A Widely Distributed Biosynthetic Cassette Is Responsible for Diverse Plant Side Chain Cross‐Linked Cyclopeptides

Abstract

Cyclopeptide alkaloids are an abundant class of plant cyclopeptides with over 200 analogs described and bioactivities ranging from analgesic to antiviral. While these natural products have been known for decades, their biosynthetic basis remains unclear. Using a transcriptome‐mining approach, we link the cyclopeptide alkaloids from Ceanothus americanus to dedicated RiPP precursor peptides and identify new, widely distributed split BURP peptide cyclase containing gene clusters. Guided by our bioinformatic analysis, we identify and isolate new cyclopeptides from Coffea arabica, which we named arabipeptins. Reconstitution of the enzyme activity for the BURP found in the biosynthesis of arabipeptin A validates the activity of the newly discovered split BURP peptide cyclases. These results expand our understanding of the biosynthetic pathways responsible for diverse cyclic plant peptides and suggest that these side chain cross‐link modifications are widely distributed in eudicots.

Cyclopeptide alkaloids are a diverse class of plants natural products, yet little is known of their biosynthesis. By combining transcriptomics and genome mining, we uncover a widespread biosynthetic cassette responsible for side chain cross‐linked cyclopeptides. We validate these observations by isolating a new cyclopeptide from Coffea arabica and reconstituting enzymatic activity for the family defining split BURP peptide cyclase.

Plants are prolific producers of cyclic, peptide derived natural products including cyclotides, ^[1] orbitides, ^[2] and cyclopeptide alkaloids.[ ² , ³ ] Detailed biosynthetic studies have demonstrated that many of these compounds are ribosomally synthesized and post‐translationally modified peptide (RiPP) natural products.[ ⁴ , ⁵ , ⁶ , ⁷ , ⁸ ] RiPPs do not have a family defining biosynthetic step or feature. Instead, they are classified by their route of biosynthesis wherein a ribosomally produced precursor peptide is modified by at least one tailoring enzyme (Figure 1A). ^[9] The majority of RiPP precursor peptides are composed of a conserved leader sequence that is responsible for recognition by the tailoring enzymes and a core peptide that is modified. In the final biosynthetic step, a peptidase will typically release the core from the precursor peptide to generate the mature natural product. Because RiPPs are a large family of natural products with over 40 distinct classes, RiPP pathways have also been shown to deviate from this general route and include precursor peptides that house an array of multiple core sequences or are fused to autocatalytic tailoring enzymes (Figure 1A).[ ⁵ , ⁶ , ⁷ , ¹⁰ ]

Figure 1.

A) Overview of RiPP biosynthetic routes. Catalyzed amino acid modifications are indicated with colored shapes. Recognition sequence, RS. B), C) Plant side chain cross‐linked cyclopeptides.

Cyclopeptide alkaloids are among the most abundant classes of plant‐produced peptides with over 200 members (Figure 1B).[ ² , ³ ] They are four or five amino acids in size and are proposed to be formed from an ether linkage between the phenolic oxygen of tyrosine and the β‐carbon of a nearby amino acid. The tyrosine also undergoes a decarboxylation and desaturation to generate the characteristic styrylamine moiety. Additional modifications including methylation and hydroxylation are also common. Cyclopeptide alkaloids have a wide range of bioactivities including antiviral, ^[11] sedative, ^[12] and analgesic. ^[13] Even though they were first identified over 50 years ago, little is known about their biosynthesis. Structurally, cyclopeptide alkaloids are related to diverse plant side chain cross‐linked cyclopeptides that include lyciumins, ^[5] selanines, ^[6] moroidins, ^[7] and hibispeptins[ ¹⁴ , ¹⁵ ] (Figure 1C). Recent work demonstrated that many of these cyclic peptides are derived from autocatalytic BURP peptide cyclases that install the characteristic amino acid side chain cross‐links.[ ⁵ , ⁶ , ⁷ , ¹⁶ , ¹⁷ ] In these systems, the core peptides are fused to the copper‐dependent BURP‐domains. While the structural similarities between these plant cyclopeptides hint at a conserved biosynthetic route for cyclopeptide alkaloids, no definitive link has yet been made.

