Isolation and selection of BMCs
BMCs were isolated from three different coral hosts, namely, Stylophora pistillata, Galaxea fascicularis, and Pocillopora verrucosa, collected at Thala (22°15′46.9″N 39°03′05.9″E), Aquarium (22°23′15.6″N 38°55′07.2″E), and Al Fahal reefs (22°18′18.4″N 38°57′52.5″E) in the Saudi Arabian Red Sea 25. Briefly, coral fragments were macerated and plated onto marine agar (Zobell 2216, HiMedia Laboratories, Mumbai, India) and Luria–Bertani broth (Sigma-Aldrich®, St. Louis, MO, USA) at 3.5% salinity for bacterial growth and isolation. To select the BMC strains from > 400 obtained isolates, the genomic DNA of the bacterial isolates was extracted using the Wizard® Genomic DNA purification kit (Promega Corporation, Madison, WI, USA) to sequence the partial 16S rRNA gene using the universal primers 27F and 1492R 36. Sequencing was conducted by Macrogen (Seoul, South Korea) using a Sanger sequencing platform. Bacterial strains identified as potential coral or clinical pathogens (e.g., Vibrio spp.) were excluded (> 300 isolates). The remaining strains were tested for their beneficial functions for corals, including catalase activity, siderophores production, urease activity, phosphate assimilation, and antagonistic effects against the well-known coral pathogen Vibrio coralliilyticus. The evaluated tests were performed following the methodologies described by Rosado et al. (2019)20, Santoro et al. (2021)21, and Delgadillo-Ordoñez et al. (2023)25. A total of six strains were selected as BMCs when positive to at least one of the aforementioned trait tests 25. In Raimundo et al. (submitted)37, the taxonomic identity of the six BMC strains was updated based on the full-length 16S rRNA gene sequences obtained in genome sequencing projects and was defined as follows: two strains of Pseudoalteromonas galatheae and Cobetia amphilecti and one strain of Halomonas sp. and Sutcliffiella sp. The selected BMCs were stored in glycerol 20% at − 80°C.
Production of the BMC consortium
In total, 1.2 L of the BMC consortium was produced weekly for the experiment, divided into three stages, namely preinoculum, inoculum, and consortium assembly25. Briefly, the preinoculum stage consisted of the activation of each bacterial strain overnight in marine broth at 26°C and 140 rpm. For the inoculum stage, 1 mL of the preinoculum of each strain was added to 250-mL Erlenmeyer flasks containing 99 mL of marine broth and incubated overnight at 26°C and 140 rpm. For the consortium assembly stage, the content of each Erlenmeyer flask was centrifuged, washed, and resuspended in 3.5% NaCl solution to a final concentration of 1 × 109 cells/mL. Next, 100 mL of each strain was added to a 1-L borosilicate flask (600 mL of BMC consortium in each flask) and stored at 4°C for ≤ 1 week until in situ inoculation on the sponges. All steps excluding centrifugation were performed in sterile conditions inside a laminar flow cabinet.
Experimental design and field sampling
The 4-week in situ experiment was performed in May 2022 and June 2022 at the Al Fahal reef (Figs. 1a and 1b). The two most common sponge species in this reef, namely S. carteri and C. crassa, were selected for the experiment (Figs. 1c and 1d). In total, 30 sponges were tagged by SCUBA diving in the field by treatments in quintuplicate: control (no inoculation), BMC inoculation once a week (BMC 1×), and BMC inoculation thrice a week (BMC 3×; Fig. 1e; Supplementary Figure S1). Because of time limitations for sample processing, the experiment for S. carteri (including sampling times) started 1 week after that for C. crassa but followed the same schedule.
Before initiating the experiment (T0), a square-like piece of sponge (approximately 25 cm2) was cut with a scalpel and placed in a 50-mL Falcon tube. For C. crassa, the top, middle, and base of the sponges were collected, whereas for S. carteri, only the top and middle ones were collected because of the small size of the individuals. In addition, 1 L of seawater was collected in a plastic bottle (cleaned with 10% bleach, 70% ethanol, and sterile Milli-Q water) next to each sponge replicate. Immediately after sampling, the sponge samples were rinsed in sterile artificial seawater, squeezed into a 5-mL cryovial, and snap-frozen in liquid nitrogen (LN2). The seawater bottles were kept in a box with ice for 1 h until laboratory processing. The seawater samples were filtered using 0.22-µm filters (47 mm, Millipore, Burlington, MA, USA) in a filtration apparatus (Sartorius [Göttingen, Germany] for C. crassa and Millipore for S. carteri). After filtration, the filters were carefully folded, placed in 5-mL cryovials, and snap-frozen in LN2. Sponge and seawater samples were transferred from LN2 to a − 80°C freezer until DNA extraction.
Three days after T0 sampling, BMC inoculation was performed every Sunday (BMC 3×), Tuesday (BMC 3×) and Thursday (BMC 3× and 1×) for 4 weeks for each species. Each plastic syringe (cleaned with 10% bleach and Milli-Q water) was filled with 30 mL of the BMC consortium and stored at 4°C 1 day before each inoculation. For each sponge individual, the content of one syringe (30 mL of the BMC consortium) was manually inoculated over the entire sponge body. No inoculation was performed for control sponges. Photographs were taken during the experiment to document the regeneration/healing process of the sponges after T0 tissue sampling.
