Precursor-guided mining of marine sponge metabolomes lends insight into biosynthesis of pyrrole-imidazole alkaloids

Abstract

Pyrrole-imidazole alkaloids are natural products isolated from marine sponges, holobiont metazoans that are associated with symbiotic microbiomes. Pyrrole-imidazole alkaloids have attracted attention due to their chemical complexity and their favorable pharmacological properties. However, insights into how these molecules are biosynthesized within the sponge holobionts are scarce. Here, we provide a multi-omic profiling of the microbiome and metabolomic architectures of three sponge genera that are prolific producers of pyrrole-imidazole alkaloids. Using a retrobiosynthetic scheme as a guide, we mine the metabolomes of these sponges to detect intermediates in pyrrole-imidazole alkaloid biosynthesis. Our findings reveal that the non-proteinogenic amino acid homoarginine is a critical branch point that connects primary metabolite lysine to the production of pyrrole-imidazole alkaloids. These insights are derived from the polar metabolomes of these sponges which additionally reveal the presence of zwitterionic betaines that may serve important ecological roles in marine habitats. We also establish that metabolomic richness does not correlate with microbial diversity of the sponge holobiont for neither the polar, nor the non-polar metabolomes. Our findings now provide the biochemical foundation for genomic interrogation of the sponge holobiont to establish biogenetic routes for pyrrole-imidazole alkaloid production.

Graphical Abstract

Introduction

Naturally produced specialized small organic molecules that do not serve roles in primary metabolism, colloquially referred to as the natural products, are privileged synthons that have evolved to mediate chemical signaling and initiate biological responses. Natural products directly provide, or inspire the development of a majority of clinically used pharmaceuticals.¹ These molecules are constructed starting from simple organic substrates by gene-encoded enzymes, and biochemical schemes underlying their construction offer new insights into biosynthetic enzymology and genomic architectures.²

Marine sponges, prolific sources of natural products, are filter-feeding sessile metazoans in which a eukaryotic sponge host is associated with a suite of microbial and other eukaryotic partners to construct a holobiont community. Sponge biomes are rich sources of microbiological novelty.^3–4 In sponges, and for other marine metazoan holobionts, the associated natural products are shown to be produced by the microbiome. However, there is increasing timorous acceptance that the eukaryotic host itself could be the site for biosynthesis of several holobiont-derived natural products.^5–7 Some chemical scaffolds are unusually well represented in sponge natural products; among these are the brominated phenols and pyrroles. The pyrrole-imidazole alkaloids are a characteristic example (Fig. 1A). Pyrrole-imidazole alkaloids are a large and chemically complex class of sponge natural products with over 150 congeners. The chemical complexity of pyrrole-imidazole alkaloids has necessitated numerous structural revisions, offered title compounds for organic syntheses, and their favorable bioactivity profiles have sustained pharmacological interest.^8–9 Retrosynthetic routes for pyrrole-imidazole alkaloids posit the enantioselective dimerization of three key monomeric building blocks, oroidin (1), hymenidin (2), and clathrodin (3) (Fig. 1A). Indeed, in in vitro biomimetic experiments, the dimerization of these monomeric building blocks was realized using enzymes extracted from pyrrole-imidazole alkaloid harboring sponges.^10–11 However, insights into the biosynthesis of 1–3 themselves have been scant; the only data that we currently possess were delivered by monitoring the incorporation of radiolabeled amino acids precursors into 1.^12–14 Specifically, while the 2-aminoimidazole moiety in 1–3 is widely present in marine metabolomes, precursors and intermediates involved in its construction have not been identified.^15–16 There are two principal challenges that have precipitated this knowledge gap. First, from the genomic perspective, the biosynthesis of pyrrole-imidazole alkaloids is not expected to resemble any of the characteristic natural product biosynthetic schemes. Hence, genes encoding their construction in sponge metagenomes are hiding in plain sight from computational tools that are trained to detect natural product biosynthetic genes. Second, from the biochemical perspective, detection of substrates and intermediates involved in pyrrole-imidazole alkaloid biosynthesis is outside the purview of metabolomic approaches that are increasingly reliant on reversed-phase chromatography which, in turn, biases the metabolomic description to non-polar molecules.

