fabB, this metabolite reduced FabB concentrations by an additional ~ five-fold (Fig. S11). By contrast, concentrations of FabR and two related proteins—FabA, an enzyme weakly regulated by FabR that contributes to unsaturated fatty acid synthesis (Feng and Cronan, 2011), and FadR, a transcriptional activator of fabB and fabA—changed by less than 1.5-fold. These results indicate that the addition of C18:1 enhances fabB repression and, in doing so, not only increases the production of short-chain fatty acids by wild-type FabF, but also amplifies the influence of FabF mutants on final product profiles.

Fig. 4. Exogenously supplied C18:1 enhances the production of short-chain fatty acids.

(A) We swapped eight mutants of FabF in place of fabF in the genome of E. coli RL08ara. Plasmid-borne control systems for TesA, FabB, and FadD allowed us to control the concentrations of these enzymes (pBad18-TesA, pFD152-FabB, and pCDM4-Pro1-FadD, respectively). The FadD system, which is new to this figure, enabled constitutive expression of fadD, which was knocked out of the RL08ara genome. Genes: β-ketoacyl-ACP synthase II (fabF), thioesterase 1 (tesA), β-ketoacyl-ACP synthase I (fabB), and long-chain-fatty acid – CoA ligase (fadD). (B) E. coli can take up long-chain fatty acids, convert them to acyl-CoAs, and use them for lipid biosynthesis. Cis-vaccenic acid, which enables lipid biosynthesis, can activate FabR, a transcriptional repressor of fabB. (C) When paired with transcriptional repression of fabB, the addition of cis-vaccenic (C18:1) to our cultures improved the production of short-chain fatty acids (C6 and C8) for both the wildtype and FabF I108F strains (RL08ara with pBad18-TesA and pFD152-FabB, 0.1% arabinose, 0 or 0.25 μg/mL aTC, and 0 or 80 μg/mL C18:1). (D) The product profiles of wild-type and mutant strains of E. coli. The unoptimized wild-type profile has no aTC or C18:1, while all other cases include 0.25 μg/mL aTC and 80 μg/mL C18:1. In C-D, plots of production depict the mean, SE, and individual measurements for n ≥ 3 biological replicates. Plots of length depict the mean and SE for n ≥ 3 biological replicates. For all plots, we measured fatty acids at 24 h after induction (30 °C).

The β-oxidation pathway can break long-chain fatty acids down into short chains and could be the source of changes in product profiles caused by C18:1 addition. A few pieces of evidence, however, help eliminate this possibility. First, FabF mutants produce different product profiles in the presence of C18:1, an indication that their mutations give rise to those differences. Second, to generate short-chain fatty acids from a long chain, the β-oxidation cycle would need to occur multiple times, generating medium-chain intermediates along the way. Because the two major thioesterases of E. coli, TesA and TesB, are most active on long-chain acyl-CoAs, β-oxidation should increase the production of long and medium chains, not short ones. Finally, β-oxidation is tightly regulated (Fujita et al., 2007; Geiger Editor et al., 2019), yet our proteomics results suggest that C18:1 addition fails to increase the expression of its constituent enzymes (Fig. S11). These findings suggest that β-oxidation does not contribute to differences in the fatty acid profiles generated in this study.

Motivated by the profile shifts afforded by cis-vaccenic acid, we extended our analysis to other mutants. In total, we examined eight that generated average chain lengths of 10 or less in vitro. For nearly all mutants, the combined effects of C18:1 addition and direct fabB repression reduced average chain length and enhanced the production of dominant in vitro products. I108M offers an exception; in vitro, this mutant produced C8 and a small amount of C6; in vivo, it generated mainly C4 and low titers. Other mutants exhibited greater consistency between the two conditions: I108F and G198F produced the largest titers of C8 fatty acids, while wild-type and L111M yielded the most C6. Interestingly, all in vivo profiles generated a high fraction of C12, which is present at low levels in vitro. We speculate that background concentrations of FabB, despite being low, might be responsible for generating this product. Our modeling results are consistent with this contribution; in the presence of I108F, concentrations of C12 fatty acids increase in proportion to FabB and become the dominant product at high concentrations of FabB (Fig. S9A).

3.5. Blocking mutations in FabB can improve the production of C8 and C12 fatty acids

Motivated by the influence of FabF mutants on in vivo product profiles, we generated mutants of FabB, which has a structurally similar active site to FabF. For this analysis, we focused on two residues: (i) G107, which acts as a gating residue for long-chain acyl-ACPs (it is analogous to I108 in FabF), and (ii) M197, which protrudes into the acyl binding pocket; this residue contacts the acyl chain in structures of FabB bound to ACP_C16 and ACP_C12 (Fig. 5A). In an attempt to block the extension of medium-chain acyl-ACPs, we mutated both residues to bulkier amino acids: G107M and M197F. In vitro, these two mutants afforded large shifts in product profiles (Fig. 5B): G107M produced C8 fatty acid as a dominant product, while M197F produced mainly C8 and C12. Intriguingly, the blocking mutations appeared more permissive than similar mutations in FabF, where they prohibited the extension of ACP_C8.

Fig. 5. FabB mutants can enhance C8 and C12 fractions.

