Logo: to the web site of Uppsala University

Marine sponges (phylum Porifera) are leading organisms for the discovery of bioactive compounds from nature. Their often rich and species-specific microbiota is hypothesised to be producing many of these compounds. Yet, environmental influences on the sponge-associated microbiota and bioactive compound production remain elusive. Here, we investigated the changes of microbiota and metabolomes in sponges along a depth range of 1232 m. Using 16S rRNA gene amplicon sequencing and untargeted metabolomics, we assessed prokaryotic and chemical diversities in three deep-sea sponge species: Geodia barretti, Stryphnus fortis, and Weberella bursa. Both prokaryotic communities and metabolome varied significantly with depth, which we hypothesized to be the effect of different water masses. Up to 35.5 % of microbial ASVs (amplicon sequence variants) showed significant changes with depth while phylum-level composition of host microbiome remained unchanged. The metabolome varied with depth, with relative quantities of known bioactive compounds increasing or decreasing strongly. Other metabolites varying with depth were compatible solutes regulating osmolarity of the cells. Correlations between prokaryotic community and the bioactive compounds in G. barretti suggested members of Acidobacteria, Proteobacteria, Chloroflexi, or an unclassified prokaryote as potential producers.

Natural product discovery by isolation and structure elucidation is a laborious task often requiring ample quantities of biological starting material and frequently resulting in the rediscovery of previously known compounds. However, peptides are a compound class amenable to an alternative genomic, transcriptomic, and in silico discovery route by similarity searches of known peptide sequences against sequencing data. Based on the sequences of barrettides A and B, we identified five new barrettide sequences (barrettides C-G) predicted from the North Atlantic deep-sea demosponge Geodia barretti (Geodiidae). We synthesized, folded, and investigated one of the newly described barrettides, barrettide C (NVVPCFCVEDETSGAKTCIPDNCDASRGTNP, disulfide connectivity I-IV, II-III). Co-elution experiments of synthetic and sponge-derived barrettide C confirmed its native conformation. NMR spectroscopy and the anti-biofouling activity on larval settlement of the bay barnacle Amphibalanus improvisus (IC₅₀ 0.64 μM) show that barrettide C is highly similar to barrettides A and B in both structure and function. Several lines of evidence suggest that barrettides are produced by the sponge itself and not one of its microbial symbionts.

Reduced representation sequencing appraches such as ddRADseq allow to assess population connectivity and infer population summary statistics, both with and without a reference genome. However, as ddRADseq employs total DNA indiscriminate of the origin, the method warrants validation prior to application in microbial rich systems. One example of a complex system are sponges such as the North Atlantic high microbial abundance sponge Geodia barretti. This species is known to maintain large, putatively disjoint populations across the deep-sea, but its dispersal capabilities remain unclear as larvae have never been observed. To study the effect of microbial contamination on data processing and population genetic inference in ddRADseq, we produced a reference genome of G. barretti and collected 163 individuals across its habitat range and bathymetry (35–1560 m) in the North Atlantic. We processed the data with Stacks2 both with and without a reference genome (de novo and hybrid/‘reference-integrated’ approach). We found that strong population structures are recovered by both approaches and across different population genetic analyses (fastStructure, PCA, F_ST). Compared to previous work using microsatellites in shallow populations, we found only very weak population structure across large geographic stretches (>1000 km). However, over a third  (34%) of the final loci produced by the de novo pipeline did not map to the reference genome indicating that these might be of microbial origin. For comparably complex systems this means that de novo RRS genotyping approaches may contain a considerable amount of off-target loci potentially biasing the results.