Fig. 1:

Sponge-derived pyrrole-imidazole alkaloids. (A) Monomers 1–3 and representative dimerization products belonging to the sceptrin, ageliferin, and axinellamine families of pyrrole-imidazole alkaloids. (B) Morphology of Stylissa, Axinella, and Agelas sponge specimens used in this study. For each genus, two biological replicates were collected. (C) Mass spectrometric detection of 1 and 2. Overlaid with the base peak chromatogram (BPC, in black), two EICs are shown that correspond to the m/z [M+H]⁺ ions for molecules 1 and 2. Molecule 3 was not detected in sponge specimens used in this study. Chemical identities were dereplicated based on comparison to reference spectra and manual annotation of MS/MS fragmentation spectra (Fig. S1). (D) PLSDA clustering of Stylissa, Axinella, and Agelas metabolomes represented as ellipses of 95% confidence level. The 4-component model had a Q² value of 0.87. Each biological specimen (two for each genus) were analyzed in triplicate. An overlap of the metabolome is observed on reducing the stringency of the principal components that describe the co-variance between the three sponge genera. (E) Bar plots denoting microbiome composition at the phylum level for sponge specimens used in this study. The microbiome composition was analyzed in duplicate for each sponge specimen used in this study. Networks demonstrating the Bray-Curtis dissimilarity indices at the (F) phylum and at the (G) genus levels for sponge microbiomes. Node color denotes sponge genera, the border color corresponds to technical replicates, in that, nodes of the same color represent the same genera while the same border color represents technical replicates for the sponge specimen. Pairwise indices were calculated for each microbiome dataset and nodes with Bray-Curtis dissimilarity indices less than 0.6 were connected. At the phylum level, all nodes are interconnected. However, at the genus level, only the nodes for each individual sponge genera are connected demonstrating overall low overlap between Stylissa, Axinella, and Agelas microbiomes.

An intriguing feature of pyrrole-imidazole alkaloids is that their production extends across different phylogenetic families of marine sponges. A comparison between metabolomic structures and microbiome communities of these sponges in light of the conservation of the natural product chemistry has not been conducted. A correlative overlap of microbiome community member(s) can possibly lend insight into the identity of the primary producer bacterium within the sponge holobiont, as was realized for the production of polybrominated phenols and polychlorinated peptidic natural products in the Dysideidae family of marine sponges.^17–19 These correlative relationships have been validated for other marine invertebrate holobionts as well, a robustly characterized example being the production of cyanobactin peptides within the microbiomes of marine ascidians.⁵ The site for the production of the pyrrole-imidazole alkaloids in the sponge holobiont has not been resolved.

Here, we provide the multi-omic profiling of the metabolomic and microbiome architectures of Stylissa, Axinella, and Agelas sponges, genera that are exceptionally prolific producers of pyrrole-imidazole alkaloids (Fig. 1B). We generate metabolomic datasets describing the polar and non-polar metabolites harbored by these sponge genera. We demonstrate that the metabolomic convergence does not translate to conversation of microbiome architectures, and neither does the microbiome diversity correlate with metabolomic richness. Within polar metabolites present in sponge extracts, we detect the presence of nonproteinogenic amino acids and derivatives that allow us to refine biosynthetic hypotheses for pyrrole-imidazole alkaloids. In line with our previous findings, we demonstrate that the repertoire of other natural products that have been isolated from these sponges remains vastly underappreciated.^20–21 Data described herein are generated using less than 1 g biomass for each sponge specimen, highlighting the applicability of contemporary -omic technologies in transcending the limitation of biomass availability.

