One Atom Matters: Modifying the Thioester Linkage Affects the Structure of the Acyl Carrier Protein

Abstract

At the center of many complex biosynthetic pathways, the acyl carrier protein (ACP) shuttles substrates to appropriate enzymatic partners to produce fatty acids and polyketides. Carrier proteins covalently tether their cargo by a thioester linkage to a phosphopantetheine cofactor. Due to the labile nature of this linkage, chemoenzymatic methods have been developed that involve replacement of the thioester with a more stable amide or ester bond. Here we explore the importance of the thioester bond to the structure of the carrier protein using solution NMR spectroscopy and molecular dynamics simulations. Remarkably, the replacement of sulfur for other heteroatoms results in significant structural changes, suggesting more rigorous selections of isosteric substitutes is needed.

Graphical abstract

Unveiling consequences of isosteres: Replacement of thioester sulfur for other heteroatoms results in significant structural perturbations of a protein with over one thousand other atoms. Rigorous evaluation can inform the proper selection of isosteric substitutes for developing chemical biology tools.

Type II acyl carrier proteins (ACPs) are involved in pathways with metabolic, pharmaceutical, and environmental significance.^1,2,3 A clear structural understanding of ACPs and their interactions with partner enzymes is essential to furthering metabolic engineering and drug discovery. Efforts in this regard have proven challenging due to transient interactions and the dynamic nature of the ACP, which transports intermediates to multiple enzymes via a covalent but labile thioester linkage. The archetypical carrier protein, AcpP, involved in type II fatty acid biosynthesis in Escherichia coli, has been well studied,^4,5 but key questions regarding conformational dynamics and interactions between AcpP and partner enzymes remain partially understood. Methods to capture ACP-partner proteins interactions have been developed, often requiring the preparation of acyl-ACPs in which the naturally occurring thioester linkage has been replaced with more stable amide or ester bond. Chemoenzymatic preparations of ACPs bearing crosslinking probes have been used to covalently trap ACP-enzyme complexes.^6,7,8 Using this strategy, crystal structures of E. coli fatty acid ACP (AcpP) crosslinked to FabA,⁹ FabB,¹⁰ and FabZ¹¹ have been determined, revealing discrete molecular interactions mediating AcpP binding to these enzymes. What remains unknown, however, are the dynamics of these interactions throughout ACP-partner protein binding events, including how intermediates attached to the ACP influence protein dynamics and molecular recognition. E. coli AcpP interacts with at least 25 different proteins^9,12,13 and transports acyl intermediates with chain lengths ranging from four to eighteen carbons that assume one of four different β-oxidation states. Given the number of proteins with which AcpP interacts and the chemical diversity of the substrates it carries, it is unlikely that the process by which AcpP delivers its cargo (chain-flipping) is stochastic, but rather is regulated by substrate-influenced protein-protein interactions.¹⁴

Due the geometric and electronic differences between thioester, ester, and amide linkages, evaluation of such modifications is necessary. We sought to determine how the structure of AcpP is affected by the chemical nature of the substrate-AcpP linkage. In this study we used solution-state nuclear magnetic resonance (NMR) spectroscopy and molecular dynamics (MD) to evaluate the influence of thioester, ester, and amide linkages upon AcpP structure and its interactions with a partner enzyme.

First, solution NMR was used to compare the structures of variants of octanoyl-AcpP (C8-AcpP) featuring either a thioester (C8-S-AcpP), amide (C8-NH-AcpP), or ester (C8-O-AcpP) linkage. The thioacyl-AcpP analogues were synthesized using a chemoenzymatic “one-pot” method,¹⁵ while the native thioacyl-AcpP was enzymatically synthesized with the enzyme VhAasS (Scheme 1).¹⁶ Substrate loading was monitored via conformationally-sensitive urea polyacrylamide gel electrophoresis (PAGE). Linkage-based differences in gel migration were observed (Figure S1B). With this preliminary indication of conformational variation between the three ACPs, we further characterized the extent and nature of these differences using ¹H-¹⁵N-heteronuclear single quantum correlation (HSQC) NMR experiments (Figure 1). Perturbations were quantified by chemical shift perturbation (CSP) calculation,¹⁷ in which the chemical shifts of each analogue were compared to naturally occurring thioacyl-AcpP. The spectra of the two non-natural AcpPs show significant chemical shift perturbations of backbone amides (Figure 1–2), indicating a change in the chemical environment of those residues. Notably, the residues that showed significant CSPs were all in regions of carrier protein spatially proximal to the phosphopantetheine (PPant) tethering the substrates. The presence of an amide linkage perturbs residues in helices II and III, while presence of the ester linkage induces both these perturbations as well as additional CSPs of residues in helix IV (Figure 2). We note that helix II is considered the “recognition helix,”¹⁸ and that AcpP primarily interacts with partner proteins through helix II and III.¹⁹

Scheme 1.

