Study site

This study was carried out during the austral spring (Sep/Oct) of 2023 at OTI Reef, a lagoonal platform reef part of the Capricorn-Bunker Group near the southern boundary of the Great Barrier Reef (23°30′S 152°06′E), Australia (Fig 1A). OTI Reef is an irregularly shaped pseudo-atoll, 5.5 km long and up to 3.5 km wide. The reef crest encloses three shallow lagoons (2–10 m deep) and a coral cay in the southeastern corner of the reef crest. All research activities were conducted within Lagoon 1, one of the three enclosed lagoons.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Locations of the in situ incubation sites and representative Porites lutea colony at One Tree Island Reef (Lagoon 1).

(A) Aerial photography of OTI Reef with coloured symbols indicating the location of the Porites lutea coral colonies and the sediment only chambers, which were selected for in situ incubation within Lagoon 1. Source of OTI Reef map: Queensland Globe, © State of Queensland (Department of Natural Resources and Mines, Manufacturing and Regional and Rural Development) 2025, licensed under CC BY 4.0 licence. (B) The transparent in situ incubation chamber over a Porites lutea coral, showing the (a) water pump for continuous water circulation, (b) oxygen logger to measure seawater dissolved oxygen and (c) light and temperature logger. (C) Example of a Porites lutea coral colony used for coral incubations. Scale bars (black-and-white triangle objects) were placed near each coral for 3D reconstruction of the coral in the modelling software.

https://doi.org/10.1371/journal.pone.0346135.g001

Benthic incubation chamber and sampling procedure

In situ incubation chambers were constructed as described by Roth et al. [45] and deployed over coral heads (Fig 1B). Briefly, each incubation chamber consisted of a clear, cylindrical polycarbonate chamber body (diameter 0.5 m, height 0.39 m), closed at one end by a removable gas-tight lid. Inside, a MiniDO₂T oxygen Logger (Precision Measurement Engineering Inc., California, USA) and a HOBO^® Pendant light and temperature logger (UA-002; Onset Computer Corporation, Massachusetts, USA) were installed in each chamber to measure oxygen flux of the seawater contained within the chamber. Water circulation in the chamber was set to ~ 2 L/min by a custom-made submersible pump (6V DC motor, 300mA, 2W) using a magnetic impeller (powered by 4 x 1.5V AA batteries). All chambers had a natural rubber skirt (3 mm thick) fitted to the chamber body to isolate the chamber from the surrounding water and improve the seal to the sandy bottom.

Two SCUBA divers deployed the incubation chambers over twenty haphazardly selected Porites lutea coral heads over nine sites within the largest lagoon at OTI (Fig 1). P. lutea is a dominant reef-building coral, widely distributed in the Indo-Pacific. The coral heads were growing on sand near patch reefs at a depth of 4–5 m, showed no visible signs of bleaching or disease, were suitably sized to fit inside the incubation chamber and a secure seal was possible around the chamber (i.e., no nearby rocks or reef structure that would prevent pushing the incubation chamber 5 cm deep into the sand to properly seal for the duration of the incubation period). First, the GPS location for each coral head was recorded, then chambers without lids were carefully positioned over the coral heads of interest and pushed into the sediment to a depth of approximately 5 cm. The silicone skirt was gently lowered until resting on top of the sediment. Chambers were left untouched for at least 1 hour to allow the suspended sediment to settle. Subsequently, the oxygen loggers and water circulation pumps were secured in brackets inside the open chambers, and temperature-light sensors were deployed on the sediment (Fig 1B). Each coral head underwent two incubation periods: one dark and one light, each lasting ~2 hours, with a one-hour interval between the two incubation periods. The dark incubation commenced once blackout material was placed over the sealed chamber, and the light incubation was initiated by sealing the chamber. The start and end times of both incubations were recorded for the posterior analysis of the community oxygen flux. Oxygen, temperature and light inside the chamber were recorded every minute.

After the dark incubation, the lid was removed, and the top 1 cm of sediment in the chamber was sampled into three 15 mL Falcon tubes for 16S rRNA gene amplicon sequencing, minimising disruption of the surface sediment layer. Each incubation chamber containing coral (n = 20) was sampled in triplicate, yielding a total of n = 60. Incubation chambers on bare sediment (i.e., sediment-only controls) were deployed as a control for the coral-associated chambers. Each sediment control incubation chamber (n = 3) was sampled in triplicate at all nine sites, for a total of n = 81. These were deployed 12 m from a patch reef to avoid potential effects from the reef. One sediment core (PVC sediment core, 5x5 cm, collected within a 50 cm radius of the incubation chamber) was sampled for grain size and nutrient analysis. All samples were stored on ice after collection, for transport back to the Research Station (within 1 hour of collection). Samples were collected under Great Barrier Reef Marine Park Authority permit G23-48125.1 (issued to SGG).

At the OTI research station, sediment samples for 16S were homogenised and sub-sampled into 1.5 mL Eppendorf tubes. RNAlater (Thermo Fisher Scientific Inc, Massachusetts, USA) was added to each replicate to preserve the DNA and samples were stored frozen at −20°C. Sediment cores were left to settle before seawater was decanted off the top and the remaining sample was transferred to a zip-lock bag and frozen at −20 °C. All remaining analyses (detailed below) were conducted at the University of Sydney.

