Use of a Scaffold Peptide in the Biosynthesis of Amino Acid Derived Natural Products

Abstract

Genome sequencing of environmental bacteria allows identification of biosynthetic gene clusters encoding unusual combinations of enzymes that produce unknown natural products. We identified a pathway in which a ribosomally synthesized small peptide serves as a scaffold for non-ribosomal peptide extension and chemical modification. Amino acids are transferred to the C-terminus of the peptide through ATP and amino acyl-tRNA-dependent chemistry that is independent of the ribosome. Oxidative rearrangement, carboxymethylation, and proteolysis of a terminal cysteine yields an amino acid derived small-molecule. Microcrystal electron diffraction demonstrates that the resulting product is isosteric to glutamate. We show that a similar peptide extension is used during the biosynthesis of the ammosamides, cytotoxic pyrroloquinoline alkaloids. These results suggest an alternative paradigm for biosynthesis of amino acid derived natural products.

One Sentence Summary:

Amino acids are added to the C-terminus of gene-encoded scaffold peptides, followed by maturation and proteolytic release.

Bacteria produce many small-molecule natural products that play important roles in communication, symbiosis, and competition (1). Historically these compounds have been discovered by activity-based screens, but an alternative avenue for their discovery starts with identification of their biosynthetic gene clusters now that bacterial genomes have revealed the tremendous diversity of natural products that remain to be discovered (2). In this study we focus on a group of biosynthetic gene clusters for which the final products were not known and could not be predicted.

Ribosomally-synthesized and posttranslationally-modified peptides (RiPPs) (3) include lantibiotics and thiopeptides that are used in food and agriculture (4). They are biosynthesized from a precursor peptide consisting of a leader peptide that serves as a recognition motif for the biosynthetic enzymes and a core peptide that is converted to the final product. During their maturation, Ser and Thr residues are glutamylated by LanB enzymes in a glutamyl-tRNA dependent mechanism (5, 6). Subsequently, the glutamate is eliminated to generate dehydroamino acids (Fig. 1A). A survey of >100,000 publicly available bacterial genomes revealed more than 600 genes encoding LanB-like proteins in which the elimination domain is not present within the cluster or genome.

Figure 1. Function of a small LanB enzyme, TglB, found in P. syringae.

(A) LanB enzymes glutamylate Ser/Thr residues and subsequently eliminate the glutamate to form dehydroamino acids. Small LanB proteins lack the elimination domain. Dha, dehydroalanine; Dhb, dehydrobutyrine. (B) Biosynthetic gene cluster in P. syringae encoding a small LanB. (C) Matrix-assisted laser desorption ionization with time-of-flight (MALDI-TOF) mass spectra of TglA coexpressed with TglB. (D) Analysis of the TglB product by tandem electrospray ionization (ESI) mass spectrometry.

In the plant pathogen Pseudomonas syringae pv. maculicola ES4326, such a protein (TglB) is encoded near an open reading frame encoding a 50-amino acid peptide (TglA; Fig. 1B). Co-expression of His₆-TglA and TglB in Escherichia coli and subsequent purification of the peptide demonstrated an increase in mass by 103 Da (Fig. 1C). This increase is inconsistent with glutamylation, but could be the result of condensation with a cysteine residue. High resolution tandem mass spectrometry analysis of the peptide suggested that the adduct was attached to the C-terminal alanine instead of the anticipated ester linkage to a Ser in the peptide (Fig. 1D). We expressed TglA and TglB individually as His₆-tagged proteins and purified them. In vitro incubation with Cys, ATP, tRNA^Cys and Cys tRNA synthetase (CysRS) resulted in the same product (TglA-Cys; Fig. 2A) as that isolated from co-expression in E. coli confirming that TglB adds a Cys to the C-terminus of TglA in a tRNA dependent manner (fig. S1A). This C-terminal peptide extension not only constitutes a previously unknown posttranslational modification but also seems counterintuitive since a more logical route to the product appears to be encoding the Cys on tglA. We next purified Cys-tRNA^Cys and showed that TglB does not transfer the Cys to the C-terminus of TglA unless ATP is present, which is converted to ADP and phosphate (fig. S1B). Performing the reaction in buffer made with H₂ ¹⁸O and subsequent MS analysis demonstrated that the product contains one ¹⁸O atom (fig. S1C), and addition of hydroxyl amine to the assay allowed trapping of C-terminally activated TglA as the hydroxamate (fig. S1D). These findings are consistent with activation of the C-terminus of TglA by phosphorylation, subsequent amide bond formation with the amino group of Cys-tRNA, and release of the tRNA by hydrolysis (fig. S1E). The observations rule out the use of the activated ester of Cys-tRNA for the non-ribosomal peptide extension (fig. S1E). TglB accepted a 12-mer peptide corresponding to the C-terminus of TglA as a minimal substrate (fig. S1F), and kinetic experiments showed that TglB has a turnover number of 28 min⁻¹ using full length TglA (fig. S1G).

