Mechanism-based cross-linking probes capture the Escherichia coli ketosynthase FabB in conformationally distinct catalytic states

Crystal structures of a cross-linked ketosynthase–acyl carrier protein complex elucidate the chain-length preference and substrate-processing mechanism of E. coli FabB.

Abstract

Ketosynthases (KSs) catalyse essential carbon–carbon bond-forming reactions in fatty-acid biosynthesis using a two-step, ping-pong reaction mechanism. In Escherichia coli, there are two homodimeric elongating KSs, FabB and FabF, which possess overlapping substrate selectivity. However, FabB is essential for the biosynthesis of the unsaturated fatty acids (UFAs) required for cell survival in the absence of exogenous UFAs. Additionally, FabB has reduced activity towards substrates longer than 12 C atoms, whereas FabF efficiently catalyses the elongation of saturated C14 and unsaturated C16:1 acyl-acyl carrier protein (ACP) complexes. In this study, two cross-linked crystal structures of FabB in complex with ACPs functionalized with long-chain fatty-acid cross-linking probes that approximate catalytic steps were solved. Both homodimeric structures possess asymmetric substrate-binding pockets suggestive of cooperative relationships between the two FabB monomers when engaged with C14 and C16 acyl chains. In addition, these structures capture an unusual rotamer of the active-site gating residue, Phe392, which is potentially representative of the catalytic state prior to substrate release. These structures demonstrate the utility of mechanism-based cross-linking methods to capture and elucidate conformational transitions accompanying KS-mediated catalysis at near-atomic resolution.

1. Introduction

Fatty-acid synthases (FASs) carry out a series of iterative biochemical transformations to produce fatty acids (FAs). Depending on their organization, FASs are classified as type I multidomain megasynthases or type II multiprotein complexes comprising discrete, monofunctional enzymes. Interestingly, type II FASs, which are commonly found in bacteria and plant plastids, are considered to be the evolutionary progenitors of polyketide synthases, which are known to produce complex, biologically active molecules (Chen et al., 2018 ▸; Hillenmeyer et al., 2015 ▸). FASs iteratively extend and reduce two-carbon malonyl-CoA-derived units. A ketosynthase (KS) begins each round of chain elongation by condensing a malonyl-CoA-derived substrate with a growing fatty acid to generate a β-keto intermediate via a decarboxyl­ative Claisen condensation reaction. Subsequently, a keto­reductase, a dehydratase and an enoylreductase catalyse three sequential reactions to fully reduce the β-ketone to a methylene (Supplementary Fig. S1). Central to the elongation cycle is the acyl carrier protein (ACP), which is required to shuttle thioester-tethered substrates to each enzyme active site (White et al., 2005 ▸). The thiol group of the ACP is derived from 4′-phosphopantetheine (PPant), which is post-translationally installed onto a conserved serine residue of the ACP. The protein–protein interactions (PPIs) of ACP and its partner enzymes are crucial for efficient substrate delivery, particularly in type II systems possessing discrete domains (Mindrebo, Patel, Misson et al., 2020 ▸).

KSs catalyse the critical carbon–carbon bond-forming reactions at the beginning of each FAS elongation cycle (Heil et al., 2019 ▸). The KS ping-pong mechanism consists of a transacylation half-reaction and a condensation half-reaction. During the transacylation step, the acyl chain of the acyl-ACP is transferred to the catalytic cysteine of the apo KS to generate an acyl-KS intermediate. During the condensation step, malonyl-ACP (mal-ACP) binds to the KS and undergoes decarboxylation to form a two-carbon enolate ion that attacks the acyl-KS thioester to form β-ketoacyl-ACP. Escherichia coli possesses two elongating KSs, FabB and FabF, that belong to the ketoacyl-ACP synthase I (KASI) and ketoacyl-ACP synthase II (KASII) families, respectively. The two KSs have partially overlapping functions (Supplementary Fig. S1), but FabB specifically elongates cis-3-decenoyl acid (C10:1), the first committed unsaturated fatty-acid (UFA) intermediate in E. coli. Strains lacking FabB are UFA auxotrophs, but those lacking FabF are viable without supplementation (Cronan et al., 1969 ▸; D’Agnolo et al., 1975 ▸; Feng & Cronan, 2009 ▸). In terms of the chain-length selectivity, FabB has reduced activity towards substrates longer than 12 C atoms, whereas FabF retains robust activity on 14-carbon and 16-carbon FA substrates (Edwards et al., 1997 ▸).

