Enzymatic Synthesis Assisted Discovery of Proline-Rich Macrocyclic Peptides in Marine Sponges

Abstract

Proline-rich macrocyclic peptides (PRMPs) are natural products present in geographically and phylogenetically dispersed marine sponges. The large diversity and low abundance of PRMPs in sponge metabolomes precludes isolation and structure elucidation of each individual PRMP congener. Here, using standards developed via biomimetic enzymatic synthesis of PRMPs, a mass spectrometry-based workflow to sequence PRMPs was developed and validated to reveal that the diversity of PRMPs in marine sponges is much greater than that has been realized by natural product isolation-based strategies. Findings are placed in the context of diversity-oriented transamidative macrocyclization of peptide substrates in sponge holobionts.

Graphical Abstract

By adopting an integrated approach of chemoenzymatic synthesis of standards and HCD-MSⁿ fragmentation, we have sequenced proline-rich cyclopeptide natural products from marine sponges. Our MSⁿ fragmentation workflow successfully differentiated between leucine and isoleucine residues in the cyclopeptides and can be used to sequence low abundance cyclic peptides from sub-gram amounts of sponge biomass.

Macrocyclization of ribosomal and non-ribosomal peptides is ubiquitous in natural product chemistry. Macrocyclization imparts pharmacologically desirable properties of stability and rigidity which reduces the entropic cost associated with target binding.^[1] From neutrophil defensins and snake venom toxins to cyclic peptide natural products produced by bacteria, across all scales of life, macrocyclic peptides demonstrate potent bioactivities and serve various different biological roles.^[2] Benthic marine invertebrates- sponges, are prolific sources of proline rich macrocyclic peptides (PRMPs).^[3] PRMPs are detected in marine sponges without restriction of geographical location and are widespread among several sponge phyla. Cell growth inhibitory cytotoxic activities of sponge derived PRMPs has sustained interest in their development as drug candidates.^[3] Unlike other natural product chemical classes, the abundance of which dominates the sponge metabolomes, PRMPs in sponges are present in trace amounts. For instance, in Stylissa spp. sponges the metabolomes of which are enriched in pyrrole imidazole alkaloids (PIAs),^[4] PRMPs were isolated in 0.01% yields (PRMP weight/sponge dry weight).^[5] Low abundance of PRMPs interspersed with much higher abundance other natural products in sponge extracts encumbers their structure elucidation using natural product isolation-based strategies. Here, we devise a mass spectrometry-based workflow to sequence sponge-derived PRMPs and validate the spectrometric sequencing using standards furnished by biomimetic enzymatic peptide macrocyclization.

We previously described the detection of PIAs in Stylissa and Axinella spp. sponges using LC/MS-based untargeted metabolomics.^[6] Screening these LC/MS data using the DEREPLICATOR tool^[7] hinted at the presence of peptidic natural products in Stylissa and Axinella spp. extracts consistent with prior reports describing the isolation of PRMPs from these sponge genera.^[3] The low abundance of PRMPs and sub-gram amounts of dry sponge biomass which were available to us precluded a natural product isolation-based effort. Thus, we decided to pursue PRMP structure elucidation from mass spectrometry fragmentation data.

In a proline containing peptide, fragmentation preferentially occurs at the N-terminus of the Pro residue.^[8] Sequencing PRMPs detected in sponge extracts thus began with the annotation of the ^NPro-Xaa^C dipeptide oxonium MS² fragment ions per the inventory shown in Table S1. Proceeding from the ^NPro-Xaa^C dipeptide, to recover the PRMP sequence, we annotated mass shifts corresponding to proteinogenic amino acids till we reached the parent ion in the MS² spectra. As shown in Fig. 1A for a PRMP detected in Axinella sp., PRMPs containing more than one proline residue yielded multiple ^NPro-Xaa^C dipeptide MS² fragments. Progressing from each of these different dipeptide fragments, ^NPro-Phe^C (MS² m/z 245) and ^NPro-Glu^C (MS² m/z 227), the same cyclic sequence, cyclo(FFPEXWP), was recovered. Here, X denotes Leu/Ile residues that cannot be differentiated based on MS² fragmentation alone. Identities of the amino acid constituents in cyclo(FFPEXWP), Phe, Pro, Glu, Leu/Ile, and Trp, were supported by the detection of the individual amino acyl immonium ions in the MS² spectra (Fig. S1, Table S2).

