Metagenomic natural product discovery in lichen provides evidence for a family of biosynthetic pathways in diverse symbioses

Significance

Remarkable chemical families are being recognized by studying diverse symbioses. We identified, through metagenomics, the first cyanobacterial trans-AT polyketide biosynthetic pathway in the Nostoc symbiont of the lichen Peltigera membranacea and showed its expression in natural thalli. An isotope-based technique designed for characterizing minute amounts of material confirmed predictions that its product, nosperin, is a distinct member of the pederin family of compounds that was previously thought exclusive to animal–bacteria associations. The unexpected discovery of nosperin in lichen expands the structural range and known distribution of this family of natural products and suggests a role associated with symbiosis.

Abstract

Bacteria are a major source of natural products that provide rich opportunities for both chemical and biological investigation. Although the vast majority of known bacterial metabolites derive from free-living organisms, increasing evidence supports the widespread existence of chemically prolific bacteria living in symbioses. A strategy based on bioinformatic prediction, symbiont cultivation, isotopic enrichment, and advanced analytics was used to characterize a unique polyketide, nosperin, from a lichen-associated Nostoc sp. cyanobacterium. The biosynthetic gene cluster and the structure of nosperin, determined from 30 μg of compound, are related to those of the pederin group previously known only from nonphotosynthetic bacteria associated with beetles and marine sponges. The presence of this natural product family in such highly dissimilar associations suggests that some bacterial metabolites may be specific to symbioses with eukaryotes and encourages exploration of other symbioses for drug discovery and better understanding of ecological interactions mediated by complex bacterial metabolites.

Symbiosis, defined by de Bary (1) as the “living together of two organisms,” includes a broad range of partnerships, from loose associations to obligate interdependencies and host–parasite interactions. Many involve microbes, with perhaps the most successful—between bacteria and early nucleated cells in the Precambrian—leading to mitochondria and chloroplasts in modern eukaryotes (2). Symbiotic interactions are being examined with increasing molecular detail, focusing not only on attributes that may be beneficial for each organism individually but also on what might be important for the association. It is increasingly being recognized that biosynthetic pathways leading to synthesis of specialized metabolites may play key roles in the biology of symbiosis (3).

Lichens are ancient and physiologically highly integrated symbioses between heterotrophic filamentous fungi (mycobionts) and cyanobacteria or coccoidal green algae (photobionts) that may date as far back as 600 Mya (4). The morphology of the characteristic and stable macroscopic body of a lichen, the thallus, typically bears little resemblance to the individual organisms that form it and, in many cases, can be highly organized: fungal cells on the periphery for physical support and protection and photobiont cells inside, providing photosynthate or fixed nitrogen or both (5) (Fig. 1 A–C). Although the photobionts can often be isolated in pure culture (Fig. 1D), most mycobionts (almost exclusively from the Ascomycota) are refractory to propagation in vitro by standard methods, and intact lichens cannot be maintained artificially for long. Nevertheless, such limitations are gradually being overcome using advanced analytical platforms, e.g., metagenomics in the characterization of mycobiont lectin genes (6), and PCR-based phylogenetics in investigation of intrathalline bacterial diversity (7).

Fig. 1.

The foliose lichen Peltigera membranacea and Nostoc symbiont. (A) Lichen in situ. (Scale bar, 5 cm.) (B) Rhizines (Rhi) on lower surface and apothecia (Apo) protruding from thallus edge. (C) Thallus cross section illustrating stratified internal structure including photosynthetic cyanobiont layer (shown with arrows) between cortical and medullary mycobiont layers (above and below, respectively). (Scale bar, 100 μm.) (D) Nostoc sp. N6 in culture. (Scale bar, 100 μm.) (photograph for Fig. 1C, courtesy of Martin Grube).

In a number of bacterial–eukaryote symbioses, bacterial partners have been implicated in the production of complex molecules derived from polyketide synthase (PKS) and nonribosomal peptide synthetase (NRPS) pathways (3, 8, 9). Examples include pederin, made by bacteria that live in rove beetles of the genus Paederus, and structurally related metabolites, the onnamides and psymberin, produced by bacteria that live in marine sponges (Fig. 2). In general, metabolites known or suspected to be of symbiont origin show remarkably low structural overlap with natural products discovered in screening programs from free-living bacteria (10). This phenomenon raises the intriguing question of whether symbiont chemistry might encompass structural scaffolds covering distinctive regions of chemical space.

Fig. 2.

Pederin family compounds and symbioses. (Upper Left) Image of Paederus fuscipes courtesy of Christoph Benisch (www.kerbtier.de). (Upper Right) Image of Theonella swinhoei courtesy of Yoichi Nakao. (Lower Right) Image of Psammocinia aff. bulbosa adapted with permission from ref. 15. Copyright 2007 American Chemical Society. (Lower Left) Image of Mycale hentscheli courtesy of Mike Page.

In this study, we applied a combination of metagenomic and natural product discovery methods to identify nosperin, the first member of the pederin family from a lichenized cyanobacterium and a further example toward the emerging concept of symbiosis-associated natural product pathways (10).

Results

Discovery of Trans-AT PKS Genes in the Lichen Metagenome.

