Trapping biosynthetic acyl-enzyme intermediates with encoded 2,3-diaminopropionic acid

Abstract

Many enzymes catalyse reactions that proceed through covalent acyl-enzyme (ester or thioester) intermediates¹. These enzymes include serine hydrolases^2,3 (encoded by one per cent of human genes, and including serine proteases and thioesterases), cysteine proteases (including caspases), and many components of the ubiquitination machinery^4,5. Their important acyl-enzyme intermediates are unstable, commonly having half-lives of minutes to hours⁶. In some cases, acyl-enzyme complexes can be stabilized using substrate analogues or active-site mutations but, although these approaches can provide valuable insight^7-10, they often result in complexes that are substantially non-native. Here we develop a strategy for incorporating 2,3-diaminopropionic acid (DAP) into recombinant proteins, via expansion of the genetic code¹¹. We show that replacing catalytic cysteine or serine residues of enzymes with DAP permits their first-step reaction with native substrates, allowing the efficient capture of acyl-enzyme complexes that are linked through a stable amide bond. For one of these enzymes, the thioesterase domain of valinomycin synthetase¹², we elucidate the biosynthetic pathway by which it progressively oligomerizes tetradepsipeptidyl substrates to a dodecadepsipeptidyl intermediate, which it then cyclizes to produce valinomycin. By trapping the first and last acyl-thioesterase intermediates in the catalytic cycle as DAP conjugates, we provide structural insight into how conformational changes in thioesterase domains of such nonribosomal peptide synthetases control the oligomerization and cyclization of linear substrates. The encoding of DAP will facilitate the characterization of diverse acyl-enzyme complexes, and may be extended to capturing the native substrates of transiently acylated proteins of unknown function.

Extended Data Fig. 1. Schematic representation and reaction cycle of a canonical NRPS.

a, Schematic representation of a generic type I NRPS. The square brackets denote a single module. b, i–vii, Synthetic cycle of a canonical elongation module. NRPSs assemble peptides from amino acyl and other small acyl building blocks using a modular and thio-templated logic. A canonical NRPS is composed of one module for every residue in the peptide product. The initiation module contains an adenylation (A) domain, which binds cognate acyl substrate and performs adenylation and transfer of that substrate as a thioester on the phosphopantetheine arm (PPE, shown as a wavy line) of a peptidyl carrier protein (PCP) domain, for transport between active sites. Each elongation module contains an A and a PCP domain, and also a condensation (C) domain, which condenses aminoacyl and peptidyl substrates bound to PCP domains, thus progressively elongating the nascent chain. Termination modules contain C, A and PCP domains, and a specialized terminating/offloading domain responsible for the release of the peptide in its final form. The most common and most versatile terminating domain in NRPSs is the TE domain. Similar TE domains terminate synthesis in polyketide and fatty acid synthases. PP_i, diphosphate; aa, amino acid.

Extended Data Fig. 2. Genetically directing DAP incorporation in recombinant proteins.

a, Structure of DAP and the protected versions investigated herein. 1, 2,3-diaminopropionic acid (DAP); 2, (S)-3-(((allyloxy)carbonyl)amino)-2-aminopropanoic acid; 3, (S)-2-amino-3-((2-nitrobenzyl)amino)propanoic acid; 4, (2S)-2-amino-3-((1-(6-nitrobenzo[d][1,3]dioxol-5-yl)ethyl)amino)propanoic acid; 5, (2S)-2-amino-3-(((1-(6-nitrobenzo[d][1,3]dioxol-5-yl)ethoxy)carbonyl)amino)propanoic acid; 6, (2S)-2-amino-3-(((2-((1-(6-nitrobenzo[d][1,3]dioxol-5-yl)ethyl)thio)ethoxy)carbonyl)amino)propanoic acid. Calculated logP values are indicated (calculated using the Molinspiration molecular property calculation services at www.molinspiration.com/cgi-bin/properties). b–f, Determining the intracellular concentration of compounds 2–6 by an LC–MS assay, performed on extracts. The dark-blue trace represents a 100 µM standard for each compound. The light-blue trace represents a 10 µM standard for each compound. The red trace results from cells grown in the absence of the compound. The brown trace results from cells grown in the absence of the compound, but spiked with the compound to 100 µM. The green trace results from cells grown in the presence of 1 mM compound. The experiments were repeated in two biological replicates with similar results. g, Phenotyping of the DAPRS/tRNA_CUA pair. Cells containing the DAPRS/tRNA_CUA pair and cat(112TAG) (encoding a chloramphenicol-resistance gene containing an amber stop codon (TAG) at codon 112) were plated in the presence or absence of 6 on the indicated concentrations of chloramphenicol. The experiment was performed in two biological replicates with similar results. h, The side chain of 6 (grey sticks) was modelled into the active site of PylRS using a co-crystal structure of PylRS and adenylated pyrrolysine (PDB accession number 2ZIM). PylRS is displayed in pale yellow and amino-acid positions randomized in DAPRSlib are shown in marine blue.

