Coral-associated nitrogen fixation rates and diazotrophic diversity on a nutrient-replete equatorial reef

The role of diazotrophs in coral physiology and reef biogeochemistry remains poorly understood, in part because N2 fixation rates and diazotrophic community composition have only been jointly analyzed in the tissue of one tropical coral species. We performed field-based 15N2 tracer incubations during nutrient-replete conditions to measure diazotroph-derived nitrogen (DDN) assimilation into three species of scleractinian coral (Pocillopora acuta, Goniopora columna, Platygyra sinensis). Using multi-marker metabarcoding (16S rRNA, nifH, 18S rRNA), we analyzed DNA- and RNA-based communities in coral tissue and skeleton. Despite low N2 fixation rates, DDN assimilation supplied up to 6% of the holobiont’s N demand. Active coral-associated diazotrophs were chiefly Cluster I (aerobes or facultative anaerobes), suggesting that oxygen may control coral-associated diazotrophy. Highest N2 fixation rates were observed in the endolithic community (0.20 µg N cm−2 per day). While the diazotrophic community was similar between the tissue and skeleton, RNA:DNA ratios indicate potential differences in relative diazotrophic activity between these compartments. In Pocillopora, DDN was found in endolithic, host, and symbiont compartments, while diazotrophic nifH sequences were only observed in the endolithic layer, suggesting a possible DDN exchange between the endolithic community and the overlying coral tissue. Our findings demonstrate that coral-associated diazotrophy is significant, even in nutrient-rich waters, and suggest that endolithic microbes are major contributors to coral nitrogen cycling on reefs.


Incubation set-up
Due to strong vertical gradients in light at these reefs, all coral fragments were collected and acclimated within ± 1 m depth from their original colonies. Corals were fragmented with a hammer and chisel, attached to a PVC rig using cable ties. The PVC fragmentation rig was coated in food-grade paraffin wax (Gulfwax) to prevent fouling.
Incubation jar lids were modified to incorporate two rubber septa (Wheaton W224100-173) and sealed with silicone, as well as a PVC attachment to hold the coral in place. Corals were retrieved from the reef, placed on new PVC attachments, rinsed briefly with filtered seawater to remove loosely associated organisms, and placed in chambers with filter-sterilized seawater. All incubation chambers were closed while submersed in water to prevent air contamination.
The 15 N 2 label was prepared by degassing 0.22 µm filtered seawater at 920 mbar for 1 hour [1,2]. Water was transferred headspace-free to serum bottles (Wheaton 223952) by siphon and crimp-sealed (Wheaton 224187-01) with rubber septa (Wheaton W224100-173). 15 N 2 gas (Cambridge Isotopes NLM-363-1LB, Lot#: I-21065/AR0664729) was added in a ratio of 2ml/100ml to each serum bottle. Serum bottles were inverted during dissolved gas addition and displaced water was captured using a separate outflow needle and syringe. After gas addition, additional filter-sterilized, degassed seawater was injected via the septa to increase pressure in the serum bottle. Serum bottles were shaken for 5 minutes and stored on their sides at room temperature for 24 hours. Final enrichment levels of all serum bottles were >90% 15 N 2 . Glassware, tubing, and septa used throughout the experiment were acid-washed with 10% HCl and rinsed throughly with ultrapure water (Elga, 18.2MΩ). More details can be found at protocols.iref.o/researchers/ molly-moynihan.

