Benthic organism exudate collections
Exudate collections from benthic organisms were conducted on board the R/V Walton Smith in November 2018 in Lameshur Bay, St. John, U.S. Virgin Islands within the Virgin Islands National Park. In brief, we collected six species of benthic organisms (n = 6 specimens), incubated these organisms in separate containers for 8 h, and harvested the incubation water to characterize the composition of dissolved metabolites in their exudates. A description of the exudate collections is included below (additional details available in Supplementary Methods).
Before each organism experiment, 58 l of surface (non-reef) seawater was collected ~1 mile offshore (18 17.127° N, 064 44.312° W, 31.6 m depth). Cells and particles were removed using peristaltic pressure through a 0.2 µm filter (47 mm, Omnipore, EMD Millipore Corporation, Billerica, MA, USA) using metabolomics-grade tubing and this filtrate (filtered seawater) was collected for the incubations. Additionally, two to three, 2 l filtrate subsets per experiment were acidified with concentrated hydrochloric acid (final concentration 1% volume/volume) and subjected to solid-phase-extraction (SPE) using a negative vacuum pressure of –3.7 to –5 100xkPA in Hg, to serve as controls. Before SPE, 6 ml, 1 gm Bond Elut PPL cartridges (Agilent, Santa Clara, CA, USA) were pre-conditioned with 6 ml of 100% HPLC-grade methanol.
For the experiments, six species of benthic organisms were collected from reefs around Lameshur Bay by SCUBA divers. Experiments were completed on three stony corals (Porites astreoides, Siderastrea siderea, and Psuedodiploria strigosa), two octocorals (Plexaura homomalla and Gorgonia ventalina), and one encrusting alga (Ramicrusta textilis) (Table S1). P. astreoides, S. siderea, and R. textilis were held in a seawater table for 24 h (hrs) before the incubations and colonies from the other three species were held for 2-3 h due to timing constraints. Coral and algal fragments were generally small (2.5-5.0 cm in length).
For each incubation, nine, acid-washed, 10 l polycarbonate bins (with lids) containing filtered seawater (4 l) were secured into an illuminated aquarium table (Prime HD, Aqua illumination, Bethlehem, PA, USA) (Photosynthetically Active Radiation = ~350–600 µmol quanta m−2 s−1). Air bubblers with sterilized Fluorinated Ethylene Propylene (FEP) tubing (890 Tubing, Nalgene, Thermo Scientific, Waltham, MA, USA) were used to inject air into each bin. Surface seawater was circulated through the aquarium table to maintain reef seawater temperature (29.5 °C). Six colonies/fragments of one species were randomly placed into 6 bins. The other 3 bins were reserved for control incubations containing filtered seawater only. A sensor (8 K HOBO/PAR loggers; Onset, Wareham, MA) monitored temperature and light conditions (data not shown). At the end of each 8 h experiment, colonies/fragments were wrapped in combusted aluminum foil and flash frozen in a charged dry shipper. The water in all incubations was re-filtered (as outlined above) and 2 l of each filtrate were acidified and subjected to SPE as described above. SPE cartridges were wrapped in combusted aluminum foil, placed in Whirl-Pak (Nasco, Madison, WI, USA) bags, and frozen at –20 °C.
Metabolomics analyses and data processing
At the Woods Hole Oceanographic Institution (WHOI), metabolites were eluted from the thawed cartridges into combusted, borosilicate test tubes using 100% methanol (Optima grade) within 3 months of collection. The eluents were transferred into combusted amber 8 ml vials and nearly dried using a vacuum centrifuge. Samples were reconstituted in 200 µL of 95:5 (v/v) Milli-Q (MQ, Millipore Sigma, Burlington, MA, USA) water: acetonitrile with a deuterated standard mix added as an internal control (Table S2), vortexed, and prepared for targeted and untargeted metabolomics analyses in both positive and negative ion modes as described previously [16]. Samples prepared for untargeted analyses were further diluted (1:200) with the reconstitution solvent. A pooled sample (technical replicate) was made by combining aliquots from all samples and was injected repeatedly to assess instrument drift over the course of the run and for downstream sample processing. Samples prepared for targeted metabolomics were analyzed using an ultra-high performance liquid chromatography system (UHPLC; Accela Open Autosampler and Accela 1250 Pump, Thermo Scientific, Waltham, MA, USA) coupled to a heated electrospray ionization source (H-ESI) and a triple stage quadrupole mass spectrometer (TSQ Vantage, Thermo Scientific), operated in selected reaction monitoring (SRM) mode. Samples prepared for untargeted metabolomics were analyzed with a UHPLC system (Vanquish UHPLC, Thermo Scientific) coupled to an ultra-high resolution mass spectrometer (Orbitrap Fusion Lumos, Thermo Scientific). MS/MS spectra were collected in a data-dependent manner using higher energy collisional dissociation (HCD) with a normalized collision energy of 35% (detailed methods provided in [16]). A Waters Acquity HSS T3 column (2.1 × 100 mm, 1.8 μm) equipped with a Vanguard pre-column was used for chromatographic separation at 40 °C for targeted and untargeted analyses. Sample order was randomized and the pooled sample was analyzed after every six samples.