Results and Discussion

Sponge phylogeny and detection of pyrrole-imidazole alkaloids

Sponge specimens were collected in Guam and in Solomon Islands and barcoded by Sanger sequencing of PCR amplicons corresponding to the internal transcribed spacer-2 (ITS-2) region between 5.8S and 28S rRNA encoding genes and the 5’-terminus of the 28S rRNA encoding gene. Sponge genera Stylissa (family Scopalinidae), Axinella (family Axinellidae), and Agelas (family Agelasidae) could be discerned by homology to sequences in the GenBank nr database. We restricted the description of sponge phylogeny to the genus level; for marine sponges, progression to species-level identification is tenuous based on DNA barcoding alone as we have demonstrated before.²¹ Two biological replicates for each sponge genera were used in this study. Replicates of the genus Agelas demonstrate highly similar ITS-2 sequences. However, the 28S rRNA sequences separate the specimens as two different species, supported by their different morphologies (Fig. 1B, Table S1).

Liquid chromatography/mass spectrometry (LC/MS) data were used to reveal the conservation of brominated alkaloids in all three sponge genera (Fig. 1C). Accurate mass determination, characteristic isotopic distribution of brominated species, comparison of the MS² fragmentation spectra to spectra available in databases, and manual annotation of observed MS/MS fragment ions lead to structural assignment of pyrrole-imidazole alkaloids 1 and 2 (Fig. S1). Molecule 3 was not detected in specimens used in this study. The extracted ion chromatograms (EICs) revealed the presence of multiple isomeric species, consistent with reported congeners that arise via intramolecular cyclization for 1 and 2 (Fig. 1C).

In light of the conservation of pyrrole-imidazole alkaloids, we interrogated the overall metabolomic overlap among the three sponge genera. Each sponge specimen was extracted in triplicate (technical replicates) and LC/MS data acquired. Thus, for each of the three genera, Stylissa, Axinella, and Agelas, six LC/MS datasets were collected. A partial least squares discriminant analysis (PLSDA) revealed that the data self-organized according to sponge genera (components 1 and 2, Fig. 1D). The score plot between components 2 and 3 reveal overlap among Stylissa and Axinella metabolomes while Agelas still remained distinct. The score plot between components 3 and 4 shows complete overlap among the three genera. Qualitatively, these findings translate to Agelas and Stylissa metabolomes being distant from each other while Stylissa and Axinella metabolomes are more similar.

Microbiome architectures

The microbiome architectures of the three sponge genera were determined. PCR amplicons for the bacterial 16S rRNA gene for each biological replicate were sequenced in duplicate. At the phylum level, microbiomes of the three sponge genera were dominated by Proteobacteria, Acidobacteria, and Thaumarchaeota, while Agelas microbiomes additionally revealed the presence of Chloroflexi, Actinobacteria, and Poribacteria that were conspicuously absent in Stylissa and Axinella (Fig. 1E). Even though the two Agelas specimens used in this study belong to different species, the overall microbiome architectures are similar. Conservation of microbiome architectures at the genus level is a recurring observation for marine sponges.^{17, 21} In sponge microbiomes, microbial diversity is a predictor of microbial abundance; the presence of Chloroflexi, Acidobacteria, and Poribacteria designate Agelas specimens to be high microbial abundance (HMA) sponges while Stylissa and Axinella specimens are designated as low microbial abundance (LMA) sponges.²²

To quantify the overlap among microbiomes of sponges investigated here, we calculated the Bray-Curtis dissimilarity matrices at the phylum and at the genus levels (Table S2). These matrices are represented as networks in which nodes corresponding to individual microbiomes are connected to other nodes with which they share dissimilarity indices less than 0.6, a cut-off value chosen in light of contemporary literature.²³ Thus, nodes with values less than 0.6 are considered to be similar and hence connected by an edge. At the phylum level, we observe a near complete interconnectedness among the microbiome network (Fig. 1F). This is likely due to the dominance of the Proteobacteria, Acidobacteria, Thaumarchaeota, and Cyanobacteria phyla which make up greater than half of the bacterial diversity for each specimen. However, at the genus level, the inter-genera interconnectedness is completely lost, and the microbiomes fall into sub-networks according to the sponge genera (Fig. 1G). At the phylum-level, a relatively higher dissimilarity index was observed between the Agelas biological replicates (Table S2), consistent with the Agelas specimens belonging to different species as determined by 28S rRNA sequences (Table S1). Still, at the genus level, the inter-species Bray-Curtis dissimilarity indices for the Agelas specimens were less than the cutoff value (0.6). It is thus instructive to observe that the production of pyrrole-imidazole alkaloids was conserved in different sponge genera with very divergent microbiome architectures.