Figure 2. The cysteine added by TglB is modified by other enzymes from the tgl cluster.

(A) Inferred biosynthetic pathway towards 3–thiaglutamate. (B) MALDI-TOF mass spectrum of in vitro reaction of TglHI with TglA-Cys. (C) MALDI-TOF mass spectrum of in vitro reaction of TglF with compound 1. Color-coding of shaded peaks in panels B and C are shown in panel A.

We next interrogated the other proteins encoded in the biosynthetic gene cluster. TglH has low homology to a structurally characterized dinuclear non-heme iron dependent protein for which no activity has been reported (7). The C-terminal domain of TglI has homology with known leader peptide binding domains in RiPP biosynthetic enzymes (Fig. 1B) (5, 8). We co-expressed TglA with TglB, TglH and TglI in E. coli and isolated a product that was decreased in mass by 14 Da from TglA-Cys (fig. S2A). We treated the peptide with trypsin to generate a C-terminal tetrapeptide. Chemical assays with thiol- and carboxylate-reactive electrophiles indicated that the product still contained these functional groups (fig. S3) suggesting structure 1 as the product of TglHI (Fig. 2A). We next repeated this experiment but using an E. coli strain that is auxotrophic for Cys and that was grown in minimal media supplemented with ¹³C-labeled Cys. Isolation of the peptide and analysis by MS showed that it is the cysteine β-carbon that is removed (fig. S4).

The biosynthetic cluster also contains a pair of genes (tglEF) encoding proteins similar to a recently characterized carboxy-S-adenosylmethionine (Cx-SAM) synthase and a SAM-dependent methyltransferase, respectively (9, 10). We added compound 1 to Cx-SAM and TglF in vitro and isolated product 2 with a mass increase of 58 Da (Fig. 2C), consistent with carboxymethylation of a thiol. This hypothesis was confirmed by treating the TglHI product with iodoacetic acid, which resulted in the same outcome, as did co-expression of TglABEFHI in E. coli (fig. S2B). The in vitro prepared peptide was treated with trypsin and the C-terminal tetrapeptide 3 was characterized by ¹H NMR spectroscopy and tandem MS, which supported structure 2 for the TglF product (Fig. 2A; fig. S5). Given the unusual architecture, we also chemically synthesized peptide 3 as two diastereomers (Supplementary Information) and demonstrated that the ¹H NMR spectrum of one isomer was identical to the enzymatic product (fig. S5). We tried to obtain crystals to assign the stereochemistry of either isomer and made several chemical derivatives but were unable to obtain crystals for X-ray diffraction.

We next turned to the cryoEM method microcrystal electron diffraction (MicroED) (11–13). A small amount of powder of the diastereomer that was obtained in higher amounts and in more pure form was placed onto an EM grid, plunged into liquid nitrogen and investigated under cryogenic conditions in an electron microscope. The seemingly amorphous powder contained numerous nanocrystals on the grid suitable for MicroED analysis, each consisting roughly of femtograms of material that diffracted to ~1 Å resolution. MicroED data was collected from each nanocrystal but the sample was highly susceptible to beam damage such that no useful diffraction was observed after the first few frames of the MicroED movie. Despite >150 data sets collected on a CMOS-based CetaD camera, nanocrystals succumbed to radiation damage too fast preventing structure determination. It is possible that the peptide was particularly susceptible to damage because of the 3-thiaglutamate, consistent with an earlier study that showed that radiation damage is particularly prevalent at Cys residues (14). We then turned to the Falcon III direct electron detector, one of the most sensitive cameras for cryoEM that was recently demonstrated to be suitable for MicroED data collection and structure determination and that minimizes radiation damage because of its high sensitivity and high frame rate (15). Atomic resolution data from seven nanocrystals were collected each covering an angular range of ~50° before damage was observed. Data from five nanocrystals were merged to yield a 96% complete data set to 1.0 Å resolution and the structure was determined by direct methods (Fig. 3; Crystallographic Table S3; Supplementary Information). The atomic-resolution MicroED structure was thus solved demonstrating the D configuration of the 3-thiaGlu in this peptide (D-3), which in turn provided the stereochemical assignment for L-3 which co-elutes with and has the same spectral data as the enzymatic product. These results demonstrate that the TglHI-catalyzed reaction occurred with retention of configuration at the α-carbon (Fig. 3B–D). These findings highlight the utility of MicroED to determine the structure and stereochemistry of a previously unknown natural product. Thus, collectively TglBEFHI convert TglA into a peptide containing L-3-thiaglutamate at its C-terminus (TglA-thiaGlu, 2; Fig. 2A).