Photometric measurements

Using photogrammetry and computer modelling techniques described in Ferrari et al. [46], the three-dimensional surface area (SA_3D) and volume (V_C) of each coral head were calculated to estimate their contribution to photosynthesis (P) and respiration (R) rates. Corals were photographed in situ by SCUBA divers from multiple angles using an Olympus Tough TG-6 digital camera and a ring light for illumination. Scale bars were placed near each coral head to enable accurate scaling for 3D reconstruction (Fig 1C). The image processing software Adobe Lightroom (Adobe Inc., California, USA) was used to adjust the white balance, and the 3D model of coral heads was built using the photogrammetry software Agisoft Metashape (Agisoft LLC, St. Petersburg, Russia) using default settings. The SA_3D (m²) and the V_C (m³) of coral heads were calculated.

For the calculation of metabolic rates, the effective volume (V_E; the volume of seawater) in each chamber during incubation was obtained by the formula:

The coral volume (V_C; converted to litres), oxygen logger volume (V_L; in litres) and water pump volume (V_P; in litres) were subtracted from the total seawater volume (V_T; in litres) enclosed by the chamber above the sediment.

Gross primary production and respiration analysis

Photosynthesis and respiration rates were processed and analysed using the respR package [47] in computing environment R [v.4.2.1; 48]. The metabolic rates were obtained in an unbiased manner using the auto_rate function, which automatically identifies the most linear regression of seawater oxygen concentration over time within each separate two-hour incubation, based on kernel density estimation. GPP (gross oxygen production rate during the light period) and R (oxygen consumption rate during the dark period) were extracted as the regression coefficient (slope). The regression coefficient represents the linear increase or decrease of seawater dissolved oxygen (DO) concentration inside the incubation chambers during both incubation periods (light and dark, respectively).

The convert_rate function was used to convert the unitless rates to the area-specific output unit: . Briefly, the seawater community DO flux was calculated as follows:

,

where ΔO₂ (mg L^-1) is the difference in DO concentration between the first (t0) and last point (t1) of the automatically identified linear regression, V_E is the effective seawater volume enclosed in the chamber, Δt is the period between the first and last point of the automatically identified linear regression, and A_E (m²) is the active photosynthetic surface area within the incubation chamber. For the coral-associated chambers, A_E was calculated as the 3D coral surface area (SA_3D) plus the sediment surface area enclosed within the chamber, minus the basal coral surface area (i.e., the area in contact with the sediment), and as sediment surface area within the chamber for the sediment control samples, to ensure only the surfaces actively contributing to photosynthesis were included in the normalisation of gross photosynthetic and respiration rates. SA_3D of the community enclosed by the chamber. V_E and the SA_3D of the coral head were used to normalise the rate values. Rate values of both GPP and R were calculated and converted to the final area-specific unit: .

Additionally, net primary production (NPP = GPP – R) and the photosynthesis-to-respiration ratio (P:R) were calculated to determine the dominant metabolic pathway, assuming that the nighttime respiration rate equals the daytime respiration rate [45,49]. A P:R ratio greater than one indicates that autotrophic activity predominates over heterotrophic activity (P:R > 1), and conversely when the ratio is less than one.

Differences between gross photosynthetic rates and respiration rates were tested using non-parametric Kruskal-Wallis tests, after checking and confirming normality and homogeneity of variances with the Shapiro-Wilk test and Levene’s test, respectively. The effects of sample type (coral-associated sediment and sediment control) and site were each tested for the gross photosynthetic and respiration rates separately.

Grain size analysis

Evaluation of the grain size distribution of the sediment is necessary to determine the availability of organic matter (OM) in the sediment. OM transport by pore-water advection is more efficient in poorly sorted calcareous sands, so its availability differs from that in permeable carbonate sand and mud sediments [50,51].

Prior to analysis, sediments were freeze-dried for 48 h (Alpha 1–4 LSCbasic; Christ, Osterode am Harz, Germany) at 1 bar and sieved to obtain the < 2 mm fraction. The total dry mass, as well as the < 2 mm and > 2 mm fractions were recorded, and the gravel content (% of grains > 2 mm) was calculated. The following sediment grain-size and nutrient analyses were conducted on the sediment fraction below 2 mm.

Sediment grain size distribution was determined by laser particle size analysis using a Malvern Mastersizer 3000 (Malvern Panalytical Ltd, Malvern, UK). Samples were sub-sampled and analysed in triplicate. Before analysis, each replicate was digested in hydrogen peroxide to remove OM (30% H₂O₂, 6 mL, 48 hours). Excess hydrogen peroxide was removed by washing the replicates twice with water. In this study, a refractive index of 1.6 and an absorbance index of 0.01 were used, as the sediment composition primarily consisted of calcium carbonate. Malvern output data were processed in GRADISTAT [v.9.1; 52] to acquire the logarithmic graphical mean (descriptive term), standard deviation (sorting), skewness and kurtosis, calculated according to the formulas of Folk and Ward [53] based on the φ scale [54]. Mean values were calculated from the three technical replicates to determine the grain size characteristics of each sample.

