Towards enhancing coral heat tolerance: a “microbiome transplantation” treatment using inoculations of homogenized coral tissues

Background Microbiome manipulation could enhance heat tolerance and help corals survive the pressures of ocean warming. We conducted coral microbiome transplantation (CMT) experiments using the reef-building corals, Pocillopora and Porites, and investigated whether this technique can benefit coral heat resistance while modifying the bacterial microbiome. Initially, heat-tolerant donors were identified in the wild. We then used fresh homogenates made from coral donor tissues to inoculate conspecific, heat-susceptible recipients and documented their bleaching responses and microbiomes by 16S rRNA gene metabarcoding. Results Recipients of both coral species bleached at lower rates compared to the control group when exposed to short-term heat stress (34 °C). One hundred twelve (Pocillopora sp.) and sixteen (Porites sp.) donor-specific bacterial species were identified in the microbiomes of recipients indicating transmission of bacteria. The amplicon sequence variants of the majority of these transmitted bacteria belonged to known, putatively symbiotic bacterial taxa of corals and were linked to the observed beneficial effect on the coral stress response. Microbiome dynamics in our experiments support the notion that microbiome community evenness and dominance of one or few bacterial species, rather than host-species identity, were drivers for microbiome stability in a holobiont context. Conclusions Our results suggest that coral recipients likely favor the uptake of putative bacterial symbionts, recommending to include these taxonomic groups in future coral probiotics screening efforts. Our study suggests a scenario where these donor-specific bacterial symbionts might have been more efficient in supporting the recipients to resist heat stress compared to the native symbionts present in the control group. These findings urgently call for further experimental investigation of the mechanisms of action underlying the beneficial effect of CMT and for field-based long-term studies testing the persistence of the effect. Video abstract Supplementary Information The online version contains supplementary material available at 10.1186/s40168-021-01053-6.


Microbiome sampling, DNA extraction, and amplicon sequencing
Throughout the two experiments microbiome samples were carefully collected from the fragments employed in the experiments using sterile clippers (Pocillopora: 1-2 cm clip of each fragment) or sterile peelers (Porites: scrape ø 1-2 cm). Prior to collecting the tissue, fragments were rinsed thoroughly with FSW (0.2 um). Then samples were flash frozen in liquid nitrogen. Porites 'start' samples were taken from extra coral fragments which were collected in the reef sites from the same colonies during coral collection. Tissue sampling was always performed after the measurements of coral response variables. At the sampling timepoints, seawater samples (1 L) were collected from each experimental tank using sterile cubitainers rinsed with 10% bleach/chlorine solution and MilliQ water. Seawater samples were vacuum-filtered over a 0.2 μm filter (Durapore PVDF filter membranes, Merck, Germany), shock frozen and stored in liquid nitrogen, and subsequently stored at -80 °C. DNA extractions followed a modified Qiagen Allprep DNA/RNA column extraction kit protocol. Filters were preprocessed by thawing (RT for 5 min) and refreezing (-20 °C for 5 min), repeating the cycle 3 times to promote cell lysis, then sliced into stripes, using a sterile scalpel before homogenization [6]. Next, for coral and filter samples, further modifications included the use of lysis tubes (2 mL Lysing Matrix E, MP Biomedicals, USA), bead-mill homogenization (2 x 1 min 30 Hz, Qiagen TissueLyser II, Germany), and centrifugation for 3 minutes at 15 000 rcf, before the clear supernatant was processed following the manufacturer's instructions, adding a second washing step prior to elution of DNA from the column. For library preparation 10-15 ng of DNA were used and the amplification performed with a Phusion HS II High-Fidelity DNA polymerase (0.5 Us) in a dual-barcoding approach [7]. Primer pair 357F [5´CCTACGGGAGGCAGCAG´3] and 806R [5´GACTACHVGGGTWTCTAAT´3] was employed at 0.28 μM. PCR cycling conditions were as follows: 30 s at 98 °C; 30 × [9 s at 98 °C, 60 s at 55 °C, 90 s at 72 °C]; 72 °C, 10 min; 10 °C on hold. PCR-products were normalized using the SequalPrep Normalization Plate Kit (Thermo Fischer Scientific, Waltham, MA, USA), pooled in equimolar amounts, and sequenced on the Illumina MiSeq v3 2x300bp with 20% PhiX (Illumina Inc., San Diego, CA, USA). Quality control samples (QC) were included, i.e., negative (DNA extraction blanks and PCR blanks) and positive controls (#ZRC 190811, ZymoBIOMICS Microbial Community DNA, Zymo Reseach).

