Active site remodelling of a cyclodipeptide synthase redefines substrate scope

Cyclodipeptide synthases form 2,5-diketopiperazines from aminoacyl-tRNAs and often exhibit promiscuity, yet predicting and reprogramming their substrate specificity remains challenging. Histidine-containing cyclic dipeptides have notable bioactivities (anticancer, neuroprotective), but among 120 functionally validated CDPSs only two accept histidine, yielding cyclo(His-Phe) and cyclo(His-Pro). How CDPSs recognize and select histidine is unclear. This work addresses the research question of how P1/P2 pocket features govern histidine selection and whether rational active-site remodelling can redirect substrate scope. The study aims to: develop a facile in vitro platform to produce histidine-containing CDPs and analogues; use minimal substrates to probe pocket occupancy and kinetics; determine a high-resolution structure of a His-Pro–producing CDPS; and rationally engineer P1 to alter electrostatics/shape to switch specificity toward nonpolar substrates (Phe, Leu), thereby expanding accessible CDP chemical space.

Foundational studies established CDPS catalysis via a ping-pong mechanism involving a catalytic serine and conserved residues (S, Y, E, Y) with two subfamilies (NYH and XYP) defined by P1/P2 motifs. Prior crystal structures (AlbC, Rv2275, YvmC, BtCDPS in NYH; Nbra-, Rgry-, Fdum-CDPS in XYP) clarified the general fold and mechanism, with a conserved tyrosine facilitating cyclization. CDPSs display inherent promiscuity, but computational specificity prediction is unreliable, necessitating empirical characterization. Previous work showed non-canonical amino acid incorporation in vivo and minimal substrate surrogates (aa-DBE) can substitute for aa-tRNAs in some contexts. However, before this study there were no reports of successfully engineering CDPS substrate scope; earlier mutagenesis often reduced activity without changing specificity. Histidinyl-tRNA synthetase literature suggests specific Tyr/Glu-mediated recognition of histidine’s imidazole, informing hypotheses for P1 interactions in CDPSs.

- Enzymes: Two histidine-utilizing CDPSs were produced and purified: ParaCDPS (from Parabacteroides sp. 20_3; produces cyclo(His-Phe)) and ParcuCDPS (from Parcubacteria bacterium RAAC4_OD1_1; produces cyclo(His-Pro) and cyclo(His-Glu)). Identity confirmed by intact protein MS and SDS-PAGE; activities verified in vitro. - tRNA supply: Compared three sources: (i) in vitro transcribed tRNAs (mutant T7 RNAP A172-173 protocol), (ii) E. coli S30 lysate, and (iii) a total E. coli tRNA pool purified by liquid–liquid organic extraction and cleanup (adapted from Mechulam et al.). The tRNA pool was used with purified aminoacyl-tRNA synthetases (aaRSs) for subsequent assays. - Non-canonical amino acid incorporation: Using the tRNA pool with specific aaRSs, screened commercially available histidine, phenylalanine, and proline analogues. A broadened phenylalanine scope utilized PheRS-A294G (widened binding pocket). Acceptance criteria for LC-HRMS product detection: unique peaks vs controls, area >100, and observed mass within 5 ppm of predicted exact mass. - Minimal substrates and hybrid reactions: Synthesized His-DBE, Phe-DBE, Pro-DBE. Conducted reactions using: aa-DBE alone, aa-tRNA alone, and hybrid combinations (aa-DBE with the corresponding or partner aa-tRNA). Quantified products by LC-MS using calibration curves. Considered aa-DBE half-life limitations. - Substrate pocket occupancy assay: Employed trapped acyl-enzyme intermediate experiments using aa-DBEs as chemical probes; analyzed covalent aminoacylation of catalytic Ser by intact protein MS to determine which amino acid binds first in P1. - Kinetics: Monitored time courses for cyclic dipeptide formation under hybrid conditions; fitted early-time data to exponential models to estimate rates. - Structural biology: Solved ParcuCDPS crystal structure at 1.90 Å resolution by iodide SAD phasing; analyzed active site and pocket topology (CASTp for pocket volumes; PROPKA/APBS for pKa and electrostatic potential). Compared with NYH/XYP structures; assessed RMSDs and secondary structure deviations. - Mutagenesis and engineering: Designed three generations of variants targeting P1 residues (Y55, E174, Y189) and catalytic residues (S26, D58, Y167, E171). Generation 1: single mutants (Y55F/V; E174A/H/L; Y189F/L; and catalytic-site controls S26A/C, D58A/N, Y167A/F, E171A/Q). Generation 2: double mutants combining P1 positions (e.g., Y55F+Y189F; Y55F+E174A; Y189F+E174A; E174L+Y189L). Generation 3: double/triple hydrophobic mutants (Y55V+E174L; Y55V+Y189L; E174L+Y189L; Y55V+E174L+Y189L) to switch specificity to nonpolar substrates. Verified mutations by intact protein MS. Solved select mutant structures by MR using WT. - Activity assays: Quantified cHP, cHF, and engineered products cLP and cFP by LC-MS with authentic standards and calibration curves; performed triplicates and reported SEM.