Sediment nutrient measurements

Sieved sediments were ground using a porcelain mortar and pestle and weighed into aluminium capsules. Samples were combusted in a macro elemental analyser (vario MACRO cube; Elementar Analysensysteme GmbH, Langenselbold, Germany), and total carbon (TC) and total nitrogen (TN) were measured as the amount of CO₂ and nitrogen released by gas chromatography. Final values are expressed as the percentage of TC or TN in dry sediment since this method does not differentiate between organic and inorganic forms of the element. In most sediments, inorganic nitrogen is a minimal fraction of TN, so it can be assumed that organic nitrogen equals TN [19]. Meanwhile, the inorganic carbon fraction in calcareous sediments can constitute a significant proportion of the total carbon in the form of calcium carbonate. Therefore, sediment organic matter was estimated by loss on ignition [55]; 250 mg of sediment was subsampled into a glass tube and heated to 375 ºC in a furnace for 24 h [56]. Lastly, the total organic carbon to total nitrogen (C:N) ratio was calculated to assess the impact of sediment biogeochemical processes.

DNA extraction, polymerase chain reaction (PCR) and amplicon sequencing

Total DNA was isolated from 0.35 g of homogenised sediment using the DNeasy PowerSoil Kit (QIAGEN, Hilden, Germany) following the manufacturer’s protocol and DNase-free water was used to elute the DNA. Genomic DNA was extracted from 60 sediment samples (from 20 individual incubations on coral heads, each with 3 replicates per incubation chamber), and 81 sediment control samples, as well as three blanks (extraction kit controls). Extracted DNA was stored at −20 °C before subsequent analyses.

Metabarcoding was performed on the hypervariable region V3–V4 of the 16S rRNA gene, which was chosen as a marker for the phylogenetic classifications of the microbial community. Amplicon libraries were created using the primer pair 341F (341 F: 5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAG-CCTACGGGNGGCWGCAG-3′) and 785R (785 R: 5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAG- GACTACHVGGGTATCTAATCC-3′) primers with Illumina overhang adapter sequences [57,58].

The DNA samples were amplified in a single-step, 35-cycle PCR using a Mastercycler^® nexus X2 gradient PCR thermal cycler (Eppendorf South Pacific Pty. Ltd., Macquarie Park, Australia) under the following conditions: (i) initial denaturation at 94 °C for 2 min, followed by (ii) 35 cycles of denaturation at 94 °C for 30 s, annealing at 55 °C for 30 s and extension at 72 °C for 40 s, with (iii) a final elongation at 72 °C for 7 min. Each PCR reaction consisted of a total final volume of 25 µL, containing 12.5 µL of EconoTaq PLUS GREEN 2X Master Mix (Astral Scientific Pty. Ltd., Tarren Point, Australia), 0.75 µL of each forward and reverse primer, 8.5 µL of DNase-free water, and 2.5 µL of extracted DNA. Indexed and pooled PCR amplicons were sequenced on the 2 x 300 bp Illumina MiSeq sequencing platform at The Ramaciotti Centre for Genomics (The University of New South Wales, Australia). Extraction kit blanks (n = 3) and PCR negative controls (n = 3) were also sequenced.

Statistical analysis of amplicon sequencing data

Adapter and primer sequences were removed using Cutadapt [v.4.4; 59] executed under Python [v.3.9.2; 60]. To ensure that only high-quality, primer-free reads and resulting amplicon sequences were analysed, data were processed in R [v.4.2.1; 48] following the bioinformatic pipeline DADA2 workflow [61]. Briefly, paired-end reads were de-noised, and chimeras were removed using DADA2 to produce amplicon sequence variants (ASVs). ASVs were taxonomically classified using a Bayesian Last Common Ancestor algorithm [BLCA; 62] against the Genome Taxonomy Database R214 (GTDB [63]; released 9 June 2023: https://gtdb.ecogenomic.org/). ASV sequences were aligned with MAFFT [64], and a phylogenetic tree was generated using FastTree2 [65] on the CIPRES v.3.3 Science Gateway [66].

Prior to the data analyses, ASV contaminants were identified using the prevalence method in the decontam package [v.1.6.0; 67]. A total of 213 ASVs were identified as potential contaminants, detected in sequenced extraction blanks and PCR-negative controls, and were removed prior to downstream analyses. ASVs were analysed with Vegan package [v.2.6.4; 68] and microeco package [v.1.15.0; 69] in R.

Coral-associated and sediment control sample comparison

These analyses were conducted on the full dataset (i.e., the sediment from inside incubation chambers with corals, n = 60, and the sediment from empty chambers, n = 81) to compare the coral-associated with sediment only control (non-coral-associated) sediments. Principal coordinates analysis (PCoA) was used to visualise dissimilarities in microbial communities between coral-associated and sediment-only samples, using weighted UniFrac distances that integrate the relative abundance of each ASV and phylogenetic distance [70]. Differences between bacterial communities of the coral-associated and sediment control samples were tested using PERMANOVA on weighted UniFrac distances with the adonis2 function in Vegan. To account for unbalanced sampling and site-level structuring, marginal (Type III) sums of squares were used (by = “margin”) in a model including sample type (i.e., coral-associated and sediment control) and site (unifrac ~ sampletype + site). A second analysis constrained permutations within sites to test whether coral-associated and sediment control samples differed consistently within sites. Homogeneity of multivariate dispersion was evaluated using betadisper to assess whether differences in sample type (i.e., coral-associated and sediment control) were influenced by variation in dispersion.