Figure 1.

(A) MS² spectra for cyclo(FFPEXWP). MS²(86)→MS³ spectra for the (B) Ile-derived MS² m/z 86 precursor ion (cyclo(FFPEIWP) standard), (C) Leu-derived MS² m/z 86 precursor ion (cyclo(FFPELWP) standard), and the (D) Leu/Ile-derived MS² m/z 86 precursor ion (cyclo(FFPEXWP) PRMP). (E) Extracted ion chromatograms (EICs) showing co-injection of cyclo(FFPELWP) (top) and cyclo(FFPEIWP) (bottom) standards with the sponge extract. (F) Inventory of single Leu/Ile containing PRMPs sequenced in this study.

To resolve Leu and Ile residues, we developed standards for the two possible PRMPs– cyclo(FFPELWP) and cyclo(FFPEIWP). For this, synthetic linear peptides FFPELWPFQA and FFPEIWPFQA were obtained. The C-terminal FQA tripeptide was proteolyzed and transamidatively macrocyclized products were furnished by the plant peptidase PCY1.^[9] Upon fragmentation of the cyclo(FFPELWP) and cyclo(FFPEIWP) products, the MS² immonium ion for Leu/Ile, m/z 86, was isolated and fragmented. Consistent with prior reports,^[10] fragmentation of the Ile-derived MS² m/z 86 immonium ion generated a characteristic MS³ m/z 69 product ion in high abundance, while the MS³ m/z 69 product ion was detected in lesser abundance when the MS² m/z 86 precursor ion was derived from Leu. By modulating the MS³ fragmentation energy (Fig. S2), we generated an unambiguous binary result in which the MS³ m/z 69 product ion was observed only for Ile (Fig. 1B), but not observed for Leu (Fig. 1C). Adopting the thus optimized methodology for the cyclo(FFPEXWP) PRMP detected in the sponge extract, we annotated X as Leu (Fig. 1D). Validation of the mass spectrometry-based result was obtained by spiking the sponge extract with enzymatically synthesized cyclo(FFPELWP) and cyclo(FFPEIWP) standards; the cyclo(FFPELWP) standard coeluted with the sponge-derived PRMP (Fig. 1E) which we term as axinellin D. PRMP congeners detected in this study are highlighted in Table S3.

Employing the now validated mass spectrometry workflow which could differentiate between Leu and Ile residues, we additionally sequenced the PRMPs stylissatin E cyclo(FVPELWP) and stylissatin F cyclo(FPWVPLTP) from Stylissa sp. specimens used in this study (Fig. 1F, S3–6, Table S3). We also detected the presence of the PRMPs cyclo(FGPELWP), cyclo(PAVMLRP), and cyclo(FPLTVPWP) in Axinella sp. extracts (Fig. S7–12). Primary sequences of these PRMPs correspond to the previously described hymenamide C, hymenamide F, and axinellin B. The presence of axinellin B serves as an internal control as it has been reported from an Axinella sponge previously.^[11] However, hymenamides C and F were isolated from Hymeniacidon sp. sponges (Table S3).^[12] Hymeniacidon sp. sponges belong to a different phylogenetic order (Suberitidia) as compared to Axinella sp. (order Axinellida) within the Demospongiae class of marine sponges.

Abovementioned workflow is applicable when a single Leu/Ile is present in the peptide sequence. When more than one Leu/Ile residues are present, multiple side chains will contribute to MS² m/z 86 ion which will complicate the MS²(86)→MS³-based Leu/Ile distinction. This scenario presents itself for the cyclo(FYSX¹AX²P) PRMP that we detected in Axinella sp.; Xⁿ denote Leu/Ile residues. As before, the PRMP was sequenced starting from the ^NPro-Phe^C MS² m/z 245 fragment ion (Fig. 2A, S13). Next, the ion corresponding to the fragment PFYSX¹A, MS² m/z 679 (Fig. 2A) was isolated and the MS³ m/z 86 Leu/Ile immonium ion was generated. Here, only a single residue, X¹, contributes to the production of the MS³ m/z 86 product ion. Further fragmentation of the MS³ m/z 86 ion did not generate the characteristic MS⁴ m/z 69 ion (Fig. 2B), denoting that X¹ was Leu.

Figure 2.