Chemoenzymatic synthesis of C8-NH-AcpP and C8-O-AcpP and enzymatic synthesis of C8-S-AcpP. The synthesized phosphopantetheine (PPant) analogs are attached to apo-AcpP with a phosphopantetheinyl transferase (PPTase), whereas an acyl-acyl carrier protein synthetase (AasS), is used to attach octanoic acid to holo-AcpP. Conversion from apo- to holo-AcpP is achieved using PPTase and coenzyme A (CoA), and is reversed by AcpH.

Figure 1.

Comparison of the NMR spectra of three different acylated AcpP species. (A) Overlay of ¹H-¹⁵N HSQC spectra of C8-NH-AcpP and C8-S-AcpP. (B) Overlay of ¹H-¹⁵N HSQC spectra of C8-O-AcpP and C8-S-AcpP. (C), (D) Selected peaks which show significant chemical shift perturbations in both spectra enlarged below each.

Figure 2.

Plots of backbone amide chemical shift perturbations (CSPs). (A) CSP plots determined by comparing chemical shifts of C8-NH-AcpP and C8-S-AcpP. The mean and mean plus one standard deviation are illustrated using solid and dotted lines, respectively. Residues with CSPs larger than one standard deviation from the mean are colored blue. (B) AcpP structure (PDB ID: 2FAD) illustrating the residues (in blue) that possess significant CSPs. (C) CSP plots determined by comparing chemical shifts of C8-O-AcpP and C8-S-AcpP. The mean and mean plus one standard deviation are illustrated using solid and dotted lines, respectively. Residues with perturbations above the standard deviation from the mean are colored red. (D) AcpP structure (PDB ID: 2FAD) illustrating the residues (in red) that possess significant CSPs.

These findings support the hypothesis that AcpP communicates substrate identity to partner proteins via changes in the recognition helix.^20,21 To further investigate the structural consequences of the perturbed residues, we performed 5 independent 500 ns conventional MD simulations of each of the spectroscopically characterized AcpPs: C8-S-AcpP, C8-NH-AcpP, and C8-O-AcpP. We analyzed how the difference in steric hindrance, electrostatics, and hydrogen bonding of the linkage atoms affected the global structure of AcpP.

The difference in migration of each species on urea-PAGE indicated that the global shape and substrate sequestration were affected by substrate linkage in these acyl-AcpPs. MD simulations were consistent with the gel migration pattern, in which the sulfur of C8-S-AcpP is most sequestered, buried deeper in the pocket and further below helix III than the nitrogen of C8-NH-AcpP or the oxygen of C8-O-AcpP (Figure S1 A,D–F). The average calculated radius of gyration and solvent accessible surface area (SASA) of the entire protein are only slightly different, though they follow the same trend increasing in order from C8-NH-AcpP, C8-O-AcpP to C8-S-AcpP (Figure S1B,C).

There are substantial differences in the AcpP pocket volume, with average volumes calculated as 95.6 ų in the native C8-S-AcpP species, compared to 102.6 ų in the C8-NH-AcpP species, and 119.2 ų in the C8-O-AcpP species (Figure 3A). These are more apparent than differences contributed by electrostatics (Figure 3B). The only difference between each species is the linkage between octanoic acid and PPant (S, O, or NH). Therefore, to investigate the steric effect of the decreased atomic radii of nitrogen and oxygen compared to sulfur, we calculated the average number of protein atoms within three different sized shells surrounding the linkage atom throughout the simulation. No atoms occupy the sphere within 1.5 Šof the sulfur in the native species, while an average of three and two atoms occupy the sphere of nitrogen and oxygen, respectively. Considering a 2.5 Šsphere, four atoms are within the sphere in the native species, and double that number of atoms are present in both of the mimicking linkages. The 3.5 Šsphere contains the most similar amount of atoms, with 15 for sulfur, 17 for nitrogen, and 16 for oxygen (Figure 3C).

Figure 3.

Analysis of AcpP pocket from molecular dynamics simulations. (A) Average calculated pocket volume of each species. (B) Comparison of pocket electrostatics of each species. (C) Schematic showing average atom count within 1.5 Å (inner sphere), 2.5 Å (middle sphere), and 3.5 Å (outer sphere) from the center of mass of the linkage atom over 2.5 μs of simulation. (D) Hydrogen bonding between the Ppant and residue T39 in each species.