For targeted metabolomics analysis, tandem MS/MS data files were converted into .mzML files using msconvert and processed with El-MAVEN [49]. Calibration curves for each compound (8 points each) were constructed based on the integrated peak areas using El-MAVEN. The concentrations of metabolites in the original samples were determined by dividing each concentration by the volume of the filtrate that passed through each PPL column. Finally, metabolite concentrations above the limits of detection and quantification were corrected for extraction efficiency using in-house values determined using standard protocols [50]. Statistical analyses of targeted metabolite concentrations were conducted using Welch’s independent t-tests and ANOVAs or Wilcoxon rank sum tests if data were not normally distributed (additional details in Supplementary Methods). We determined the mass of each colony and conducted Pearson correlations to investigate if colony size significantly correlated with concentrations of targeted metabolites, but no correlations were found.
For the untargeted metabolomics analyses, raw files containing MS1 and MS/MS data were converted into .mzML files using msconvert and processed using XCMS [51]. Ion modes were analyzed separately. Before processing with XCMS, the R package AutoTuner [52] was used to find XCMS processing parameters appropriate for the data. In XCMS, the CentWave algorithm picked peaks using a gaussian fit. The specific parameters for peak picking for both ion modes were: noise = 10,000, peak-width = 3–15, ppm = 15, prefilter = c(2,168.600), integrate = 2, mzdiff = –0.005, snthresh = 10. Obiwarp was used to adjust retention times and this step was followed by correspondence analysis. For statistical analyses, including permutational PERMANOVA adonis tests and non-metric multidimensional scaling analysis (NMDS), MS1 features (defined as unique pairings of mass-to-charge (m/z) values with retention times) in both ion modes were culled following XCMS if they: (1) had >1 average fold change in the MQ blanks compared to the other samples, (2) occurred in less than 20% of samples (excluding pooled controls), and/or (3) were invariant (relative standard deviation of <15%) across the samples [53]. This feature-removal strategy retained 27% (2317) of MS1 features ionized in positive mode and 84% (4725) of features ionized in negative mode. The R packages ggplot2 [54] and vegan [55] were used to visualize trends and inspect compositional differences among samples in the untargeted datasets using Bray-Curtis dissimilarity, non-metric multidimensional scaling analysis, and permutational PERMANOVA (adonis) tests with and without constraining permutations across the different incubations. Non-Euclidean NMDS ordination and permutational multivariate analysis of variance (adonis) methods were used because these approaches are well-suited for investigating compositional dissimilarities among samples in sparse, semi-quantitative datasets.
The MetaboAnalyst 5.0 web browser [56] was used to putatively identify enriched metabolic pathways in the untargeted MS1 features ionized in positive mode resulting from the XCMS processing (as described above) using the mummichog algorithm [57]. Prior to analysis, zero or missing feature intensities were replaced by the 1/5 minimum positive feature intensity and feature intensities were log10-normalized. Enrichment analysis of metabolic pathways was conducted by comparing control samples containing no organisms to organism incubation samples (all species). Additional details are reported in Supplementary Methods.