Interactions between ACP and its partner enzymes (PEs) are transient, thus complicating the visualization of PE–ACP complexes that approximate key catalytic states. In recent years, the application of phosphopantetheine analogs outfitted with mechanism-based cross-linking probes mimicking fatty-acyl substrates have advanced our understanding of the multistep KS reaction mechanisms and substrate specificities for both FAS and PKS (Chen et al., 2022 ▸; Milligan et al., 2019 ▸; Mindrebo, Patel, Kim et al., 2020 ▸; Du et al., 2020 ▸; Mindrebo et al., 2021 ▸; Herbst et al., 2018 ▸). These probes contain selectively reactive warheads that form a covalent bond with the catalytic cysteine of KSs, resulting in the formation of a covalently trapped KS–ACP complex for structure elucidation using X-ray crystallography. Two classes of KS warheads, chloro­acrylates and α-bromocarbonyl-containing molecules (Worthington et al., 2006 ▸, 2008 ▸), have successfully been applied in previous structural studies. Chloroacrylate cross-linked complexes, which tether the C₃ position of the acyl chain to the KS active site via conjugate substitution, mimic an intermediate formed during the condensation half-reaction, as demonstrated by previously published ACP–KS cross-linked structures of FAS and PKS (Milligan et al., 2019 ▸; Du et al., 2020 ▸; Mindrebo et al., 2021 ▸). The FabB–ACP α-bromo cross-linked complexes, which tether the C₂ position of the acyl chain to the KS active site, mimic a transacylation intermediate. In contrast, FabF–ACP complexes trapped with α-bromo probes resemble an intermediate substrate-delivery state typified by large conformational rearrangements of two KS active-site loops (Supplementary Fig. S2; Mindrebo, Patel, Kim et al., 2020 ▸; Mindrebo et al., 2021 ▸). The two loops, loop 1 and loop 2, form a double drawbridge-like gate (Gora et al., 2013 ▸) at the entrance of the active-site pocket that regulates substrate processing. Current models suggest that this gate is likely to be triggered by the interaction of loop 2 and acyl-ACP and subsequent chain flipping (Cronan, 2014 ▸) of ACP-sequestered cargo in the active site. The gate-open conformation has only been captured in cross-linked crystal structures, highlighting the utility of mechanism-based cross-linking probes.

Despite extensive structural and biochemical research, the specific molecular mechanisms that govern the substrate specificity of FabF and FabB remain elusive. FabF and FabB share 40% identity, and representative structures of FabF (PDB entry 2gfw; Wang et al., 2006 ▸) and FabB (PDB entry 1g5x; Zhang et al., 2001 ▸) overlay with an r.m.s.d. of 0.72 Å. Our recent structural findings indicate that FabF and FabB have divergent loop 2 gating sequences (Mindrebo, Misson et al., 2020 ▸; Mindrebo et al., 2021 ▸). However, a FabF loop-swap mutant, in which loop 2 of FabB replaces loop 2 of FabF, has dramatically reduced condensation and transacylation rates, suggesting that these elements are not readily transferable between the two synthases (Mindrebo et al., 2021 ▸). Additionally, the fatty-acid binding pockets of FabF and FabB possess minor differences. An isoleucine residue in FabF (Ile108), which undergoes a conformational change to accommodate the thioester-tethered acyl moiety (Wang et al., 2006 ▸), is not present in FabB. Interestingly, mutations of this residue can lead to antibiotic resistance and modify the fatty-acid profile in both bacterial and fungal FAS (Val et al., 2000 ▸; Schujman et al., 2001 ▸; Gajewski et al., 2017 ▸). Structural overlays of FabF and FabB bound to dodecanoic acid do not clearly indicate how this residue would alter specificity between the two homologs, but these structures represent states after successful acyl-chain offloading and may not capture conformations induced by ACP binding and substrate delivery.

In this study, we develop and apply two cross-linking probes bearing long-chain fatty-acyl substrate mimetics to investigate the molecular basis of the substrate preference of FabB. The first probe, C14-chloroacrylate-pantetheinamide (C14Cl), mimics the final condensation reaction catalysed by FabB in the elongation cycle, while the second probe, C16:1-α-bromo-pantetheinamide (C16:1Br), represents the acylation reaction of FabB with a disfavoured substrate. Two crystal structures, FabB=C14-ACP and FabB=C16:1-ACP (where the ‘=’ sign denotes a cross-linked complex), were determined using C14Cl and C16:1Br as cross-linkers, respectively. Both structures possess a back gate comprised of a glutamine and a glutamate residue that can alternate between conformers to expand or limit the size of the fatty-acyl binding pocket. The FabB=C14-ACP structure possesses a relaxed back-gate arrangement that accommodates the C14 chain well in both protomers. In contrast, the FabB=C16:1-ACP complex has distinct substrate-binding pocket structures in each of the protomers that either limit or accommodate an extended substrate conformation. Importantly, the pocket conformation affects thioacyl coordination and alters the conformation of key catalytic residues related to KS gate function. Insights from these structures correlate well with FabB function and provide a new perspective on E. coli FAS and KS-directed metabolic engineering.