Discovery of homoarginine and homoagmatine in sponge metabolomes

The high abundance of pyrrole-imidazole alkaloids among the three sponge genera investigated here has been noted above. Monomeric pyrrole-imidazole alkaloids, such as 1, can be rationalized to be derived via the condensation of the pyrrole-2-carboxylic acid 4 and aminoalkyl-2-aminoimidazole 5, both of which are reported natural products isolated from Axinella and Agelas genera (Fig. 2A).^24–27 By tracing the incorporation of radiolabeled precursors, different groups have presented convergent data that the pyrrolyl moiety in 1 is derived from proline (6).^12–14 The enzymatic oxidation of the pyrrolidine heterocycle in 6 and pyrrole halogenation are well established transformations.^28–29 The biosynthetic elaboration of 5 is more contentious. Based on radiolabeling experiments, differing data for the incorporation of lysine (7) and histidine into 1 has been presented and the involvement of other polar amino acids such as arginine and ornithine has been proposed.^{8, 30–32}

Fig. 2:

Proposed biosynthesis of pyrrole-imidazole alkaloids. (A) Retrobiosynthetic scheme rationalizing the elaboration of 1 from amino acid precursors. Chemical structures of pyrrole-imidazole alkaloids in which 8 and 9 are rationalized to be directly coupled with pyrrole carboxylic acids are shown in the dashed box. (B–E) EICs and MS² spectra mirror plots comparing 7–9 and 11 detected in the Stylissa metabolome against synthetic standards. Key MS² ions are highlighted.

To furnish 5, we rationalized a biosynthetic scheme starting from 7. In this hypothetical scheme, amidinotransfer from arginine to the side chain primary amine of 7 generates the non-proteinogenic amino acid homoarginine (8, Fig. 2A). Ornithine, a product of this reaction, can, in turn, furnish 6. Decarboxylation of 8 furnishes homoagmatine (9). Support for this hypothesis is derived from the observation that pyrrole-imidazole alkaloids arising from direct condensation of 8 and 9 with (di)bromo-1H-pyrrole-2-carboxylic acids have been described (Fig. 2A).^33–35 While 8 was proposed as a pyrrole-imidazole alkaloid biosynthetic intermediate, it had not been detected in sponge metabolomes previously; invoking it as a biosynthetic intermediate was thus not supported by prior literature.^{13, 36}

To query for the presence of 8 and 9, first, we acquired LC/MS data used using altered sample extraction and chromatographic procedures that allowed us to interrogate the polar metabolites present in the sponge metabolomes that were outside the purview of the reversed-phase chromatographic techniques used above. Comparison of retention time and MS² fragmentation spectra to an authentic standard validates the presence of the proteinogenic amino acid 7 (Fig. 2B). In an identical fashion, by comparison to a commercially available standard, we could detect the presence of 8 in all three, Stylissa, Axinella, and Agelas sponges (Fig. 2C, data shown for Stylissa metabolomes only, see below). Next, we synthesized a standard of 9 by the guanidinylation of 1,5-diaminopentane (Fig. S2–4). Presence of 9 in sponge metabolomes was similarly confirmed by comparison of retention time and MS² spectra with this synthetic standard (Fig. 2D). This is the first report for the detection of the non-proteinogenic amino acid 8 and its decarboxylated derivative 9 in marine sponges.