- tRNA sourcing: The purified E. coli tRNA pool enabled robust CDP formation and outperformed the S30 extract; it was cheaper, easier, and faster than in vitro-transcribed tRNA while yielding the same products. - Non-canonical substrate incorporation: Both CDPSs accepted histidine analogues H-β-(2-thiazolyl)-DL-alanine and 3-(2-pyridyl)-L-alanine; ParcuCDPS also accepted β-(1,2,4-triazol-3-yl)-DL-alanine. Only the 2-pyridyl isomer was incorporated among pyridyl variants, highlighting the importance of a ring nitrogen near the α-carbon for recognition. ParaCDPS, with PheRS-A294G, incorporated most reported Phe analogues, readily tolerating para-halogen substitutions but not para polar groups (amine, nitro, azido). ParcuCDPS tolerated conservative proline derivatizations (including 4-bromo-L-proline) but not hydroxyl/amine ring substitutions; it did not accept tested glutamate analogues, indicating narrow specificity. - Minimal substrates and hybrid reactions: Reactions using only aa-DBEs gave very low yields, whereas hybrid aa-DBE + aa-tRNA combinations produced more product though less than tRNA+tRNA reactions due to aa-DBE instability. - Pocket occupancy (acyl-trap): ParaCDPS binds phenylalanine first in P1, whereas ParcuCDPS binds histidine first in P1, as shown by trapped acyl-enzyme intermediates detected by intact protein MS. - Kinetic rates (min−1) for product formation (early phase fits): For CHF (ParaCDPS): His-tRNA + Phe-tRNA = 0.0024; His-tRNA + Phe-DBE = 0.0048; His-DBE + Phe-tRNA = 0.0022; His-DBE + Phe-DBE = too low to fit. For cHP (ParcuCDPS): His-tRNA + Pro-tRNA = 0.05; His-DBE + Pro-tRNA = 0.00002; His-tRNA + Pro-DBE and His-DBE + Pro-DBE = too low to fit. These indicate different rate-limiting steps for the two enzymes. - Structure of ParcuCDPS: 1.90 Å resolution structure revealed a Rossmann-like fold with conserved catalytic residues (S26, Y167, E171, Y191). Unique deviations include a bend splitting α3 (via G84) and a direction change in β3 near residue 156, positioning D58 within H-bond distance to S26. Mutating D58 (to A or N) disrupted the S26–D58 H-bond and reduced activity by >90% while preserving overall structure, suggesting a functional Ser/Asp dyad. - Pocket properties: CASTp showed P1 is deeper and narrower than P2; PROPKA/APBS indicated P1 is mostly neutral with a negatively charged microenvironment from Y55 and E174, whereas P2 is more positively charged, consistent with accommodating negatively charged Glu-tRNA in P2 for cHE formation. - Mutational analysis and engineering: Active-site mutants S26C and Y167F retained reduced activity; D58A/N and other catalytic residue mutations largely abolished activity, underscoring D58’s essential role in activating S26. P1 single mutants E174A and E174H abolished histidine acceptance (cHP formation), implicating E174 in H-bonding/electrostatics for His recognition; broader P1 mutations (Y55, Y189) reduced activity, with combined mutations (Generation 2) further diminishing cHP production, indicating cooperative contributions to His binding. - Specificity switch (Generation 3): Hydrophobic P1 remodeling enabled acceptance of leucine and phenylalanine (not valine/isoleucine). Using Pro-tRNA with Leu-DBE or Phe-DBE yielded higher product levels than with Leu-tRNA/Phe-tRNA, indicating residual tRNA-body selectivity. Product distributions: Y55V+E174L produced 84% cyclo(Leu-Pro) among total products; E174L+Y189L produced 70% cyclo(Phe-Pro). Trapped acyl intermediates confirmed switched P1 selection. This constitutes the first demonstration of engineered substrate scope change in a CDPS, yielding bioactive products cLP and cFP.

