Control of unsaturation in de novo fatty acid biosynthesis by FabA

Abstract

Carrier protein dependent biosynthesis provides a thiotemplated format for the production of natural products. Within these pathways, many reactions display exquisite substrate selectivity, a regulatory framework proposed to be controlled by protein-protein interactions (PPIs). In Escherichia coli, unsaturated fatty acids are generated within the de novo fatty acid synthase by a chain length-specific interaction between the acyl carrier protein AcpP and the isomerizing dehydratase FabA. To evaluate PPI-based control of reactivity, interactions of FabA with AcpP bearing multiple sequestered substrates were analyzed through NMR titration and guided high-resolution docking. Through a combination of quantitative binding constants, residue-specific perturbation analysis, and high-resolution docking, a model for substrate control via PPIs has been developed. The in silico results illuminate the mechanism of FabA substrate selectivity and provide a structural rationale with atomic detail. Helix III positioning in AcpP communicates sequestered chain length identity recognized by FabA, demonstrating a powerful strategy to regulate activity by allosteric control. These studies broadly illuminate carrier protein dependent pathways and offer an important consideration for future inhibitor design and pathway engineering.

Graphical Abstract

Introduction

Fatty acid biosynthesis (FAB), an essential primary metabolic pathway found across all domains of life, which produces not only the fatty acids required for membrane biogenesis and energy storage, is central to the biosynthesis of essential cofactors and cellular regulatory processes^1,2. In E. coli, FAB exists as a precisely coordinated multi-enzyme pathway that must maintain control over both iterative and substrate-selective reactivity in order to maintain cellular homeostasis and environmental response^3–5. This type II FAB expresses all of the protein components as stand-alone and freely diffusing enzymes/domains (Fig 1a), with E. coli FAB associated with more than 25 known participating enzymes^6–8. Still, the membranes of E. coli, and by extension the resulting FAB products, are evolutionarily selected and dynamically tuned to the environment^9,10. While the microbiological phenomena of fatty acid composition specificity have been observed^11,12, the molecular and enzymatic details of this control is just beginning to be unraveled. There remain multiple unknowns around how these enzymes can maintain selectivity over the dozens of possible substrates upon which they act to regulate such a complex and essential pathway.

Figure 1.

A) The type II fatty acid biosynthesis cycle in E. coli. KS: ketosynthase, KR: Ketoreductase, DH: Dehydratase, ER: enoylreductase, TE: Thioesterase, AT: acyltransferase. The secondary role of FabA as an isomerase is specific to C10 acyl chains. B) AcpP as a 3 step process, wherein transient binding and unbinding can be performed separate from the chain flipping of substrates into a partner protein.

An important mechanism may lie in the protein-protein interactions (PPIs) of the central FAB player, the acyl carrier protein (ACP), in its interactions with FAB enzymes^13,14. The E. coli ACP, AcpP, is a small (~10 kDa), four-helical protein, with high homology to many other organisms’ FAB systems. As such, AcpP has served as a model for other thiotemplated, or carrier protein-dependent, pathways^15,16.

Through each step, substrates that are covalently bound to the AcpP by a 4’-phosphopanthetheine arm are accessed by partner enzymes through PPIs in order to elongate the cargo or perform tailoring reactions¹⁷. In FAB, elongation adds two carbon units to provide a β-ketone, which is fully reduced to the saturated alkane in three steps that involve formation of β-hydroxy and trans-α/β unsaturated intermediates. The trans-α/β unsaturated intermediate can undergo reduction to the fully reduced alkane. Alternatively, the trans-α/β-unsaturation can be isomerized to the cis-α/β-unsaturation and enter back into elongation in order to immortalize the cis double bond^18,19. Throughout FAB elongation, ACP-bound intermediates can also be diverted to other enzymes to source fatty acids for the biosynthesis of molecules such as lipoic acid²⁰, acyl homoserine lactones²¹, and lipid A²².