To compare alpha diversity indices for the coral-associated and sediment-only samples, samples were rarefied to the smallest sample (1,134 reads). Homogeneity of variance was tested using Levene’s test, and normality assessed using Shapiro-Wilk test, before running a non-parametric Kruskal-Wallis test for all indices (i.e., observed number of ASVs, Shannon, Simpson and Chao1). The composition of the sediment bacteria community in each sample for the coral-associated and sediment only samples was shown for the top ten families (relative abundance).

To test for differentially abundant taxa between the coral-associated and sediment control samples, we used DESEq2 package [v.1.38.3; 74], which uses negative binomial generalised linear model (GLM). Analysis was conducted at the family level taxonomic aggregation and the top ten with the largest positive and negative log₂ Fold Changes (log₂ FC) were plotted to illustrate the most representative bacterial families.

Core bacterial families, i.e., including only taxa found in 100% of the samples at the family level, were identified for the coral-associated versus the sediment-only samples. The core community was summarised as the mean relative abundance at the family level and visualised using a stacked bar plot. Core microbiome patterns were plotted using the ggvenn package [v.0.1.10; 71] to show the shared and unique core bacterial taxa between sample groups.

Coral-associated sediment sample analyses

For the remaining analyses using the coral-associated samples (n = 60, with the 81 sediment-only samples removed), principal coordinates analysis (PCoA) was used to visualise dissimilarities in microbial communities using weighted UniFrac distances. Differences in bacterial community composition among coral-associated samples were assessed using permutational multivariate analysis of variance (PERMANOVA) based on weighted UniFrac distances with the adonis2 function (Vegan), using 9,999 permutations. To account for the unbalanced, hierarchical design of the chambers nested within sites, marginal (Type III) sum of squares (by = “margin”) were calculated in the model (adonis2(unifrac ~ SA + Site/Chamber-replicate, data = sampledf, permutations = 9,999, by = “margin”). This allowed independent assessment of the influence of coral surface area on microbial community structure while controlling for site and chamber-level variation.

Distance-based redundancy analysis (dbRDA) models were run based on a weighted UniFrac distance ordination for the Porites lutea coral-associated sediment dataset. Pearson correlation plots were inspected using corrplot package [v.0.95; 72], and collinear variables were removed when r > 0.6 (S1 Fig). Model parameters were adjusted to keep the most significant factors, while concurrently removing any variables with high correlation (VIF score > 5) to reduce the effects of high collinearity in the model. A backward stepwise model was run to further reduce the number of model parameters using the ordistep function with 999 permutations (Vegan). Significance was assessed through PERMANOVA (999 permutations) on the ordistep output, and dbRDA plots were created in ggplot2 [73]. An additional dbRDA analysis was run using the same model parameters and backward stepwise model, after removing four outliers (13_9a, B6_7a, B6_7b, B6_7c) based on two diagnostic steps. First, relative abundance bar plots show these samples had marked differences in community composition compared with other replicates from the same sites. Second, an unconstrained UniFrac-based ordination indicated these samples were inconsistent with the expected within-site replication (i.e., they were far from their respective site clusters), which can disproportionally influence the model.

Finally, to test for differentially abundant taxa based on continuous variables, we used DESEq2 package [v.1.38.3;[74]. Separate models were constructed for each environmental covariate (i.e., seawater dissolved oxygen flux, sediment grain size and sediment nutrient variables), using family level taxonomic aggregation. log₂ Fold Changes were shrinkage-adjusted with the apeglm [v.1.20.0; 75] method to improve effect size estimation and control for variance and taxa with adjusted p-values ≤ 0.05 were considered significantly associated with each covariate. The top three taxa with the largest positive and negative log₂ fold changes were plotted to illustrate the most representative bacterial families to each environmental variable.

Environmental variables

Gross photosynthetic rates were significantly higher for the seawater contained in the chambers with Porites lutea coral-associated sediments (0.52 ± 0.03 g O₂ m ⁻ ² h ⁻ ¹) compared with seawater in the sediment control chambers (0.32 ± 0.01 g O₂ m ⁻ ² h ⁻ ¹; χ² = 10.88, p < 0.001; S2A Fig), similarly, respiration rates were also higher for the seawater in the coral-associated chambers (0.24 ± 0.02 g O₂ m ⁻ ² h ⁻ ¹) compared with sediment control chambers (0.12 ± 0.01 g O₂ m ⁻ ² h ⁻ ¹; χ² = 16.50, p < 0.001; S2A Fig). There was no significant effect of site for gross photosynthetic rate (χ² = 10.13, p = 0.340) or respiration rate (χ² = 9.69, p = 0.380; S2A Fig).