To discover the biosynthetic basis for cyclopeptide alkaloids, we examined Ceanothus americanus L. (Rhamnaceae), which is commonly known as New Jersey tea. This plant is a well‐known producer of cyclopeptide alkaloids with at least ten different molecules isolated, such as frangulanine (Figure 1B).[ ¹⁸ , ¹⁹ , ²⁰ ] Study of C. americanus would not only give insight into the biosynthesis of bioactive cyclopeptide alkaloids, but would also explain how a single plant is able to make so many diverse analogs.

In order to link the known cyclopeptide alkaloids to their precursor peptides, we generated transcriptomic data sets from the leaf, stem, and root of a single C. americanus individual. We first searched our translated transcriptomic data sets for BURP‐domain containing transcripts using an available Hidden Markov Model (HMM) for the protein family (PF03181) to identify possible autocatalytic precursors (Supporting Information Table). Surprisingly, there were no clear correlations between the amino acid sequences in the BURPs and the predicted core sequences of the cyclopeptide alkaloids that we confirmed to be present in the roots of our C. americanus samples by LC‐MS (Figures S1–S8).

As the observed cyclopeptide alkaloids did not appear to be derived from autocatalytic BURPs, we queried the entire translated transcriptome for all four‐amino acid sequences that could lead to the observed natural products (Figure S1). Clustering these candidates by sequence identity revealed related transcripts that corresponded to the observed cyclopeptide alkaloids (Figures 2A and S9). Each of these precursors contained an N‐terminal signal sequence and a conserved leader peptide‐like region (Figure S10). This is followed by a repeating recognition sequence and core structure that has previously been observed in cyanobactins ^[10] and numerous cyclic peptides from plants.[ ⁵ , ⁶ ] The core sequences themselves are typically demarcated by a conserved N‐terminal Asn and a C‐terminal His (Figures S9 and S11). To confirm our hypothesis that these precursor peptides are directly linked to the observed cyclopeptide alkaloids, we identified E‐L‐W‐Y and F‐F‐F‐Y cores for which no known cyclopeptide alkaloids correspond (Figure S9). We searched the feature list obtained from LC‐MS analysis of a C. americanus root extract for ions consistent with these new cyclopeptides and found features with the expected m/z and MS/MS fragmentation patterns (Figures S12 and S13).

Figure 2.

Stand‐alone precursor peptides from C. americanus (A) and Z. jujuba (B) contain multiple cores that directly map to known cyclopeptide alkaloids (red, bold). B) Precursor peptides are co‐clustered with BURP peptide cyclases in the genome of Z. jujuba. Predicted amino acid sequence is shown below the molecule name.

After linking C. americanus cyclopeptide alkaloids to dedicated precursor peptides, we sought to evaluate if these stand‐alone precursor peptides correspond to other known plant side chain cross‐linked cyclopeptides. We used BLAST to query the genome of the prolific cyclopeptide alkaloid producer Ziziphus jujuba for homologs of the C. americanus precursor peptides. The resulting hits were compared to the anticipated core sequences of the known Z. jujuba cyclopeptide alkaloids and 15 molecules could be directly linked to a core sequence including jubanines F, G, H, I, J, and nummularine B (Figures 2B and S14). ^[11] As with C. americanus, most precursor peptides contained an N‐terminal signal sequence followed by a conserved leader‐like region and repeating recognition sequence+core architecture (Figures 2B, S11, and S15).

Examination of the Z. jujuba genome revealed that 30 of the 35 annotated precursor peptides were co‐clustered with a BURP‐domain containing protein. Moreover, there were genomic loci containing multiple precursor peptides and BURPs arrayed together (Figure 2B). When aligned to the autocatalytic BURPs, the split BURPs are shorter and lack the N‐terminal extension that contains the core peptide(s) (Figure S16). The proximity of the split BURPs to the precursors in the genomes along with chemistry consistent with a post‐translational modification known to be catalyzed by BURP‐domains indicates the presence of a new split BURP biosynthetic system that parallels the fused autocatalytic BURPs.