Experimentally establishing the presence of 8 and 9 in sponge metabolomes allows us to now progress along the rationalized biosynthetic scheme illustrated in Fig. 2A. Conceivable Cδ-hydroxylation for 9 (carbon atom naming convention for amino acid 8) furnishes 10. The decarboxylation-hydroxylation sequence can be inverted to furnish 10 via 11 rather than via 9. Thus, we proceeded to interrogate the presence of 11 as well. Using established procedures,³⁷ a racemic standard for 11 was synthesized by the specific guanidinylation at the ε-NH₂ of the cupric salt of commercially available racemic 5-hydroxylysine (Fig. S5–7). As for 7–9, comparison of retention time and MS² fragmentation spectra establishes the presence of 11 in the sponge metabolomes (Fig. 2E). Thus, at this stage, metabolomic evidence has been presented pointing towards guanidinylation, decarboxylation, and Cδ-hydroxylation of 7 as participating events in pyrrole-imidazole alkaloid biosynthesis. Lysine Cδ-hydroxylation by α-ketoglutarate dependent enzymes has biosynthetic precedent; this modified amino acid is detected in eukaryotic collagen including that in marine sponges.^38–39 Overall, data presented above is consistent with prior experiments demonstrating the incorporation of radiolabeled 6 and 7 in 1.¹⁴ Proceeding along the proposed biosynthetic route, oxidation of the secondary alcohol in 10 to a ketone will facilitate a spontaneous dehydration to install the 2-aminoimidazole heterocycle in 12, as has been realized synthetically,⁴⁰ followed by Cβ-Cγ oxidation to afford 5.

Molecule 8, which had not been described from sponges previously, is a critical branch point which connects the primary metabolite 7 to a natural product biosynthetic process (Fig. 2A). In addition to the proposed amidinotransfer, 7 can be converted to 8 by carbamoylation at the ε-NH₂. At present, we cannot eliminate this possibility. We are biased towards amidinotransfer being the operative transformation because we do not detect carbamylated lysine (homocitrulline) in the sponge metabolomes. Amidinotransfer upon 7 is a biosynthetic reaction validated to be present in the eukaryotic and bacterial metabolic toolkits.^41–42 In vertebrates, 8 and 9 serve as metabolic markers for cardiovascular disease with roles in the production of nitric oxide.⁴³ It is tantalizing to posit that biosynthetic ingenuity has repurposed these metabolites for the elaboration of natural products in sponge holobionts.

Enrichment of pyrrole-imidazole alkaloid biosynthetic intermediates

The abovementioned transformations are all supported by biosynthetic precedent. The conceptual challenge here is that amino acids are primary metabolites and their derivatives could thus be present at a basal level in all sponge holobionts regardless of the production of pyrrole-imidazole alkaloids. Thus, to interrogate the specific enrichment of metabolites 7–12 in sponges that produce pyrrole-imidazole alkaloids, we chose the marine sponge, Dysidea sp. to serve as a comparative negative control (Fig. 3A). Though dominated by amino acid-derived natural products, the Dysidea metabolome has scant overlap with Stylissa, Axinella, and Agelas metabolomes, in that, no brominated alkaloids were detected in its metabolome. Instead, in Dysidea, we detected the presence of polychlorinated natural products in high abundance. Structural annotation of the MS² spectra led to the dereplication of previously reported natural products barbaleucamides A and B (Fig. 3A, S8).⁴⁴ In addition to barbaleucamides, characteristic Dysidea molecules- the polychlorinated diketopiperazine dysamides were also detected (Fig. S8). By analogy to other Dysidea-derived natural products that also possess the leucine-derived trichloromethyl moiety, biosynthesis of barbaleucamides and dysamides is expected to reside within cyanobacterial symbionts of the genus Oscillatoria.¹⁹ Distinct from Oscillatoria, cyanobacterial symbionts within Stylissa, Axinella, and Agelas sponges belong to the Synechococcus and Prochlorococcus genera (Fig. S9). To validate the pyrrole-imidazole alkaloid biosynthetic scheme illustrated in Fig. 2A, we searched for the enrichment of precursor 7 and intermediates 8–12 in the Stylissa, Axinella, and Agelas sponge metabolomes, the key metric being that they should not be present at comparatively high levels in Dysidea with metabolomic data generated using identical extraction and LC/MS procedures.