Extended Data Fig. 3. Stably trapping the acyl–enzyme intermediate of a cysteine protease.

a, Different variants of TEV protease (shown at the top) were reacted with Ub–tev–His. The use of TEV(wt) results in cleavage of the TEV cleavage sequence. The use of TEV(C151A) results in minimal cleavage. The presence of DAP in the active site of TEV results in the presence of an extra band in the Coomassie gel, representing the isopeptide-linked TEV(C151DAP)–Ub complex. b, c, Anti-streptavidin (α-strep; b) and anti-Ub (α-Ub antibody P4D1; c) western blots of the reactions confirm the identity of the complex. For panels a–c, the experiment was repeated in two biological replicates with similar results. d, Tandem mass spectrometry following tryptic digest of the TEV(C151DAP)–Ub conjugate confirms amide-bond formation at the expected position. Top, the sequence of the branched peptide subject to fragmentation. Fragmentation of the substrate chain is predicted to lead to a series of y ions (yellow) and a series of b ions (green); the ions from this chain are labelled as ‘β’. Fragmentation of the TEV(C151DAP)-derived chain is predicted to lead to a series of y ions (blue) and a series of b ions (red); the ions from this chain are labelled as ‘α’. Bottom, MS/MS spectra with peak assignments. Ions in the α-chain were assigned by treating DAP and the β-chain as a modification of known mass. Ions in the β-chain were manually assigned. The mass-spectrometry analysis was performed once.

Extended Data Fig. 4. Chemical structures of key Vlm TE substrates described here.

The chemical structures and the numbers used to refer to them are shown.

Extended Data Fig. 5. The mechanism by which by Vlm TE catalyses oligomerization.

Oligomerization could conceivably take place in two ways. a, In the first scenario, ‘forward transfer’, the distal hydroxyl group of the tetradepsipeptidyl–O-TE complex attacks the thioester group in the tetradepsipeptidyl–S-PCP enzyme intermediate, directly forming octadepsipeptidyl–O-TE as a product. b, In the second scenario, ‘reverse transfer’, the distal hydroxyl group of the tetradepsipeptidyl–S-PCP complex attacks the ester group in the tetradepsipeptidyl–O-TE enzyme intermediate, forming octadepsipeptidyl–S-PCP as a product, which would then need to be transferred onto the TE-domain serine (here labelled as ‘re-capture’). c, d, Analogous scenarios involving tetradepsipeptidyl–SNAC (7) as the substrate instead of tetradepsipeptidyl–S-PCP. e, f, EICs (HR LC–ESI–MS) of a mix of 7 (1.7 mM) and buffer (e), or the products of a reaction between 7 (1.7 mM) and Vlm TE_DAP (6.5 μM) (f). g–i, EICs (low-resolution (LR) LC–ESI–MS) of reactions using a higher-volume injection into an ion-trap MS instrument. g, The higher-volume injection of a reaction of 7 (1.7 mM) and Vlm TE_wt (6.5 μM) enabled detection of a peak consistent with the 20-mer depsipeptidyl–SNAC (24). h, LC-Ion-trap MS of reaction of 7 (1.7 mM) and Vlm TE_DAP (6.5 μM). i, Small amounts of the cyclic 16-mer depsipeptide 29 elute during post-run column clean-up of experiment shown in g. j, EICs (HR LC–ESI–MS) of products of reactions between Vlm TE_wt (6.5 μM) and a mix of 7 and deoxy-tetradepsipeptidyl–SNAC (8; 1.7 mM of each). TE_wt produces the intermediates deoxy-octadepsipeptidyl–SNAC (12), deoxy-dodecadepsipeptidyl–SNAC (16) and deoxy 16-mer depsipeptidyl–SNAC (20), confirming the reaction pathway shown in panel b. See ‘Supplementary Methods for Statistics and Reproducibility’ for accurate mass analysis and deviations from calculated m/z values of each compound. The experiments in panels e–i were repeated independently two times with similar results. Mass-spectrometry analysis of the experiment in panel j was performed once.