(A) Top: Crystal structures of FabB bound to C16 crypto-ACP (PDB: 6OKF, left) or C12 crypto-ACP (PDB: 6OKC; right). Highlights: FabB (gray), ACP (teal), the acyl chain (blue), the catalytic residues (dark blue), G107 (red), and M197 (purple). Bottom: Two-dimensional illustrations depict the active site of FabB with loop 1 in the “closed” conformation. (B) The product profiles of FASs reconstituted in vitro from purified enzymes. Each reaction consists of 2.6 mM NADPH, 1.5 mM malonyl-CoA, 0.3 mM acetyl-CoA, 10 μM ACP, 10 μM TesA, 1 μM of each Fab, 0 μM FabF, and 1 μM of a FabB variant: wild-type (WT) or mutant. Plots depict the mean, standard error (SE), and individual measurements for n ≥ 3 biological replicates. Fatty acid profiles were measured at 2 h. (C) We swapped two mutants of FabB in place of fabF in the genome of E. coli RL08ara. Plasmid-borne control systems for TesA, FabB, and FadD allowed us to control the concentrations of these enzymes (pBad18-TesA, pFD152-FabB, and pCDM4-Pro1-FadD, respectively). Genes: β-ketoacyl-ACP synthase II (fabF), thioesterase 1 (tesA), β-ketoacyl-ACP synthase I (fabB), and long-chain-fatty acid—CoA ligase (fadD). (D) When paired with transcriptional repression of fabB, the addition of cis-vaccenic (C18:1) to our cultures improved the production of C8 and C12 fatty acids by FabB mutants (RL08ara with pBad18-TesA and pFD152-FabB, 0.1% arabinose, 0.25 μg/mL aTC, and 80 μg/mL C18:1). (E) Under these conditions, G107M afforded higher titers of C8 fatty acid than any other mutant. In D-E, plots of production depict the mean, SE, and individual measurements for n ≥ 3 biological replicates; we measured fatty acids at 24 h after induction (30 °C).

Next, we evaluated the influence of FabB mutants in vivo. Given the essentiality of fabB, we integrated each mutant into the genome in place of fabF (Fig. 5C). As expected, C18:1 addition and fabB repression allowed both mutants to enhance the production of short-chain fatty acids (Fig. 5D). Both generated C8 as dominant product with C12 as a minor product. Intriguingly, G107M afforded higher titers of C8 fatty acids than any other strain in our study (2340 + 150 μM; Fig. 5E); it generated 120% more than the wild-type strain with both C18:1 addition and fabB repression, and an astounding 730% more than the wild-type strain without these adjustments. In fact, this FabB mutant produced as much C8 as a mutant of TesA (TesA R3M4) engineered for this purpose (Fig. S12). This high titer highlights the pronounced influence that changes in KS substrate specificity can have—at least within specific biochemical contexts—on the product profiles of FAS systems. Broadly, it suggests that KS engineering can be as effective as thioesterase engineering, a more thoroughly explored strategy, in controlling fatty acid profiles.

4. Discussion

Fatty acid synthesis is a cyclic biocatalytic process in which elongating KSs control the distribution of acyl-ACPs available for downstream products. In this study, we generated KS mutants with altered substrate specificities and used in vitro, in vivo, and in silico analyses to identify biochemical adjustments necessary to amplify their influence on fatty acid profiles. Our analysis of in vitro systems helped confirm the contribution of conserved structural motifs to the substrate specificity of KSs. In FabF, mutations that disrupted stabilizing interactions in loop 1 had little influence on product profile, while similar mutations in loop 2 decreased the production of long chains (≥14 carbons). Blocking mutations in the acyl binding pocket had the strongest effect; in general, these mutants generated mostly C8, which they could not extend further, and some shorter fatty acids, depending on the mutant. Analogous mutations in FabB had a similar but less dramatic effect; they inhibited—but did not prohibit—the elongation of C8. In general, our in vitro analyses indicate that single amino acid substitutions in loop 2 and the acyl binding pocket of elongating KSs can shift FAS product profiles toward short chains without inhibiting overall rates of fatty acid synthesis.

FabB, which is essential for unsaturated fatty acid biosynthesis, blunts the influence of KS mutants on fatty acid profiles. The importance of these unsaturated fatty acids, which FabF helps elongate (Ruppe et al., 2020), is highlighted by the increased expression of FabB that occurs when fabF is deleted from the genome. Vexingly, the repression of FabB inhibits cell growth and reduces titer, and higher concentrations of FabF mutants have little effect on product profiles. Findings indicate that FabB must be rendered nonessential to achieve a level of FabB depletion sufficient to resolve the influence of KS mutants on FAS outputs. This observation is consistent with prior work, where inducible degradation of FabB enhanced the influence of a FabF mutant expressed alongside thioesterase variants (Torella et al., 2013). The addition of inducers or unsaturated fatty acids may be impractical for an industrial process; however, heterologously expressed desaturases that act on free fatty acids or membrane lipids offer a potential alternative. Broadly, this work shows that fatty acid synthesis must be decoupled from microbial growth to enable the use of KS mutants for profile tuning, but it illustrates that these mutants can be extraordinarily influential. In future work, they could accelerate efforts to control the product profiles of other termination enzymes that act directly on acyl-ACPs (acyl-thioester reductases, acyl-ACP reductases, and methyl ketone synthases, to name a few (Peoples et al., 2022)).

Data availability

Data will be made available on request.