Coral thermal stress and bleaching enrich and restructure reef microbial communities via altered organic matter exudation

Coral bleaching is a well-documented and increasingly widespread phenomenon in reefs across the globe, yet there has been relatively little research on the implications for reef water column microbiology and biogeochemistry. A mesocosm heating experiment and bottle incubation compared how unbleached and bleached corals alter dissolved organic matter (DOM) exudation in response to thermal stress and subsequent effects on microbial growth and community structure in the water column. Thermal stress of healthy corals tripled DOM flux relative to ambient corals. DOM exudates from stressed corals (heated and/or previously bleached) were compositionally distinct from healthy corals and significantly increased growth of bacterioplankton, enriching copiotrophs and putative pathogens. Together these results demonstrate how the impacts of both short-term thermal stress and long-term bleaching may extend into the water column, with altered coral DOM exudation driving microbial feedbacks that influence how coral reefs respond to and recover from mass bleaching events.

daily average water temperature data.Time-series data were collected from 3 sites on the MCR 7 LTER fore-reef: FOR1, FOR4 and FOR5 (GPS location: 17°28'30.0"S149°50'13.2"W;8 17°32'49.2"S149°46'08.4"W;17°34'55.2"S149°52'30.0"W;respectively).From each location, 9 measurements from five sensors ("upper water column", "middle water column", "bottom water 10 column", "temperature shallow", and "temperature deeper") was used to calculate the average 11 temperature +/-one standard deviation.Bleaching was first observed in the corals adjacent to 12 Gump Station, Mo'orea, in April 2019 (Leinbach et al., 2021) (Figure 1B).Accumulated degree 13 heating days reached a maximum of 17°C-Days in mid-April before rapidly decreasing (Burgess 14 et al., 2021).By the start of field collection on May 8th, 2019, the temperatures dropped below 15 the 29 °C threshold.Corals had experienced a total of 110 days of temperatures exceeding the 16 threshold in a period of five months (151 days).17 The BL1 detector for the blue laser (488 nm) and the SSC detector for the violet laser 26 (405 nm) were used in conjunction to elucidate bacterial abundances.Voltages and gating were 27 manually determined to enable easy identification of SYBR green stained bacterioplankton 28 populations using a BL1 voltage of 2,625 mV and a SSC voltage of 2,500 mV.Density plots of 29 BL1 vs. SSC were gated on the easily distinguishable population of SYBR green stained 30 bacteria (Figure S1a).31 32

Symbiodiniaceae Quantification 33
Samples were frozen at -40 °C for 14 days prior to transportation to University of Hawaiʻi 34 at Mānoa campus where they were frozen at -80 °C for 23 months prior to flow cytometry 35 processing.In brief, samples were thawed, briefly homogenized using a vortex machine, and 36 200 µL of each sample being aliquoted into 96-well round-bottom flow cytometry plates.37 Symbiodiniaceae slurries were run on a Beckman Coulter CytoFLEX S Flow Cytometer 38 (Beckman Coulter, Product No: B78560).39 The Chlorophyll-a emissions PMT and the two scatter detectors for the violet laser were 40 used in conjunction to count Symbiodiniaceae cells, delineated as distinct populations of large 41 cells with high specific chlorophyll a content.Voltages and gating were manually determined to 42 enable easy identification of this population using a FSC gain of 100, a SSC gain of 100, and a 43 Chla gain of 50.Chla values had a lower threshold of 20,000.Density plots of Chla vs. FSC 44 were gated on the easily distinguishable population of SYBR green stained bacteria (Figure 45 S1b).46 47 Metabolomics Sample Collection field, PPL cartridges were activated with 1x methanol wash followed by 2x pH 2 LC-MS grade 52 water washes.Samples were loaded on the cartridges by pumping the acidified filtrate over the 53 PPL cartridges at 8mL/min.Cartridges were desalinated using a 3x pH2 LC-MS grade water 54 wash and dried using N2 gas.Sequences with a start or stop position outside the 5th-95th percentile range (over all 98 sequences) were discarded.We removed potential chimeras with chimera.vsearch().99 Taxonomies were assigned using classify.seqs()and classify.otus().We removed all 100 mitochondrial or chloroplast OTUs, as well as sequences with no annotations at the domain 101 level.Using sub.sample(), we normalized the abundance in each sample by subsampling to 102 12,000 sequences.OTUs were defined as unique "amplicon sequence variants" (100% clustering OTUs) by DADA2 (Callahan et al., 2016).We used the lulu R package to remove 104 artefactual OTUs (Frøslev et al., 2017): we merged two OTUs if all of the 3 following conditions 105 were satisfied: 1) They co-occur in every sample, 2) One of the two OTUs has a lower 106 abundance than the other in every sample and 3) they share a sequence similarity of at least 107 97%.Finally, we discarded OTUs represented by two or less reads across the 243 samples 108 included in this library.UniFrac distance matrices were constructed from the OTU data and 109 used to assess multivariate differences between microbial communities (Lozupone & Knight, 110 2005).At the final time point, two outlier samples were identified and removed from downstream 111 16S analysis (outliers were defined as samples whose log10 distance from the centroid of a 112 treatment ≥ 1.5 SD above the mean log10 distance from the centroid for a given treatment).113 114

