All collections were conducted in October 2020 from the back reef (1-2 m depth) and fore reef (5-10 m depth) in Moorea, between LTER sites 1 and 2 of the Moorea Coral Reef (MCR) Long Term Ecological Research (LTER) site. Obligate corallivores were defined as fish that eat corals nearly exclusively (>90% of stomach content or number of bites; Harmelin-Vivien and Bouchon-Navaro, 1983; Rotjan and Lewis, 2008), and facultative corallivores were defined as fish that were observed to feed on coral for a minor to major part of their diet, while also feeding on algae or other invertebrates (~5-80% of bites from corals; Harmelin-Vivien and Bouchon-Navaro, 1983; Rotjan and Lewis, 2008; Viviani et al., 2019; Ezzat et al., 2020). The selected species included four obligate corallivore species (butterflyfishes Chaetodon ornatissimus, "CHOR"; Chaetodon lunulatus, “CHLU”; Chaetodon reticulatus, “CHRE”; and the filefish Amanses scopas, “AMSC”), three facultative corallivore species (butterflyfishes Chaetodon citrinellus, “CHCI”; and Chaetodon pelewensis, “CHPE”; and the parrotfish Chlorurus spilurus, “CHSP”), and three grazer/detritivore species (surgeonfishes Ctenochaetus flavicauda, “CTFL”; Ctenochaetus striatus, "CTST"; and Zebrasoma scopas, “ZESC”). Coral and algal samples were collected from locally abundant coral and algae genera (Pratchett, 2014; Burkepile et al., 2020), including the corals Acropora hyacinthus (“ACR”), Pocillopora species complex (“POC”, Gélin et al., 2017; Johnston et al., 2018), and Porites lobata species complex (“POR”, Forsman et al., 2009; Forsman et al., 2015); and mixed communities of turf algae (“Turf”) as well as macroalgae in the general Asparagopsis (“Asp”), Dictyota (“Dict”), Lobophora (“Lob”), Sargassum(“Sarg”), and Turbinaria sp. (“Turb”). Sediment (“SED”; 250 ml) and water (“WAT”; 1.9 L) were collected concomitantly with fish and environmental samples using sterilized containers. Following collection, all samples (fish, coral, algae, sediments and water) were immediately transported on ice to the lab, where they were processed using sterile methods as described in Grupstra et al. (2021) and preserved in DNA/RNA Shield (Zymo Research, CA).

Sequencing and analysis of 16S rRNA gene amplicons from in situ samples

DNA was extracted from all samples using the ZymoBIOMICS DNA/RNA Miniprep kit (Zymo Research, CA) according to the manufacturer’s instructions, but with a 1hr proteinase K incubation (35°C) before the lysis buffer step. Library preparation and sequencing was conducted at the Genomics Core Lab at the Institute of Arctic Biology of the University of Alaska Fairbanks. All samples were sequenced using the 16S rRNA gene V4 primers 515f (GTGYCAGCMGCCGCGGTAA) and 806rB (GGACTACNVGGGTWTCTAAT) on Illumina MiSeq using v3 2x300bp chemistry (Apprill et al., 2015; Parada et al., 2016; Walters et al., 2016). Mock communities (HM-782D, BEI Resources, VA), extraction negatives, and plate negatives were included to facilitate the identification and removal of potential contaminants in silico (Davis et al., 2018).

Quantitative PCR of Vibrio coralliilyticus genes in fish feces

Quantitative PCR (qPCR) was conducted using vcpARTF and vcpARTR qPCR primers developed for the bacterial coral pathogen Vibrio coralliilyticus (Wilson et al., 2013) and results were standardized using primers for general bacteria 967F and 1046R (Sogin et al., 2006; Chen et al., 2011; Shiu et al., 2020). Standard curves (for V. coralliilyticus and general bacteria) were made from a Vibrio coralliilyticusculture (AS008) isolated from corals collected in the Flower Garden Banks National Marine Sanctuary (northwest Gulf of Mexico). Sequencing of the full-length 16S gene region of bacterial DNA with primers 8F and 1513R resulted in 97.5% percent identity with V. coralliilyticus strain U2 (accession MK999891.1) with 100% query cover. The primers vcpARTR and vcpARTF were used to amplify metalloprotease genes and the amplicon was then cleaned using a genejet PCR purification kit (Thermo Fisher, MA); resultant DNA concentrations were acquired using Qubit (Thermo Fisher, MA). A standard curve was made using serial dilutions from 109 to 100 gene copies per µl template; the standard curve for V. coralliiyticus primers vcpARTR and vcpARTF had an efficiency of 106.2%; the standard curve for general bacteria primers 967F and 1046R had an efficiency of 92.3%. Sanger sequencing of gene fragments amplified using vcpARTF and vcpARTR primers from DNA extracted from C. striatus feces resulted in a top hit against a V. coralliilyticus strain P4 metalloprotease gene (accession JQ345042.1) with an e-value of 3x10-12 (query cover 58%, percent identity 85%).

Analysis of 16S rRNA gene amplicons

Bacterial 16-S reads were processed in RStudio (version 1.1.456) through the DADA2 pipeline (version 1.11.0, Callahan et al., 2016) and using phyloseq (v1.38). The DADA2 pipeline generated a table of amplicon sequence variants (ASVs), and bacteria taxonomy was assigned using the SILVA rRNA database (version 132, Quast et al., 2012). Non-target (e.g., mitochondrial, chloroplast) reads and singletons were removed. A total of 865 potential contaminant ASVs were identified and removed with the decontam package (v. 1.10.0) using the prevalence method (Davis et al., 2018). A total of 160 samples with >1,000 reads remained after these initial quality control and filtering steps (excluding negatives and mock communities), with a mean of 15,263 reads per sample. All analyses outlined below were conducted based on this dataset, but additional ASV and sample filtering steps were included for several of the approaches. Raw sequencing data are deposited on the SRA (BioProject PRJNA935035).

Quantitative PCR of Vibrio coralliilyticus genes in fish feces

Amplification of the V. coralliilyticus target gene at <30 cycles was counted as a positive detection, which roughly corresponded to amplification of the 102 gene copies µl-1 standard (Ct=30.4). Delta cycle threshold (dCT) values were calculated by subtracting the cycle threshold at which the signal from general bacteria primers was detected from the cycle threshold at which V. coralliilyticus was detected. Higher dCT values indicate lower relative abundances of V. coralliilyticus.