ACPs carry the growing acyl chains covalently attached to a 4’-phosphopantetheine cofactor via a thioester linkage. The ACP has been found to sequester the acyl cargo inside of a hydrophobic pocket that is created by the four α-helices of the ACP structure (Fig 1b). Upon interaction with a partner enzyme, the cofactor and cargo are “chain-flipped” out of the pocket and into the active site of the partner enzyme^23,24. After the reaction, the acyl chain is flipped back to sequester within ACP, thus protecting the substrate from hydrolysis. The PPI binding events have been previously demonstrated to occur rapidly, with the ACP and partner proteins forming transient interactions^25,26. Published studies have demonstrated that substrate changes translate into structural modifications within the ACP, particularly upon helix III²⁷. It has been thought that these structural changes effect partner enzyme interactions²⁸.

One of the most remarkable transformations within unsaturated FAB is the production of a C10:1 fatty acid. The first control step in unsaturation is catalyzed by 3-hydroxy-decanoyl dehydratase, FabA in E. coli. FabA’s role in chain length-specific unsaturation has been established through both in vitro and metabolic flux experiments.^29,30 Specifically, FabA produces the C10:1 cis unsaturation by scavenging 3-hydroxy-decanoyl AcpP from the saturated FAB cycle (Fig 1a). For this, FabA performs a dual purpose. First, dehydrating to form trans-2-decenoate, then FabA performs an isomerization to cis-3-decenoyl AcpP. This substrate then continues to elongation by FabB. This process is the only entry to unsaturated fatty acids for E. coli. Previous studies have demonstrated that substrate analogs bearing a reactive chemical crosslinker loaded onto the AcpP will not react with partners when the probe does not mimic the natural ten-carbon substrate³¹, suggesting a point of substrate control preceding full chain flipping into the partner active site (Fig 1B). The first crosslinked crystal structure of E. coli AcpP was reported of the AcpP=FabA complex (PDB: 4KEH) using crosslinking probes attached to AcpP in order to trap the active site histidine residue²⁶. The crosslinked structure established the catalytic conformation of the AcpP=FabA complex, and associated NMR titration experiments demonstrated the ability to probe dynamic PPIs to understand pre-catalytic interactions.

FabA has a unique and important secondary role as an isomerase, isomerizing trans-2-decenoyl-AcpP into cis-3-decenoyl-AcpP^29,30. We recently elucidated the mechanism and specificity of isomerization through comparison of the crosslinked structure with molecular dynamics (MD) analysis³². This reaction has been demonstrated to be highly selective, and we have reported that chemically reactive probes demonstrate specificity for C10 acyl chain lengths³¹ We observed significant PPI-based substrate control by FabA using crosslinking probes that mimicked C6, C8, and C10- chain lengths tethered to AcpP³³. The corresponding magnitude and specificity implicit in this interaction indicated that selectivity surpassed active site recognition and implicated a substrate-controlled PPI. This inspired us to study the phenomenon in more detail, beginning with solution-state NMR.

Here we report the evaluation of FabA chain length specificity through an analysis of AcpP structure bearing three acyl chain lengths: six, eight, and ten carbons. By uniting NMR titration analysis, MD minimized acyl-AcpP structures, and high-resolution protein docking, we evaluate the role of chain length to modify AcpP structure and the ability of those structural modifications to regulate FabA activity. This study presents a novel combination of structural and quantitative techniques to elucidate mechanisms of chain-length regulation in FAB.

Results

FabA titration with hexanoyl, octanoyl, and decanoyl AcpP

¹H-¹⁵N HSQC NMR has found a role in the study of in solution interactions of AcpP with proteins^25,26, and we employed it here to observe the subtle differences in binding of the substrate bearing AcpPs with FabA. With the movement of peaks representing the average of the populations of states present in solution³⁴. Uniformly labeled ¹⁵N-AcpP was prepared and loaded with the three different probes with C6, C8, and C10 attached through aminopantetheine linkage³³. These were titrated with increasing concentrations of unlabeled FabA to observe the residues which experienced peak migration upon ¹H-¹⁵N HSQC NMR (Fig 2A). Saturated acyl chains were chosen as the tethered analogs to avoid reaction with FabA during experiments and remain consistent with prior crosslinking studies. However, future studies with catalytically inactive enzymes and the native substrate will be necessary to further study specificity.