Hydrogen bonding analysis revealed that the amide linkage in C8-NH-AcpP acted as a hydrogen bond donor that displaced native hydrogen bonding between the first proximal PPant amide (linking β-Ala and cysteamine) and AcpP. In the native C8-S-AcpP species, the first proximal amide, labeled α in Figure 3D, donates a hydrogen bond to backbone carbonyls of A34 (18% of the simulation), A59 (18%), D56 (3%), I62 (2%) and side chain hydroxyl of T39 (7%) (Figure S8, Table S2). In C8-NH-AcpP, hydrogen bonding to the backbone oxygen of I62 and the α amide is completely displaced (0% of simulation data features this contact) by the linkage amide, which spends 21% of the simulation hydrogen bonded to I62. This results in the α amide of C8-NH-AcpP forming a new hydrogen bond, which is not observed in any of the other simulations, with the side chain oxygens of E60 for 4% of the simulation (S8, Table S2). These findings can explain the chemical shift differences observed between C8-S-AcpP and C8-NH-AcpP at residues E60 and I62 (Figure 2). In C8-NH-AcpP, hydrogen bonding to hydroxyl group of T39 spent 27% of the simulation with the α amide and was displaced by the amide linkage during 4% of the simulation. The amide linkage also formed hydrogen bonds with the backbone oxygen of T39 for 7% of the simulation (Figure 3D, S8, Table S2). T39 is another residue which showed a significant chemical shift perturbation when comparing the C8-S-AcpP and C8-NH-AcpP species (Figure 1–2). On average the phosphopantetheine of C8-NH-AcpP has one more hydrogen bond than C8-S-AcpP or C8-NH-AcpP (Figure S9). Calculating the SASA of just the ligand portion shows three distinct states in the native C8-S-AcpP and the C8-O-AcpP. The C8-NH-AcpP has only two distinct states, likely due to stable hydrogen bonding of the amide decreasing conformational transitions. The most occupied SASA state shifts from the larger in the native (and C8-O-AcpP) to a smaller SASA, indicating more sequestration inside the pocket, shielding water accessibility. Due to the partial double bond character of the amide bond, we sought to compare rotation about the linker atom and carbonyl carbon bond (Figure S10), yet found similar rigidity in the thioester and ester species. This is likely due to stereoelectronic effects and dipole minimization restricting rotation of the thioester and ester linkages.^22,23

Based on the differences in the position of PPant on the surface of these acyl-AcpPs and overall shape and substrate sequestration of each AcpP, we hypothesized that the thioester isosteres may lead to a functional difference in AcpP binding to partner proteins.

To test this, we performed Cluspro^24,25,26 protein-protein docking simulations using representative structures from the MD data of either C8-S-AcpP, C8-NH-AcpP, or C8-O-AcpP and apo-FabA (PDB ID: 1MKB). The docking results confirm a canonical AcpP-partner protein binding mode which is primarily stabilized by electrostatic interactions between AcpP and FabA (Figure S11). The native C8-S- AcpP forms the most favorable interface with the lowest docking score (Table S5) and possesses the greatest number of total residues participating in the protein-protein interface (Table S3, Table S4). The next most favorable score and total interface residues is C8-NH-AcpP, followed by C8-O-AcpP.

Though the residues within a 3.5 Å distance at the interface of each AcpP-FabA structure appear slightly different, there is only one key difference in the hydrogen bonding between FabA and AcpP. FabA residue K112 forms a hydrogen bond with Q14 in the native C8-S-AcpP docked structure, while with C8-NH-AcpP, the hydrogen bond is formed with E11. In the C8-O-AcpP structure, no hydrogen bonds are formed with K112 (Figure 4). This particular interaction is unique because it is the only hydrogen bond between AcpP and the non-primary homodimer subunit of FabA. This interaction is not present in the AcpP-FabA crosslinked crystal structure (PDB ID: 4KEH), which captures the AcpP binding to FabA in a mechanistically relevant conformation, as opposed to the pre-chain flipping conformation detected by these docking simulations. Most of the interface residues elucidated by the crosslinked crystal structure are conserved between each of the docking simulations (Table S3, Table S4).

Figure 4.

Differences in docking interface between FabA-AcpP as a result of changing the acyl linker. The difference in all three structures between the K112 salt bridge is shown as yellow dashes to either Q14 in (A) C8-S-AcpP-FabA, E13 in (B) C8-NH-AcpP-FabA, or neither in (C) C8-O-AcpP-FabA.

This study demonstrates that minor chemical modifications of biomolecules can have dramatic effects on protein structure, perhaps explaining how the 77 amino acid AcpP protein can deliver the dizzying variety of possible intermediates to the appropriate partner enzymes in such a highly choreographed fashion. Although the amide linkage significantly alters the position of the fatty acid through hydrogen bonding, the oxoester linkage has a greater magnitude of chemical shift values, significantly expands the pocket of AcpP, and results in poorer docking of AcpP to a partner enzyme. Therefore, we conclude the amide linkage is a more suitable substitute for the thioester than the oxoester in this system, and likely other systems involving sequestration of a chemical intermediate inside a protein cavity. Crosslinking experiments and the structural resolution of crosslinked AcpP-FabA, AcpP-FabZ, and AcpP-FabB suggest that modification of the thioester bond does not disrupt chain flipping and substrate localization into the active site of each enzyme. Though these differences do not seem to prohibit this primary function of AcpP, the finer details of the structure, dynamics, and interaction of AcpP with partner proteins are influenced by the chemical nature of the substrate-cofactor linkage. Such a fine understanding of AcpP will be required to further long-standing goals of metabolic engineering and drug discovery. Consequently, efforts should be directed towards developing more chemically similar analogs and methods to study AcpP with substrates attached via the native thioester.