Dynamics and mechanistic interpretations of nonribosomal peptide synthetase cyclization domains

Abstract

Ox-/thiazoline groups in nonribosomal peptides are formed by a variant of peptide-forming condensation domains called heterocyclization (Cy) domains and appear in a range of pharmaceutically important natural products and virulence factors. Recent cryo-EM, crystallographic, and NMR studies of Cy domains make it opportune to revisit outstanding questions regarding their molecular mechanisms. This review covers structural and dynamical findings about Cy domains that will inform future bioengineering efforts and our understanding of natural product synthesis.

Introduction

Nonribosomal peptide synthetase (NRPS) heterocyclization (Cy) domains (Fig. 1) afford ox-/thiazoline groups in a variety of natural products [1–3]. Found in three oxidation states (Fig. 1A, bond-line structures), these heterocycles can increase potency for biological targets [4,5]. Nature has developed alternative strategies for incorporating ox-/thiazoline groups within peptides depending on whether they are synthesized ribosomally or nonribosomally [6–9]. Ribosomal systems use in trans interactions after polymerization [10], whereas nonribosomal systems install ox-/thiazoline groups as the peptide is produced [11].

Figure 1. Structures of NRPS Cy domains.

A) Crystal structures of Cy domains [12,13,16,17] alongside their products. Heterocycles are colored pink. In BmdB, the N- and C-terminal subdomains are shaded blue and red, respectively. The latch and floor loop are purple and orange, respectively. B) Three conformations of homodimeric PchE modules determined by cryo-EM [15], with the dimer interface marked by a dashed line. ArCPs are represented in surface. C) Catalytic scheme for Cy domains, including currently available structures resembling proposed catalytic states. Atoms for the heterocycle are colored magenta, and atoms of the water leaving group are blue.

Recent studies of JBIR-34/35 [12], yersiniabactin [13,14], pyochelin [15], epothilone [16], and bacillamide [17,18] synthetases have expanded our knowledge of Cy domain structure and function (Fig. 1A, B), complementing previous biochemical studies of NRPS Cy systems ([19] and Table S1 from [17]). These structures are of interest for bioengineering [20–24] as type 1 NRPSs follow a modular, assembly-line logic involving phosphopantetheine (Ppant)-loaded carrier proteins (CPs) that tether amino acid substrates installed by adenylation (A) domains [25,26], and modifying substrate incorporation affords new products [27,28]. Peptide bond formation requires two CPs and intervening condensation (C) domains (Fig. 1C, first step) or Cy domains. Thiazoline-producing Cy domains catalyze both condensation and cyclodehydration steps (Fig. 1C) [29], whereas oxazoline-producing Cy domains are typically found in tandem, with each domain catalyzing condensation or cyclodehydration [12,30].

C and Cy domains share little-to-no sequence similarity [16,17,31], enabling exaptation of the parent C-domain fold to additionally catalyze cyclodehydration. Both domains adopt a fold comprising two chloramphenicol acetyltransferase (CAT)-like subdomains (N- and C-terminal) connected by a hinge linker with two additional points of subdomain crossover, known as the floor loop and latch (Fig. 1A). The Cy-domain active site lies between the β-sheets of the N-terminal and C-terminal subdomains and is connected by tunnels to acceptor and donor CP binding sites. Cy domains contain a DxxxxDxxS motif [29], with DxxS in helix α4, identified with a structural role, an aspartate-threonine catalytic dyad required for cyclization, and a gating phenylalanine blocking the active site from bulk solvent, presumably to facilitate dehydration [16,17]. Here, we highlight observations in the current body of literature that link global conformational changes to local remodeling of CP and substrate binding sites and propose a remodeling of the active site architecture for condensation and cyclodehydration within a dynamic protein scaffold.