Across coral-associated sediment samples, gross photosynthesis rates for the seawater ranged from 0.11 to 1.29 g O₂ m ⁻ ² h ⁻ ¹, while seawater respiration rates varied between 0.01 and 0.53 g O₂ m ⁻ ² h ⁻ ¹ (S2 Table, S2B Fig), with the photosynthesis to respiration (P:R) ratio ranging between 1.94 to 8.66 (S2 Table, S2C Fig). Total carbon content ranged from 8.59% to 11.45%, total nitrogen from 0.03% to 0.06%, and total organic carbon from 0.08% to 0.29% (S2 Table) while the C:N ratio varied between 1.60 and 6.40 (S2 Table). These coral-associated sediment samples were sand dominated (medium to fine sand), where the gravel fraction (> 2 mm) represented between 0 and 0.08%, sand fraction (2 mm – 63 µm) between 0.81 and 0.99%, and mud fraction (< 63 µm) between 0.01 and 0.19%, with the mean grain size ranging from 0.534 to 2.47 φ (S2 Table).

Bacterial community composition

The bacterial dataset comprised 141 amplicon libraries (n = 60 sediment samples taken inside incubation chambers with corals and n = 81 sediment samples taken inside incubation chambers with sediment only) of the hypervariable region V3–V4 from the 16S rRNA gene, with an average length of 500 bp. After quality filtering and removing chimaeras and potential contaminants, 8,046,234 total sequences were assigned to the kingdom Bacteria, corresponding to 70,230 unique ASVs, comprising 74 phyla, 179 classes, 555 orders, 1,287 families, and 3,464 genera.

Comparison between coral-associated and sediment control samples

Principal coordinates analysis (PCoA; weighted UniFrac) showed coral-associated and sediment-only bacterial communities partially overlapped (Fig 2A), with the sediment samples showing greater dispersion (within-group variability) than coral-associated sediment samples (axis 1: 21.1% variance and axis 2: 10.7% variance; Fig 2A). PERMANOVA indicated that bacterial community composition differed significantly between coral-associated and control sediments (adonis2: F = 16.60, R² = 0.087, p < 0.0001; Fig 2A) after accounting for site effects. Site-level variation also explained a substantial portion of community differences (F = 5.38, R² = 0.225, p < 0.0001). When permutations were constrained within sites, the difference between coral-associated and control sediments remained significant (F = 13.88, R² = 0.091, p < 0.0001; S3A Fig), confirming that this pattern was consistent within sites rather than driven by among-site variation. However, multivariate dispersion differed between sample type (betadisper: F = 51.47, p < 0.0001), indicating that sediment control samples exhibited greater variability in community composition than coral-associate sediments.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Comparison of coral-associated sediment and sediment control samples.

(A) Principal coordinates analysis (PCoA) using weighted UniFrac distances for the Porites lutea coral (dark blue triangle symbols) and sediment (orange diamond symbols) samples, with the ellipses around the sample type (i.e., coral-associated sediment or sediment control samples). (B) Alpha diversity indices for the Porites lutea coral (dark blue) and sediment (orange) samples including the observed number of ASVs (Observed), species richness (Chao1), Shannon evenness (Shannon) and Simpson diversity (Simpson). Significant difference indicated by asterisk (C) Relative abundance of the top ten families in each replicate for the sediment and Porites lutea coral samples, across each site.

https://doi.org/10.1371/journal.pone.0346135.g002

Alpha diversity indices comparing the coral-associated and sediment only samples revealed coral-associated samples were significantly higher for all metrics (the observed number of ASV’s (X² = 50.373, p < 0.001), Chao1 (X² = 65.173, p < 0.001), Shannon (X² = 45.465, p < 0.001), Simpson (X² = 36.232, p < 0.001; S1 Table, Fig 2B)).

At the family level, the most abundant families for both the coral-associated and sediment-only samples were Rhodobacteraceae, Flavobacteriaceae, and Woeseiaceae. In the coral-associated samples, Rhodobacteraceae comprised 10.56% ± 0.29 SE average relative read abundance, followed by Flavobacteriaceae (6.24% ± 0.17) and Woeseiaceae (4.50% ± 0.18; Fig 2C), while the sediment only samples were dominated by Flavobacteriaceae (6.69% ± 0.30), followed by Rhodobacteraceae (5.97% ± 0.54) then Woeseiaceae (4.49% ± 0.26; Fig 2C).

Differential abundance analysis of relative read abundance (RRA) identified 283 bacterial families with significant differences in abundance between the coral-associated sediment versus sediment control samples. In the coral-associated sediment, there was a 24.64 ± 1.27 log₂ FC increase in relative read abundance of Flavobacteriaceae compared with the sediment controls (p < 0.001; S3B Fig). In contrast, Sphingomonadaceae was significantly enriched in the sediment control samples compared with the coral-associated samples (27.21 ± 6.94 log₂ FC, p < 0.001; S3B Fig).