Amplicon raw data processing
Amplicon data produced in two Illumina runs were demultiplexed based on 0 mismatches in the barcode sequences. Raw sequence data was processed using a QIIME2 V2019.7 pipeline. First, PCR primer sequences were removed (cutadapt, [8]) and quality of paired-end sequence reads assessed and truncation parameters set at a read quality of Qscore ≥ 20 (demux). Assembly of reads, denoising, and generation of bacterial amplicon sequence variants (ASVs) were carried out using the DADA2 plug-in [9] under default settings, truncating poor quality bases of the forward read at 277 bp and reverse read at 220 bp resulting in an contig overlap of 31 bp for the first library, and truncating the forward read at 278 bp and the reverse read at 230 bp (i.e., overlap of 42 bp) for the second library. This step removed 31% and 26% of sequences from the two libraries, respectively. Now, libraries were merged ('feature-table' merge options). A naïve-Bayes classifier object was trained based on the 16S region V3-V4 and SILVA database V132 (99%; [10]) and subsequently employed for the classification of the sequences (classify-sklearn, feature-classifier; [11,12]). Unassigned, mitochondrial, archaeal and chloroplast reads were removed (feature-table, filter-features). ASV count tables were exported as 'biome' files to be also used in R (export, and biom convert). Contaminant bacterial taxa aka amplicon sequence variants (ASVs) were identified through examination of four PCR negative control samples (10,080 reads over 31 ASVs with 95 -8 854 reads per sample) and 13 DNA extraction kit blank samples (14,328 reads over 92 ASVs and 32 -5,841 reads per sample). ASVs were scored as contaminants, once they occurred in >1 sample and had a relative abundance higher than 5% or 1% within all control samples, respectively (i.e., a read count of 200-500). 10 contaminant ASVs resulted from PCR negative controls and 27 from kit blank controls (in total 33 contaminant ASVs, Dataset S3). Next, to further identify and exclude coral origin sequences, 241 ASV sequences showing no higher classification level than 'domain: Bacteria' (SILVA database) were compared with GenBank (https://www.ncbi.nlm.nih.gov). 96 such sequences were identified as coral origin sequences. The lists contaminant ASVs and coral origin sequences were subtracted from the full data set in a final clean up step in R environment.

Amplicon data overview
After denoising, classification, and removal of unclassified reads (QIIME2) the full 16S rRNA gene amplicon data had 2,560,337 reads across 9,593 amplicon sequence variants (ASVs) and 312 samples with an average of 8 206 reads per sample (including coral and seawater samples from two CMT experiments and DNA-extraction kit blanks). After removal of further unrelated sequences (bacterial contaminants and sequences of coral host origin) samples reached an asymptote at subsampling depth of 4,000 reads, while retaining essential replicate samples. Seawater samples collected from the source tank did not reach asymptote at 4,000 reads and were excluded from α-and β-diversity analyses (three samples from the Pocillopora experiment and two samples from the Porites experiment, Fig. S3). The data set contained 840,000 reads over 7,177 ASVs and 210 samples after rarefying. In parallel, the non-rarefied data set was filtered by removal of rare ASVs (< 10 reads,'filt-10') resulting in 2 335,885 reads over 4,604 ASVs and 293 samples (for ASV count tables see Dataset S1 This filtering translates to the removal of 0.01% of total reads and 51% of all ASVs and demonstrates a significant proportion of rare ASVs in the sequencing data.

Fig. S1 Temperature profiles of experimental treatments and sampling timepoints. (A-B)
Temperature profiles of heat tolerance assays were conducted to identify suitable donors and recipients. Temperature treatments were adjusted accounting for the specific environmental sensitivity of each coral genus: Pocillopora was exposed to a single heat-peak over one day; Porites required two heat-peaks over two days to show a heat stress response. (C-D) Temperature profiles during coral microbiome tranplantation experiments: The inoculation phase was performed at 29 °C and subsequent heat tolerance reassessment peaked at 34 °C. Inoculation was performed once for (C) Pocillopora, and repeated over three days for (D) Porites. Timepoints of coral response measurements and sample collection are indicated: Start (①) and end (②) of heat tolerance assessment; start (③) and end (④) of inoculation, end (⑤) of heat tolerance reassessment. Types of data and sample collection: ✱ = photosynthetic efficiency and bleaching score measurements; ○ = DNA sampling; syringe icon = CMT inoculation event; branching coral = Pocillopora sp.; massive coral = Porites sp.; light green = 'HighVar' west shore corals; orange = 'HighVar' reef flat corals; teal = 'LowVar' east shore corals; blue line = ambient '29 °C' treatment; red line = heat stress '34 °C' treatment. . Subsequently, (E-F) the temperature effect on the photosynthetic efficiency of the recipient group and the FSW control group are shown. Plots visualize ∆-effective quantum yield (i.e., the difference in photosynthetic efficiency at endstart of each experimental part). Swarm plots (left side plot) show raw data points and Cumming estimation plots (right) depict the effect sizes as the mean differences between the treatment groups using Cohen's d and a 95% confidence interval. Significant differences are indicated by connecting lines (p < 0.001***, < 0.01**, < 0.05* from generalized linear/linear mixed effect models). Vertical error bars = 95% CI; N = individuals per treatment group; Branching coral = Pocillopora sp.; massive coral = Porites sp.; light green = 'HighVar' west shore corals; orange = 'HighVar' reef flat corals; teal = 'LowVar' east shore corals; colored circles represent the donor inoculum used: light green = 'HighVar' Pocillopora donor, orange = 'HighVar' Porites donor.               Table S6 Summary of tank conditions during initial heat tolerance (HT) assessment. Temperature, oxygen, light intensity and salinity data (mean ± SD) are presented for the duration of the HT assay for both coral species. Temperature is specifically summarized for the temperature-peak period for each treatment (i.e., '29 °C' and '34 °C'). Two HT assays for Porites corals were run separately, as Porites corals from each respective high and low variability habitat required different light regimes. Sites of origin: 'HighVar' = high variability site; 'LowVar' = low variability site.

Corals 'Site of origin'
Treatments Temperature (°C)