Cy structures provide insight into global conformational change

The first Cy-domain crystal structures revealed a closed donor site entrance, blocked by a gating phenylalanine [16,32] which we refer to as a “closed-for-cyclodehydration” state. Recently, cryo-EM structures of PchE [15] captured a Cy domain bound to a substrate-loaded donor (Fig. 1C). Here, an “open-for-condensation” state displays a bound donor Ppant, with a C-terminal subdomain mobile lobe (containing strand β10) splayed away from strands β8 and β9 flanking the floor loop (Fig. 2A). Aligning all existing models [12–18] using β8 and β9 (Fig. 2A) displays rotation of the N-terminal subdomain, a twisting of the floor loop, and displacement of the mobile lobe (Fig. 2) [13,14], motions observed by normal mode analysis of HMWP2-Cy2 [13] and in C domains [32,33]. In some normal modes, these motions accompany rearrangement of α1, α10, and the β-turn (BT) at the acceptor site entrance (ASE, Fig. 2B) as well as opening of the donor site entrance (DSE) and rearrangement of C-terminal subdomain helices (e.g. α9/α11, Fig. 2) [13,14,17]. Thus, global rearrangements can remodel entrances to the active-site tunnel.

Figure 2. Motions common to Cy domains.

A,B) Structures [12,13,15–18] aligned by strands flanking the floor loop, viewed from the donor site entrance (A) and acceptor site entrance (B). Pink arrows highlight variations between models. MLS = mobile lobe strands, DSE = donor site entrance, ASE = acceptor site entrance.

The Cy-specific DxxxxDxxS motif, which when mutated led to decreased structural integrity or activity [12,15–18,30,34–37], appears to be part of the open-for-condensation to closed-for-cyclodehydration transition. In absence of CP, the second aspartate of the DxxxxDxxS motif usually engages the serine within the motif and an arginine in the floor loop (FL Arg) (Fig. 3A). Structures of the donor-bound and acceptor-bound PchE-Cy1 models (Fig. 3B) reveal conformational changes in an open-to-closed transition, where the second motif aspartate interacts with the Ppant arm. The serine then interacts with the backbone at the end of the floor loop, substituting for the lost interaction between the FL Arg and the aspartate, leading to a conformation that could indirectly affect the gating phenylalanine and open the DSE. The first aspartate of the DxxxxDxxS motif appears to transiently interact with a conserved arginine in α1 near the ASE, suggesting a possible role in sensing acceptor CP binding (Fig. 3D). Thus, DxxxxDxxS conformational changes accompany remodeling of the DSE and ASE upon CP and tethered Ppant binding, perhaps participating in communication between both sites.

Figure 3. Cy-domain motions may participate in substrate binding and allostery.

Comparison of donor site entrances of A) HMWP2-Cy2, B) PchE-Cy1 and C) LgrA-C [13,15,38] highlights twisting of the floor loop, with different floor loop arginine and motif aspartate conformations, as well as different Ppant and thioester positions (C). Electrophile positions are marked by asterisks and floor loops are colored orange. D) Acceptor site entrances of PchE-Cy1, HMWP2-Cy2, EpoB-Cy, FmoA3-Cy2 and BmdB-Cy2 [12,13,15–17]. A semi-conserved phenylalanine in loop 1 is displayed to help emphasize variation in loop 1 structure. E) Docking model of a cyclodehydration intermediate-bound state for HMWP2-Cy2 (acceptor PCP residues and Ppant are in yellow, cartoon for PCP α3 containing the displayed phenylalanine is omitted for clarity) F) Loaded and unloaded holo acceptor CP complexes in PchE highlight differences in pantetheine placement. Acceptor CP and Ppant are displayed as in panel E, and the pantetheine-only model is displayed with transparency.

Features of condensation and cyclodehydration

The open-for-condensation state of PchE [15] contrasts a prior CP-C donor-mimic LgrA complex with respect to donor electrophile positioning (Fig. 3B,C) [38]. Whereas the LgrA complex positions the electrophile near the N-terminus of α4, the PchE model instead places its thioester oxygen near the putative catalytic aspartate, inconsistently positioned for nucleophilic attack. In the closed-for-cyclodehydration state in HMWP2-Cy2 (Fig. 3A), the leaving group oxygen of a modelled cyclodehydration intermediate sits at the N-terminus of α4 [13]. The dipole of ⍺4 could then stabilize negative charge accumulation on the oxygen in the tetrahedral intermediates of both condensation and cyclodehydration steps [13,38]. These observations support an amine-first condensation near α4 (mirroring C-domain activity), followed by cyclization via the Cys acceptor’s nucleophilic thiol, with deprotonation by the putative catalytic dyad [13]. Future investigations should consider these alternative electrophile orientations before interpreting mechanisms.