2. Materials and methods

2.1. Synthesis of cross-linking probes

C16:1Br (Mindrebo et al., 2021 ▸), (E)-C3Cl (Worthington et al., 2006 ▸) and (E)-C8Cl (Worthington et al., 2008 ▸) were prepared according to previous literature protocols. (E)-C14Cl and (Z)-C14Cl were prepared according to Supplementary Scheme S1. See the supporting information for complete synthetic details.

2.2. Protein expression and purification

N-terminally His-tagged recombinant FabB protein was expressed in E. coli BL21(DE3) cells cultured in LB broth containing 50 mg l⁻¹ kanamycin. The cells were grown at 37°C until the OD₆₀₀ reached ∼0.7 and were induced with 0.5 mM isopropyl β-d-thiogalactopyranoside (IPTG) for 4 h at the same temperature. The cell pellets were collected by centrifugation, lysed by sonication in lysis buffer (50 mM Tris pH 8.0, 150 mM NaCl, 10% glycerol) and underwent further centrifugation at 17 400g for 1 h to clarify the lysate. The His-tagged protein was pulled down by Ni–NTA resin (0.5 ml resin per litre of culture), washed with wash buffer (50 mM Tris pH 8.0, 150 mM NaCl, 10% glycerol) and 20 mM buffered imidazole and eluted with 250 mM buffered imidazole. His-tag cleavage by bovine thrombin (2 U per milligram of FabB) was performed while dialyzing against dialysis buffer (50 mM Tris pH 8.0, 150 mM NaCl, 10% glycerol, 0.5 mM TCEP) for 16 h at 6°C. The resulting solution was passed over an Ni–NTA resin column to remove the uncleaved protein and was further purified by FPLC with a HiLoad Superdex 200 size-exclusion column (GE Bio­sciences). Pure tag-free FabB was collected and concentrated to 2–4 mg ml⁻¹ using Amicon Ultra centrifugal filters (MilliporeSigma) with a 10 kDa molecular-weight cutoff.

Native recombinant ACP protein was expressed in E. coli BL21(DE3) cells cultured in LB broth containing 100 mg l⁻¹ ampicillin. The cells were grown at 37°C until the OD₆₀₀ reached ∼0.7 and were induced with 0.5 mM IPTG for 16 h at 18°C. The cell pellets were collected by centrifugation, lysed by sonication in lysis buffer (50 mM Tris pH 7.4, 150 mM NaCl, 10% glycerol) and underwent further centrifugation at 17 400g for 1 h to clear the lysate. Taking advantage of the high stability of ACP in 50% 2-propanol solution, irrelevant proteins were precipitated by titrating 2-propanol into the lysate at a speed of 0.1 ml s⁻¹ [to 50%(v/v) 2-propanol] and were removed by centrifugation at 17 400g for 1 h. The supernatant was purified by FPLC via a HiTrap Q HP anion-exchange column with a linear gradient of 0–1 M NaCl buffer. ACP eluted around 0.3 M NaCl and was collected for use in a second round of FPLC purification with a HiLoad Superdex 75 pg size-exclusion column (GE Biosciences). Pure ACP was collected and concentrated using Amicon Ultra centrifugal filters (MilliporeSigma) with a 3 kDa molecular-weight cutoff.

Actinorhodin KS (actKS) was prepared as described previously (Taguchi et al., 2017 ▸). C-terminally His-tagged ACP hydrolase (Pseudomonas aeruginosa AcpH) was prepared as described previously (Kosa et al., 2012 ▸) except that the purification was performed in a cold room (6°C). The four enzymes used for loading the cross-linking probes, Sfp, CoaA, CoaD and CoaE, were prepared as described previously (Finzel et al., 2017 ▸).

2.3. ACP modification and cross-linking reaction

ACP is overexpressed as a 1:1 ratio of the apo and holo forms. Since the cross-linking probes contain the PPant moiety, apo ACP is the desired form, and thus the PPant on holo ACP is removed by enzymatic reaction using the ACP hydrolase AcpH. 2–10 mg ml⁻¹ of ACP mixture can be totally converted to apo ACP with 0.01 mg ml⁻¹ AcpH in the reaction buffer (50 mM Tris pH 7.4, 150 mM NaCl, 10% glycerol, 10 mM MgCl₂, 5 mM MnCl₂, 0.25% DTT) at 25°C in 16 h. The resulting apo ACP was purified by size-exclusion chromatography (HiLoad Superdex 75 pg) and concentrated (Amicon Ultra centrifugal filters, 3 kDa molecular-weight cutoff) to 2–10 mg ml⁻¹. To load the cross-linking probes onto apo ACP, they were first converted to CoA analogues through the activities of three CoA biosynthetic enzymes (CoaA, CoaD and CoaE) and transferred onto the serine residue (Ser36) of apo ACP by the phosphopantetheinyl transferase Sfp. Modification and loading were performed as a one-pot chemo­enzymatic reaction with the following conditions: 1 mg ml⁻¹ apo ACP, 0.04 mg ml⁻¹ Sfp, 0.01 mg ml⁻¹ of each CoA enzyme and 0.2 mM cross-linking probe in reaction buffer (50 mM potassium phosphate pH 7.4, 12.5 mM MgCl₂, 1 mM DTT) at 37°C for 16 h. The stock solution of the probes was prepared by dissolving the probes in DMSO to a final concentration of 50 mM for C14Cl and 25 mM for C16:1Br. The probe-loaded ACP, or crypto ACP, was then purified by size-exclusion chromatography and concentrated as above to 2–5 mg ml⁻¹. All of the different forms of ACP were tracked and confirmed using conformationally sensitive urea-PAGE.