(A) MS² spectra for the PRMP cyclo(FYSX¹AX²P) detected in Axinella sp. where X corresponds to Leu/Ile. MSⁿ fragmentation strategies to query X¹ and X² identities are illustrated. (B) Confirmation of X¹ as Leu as MS⁴ m/z 69 product ion is not detected. (C) Confirmation of X² as Ile as MS⁴ m/z 69 product ion is detected. (D) EICs showing co-injection of enzymatically synthesized PRMP standards with the sponge extract. (E) Inventory of multiple Leu/Ile containing PRMPs sequenced in this study.

Next, the cyclo(FYSLALP), cyclo(FYSIALP), cyclo(FYSLAIP), and cyclo(FYSIAIP) standards were generated by PCY1-mediated macrocyclization of the corresponding -FQA appended linear peptides. Fragmenting cyclo(FYSIALP) and cyclo(FYSIAIP) using the MS²(679)→MS³(86)→MS⁴ procedure yielded the characteristic MS⁴ m/z 69 ion while fragmentation of cyclo(FYSLALP) and cyclo(FYSLAIP) did not (Fig. S14). To query the identity of X² in the PRMP cyclo(FYSX¹AX²P), the MS² m/z 185 ion was isolated. This MS² ion can be generated by the X¹A, or the AX² dipeptides. The MS²(185)→MS³(86)→MS⁴ fragmentation scheme demonstrated the production of the characteristic MS⁴ m/z 69 ion (Fig. 2C). As X¹ was Leu (and hence the X¹A fragment could not contribute to the production of the MS⁴ m/z 69 product ion), these data denote that X² was Ile. Analogous MS²(185)→MS³(86)→MS⁴ fragmentation for the cyclo(FYSLAIP) standard generated the MS⁴ m/z 69 ion while the cyclo(FYSLALP) standard did not (Fig. S15). Validation of the mass spectrometry-based PRMP sequencing scheme was obtained by spiking the sponge extract with the PCY1 generated standards; the cyclo(FYSLAIP) standard coeluted with the sponge PRMP which is termed axinellin E (Fig. 2D, Table S3).

Using the workflow developed above, we additionally sequenced the multiple Leu/Ile containing PRMPs axinellin F cyclo(FFPELLP) from Axinella sp., and stylissatin G cyclo(WLPLTPLP) from Stylissa sp. (Fig. 2E, Fig. S16–19, Table S3). In addition to these PRMPs, we detected the presence of the previously described PRMPs hymenamide H^[13]- cyclo(WVPLTPLP) and phakellistatin 18^[14]- cyclo(YPIFPIP) from Stylissa sp. (Fig. 2E, Fig. S20–23). Hymenamide H was described from Hymeniacidon sp. (order Suberitidia) while phakellistatin 18 was described from Phakellia fusca (order Axinellida), sponges belonging to orders different from Stylissa sp. (order Scopalinida). Together with the detection of hymenamide C in Axinella sp., it is evident that PRMP congeners are shared among phylogenetically different sponges.

Unlike axinellin F, cyclo(FFPELLP), where Ile residues occur next to Leu/Ile residues in a PRMP sequence, mass spectrometry alone may be insufficient to provide unambiguous results. This scenario presented itself during the sequencing of the PRMP cyclo(X¹PHPYX²X³GP) that we detected in Axinella sp. As before, the PRMP was sequenced starting from the ^NPro-His^C (m/z 235), ^NPro-Tyr^C (m/z 261), and ^NPro-X^1C (m/z 211) dipeptide MS² ions (Fig. 3A). Using PCY1 and appropriately designed linear peptide substrates, we generated standards for all eight possible Leu/Ile combinations for cyclo(X¹PHPYX²X³GP). Residue X¹ was annotated as Ile by isolating the MS² m/z 211 ion corresponding to the PX¹ dipeptide and following the MS²(211)→MS³(86)→MS⁴ fragmentation scheme (Fig. 3B). Annotation of X¹ as Ile was supported by analogous fragmentation of the cyclo(LPHPYX²X³GP) and cyclo(IPHPYX²X³GP) standards (Fig. S24–25). Similarly, identity of X² as Ile was decided by isolating the MS² m/z 608 ion corresponding to the PHPYX² pentapeptide and following the MS²(608)→MS³(86)→MS⁴ fragmentation scheme (Fig. 3C) and comparing the MS⁴ spectra to cyclo(IPHPYLX³GP) and cyclo(IPHPYIX³GP) standards (Fig. S26). Because X² was Ile (which generates the MS⁴ m/z 69 diagnostic ion) and no MS² fragment could be detected which contained only X³ without X¹ or X², identity of X³ could not be determined by mass spectral fragmentation. To annotate X³, retention times for the cyclo(IPHPYILGP) (X¹,X²=Ile; X³=Leu) and cyclo(IPHPYIIGP) (X¹,X²=Ile; X³=Ile) standards were compared to reveal the sponge PRMP as cyclo(IPHPYIIGP) which we term axinellin G (Fig. 3D, Table S3). Among the inventory of PRMPs, cyclo-nonapeptides such as axinellin G are relatively rare (Table S3). Overall, we demonstrate the utility of enzymatic synthesis of cyclic peptide standards for discovery and sequencing of sponge PRMPs. At present, we cannot determine the cis/trans conformation of the Pro residues.