Importantly, modeling of cysteinyl-PCP from structural comparison to the C domain of LgrA places the acceptor cysteine past the Cy-domain catalytic dyad [13,38], indicating rearrangements are necessary for cyclodehydration. At the acceptor site entrance, loop 1 (Fig. 3D) forms an extended surface where PCP and Ppant interactions may occur in the cyclodehydration complex. In PchE, a conserved glutamine within the latch is found to interact with the acceptor Ppant. In HMWP2-Cy2, this glutamine instead interacts with loop1, which forms part of a basin of attraction in which the Ppant can reside in a conformation specific for cyclodehydration [13]. Notably, the residues of this basin differ in PchE (Fig. 3E,F) compared to other systems, which may reflect different classes of Cy domains utilizing alternate interactions. Nonetheless, these observations suggest that Cy domains have two separate binding states for the acceptor Ppant: a condensation state that positions the loaded Ppant further into the active site (as in the C domain) in proximity to the donor substrate, and a cyclodehydration state that places the reactive functional groups of Cys/Thr/Ser sidechains near the catalytic dyad.

In comparison, tandem Cy domain architectures within (methyl)oxazoline systems may contain one Cy domain that lacks features that are important for facilitating cyclodehydration and a second Cy domain that lacks the appropriate interface for binding an acceptor CP [12,17]. The utility of tandem Cy domains was proposed to increase efficiency for cyclodehydration by serine/threonine in comparison to cysteine [33,39]; however, thiazoline systems containing single Cy domains have demonstrated the ability to generate oxazoline [40], and there are oxazoline-forming systems that contain Cy domains functional for both condensation and cyclodehydration (e.g. mycobactin) [30,37]. It seems likely that the Cy domains containing all of the prerequisite structural motifs and conserved amino acids can accept and cyclize Ser, Thr, or Cys, albeit with some selectivity for a particular donor substrate [41], and that overall specificity in turn comes from the coupled A domain. Future comparative studies will be necessary to better understand the evolutionary and molecular differences that favor tandem Cy domain architectures.

Global dynamics within Cy domains

Expanding upon static global conformations detailed above, recent solution studies captured Cy-domain global dynamics that appear necessary for domain communication. Here, dynamics is defined as structural fluctuations granting access to diverse local or global conformations, which can play a role in enzyme catalysis [42,43] or allosteric communication [44,45]. Global dynamics were observed in the Cy1 domain of HMWP2-Cy1 [14], with a footprint suggesting that the conformational changes of Figs. 2 and 3 reflect fluctuations in solution (Fig. 4A). Cy1 only engages with its substrate-loaded donor and not its holo form, thus avoiding unproductive interactions, as also reported for condensation domains [46,47]. Surprisingly, domain engagement generated an allosteric response reaching the acceptor site 40 Å away, suggesting communication between acceptor and donor sites (Fig. 4). To demonstrate a tie between structural dynamics and recognition of substrate-loaded CPs, the conserved dyad aspartate was mutated, which yielded a global response that impeded both the domain dynamics and carrier protein recognition. The results highlighted the importance of structural fluctuations in domain communication, in line with the interpretation of conformational changes discussed above. They also cautioned that, amidst dynamics, mutations may impart global responses confounding traditional interpretations using a single structure.

Figure 4. Cy structural dynamics modulate transient contacts for function.

A) The dynamic footprint of HMWP2-Cy1 [11]. Carrier proteins from PchE [15] are superimposed to mark donor and acceptor sites. Cy residues exhibiting relaxation dispersion in solution NMR experiments are shown as spheres. B) Substrate funneling at the donor site may be mediated by a network of dynamic residues. The pink arrow labeled “Mobile Lobe” indicates the direction the mobile lobe would shift to open the donor site entrance. The cyclodehydration intermediate (pink sticks) and the PchE product-bound state (light pink sticks) are shown. Yellow dotted lines are polar contacts, including interactions linking the acceptor site (α1, in gray) to the floor loop (orange) and subsequently α4, which may mediate allosteric effects. C) Proposed role of conformational fluctuations in the catalytic cycle. Gray icons resemble observed states shared by C and Cy domains, whereas the purple icon represents an anticipated Cy domain-specific state. Carrier proteins are salmon and yellow orange.