The cross-linking reaction was set up by mixing crypto ACP with FabB in a 3:1 ratio in reaction buffer (50 mM Tris pH 7.4, 150 mM NaCl) at 37°C for 16 h. The reaction was monitored by 12% SDS–PAGE and purified by size-exclusion chromatography with minimal buffer (12.5 mM Tris pH 7.4, 25 mM NaCl). The cross-linked complex was concentrated to 8–10 mg ml⁻¹ using Amicon Ultra centrifugal filters with 30 kDa molecular-weight cutoff and flash-frozen for storage at −80°C if not immediately used for crystallography.

2.4. Protein crystallography, data processing and structure refinement

Crystals of all of the cross-linked complexes were grown by vapour diffusion in a cold room kept at 6°C. 1 µl cross-linked complex (8–10 mg ml^–1) was mixed with 1 µl of the corresponding crystallographic condition and the mixture was placed inverted over 500 µl well solution (hanging-drop method). The FabB=C14Cl-ACP and FabB=C16:1-ACP crystals were grown in 18–24%(w/v) PEG 8K, 0.2 M magnesium acetate, 0.1 M sodium cacodylate pH 6.5. These conditions produced square plates and required two weeks for complete crystal growth. X-ray diffraction data (Table 1 ▸) were collected at the Advanced Light Source synchrotron at Berkeley. Data were indexed using iMosflm (Battye et al., 2011 ▸) and were processed and scaled using AIMLESS from the CCP4 software suite (Winn et al., 2011 ▸). Scaled reflection output data were used for molecular replacement and model building in Phenix (Liebschner et al., 2019 ▸). Initial phasing via molecular replacement was performed by searching for the full FabB=ACP complex using PDB entry 5kof (Milligan et al., 2019 ▸) as a search model. Subsequent refinement and model refinement were performed in Phenix and Coot, respectively. The parameter files for the covalently bound 4′-phosphopantetheine were generated using JLigand (Lebedev et al., 2012 ▸). Manually programmed parameter restraints were used to create the associated covalent bonds between 4′-phosphopantetheine and Ser36 of ACP and Cys163 of FabB during refinement.

Table 1. Data-collection and refinement statistics.

Structure (PDB code)	FabB=C14-ACP (7sqi)	FabB=C16:1-ACP (7sz9)
Wavelength (Å)	1	1
Resolution range (Å)	36.86–1.7 (1.761–1.700)	47.9–2.2 (2.279–2.200)
Space group	P212121	P1211
a, b, c (Å)	58.99, 112.38, 141.64	59.45, 103.58, 79.28
α, β, γ (°)	90, 90, 90	90, 108.28, 90
Unique reflections	104222 (10240)	38076 (2494)
Reflections used in refinement	103901 (10155)	37907 (2488)
Reflections used for R free	5103 (466)	1808 (122)
R work	0.1608 (0.3039)	0.2084 (0.3117)
R free	0.1981 (0.3281)	0.2553 (0.3461)
Protein residues	965	957
Solvent atoms	1069	165
R.m.s.d., bond lengths (Å)	0.008	0.002
R.m.s.d., angles (°)	0.97	0.47
Ramachandran favoured (%)	96.33	94.84
Ramachandran allowed (%)	3.36	4.43
Ramachandran outliers (%)	0.31	0.74
Rotamer outliers (%)	0.80	1.46
Clashscore	3.80	4.84
Average B factor (Å2)
Overall	23.68	57.06
Protein	24.48	57.11
Cross-linker	19.97	78.61
Water	30.89	45.80

