The Cyanobactin Heterocyclase Enzyme: A Processive Adenylase That Operates with a Defined Order of Reaction

Abstract

Rückwärts zählen: Die Cyanobactin-Klasse der Heterocyclasen, mit TruD als Beispiel, zeichnet sich durch eine nahezu einzigartige Kombination von Prozessivität, Spezifität, chemischer Vielseitigkeit und Promiskuität aus. Biochemische Assays zeigen, dass TruD eine Adenylase ist und Cysteine prozessiert. Selbst wenn die komplette Führungssequenz des Substrats entfernt wird, kann TruD einen einzelnen spezifischen Cysteinrest prozessieren; die Funktion der Führungssequenz besteht aber in der Aufrechterhaltung der Prozessivität durch Ausbalancieren der molekularen Erkennung. Heterocyclic rings are a recurring and iconic motif in organic chemistry. The introduction of five-membered rings into protein backbones is synthetically challenging, but many peptide-based biologically active natural products contain such Ser-, Thr-, and Cys-derived heterocycles1 and therefore a facile route to their introduction is highly desirable. More generally thiazolines and oxazolines, as well as their oxidized forms (thiazoles and oxazoles), are found in a variety of approved drugs, drug leads, and toxins.2 The cyanobacterium Prochloron spp. produces multiple macrocyclic cyanobactins, which are known as the trunkamides (seven or eight residues) and the patellamides (eight residues). These natural products originate from two different ribosomal precursor peptides (denoted TruE and PatE), and each ribosomal peptide contains multiple distinct core peptides. In TruE (and PatE), each of the core peptides is flanked at the N-terminus by a conserved five-residue protease signature3 and at the C-terminus by a conserved macrocyclization signature (A/SYDG; Scheme 1). It is the core peptides that go on to become the different natural products. In trunkamides and patellamides (as well as other cyanobactins) these products possess multiple heterocyclic cysteine, and/or serine and threonine amino acids. All cyanobactin ribosomal precursor peptides possess (along with the core peptides and their flanking regions) a conserved thirty- to forty-residue N-terminus (Scheme 1). This N-terminal leader, which is discarded during processing is thought to be essential for heterocyclization.4 Heterocyclization of the multiple cysteine residues in TruE is carried out by the single enzyme, TruD, which by definition is processive. The amino acid sequence of the core peptide is variable (Supporting Information, Figure S1) and thus TruD must be in part insensitive to the immediate sequence context of the target cysteine (Scheme 1). Further, the positions of the target cysteines in TruE relative to the leader vary, both within and between core peptides. This flexibility suggests that TruD could be valuable for synthetic chemists. The corresponding enzyme in the patellamide pathway is denoted PatD. TruD and PatD process cysteine residues4, 5 in the other enzymes’ precursor peptide substrate equally well; TruD processes serine and threonine residues less well than PatD.4, 5 The interchangeability of enzymes mirrors the high degree of sequence identity in the enzymes (Figure S2) and substrate N-terminal leader (Figure S1). Heterocyclization reaction catalyzed in the cyanobactins. TruD efficiently processes cysteines and selenocysteine.8 TruD operates on core peptides of seven or eight residues; a single precursor peptide can contain up to four core peptides each with different sequences. The immediate sequence context and position (relative to the leader) of the target cysteine is variable. The heterocyclization of cysteine to form thiazoline as part of the microcins (thiazole/oxazole-modified microcins, or TOMMs) has recently been studied and the responsible ATP Mg2+ dependent heterocyclase BalhD identified.6 BalhD is homologous to the C-terminal 400 residues of TruD, whilst a second protein BalhC matches the N-terminal 300 residues (Figure S2). When both BalhD and BalhC are present heterocyclization of the microcin substrate is accelerated and only when both proteins are present can phosphate release be robustly measured. Pyrophosphate production by BalhD was ruled out and BalhD was proposed to operate by a kinase type mechanism in which a hemiorthoamide intermediate is phosphorylated6, 7 (Figure S3). The cyanobactin class of heterocyclases, exemplified by TruD, possess an almost unique combination of processivity, specificity, chemical versatility, and promiscuity, which thus makes them a fascinating subject for study. We show by biochemical assay that TruD is an adenylase, not a kinase, and that TruD processes cysteines in a defined order, which we describe. We demonstrate that entire substrate leader can be removed and TruD will process a single specific cysteine residue. However, we establish the role of leader is to permit processivity through a balance of recognition. TruD overexpressed in E. coli BL21 (DE3) cells and purified to homogeneity (as judged by SDS-PAGE) using established procedures8 was used for all analyses. TruD (5 μM), when incubated with Mg2+, ATP (5 mM) and an engineered PatE2 (single core peptide sequence ITACITFC, Figure S1; 100 μM) results in a loss of mass in PatE2 of two water molecules (Figure S4a). Mass spectrometry (MS) fragmentation data confirms the formation of heterocycles,5, 8 validating TruD activity and substrate. 