The .mzML files containing MS1 and MS/MS spectra from untargeted data analyzed in positive ion mode were interrogated using classical molecular networking (version release 28.1) [58] available through the Global Natural Products Social Molecular Networking (GNPS) database. Classical molecular networking was conducted using default parameters (pre-cursor ion mass tolerance = 2.0 Da, fragment ion mass tolerance = 0.5 Da, minimum cosine pair = 0.7, minimum matched fragment ions = 6) and the results can be viewed using the following link: https://gnps.ucsd.edu/ProteoSAFe/status.jsp?task=8abc6ce7fd334ff19eee1d0df12dcfe0. The average number of input MS1 features for classical molecular networking was 1014 with 4748 MS/MS spectra per sample. We also tested feature based molecular networking (version release 28) [59] using filtered MS1 data for both positive and negative ion modes. Raw and derived (.mzML) files from targeted and derived files from untargeted metabolomics analyses can be accessed in the MetaboLights database under accession MTBLS2855.
Exudate uptake experiments
To assess the lability of organism exudates, we examined the responses of reef seawater microbial communities to exudates from P. astreoides and G. ventalina. After the organism incubations described above, excess filtrate (2 l) from 3 of the 6 organism incubations (selected randomly) was pooled into an acid-washed 10 l carboy, and excess filtrate (~2 l) from the three control incubations was pooled into a second, acid-washed 10 l carboy.
For the P. astreoides experiment, surface seawater was collected from the offshore site and coarsely filtered through a GF/A filter (1.6 µm nominal pore size) using peristalsis to remove larger cells and minimize heterotrophic grazing, while retaining bacteria and archaea. Approximately 2.4 l of this inoculum was added separately to each 10 l carboy of either pooled coral or control filtrate to create a 5:2 ratio of filtrate: inoculum. After this addition, each carboy was mixed and a suite of samples were collected for different analyses including cell abundance enumeration, inorganic and organic macronutrient quantification, and 16 S rRNA gene sequencing. This initial collection was the first time point (0 h) for the exudate uptake experiment. For the G. ventalina experiment, reef seawater inoculum was collected from Tektite reef (Table S1).
For each experiment, coral metabolite or control filtrate seawater was transferred into 1 l acid-washed polycarbonate bottles (6 bottles per treatment). Within each treatment (coral and control), 3 of the 6 bottles were blackened to block light. The bottles were placed into a flow-through seawater table. PAR readings in the seawater table for the P. astreoides and G. ventalina experiments were 250–1000 and 164–530 µmol quanta m−2 s−1, respectively, with variation caused by changes in cloud cover. Over 48 h, samples were collected for cell enumeration (1 ml) at all time points (0, 12, 24, 36, and 48 h), macronutrient analyses (30–40 ml) at 0 and 48 h, and microbial community analyses (60–300 ml) at 0, 24, and 48 h.
Samples collected for cell enumeration were fixed to 1% (v/v) paraformaldehyde, refrigerated for 20 minutes in the dark, and frozen in a charged dry shipper. Abundances of Prochlorococcus, Synechococcus, picoeukaryotes, and unpigmented cells were enumerated via flow cytometry (Supplementary Methods). Cell abundances collected at 12, 24, 36, and 48 h were normalized using initial cell abundances at 0 h.
Samples (40 ml) were collected for total organic carbon (TOC) and total nitrogen (TN) analyses into combusted, borosilicate EPA vials and acidified using 75 µl of concentrated phosphoric acid. Samples were stored at room temperature for two weeks and then refrigerated at 4 °C until analysis. Samples were analyzed at WHOI using a Shimadzu TOC-VCSH total organic carbon analyzer with a TNM-1 module [60].
For inorganic nutrient analyses, seawater (30 ml) was allocated into acid-washed, polypropylene bottles (Nalgene) and frozen at −20 °C. These samples were shipped to Oregon State University and the concentrations of nitrite, nitrite + nitrate, ammonium, silicate, and phosphate were obtained using a continuous segmented flow system (described in [61]). Total organic nitrogen concentrations were obtained by subtracting the sum of the inorganic nitrogen species (nitrite + nitrate and ammonium) from the total nitrogen concentrations. Values measured below the detection limits of the instruments (ammonium = 0.02 μM, phosphate = 0.01 μM, nitrite + nitrate = 0.07 μM, nitrite = 0.01 μM) were reported as zero. Cell abundances and nutrient concentrations can be accessed via the Biological and Chemical Oceanography Data Management Office (BCO-DMO) under dataset accessions 865739 and 865193, respectively.