Metabolomics Chemoinformatic Methods 115
Untargeted LC-MS/MS data pre-processing was performed with MzMine3 v3.2.8 116 (Pluskal et al., 2010).Mass detection was performed using the "centroid" algorithm.Intensity 117 thresholds of 1E5 and 1E3 were for used for MS1 and MS2, respectively.Chromatograms were 118 built using the ADAP chromatogram builder with a min group size of four, group intensity 119 threshold of 2E5, minimum peak intensity of 1E5, and m/z tolerance of 0.0015 Da or 10 ppm.120 Extracted Ion Chromatograms (XICs) were deconvoluted using the local minimum search 121 algorithm with a chromatographic threshold of 85% , a search minimum in RT range of 0.08 min, 122 and a median m/z center calculation with m/z range for MS2 pairing of 0.1 and RT range for 123 MS2 scan pairing of 0.15.Isotope peaks were grouped and features from different samples 124 were aligned with 0.001 Da or 5 ppm mass tolerance and 0.1 min retention time tolerance.MS1 125 peak lists were joined using an m/z tolerance of 0.0015 Da or 10 ppm and retention time 126 tolerance of 0.15 min.Alignment was then performed by placing a weight of one on RT and a 127 mobility weight of one.The feature table of peak areas were exported as a .csvfile and the 128 corresponding consensus MS/MS spectra were exported as a .mgffile.129 130

Bacterial Differential Abundance Analysis 131
In order to directly elucidate which specific bacterial taxa were driving these differences, 132 we performed DESeq2, a method for analysis of differential expression of count data derived 133 from high throughput sequencing, on a subset of the data that only included the four coral DOM 134 treatments (Love et al., 2014).DESeq2 requires raw read inputs prior to reads per sample 135 normalization and was thus run on raw read counts prior to the subsampling and Lulu steps of 136 our bioinformatic pipeline.In order to eliminate low and prevalence of OTUs prior to 137 DESeq2, OTUs were removed so that only those with raw abundance ≥ 50 in three or more 138 samples or a raw abundance ≥ 1000 in one or more samples were included, which comprised a 139 subset of 187 OTUs.Given that OTU abundances in the stressed coral treatments were going 140 to be compared to the coral controls, we further removed 28 highly variable OTUs within the 141 Ccontrol coral treatment.Specifically, all OTUs with a coefficient of variation (CV) greater than 1 142 standard deviation of the mean CV of all OTUs were culled, yielding a final count of 159 OTUs 143 to be run through DESeq2.144

Figure S1 :Figure S2 :
Figure S1: A) Representative density plots of gated SYBR polygon derived bacterial counts for a SYBR stained .2µmfiltered milliq control and a SYBR stained sample and B) a representative dotplot of a gated Symbiodiniaceae population in orange.

Figure S3 :
Figure S3: Box and whisker plots of bacterial specific growth rate, in log10 cells per hour, for the 6 treatments.Significant differences between treatments (Tukey post-hoc test, p<0.05) are denoted by letters above each boxplot.

Figure S4 :
Figure S4: Non-metric multidimensional scaling plot of bacterial communities from start and end of bottle incubation using unifrac dissimilarity.

Figure S5 :Figure S6 :
Figure S5: Box and whisker plots of the alpha diversity of the bacterial communities at the end of the incubation

Figure S7 :Figure S8 :
Figure S7: Stacked barplots of the relative abundance of significant OTUs (p≤.05 after FDR) enriched or depleted in any of the 3 coral stress treatments relative to the Control treatment according to DESEq2.Column facets denote if a given OTU is enriched or depleted relative to the Control.Row facets denote which treatments a group of OTUs is either significantly enriched or depleted in.Relative abundance was derived from the non-subsampled, raw abundance data used in DESEq2.Bars are colored according to bacterial family.

