Coral larval settlement preferences linked to crustose coralline algae with distinct chemical and microbial signatures

The resilience of coral reefs is dependent on the ability of corals to settle after disturbances. While crustose coralline algae (CCA) are considered important substrates for coral settlement, it remains unclear whether coral larvae respond to CCA metabolites and microbial cues when selecting sites for attachment and metamorphosis. This study tested the settlement preferences of an abundant coral species (Acropora cytherea) against six different CCA species from three habitats (exposed, subcryptic and cryptic), and compared these preferences with the metabolome and microbiome characterizing the CCA. While all CCA species induced settlement, only one species (Titanoderma prototypum) significantly promoted settlement on the CCA surface, rather than on nearby dead coral or plastic surfaces. This species had a very distinct bacterial community and metabolomic fingerprint. Furthermore, coral settlement rates and the CCA microbiome and metabolome were specific to the CCA preferred habitat, suggesting that microbes and/or chemicals serve as environmental indicators for coral larvae. Several amplicon sequence variants and two lipid classes—glycoglycerolipids and betaine lipids—present in T. prototypum were identified as potential omic cues influencing coral settlement. These results support that the distinct microbiome and metabolome of T. prototypum may promote the settlement and attachment of coral larvae.

control samples and methanol blanks to minimize any potential bias due to instrumental drift. Molecular formula and fragmentation spectra were confirmed by analysis on a QEXactive plus Orbitrap mass spectrometer (ThermoScientific, MA, USA). The chromatographic separation was carried out as above. The mass spectrometer analyzer parameters were set as follows: sheath gas flow rate: 35 a.u (arbitrary units), aux gas flow rate: 10 a.u, sweep gas flow rate: 0 a.u, capillary All data files were converted to mzXML files with MSConvert (ProteoWizard 3.0). Preprocessing, normalization and quality checks were performed using Workflow4metabolomics version 3.3 [1]. Processed data were analysed with MetaboAnalyst 3.0 after having been Pareto scaled (i.e., mean-centered and divided by the square root of the standard deviation of each variable) to provide equivalent weight among variables [2]. Species profiles were compared using principle components analysis (PCA). PERMANOVA followed by pairwise comparisons was run to reveal differences in metabolomics fingerprints between CCA species. Shannon index and number of ions were analysed using one way ANOVAs with CCA species as fixed factor, followed by Tukey posthoc tests. To compare metabolomic richness, ions were classified as 'union' if present in at least one sample of a given species and classified as 'core' if present in all samples of that species. Union and core ions were classified as 'unique' to a species if they were absent in all samples outside that species.
Partial least square discriminant analysis (PLS-DA) was used to find the metabolites that contributed most to the discrimination of T. prototypum. To increase the discriminative power of the model, the number of groups was reduced using habitat instead of CCA species as factor.
Significance of PLS-DA model was assessed with permutation tests (consisting of 1000 permutations) and leave one out cross-validation (LOOCV). Robustness of PLS-DA model was validated by calculating Q2. Variable Importance in Projection (VIP) was used to summarize the importance of each variable (i.e., metabolite) in driving the separation among habitats. Using the exact mass, the molecular formulas were estimated and putative identifications were assessed for the VIPs which were present at significantly higher concentrations in T. prototypum (i.e., the cryptic group) relative to the subcryptic and exposed groups. Identifications were strengthened by fragmentation spectra issued from MS/MS analyses and according to literature and databases, including m/z cloud, KEGG, LIPID MAPS, Metlin and MarinLit.

Supplementary Methods 2, DNA extraction, amplicon sequencing and sequence analyses:
Fragments from the same four specimens used for metabolome analysis were processed for microbiome analysis (n = 4 replicates). DNA extraction followed the protocol of Meistertzheim et al. [3]. Briefly, this protocol starts with a mechanical lysis using a FastPrep Instrument (MP Biomedical, CA, USA) with Y Matrix tubes, followed by a chemical lysis phase by incubation Sequences were analysed using the standard Dada2 pipeline in R [4]. The R1 and R2 reads were filtered, trimmed, and merged to create an amplicon sequence variant (ASV) table.
This is a higher resolution analogue to the traditional OTU table, which records the number of times each exact amplicon sequence variant is observed in each sample. Chimeras were removed and taxonomy was assigned using the SILVA v132 database [5]. The taxonomic affiliation of ASVs of interest was further verified against sequences from the NCBI database using BLAST.
Sequences that belonged to algal chloroplast and mitochondria were removed.
Sequence data were analysed using the R package vegan after Hellinger transformation [6] and using the STAMP software [7]. Alpha diversity was calculated at the ASV level using the Shannon diversity index and analysed using one way ANOVA with CCA species as fixed factor, followed by TukeyHSD posthoc tests with Bonferroni correction. A non-metric multidimensional scaling ordination (NMDS), based on the Bray-Curtis similarity, was used to visualize community composition between species. PERMANOVA followed by pairwise comparisons was run to reveal differences in bacterial community composition between CCA species at the ASV level. Similarity percentage analysis (SIMPER) was used to determine the ASVs that contributed most to the dissimilarity between CCA species.   Closest order match is based on percent similarity in the SILVA database. Closest match source is based on BLAST search of the NCBI database.