3. Results and discussion

3.1. Rationale behind cross-linker choice and the stereospecificity of KS for chloroacrylate probes

To understand the molecular basis of the substrate preference of FabB towards long-chain FAs, we developed a 14-carbon chloroacrylate cross-linker, C14Cl, that we hypothesized would mimic the condensation step between the longest favoured FabB substrate, C12, and malonyl-ACP. Probes containing two different configurations, E- and Z-form, of the acrylate double bond were synthesized and tested (Supplementary Fig. S3). Only the Z-form showed cross-linking activity on FabB (Supplementary Fig. S4a ) and the other two tested KSs: FabF and actinorhodin KS (actKS), a type II polyketide synthase (Keatinge-Clay et al., 2004 ▸). Given that there are published cross-linked structures of FabB–ACP and FabF–ACP utilizing (E)-C3Cl and (E)-C8Cl, respectively, we hypothesized that while the Z-form is the preferred configuration, the selectivity only develops at longer chain lengths due to restrictions in the substrate pocket. To test this hypothesis, we performed cross-linking assays using (E)-C8Cl and actKS, which has a similar active-site arrangement to FabF and FabB but possesses a narrower pocket (Chen et al., 2022 ▸) that is more likely to restrict the rearrangement of the E-form stereoisomer. The results from these assays showed no cross-linking activity (Supplementary Fig. S4b ). However, (E)-C3Cl was able to cross-link with our entire panel of KSs, indicating that actKS is generally active towards unsubstituted E-form chloroacrylate cross-linkers. As a result, the data suggest stereospecificity of KSs towards the Z-form long-chain chloro­acrylate cross-linker. After confirming the preferred stereochemistry, we loaded ACP with (Z)-C14Cl and performed large-scale cross-linking to obtain FabB=C14-ACP for structural analysis.

To explore the mechanisms governing FabB substrate selectivity, we also obtained a cross-linked complex using the C16:1-α-bromo cross-linker (C16:1Br; Supplementary Fig. S3), which is likely to mimic the transacylation step of an unfavoured substrate, namely C16:1-ACP. FabB has poor activity with C16:1 in vitro and only produces appreciable amounts of C18:1 in vivo when overexpressed (Mendoza et al., 1983 ▸). Importantly, this substrate is readily accepted by FabF in vitro and in vivo, and its elongation to C18:1 is critical for the rapid homeoviscous adaptive response of E. coli to regulate membrane fluidity due to changes in environmental temperature (Garwin et al., 1980 ▸; de Mendoza & Cronan, 1983 ▸). Previously, we utilized C16:1-ACP to investigate the molecular mechanisms governing the unique capacity of FabF to extend C16:1 to C18:1 (Mindrebo et al., 2021 ▸). Therefore, we reasoned that a FabB=ACP structure using the same cross-linker could potentially inform on how FabB discriminates against longer acyl substrates.

3.2. FabB–ACP trapped in different catalytic states by the cross-linkers

The FabB=C14-ACP complex was crystallized and diffracted to a nominal resolution of 1.7 Å. The asymmetric unit contains the biologically relevant dimeric structure (FabB–ACP)₂ with well defined electron density for the cross-linker (Fig. 1 ▸). The active-site cysteine, Cys163, is covalently attached to the C₃ position of the acyl chain, which resembles the putative tetrahedral intermediate formed during the condensation half-reaction. The carbonyl group of the acyl chain coordinates with the two active-site catalytic histidine residues, His298 and His333, as seen in all other KS–ACP cross-linked structures using chloroacrylate cross-linkers (Milligan et al., 2019 ▸; Du et al., 2020 ▸; Mindrebo et al., 2021 ▸). Interestingly, the double bond of the acrylate is in the same trans configuration as observed in other structures despite the use of E-form cross-linkers. This suggests a stereoselective addition–elimination reaction mechanism governed by the active-site arrangement during the reformation of the double bond. Overall, the two protomers of the structure align with an r.m.s.d. of 0.872 Å and show no obvious structural differences.

Figure 1.

FabB trapped in different states by cross-linking with ACP. FabB=C14-ACP (PDB entry 7sqi), the blue structure on the left, possesses two identical active sites with the cross-linker solved in alternative conformations. One of its active sites is enlarged in the black frame, which depicts a condensation half-reaction arrangement. The FabB=C16:1-ACP structure (green, PDB entry 7sz9) on the right has two distinct protomers. Protomer A (blue frame) depicts a transacylation half-reaction arrangement, while protomer B (red frame) features a partially resolved and a catalytically incompetent Phe392 rotamer, potentially mimicking a substrate-release state. The mesh surface depicts the F _o − F _c polder omit map calculated by omitting the cross-linker (contoured at 3.0σ, 2.0σ and 3.0σ, respectively, from left to right). The PPIs of both structures are highly similar to those characterized in a previous publication (Milligan et al., 2019 ▸).