2.2 molar excess of ATP (1.1 for cysteines) was added to PatE2 at 20 °C and incubated with TruD for 8 h and immediately the product of the reaction was analyzed by 1D 31P NMR spectroscopy. Authentic AMP, ADP, ATP, pyrophosphate, and phosphate in the same buffer at the same concentration and pH were used as standards (Figure 1 a,b). Two new peaks matching AMP and pyrophosphate were identified in the product mixture. Unreacted ATP and a smaller amount of phosphate were observed, but no ADP. We repeated the experiment at 10 °C and recorded spectra every 30 min for the first 12 h and every 2 h thereafter (Figure S5a). This time-course confirms AMP and pyrophosphate production (and small amount of phosphate). A control incubation of the mixture with ADP reveals that it only slowly degrades to AMP. These data can only be explained by an adenylation mechanism and rules out a kinase mechanism. Repeating the NMR after overnight incubation or at a temperature >25 °C shows only phosphate (Figure S5b), a fact we attribute to pyrophosphate degradation rather than a time- or temperature-dependent change in mechanism. PatE2 TruD reaction. a) 31P NMR of a heterocyclization reaction, authentic AMP, pyrophosphate (PPi), ADP, phosphate (P), and ATP. AMP and PPi predominate at end of reaction with small amounts of phosphate and unreacted ATP present; no ADP is observed. b) Expansion of the PPi peak in the final reaction and PPi standard. c) Decrease of absorbance in the coupled NADH assay in the presence (blue) and absence (black) of myokinase (MK). Reaction rates were calculated from fitted curves (red and magenta, respectively). Addition of MK accelerates the apparent rate by a factor of about 8. ATP, ADP, and AMP bind to TruD with KD values of 19, 20, and 12 μM, respectively (Figure S5c), but as ADP decomposes slowly in a dissipated exothermic manner, the corresponding KD value is an apparent value obtained by analyzing the rapidly emerging endothermic parts of the ITC signals. Although dTTP does not bind in ITC experiments, we find it and GTP can catalyze heterocyclization of cysteine (Figure S5 d–f); BalhD has also been shown to use GTP.6 The coupled NADH assay reported in the study of the adenylase enzyme AcsD10 uses lactate dehydrogenase to oxidize NADH during the reduction of pyruvate. Pyruvate is generated by pyruvate kinase, which requires ADP. The assay has a background rate in the absence of any substrate,10 which for a slow enzyme such as TruD is significant. The addition of TruD and/or PatE2 to the mixture did not increase the rate of NADH consumption above background, indicating no enzymatic production of ADP. Addition of myokinase, which converts AMP and ATP into ADP, accelerates the rate of NADH consumption to 5.04 μM min−1 (Figure 1 c), confirming enzymatic production of AMP consistent with an adenylase mechanism. Time-course MS experiments showed that although both cysteines were heterocyclized after about 60 min at 37 °C, there appeared to be an accumulation of a mono-heterocyclized intermediate during the reaction (Figure S4b). Using standard triple-resonance experiments, we assigned the spectra (except four of the six histidine residues of the C-terminal tag) of uniformly 13C, 15N-labeled PatE2 (Figure S6). The sharp, poorly dispersed NMR signals and 13C chemical shifts indicate PatE adopts a natively unfolded state in aqueous buffer (compare with the helical arrangement of N-terminal leader peptide in organic solvent11). Real-time 1H-15N HSQC-NMR reaction monitoring using uniformly 15N-labeled PatE2 showed a specific order with the first heterocyclization event characterized by the disappearance of eight cross-peaks (assigned to residues 48–55) and the emergence of seven new cross-peaks (Figure 2). In the second heterocyclization event, eight cross-peaks (residues 43–50) disappeared (three of which appeared with first heterocyclization) and seven new peaks appeared (Figure 2). PatE2-TruD reaction monitoring and binding analysis. Overlay of HSQC spectra of 15N-PatE2 in presence of ATP/Mg2+ without TruD (black cross-peaks) and 8 h (left panel, blue cross-peaks) or 44 h (right panel, red cross-peaks) after addition of the enzyme. Resonances that disappear in the course of the transformation are labeled black; emerging resonances are labeled blue (8 h) and red (44 h). The cross-peak for C-terminal His64 undergoes a gradual chemical shift change in the course of the reaction (fast chemical exchange rather than the slow exchange observed for other affected resonances). Heterocyclization of cysteine induces large changes in 1HN, 15NH, Cα, and Cβ chemical shifts (Figure S6). This allowed us to establish that the terminal cysteine, C51, heterocyclizes first, followed by the slower heterocyclization of the internal cysteine C47 (Figure 2; Supporting Information, Figure S6). The reaction was highly temperature-dependent with completion in under 5 h at 20 °C, but still incomplete after 44 h at 10 °C (Figure 2). The internal cysteine of the PatE2 mutant C51P (chosen to mimic the mono-heterocyclized intermediate) was processed (within error) at the same rate as the internal cysteine in PatE2 (Figure S7). However, the internal cysteine of the C51A PatE2 mutant is processed very slowly (Figure S7). The conformational change at the C-terminus of core peptide introduced upon formation of the five-membered ring is thus critical for TruD to be able to process the internal cysteine of the core peptide. We incubated a uniformly 15N-labeled PatE with a core peptide with three cysteines (sequence ICACITFC (PatE3C)) and monitored the reaction by HSQC NMR at 10 °C and observed that C51 reacts first, followed by C47 and finally C45 (Figure S8). TruD thus works from the C-terminus within the core, but the accumulation of the intermediates shows that they are released and rebound, rather than held on the enzyme. The first two heterocyclizations of PatE3 occur on an identical timescale (at both 10 °C and 37 °C) to the two-cysteine PatE2 substrate, but the third heterocyclization is slower than the second heterocyclization (which is slower than the first). PatE2 binds very tightly to TruD (KD 80 nM) (Figure 3), but titrating TruD into 15N-PatE2 shows that 1H, 15N-HSQC cross-peaks of residues 1–15 are unaffected and thus do not bind to TruD (Figure S6e); they might therefore have no role in substrate recognition. ITC data obtained for the injection of PatE and PatE variants into TruD solutions. The top panels show raw data representing the response to injections, the bottom panels show integrated heats of injections (□) and the best fit (—) to the one-site model (origin). Using MS fragmentation analysis and ITC (where the substrate was sufficiently soluble), we sought to define the role of precursor peptide residues outside the core peptide in controlling substrate recognition. A synthetic peptide with the first 25 residues of PatE2 deleted (Δ25PatE) retains the most conserved leader region, residues 26-35 (denoted “minimal leader”), and was processed by TruD only slightly more slowly than PatE2 (judged by MS; Figure S9). Insertion of up to six amino acids between the N-terminal protease signature, and the core peptide permitted normal heterocyclization (Figure S7, Table S1). Four further synthetic peptides were tested: Δ37PatE (retains five-residue protease signature) Δ42PatE (no leader), the eight-residue core peptide itself and the eight-residue core peptide with an additional C-terminal Gly (Figure S9). No reaction is seen with core peptide alone or core peptide with C-terminal Gly. The terminal cysteine only is however processed in both Δ37PatE and Δ42PatE peptides (more slowly than Δ25PatE but within an order of magnitude). From these data we conclude that recognition of the C-terminal cysteine of the cassette does not require the leader. Radical mutants of individual residues within the center of the minimal leader region revealed S30F had no effect but mutants L29R and E31R did not allow processing of the PatE2 to completion. In both cases, we observed the mono-heterocyclized intermediate along with the product but no unreacted substrate, indicating that the rate of the second heterocyclization is significantly slower in these mutants but the rate of the first heterocyclization is largely unaffected. Both mutations significantly decreased binding affinity (Figure S7). Mutations G38I, L39N, A41I, and S42C within GLEAS protease signature had little or no effect on processing (first and second heterocyclization; E40R was insoluble in our hands); although the binding was decreased (Figure S7). The “new” cysteine of S42C was not processed by TruD. Y53A and D54R (in the C-terminal macrocyclization signature) were processed normally. Processivity (heterocyclization of the second or internal cysteine of the core peptide), although very sensitive to changes in the minimal leader, is insensitive to changes in the core peptide flanking regions. We explored more radical mutants S42Q and A52D, which immediately flank the core peptide; both mutants significantly disrupted binding and processing. In contrast to changes in the leader peptide however, these mutations primarily produced either fully heterocyclized product or unreacted substrate. This suggests that these mutations affect the rate of the first heterocyclization (C-terminal cysteine), which has now become rate-limiting, but the mutations do not significantly perturb heterocyclization of the internal cysteine (the processivity). The PatE2 A52P mutant has the internal cysteine processed first and only very slowly has the terminal cysteine processed. ITC shows that PatE2 C51A binds almost as tightly as PatE2, but as the internal cysteine is processed extremely slowly, we conclude this binding is non-productive; the Ala at position 51 is bound at the active site. The PatE2 C51P mutant, which mimics the mono-heterocyclized intermediate and permits normal processing of the internal cysteine, binds much more weakly (1500 nM). We suggest this weaker binding allows the internal cysteine to access the active site. The structure of TruD was determined to 2.9 Å (PDB 4BS9) and can be decomposed into three domains (domain 1 residues 2–85, domain 2 residues 86–321, domain 3 residues 323–781), which combine to form an extended, curved molecule that exists as a dimer (Figure 4 a). Structure of TruD. a) Structure of the biological TruD dimer. Domain 1 (blue), domain 2 (yellow), domain 3 (red), and the covalently bound Zn (gray sphere in domain 2, near domain 3) are shown. The dimer partner is represented