Examining structures of Cy domains in light of dynamics provides alternative interpretations of mutagenesis. The dyad aspartate is remote from both donor and acceptor CP binding sites [13,16,17]. How then can mutating this residue lead to a dramatic global response and affect donor recognition? Mishra et al. noted that, in some Cy conformations, the aspartate can interact with a semi-conserved serine/threonine (Fig. 4B, Latch Ser), as confirmed by the structures of PchE Cy [15] (Fig. 4B, Latch Thr). These residues immediately precede the latch glutamine, which was found to interact with the acceptor Ppant in PchE (Fig. 4B, Latch Gln) but displays different conformations in unliganded Cy domains. Within a dynamic scaffold begetting an ensemble of conformations, mutating a residue destabilizes specific conformations. Mutating the dyad aspartate would remove contacts with the latch serine and threonine residues, destabilizing conformers; the global dynamics are then altered and apparently impede access to conformations required for donor recognition. Because fluctuations sample a variety of conformers, these mechanisms necessarily supplement our understanding of Cy function. Here, the dyad aspartate can act both as a general base and in mediating allosteric networks.

Dynamics may help form fully engaged complexes featuring a substrate-tethered Ppant in the active-site tunnel. Brief encounter complexes are increasingly thought to precede and promote fully bound forms [48], and such mechanisms appear to be involved in Cy-donor recognition [14]. Dynamics of HMWP2-Cy1 in solution (Fig. 4A) reveals a network of residues that could couple the donor site entrance with tunnel residues known to interact with the Ppant arm (Fig. 4B). That the entire region surrounding the donor site entrance is dynamic suggests that fluctuations in the region can couple the conformational changes we described at the donor site (Fig. 3A,B), such that contacts between substrates and the gating phenylalanine redistribute conformations at the second aspartate of the DxxxxDxxS motif to help funnel the substrate into the tunnel. Highly conserved residues in α1 that interact with the DxxxxDxxS motif, substrate and the floor loop (as described in Fig. 3D) also display relaxation dispersion profiles.

Importantly, global dynamics are also inferred for the C domain from static structures and normal mode analysis demonstrating variable orientations of the N- and C-terminal subdomains [32], which similarly present an opportunity for CP binding to affect the state of the donor and acceptor entrances [14]. Only open donor entrances with the mobile lobe splayed away from the floor loop strands have been experimentally observed for C domains [33], and it would not be necessary for the C domain to reach a closed-for-cyclodehydration state. However, details of C-domain catalysis also remain unresolved, so global conformational changes—perhaps including mobile lobe rearrangement—could be involved in yet unknown ways such as product release trajectories.

Overall, consideration of structural dynamics may resolve open questions around Cy domains (Fig. 4C), and future research should account for dynamics when determining molecular mechanisms. Similarly, future bioengineering efforts should consider aspects of the Cy-domain conformational landscape, which appear intimately connected to domain communication and catalysis of peptide bond formation and cyclodehydration. Indeed, the interactions between A and C domains within a module have been shown by Townsend and colleagues to have dramatic effects on reactivity, with different A domain specificity observed in the presence and absence of the neighboring C domain [49]. These findings echo the global conformational responses of Cy domains towards carrier proteins binding and speak to how allostery is likely involved in regulating NRPS productivity.

Final thoughts

An understanding of the molecular determinants of Cy reaction mechanisms is beginning to take shape, although many questions remain. Excitingly, we are now at the point of building an understanding of allosteric communication between CP binding events and remodeling of the Cy-domain active site. Further studies will be needed to detail these features for improved bioengineering of Cy domain-containing systems.

Highlights.

• Heterocyclization domains catalyze both peptide bond formation and cyclodehydration in nonribosomal peptide synthetases

• Heterocyclization domains appear to change conformations for distinct catalytic steps

• Structural dynamics likely couple molecular communication with fold remodeling

• Conformational changes and structural dynamics should be considered in future bioengineering of heterocyclization domains