The FabB=C16:1-ACP complex was crystallized and diffracted to 2.20 Å resolution, and the asymmetric unit contains the same dimeric arrangement (Fig. 1 ▸). The protomer that contains the chain A monomer of FabB, referred to as protomer A, has the acyl-chain carbonyl positioned in the oxyanion hole formed by the backbone amides of Cys163 and Phe392. This coordination is essential for the transacylation half-reaction (Olsen et al., 2001 ▸; Rittner et al., 2020 ▸). The other protomer of the complex, referred to as protomer B, captures a unique active-site arrangement in which the acyl carbonyl is not coordinated in the oxyanion hole but instead positions itself so that it repels Phe392 away from the active site towards the second gating loop (loop 2). It is difficult to unambiguously assign coordinates for Phe392 beyond C^γ in 2F _o − F _c maps contoured at or above 1.0σ, which is likely due to the absence of the carbonyl in the oxyanion hole coordinated by Phe392 and rotation around the χ₂ angle. To increase confidence in our interpretation of the Phe392 conformer, we further validated the modelled coordinates using Phenix-generated polder omit maps (Supplementary Fig. S9; Liebschner et al., 2017 ▸). Importantly, Phe392 is a crucial gating element and its rotamers regulate the entry of the malonyl substrate (Zhang et al., 2006 ▸; Luckner et al., 2009 ▸; Witkowski et al., 1999 ▸). Furthermore, it also plays a critical role in the KS gating mechanism during substrate delivery and undergoes an ∼8 Å residue shift (C^α to C^α) when transitioning to the gate-open conformation. Due to the unfavourable substrate conformation, protomer B resembles a catalytically unproductive state (unoccupied oxyanion hole) with an altered Phe392 gating-residue position. It is possible that this conformation represents an active site primed for substrate release. An overlay of the two protomers with our recently published gate-open FabF=C16:1-ACP structure (PDB entry 7l4e; Mindrebo et al., 2021 ▸) suggests a potential trajectory for gate opening, starting from the gate-closed protomer A to the disrupted protomer B and finally to the gate-open FabF=ACP structure (Fig. 2 ▸). The proposed transition features a 90° rotation of the Phe392 side chain, which is likely to be driven by the acyl carbonyl flipping out from the oxyanion hole, which places the side chain underneath the PPant arm to facilitate the subsequent transition. The loop then undergoes a large conformational change to the gate-open structure, with Gly391 and Gly394 acting as hinges, creating space for product dissociation. During the gate transition, loop 2 undergoes a conformational change in concert with loop 1, as observed in FabF. The adjacent loop 2 residue, Val270, abuts Phe392 and would sterically impede a transition to the gate-open conformation. Val270 from protomer B has a notably higher B factor (135.5 Ų) than Val270 from protomer A (93.1 Ų), indicating a less ordered loop 2 potentially primed for transition to the open state. To better illustrate the gate transition, we used PyMOL to generate an interpolated trajectory between the three states in Fig. 2 ▸ (Supplementary Movie S1).

Figure 2.

A potential trajectory for substrate desequestration is illustrated by overlaying three active sites from two cross-linked KS–ACP structures utilizing the same C16:1Br cross-linker. The arrow indicates the transition of the gating residue phenylalanine from protomer A (dark green, gate closed) to protomer B (light green) and finally to the gate-open form (magenta, FabF–ACP structure; PDB entry 7l4e). An animation (Supplementary Movie S1) with interpolated trajectories between the three states can be found in the supporting information.

3.3. Asymmetric pockets of the FabB homodimer suggest negative cooperativity

The substrate pocket of FabB extends from the active-site Cys163 to the homodimer interface. At the interface, Glu200 and Gln113′ (where the prime denotes a residue from the dyad-related protomer) sit near the bottom of the pocket, forming side-chain polar contacts that modulate the size of the binding pocket. However, these residues can also adopt different rotamers where the side chains turn away from each other in opposite directions, breaking the interaction and resulting in pocket expansion. This double side-chain rotation can be classified as a ‘swinging-door enzyme gate’ (Gora et al., 2013 ▸) and will hereafter be referred to as the ‘E-Q gate’. In the FabB=C14-ACP structure, in which a 12-carbon aliphatic chain extends into the binding pocket, two sets of alternative conformations were refined for both the acyl chains and the E-Q gates (Fig. 3 ▸ a). The E-Q gates from both pockets can be modelled in open (expanded pocket) or closed (normal pocket) states. The open E-Q gate places the Gln113′ residue away from the pocket towards the dimer interface and expands the size of the binding pocket when compared with the closed state. However, the modelled C12 substrate conformations are readily accommodated by both states of the E-Q gate (Fig. 3 ▸ a). Interestingly, there is only enough space at the dimer interface to accommodate Gln113 from a single protomer, indicating that the two E-Q gate-open states cannot coexist.

Figure 3.

Asymmetric pockets and the E-Q gate. Residues in the same colour belong to the same FabB monomer. (a) FabB=C14-ACP possesses two sets (conformer A in an orange frame and conformer B with no frame) of alternative conformations of the acyl chain and the E-Q gate. (b) FabB=C16:1-ACP has an expanded pocket (left, protomer A, open E-Q gate) and a normal pocket (right, protomer B, closed E-Q gate).