Seawater samples (60 ml for P. astreoides) collected for microbial community analyses were obtained using 60 ml sterile, Luer-lock syringes (Becton Dickinson, Franklin Lakes, NJ, USA) and filtered using positive pressure onto 25 mm, 0.2 μm pore-size filters. The amount of volume filtered was increased for the G. ventalina exudate experiment because of biomass concerns and ranged from 180 (0 and 24 h) to 300 ml (48 h). Previous work has demonstrated that sample volume does not significantly influence microbial community composition when beta diversity comparisons are made [24]. Alpha diversity analysis was avoided due to the differences in seawater volumes. Filters were then transferred into cryovials and placed in the charged dry shipper, followed by storage at –80 °C, until DNA extraction.
DNA was extracted from filters using the Qiagen PowerBiofilm extraction kit (Qiagen, Germantown, MD, USA) following default instructions. Five DNA extraction controls were created alongside the samples by performing the extractions without filter biomass. Purified DNA from a mock community (BEI Resources, NIAID, NIH as part of the Human Microbiome Project: Genomic DNA from Microbial Mock Community B (Even, Low Concentration), v5.1 L, for 16 S rRNA Gene Sequencing, HM-782D) was included in the sequencing library to assess PCR performance. Samples were amplified using the primers 515F-Y [62] and 806R-B [63] with conditions outlined in the Supplementary Methods and sequenced using a benchtop iSeq 100 sequencer (Illumina, San Diego, CA, USA). Two sequencing libraries were built to maximize sequence coverage per sample.
Across both sequenced libraries, the average number of initial reads in non-control samples was 139,508 ± 55,490 (standard deviation). To analyze the sequencing data, all the demultiplexed fastq files across both libraries were compiled and run together using the DADA2 R package [64]. Reverse reads were dropped from the analysis due to lower read quality and concatenation difficulties due to the shorter reads. Based on previous work comparing phylogenetic resolution from different lengths of V4 reads [65], our general taxonomic results are likely comparable between the iSeq and MiSeq-based (described below) results. Forward reads were then filtered using the ‘filterAndTrim’ parameter (trimLeft = 19, truncLen = 145, maxN = 0, maxEE = 1, rm.phix = TRUE). Prior to inferring amplicon sequence variants (ASVs), error rates were obtained and screened. Chimeras were removed (~2% of all remaining sequences after the filtering), and the average number of reads per sample after the filtering, denoising, and chimera removal steps dropped to 123,867 (14,703–281,252 range). The average number of reads in the 5 DNA extraction control samples was 3351(1092–6600 range) and the negative PCR control had 1471 reads. Taxonomy was assigned using the Silva v.138 database [66] and the “addSpecies” function was used to assign more specific taxonomy designations, resulting in 255 species-level assignments out of 7364 ASVs. Mock community performance was also assessed, and 43 and 26 ASVs were detected in the sequenced mock community samples. All ASVs (consisting of forward reads only) detected in the mock communities were continuous partial matches to the longer sequence reads included in the mock reference file. The R package decontam [67] was used to remove contaminating sequences from the samples using either their prevalence or frequency in the 5 DNA extraction controls, decreasing the number of ASVs from 7364 to 7333. Samples with <7000 sequences and sequences identifying as NA or uncharacterized at the phylum level were removed. Sequences identifying as chloroplasts at the order level were also removed. After conducting these filtering steps, we detected an average of 388 ASVs (standard deviation ±227 ASVs) across control and coral exudate addition treatments.
The R package corncob [68] was used to identify ASVs that significantly covaried based on differential abundance with treatment type (coral exudate addition vs. control) at each individual time point across the two experiments. Corncob was run using the Wald test with an FDR cut-off of 0.05. Sample counts were transformed to relative abundance for taxonomic comparisons and boxplots were generated to verify the corncob results. To summarize these results, ASVs were selected that displayed consistent patterns of increase or decrease in relative abundance over time. In addition, only ASVs with coefficients of variation less than –2 (depleted in coral exudate additions) and greater than 0 (enriched in coral exudate additions) were included in this summary analysis to focus on ASVs displaying the most distinct changes. The relative abundance of each ASV was averaged across replicates and normalized to the initial abundance of that ASV at the first time point (0 h). Using these averages, fold changes (coral addition/control) were calculated for each ASV and the log10 of the fold changes were plotted. To prevent the formation of infinite values, relative abundances of 0 were substituted with low values (1e-7). Non-metric multidimensional scaling was completed on the Bray-Curtis dissimilarity matrix to compare overall compositional differences between samples. Adonis tests were completed on the Bray-Curtis dissimilarity matrix to test which factors significantly influenced dissimilarity between samples. The raw fastq files containing sequences for these samples can be accessed on the NCBI Sequence Read Archive (SRA) under BioProject PRJNA739882.