To explore the distribution of this split BURP system, we built a custom Hidden Markov Model (HMM) using PSI‐BLAST results from the C. americanus precursor peptide (Supporting Information Table) and analyzed all the Viridiplantae sequences available from the NCBI identical protein group database. Unexpectedly, a significant portion of the resulting sequences were fused to a BURP‐domain. The detection of BURP proteins is likely due to the presence of a highly conserved Ax₆YWx₇PMP motif that follows the signal peptide in many autocatalytic BURP‐domains and stand‐alone precursor peptides (Figure S17). We used an existing BURP‐domain HMM (PF03181) to differentiate between these fused and split biosynthetic systems. This ultimately allowed for the identification of 1423 stand‐alone precursor peptides and 1099 fused BURP‐domain systems. We analyzed the genomic context of the stand‐alone precursor peptides to determine how often they co‐cluster with BURP‐domains. At least 58 % of the precursor peptides were found within ten annotated genes of a BURP (Supporting Information Table). This overall high level of co‐occurrence suggests that the precursor peptide and BURP pairs form a conserved biosynthetic cassette.

In addition to co‐occurrence, precursor peptides appear to be transcriptionally co‐regulated with a split BURP. Published work from Z. jujuba showed that sanjoinine A content was most closely associated with transcript levels of a genomic locus containing both a split BURP and methyltransferase. ^[21] We also analyzed the ATTED‐II database ^[22] for genes co‐expressed with putative precursor peptides from Arabidopsis thaliana, Vitis vinifera, Populus trichocarpa, Glycine max, Medicago truncatula, and Solanum lycopersicum. In each case, split BURPs are amongst the three most co‐regulated genes (Supporting Information Table).

The precursor peptides for the split systems were further explored by generating sequence similarity networks (SSNs) ^[23] (Figures 3 and S18). Putative core sequences were manually annotated using the observed recognition sequence elements from C. americanus and Z. jujuba (Figure S11 and Supporting Information Table). Throughout the entire sequence similarity network, a C‐terminal tyrosine is a highly conserved feature of the core. The largest cluster is characterized by putative four‐membered core sequences with a Ser or Thr in the second position and an almost completely conserved Tyr (Figure S11). Even though examination of the genomic context reveals a 59 % co‐localization rate with BURP proteins (Supporting Information Table), no small molecules have been isolated from plants that match this motif to our knowledge. ^[2] The vast majority of these precursor peptides belong to the organ specific protein family (PF10950). ^[24] Members of this family contain an N‐terminal signal peptide followed by a series of features exhibiting RiPP precursor peptide characteristics, including a conserved leader peptide‐like motif and alternating hypothetical recognition sequence and core peptide repeats containing a completely conserved tyrosine. ^[25] While the exact function of organ specific proteins remains unclear, they are hypothesized to play a role in abiotic stress response, root development, nodule formation, and establishment of mycorrhizal interactions.[ ²⁴ , ²⁵ , ²⁶ , ²⁷ ] Whether these proteins are processed to secondary metabolites or simply post‐translationally modified requires further study.

Figure 3.

Sequence similarity network of stand‐alone precursor peptides using an alignment score of 70. Singlets are shown in Figure S18. Precursor peptides with cores containing serine or threonine in the second position are represented in navy blue. Clusters from H. syriacus (pink), Z. jujuba (light blue), Coffea (red), and C. americanus (purple) are indicated along with those with lyciumin‐like (green) and moroidin‐like (brown) core sequences. Amino acid probabilities of the core sequences for these clusters are depicted.