Figure 3. In vitro TglHI reacts with ¹³C-labeled TglA-Cys to produce ¹³C-formate and 1 with retention of configuration.

(A) Diffraction pattern of D-3 with resolution ring at 0.9 Å. (B) Atomic MicroED structure of D-3 determined at 1.0 Å resolution. X refers to 3-thiaGlu. (C) Structure of chemically synthesized tetrapeptides (VFAX) containing D-thiaGlu (D-3) and L-thiaGlu (L-3). (D) Determination of stereochemical configuration of thiaGlu by comparison with synthetic standards. High-performance liquid chromatograms are shown. VFAX_enz was obtained by TglHI modification of TglA-Cys followed by 2-iodoacetic acid alkylation and trypsin digest. (E) ¹³C NMR spectra showing the β-carbon of the C-terminal cysteine in ¹³C-labeled TglA-Cys (26.3 ppm, top), and a new signal at 171.0 ppm that corresponds to ¹³C-formate after reaction with TglHI (bottom).

We next investigated the TglHI-catalyzed reaction with purified proteins. Neither protein could be expressed in soluble form individually, but co-expression resulted in co-purification and metal analysis indicated TglHI contained 2.5 Fe. In vitro TglHI converted TglA-Cys to 1 under aerobic conditions with a turnover number of 1.1 min⁻¹ (Fig. 2B) whereas under low oxygen concentrations product formation was negligible, confirming oxygen-dependency of the reaction (fig. S6A). To investigate if TglHI can functionalize internal cysteine residues, the extension mutant TglA-CysAla was prepared. This peptide was not modified by TglHI (fig. S6B). TglHI also did not modify other unrelated peptides that end in Cys (fig. S6C), and N-terminal truncation of TglA-Cys led to diminished or abolished TglHI activity (fig. S6D). Thus, the enzyme has high specificity for TglA-Cys. To identify the fate of the lost carbon atom, ¹³C-labeled TglA-Cys was reacted with TglHI and formate was observed by ¹³C NMR spectroscopy (Fig. 3E). Moreover, when [2,3,3,−²H]-Cys was used, the product contained one deuterium illustrating that the α-hydrogen is likely not removed during the transformation (fig. S4D). Thus, TglHI catalyzes a net four-electron oxidation of TglA-Cys, modifying the redox states of both the α and β carbons of the C-terminal cysteine installed by TglB. Based on the in vitro studies, we propose a mechanism for the formation of 1 and formate from TglA-Cys (fig. S7). The chemistry catalyzed by TglHI expands the range of post-translational modifications in natural product biosynthesis (16) to include a remarkable excision of a methylene group from cysteine. Additional TglHI-like enzymes are present in the genomes (fig. S8) including in the biosynthetic gene cluster encoding the methanobactin precursor (17–19).