Fig. 3:

Enrichment of rationalized pyrrole-imidazole alkaloid biosynthesis intermediates. (A) Dysidea sp. metabolome BPC with overlaid EICs demonstrating presence of barbaleucamides A–B. (B) EICs, generated within 1 ppm error tolerance for 7–12 across Stylissa, Axinella, Agelas, and Dysidea polar metabolomic LC/MS datasets. The y-axes for each EIC is identical as indicated, except for 7 in which the high abundance in Stylissa is adjusted with a different y-axis scaling. (C–G) MS² spectra for 8–12 with rationalized structural annotations of fragment ions. The ppm error associated with each structural annotation is listed. The MS² spectra shown here were recorded by highly accurate Fourier transform mass spectrometric fragmentation on an orbitrap mass spectrometer.

Within 1 ppm tolerance, EICs for [M+H]⁺ ions demonstrate that all 7–12 are indeed detected, and are enriched in Stylissa, Axinella, and Agelas metabolomes as compared to the Dysidea metabolome (Fig. 3B). Genus Stylissa demonstrated the highest enrichment of metabolites 7–12 with the relative abundance of different molecules changing between Axinella and Agelas.

The presence of 8, 9 and 11 have been confirmed by comparison to standards; structural assignment of 10 and 12 is supported by annotation of fragmentation spectra (Fig. 3C–G). Fragmentation of 8 demonstrated the neutral loss of -CONH₂, rationalized as imine formation by decarboxylation followed by deamination to install an aldehyde at Cα (m/z 144) (Fig. 3C). Additional fragment ions corresponding to oxidative deamination (m/z 172) and loss of the guanidino group (m/z 130) were also observed. The characteristic C₅-aminoalkyl chain of 8 manifests as the oxidized piperidine (m/z 84). Fragmentation spectra for 9, identical to that of the synthetic standard (Fig. 2D), demonstrated the presence of an oxidatively deaminated product (m/z 128) which established the presence of the primary amine, while the oxidatively deguanidinylated product ion (m/z 86) established the presence of the guanidino group, consistent with 9 being the decarboxylated derivative of 8 (Fig. 3D). MS² spectra for 10 likewise possesses the oxidatively deaminated product ion (m/z 144), and further deamination at the guanidino group (m/z 127) (Fig. 3E). Instructive to observe is m/z 101 product ion, the structure for which is rationalized by dehydration at Cδ (vide infra) followed by ring closure.

Fragmentation spectra for 11, richest and most descriptive of all species queried here, is described in light of that of 8 (Fig. 3F). Molecules 8 and 11 differ only by single hydroxylation, however, the MS² spectra are quite different (Fig. 3C,F). Consistent with the fragmentation of 8, we observed the oxidatively deaminated fragment ion (m/z 188) and the product ion corresponding to the -CONH₂ neutral loss (m/z 160). Unexpectedly, formulae determination for the more abundant fragment ions, m/z 145 and m/z 128, demonstrated one, and two deaminations occurring at the guanidino group after the -CONH₂ neutral loss at Cα. This observation is in contrast to the fragmentation spectra for 8 where no guanidino deaminations were observed (Fig. 3C), but, consistent with fragmentation spectra for 10, another hydroxylated derivative where guanidino deaminations were also detected (Fig. 3E). Progressing from the annotation of the m/z 160 product ion, we rationalized that the guanidino deaminations are facilitated by the formation of oxazoline and oxazolidine heterocycles via the nucleophilic attack of the Cδ-hydroxy oxygen on to the guanidino carbon followed by successive deaminations. Hence, retrospectively, hydroxylation is likely affected upon the homoarginine Cδ. If hydroxylation occurs at Cγ, deamination must progress via the formation of the six-membered 1,3-morpholine ring, a possibility that cannot be entirely discounted. A similar scenario with hydroxylation occurring at Cβ and Cε will lead to formation of less likely seven- and four-membered heterocycles, respectively. The fragmentation spectra for 12 demonstrates the oxidatively deaminated product ion (m/z 124), and deamination at the 2-aminoimidazole ring (m/z 99) (Fig. 3G).