Three bacterial families showed significant responses to gross photosynthesis and respiration rates. Woeseiaceae, Pirellulaceae and Streptomycetaceae were significantly enriched with increasing gross photosynthesis and respiration (all p < 0.002; S3 Table, S4 Fig). There was a negative association for Rhodobacteraceae with a 6.8 log₂ FC decrease in relative read abundance (RRA) with increasing total carbon content (p < 0.001; S3 Table, S4 Fig), whereas there was a positive association for total nitrogen, with a 9.07 log₂ FC increase in abundance with increasing total nitrogen (S3 Table, S4 Fig). We also found an 11.32 log₂ FC decrease (p < 0.001; S3 Table, S4 Fig) in the RRA of Alteromonadaceae with increasing total nitrogen, but a 16.35 log₂ FC increase with increasing C:N ratio (p < 0.001; S3 Table, S4 Fig). For the sediment grain size variables, there was a negative association for Alteromonadaceae) with mean grain size and gravel content (17.55 and 8.58 log₂ FC decrease, respectively, both p < 0.001; S3 Table, S4 Fig), but a positive association with the sand fraction with a 26.33 log₂ FC increase (p < 0.001; S3 Table, S4 Fig). Additionally, family GCA-2696645 was enriched with increasing sand fraction (24.19 log₂ FC increase), with a 28.26 log₂ FC decrease in the mud fraction (both p < 0.001; S3 Table, S4 Fig). Finally, we also found Enterobacteriaceae was negatively associated with increasing mean grain size (14.39 log₂ FC decrease, p < 0.001) and gravel content (6.12 log₂ FC decrease, p < 0.001; S3 Table, S4 Fig).

We found four shared families across the coral-associated and sediment only samples, with the largest abundance of Rhodobacteraceae (relative read abundance (RRA): 11.33 ± 0.35%), followed by Flavobacteriaceae (6.50 ± 0.19% RRA). There were 30 families of bacterial taxa that were found in all the coral-associated samples (representing 58% of total relative abundance), of these there were 26 exclusive to the coral-associated samples with four shared with the sediment controls (S3C Fig). There were no unique families for the sediment controls (S3C Fig).

Following the above analyses highlighting the differences between the Porites lutea coral-associated and sediment control samples, we present the remaining analysis with the coral-associated chambers only to further explore variation and environmental drivers of community structure.

Bacterial community structure of the porites lutea coral-associated sediment

To examine the coral-associated sediment community structure, we first performed a principal coordinate analysis (PCoA) based on weighted UniFrac distances and found clear separation of site (B6), where the first two axes explained 28.3% and 13.3% of the total variation (Fig 3A). PERMANOVA revealed coral surface area had a small but significant effect on bacterial community structure (R² = 0.027, F = 2.06, p = 0.027), whereas nested site and chamber replicate factors accounted for a larger proportion of variation (R² = 0.179, F = 1.94, p < 0.001). Together this suggests that microbial communities are predominantly influenced by site effects.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Patterns and predictors of sediment microbial community structure.

(A) Principal coordinates analysis (PCoA) of microbial community structure using weighted UniFrac distances. Each point represents a coral-associated chamber, coloured by site. Axes indicate the first two principal coordinates, with percentages reflecting the proportion of variance explained. (B) dbRDA for the coral-associated sediment samples with the environmental parameters displayed as vectors (black lines) driving the community structure. Asterisk (TN content) indicates statistical significance (p < 0.05) derived from a dbRDA back wise step model. (C) dbRDA for the coral-associated samples after removing four outlier samples, with the environmental parameters as vectors (black lines) driving the community structure and asterisks indicating statistical significance (p < 0.05) derived from a dbRDA back wise step model.

https://doi.org/10.1371/journal.pone.0346135.g003

Distance-based redundancy analysis (dbRDA) revealed clear differentiation between microbial communities from sites B6 and A1, with dbRDA1 and dbRDA2 explaining 31.2% and 14.2% of total variation, respectively (Fig. 3B). There was a significant effect for total nitrogen (F = 2.17, p = 0.024, Table 1, Fig 3B), explaining 4.1% of the total variation in microbial communities (adjusted R² = 0.041). After removing four outlier samples, dbRDA revealed gravel content (F = 5.73, p = 0.001), sand % (F = 2.74, p = 0.004), mean grain size (F = 2.77, p = 0.002) and total nitrogen (F = 3.69, p = 0.001; Table 1, Fig 3C) significantly influenced microbial community structure, collectively explaining 18.2% of the total variation (adjusted R² = 0.182).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary of distance-based Redundancy Analysis (dbRDA) results following backward model selection using ordistep. The table presents the environmental variables retained in the final model, their degrees of freedom (df), sum of squares, F statistic and significance level (p value) for (A) all Porites lutea coral-associated sediment samples and (B) a subset of the Potites lutea coral-associated sediment samples with four outlier replicates removed.