The asymmetric pocket features of FabB are further illustrated in the FabB=C16:1-ACP structure, in which protomer A has an expanded pocket and protomer B does not (Fig. 3 ▸ b). In protomer A the 16-carbon aliphatic chain penetrates through the open E-Q gate, with the last four carbons flanked by the side chains of the two gating residues. The expanded pocket extends to the α-helix (residues 110′–121′) that contains Glu113′. Protomer B, on the other hand, has a closed E-Q gate that distorts the C16 chain, which prohibited the modelling of more than 13 C atoms into the available density. As a result, a FabB binding pocket in the E-Q gate-closed state cannot comfortably accommodate an acyl chain of longer than 12 C atoms. Importantly, the sterically occluded chain in protomer B is likely to contribute to the distorted active-site arrangement noted above. These observations suggest that the inability to accommodate the acyl substrate results in an improper positioning of the thioester moiety for catalysis, prohibiting the transfer of unfavourable substrates to the KS cysteine residue during the transacylation half-reaction.

FabB shows robust activity in elongating FAs of up to 12 C atoms in length, but reduced activity towards substrates of 14 C atoms in length or longer (Edwards et al., 1997 ▸). This is supported by the FabB=C14-ACP structural data, as both protomers can accommodate 12-carbon substrates regardless of the E-Q gate conformation. However, only the expanded pocket can readily accommodate acyl substrates of beyond 12 C atoms in length. These results explain the low, but existing, FabB activity towards long-chain acyl substrates in vitro and in vivo (Edwards et al., 1997 ▸; Mindrebo, Misson et al., 2020 ▸). Therefore, the observed pocket and the E-Q gate mechanism align well with the reported activity of FabB and suggest negative cooperativity towards long-chain acyl substrates, as observed in the FabB=C16:1-ACP structure, where protomer B ‘struggles’ to accept the substrate. We propose a model in which this negative cooperativity discourages dual acylation with long-chain substrates, ensuring that one active site remains available for medium-chain acyl-ACP extension. In this scenario, both FabB active sites are initially available to accept acyl groups, but if a longer chain fatty acid is transferred to one active site the other protomer will have a stronger preference to accept substrates shorter than 12 C atoms. Ensuring that one active site is always available for medium-chain substrates, such as a cis-3-decenoyl-ACP substrate, could potentially explain the unique properties of FabB in establishing the UFA branch of FAS. Additionally, previous studies indicate that FabF and FabB have similar kinetic parameters towards cis-3-decenoyl-ACP (D’Agnolo et al., 1975 ▸), which is counterintuitive as only fabB knockouts are fatty-acid auxotrophs. However, the presence of both FabF and FabB in E. coli FAS represents an interesting case of partial redundancy, whereby two enzymes with overlapping specificity can tightly regulate distinct metabolic objectives (Ruppe et al., 2020 ▸). In this case, FabF would be ‘distracted’ by a larger substrate pool and would serve to regulate the final fatty-chain length, while FabB could utilize an additional mechanism, such as the E-Q gate, to minimize its potential substrate pool to ensure the recognition and extension of UFA precursors to establish the UFA pathway.

Multiple sequence alignments (MSAs) of 461 FabB orthologs indicate that the E-Q gate is only modestly conserved. The Gln113 residue is present in 78% of sequences, while the equivalent Glu200 position is found in 47% of the orthologs, but interestingly all sequences with Glu200 also possess a corresponding Gln113 residue. These data suggest that half of the analyzed FabB sequences have an E-Q gate similar to E. coli FabB, indicating that this structural element is not strictly conserved across all orthologs. However, it is important to note that the anaerobic FabA/FabB route to UFA production is only common in γ-proteobacteria (Parsons & Rock, 2013 ▸). Most bacteria have only a single elongating KS of the FabF family or employ membrane-localized oxygen-dependent desaturases that are capable of directly introducing the double bond into phospholipid acyl chains (Parsons & Rock, 2013 ▸). Aerobic and anaerobic pathways can also coexist, as is the case for P. aeruginosa, which can utilize the anaerobic FabA/FabB route or one of two aerobic membrane desaturases (Zhu et al., 2006 ▸). Additionally, the FabF/FabB dichotomy is not strictly conserved and there are several FabF orthologs that are capable of complementing E. coli fabB-knockout cell lines and FabB orthologs capable of extending C16:1 to C18:1 (Morgan-Kiss & Cronan, 2008 ▸; Zhu et al., 2009 ▸; Wang & Cronan, 2004 ▸; Luo et al., 2016 ▸; Li et al., 2016 ▸; Dong et al., 2021 ▸). Therefore, the metabolic and genetic background of a particular bacterial species may dictate the selective pressure to maintain specific FabB structural elements that allow the recognition and extension of the cis-3-decenoyl-ACP substrate. Future mechanistic studies using in vitro assays coupled with in vivo analyses of temperature-sensitive fabB cell lines (Harder et al., 1972 ▸) will be needed to examine the functional relevance of the E-Q gate and its potential role in regulating substrate specificity.