Comparative polar metabolomes

Next, we sought to query which other polar metabolites could be identified in the sponge metabolomes. With the greatest enrichment of 7–12 in Stylissa, as compared to Axinella and Agelas, we compared the Stylissa and the Dysidea polar metabolomes. From the LC/MS datasets that were acquired in technical triplicate for each sponge specimen, mass spectral features were extracted using MZmine2.⁴⁵ Along with the MS¹ m/z ion, mass spectral features are appended with the metric of abundance, as determined by area under the EIC. The Stylissa and Dysidea features are illustrated in Fig. 4A in which the logarithmic fold-change in ion abundance is plotted along the horizontal axis and the logarithmic statistical significance (p-value) of the mass spectral feature in differentiating between the Stylissa and Dysidea metabolomes is plotted along the vertical axis.

Fig. 4:

Comparative polar metabolomes. (A) Mass spectrometric features for the polar metabolomes of Stylissa (left) and Dysidea (right) represented as a volcano plot. Key features described in text are highlighted. Retention time comparison and MS² mirror plots for (B) 13 and (C) 14 detected in the Stylissa metabolome against authentic standards. Note that EICs corresponding molecular formulae for both 13 and 14 demonstrate more than one peak.

Features corresponding to 7–12 are some of the most enriched features in the Stylissa metabolome (blue dots, Fig. 4A). We also observe the enrichment of 6, precursor for 1 in Stylissa. On the other hand, other amino acids are either evenly distributed (Gln, Orn, Val), or are not highly differentiated (Gln, Trp, Phe, Leu, Arg; green dots). Both, histidine and histamine (labeled His(-CO₂)), are not highly differentiated among the two metabolomes.

Instructive to compare is the distribution of amino acids 7 and 8 against that of leucine in Stylissa and Dysidea polar metabolomes (Fig. 4). Given the abundance of leucine-derived barbaleucamides and dysamides in the Dysidea non-polar metabolome, we expected leucine to be highly enriched in the Dysidea polar metabolome. However, this was not the case. In the Axinella vs Dysidea, and Agelas vs Dysidea comparative plots, we observed a similar equitable distribution of leucine with enrichment of 7 and 8 in Axinella and Agelas as compared to Dysidea (Fig. S10). We rationalize the equitable distribution of leucine by the fact that biosynthesis of barbaleucamides and dysamides resides within a cyanobacterial symbiont inside the Dysidea sponge host.¹⁹ However, the polar metabolome, which is representative of primary metabolic state of the entire holobiont, will be dominated by the eukaryotic sponge host. In context of this argument, does the enrichment of precursors 7 and 8 in the polar metabolomes suggest that the biosynthesis of the pyrrole-imidazole alkaloids is catalyzed by the sponge host rather than a bacterial symbiont? Two lines of evidence support this assertion. Firstly, by culturing the archaeocyte cells from the sponge Teichaxinella morchella, phylogenetic assignment revised to Axinella corrugata,⁴⁶ Kerr successfully recovered the production of pyrrole-imidazole alkaloids.¹² Secondly, using enzyme extracts from the LMA sponge Stylissa caribica, Molinski reconstituted the late stage intermolecular coupling of 1 to dimeric pyrrole-imidazole alkaloids (Fig. 1A).^10–11 The hypothesis that the sponge host, and not the microbiome is the biosynthetic source would also support the observation that Stylissa, Axinella, and Agelas sponge genera bearing very diverse microbiome structures support pyrrole-alkaloid production (Fig. 1E). This observation is in contrast to other marine invertebrate systems in which bacterial symbionts have been validated to be sites for natural product biosyntheses; in all such systems the microbiome architecture remains correlatively conserved with the natural product chemistry across the eukaryotic host phylogeny.^{4–5, 17, 47}