In contrast to the prevalence of Ser/Thr containing cores without a clear link to a known natural product or peptide modification, other clusters could be mapped to known plant side chain cross‐linked cyclopeptide natural products. For example, the cyclopeptide alkaloid precursor peptides from C. americanus and Z. jujuba formed discrete clusters (Figure 3).

In addition to clusters from the known cyclopeptide alkaloid producers, a large cluster of putative precursor peptides from Hibiscus syriacus was also present in the SSN (Figure 3). Precursor peptides from this cluster were rich in the core sequences Q‐I‐P‐L‐F‐Y and Q‐I‐P‐L‐L‐Y, which match the known side chain cross‐linked cyclopeptides hibispeptin A and B, respectively (Figures 1C, 3, and S19).[ ¹⁴ , ¹⁵ ] As observed with the cyclopeptide alkaloids, BURPs were also found clustered with these precursor peptides (Supporting Information Table). In addition to the known hibispeptins, the presence of uncharacterized core sequences suggested that undiscovered analogs may exist (Supporting Information Table). Therefore, we analyzed H. syriacus root extracts by LC‐MS. As expected, detected ions matching the known molecules hibispeptin A and B were observed and supported by MS/MS fragmentation (Figures S20 and S21). To discover unreported hispeptin analogs that could be explained by the precursor peptides, we generated a GNPS molecular network ^[28] (Figure S22). A six‐node cluster was found that contained both hibispeptin A and B. In addition to suspected derivatives of hibispeptin A and B, a node with an m/z of 669.3638 was identified. This m/z matches the predicted value (669.3606 m/z, Δ 4.77 ppm) for the [M+H]⁺ ion of the Q‐V‐P‐L‐V‐Y core peptide which is found in the same gene as the hibispeptin A and B cores (Figure S19). MS/MS analysis supported the presence and structure of this new hibispeptin analog (Figure S23).

Along with putatively linking known molecules to their precursor peptides, we also sought to use our SSN to discover new side chain cross‐linked cyclopeptides from a previously unknown producer. Multiple members of the Coffea genus, including Coffea arabica, had potential precursor peptides with clear tyrosine‐containing sequence repeats (Figures 3, S11, and S24). Notably, over 50 % of these Coffea precursor peptide sequences co‐localized with a BURP protein (Supporting Information Table). Therefore, we analyzed methanol extracts of C. arabica by LC‐MS/MS and generated a combined GNPS molecular network with C. americanus to identify features related to cyclopeptide alkaloids (Figure S25). Twelve nodes from C. arabica clustered together and we could clearly map five of these mass features to core sequences (Figures S25 and S26–S30). Based on the MS/MS fragmentation and analogy to known cyclopeptides, these new secondary metabolites from C. arabica were proposed to be four or five amino acids in size and cyclized through an ether bond between the conserved C‐terminal Tyr and the β‐carbon of Ile/Leu. No decarboxylation was observed at the C‐terminal Tyr, similar to cyclo‐[VPIFY] from G. max. ^[6] We confirmed the predicted structure for the 711.3503 m/z ion by using spectroscopic analysis. Key HMBC and NOESY correlations supported the C−O crosslink between the β‐carbon of leucine and the phenolic oxygen of tyrosine (Figure 4C). We named this new cyclic peptide arabipeptin A. As initial assays testing for antifungal activity against Aspergillus fumigatus and antibacterial activity against Staphylococcus aureus (SA1199) and Escherichia coli (ATCC14948) were negative, the biological function of arabipeptin A remains unknown (Figure S31).

Figure 4.

A) Arabipeptin A gene cluster from C. arabica. B) In vitro reconstitution of the ArbB2 split BURP showed modification of the arabipeptin A core found in ArbA2 (highlighted in red). Trypsin digestion of the reaction revealed a loss of 2 m/z (z=2) in the ArbA2 core peptide. C) Structure of arabipeptin A. Pink and blue arrows represent key NOESY and HMBC correlations, respectively. D) Cladogram of Viridiplantae species with a stand‐alone precursor peptide. Orange dots indicate the presence of known structures derived from a stand‐alone precursor peptide. Clades are highlighted and orders for the core eudicots are labeled adjacent to the corresponding black bar.