Peltigera membranacea is a widely distributed terrestrial lichen carrying Nostoc sp. as its photobiont (Fig. 1 A–D). Total lichen DNA extracted from field samples collected in Iceland processed for whole genome sequencing (WGS) (11) revealed approximately equal contributions from the mycobiont, the photobiont, and the community of intrathalline microbes. Bioinformatic mining of the initial metagenome assembly yielded 18 candidate clusters containing genes that encode PKS enzymes (SI Appendix, Table S1). Among the putative bacterial gene clusters, two were members of the trans-acyltransferase (AT) PKS family (Fig. 3; SI Appendix, Fig. S1) in which AT domains are not encoded by the PKS genes but rather by a separate gene elsewhere: i.e., the ATs that load the polyketide building blocks are not integral parts of the modules but act as free-standing units (10). This group of enzymes is particularly interesting because many of them are responsible for products made specifically by symbiotic bacteria (10). These gene clusters in the lichen are most likely derived from the photobiont, as only Nostoc exhibited a high level clonal presence, indicated by DNA sequence coverage in the WGS, and a commensurate level of coverage was found for diagnostic markers of the Nostoc genome, such as hgl (involved in heterocyst glycolipid biosynthesis; SI Appendix, Table S1). The longer of the two gene clusters in P. membranacea, designated “nsp” (Fig. 3) had significant homology to the gene clusters for the biosynthesis of pederin family compounds and therefore was selected for further investigation.

Fig. 3.

Nosperin biosynthetic gene cluster nsp and flanking regions. Microsynteny and homology with pederin and onnamide biosynthetic gene clusters are indicated in gray. Similarity of nsp to other PKS biosynthetic gene clusters is indicated by double-headed arrows. Numbers denote individual modules. Genes with similar proposed functions (Table 1) are indicated with identical colors. β, genes involved in β-branch formation; T, transposon. See SI Appendix, Figs. S15 and S16 for details of regions flanking the nsp locus.

The nsp gene cluster consists of a 59-kb region with 3 large genes (nspA, nspC, and nspD) that encode multidomain PKS or PKS/NRPS proteins and a suite of 10 smaller genes that encode accessory enzymes (Fig. 3; Table 1). The multidomain proteins together comprise a “starter” module 0, followed by nine PKS or PKS/NRPS elongating modules (modules 1–9). The 5′ end of the gene cluster, i.e., nspA (modules 0–3), nspB, and the beginning of nspC (module 4 and the KS region of module 5), as well as accessory genes at the 3′ end of the cluster, have closely related counterparts in biosynthetic gene clusters of pederin-type compounds (Fig. 3). The middle region, however, has primary affinities to NRPS–PKS biosynthetic pathways from other members of Proteobacteria or Cyanobacteria, viz., the end of nspC (modules 5–7) is similar to the PKS genes of the rhizoxin (rhi) biosynthetic gene clusters from Burkholderia sp. (12) and Pseudomonas sp. (13). Further downstream, the PKS genes have resemblances to gene clusters reported from various Nostocales or Oscillatoriales. An ∼3-kb region at the junction of nspC and nspD is especially intriguing in bearing ∼80% identity at the DNA level to a portion of the nos-like gene cluster (a cis-AT PKS pathway) in P. membranacea (SI Appendix, Fig. S2); a homolog of this cluster in Nostoc sp. GSV224 is responsible for biosynthesis of nostopeptolide (14), a cyclic peptide-polyketide (SI Appendix, Table S2). Altogether, the nsp locus appears to be an evolutionary mosaic of trans- and cis-AT PKS fragments from diverse sources.

Table 1.

List of the genes present in the nsp gene cluster and their predicted functions

ORF	Protein size	Proposed function	Closest homolog (protein,origin)	Percent identity	Accession number*
nspA	5,320	PKS	PedI, Paederus fuscipes symbiont	42	AAR19304
nspB	371	Flavin-dependent oxygenase	PedJ, P. fuscipes symbiont	66	AAR19305
nspC	8,252	PKS-NRPS	OnnI, Theonella swinhoei symbiont	49	AAV97877
nspD	2,206	PKS	JamP, Lyngbya majuscula	60	AAS98787
nspE	474	MatE efflux transporter	SxtM1, Lyngbya wollei	54	ACG63829
nspF	285	O-Methyltransferase	OnnH, T. swinhoei symbiont	57	AAV97876
nspG	86	ACP	Cpap_1683, Clostridium papyrosolvens DSM 2782	58	EGD47495
nspH	411	β-Ketoacyl synthase	Cpap_1682, C. papyrosolvens DSM 2782	60	EGD47494
nspI	420	HMG-CoA synthase	PksG, Bacillus subtilis subsp. subtilis SC-8	72	EHA29460
nspJ	262	Enoyl-CoA hydratase	Cpap_1678, C. papyrosolvens DSM 2782	52	EGD47490
nspK	442	Acyltransferase	PedD, P. fuscipes symbiont	49	AAS47563
nspL	464	Cytochrome P450	PPSIR1_33239, Plesiocystis pacifica SIR-1	35	EDM78481
nspM	647	Asparagine synthase	Acid_5610, Candidatus Solibacter usitatus Ellin6076	48	ABJ86557

^*Accession numbers are for the GenBank database.

Expression of the nsp pathway was detected by RNA-seq analysis in P. membranacea thalli freshly collected from the same location as the source material for the WGS. Consistent with expectations for a photobiont-specific gene cluster, nsp transcripts were observed in the main thallus tissue that contains both mycobiont and photobiont cells, but not in apothecia or rhizines, which are lichen structures that are derived only from the mycobiont (Fig. 1B; SI Appendix, Table S3). Although trans-AT PKS systems have been found in a wide range of bacteria (16, 17), none have been reported for cyanobacteria, which are otherwise rich sources of cis-AT PKSs (17, 18). These observations suggested the possibility of metabolic products that might be novel, not only from structural but also from ecological and evolutionary perspectives.

Prediction of the Compound Structure.

Detailed examination of the ketosynthase (KS) domains in PKS gene clusters using phylogenetic methods and comparisons of module architecture in pathways with similar products can often facilitate prediction of natural product structures generated by trans-AT PKSs (16, 19). When the Nsp KS sequences (KS1–KS9, referring to the module number in which the domain occurs) were aligned and compared with 494 homologs using KSs from cis-AT systems as an outgroup, the resulting clades were generally consistent with respect to KS functions (SI Appendix, Fig. S3–S5). For example, all KSs with known function in the same group as KS1 accept acetyl starters incorporated by domains of the GCN5-related N-acetyltransferase family (GNAT) (20). In this way, partial structures for the substrates of KS1–3, 5, 7, and 9 were predicted (Table 2). As expected from the earlier analyses, KS1–5 (nspA, nspC) were most similar to KSs of the pederin (21, 22) and/or onnamide (23) PKS, and a full domain analysis revealed virtually complete architectural identity with corresponding portions of the ped and onn PKS–NRPS clusters over the first six modules, ending with KS5. The region also included an NRPS (module 4a) that catalyzes the insertion of a glycine residue (Fig. 4). This observation indicated that a large part of the polyketide product would resemble pederin and onnamides. The remainder of the core structure was more difficult to predict, because two of the four KSs (KS6 and KS8) fell into clades consisting of KS⁰s, which are nonelongating KS variants that usually show little consistency between phylogeny and substrate structure (16). KS⁰6 was positioned in a small subclade containing homologs from the rhizoxin and bacillaene PKSs that are involved in shifting double bonds from the α,β- to the β,γ-position (24, 25). These KSs are found in modules harboring, in addition to the KS⁰ and the acyl carrier protein (ACP), a dehydratase (DH) domain postulated to catalyze double bond isomerization and characterized by a NSAF/YL instead of the usual DxxxQ/H motif involved in dehydration (26). The same elements are present in the nsp module encoding KS⁰6; moreover, KS7, encoded by the module immediately downstream, is highly similar to rhizoxin KS12, which accepts a substrate with a shifted double bond (25). Together with the upstream NRPS module, these features strongly suggested the presence of an enamide moiety, which is not present in pederin or onnamides. KS9, associated with KSs elongating chains of amino acid residues, consistent with its position C-terminal to a second NRPS module (8a), was the only other KS with predictable function. An analysis of residues lining the substrate pocket of the adenylation domain, known as the nonribosomal code (27, 28), returned a perfect match to proline-activating domains (SI Appendix, Table S5). In the absence of diagnostic downstream KS domains, the portions of the polyketide generated by modules 5, 7, and 9 were predicted using classical PKS colinearity rules (29), although they often apply poorly to trans-AT PKSs. These rules indicated the presence of two additional methyl groups and a hydroxyl function. The terminal elongation step was predicted to be catalyzed by module 9 in NspD, a cis-AT PKS module with an integrated AT domain. This step remained obscure, because the module architecture (KS-AT-KR-ER-ACP-TE) contrasts with the canonical order (KS-AT-DH-ER-KR-ACP-TE) and lacks a DH domain to provide the substrate for the subsequent enoyl reduction. This feature suggested either that the ER domain is nonfunctional, despite the presence of key amino acid residues, or that the DH activity is provided in trans.

Table 2.

Analysis of KS domains present in the Nsp PKSs

Domain	Closest characterized relative (substrate specificity)	Predicted specificity of KS clade	Moiety present in nosperin
KS1	pederin KS1 (acetyl starter)	Acetyl	Acetyl
KS2	onnamide KS2 (α-l-methyl + β-d-OH)	α-l-methyl + β-d-OH	α-l-methyl + β-d-OH (anti configured)
KS3	onnamide KS3 (β-exomethylene)	mostly β-exomethylene	β-exomethylene
KS4	onnamide KS4 (KS0)	KS0	KS0
KS5	pederin KS5 (amino acid)	amino acid	glycine
KS6	rhizoxin KS11 (KS0, double bond)	KS0, double bond	KS0, double bond
KS7	rhizoxin KS12 (shifted double bond)	shifted double bond	shifted double bond
KS8	bryostatin KS8 (KS0)	KS0	KS0
KS9	oxazolomycin KS9 (serine)	amino acid	proline

Fig. 4.

The nsp gene cluster, deduced architecture of the PKS proteins NspA, NspC, and NspD, and proposed biosynthesis of nosperin. GNAT, GCN5-related N-acetyltransferase family (20); KS, β-ketoacyl synthase; KR, ketoreductase; MT, C-methyltransferase; CR, crotonase superfamily (also known as enoyl-CoA hydratase) (30); KS⁰, nonelongating KS; C, nonribosomal peptide synthetase (NRPS) condensation domain; A, NRPS adenylation domain; DH, dehydratase; AT, acyltransferase; ER, enoyl reductase; TE, thioesterase; ?, unknown. Small black circles symbolize acyl and peptidyl carrier proteins. The positions of amplicons used for the nsp screening are shown with black boxes and roman numerals.

Further structural predictions were possible by comparison of the accessory and post-PKS nsp genes to known pathways. The genes nspGHIJK resembled those typically involved in the generation of polyketide β-branches (Table 1), indicating the presence of a pederin-type exomethylene bond (30). Because the closest relatives of nspB and nspF in the ped and onn clusters encode an oxygenase and a methyltransferase (31), responsible, respectively, for one oxygenation (at C7) and one methylation (at the C6 acetal oxygen) within the corresponding moiety of pederin, similar units in the nsp product were expected. The putative asparagine synthetase (NspM) and cytochrome P450 (NspL) enzyme homologs, however, remained without counterpart in pederin-type pathways.

Isolation and Characterization of the Polyketide Nosperin.

Using the preliminary structural information as a guide, total extracts of whole lichens were examined for the presence of the predicted metabolites. Due to copious amounts of diverse glycolipids and other metabolites, however, LC-MS and extensive NMR-guided subfractionation failed to detect a pederin-type polyketide. As Nostoc symbionts have often been cultured from lichens (5), an alternative approach focusing only on the cyanobacterium was taken. Macerated thalli of P. membrancacea were plated on BG-11₀, a minimal medium lacking nitrogen, and cyanobacteria identifiable as Nostoc sp. by microscopic examination were established in pure culture (Fig. 1D). The presence of the nsp cluster in three random isolates was confirmed by PCR for amplicons representing the PKS genes nspA and nspC, and the accessory gene, nspF (Fig. 4; SI Appendix, Table S7). One strain, designated N6, also characterized by sequencing of 16S and 23S ribosomal RNAs, was grown in BG-11 liquid medium for 4 wk to evaluate gene expression in culture. Transcription of the nsp gene cluster in Nostoc sp. N6 was confirmed by mapping RNA-seq data and found to be fivefold higher than in the thallus, relative to expression of rbcLXS, a Nostoc reference marker (SI Appendix, Table S3). When extracts were prepared from scaled up cultures, numerous metabolites were observed in small amounts. However, due to the unusual architecture of the terminal Nsp domains and the unknown nature of post-PKS modifications, prediction of the mass of the compound was challenging and convincing candidates were not identified by MS analysis.

In light of the challenges imposed by multicomponent trace mixtures and the absence of a known mass, a strategy of stable isotopic enrichment followed by HPLC-SPE-NMR to address problems of sensitivity and complexity while allowing detection of predicted structural moieties was used. Nostoc sp. was cultured in 25 L of BG-11 supplemented with ¹³C-labeled NaHCO₃, and after 5 wk, cyanobacterial biomass from 10 L of culture was freeze-dried and extracted. HPLC-electrospray ionization (ESI)-MS analysis confirmed that most components in the extracts had multiple ¹³C atoms incorporated into individual molecules. To obtain insights into structural features of these compounds, the crude extract was subjected to repetitive HPLC-solid phase extraction (SPE) purification with subsequent NMR analysis of target molecules eluted with fully deuterated solvent (SI Appendix, Figs. S6–S9). This method allowed collection of high-quality ¹H- and ¹³C-spectra, as well as COSY-, HSQC-, and HMBC-2D-NMR data from microgram amounts of the ¹³C-labeled material (SI Appendix, Figs. S8–S13). NMR signals characteristic of the predicted exomethylene, methoxy, and C-methyl functions were detected for a component eluting at 30.3 min (SI Appendix, Figs. S6–S9), of which only 30 μg were obtained. Further support for the identity of the compound came from ESI(+)-MS analysis, which indicated a multiply labeled minor component with an exact unlabeled mass of m/z = 564.2926, well within the predicted range and best fitting a calculated atomic composition of C₂₆H₄₃N₃O₉Na (calculated: m/z = 564.2897). MS/MS analysis of the molecular ion peak additionally revealed a daughter ion consistent with the formal loss of MeOH (m/z = 532.2664; calculated for C₂₅H₃₉N₃O₈Na: m/z = 532.2635), suggesting a methoxy group in the predicted structure. Several other indicative fragments were also visible in the MS/MS data, e.g., an additional cleavage of an acetamide functionality (m/z = 473.2260; calculated for C₂₃H₃₄N₂O₇Na: m/z = 473.2264). The NMR data fully supported the identity of the compound as a member of the pederin group and allowed elucidation of its constitution. In combination with bioinformatic analysis it was possible to predict most of the stereogenic elements, except the configuration at C12 and C14 (Fig. 5; for a full description of the MS and NMR-based characterization see SI Appendix, Data S2). The structure of the compound is in almost perfect agreement with the predicted features and represents a hybrid of pederin and an unusual proline-containing terminal moiety not previously observed in this group of natural products. Two deviations from the incomplete product prediction are the terminal amide function, most likely generated by the asparagine synthase-like protein NspM, and the hydroxyl moiety at C20, indicating post-PKS oxidation by the P450 homolog NspL. Altogether, the data show that this compound, which has been designated nosperin, represents a unique member of the pederin family of natural products.

Fig. 5.

Stereochemical characterization of nosperin by NMR and bioinformatic analysis. The absolute configurations at the chiral centers were predicted by analysis of the stereospecificity of KR domains (blue), the NRPS domain structure (orange), the overall domain organization in comparison with other pederin-type biosynthetic gene clusters (red), and NMR coupling constants and/or chemical shifts (green).

Distribution of the nsp Locus in Cyanobacteria.

Although the cultivated strain Nostoc sp. N6 carried the nsp cluster, it also exhibited a distinctive 23S rRNA polymorphism and originated from a specimen of P. membranacea independent from that used in metagenome sequencing. These observations suggested that the nsp pathway might be common in P. membranacea photobionts or in Nostoc and possibly even other cyanobacteria. A PCR-based survey for nspA, nspC, and nspF indicated that the nsp cluster was present in P. membranacea from several locations in Iceland, but samples of this lichen in British Columbia, Canada, also included specimens where the targeted sequences were not detected (Table 3). Three nsp amplicons obtained from a specimen near Vancouver on the mainland were nearly identical (99.9%) to those from a Vancouver Island sample, but showed ∼3% divergence from the Icelandic reference (SI Appendix, Fig. S14).

Table 3.

Detection of nsp amplicons in P. membranacea from Iceland and two locations in British Columbia, Canada

Region	Locality	Number of samples	nspA	nspC	nspF
Reykjavík	Grafarholt	1	+	+	+
	Keldur	3	+	+	+
	Mosfellsbaer	1	+	+	+
	Öskjuhlid	5	+	+	+
	Raudavatn	1	+	+	+
	Ulfarsfell	1	+	+	+
Vancouver (Mainland)	Belcarra	1	+	+	+
	Black Mountain	1	−	−	−
	Brothers Creek	1	+	+	+
	Eagle Ridge	2	+	+	+
	Eagle Ridge	3	−	−	−
Vancouver Island	Horth Hill	2	+	+	+
	Roche Cove	2	−	−	−

For positions of the amplicons see Fig. 4. +, a PCR product of the expected size was observed; −, no PCR product was observed.

A bioinformatic search for nsp-like sequences in GenBank (SI Appendix, Table S2) and a PCR survey for amplicons of nspA, nspC, and nspF in 26 cyanobacterial strains representing 15 genera from all cyanobacterial orders (Materials and Methods) did not return any positive results except in the genus Nostoc, suggesting that the nsp pathway per se may have a phylogenetically restricted distribution. Within Nostoc there appeared to be an association with lichens: results from PCR testing of four strains (Nostoc sp. PCC9709, AR10B, AR9A, and WL-1) isolated from Peltigera spp. (32, 33) were positive for nsp, whereas four other strains not associated with lichens (Nostoc muscorum PCC7906, Nostoc punctiforme PCC73102, Nostoc spp. PCC6705, and PCC7107) were negative.

The limited distribution within Cyanobacteria and the apparent absence of the nsp gene cluster in some samples from Canada suggested that rather than being a core part of the genome, the nsp genes may have been introduced horizontally. Investigation of regions flanking the nsp cluster revealed IS4 elements on both sides and linkage to genes associated with plasmid replication and partitioning (including parA, parB, and parM homologs and a gene encoding a DNA helicase), suggesting the possibility of an extrachromosomal source (Fig. 3; SI Appendix, Figs. S15 and S16).

Discussion

In this report, we describe identification of the nsp genes in the P. membranacea lichen metagenome, the first trans-AT PKS gene cluster from a cyanobacterium, and the application of a strategy consisting of bioinformatic prediction, symbiont cultivation, isotope enrichment, and ¹³C-NMR that enabled characterization of a unique symbiosis-associated natural product, nosperin, from the photobiont.

Lichens have long been known for distinctive mycobiont-produced compounds, such as depside and depsidone polyketides (34), but unique structures and pathways are now also emerging from studies of the photobionts. Two conventional cis-AT PKS–NRPS biosynthetic pathways have recently been described from cyanobacteria associated with lichens: the mcy gene cluster (35, 36) involved in synthesis of microcystins, notorious hepatotoxins typical of many cyanobacteria, and the crp gene cluster, responsible for production of cryptophycins (37), anticancer agents of more limited distribution. The elements of the nosperin biosynthetic pathway, in contrast, are similar to the less common trans-AT PKS–NRPS systems responsible for a number of animal–bacteria symbiosis-associated compounds including pederin, theopederins, onnamides, mycalamides, psymberin (irciniastatin A), and others (10) (Fig. 2). These compounds, with almost identical core regions but different biosynthetic starter regions and/or termini, are often highly toxic to eukaryotes and some have been considered promising candidates for anticancer drug development (38–40). Notably, this group of compounds has never been recovered from screening free-living bacteria, despite conspicuous pharmacological activities. Studies of mycalamide A (Fig. 2), which binds in the E-site of the ribosome normally occupied by the tRNA-terminal CCA (41), and synthetic analogs together with molecular modeling, have identified the N-acyl linked tetrahydropyran structure as central to binding and activity (42). The presence of the N-acyl linked tetrahydropyran in nosperin suggests it might have similar bioactivity; however, the amounts available were too small for testing.

The discovery of nosperin not only increases the number of chemical scaffolds and biosynthetic enzymes encompassed by the pederin group but also expands the remarkable range of symbioses associated with this natural product family (Fig. 2). Furthermore, although the taxonomic identities of their producers are unknown, with the exception of pederin, which is produced by a close relative of Pseudomonas aeruginosa (21, 43, 44), both lichen metagenomic data and expression and product characterization in Nostoc sp. N6 clearly show that nosperin derives from the cyanobacterial photobiont of P. membranacea. Although this cyanobacterium is essential to every phase of thallus growth and development, it is also a facultative symbiont, being culturable by itself on basic mineral salts media. This first individually identified and culturable producer of a pederin family natural product provides new opportunities to study the biochemistry and physiology of the biosynthetic pathway in vivo, as well as to improve metabolite yield through optimization of production protocols and strain improvement.

Study of genes, gene clusters, and biosynthetic pathways in diverse symbiotic associations may help clarify their functions or identify metabolic products that are essential. Some polyketides produced by trans-AT PKSs, such as pederin (21, 22, 43), bryostatin (45, 46), and rhizoxin (12, 47), are known to participate in host defense and pathogenicity in symbiotic associations. It has also been suggested that PKS–NRPS compounds such as the microcystins, sometimes produced by cyanobacterial symbionts, may contribute to the chemical defense of lichens against grazers (35, 36). Expression of the nsp genes in P. membranacea and their presence in all Icelandic specimens tested suggest that it is a beneficial trait, although its role is unclear. In this regard, it may be significant that no microcystin pathway homolog was identified in P. membranacea, leaving open the possibility that nosperin might have a similar function in the lichen. Examination of geographically distant populations of P. membranacea was informative, as absence of nsp amplicons from some samples from Canada indicates that these genes are not essential for the lichen symbiosis although nosperin may confer advantage under some conditions. This provision may also apply to pederin, onnamides, and psymberin, where the metazoan hosts can be found with or without the metabolites (23, 43, 48). In the case of P. membranacea, additional field studies may help elucidate whether there is a primary cause, e.g., a founder effect, a particular environmental condition, or an interplay of other factors that underlie the distributional differences observed.

The presence of similar trans-AT PKS–NRPS gene clusters in different groups of bacteria has suggested that these clusters are horizontally transferred (44). The flanking of the nsp cluster by transposable elements is consistent with this hypothesis, and the mosaic of homologies across the gene cluster suggests involvement of several intergenomic and intragenomic recombination events. The homology of NspE and part of NspD to proteins from Oscillatoriales (Fig. 3) suggests that an ancestral ped-like operon, specifying the conserved core part the molecule, may have been introduced into and modified by oscillatorean cyanobacteria: the position of nspE and nspD between sequences with high homology to the ped gene cluster and the orientation of nspE opposite to the ped-like genes suggest an intragenomic rearrangement mediated by genetic similarity of PKS–NRPS modules. Transfer to Nostocales and subsequent recombination resulted in the present domain organization that includes the ∼3-kb cis-AT containing fragment from a Nostoc nos-like cluster. The presence of a cis-AT domain is unusual in a PKS relying on trans-ATs, with few occurrences among the ∼40 large trans-AT PKS complexes with known products (10, 49). A relatively recent insertion of this ∼3-kb fragment is indicated by the high amino acid and nucleotide similarities to the nos cluster and suggests that near relatives of the nsp pathway may exist. It will be interesting whether a gene cluster similar to nsp, but without the cis-AT encoding region, or with other types of inserts and substitutions, will be found in other bacteria. Study of such examples of naturally engineered multidomain genes and gene clusters involving distantly related participants may not only generate useful hypotheses for further understanding their evolution, but the phylogenetic reconstruction may also be informative in identifying models of successful architectures for application in combinatorial biosynthesis industrially.

The metabolic options offered by symbiotic associations provide exciting potential for drug development and highlight the need for new discovery strategies applicable to these complex systems. Although individual steps of the present procedure have been used previously in natural product research (15, 19, 47, 50–52), the combination of methods has not been reported and should be applicable to many further organisms. The ¹³C-NMR–based technique and other recent methods such as imaging MS (53) can detect low concentration signatures of nosperin and facilitate investigation of its role in symbiosis. These approaches could also identify molecular variants: e.g., recently studied specimens of P. membranacea that appear negative for only one or two of the three primer sets used for nsp screening may present variants of nosperin. This possibility is akin to the situation of the microcystins and cryptophycins for which a large number of structural variants have been found (36, 37). A thorough study of the >1,500 species of cyanobacteria-bearing lichens and the multitude of other organisms including bryophytes, ferns, cycads, and angiosperms (54) that harbor cyanobacterial symbionts may yield many new biosynthetic pathways and metabolites to provide both alternative chemistry for potential pharmacological applications and a wealth of information on the chemical biology of symbiosis.

Materials and Methods

Identification of PKS Gene Clusters in the P. membranacea Metagenome and Expression Analysis of Whole Thalli.

Metagenomic DNA was processed for sequencing at commercial facilities via Roche 454 and Illumina Solexa 2 × 35-bp methodology generating 1.76 GB of 454 data and 1.4 GB of Illumina data, yielding ∼50× coverage of the Nostoc genome. A draft assembly of the P. membranacea metagenome was constructed with MIRA v3.2.1 (www.chevreux.org/projects_mira.html). To search for PKS gene clusters, concatenated consensus sequences of the KS (N terminus, pf00109; C terminus, pf02801; http://pfam.sanger.ac.uk/) and ACP domains (pf00698) were used in a TBLASTN search (55) to retrieve all relevant contigs from the metagenomic database. Accuracy of the assembly was verified by visual inspection of the contigs in GAP4 (Staden package) (56) based on a mapped 3.5-kb paired-end library. Portions of the nsp sequence were verified by PCR amplification and sequenced directly using BigDye chemistry (Applied Biosystems; MacroGen). RNA-seq data sets from field samples of lichen thalli, apothecia, and rhizines were previously generated (6) and used for mapping in this study with Bowtie (57).

Structure Prediction.

Amino acid sequences of 503 KS domains from trans-AT and cis-AT PKSs were retrieved from GenBank and aligned using the MUSCLE algorithm with a gap open score of −1, as implemented in Geneious 5.5.3 (Biomatters Ltd.). After manual improvement of the alignment, phylogenetic reconstruction was performed by means of the Geneious software using the neighbor joining algorithm with a Jukes-Cantor distance method. KS domains of cis-AT PKSs were used as an outgroup. Bootstrap analysis was done with 1,000 pseudoreplicate sequences.

Chemical Analysis of Whole Lichen.

Air-dried lichen (30 g) was ground to a fine powder in liquid nitrogen using a mortar and pestle and stirred for 24 h at room temperature in MeOH. The mixture was filtered, and the solid material was extracted a second time. The solvent of the combined MeOH extracts was removed under reduced pressure. The crude extract was partitioned between 10:1 MeOH/H₂O (300 mL) and n-hexane (3 × 100 mL). The solvent was removed from the aqueous MeOH layer under reduced pressure, and the residue was further fractionated by silica gel column chromatography. The following solvents (0.5 L each) were used to elute compounds: petroleum ether, petroleum ether/EtOAc (1:1), EtOAc, EtOAc/MeOH (9:1; 8:2; 7:3, 1:1), and MeOH. The fractions were evaporated under reduced pressure and analyzed by LC-MS using a Phenomenex Luna C18 column with a mobile phase gradient of 1:9 CH₃CN/H₂O + 0.1% TFA to 100% acetonitrile over 30 min and a flow rate of 1 mL/min.

Isolation of Nostoc sp. N6.

Lichen thalli collected from the same location as material used for WGS and expression studies were macerated between sterile microscope slides (58), and cells were plated on BG-11₀ (without NaNO₃) agar medium (59) and incubated at 20 °C with a 12/12-h day/night cycle. Nostoc colonies were purified by repeated streaking on the same medium and maintained at room temperature. Analysis of the 16S and 23S rRNA sequences in the RNA-seq library (below) confirmed both the purity of the culture and its identification as a Nostoc sp.

RNA Extraction and RT-PCR of Nostoc sp. N6.

Total RNA was isolated from 1 L of BG-11 medium incubated at 20 °C under constant illumination for 4 wk. Cyanobacteria were retained on Miracloth (Calbiochem) after culture filtration, rinsed with water, blotted with paper towels, flash frozen in liquid nitrogen, and crushed to a fine powder. TRIzol reagent (Life Technologies) was added to the powder, and it was ground again. The mixture was transferred to a 15-mL polypropylene tube and processed according to the TRIzol protocol. Before RT-PCR, the RNA was treated with DNase I (RNase-free) (Fermentas) to remove residual genomic DNA. First-strand cDNA was synthesized from 1 μg total RNA using SuperScript II Reverse Transcriptase (Invitrogen). RNA-seq data were obtained using Illumina Solexa Genome Analyzer IIx at the deCODE Genetics facility (Reykjavik, Iceland). RNA-seq mapping was done with Bowtie (57) and visualized in Geneious 5.5.3.

Culture of Nostoc sp. N6 for Natural Product Analysis.

Twenty-five liters of cells were grown in an illuminated (5,200 lm) bubble-column bioreactor in BG-11 liquid medium, optionally enriched with 3 mM ¹³C-labeled NaHCO₃, for 5 wk at pH 7.8 and 25 °C. The cyanobacteria from 10- and 5-L portions of the culture were collected by filtration, frozen in liquid nitrogen, freeze-dried, and stored at −20 °C.

Chemical Extractions and Analysis of Nostoc sp. N6, Unlabeled Culture.

Freeze-dried cyanobacteria (above) were homogenized in 50 mL CH₂Cl₂/MeOH (2:1) and stirred for 15 min at room temperature. Biomass was filtered and treated again with the same amount of CH₂Cl₂/MeOH for 30 min at 30 °C. This procedure was repeated twice. The combined extracts were dried under reduced pressure. The crude extract was dissolved in MeOH and subjected to LC-MS analysis using an Agilent 1200 series HPLC and Bruker Daltonics micrOTOF-Q-spectrometer. HPLC was carried out with a Phenomenex Luna C18 column (5 μm, 250 × 2.00 mm), a mobile phase gradient of CH₃CN/H₂O (20:80) to (80:20) over 45 min, and a flow rate of 1 mL/min.