Figure 2.

A) The chemical shift perturbations (CSPs) of C6, C8, and C10 after titration to saturation with FabA. Each titration was performed to at least 1.5 molar equivalents of FabA. Individual perturbation graphs and spectra are presented in SI figures 1–4. B) The difference in residues between the C6 and C10-AcpP. C) The effect of FabA titration on the C6-AcpP D) The effect of FabA titration the C8-AcpP E) The effect of FabA titration C10-AcpP F) The starting MD derived structures of AcpP overlaid. Helix III exhibits the largest structural change between the chain lengths. G) The APBS electrostatics of the highly positive binding patch of FabA and negative face of AcpP. The left portion of the highlighted region is responsible for the majority of the binding interactions, binding helix III of AcpP.

The effect of the different chain lengths alone on the chemical shifts of AcpP led to the largest chemical shift perturbations (CSPs) at residues F28, D35, S36, E47, I54, D56, A59, I62, T64, and Q66 (Fig 2B, S1&2). These residues all reside along the hydrophobic pocket of AcpP, with many occurring at helix III or along the preceding loop. Across the three titrations we observed CSPs in similar regions of the AcpP (Fig 2C–E). The bulk of CSPs occurred in the helix II and III region and the helix II/III loop. However, the CSPs extended through the top of helix IV. Perturbation magnitude increased with each successive two-carbon increase in substrate size. From a mean CSP of 0.024 for C6, to 0.032 for C8, and 0.043 for C10 (Table 1).

Table 1.

The average CSPs, TITAN analysis, and docking energetics of the AcpP • FabA interaction.

Chain Length	Average CSP	Kd (μM)	koff (s−1)	Electrostatics (kcal/mol)	Van der Waal’s (kcal/mol)	Total energy (kcal/mol)
C6	0.0241	34.8±5.9	311±64	−6.936	−11.18	−2.512
C8	0.0319	–	–	−5.149	−16.11	−11.79
C10	0.0434	8.6±3.0	4535±647	−8.065	−28.13	−22.36

The most perturbed residues of the C10-AcpP•FabA titration were located at D35, S36, D38, A45, E47, T52, I54, E57, A59, T63, Q66, and A68. The largest chemical shift differences between the C6 and C10-AcpPs alone correspond closely to these same most-perturbed residues in the titration with FabA. The most perturbed residues are neighboring or close to one another, except for F28, which remained one of the most perturbed residues. To compare the C6 and C10 titrations, the CSP values were normalized to the largest CSP within the data set (Fig S3). It was noted that CSPs unique to the C10-AcpP•FabA titration were I54, A59, T63, Q66, and A68, while the C6-AcpP•FabA titration had unique CSPs at residues L42, V43, V65, and Y71. Despite largely shared surface interactions, there are unique internal effects of FabA interacting with different cargo bearing AcpPs.

Thermodynamic and kinetic parameters vary between AcpP substrates

TITAN line shape analysis was utilized to generate quantitative data from the NMR titrations performed³⁵. Characterization of the titrations of C6 and C10-AcpP with FabA found that the C6-AcpP bound with 34.8 ± 5.9 μM affinity and 311 ± 64 s⁻¹ off rate (Table 1, Fig S4). The C10-AcpP bound FabA with a 8.6 ± 3.0 μM Kd and 4535 ± 647 s⁻¹ off rate. This demonstrates a significant effect upon increasing the chain length. It is compelling that an addition of four carbons in cargo size could affect such a significant change to the binding affinity. A structural rationale for this difference was sought further through high resolution docking.