https://doi.org/10.1371/journal.pone.0346135.t001

Differential abundance of bacterial taxa in response to environmental covariates

Differential abundance testing of relative read abundance (RRA) data identified the number of bacterial families and the magnitude of significant responses among environmental variables. For gross photosynthesis and respiration rates, there were three notable bacterial families that significantly responded to these covariates. Specifically, Woeseiaceae, Pirellulaceae and Streptomycetaceae were significantly enriched with increasing gross photosynthesis and respiration (all p < 0.002; S3 Table, S4 Fig). There was a negative association for Rhodobacteraceae with a 6.8 log₂ FC decrease in relative read abundance (RRA) with increasing total carbon content (p < 0.001), whereas there was a positive association for total nitrogen, with a 9.07 log₂ FC increase in abundance with increasing total nitrogen (S3 Table, S4 Fig). We also found an 11.32 log₂ FC decrease (p < 0.001) in the RRA of Alteromonadaceae with increasing total nitrogen, but a 16.35 log₂ FC increase with increasing C:N ratio (p < 0.001). For the sediment grain size variables, there was a negative association for Alteromonadaceae with mean grain size and gravel content (17.55 and 8.58 log₂ FC decrease, respectively, both p < 0.001; S3 Table, S4 Fig), but a positive association with the sand fraction with a 26.33 log₂ FC increase (p < 0.001; S3 Table, S4 Fig). Additionally, GCA-2696645 was enriched with increasing sand fraction (24.19 log₂ FC increase), with a 28.26 log₂ FC decrease in the mud fraction (both p < 0.001; S3 Table, S4 Fig). Finally, we also found Enterobacteriaceae was negatively associated with increasing mean grain size (14.39 log₂ FC decrease, p < 0.001) and gravel content (6.12 log₂ FC decrease, p < 0.001; S3 Table, S4 Fig).

Enhanced metabolic activity in coral-associated sediments

The higher gross photosynthesis and respiration rates in the seawater contained in the P. lutea coral-associated chambers, compared with the seawater from the sediment only control chambers, suggests that coral proximity increases benthic metabolic activity. This may be due to increased inputs of organic matter close to coral colonies, as the result of corals’ shedding mucus. Coral mucus dissolves in seawater where it fuels the microbial loop or can sink to be degraded by sediment microbial communities, which results in the recycling of essential nutrients and energy back into the reef system [27]. This is also supported by the distinct bacterial communities found in the coral-associated versus the sediment control chambers (discussed in greater depth below), because coral mucus creates an ideal environment for microbial growth and represents one of the largest sources of organic matter in the coral reef ecosystem [28]. This is further supported by the dominance of medium to fine sands in our study, as permeable sediments typically facilitate higher rates of interstitial water exchange and microbial turnover [77,78]. Further laboratory and in situ studies would help clarify how coral-derived organic matter, sediment permeability, and microbial communities interact to regulate benthic metabolism and nutrient cycling.

Distinct bacterial community structure between coral-associated and control sediments

Bacterial communities differed significantly between coral-associated sediments and sediment-only controls, with the coral-associated samples exhibiting smaller within group variability and higher alpha diversity indices compared with the sediment-only controls. This suggests the P. lutea coral promotes a more biodiverse and compositionally consistent bacterial assemblage in the closely-associated sediment. This compliments the enhanced metabolic activity in the coral-associated chambers, where the coral exudates, mucus and organic matter create a nutrient enriched microenvironment in the closely-associated sediment that supports a broader range of microbial taxa, as discussed above. Furthermore, we found distinct bacterial communities in the coral-associated versus the sediment control samples, consistent with previous work describing niche specialisation by marine microbes [14,79]. Niche partitioning or variability in the microbial community composition and metabolic function, has been identified for coral, sediment and seawater [14,79]. Coral colonies have previously been shown to host diverse and species-specific microbial communities [80,81] that are distinct from those in adjacent seawater [14,81–83], and individual sand grains have been reported to harbour several thousand bacteria each [77]. Additionally, microbes associated with other benthic organisms on coral reefs have shown patterns of host specificity – for example, sponges harbour diverse and abundant microbial communities, which generally exhibit species specificity regardless of geography [84].

The dominant families Rhodobacteraceae, Flavobacteriaceae, and Woeseiaceae were shared across sample types, yet their relative abundance differed for both sample groups, highlighting microbial niche differentiation between coral-associated and adjacent sediment control samples. The contribution of Rhodobacteraceae as the dominant shared family across coral-associated and sediment control samples is notable, as this group is known for their role in organic matter degradation and sulfur cycling [85]. Despite no exclusive taxa for the sediment controls, compared with the 26 unique families for the coral-associated samples, we found the influence of the coral provides a distinct and diverse microbial assemblage.

The two most abundant families present across both sample types were Rhodobacteraceae and Flavobacteriaceae. The family Rhodobacteraceae, within the class Alphaproteobacteria, are widespread in marine habitats and are typically considered ecological generalists [85,86]. Bacteria of this family are categorised as extreme and moderate oligotrophs [87], and have relatively flexible requirements for carbon and energy metabolism, which may allow them to respond to a diverse range of environmental conditions [88]. HIMB11 is a member of the family Rhodobacteraceae, and despite its comparatively more streamlined and smaller genome size (indicating a more specialised lifestyle [89]), it has been found to be capable of degrading dimethylsulfoniopropionate [90], a major source of organic sulfur found in significant concentrations within corals, with the ability to harvest energy from carbon monoxide [91] and light [92].

The family Flavobacteriaceae are found in a broad range of marine and non-marine habitats and are key heterotrophic bacteria in marine ecosystems commonly found in sediment [93,94]. Regardless of their habitats, Flavobacteriaceae are generally considered as specialists for the degradation of complex organic matter, including polysaccharides [93], with these degradation products subsequently made available for other microorganisms [93]. The genus Aequorivita from Flavobacteriaceae plays a functional role in nutrient cycling and the breakdown of organic matter, with some strains associated with antimicrobial activity [95].