Extended Data Fig. 6. Structures of Vlm TE_wt and tetradepsipeptidyl–TE_DAP, and top-down LC–ESI–MS of Vlm TE_wt.

a, Secondary-structure elements of Vlm TE; the naming is based on the convention for α/β-hydrolase proteins. b, Comparison of two TE_wt structures (PDB accession numbers 6ECB and 6ECC). The active-site lid of the first structure (light grey) is nearly completely ordered, while the lid of second structure (dark grey) shows density for Lα3, Lα4 and Lα5 only. In the second structure, Lα3 is rotated 10° towards the active site. c, Deconvoluted mass spectra of TE_wt incubated with different substrates. Solid line, buffer control: expected molecular mass 31,028.22 Da; observed 31,028.75 Da. Dashed line, TE_wt incubated with tetradepsipeptidyl–SNAC: expected 31,028.22 Da (unmodified) and 31,399.44 Da (modified); observed 31026.29 Da. Dotted line, TE_wt incubated with valinomycin: expected 31,028.22 Da (unmodified) and 32139.86 Da (modified); observed 31,027.01 Da. Experiments were repeated independently two times with similar results. d, Comparison of near-identical conformations of TE_wt (light grey; 6ECB) and tetradepsipeptidyl–TE_DAP (tan and dark grey; 6ECD).

Extended Data Fig. 7. Expression and substrate conjugation to Vlm TE containing DAP at position 2,463.

a, Following expression and purification of Vlm TE_DAP, the protein was loaded on an SDS–PAGE gel and Coomassie stained; the experiment was repeated in two biological replicates with similar results. b, The deprotection of 6 in TE_DAP–strep was followed by ESI–MS analysis. Green trace, purified TE_DAP–strep containing 6 at position 2,463: expected mass 32,364.6 Da, observed 32,365.78 Da. Red trace, TE_DAP–strep containing 6 at position 2,463 following illumination to convert 6 to the intermediate: expected 32,171.56 Da, observed 32,168.48 Da; and further incubation (1 h, 4 °C) to convert the intermediate to product: expected 32,067.62 Da, observed 32,068 Da). Blue trace, TE_DAP–strep containing 6 at position 2,463 following illumination (to convert 6 to the intermediate) and further incubation (10 h, 4 °C) to convert the intermediate to DAP (1): expected 32,067.62 Da; observed, 32,067.84 Da. The experiment was repeated in two biological replicates with similar results. c, Purified TE_DAP after illumination and intermediate fragmentation: expected 31,027.24 Da, observed 31,026.95 Da and 31,131.82 Da. d, TE_DAP incubated with tetradepsipeptidyl–SNAC 7: expected 31,027.24 Da (unmodified) and 31,398.69 Da (modified); observed 31,025.92 Da and 31,396.55 Da. The experiments in panels c, d were repeated independently two times with similar results.

Extended Data Fig. 8. Electron density of the active site of covalent depsipeptidyl–TE_DAP complexes.