AcpP•FabA 3D model

To evaluate these observations structurally, high-resolution docking was performed in the ICM fast fourier transform docking protocol using NMR titration data to guide the docking algorithm^36–39. To ensure that the most relevant acyl-AcpP structure is used for docking, MD derived structures of AcpP with tethered acyl chains of differing chain lengths were docked with explicit acyl cargo (Fig 2F). Next, the partner FabA enzyme was prepared through solvation of the reported crystal structure (PDB: 1MKB)⁴⁰, optimizing the hydrogen bonding and angles to the water molecules. The interfaces of the AcpP are highly acidic, and those of FabA are basic (Fig 2G), making accurate preparation of structures for hydrogen bonding important to accurate modeling. FabA was docked to the three different MD derived structures using identical methodologies except for varying the beginning AcpP inputs. The structures were sorted based on their RMSD to the post-catalytic crosslinked crystal structure 4KEH, and the most thermodynamically favorable model within 5Å of the crosslinked structure was chosen for each chain length (Fig S6). This cutoff was chosen to accommodate the differences between a crosslinked complex and docked complex, both due to the unbound FabA and the un-chain flipped and precatalytic AcpP. Based on this analysis, we have noted that the energetics of the bound complexes agrees with the TITAN analysis and CSPs (Table 1). The AcpPs from the MD study were observed to vary most heavily and distinctly in the helix III region of AcpP (Fig 3a,b), matching the same regions identified by NMR titration studies above. Combined with the observation that the FabA interface is most heavily electropositive in the region which binds helix III, we identified an important role in recognition at helix III of AcpP.

Figure 3.

A) The left face of the AcpP • FabA interaction. The helix III interactions are highlighted and focused on in panels D, E, and F. B) The right face of the AcpP • FabA interaction. E41 is highlighted and focused on in panel C. C) The interactions of E41 with K161’ in C6, C8, and C10-AcpP. D) The interactions of helix III of C10-AcpP with FabA. Displaying the geometric complementarity of the C10-AcpP for the FabA residues E) The interactions of helix III of C8-AcpP with FabA. Displaying the non-complementarity of the binding region for the C8-AcpP helix III. F) The interactions of helix III of C6-AcpP with FabA. C6-AcpP has a helix III angled up and away from the region, such that only D56 at the base of helix III can form any interaction.

Next it was examined whether the AcpPs would demonstrate different ability to bind FabA based on the docked thermodynamics. The energies of the three interactions were well in line with the known substrate preferences (Table 1), with C6 binding with an overall energy of −2.5 kcal/mol, C8 with an overall energy of −11.8 kcal/mol, and C10 with an overall energy of −22.4 kcal/mol³⁸. These energies are not definitive alone, but they are useful for examining how well the AcpP conformations complement the surface of FabA. To appreciate exactly how these minor changes translate to differences in surface binding, we next examined the interface in detail, focusing on helix III.

Identifying structural features which facilitate specificity

To examine exactly which AcpP residues were most important to the interaction, the docked models were examined relative to one another (Fig 4, S4&7). There were several regions of the protein which were identified as important to the interactions, with the most significant occurring on helix III.

Figure 4.

The poses of C10-AcpP mapped onto the AcpP-FabA model. These residues are in agreement and further highlight the importance of I54, which most likely pushes the cargo out during chain flipping. Also, the importance of D35 and D38 in binding the poorly resolved loop region is shown in the CSPs, though not possible in the model.

In the C10 docked structure, there appeared to be a coordinated network of interactions with three acidic residues, D56, E57, and E60, within a distance to form interactions with R136’, the R137’ backbone, and the R137’ side chain, respectively. (Residues on FabA will be noted by a “ ‘ ” throughout the text.) These docked models likely represent a bound but non-chain-flipped encounter complex, representing an initial binding interaction. The C8 helix III structure demonstrates poorer binding, with helix III out of orientation and only one coordinated residue at E57 nearby R136’. E60 appears to be oriented such that there is no space for R137’ to rotate and bind. C6 binds similarly, but with D56 successfully coordinating R136’, reflecting how the orientation of helix III determines the creation of three critical interactions in C10. It appears that shorter chain lengths cannot form all three of these encounter complex interactions. Binding would require structural rearrangement and hinder rapid binding.