To better understand the biosynthesis of arabipeptin A, we examined the genomic context of the precursor peptide (NCBI accession: XP_027066141.1) and identified a putative biosynthetic gene cluster, which we named arb (Figure 4A). This genomic region comprises two precursor peptides containing the arabipeptin A core peptide (arbA1 and arbA2), three split BURP peptide cyclases (arbB1, arbB2, and arbB3), a peptidase, and two methyltransferases (Figure 4A). We targeted this cluster to validate our hypothesis that a split BURP protein was indeed responsible for the macrocyclization of the arabipeptin A core peptide. We heterologously expressed an N‐terminally truncated ArbB2 (ΔMet75) as a maltose binding protein fusion construct in E. coli. Following purification and refolding, enzymatic activity was assayed with a truncated ArbA2 Arg23‐Leu102 precursor peptide in the presence of Cu^II. ^[6] LC‐MS/MS analysis of the reaction revealed a loss of 2 m/z as compared to a control reaction completed in the presence of the copper chelator EDTA or in the absence of ArbB2 (Figure S32). Further LC‐MS/MS characterization of a trypsin digest of the modified ArbA2 Arg23‐Leu102 peptide revealed that the arabipeptin A core‐containing fragment had a loss of two hydrogens compared to the EDTA‐containing control reaction (Figures 4B). MS/MS analysis showed that the modification occurred specifically in the core peptide sequence FLWGY and agreed with the proposed Leu‐Tyr side chain macrocyclization observed in arabipeptin A (Figures S33 and S34).

Even though C. americanus, C. arabica, and H. syriacus all contain these split BURP pathways, they are not closely related by phylogeny. To better understand the prevalence and distribution of the split BURP systems in Viridiplantae, a cladogram was constructed from all species identified as having a split BURP peptide cyclase by our custom HMM (Figures 4D and S35). This analysis showed that the split precursor peptides BURPs are nearly ubiquitous among eudicots (Supporting Information Table). We also mapped natural products with known structures that can be attributed to stand alone precursor peptides onto the cladogram. These molecules were distributed across multiple orders including rosales (cyclopeptide alkaloids), malvales (hibispeptins), and gentianales (arabipeptins).

In summary, we linked cyclopeptide alkaloids from C. americanus to their precursor peptides and related these results to additional cyclopeptide alkaloids from Z. jujuba. Unexpectedly, the precursor peptides were not found fused to BURP peptide cyclases as has been the case for other plant side chain cross‐linked cyclopeptides. Instead, the cyclopeptide alkaloids appear to be the first example of a split BURP biosynthetic system. A similar split and fused enzyme/precursor system was recently shown to exist in the borosin family of RiPPs.[ ²⁹ , ³⁰ ]

In addition to cyclopeptide alkaloids, we found that the known molecules hibispeptin A and B are formed from stand‐alone precursor peptides. Our isolation of arabipeptin A from the roots of C. arabica validates that the core peptide sequences can indeed be used as a guide for the discovery of new plant metabolites. Moreover, our reconstitution of ArbB2 illustrates that these split BURPs are active peptide cyclases and form biosynthetic cassettes with their cognate stand‐alone precursor peptides. It is now apparent that numerous plant cyclic peptide scaffolds are produced using a conserved BURP biosynthetic enzyme. Ultimately, our results lay the foundation for future studies to completely elucidate and re‐engineer the biosynthesis of split plant side chain macrocyclized peptides.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Supporting Information

Lima S. T., Ampolini B. G., Underwood E. B., Graf T. N., Earp C. E., Khedi I. C., Pasquale M. A., Chekan J. R., Angew. Chem. Int. Ed. 2023, 62, e202218082; Angew. Chem. 2023, 135, e202218082.

Data Availability Statement

The data that support the findings of this study are available in the Supporting Information of this article.