In addition to primary metabolites and amino acid derivatives, by comparison to authentic standards, the quaternary ammonium betaines stachydrine (13) and trigonelline (14) were also identified in the sponge polar metabolomes (Fig. 4B,C). Among these, 14 has been described from marine sponges previously.⁴⁸ While 13 was enriched in Dysidea, 14 had a higher abundance in the Stylissa metabolome (red dots, Fig. 4A) Molecule 14, in particular, is widely distributed in the marine metabolome. In addition to sponges, 14 has been detected in corals, seaweeds, and phytoplankton.^49–51

Correlative microbiome-metabolome richness

At this stage, we have an inventory of the polar and the non-polar metabolites of Stylissa, Axinella, and Agelas sponge specimens. Next, we queried whether the microbiome diversity correlates with the metabolomic diversity for these marine sponges; in other words, do more diverse sponge holobionts produce more polar and non-polar metabolites? To address this question, we plotted the Shannon indices, a measure of α-diversity of the sponge microbiome, against the detected polar and non-polar mass spectrometric features. Surprisingly, we observed that Agelas, an HMA sponge with the highest Shannon index among the three sponge genera was associated with the lowest metabolomic diversity (Fig. 5A). On the other hand, LMA sponges, Axinella and Stylissa, possess higher relative number of polar and non-polar metabolites. It is instructive to observe that the numerical spread of non-polar metabolites is higher than that for the polar metabolites (Fig. 5A, y-axes). We additionally observe that a greater fraction of the polar metabolome is shared among the three sponge genera (Fig. 5B). We rationalize these observations with the assertion that the polar metabolome is more descriptive of primary metabolites as compared to the non-polar metabolomes which will be biased towards more specialized and thus non-overlapping secondary metabolites and natural products.

Fig. 5:

Microbiome and metabolome diversity. (A) Plot for Shannon indices representing the microbial diversity (along x-axis) and metabolite diversity (along y-axes) for Stylissa, Axinella, and Agelas specimens used in this study. Metabolomic diversity for both, non-polar and polar metabolomes is represented. The horizontal error bars denote standard deviation across four microbiome datasets recorded for each sponge genera, as shown in Fig. 1E. The vertical error bars denote standard deviation in number of mass spectral features detected across six LC/MS datasets for each sponge genera. (B) Euler diagrams demonstrating the overlap of non-polar (left) and polar (right) metabolomic features across the three sponge genera.

Metabolomics guides contemporary investigations in chemical ecology.⁵² An open question is in which of the two fractions, the polar or the non-polar metabolomes, do the ecologically relevant metabolites lie? Specific for sponge genera investigated here, conflicting roles for pyrrole-imidazole alkaloids in protecting sponges against herbivory and predation have been presented. While their role in defending Agelas sponge species in the Caribbean and Axinella species in the Mediterranean has been established, no such protective function for pyrrole-imidazole alkaloids was realized for Stylissa sponges in the Indo-Pacific.^53–55 On the other hand, zwitterionic betaines, such as 14 identified here in the Stylissa polar metabolome, have been shown to be primary waterborne cues that mediate predator-prey interactions.⁵⁶ As a class of molecules, the primacy of zwitterionic betaines in mediating ecological interactions is well established. In terrestrial plants, 13 and 14 were shown to modulate interactions between plants and bacteria in the rhizosphere.⁵⁷ While it is tantalizing to propose that sponge natural products serve defensive roles in their physiological setting, this hypothesis may not be universally valid. Indeed, due to paucity of natural products detected in their metabolomes, some sponges are classified as ‘chemically undefended’.⁵⁸ We posit that polar metabolites that have stayed outside the purview of isolation and structural characterization efforts may mediate core ecological functions, perhaps in synergy as multicomponent cues with non-polar natural products. A comprehensive inventory of both, the polar and the non-polar metabolites in sponge holobionts is critical in discerning chemical crosstalk in organismal interactions.

Materials and Methods

Materials and methods used in this study are described in the Supporting Information.