Two additional identified interactions lie on residues E41 and E47. E41 interacts with K161’, and only C10 and C6 docked models, but not C8, are within approximate range (within ~3Å) to interact. E47 appears to be important in stabilizing the bottom of helix II, though it is out of interacting distance in the docked pose at 6Å, and it may interact upon full binding. Next, to complement these two studies, the CSP values were projected onto the docked AcpP structures (Fig 5). In addition to noting several strong internal perturbations, most significantly I54, two interactions were noted at D35 and D38. These residue bridges have been identified previously as participating in the stabilizing of the chain-flipped acyl chain²⁶. However, it was noted that D35 was also within a proper hydrogen bonding distance to the backbone of A170’. These interactions may also anchor helix II, along with the important salt bridges identified at E41 and E47. This interaction would have been impossible to identify in the crosslinked structure, where the loop with A170 was not resolved in the crystal structure. I54 has been identified as involved in chain flipping by examining its distance from interactions and positioning of the side chain directly into the acyl pocket. We have additionally examined the docked models of C6 and C8-AcpP in the Supplemental Information.

Figure 5.

A) The ensemble of conformations around the C10-AcpP • FabA interface. A narrowed RMSD range is presented in figure S7. Specific conformations at high RMSD are color coded blue, maroon, green, yellow, and pink panel F. B) The ensemble of conformations around the C8-AcpP • FabA interface. The conformations are presented in further detail in figure S8. C) The ensemble of conformations around the C6-AcpP • FabA interface. The conformations are presented in further detail in figure S8. D) An overlay of the conformations of the C6, C8, and C10-AcpP• FabA interface. The low RMSD displayed in panel E are circled in blue. E) The ensemble of low RMSD states of the C10-AcpP • FabA interface. F) A sample of low energy states of the C10-AcpP • FabA interface. Colors are matched to the model in panel A

Taken together, these data create a compelling picture of FabA’s chain length selectivity as determined by the PPIs with C6, C8, and C10 acyl chain sequestered AcpP. The CSPs reflect different surface interactions from the crosslinked crystal structure, while the magnitude of the CSPs with increasing chain length is in agreement with the TITAN analysis. This demonstrates a mechanism of substrate-dependent selectivity and regulation based upon the substrate-induced structure of AcpP. Here, the unique positioning of helix III is dictated by the respective influence of sequestered chain lengths. These structural differences, though subtle, are significant enough that the most energetically stable docking poses appear to be occluded in shorter chain lengths. This conclusion is further supported by the corroborating thermodynamic and CSP data.

Expanding analysis of FabA docked states

To explore the full structural space of FabA • AcpP interactions, docking was performed to observe secondary binding modes. Using the C6, C8, and C10 models examined above as reference structures, a second docking calculation was performed with all docked poses examined based on their RMSD to the minimized model. The docked states were then graphed by RMSD vs energy (Fig 5) to observe other potential binding modes. In the C10 binding, a cluster of low RMSD conformations (~5Å) were observed (Fig 5A), along with a second cluster of stable secondary states at higher RMSD (~15Å) (Fig 5E). A second, lower energy state, was also observed at ~13Å RMSD from the preliminary model (Fig 5F). This model has FabA adopting a similar docked site (Fig S5), but with the AcpP directed more sharply into the FabA face. In addition, multiple low energy, higher RMSD states sampled features of the FabA interface. The C8 docked data displayed that the model closest aligned to the crosslinked-like structure was higher energy than a set of ~7Å RMSD models (Fig 5B,S5). The lowest energy of these adopts a similar structure to the crosslinked-like model examined. Lastly, C6 showed few low energy states, and low RMSD to the crosslinked-like state. The lowest energy model occurred at ~17Å RMSD from the docked model examined above (Fig 5C,S5). However, this state, as well as multiple similar energy states, docked “upside-down,” likely with no ability to perform a functional interaction that could result in chain-flipping.