3.4. The implication of the side pocket in the KS mechanism

In FAS KSs, there is a ‘side pocket’ located adjacent to the catalytic active site and on the opposite side to the substrate-binding pocket. The PPant binding tunnel, the acyl substrate-binding cavity and the side pocket form a τ-shaped space (Supplementary Fig. S5). The side pocket has a surface area of 180 Ų and a volume of 95 ų and is larger than the acyl substrate pocket, which has a surface area of 145 Ų and a volume of 58 ų, as calculated by CASTp (Tian et al., 2018 ▸). Despite its proximity to the active site, the side pocket has only been examined in passing in previous studies and has not been assigned a role in catalysis. The recently discovered gating loops reveal that the side pocket, which is partially formed by loop 1 and loop 2 when in the gate-closed conformation, is required to accommodate the loop 1 phenylalanine gating residue in the gate-open conformation (Supplementary Fig. S6). Introducing bulky residues by site-directed mutagenesis that fill the available side-pocket space significantly impairs the activity of FabF (Mindrebo et al., 2021 ▸) by restricting access to the gate-open conformation.

Interestingly, the high-resolution FabB=C14-ACP structure allows us to unambiguously assign ten water molecules in the side pocket that form a water-mediated network connecting loop 1, loop 2 and the PPant moiety (Fig. 4 ▸). In the proximity of the catalytic triad, two water molecules, WAT1 and WAT2 (Supplementary Fig. S7), abut the catalytic triad, forming another water-mediated network linking Thr300, Asp306 and Glu309. Notably, these three residues are absolutely conserved across 461 FabB sequences spanning bacterial species. Previous work suggests that a nucleophilic water molecule, activated by the N^ɛ atom of the catalytic histidine residue (His298 in FabB), is required for malonyl-ACP decarboxylation in order to form the enolate anion that subsequently attacks the acyl-KS intermediate. In this mechanism, a hydroxide ion attacks the malonyl carboxylate moiety to form bicarbonate as the released product instead of CO₂ (Supplementary Fig. S7; Zhang et al., 2006 ▸; Witkowski et al., 2002 ▸). The observed water network in the side pocket creates a hydrophilic and water-rich site adjacent to the reaction chamber. This network might assist the decarboxyl­ation of malonyl-ACP by providing the water molecule needed for bicarbonate formation, as well as by accommodating the bicarbonate product. This observation would support the bicarbonate-forming mechanism, as opposed to the mechanism that posits the release of carbon dioxide (Chisuga et al., 2022 ▸). Further biochemical and structural studies are required to definitively address this longstanding question (Heil et al., 2019 ▸).

Figure 4.

Ten water molecules in the side pocket form a water network between highly conserved residues and the PPant arm (FabB=C14-ACP; PDB entry 7sqi). A cartoon representation is shown for the gating loop 1 (orange) and loop 2 (blue). All polar contacts (yellow dashed lines) are within 3.2 Å.

4. Conclusion

Since its initial discovery (Cronan et al., 1969 ▸) E. coli FabB has been studied for more than 50 years, with at least 17 structures being published and deposited in the Protein Data Bank (apo, mutants, substrate-bound, inhibitor-bound and cross-linked with ACP). Despite the extensive knowledge we have regarding this enzyme, some fundamental questions, such as its substrate specificity and the exact decarboxylation mechanism, have yet to be fully understood. In this study, we leveraged our understanding of the KS reaction mechanism and applied mechanism-based cross-linkers to elucidate two FabB=ACP cross-linked complexes in both favoured and disfavoured catalytic states. The C14Cl cross-linker captures FabB in an on-pathway state, representing the condensation reaction with a favoured substrate, C12, and malonyl-ACP. Alternatively, the complex trapped using the C16:1Br cross-linker represents the acylation reaction with a disfavoured long-chain acyl-ACP substrate. Analysis of these structures led to the identification of an E-Q gate at the back of the substrate-binding pocket that enforces asymmetry between the two protomers. A 12-carbon substrate can readily be accommodated in both pockets. However, the E-Q gate must adopt the open conformation to accommodate longer fatty-acid substrates to appropriately position the thioester moiety for the transacylation reaction. The asymmetric nature of the E-Q gate suggests negative cooperativity of FabB towards long-chain substrates, which may explain the reduced activity of FabB towards C14 and C16:1, two substrates that are readily extended by FabF. In addition to these mechanistic insights, we have characterized a water-rich side pocket that plays an important role in the dual-loop KS gating mechanism and supports the bicarbonate-forming decarboxylation mechanism. The results of these studies provide important insights into E. coli FAS and broadly inform on mechanisms governing KS-mediated catalysis.

5. Related literature

The following references are cited in the supporting information for this article: Gottlieb et al. (1997 ▸) and Meier & Burkart (2009 ▸).

Supplementary Material

PDB reference: FabB cross-linked with C14-ACP, 7sqi

PDB reference: FabB cross-linked with C16:1-ACP, 7sz9

Funding Statement

This work was supported by NIH R01 GM095970. JTM was supported by T32 GM832626. TDD was supported by NIH K12 GM068524 and K99 GM12945.