Unbiased mF_o–DF_c maps (green mesh, contoured at 2.5σ), calculated before depsipeptide residues were placed in the model. DAP (brown) and depsipeptide residues (cyan) are depicted as sticks. a, Tetradepsipeptidyl–TE_DAP (PDB accession number 6ECD). b–g, Dodecadepsipeptidyl–TE_DAP P₁ space-group structure (6ECF), with crystallographically independent molecules A to F shown in sequential order. h, i: Dodecadepsipeptidyl–TE_DAP H₃ space group (6ECE), for crystallographically independent molecules A and B. j–l, Electron density of the active site of covalent depsipeptidyl–TE_DAP complexes extends beyond modelled depsipeptides. Unbiased mF_o–DF_c maps (green mesh, contoured at 2.5σ), calculated before depsipeptide residues were placed in the model, for dodecadepsipeptidyl–TE_DAP P₁ space-group structure, with crystallographically independent molecules A, B and D in sequential order. The observed electron density that extends beyond the modelled depsipeptides (cyan sticks) could accommodate extra depsipeptide residues in different orientations. However, unambiguous modelling into this density could not be achieved.

Extended Data Fig. 9. Modelling of interaction between the PCP domain and TE domain and putative pathway.

a, Superimposition of dodecadepsipeptidyl–TE_DAP with the structure of the EntF PCP–TE didomain (PDB accession number 3TEJ) shows the path of the PPE moiety to the active site. b, Hypothetical pathway for oligomerization and cyclization, starting from octadepsipeptidyl–TE. i, The position of Lα1 in the observed apo/tetradepsipeptide conformation promotes an extended peptide conformation. ii, The tetradepsipeptidyl–PCP accepts the octadepsipeptide onto its terminal hydroxyl, perhaps using a dodecadepsipeptide-like lid conformation which could accommodate the roughly 30-Å tetradepsipeptidyl–PPE bound to the PCP domain and guide it towards the active site. iii, The PCP domain presents the thioester for transfer back to serine 2,463. iv, Finally, the lid conformation observed in the dodecadepsipeptide–TE_DAP structures could help to curl the dodecadepsipeptide back towards serine 2,463 for cyclization.

Fig. 1. Capturing transient acyl–enzyme intermediates with DAP, and the proposed biosynthesis of valinomycin.

a, Active-site serine or cysteine residues react with carbonyl groups to form tetrahedral intermediates (not shown) that collapse to acyl–enzyme intermediates by loss of R₁–YH. Attack by nucleophilic R₃ groups (commonly a hydroxyl, amine or thiol) releases the bound substrate fragment and regenerates the enzyme. R₁, R₂, and Y each represent the diverse chemical groups that may be found in distinct reactants. b, Replacing cysteine or serine with DAP may result in a first acyl–enzyme intermediate that is resistant to cleavage. c, Valinomycin synthetase (Vlm) condenses d-α-hydroxyisovaleric acid (d-α-hiv), d-valine (d-val), l-lactic acid (l-lac) and l-valine (l-val) to form the tetradepsipeptidyl (d-hiv–d-val–l-lac–l-val) intermediate. d-α-hiv and l-lac arise from the reduction of precursor ketoacyl moieties by ketoreductase (KR) domains. Tetradepsipeptidyl intermediates are oligomerized to a dodecadepsipeptidyl intermediate which is cyclized, by the terminal TE domain, to produce valinomycin. Vlm1 and Vlm2 are the two protein subunits that form valinomycin synthetase. A module is a set of domains which work together to add one monomer to the growing depsipeptide. A, adenylation domain; C, condensation domain; PCP, peptidyl carrier protein domain. See Extended Data Fig. 1 for a synthetic cycle of an NRPS.

Fig. 2. Genetically directing DAP incorporation in recombinant proteins and stably trapping the acyl–enzyme intermediate of a cysteine protease.