18
Bacterioplankton Abundance and Flowcytometry Settings 19Samples for bacterioplankton abundance were flash-frozen at -40 °C for 14 days prior to 20 transportation to University of Hawaiʻi at Mānoa campus where they were frozen at -80 °C for 21 six months prior to flow cytometry processing.Fixed microbial abundance samples were 22 thawed, and 200 µL of each sample aliquoted and stained with 2 µL 100X SYBR Green to be 23 run on an Attune Acoustic Focusing Cytometer (Applied Biosystems, Part No. 4445280ASR) at

55 56Sample
Storage for Bacterial Community Composition, Dissolved Organic Carbon and 57 Metabolite Solid Phase Extraction 58 Sterivex filters were frozen at -40 °C for 14 days prior to transportation to University of 59 Hawaiʻi at Mānoa campus where they were transferred to -80 °C for six months prior to DNA 60 extraction.DOC samples were kept at room temperature for 14 days prior to transportation to 61 the University of California, Santa Barbara (UCSB).At UCSB the DOC samples were stored at 62 15 °C in the dark for six months prior to sample processing.PPL Cartridges were stored at -80 63 °C until samples were processed at UCSD in spring 2021.64 65 Microbial Community DNA Extraction, Library Prep, and Sequencing 66 To extract bacterial DNA from the Sterivex filters, the filter was removed from the plastic 67 casing using sterile pliers, scalpels, and tweezers, and added to MP Biomedicals Lysing Matrix 68 A (No. 116910100) tubes with 0.5 mL MC 1 lysis buffer and homogenized using a MP 69 Biomedicals FastPrep-96 bead beater.A portion of the homogenate (0.4 mL) was recovered 70 and DNA extractions were completed using the Macherey-Nagel NucleoMag Plant Extraction Kit 71 (No. 744400.4)with KingFisher Accessory Kit (No. 744951).DNA samples were eluted to a final 72 volume of 110 μL.73 515F (Parada et al., 2016) and 806R (Apprill et al., 2015) Earth Microbiome Project 74 primers were used according to Walters et al., 2016 with barcodes on the 515F primer.75 Amplicons were generated from a single round of PCR using primers that include Illumina 76 spacers, Illumina adapters, index sequences (on the forward primers), and 16S rRNA gene template region.PCR reagents included 7.2 µl DNase free water, 10 µl PLAT II MM (2X) 78 (Invitrogen, Platinum II Hot-Start PCR Master Mix, Catalog No: 14000012), 0.4 µl forward primer 79 (10 µM), 0.4 µl reverse primer (10 µM), and 2 µl DNA template.PCR was performed on an 80 Applied Biosystems SimpliAmp (Catalog No: A24811) system using an initial denature of 94 °C 81 for 2:00 min followed by 35 cycles of 94 °C for 15 s, 54 °C for 15 s, 68 °C for 7 s, followed by a 82 final extension at 68 °C for 3 min.DNA extraction blanks and no-template control blanks were 83 included as negative controls.Mock communities (ZymoBIOMICS Microbial Community DNA 84 Standard, Cat No: D6305) were included as positive controls to detect contaminants from kits or 85 library preparation.Total amplicons per sample were normalized to between 12.5 and 15 ng 86 using Charm Biotech Just-a-Plate PCR purification and normalization kit.87 88 16 Amplicon Bioinformatics 89 Raw paired fastq reads were preprocessed using the DADA2 R package (Callahan et 90 al., 2016a).We truncated forward reads at position 220 and reverse reads at position 190 and 91 discarded them if they contained a number of expected errors above three using the 92 filterAndTrim() function.Denoising was performed with the learnError() and dada() functions 93 with default parameters.Using the mergePairs() function, we merged reads if they overlapped 94 by at least 20 bases, and allowed for 1 mismatch at most.Duplicate technical replicates were 95 then merged bioinformatically.We used mothur (Schloss et al., 2009) along with the Silva 96 (release 132) database (Quast et al., 2013) to align and annotate the sequences, respectively.97

Table S1 :
Mean relative abundance of dominant Families in the 6 treatments