Microbial populations that are consistently present across assemblages [96], are thought to contribute to the overall health and metabolic functioning of their hosts [97]. Consistent with the overall taxonomic patterns described above, Rhodobacteraceae and Flavobacteraceae were consistently present in both the coral-associated sediment and sediment control (S3b Fig). Previous work has identified Rhodobacteraceae and Endozoicomonadaceae as key families associated with P. lutea [98,99]. Although Endozoicomonadaceae was not consistently present in the coral-associated sediment in our study, it was observed at low overall total relative abundance (0.04%) across these samples. The absence of Endozoicomonadaceae from the sediment only controls, supports previous work indicating coral host specificity [100], and its low abundance in coral-associated sediment may reflect localised seeding from the coral microbiome, potentially via mucus shedding. This further suggests corals can influence adjacent sediment communities, yet the spatial extent and ecological significance of this influence remain important areas for further investigation.

Although our approach in this study enabled us to identify unique bacterial communities for the coral-associated and sediment control samples, we acknowledge there may be a contribution from the associated communities in the overlying and interstitial seawater, which was not described in this study. Further, it’s also worth noting that the difference between replicates at each site could be due to the possibility of sampling clonal colonies at each site, and without genetic testing to confirm coral heads weren’t from the same parent colonies we are unable to confirm variability due to host genotype [101]. This may be an issue because differences in mucus production have been described between and within coral species [102], therefore genotypes potentially have different rates of mucus shedding.

Environmental drivers of coral-associated microbial community structure

Total nitrogen content was a key environmental driver explaining variation in coral-associated sediment bacterial assemblages, consistent with previous studies showing strong associations between microbial communities and sediment habitat properties [103]. While total nitrogen does not represent bioavailable nitrogen, it may covary with sediment characteristics such as permeability and organic matter content. Sediment permeability can influence nitrogen transformations and exchange with the overlying water column and has been shown to be positively associated with the bioactivity of calcareous sediments [30,33]. In permeable sediments, enhanced denitrification can lead to the loss of bioavailable fixed nitrogen from the sediment to the water column [29,30]. Furthermore, elevated total nitrogen may reflect increased organic matter content, which could indirectly influence the distribution of putative nitrogen-fixing bacterial taxa; however, because nitrogen speciation or production rates were not measured in the current study, these interpretations remain speculative. This aligns with previous work on marine sediments, where dose-dependent effects of total nitrogen were linked to shifts in microbial composition [104]. Although total nitrogen content only explained ~4% of community variation in our study, complex benthic systems such as the marine sediments surrounding coral colonies are typically shaped by multiple, interacting gradients [105,106]. This was evident from the increased explanatory power observed (~18% of the community variation) when sediment physical properties were included (after removing four outliers), highlighting that both sediment nutrients and physical characteristics influence microbial assemblages.

Differential taxonomic responses to environmental variables

Differential abundance analyses revealed key bacterial families exhibited distinct responses to environmental gradients, highlighting the ecological versatility of sediment microbial assemblages to changes in nutrient availability, productivity, and sediment properties (S3 Table, S4 Fig). The enrichment of Woeseiaceae, Pirellulaceae, and Streptomycetaceae with increasing gross photosynthesis and respiration within the chamber suggests that these taxa are metabolically linked to elevated organic carbon inputs and oxygen fluxes. Members of Woeseiaceae and Pirellulaceae have previously been reported as dominant families in marine sediment, including in Caribbean reef systems [41], and are often associated with oligotrophic marine sediments playing central roles in organic matter degradation and nitrogen cycling. Similarly, Streptomycetaceae are well-known decomposers in both marine and terrestrial environments, that increase in abundance under higher carbon turnover and aerobic conditions [107].

Nutrient related variables further structured the microbial assemblages, with Rhodobacteraceae and Alteromonadaceae exhibiting opposing responses to C:N ratios. The enrichment of Rhodobacteraceae under increasing nitrogen, but their decline with higher carbon, suggests divergent nutrient acquisition strategies that align with their known photoheterotrophic and denitrifying metabolisms [85]. In contrast, Alteromonadaceae increased under high C:N ratios but declined under elevated nitrogen, consistent with a copiotrophic, carbon-responsive lifestyle [108]. These patterns are supported by findings from deep-sea [108] and coral reef sediments [109] where the genus Pseudoalteromonas within family Alteromonadaceae thrives under carbon-rich, nitrogen-limited conditions. Despite the common occurrence of Pseudoalteromonas in sediments, little information exists on how they thrive in marine sediments [but see Qin et al., 107].

Sediment grain-size variables also exerted a strong influence on community structure, by influencing organic matter retention, oxygen penetration, permeability and the physical surface area available for microbial colonisation, all of which shape microbial habitats environment [110]. The decrease in Enterobacteriaceae with increasing grain size and gravel content suggests these facultative anaerobes prefer finer, organic-rich sediments, consistent with patterns in lake sediments [111]. Similarly, Alteromonadaceae increased with higher sand fractions and decreased in gravel-dominated fractions, highlighting their affinity for oxic, permeable substrates. Our findings also align with work in seagrass sediments, showing sediment physical structure is a key determinant of microbial community composition [112]. Taken together, the interaction between sediment physical properties, productivity and nutrient concentrations in shaping microbial communities is complex and requires further investigation to better understand how coral reef sediments regulate microbial diversity and ecosystem functioning.