Discussion

Evaluation of this data demonstrates several key concepts. First, the “active” cross-link-like conformation, which is stable in the C10-AcpP docked model, is less energetically favorable for shorter chain lengths. The C8-AcpP docked model was observed to have a small set of favorable structures which were relatively similar to the active model; and C6’s were significantly different. This could explain the observation that FabA crosslinking experiments with C8 analogs were more favorable than with C6 variants³¹, given that C8 can form more stable PPI conformations. Finally, it has been long known that the AcpP • partner enzyme interface is a dynamic interaction^26,32,41. We predict that the various energetically favorable bound states can constitute encounter complexes^42,43, or preliminary interactions that bind the AcpP transiently and allow rearrangement of the AcpP into the active state. This PPI mechanism would eliminate the need for many transient binding events, allowing the AcpP to associate with the interface in order to form a catalytic interaction from more than just one perfectly aligned binding event with a partner enzyme.

These experiments demonstrate the role of the AcpP sequestered substrate in the PPI-controlled catalysis of FabA. Here, the identity of the substrate is pivotal to the positioning of helix III to attain favorable FabA binding. By merging our knowledge of the structural effects of chain length with observations of substrate specificity, we show how FabA, and presumably all other AcpP partner enzymes, maintain a control step over possible substrates tethered to AcpP before the chain flipping process commits the substrate into the active site. This PPI-based specificity mechanism has been hypothesized as an explanation for substrate specificity, although it had not been demonstrated for FabA⁴⁴. While prior demonstration of selective crosslinking of FabA by a substrate-mimicking crosslinker favored C10-AcpP over C6 or C8^31,45, the phenomenon of substrate controlled PPI has not been specifically identified. This study presents a structural model for explaining this specificity, with additional work necessary to fully understand the mechanism. However, this can serve as a foundation to build the understanding of this phenomenon. This first step required leveraging dynamic studies by NMR to examine the solution interactions of acyl-AcpP with FabA, to inform docking simulations, and to provide confirmatory thermodynamics.

Further, we observe the formation of several stable conformations of the acyl-AcpP • FabA interaction that are near or in the path of the crosslinked conformation. These states around the catalytic structure of AcpP may present an expanded paradigm for AcpP • partner interactions that capture encounter complexes that form prior to the chain-flipping event. Stable secondary interfaces could help lead acyl-AcpP, which may initially bind non-ideally, to the active bound conformation required for chain-flipping. This explains the plasticity of the AcpP interface, with few studies identifying single mutations sufficient to abolish activity^46,47. These findings also explain the remarkable efficiency of FAB, with the ability of acyl-AcpP to be “funneled” from improperly coordinated binding events into the required conformation.

Previous microbiological and biochemical studies have demonstrated the role of FabA in scavenging acyl-AcpP pools to perform unsaturation. This work now presents a model for a first control step, maintaining the essential specificity by PPIs evolved to selectively recognize C10-AcpP structure. Unsurprisingly, these findings required a highly interdisciplinary methodology that relied on a fusion of experimental and computational analyses. This model for rapidly sampled chain length specificity must certainly extend beyond FabA, and likely accounts for acyl chain selectivity across some of the 25 known AcpP-dependent enzymes in E. coli, explaining both the speed and precision evolved into FAB. We suspect that these phenomena are involved in all thiotemplated pathways that sequester intermediates within the carrier protein, providing a powerful and fundamental control mechanism.

ABBREVIATIONS

ACP

Acyl carrier protein

AcpP

The E. coli carrier protein

FAB

Fatty acid biosynthesis

FabA

The E. coli isomerizing 3-hydroxydecanoyl dehydratase

HSQC

Heteronuclear single quantum coherence

NMR

Nuclear magnetic resonance spectroscopy

MD

molecular dynamics