The last four genes in the biosynthetic cluster encode a putative membrane bound protease (TglG), a putative pyridoxal-phosphate dependent enzyme (TglC) that is sometimes missing in homologous clusters, and two putative transporters (TglD and TglJ). Like TglB and TglI, TglG contains a RiPP leader peptide recognition motif suggesting it will act on a TglA-derived peptide (Fig. 1B) and homologous enzymes have cytoplasmic active sites (20). When TglA-thiaGlu was exposed to the membrane fraction of cell lysate of E. coli expressing GFP-TglG, the peptide was cleanly converted into TglA (fig. S9). TglA-Glu was also a substrate but not TglA-GluAla, illustrating that the protease cannot distinguish Glu and 3-thiaGlu but does not tolerate extension of the peptide. Thus, TglA appears to be a scaffold on which 3-thiaGlu is assembled and final proteolytic release regenerates TglA for another round of biosynthesis (Fig. 2A). Were cysteine merely encoded in tglA, then each ribosomally produced peptide could make only a single 3-thiaglutamate. Instead, the use of TglA as a scaffold peptide is conceptually more efficient than the stoichiometric use of leader peptide in other RiPP pathways (4). At present we do not know the function of 3-thiaGlu, nor whether this unstable compound is further chemically modified. Plants were recently shown to use Glu for a systemic signaling response to pathogens (21), and it is possible that 3-thiaGlu or a product derived from it interferes with Glu signaling similarly to other anti-metabolite toxins made by P. syringae that block jasmonate and ethylene signaling pathways (22).

We note that 3-thiaGlu is not a RiPP because it is not ribosomally synthesized, but it is made by posttranslational modification reactions. Perhaps this unusual pathway evolved because of the significant relative burden of leader peptide production for a single amino acid product. Bioinformatic prediction of TglA transcriptional regulation (23) suggests precursor production is not driven by a separate promoter, which is consistent with putative catalytic use of the peptide (fig. S10). This contrasts with most RiPP pathways in which expression of the substrate peptide is controlled by its own promoter followed by a read-through transcriptional terminator to allow the precursor peptide to be present in excess over the biosynthetic machinery (24, 25).

It is the Cys-tRNA-dependent enzyme TglB that allows the proposed catalytic use of TglA. Similar small LanB-encoding genes are found in several bacterial phyla, with some clusters encoding multiple such proteins and a range of additional putative modification enzymes (fig. S11). To assess the generality of the function of small LanB proteins and provide further support for a catalytic role of the scaffold peptide, we investigated ammosamide biosynthesis. A previous study of these Trp-derived pyrroloquinoline natural products (Fig. 4A) hinted that the compounds could be derived from a small peptide AmmA ending in Trp encoded in the gene cluster (Fig. 4B) (26). However, when this Trp was mutated to Ser or deleted altogether, ammosamide was still produced (26). The ammosamide gene cluster encodes four small LanB proteins. We tested all four for activity in vitro and in E. coli with AmmA (previously annotated Amm6) and AmmA lacking the C-terminal Trp but observed no activity. We noted that AmmA has homology with other peptides encoded in clusters with small LanB proteins (Fig. 4B), but that AmmA appears to have a C-terminal extension. When we removed this extension, AmmB2 (previously annotated Amm9), but not the other three AmmB proteins, added a Trp in a Trp-tRNA dependent fashion to the C-terminus of the peptide in vitro and in E. coli (Fig. 4C). This finding explains the observation that mutation or deletion of the C-terminal Trp still resulted in ammosamide production, and supports catalytic use of the peptide. Such use provides an attractive explanation for the 134 mg/L of ammosamide C produced by the producing bacterium (26) because stoichiometric use would require production of 3.0 g of AmmA. Given this second example of tRNA-dependent activity, we suggest the name peptide-amino acyl tRNA ligase (PEARL) for the small LanB proteins. The biosynthesis of a metabolite on a small peptide scaffold is uncommon, with the closest similarity found in the biosynthesis of amino acids linked by isopeptide bonds to a glutamate residue on amino-carrier proteins in some bacteria (27, 28).

Figure 4. Ammosamide biosynthesis involves addition of L-Trp to the C-terminus of a ribosomally synthesized peptide.

(A) Pyrolloquinoline alkaloids ammosamides A-C. (B) Sequence alignment of the C-terminus of the AmmA precursor peptide and its homologs showing a C-terminal extension for AmmA relative to most homologs. The gene cluster for ammosamide biosynthesis in Streptomyces sp. CNR698 comprises 27 orfs. The encoded proteins include four small LanBs, two proteases, one halogenase and a transporter. (C) AmmB2 adds L-Trp to AmmA*, a truncated peptide of AmmA, to afford AmmA*W in vitro in an ATP, tRNA^Trp and Trp-RS dependent reaction. Red MALDI-TOF mass spectrum is AmmA* and the blue spectrum shows the product of the reaction. HR-ESI MS/MS confirms addition of L-Trp to the C-terminus.