DDN assimilation calculations
To determine diazotroph-derived nitrogen (DDN) assimilation rates, particulate atom % 15 N from paired experimental (A e ) and control (A c ) replicate samples were subtracted for each fraction (i.e. host tissue, algal symbionts, released mucus, skeleton, and seawater). Results were only considered significant if this difference was greater than 3x the standard deviation of repeated A c measurements, following [3,4,5]. Rates below the detection limit were assumed to be zero when calculating average rates, resulting in more conservative estimates. V (T −1 ) is defined as the particulate difference (A e -A c ) divided the amount of dissolved gas label (A 15N (refs. 5-6), according to the sample type. DW x is the dry weight of the specific fraction (host tissue, algal symbionts, released mucus, and seawater particulate organic matter), and ρ s is the skeletal density (g cm −3 ). Genera-specific density values from Singapore were compiled from Januchowski-Hartley et al. [7]. For coral nitrogen budget calculations, unseparated tissue (host and symbiont) DW x cm −2 was multiplied by %N to determine N content.
1.3 Nucleic acid extraction, PCR, sequencing and bioinformatics DNA was extracted from coral tissue and skeletal samples using the DNeasy PowerBiofilm kit (Qiagen 24000-50, Hilden, Germany). The following modifications were made based on Sunagawa et al. [8]: (i) after the addition of lysis solution (FB), 10 µl of lysozyme (12.5mg/ml) were added to each reaction and samples were incubated at room temperature for 10 minutes; subsequently, (ii) 20 µl of proteinase-K (20mg/ml) were added and samples were incubated at 65°C for 60 minutes. The remaining protocol followed the manufacturer's instructions. From seawater samples, DNA was extracted from Sterivex filters (1 l seawater) using a modified phenol-chloroform method with the DNeasy PowerSoil kit (Qiagen 12888-50). After 2x lysis incubations with solution C1 (70°C, 500rpm, 10 minutes), 700µl of phenol-chloroform-isoamyl alcohol were added to each reaction (adapted from [9]). Samples were vortexed for 10 minutes and centrifuged at 10,000 x g for 30 seconds. The resulting supernatant was transferred to a clean tube, and remaining steps followed the manufacturer's protocol. RNA was extracted using Trizol (Invitrogen 15596018, Carlsbad, CA, USA). After phase separation, RNA was precipitated for 10 minutes at room temperature with 250 mL isopropanol, 250 mL high salt solution (0.8 M Na citrate, 1.2 M NaCl), and 1 µl GlycoBlue coprecipitant (Ambion AM9515, Austin, TX, USA) for every 1 mL of Trizol. After solubilization, RNA was cleaned and concentrated (Zymo, R1017). Samples were checked for DNA contamination by performing PCR with 16S and 18S rRNA primers (see below). cDNA was synthesized from RNA using the SuperScript VILO Master Mix (Invitrogen 11766050) following manufacturer's instructions. Both cDNA and no-reverse transcription controls were amplified, along with extracted DNA, following the conditions below. RNA samples with DNA contamination were treated with ezDNase (Invitrogen 11766051) and re-amplified. Protocols of sample preparation, nucleic acid extractions, and PCR conditions can be found online (protocols.io/researchers/molly-moynihan).
Negative controls were sequenced and used to control for contamination using the prevalence method of decontam with a threshold of 0.1 [22]. Additional seawater samples, processed together with the seawater samples used in this study, were included in the decontam step to avoid bias in identifying contaminants. In the 16S rRNA dataset, 22 ASVs were identified by decontam and removed. No contaminating sequences were identified in the 18S rRNA and nifH datasets.
The nifH phylogeny was made using reference sequences from Raymond et al. [23], Meheust et al. [24] and Heller et al. [25]. Cluster definitions were based on Meheust et al. [24]. The representative sequences of nifH ASVs (3,632) were translated into protein sequences using Geneious Prime 2019.2.3 with standard genetic code (https://www.geneious.com), and correct reading frames were checked manually for all sequences. Sequence alignment was performed with the MAFFT Alignment plugin with the following settings: auto algorithm, BLOMSUM62 scoring matrix, a gap open penalty of 1.53, and an offset value of 0.123. Cluster assignment done manually by comparing groups of aligned ASVs (≈500 ASVs) with reference sequences [23,24]. In total, 5.08% of ASVs belonged to Cluster I, 0.66% to Cluster II, 4.15% to Cluster III, 1.11% to Cluster IV, 88.91% to Cluster V, 0.01% to Cluster VI, and 0.08% were not assigned a cluster. To make the final tree (Fig. 4), all Cluster I -III ASVs from coral DNA-and RNA-based samples were added to the tree, as well as ASVs from Clusters VI-VI with a normalized abundance ≥ 200 for each species, site, and sample type (i.e. tissue or skeleton), respectively. The tree was assembled using the FastTree plugin. Both FASTA formatted sequence alignment and Newick files can be found on https://github.com/moyn413/Singapore-coral-microbes. Table S1: Rate measurements of nitrogen fixation in corals using the acetylene reduction assay (ARA), 15 N2 bubble method, or 15 N2 dissolution method. The 15 N2 bubble method underestimates rates due to a delay 15 N2 in equilibration [1]. Dissolved    4 ] and N:P as the ratio of DIN to dissolved inorganic phosphorus (DIP) (i.e. phosphate). Samples were collected at 5 m depth. Temperature, salinity, and chlorophyll-a were measured at the experimental site using a CTD profiler at 5 m depth. The daily light integral (DLI) (mol photons m −2 d −1 ) of photosynthetically active radiation (PAR) was measured using loggers deployed approximately 1 m above the incubation site (3-4 m depth) [44]. ( ⋆ ) denotes the date water was collected for filter-sterilization, ( ○ ) denotes the date water was sampled for water column incubations and that filter-sterilized water was used in coral incubations.     [3,4]). Rates below the detection limit were considered to be zero when determining replicate averages. Here, rates are also presented only considering rates above the detection limit ( 15 N excess >0). Table S6: Reef-scale estimations of N 2 fixation biogeochemical significance. Reef surface area was estimated using Google Earth and multiplied by an average water column depth of 5 m to determine the water column volume over the reef. Coral cover was measured by Bauman et al. [45]. Effective hard coral surface area was estimated by multiplying the surface area by 2.2 to account for macro-scale rugosity [7] and by 4 to account for micro-scale rugosity (effective coral surface area) [47]. Released mucus-associated and water column N 2 fixation rates are based on data in this study. DIN at each site is provided in the Hantu Calculations: