Structural Elucidation of Cryptic Algaecides in Marine Algal-Bacterial Symbioses by NMR Spectroscopy and MicroED

Abstract

Microbial secondary metabolite discovery is often conducted in pure monocultures. In a natural setting, however, where metabolites are constantly exchanged, bio-synthetic precursors are likely provided by symbionts or hosts. In the current work, we report eight novel and architecturally unusual secondary metabolites synthesized by the bacterial symbiont Phaeobacter inhibens from precursors that, in a native context, would be provided by their algal hosts. Three of these were produced at low titres and their structures were determined de novo using the emerging microcrystal electron diffraction method. Some of the new metabolites exhibited potent algaecidal activity suggesting that the bacterial symbiont can convert algal precursors, tryptophan and sinapic acid, into complex cytotoxins. Our results have important implications for the parasitic phase of algal-bacterial symbiotic interactions.

Microalgal–bacterial symbioses have emerged as a model system for exploring the structures and functions of secondary metabolites, especially those that are synthesized in response to intercellular signals.^[1–3] We have focused on the interaction between the haptophyte Emiliania huxleyi, which forms massive seasonal blooms in the oceans, and the α-proteobacterium Phaeobacter inhibens, a member of the wide-spread and metabolically versatile roseobacter group bacteria. The interaction in laboratory culture is biphasic and consists of a mutualistically beneficial phase, during which the algae secrete food molecules such as amino acids and dimethylsulfoniopropionate (DMSP), and the roseobacter produce vitamins, growth hormones, and the broad-spectrum antibiotic tropodithietic acid (TDA) to promote and protect algal growth (Figure 1).^[4–12] However, the interaction changes when the algae release phenylpropanoids, such as p-coumaric acid or sinapic acid (SA). Under these conditions, the bacteria biosynthesize the algaecidal roseobacticides and roseochelin as well as the cryptic side-rophore roseobactin, which lyse algal cells and help secure any iron released by the host (Figure 1).^[4,13–16] A mutualist-to-parasite switch has also been detected in other algal-bacterial symbioses in laboratory co-culture, suggesting it is a common interaction paradigm between these partners.^{[10,11,17–19]} The studies thus far also suggest that roseobacticides are not the only bacterial toxins and that other, yet unknown ones, may be at play. In the current work, we shed light on this aspect and report algaecidal molecules with unprecedented architectures, which we characterized structurally using, among other techniques, the emerging microcrystal electron diffraction method (MicroED).^[20–22]

Figure 1.

Working model for microalgal–bacterial symbioses. In the mutualistic phase (green arrows), the microalgae secrete DMSP, amino acids, and sugars, and provide an attachment surface. The roseobacter use some of these precursors to generate the growth hormones phenylacetic acid (PAA) and indole-3-acetic acid (IAA) as well as the protectant TDA. The interaction turns parasitic (red arrows) when the algae release phenylpropanoids, such as sinapic acid, which stimulates production of the algal toxins roseobacticide B and roseochelin B. Other algaecides have been proposed to exist as well.

Aside from providing an attachment surface and supplying the bacteria with DMSP (Figure 1), E. huxleyi and other microalgal hosts also release aromatic amino acids. Tryptophan (Trp) is thought to play an especially important role in the mutualistic phase as it creates a positive feedback loop wherein algal Trp is converted to indoleacetic acid (IAA) by the bacteria, which is then secreted to promote algal growth.^[10,23] Hypothesizing that Trp may also be involved in the parasitic phase, we exposed P. inhibens to sinapic acid (1 mM) in the presence of Trp (0.6 mM) and examined the resulting secondary metabolome. These concentrations were chosen based on prior work.^[4,10,13] The bulk concentration of phenylpropanoids and Trp are in the 20–100 μM range in laboratory cultures.^[4,13,24,25] However, because the bacteria form a physical association with the microalgae, the local concentration they sense is believed to be considerably higher.^[10] In addition to roseobacticides B and C,^[13] we found several new pigmented natural products, notably with m/z 335, 349, 485, 691, 788, and 802 (see Table S1 for high-resolution mass spectral data; Figure S1), which were not observed with Trp alone and detected only minimally with sinapic acid alone (Figure 2A). These compounds did not match reported ones in natural product databases. As P. inhibens can use algal Trp as a precursor for IAA synthesis,^[10] we repeated the experiment with isotopically labelled ring-²H₅-L-Trp and sinapic acid. The newly identified metabolites revealed incorporation of the isotopologue with m/z shifts of +5, +8, and +13, suggesting that Trp and/or fragments of Trp may be incorporated into the resulting natural products (Figure 2B). We focused on these “hybrid” molecules, which in a natural setting could be synthesized from algal and bacterial precursors.^[3] They were purified from large-scale cultures and then assessed for their structures and biological properties.

Figure 2.

Induction of cryptic bacterial metabolites upon treatment with Trp and sinapic acid. A) HPLC-MS analysis of P. inhibens cultures grown in the presence of Trp, SA, or both. The m/z of new molecules are marked. Roseobacticide B and C are labelled. B) HPLC-MS analysis of P. inhibens cultures grown in the presence of isotopically labelled ring-²H₅-L-Trp and SA. The insets show the mass shifts observed (in red) for each compound.

The structures of compounds with m/z 335 and 349, which we have termed sinacidin A and B, respectively, were successfully elucidated using multi-dimensional NMR spectroscopy (Figures S2, S3, Table S2). ¹H and HMBC spectra revealed an indole spin system in both variants that was connected to a central 3-pyrrolin-2-one substituted with 3-methoxy-5-hydroxy-para-quinone methide (sinacidin A) or 3,5-dimethoxy-para-quinone methide (sinacidin B) at the C-5 of the substructure (Figure 3A). The ¹H/¹³C chemical shifts and the UV/Vis spectra were consistent with the quinone oxidation state (Figures S1, S2, Table S2). Sinacidins are new secondary metabolites notable for the uncommon central 3-pyrrolin-2-one moiety, which is shared with the natural product violacein first isolated from Chromobacterium violaceum.^[26] The indole substituent provided an explanation for the +5 Da shift, when the bacteria were cultured with ring-²H₅-L-Trp. The results suggest that P. inhibens can take up Trp and sinapic acid—both presumably provided by the host in a natural setting—and generate the sinacidins through a series of complex reactions, some of which may mimic those in the violacein pathway.

Figure 3.

Structural elucidation of metabolites induced by Trp and SA. A) Structures of eight new metabolites from P. inhibens. NMR correlations from 1D/2D NMR spectra or electron density maps from MicroED for sinatryptin B and sinamicin C are shown (top) along with all structures (bottom). B) Experimental and calculated CD spectra to determine the C-14 absolute configuration in sinatryptin A. C) Approach used for MicroED to solve structures of sinatryptin B and sinamicin B and C. Microcrystals (shown for sinatryptin B; red bar, 5 μm) are grown by slow evaporation on the grid, then subjected to an electron beam, and the diffraction images (shown for sinatryptin B) used for structure elucidation via direct methods. Structures of sinamicin B and C were determined in a similar manner.

The induced metabolites with m/z 485 and 691 showed a +8 Da shift in the presence of ring-²H₅-L-Trp. Upon purification of m/z 485, we found two distinct chromatographic peaks with identical HR-MS features. We have assigned these compounds sinatryptin A (major m/z 485 component), B (minor m/z 485 component), and C (m/z 691). The structure of sinatryptin A was determined using 1D/2D NMR spectroscopy. The observed chemical shifts, COSY, and HMBC correlations were consistent with the presence of the 3-indolyl-3-pyrrolin-2-one substructure (Figures 3A, S4, Table S3). The quaternary C-14 was substituted with 3,5-dimethoxyphenol in its reduced phenolic form (rather than the quinone) and with the carboxylic acid oxygen of 3-hydroxybenzoic acid. The latter moiety also formed a C21–N10 bond. These data completed the planar structure of sinatryptin A leaving the absolute configuration of C-14 to be assigned. Repeated attempts to crystallize sinatryptin A for analysis by X-ray crystallography or MicroED, a recent method in which structures can be obtained by analysis of diffraction patterns upon exposure of microcrystals to a low-dose electron beam,^[20–22,27] failed. We therefore resorted to electronic circular dichroism (eCD) calculations and experiment,^[28–30] a method that has recently emerged for configurational assignment of chiral centres. The experimental CD spectrum matched the spectrum calculated for 14S-sinatryptin A, but not the 14R-epimer (Figure 3B, Tables S4, S5), thereby completing structural elucidation of this congener.

The isolated yields of sinatryptin B, 0.2 mg from 16 L, proved too low for spectroscopic analysis or growth of suitably sized crystals for X-ray crystallography. We again turned to the MicroED method and, in contrast to attempts with sinatryptin A, obtained needle-like crystals (Figure 3C). Merging two different datasets allowed structure elucidation to 0.95 Å resolution (79% data completeness, Table S6, Figure S5), revealing sinatryptin B as the quinone derivative of congener A in which the benzoic acid substituent is only connected at N-10; it is perpendicular to the plane generated by the central 3-pyrrolin-2-one core and the indole/3,5-dimethoxy-para-quinone methide substituents (Figure 3A, C). Importantly, sinatryptin B is not a degradation product as both metabolites appeared simultaneously, and as extended incubation of sinatryptin A did not result in variant B. Sinatryptin B is only the second new natural product that has been structurally elucidated de novo using MicroED.^[22]

NMR analysis of sinatryptin C revealed a central molecule that was similar to sinatryptin A and the sinacidins. In this case, a separate spin system, which NMR data indicated to be sinapic acid, was detected. From chemical shift and 2D NMR analysis, we inferred that sinapic acid esterified (Figure S6, Table S3), thereby completing the structure of sinatryptin C. Sinatryptins contain a 3-hydroxybenzoic acid component that we suspected may be derived from Trp and explain the +8 Da shift that we observed in the isotope labelling studies (Figure 2B). To test its origin, we isolated sinatryptin A from large-scale cultures grown in the presence of sinapic acid and ring-²H₅-L-Trp. Using NMR analysis, we observed suppression of the benzoic acid protons and the indole protons, whereas other protons in the molecule were not substituted, thereby confirming that the benzoic acid unit is indeed derived from Trp (Figures S7, S8). How Trp is transformed into this moiety remains to be shown. Homologs of enzymes that could be involved, indoleamine-2,3-dioxygenase and kynurenase, can be identified in the genome of P. inhibens, though these are not clustered in the P. inhibens chromosome.

Finally, we turned our attention to the largest compounds with m/z 788 and 802 consisting of two congeners; we have termed these sinamicin A (major m/z 802 component), B (minor m/z 802 component) and C (m/z 788). The structure of sinamicin A was completed using 1D/2D NMR spectroscopic analysis (Figure 3A, Figure S9, Table S7), which showed it to consist of a heterodimer of sinatryptin A and a variant of sinacidin B in which N-10 is replaced with carbon, with the fusion occurring via an unusual spiro 3-pyrroline linkage. Sinamicin A contains a single chiral centre and CD spectral analysis indicated the sample to be racemic (Figure S10). The presence of two indole substituents and the 3-hydroxybenzoic acid group, which is derived from Trp (Figures S7, S8), explained the observation of a +13 Da shift, when sinamicins are produced in the presence of ring-²H₅-L-Trp. Sinamicin B and C were difficult to separate and were produced at miniscule quantities (≈0.2 mg combined), which together with the low H/C ratio, precluded structure determination by NMR spectroscopy. We were able to grow microcrystals and solve structures of both congeners using MicroED to 0.98 Å resolution upon merging four datasets (Figure 3A, Table S6). Sinamicin B is the ring-opened form of congener A, in which the C14–O bond is no longer present (Figure 3A, C). Sinamicin C is the C-17 hydroxy form of sinamicin B. Interestingly, the two variants crystallized together, and we could see alternate stacking of sinamicin B and C in the asymmetric unit (Figure S11). In this case, MicroED allowed us to solve the structures of two mass-limited natural products in one experiment. Characterization of sinamicins completed structural elucidation of the eight novel alkaloids that are produced by P. inhibens in the presence of sinapic acid and Trp.

The production of “cryptic” metabolites^[31,32] in the parasitic phase suggests they may be algaecidal or provide another competitive advantage for P. inhibens during this phase of the interaction. We conducted broad antimicrobial and antialgal assays with the three most abundant compounds, the sinacidins and sinatryptin A, in an attempt to explore their biological properties (Table S8). In assays against a panel of 27 microbial strains consisting of fungi, marine bacteria, and common human Gram-positive and Gram-negative pathogens, the only noteworthy activity observed was that of sinatryptin A against Neisseria gonorrhoeae, the causative agent of gonorrhoea, with a minimal inhibitory concentration of 8 μgmL⁻¹ (16.5 μM). In E. huxleyi antialgal assays, however, four of the six compounds exhibited cytotoxicity with half-maximal inhibitory concentrations (IC₅₀) in the 5.7–8.2 μM range (Figure 4). Images before and after treatment of E. huxleyi with sinatryptin A show clear cell lysis after only short treatment periods (Figure 4A). As the sinacidins and sinatryptin A are produced at bulk concentrations (≈0.5 mgL⁻¹, ≈1–2 μM) that are analogous to these IC₅₀ values (Figure 4B), they may provide the sought-after algal toxin that have been suggested based on co-culture experiments.^[10] Additional studies are necessary to evaluate the physiological relevance of these compounds in co-culture, mesocosm, or field experiments.

Figure 4.

Antialgal effect of sinatryptin A on E. huxleyi CCMP374. A) Shown are light micrographs of E. huxleyi exposed to DMSO control or to 100 μM sinatryptin A after 6 h, 12 h, and 24 h. Scale bar, 5 μm. B) IC₅₀ values of compounds identified in this work against E. huxleyi. The average of two independent measurements is shown.

The present work contributes two key insights. In the microalgal–roseobacter association and in other algal–bacterial symbioses, Trp provides the basis of a positive feedback loop in the mutualistic phase, where it is converted by the bacteria into the algal growth hormone IAA (Figure 5).^[10,23] We show that a negative feedback loop exists in the parasitic phase, in which bacteria can utilize algal Trp to synthesize molecules that ultimately lead to the demise of the host. Thus, the mode of interaction dictates how algal Trp is utilized by the bacteria. All three compound groups (Figure 3A) also incorporate fragments of sinapic acid, as can be seen in the 3,5-dimethoxyphenol group, suggesting that these toxins are produced largely from algal precursors. Given that the symbiosis occurs in the nutrient-poor pelagic zone of the ocean, the bacterial biosynthesis of algaecides largely from algal precursors provides a remarkable example of metabolic economy. This aspect, derived in our laboratory setup, can now be examined in ecology-based studies. Our work also provides one of the first examples of de novo natural product structure elucidation by MicroED and simultaneously highlights the method’s advantages and limitations. A clear advantage is that small crystals, not suitable for X-ray methods, can be analysed by MicroED. Miniscule amounts of a natural product, therefore, are sufficient for complete structure determination and mixtures can be subjected to analysis, as seen with sinatryptin B and sinamicin B/C. However, the method still relies on well-diffracting crystals, which can be difficult to come by, as seen with sinatryptin A. Moreover, instrument access is not routine and instrument time remains costly. Here NMR spectroscopy offers an advantage. Presently, MicroED provides a complementary approach to existing methods for molecules that readily crystallize. Additional advances that are sure to come could make it competitive with multi-dimensional NMR spectroscopy for de novo structure elucidation of natural products.

Figure 5.

Model of algal–bacterial symbiosis focusing on the role of Trp. In the mutualistic phase, a Trp positive feedback loop results in conversion of algal Trp into the growth hormone IAA, as previous studies show. In the parasitic phase, a Trp negative feedback loop operates, wherein the amino acid serves as a building block for production of complex algal toxins that lead the host’s demise. Other algaecides that incorporate Trp or fragments thereof are not shown.