in corresponding faded colors. b) Structural alignment of TruD (domain 1 blue, domain 2 yellow) with MccB (magenta, PDB 3H5A). The MccB ATP binding site is boxed. Domains 1 and 2 of TruD share 21 % sequence identity and structural fold with the enzyme MccB, an adenylase found in the microcin C7 antibiotic production pathway (MccB catalyzes two successive adenylation reactions; Figure S10).12 Pairwise of TruD with the MccB structure a of Å residues (MccB the third Figure 4 Supporting Information, Figure individual domain are much to a domain Å 80 residues and Å residues). TruD and MccB possess a arrangement (Figure 4 a) and both share two in domain 2 that bind a in a arrangement of Both are conserved in cyanobactin and these (and are a conserved of class of The a structural role in this class of Domain 2 of TruD the fold of this class of other enzyme 5 Å and pyrophosphate enzyme Å however, a all the enzymes that share a fold with TruD have an between the first and second of the relative to TruD domain this that the ATP binding site of MccB is not fully conserved in TruD (Figure The structure of MccB (PDB suggest and of TruD as ATP binding activity and the three other TruD mutants were in TruD is conserved both in and in the adenylase but the TruD mutant Structural data of course do not establish the mechanism but they are consistent with TruD an adenylase enzyme. The 20 % sequence identity of BalhC to TruD the two Figure that BalhC will have the same fold as domains 1 and 2 of TruD and thus fold will to the same class of Domain 3 of TruD is a fold and has a large (Figure In our the third domain of TruD or is but by of PatD homologous to we were able to part of the third domain (residues In contrast to which is the no or substrate binding activity (Figure with % identity will have the same fold as TruD domain 3 and thus an new fold for kinase chemistry. analysis of TruD rules out a kinase mechanism for this enzyme as there is no ADP both AMP and pyrophosphate are during the data can only be as an mechanism. We did phosphate than at temperature or after but this is consistent with degradation of the suggest the of is a to with ATP. In any event, such a mechanism to ATP consumption of followed by which is not observed. The most mechanism therefore requires the formation of a five-membered hemiorthoamide ring in a (Scheme proposed by in the BalhD mechanism. The for the first could from PatE but the that the entire leader region can be and the conserved flanking residues We suggest the hemiorthoamide intermediate the of ATP pyrophosphate an This highly intermediate decomposes to thiazoline and by of AMP (Scheme 2). mechanism for heterocyclization by the TruD class of enzyme. of the to the formation of a which is of AMP to the formation of the The in mechanism of adenylation between BalhD and TruD is their sequence and by structural and their chemical reaction. We that BalhD and the third domain of TruD both catalyze the formation of the hemiorthoamide In TruD, domain 2 the intermediate whilst in the BalhC to BalhD and accelerates the These data suggest that a very small of enzymes exemplified by the type and type which catalyze the same reaction, share a fold operate by different chemical TruD and have to process multiple cysteines a precursor peptide. NMR showed TruD operates in order with the terminal cysteine in the core peptide the most C-terminal cysteine and on (Scheme 3), thus a of The binds tightly nM) whilst the mono-heterocyclized intermediate by PatE2 binds more weakly (KD nM). the intermediate only binds all the substrate is processed. The for a five-membered ring at the end of the core peptide for processing mirrors the for the but for the In the the ring is essential for binding of the substrate, but for the heterocyclase it is essential to substrate binding at an processed a form of chemical (Scheme residues for and order of a) The precursor peptide a single core peptide. essential for the formation of internal are in the red residues with the formation of the first are in are b) The order of heterocyclization reaction. substrate binds tightly to the enzyme the first heterocyclization; recognition of substrate does not on leader. of the first binding affinity successive heterocyclization within the core peptide from the recognition now on leader (red of the terminal core peptide cysteine to to binding in a for The entire leader peptide could be with and the terminal cysteine still processed has been reported in the This suggests recognition for the terminal cysteine on immediate consistent with our analysis (Scheme is the role of the leader We have shown that the leader sequence residues is for processivity (heterocyclization of We a model in which substrate recognition both the target and the leader, the of each of these to substrate binding shifts during processing and it is this binding that the order of multiple core peptides share a single leader, the flanking the core peptide can be and of residues between the core and leader can be we conclude that the between the minimal leader and target is highly This of recognition and is and the use of the enzyme in chemical a to our and this by the are and be for but are not or from than be to the The is not responsible for the or of any by the than be to the corresponding for the