The study elucidates how a histidine-accepting CDPS recognizes its polar substrate in P1 and demonstrates that targeted remodeling of P1 electrostatics and shape can redirect substrate selection to nonpolar amino acids. Structural insights (unique β3 reorientation and α3 bend; S26–D58 interaction) clarify catalytic architecture and rationalize the impact of D58 on nucleophile activation. Electrostatic mapping and mutagenesis implicate Y55, E174, and Y189 in forming a negatively biased microenvironment and H-bonding network compatible with histidine’s imidazole; disrupting these interactions suppresses histidine binding and activity. Hybrid aa-DBE/aa-tRNA kinetics and acyl-trapping experiments disentangle binding order and rate-limiting steps, revealing that ParaCDPS and ParcuCDPS differ in which reaction half is slow and which substrate occupies P1 first. By progressively attenuating histidine recognition (Generations 1–2) and then introducing hydrophobic substitutions (Generation 3), the authors reprogrammed P1 to accept leucine and phenylalanine, while maintaining substrate selectivity against other hydrophobics (valine/isoleucine). These results address the central question of substrate selection in CDPSs and provide a roadmap for engineering enzymes to produce desired cyclic dipeptides, with implications for expanding bioactive DKP libraries and improving functional prediction across the CDPS family.

The authors established an efficient in vitro platform using a pooled E. coli tRNA extract to generate histidine-containing cyclic dipeptides and diverse analogues, including non-canonical amino acids. Structural characterization of ParcuCDPS revealed distinctive pocket topology and an S26–D58 interaction essential for catalysis, clarifying histidine recognition in P1. Acyl-trap experiments showed differing P1 occupancy for ParaCDPS (Phe) and ParcuCDPS (His), aligning with distinct kinetic bottlenecks. Guided by structure and electrostatics, rational mutagenesis of P1 (Y55, E174, Y189) first disrupted histidine selection and then successfully switched specificity to leucine and phenylalanine, yielding cLP and cFP—constituting the first engineered substrate scope change in a CDPS. Future work could expand to other residues and substrates, explore tRNA-body recognition to improve yields with aa-tRNA for new substrates, and determine complex structures with bound substrates/tRNAs to refine engineering and predictive models.

- Reactions employing minimal aa-DBE substrates yielded low product due to limited aa-DBE half-life; hybrid aa-DBE + aa-tRNA conditions improved but did not match tRNA+tRNA yields. - Engineered variants accepted leucine and phenylalanine but did not accept other small hydrophobics (valine, isoleucine), indicating that specificity broadening remains limited and context-dependent. - ParcuCDPS displayed narrow tolerance toward glutamate analogues despite producing cHE, suggesting constraints not fully resolved by current structural data. - Structural insights are from apo and mutant structures; lack of high-resolution complexes with tRNA/substrates limits direct visualization of binding interactions and tRNA-body contacts.