Chemical Extractions and Analysis of Nostoc sp. N6, Labeled Culture.

Freeze-dried cyanobacteria were extracted with stirring for 24 h in 2 L MeOH at room temperature. After filtration, the methanolic fraction was dried by evaporation and redissolved in 0.5 L MeOH/H₂O (10:1) followed by liquid-liquid extraction with 0.5 L cyclohexane. The cyclohexane fraction was discarded. The remaining MeOH/H₂O fraction was dried and stored at −20 °C. This material was directly used for LC-SPE-NMR analyses.

HPLC-SPE-NMR.

The solvent system consisted of eluent A (H₂O + 0.1% deuterated formic acid) and eluent B (acetonitrile) with a linear gradient starting with 10% of B up to 90% B in 30 min. The flow rate was 0.8 mL/min at 25 °C, and the injection volume was 50 μL. The chromatography was monitored at 210, 220, and 254 nm, and these wavelengths were used to define absorbance thresholds to trigger SPE trapping. The HPLC eluate was diluted with H₂O (2.4 mL/min) before trapping on SPE cartridges (Spark Holland), and individual peaks were trapped four times to increase concentration on cartridge. The cartridges were dried with pressurized nitrogen gas for 30 min each, and the analytes were eluted with 190 μL CD₃CN (99.8 atom %; Deutero GmbH) into 3-mm match tubes from Bruker BioSpin GmbH.

NMR.

All NMR experiments were acquired on an AVANCE III 600 MHz NMR spectrometer equipped with a 5-mm QNP cryo probe head (Bruker Biospin). Standard parameter sets created for the Bruker SELU (structure elucidation) program were uniformly used. Gradient correlation spectroscopy (COSY) and heteronuclear multiple-bond correlation spectroscopy (HMBC) were carried out using 4,000 complex data points in F2 and 512 points in the F1 dimension. The multiplicity edited gradient heteronuclear single quantum correlation (HSQC) was acquired with 2,000 data points in F2 and 400 points in the F1 dimension. The COSY experiment was acquired with 32 scans, the HSQC with 64 scans, and the HMBC with 128 scans per increment, resulting in experiment times of 8 h 46 min (COSY), 12 h 4 min (HSQC), and 1 d 11 h (HMBC). A C13 spectrum with composite pulse decoupling on the proton channel was acquired by collecting 4,096 scans with 131,072 complex data points at a sweep width of 40,761 Hz and with a relaxation delay of 5 s. The experiment time was 7 h 11 min.

Distribution Survey.

P. membranacea thalli were collected at several localities in Iceland (Reykjavik area) and in British Columbia (North Shore mountains near Vancouver; Vancouver Island), and DNA was extracted using the previously described methods (60). DNA samples representing cyanobacterial strains other than those newly isolated from P. membranacea for this study were prepared and described previously (32, 61) and stored at −20 °C. They include Anabaena sphaerica UTEX1616, Chlorogloeopsis fritschii PCC6718, Cylindrospermum stagnale PCC7417, Fischerella muscicola PCC7414, Geitlerinema sp. PCC7105, Gloeobacter violaceus PCC7421, Leptolyngbya sp. PCC7104, Leptolyngbya sp. PCC7375, Lyngbya kuetzingii UTEX1547, Myxosarcina sp. PCC7325, Nodularia spumigena PCC73104, Nodularia harveyana UTEX2093, Pleurocapsa sp. PCC7315, Pleurocapsa sp. PCC7324, Pleurocapsa sp. PCC7321, Scytonema hofmanni PCC7110, Synechocystis PCC6803, Nostoc punctiforme PCC73102A, Nostoc sp. PCC6705, Nostoc sp. PCC9709, Nostoc sp. AR10B, and Nostoc sp. AR9A. In addition, DNA samples from Calothrix sp. PCC7601, Nostoc muscorum PCC7906, and Nostoc sp. PCC7107 (originally obtained from the Pasteur Culture Collection of Cyanobacteria) and Nostoc sp. WL-1 (kindly provided by E. Loos, University of Regensburg, Regensburg, Germany) were prepared from cultures using similar methods (32). Amplification of rbcLX (62) or rnpB (63) regions (SI Appendix, Table S7) (and in some cases, also the 16S rRNA gene) was used as a positive control to ensure DNA quality before screening with primer sets targeting the nsp gene cluster (Fig. 4). DNA from the Nostoc sp. N6 strain was used as an nsp positive control. Conditions for nsp screening were 94 °C for 2 min, then 94 °C for 10 s, 55 °C for 30 s, and 72 °C for 30 s (35×), and then 72 °C for 7 min. For rbcLX primers, the extension time was 1 min. Eppendorf MasterMix 2.5× (Eppendorf) was used according to the manufacturer’s protocol in a final volume of 50 μL. All negative samples were repeated at least once. PCR amplicons from two samples were sequenced directly (MacroGen). The bioinformatic search was conducted in November 2012.