a, SDS–PAGE gels of GFP150 (6) and GFP150 (BocK), with protein detected by Coomassie staining (top gel) or anti-His₆ antibody (bottom gel); the experiment was performed in two biological replicates with similar results. We used the indicated enzymes (DAPRS and PylRS) with their cognate tRNA_CUA and amino acids (6 and BocK (Nε-[(tert-butoxy)carbonyl]-l-lysine) together with an sfGFP150TAG reporter construct. b, Encoded 6 was photo-deprotected, leading to an intermediate, which spontaneously fragments to reveal DAP. c, Deprotection of 6 in sfGFP followed by electrospray ionization–mass spectroscopy (ESI–MS) analysis. Green trace, purified GFP150(6): expected molecular mass 28,096.27 Da; observed 28,097.21 Da. Light blue trace, intermediate: expected 27,902.22 Da; observed 27904.14 Da. Dark blue trace, incubation (10 h, 37 °C) converts the intermediate to DAP (1): expected 27,798.23 Da; observed 27,800.88 Da. Minor peaks resulting from loss of the N-terminal methionine are also observed. The experiment was performed in two biological replicates with similar results. d, TEV protease variants were incubated with Ub–tev–His. TEV(C151DAP)–Ub is the amide-bond-linked complex. Anti-Ub and anti-strep western blots confirm the identity of the complex (TEV constructs contain a streptavidin tag). The experiment was performed in two biological replicates with similar results.

Fig. 3. Vlm TE produces valinomycin and intermediates that delineate the oligomerization pathway from tetradepsipeptidyl–SNAC.

a, Extracted ion chromatograms (EICs) from high-resolution (HR)–liquid chromatography (LC)–ESI–MS of reactions of tetradepsipeptidyl–SNAC (7; 1.7 mM) and Vlm TE (6.5 μM); TE_wt produces valinomycin as its major product. The experiment was performed two independent times with similar results. See ‘Supplementary Methods for Statistics and Reproducibility’ for mass analysis and deviations from calculated m/z values. b, Two scenarios for oligomerization^,. In the ‘forward transfer’ scenario, the distal hydroxyl group of tetradepsipeptidyl–O-TE (TE) attacks (dotted line) the thioester group in tetradepsipeptidyl–S-PCP (PCP), directly forming octadepsipeptidyl–O-TE (right). In the ‘reverse’ scenario, the distal hydroxyl group of tetradepsipeptidyl–S-PCP attacks the ester group in tetradepsipeptidyl–O-TE, forming octadepsipeptidyl–S-PCP (left), which would later be transferred onto the TE domain serine. Our data are consistent with the ‘reverse’ oligomerization scenario; see also Extended Data Fig. 5.

Fig. 4. Crystal structures of complexes of TE_DAP.

a, b, Deconvoluted mass spectra. a, TE_DAP incubated with deoxy-tetradepsipeptidyl–SNAC (8). Expected molecular masses 31,027.24 Da (unmodified) and 31,383.70 Da (modified); observed 31,029.69 Da and 31,382.69 Da. b, TE_DAP incubated with valinomycin. Expected molecular masses 31,027.24 Da (unmodified) and 32,139.11 Da (modified); observed 31,024.12 Da and 32135.94 Da. The experiments were repeated independently five times with similar results. c, d, Unbiased electron-density (mF_o–DF_c) maps (green mesh, 2.5σ) for depsipeptide residues of tetradepsipeptidyl–TE_DAP (c, Protein Data Bank (PDB) accession number 6ECD) and dodecadepsipeptidyl–TE_DAP (d; 6ECE and 6ECF). An amide bond links DAP (brown) and depsipeptide residues (cyan). e, f, The active sites of tetradepsipeptidyl–TE_DAP (e) and dodecadepsipeptidyl–TE_DAP (f). The carbonyl oxygen of the amide formed by DAP and valine 4 (e) or valine 12 (f) is positioned close to the oxyanion hole formed by the main chain of A2,399 and L2,464. Catalytic triad residues H2,625 and D2,490 are shown as sticks. g, The lid of tetradepsipeptidyl–TE_DAP (6ECD) is in a similar position to that seen in TE_wt (6ECB; not shown). h, All crystallographically independent molecules of the dodecadepsipeptidyl–TE_DAP (6ECE and 6ECF) are in a set of similar conformations, distinct from that seen in TE_wt. i, Substantial conformational changes occur in lid helixes Lα1–Lα4 between the conformations of tetradepsipeptidyl–TE_DAP (g) and dodecadepsipeptidyl–TE_DAP (h). See also Supplementary Videos 1 and 2. j, In the dodecadepsipeptidyl–TE_DAP structure, the lid sterically prevents the dodecadepsipeptide from extending out in a linear fashion, instead favouring it curling back through this steric block and forming largely hydrophobic, nonspecific interactions with the lid.