Figure 3.

(A) Sequencing of the cyclo(X¹PHPYX²X³GP) based on MS² spectra. Note that three Pro residues in this octapeptide generate three different ^NPro-Xaa^C MS² dipeptide product ions. Fragmentation strategies to query identities of X¹ and X² residues are illustrated. Annotation of (B) X¹ and (C) X² residues as Ile by the detection of the MS⁴ m/z 69 product ion. (D) Annotation of X³ as Ile by retention time comparison of the sponge detected PRMP against enzymatically synthesized cyclo(IPHPYILGP) and cyclo(IPHPYIIGP) standards.

Amino acid residues reported to date for sponge PRMPs are proteinogenic amino acids in the L- configuration (to the best of our knowledge the only exception is Trp derived L-kynurenine in phakefustatins A–C^[15]). It is thus conceivable that sponge-derived PRMPs are ribosomally synthesized and posttranslationally modified peptides^[16] and that the PCY1 catalyzed proteolysis of a C-terminal recognition sequence (RS in Fig. 4A; PCY1 RS is FQA^[9]) and intramolecular transamidative macrocyclization mimics their physiological biosynthetic route. Peptide macrocyclization by prolyl oligopeptidases and cyanobactin macrocyclases is well established.^[16] A recurring feature in biosynthesis of macrocyclic peptides is the requirement of a proline residue or a Ser/Thr/Cys-derived azol(in)e heterocycle immediately upstream of the RS (shaded green in Fig. 4A).^{[9, 17]} Organizing the sponge PRMP octa- and heptapeptide sequences containing a single Pro residue such that the Pro is at the C-terminus leads to the observation that the transamidating N-terminal residues in PRMP linear peptides are hydrophobic amino acids with Trp, Phe, and Tyr predominating (Fig. 4B). Further organizing PRMP sequences containing multiple Pro residues in this sequence alignment then identifies position 2 to be a site for hypervariability.

Figure 4.

(A) Intramolecular transamidative peptide cyclization catalyzed by serine proteases such as PCY1. (B) WebLogo representation of PRMP octa- (top) and heptapeptides (bottom) with a C-terminal Pro. (C) Representative phylum-level microbiome architectures for Axinella,^[6] Stylissa,^[6] Phakellia,^[20] and Hymeniacidon^[20] spp. sponges. Microbiome diversity is denoted by Shannon indices. Shannon indices for HMA sponges are typically greater than 4.

It is as yet unknown whether the sponge host or a bacterial symbiont produces PRMPs in sponge holobionts. To date, natural products that have been shown to be produced by bacterial symbionts within marine sponges are all derived from high microbial abundance (HMA) sponges.^[18] As microbial diversity correlates with microbial abundance for sponge microbiomes that are not dominated by cyanobacteria,^[19] PRMP harboring sponges of the genera- Axinella, Stylissa, Phakellia, and Hymeniacidon spp. are all LMA sponges (Fig. 4C).

The diversity of natural products in marine sponges is much greater than that has been realized.^[21] Here, with the sequencing of 12 macrolactam cyclopeptide sequences of which seven were novel PRMP sequences including a rare cyclononapeptide (Table S3), we demonstrate that inventorying natural product diversity using mass spectrometry and supplementing structural information using enzymatic synthesis can rapidly expand the natural product chemical space without recourse to sacrificing large amounts of sponge biomass which is otherwise necessary for the isolation of low abundance molecules.