At the end of the experiment (T1), the sponges and seawater were sampled as described for T0. Approximately 1.5 months after the end of the experiment, a nontoxic fluorescent dye (Fluorescein F2456, Sigma-Aldrich) was applied to the sponges to verify that they were pumping properly. The in situ seawater temperature was continuously recorded using a Multiparameter CTD probe (Ocean Seven 310, Idronaut, Brugherio, Italy). The seawater temperature ranged from 27.71°C (lowest daily minimum temperature) to 30.71°C (highest daily maximum temperature) throughout the experiment. The daily mean temperature was 28.26°C–30.34°C (Supplementary Figure S2).
Microbiome analysis
Total DNA from the sponges and seawater was extracted using the DNeasy PowerSoil kit (Qiagen, Hilden, Germany) following the manufacturer’s protocol and stored at − 80°C. DNA was quantified using Qubit 4 (Invitrogen, Thermo Fisher Scientific Inc., Waltham, MA, USA) and NanoDrop One (Thermo Fisher Scientific Inc.). The hypervariable region V3-V4 of the 16S rRNA gene was sequenced on the Illumina NovaSeq Platform (2 × 250 bp) by Novogene (Beijing, China) using the universal primers 341F (CCTAYGGGRBGCASCAG; 38) and 806R (GGACTACNNGGGTATCTAAT; 39). The fourth replicate for the S. carteri seawater sample failed Novogene’s quality control; thus, it was not sequenced or considered for subsequent analyses.
Raw sequences were quality-filtered using Trimmomatic (version 0.38) in a sliding window of 4 bp and reads with a quality score below 15 or shorter than 100 bp were discarded 40. Paired-end reads were merged using USEARCH (version 11.0.667), and contigs < 325 bp or > 550 were removed. Low-quality contigs (Q < 15) and contigs with more than one ambiguous base were also removed 41. The UNOISE algorithm was used to cluster, denoise, and generate unique amplicon sequence variants (ASVs). Chimeras were removed using the UCHIME2 algorithm (high confidence mode; 42). Taxonomy was assigned to the ASV sequences using the BLCA algorithm (43; identity and coverage intervals set to 95–100%) against the genome taxonomy database (version 207; 44). All nonbacterial and nonarchaeal ASV sequences were removed, and reads were mapped to the remaining ASV sequences to estimate relative abundance. Data were normalized for read counts using the DESeq2 package 45. Alpha (ASV richness and Shannon diversity H′) and beta diversities were calculated based on normalized data using the R package vegan (version 2.6-4; 46) in R software (version 4.1.3).
Data analysis
To potentially identify BMCs in sponge and seawater samples, their 16S rRNA gene sequences were searched using BLAST+ (version 2.9.0; 47 against the ASV dataset. Only hits with a nucleotide identity of 100% against at least one of the 16S rRNA gene copies of BMC isolates were considered.
PERMDISP analysis was performed using the R package vegan (version 2.6-4; function betadisper; 46 to check data homogeneity. To compare alpha diversity metrics and detect differences in the relative abundance of the BMC strains in sponge and seawater samples, three-way univariate PERMANOVA based on Euclidean distance matrices was performed using the R package vegan (version 2.6-4; function adonis2; 46) with the factors “sponge body part” (levels: top, middle and base), “time” (levels: T0 and T1) and “treatment” (levels: control, BMC 1×, and BMC 3×) with 9,999 permutations for the sponges, and two-way univariate PERMANOVA was performed with time and treatment as factors for the seawater samples. The post hoc pairwise test was performed with the R package RVAideMemoire (function pairwise.perm.manova; Hervé 2022) once PERMANOVA generated significant results (p < 0.05).
For beta diversity, nonmetric multidimensional scaling was performed using the R package viridis (version 0.6.4) to visually compare the Bray–Curtis dissimilarities of microbial communities. Three-way PERMANOVA based on Bray–Curtis dissimilarity using 9,999 permutations was performed with the R package vegan (version 2.6-4; function adonis2; 46) to detect significant differences in the microbial communities structure over time, treatment, and sponge body part. To determine which ASVs were affected by the BMC 1× and BMC 3× treatments relative to the control, univariate generalized linear models (GLMs) with a negative binomial or Poisson distribution (according to the abundance distribution of each ASV) were used using the R package mvabund, correcting multiple testing, correlations between ASVs, and compositional data 48. Univariate GLMs were generated using “sponge body part”, “time,” and “treatment” as fixed factors. Differentially abundant ASVs between BMC treatment and control groups (regardless of the sponge body part) were analyzed in additional univariate GLMs to assess which microbial species were changing between these groups (control × BMC 1× and control × BMC 3×). Differences were considered statistically significant at p < 0.05. Significant ASVs related only to the factor time were not considered in the results, as they represent the natural variation of the microbiome. An increase or decrease in the relative abundance of the microorganisms for BMC treatments was only considered if the variation between T1 and T0 in the treatment samples exceeded the natural variation in the control samples during the same period. Results from univariate GLM analysis for ASVs with relative abundance ≥ 0.1% were plotted using Microsoft Excel, and analyses were conducted using R software (version 4.1.3).