Metabolite uptake incubations
Three metabolite uptake incubation experiments were conducted in St. John, USVI in January 2021 based on the results of the benthic organism incubations. These experiments assessed if reef seawater microbial communities would respond distinctly to different metabolites. Experiments were conducted with three metabolites (riboflavin, pantothenic acid, and caffeine) in separate incubations occurring on different days. The B vitamins riboflavin and pantothenic acid were chosen because they were released by at least 4 of 6 species during our benthic organism incubations. Caffeine was also selected because it was released in high quantities by the invasive alga R. textilis. All three metabolites were also detected and quantified in seawater collected from reefs in the Jardines de la Reina reef-system in Cuba [16]. Samples were collected at three time points over 24 h: 0, 6, and 24 h. Incubations were conducted in the dark and processed in low-light conditions to minimize photodegradation of the metabolites, especially riboflavin.
Metabolite standards were diluted with MQ water to 5 nM (riboflavin, pantothenic acid) and 10 nM (caffeine) one month prior to the incubations, frozen, and shipped to the USVI. Metabolite dilutions were prepared in combusted amber vials and kept in the dark to minimize photochemical degradation. To set up the incubations, 5.5 l of seawater was collected approximately 1 m above Tektite reef (9 m depth) using Niskin bottles deployed by divers. Half of the seawater was filtered using a 0.1 µm, 47 mm Omnipore filter to create filtered seawater, while the rest of the seawater was filtered using a 1 µm, 47 mm Omnipore filter to remove larger phytoplankton and protistan grazers but retain picoplankton. The filtrate treatments were poured into 36 acid-washed and autoclaved 125 ml polycarbonate bottles. To account for small volumes and minimize the chance of sample contamination, incubations were designed so that bottles could be sampled by sacrifice at each time point. Each incubation had four different experimental conditions: 3 different controls and one experimental treatment. The three controls included filtered seawater (F), filtered seawater with the addition of a metabolite spike (F + Mtb), and 1 µm filtered seawater containing microbes (M). The experimental condition was 1 µm filtered seawater containing microbes with the addition of the metabolite (M + Mtb). Metabolite spikes (20 pM for riboflavin and pantothenic acid and 70 pM for caffeine) were added to the F + Mtb and M + Mtb treatments by removing 500 µl (riboflavin and pantothenic acid incubations) or 875 µl (caffeine incubation) of seawater from each incubation bottle, replacing the lost volume with the appropriate metabolite spike, and inverting the bottles to mix. After the metabolite spikes were added, 24 of 36 incubation bottles were placed in a flow-through seawater table covered with a tarp and equipped with a HOBO data logger to monitor relative light levels and water temperature (data not shown), while the remaining 12 bottles, accounting for triplicate bottles of each of the four experimental conditions, were immediately processed for the initial time point.
After each time point, samples from the incubation bottles were re-filtered using 0.2 µm, 47 mm Omnipore filters and filtrate was collected and acidified with 125 µl of concentrated hydrochloric acid. The filters were placed into cryovials and frozen in a dry shipper. All samples were filtered and acidified within one hour of collection. SPE was conducted on filtrate using the methods outlined above. Bottle masses were obtained prior to and after SPE to account for sample volume. Additionally, subsamples of the initial and final incubation samples were obtained to assess microbial cell counts across the different treatments using flow cytometry (described in Supplementary Methods).
Metabolite uptake incubation data processing and analyses
The previously frozen PPL cartridges were eluted using 6 ml of 100% methanol within 3 weeks of sample collection and samples were prepared for targeted metabolomics analysis in positive ionization mode. Samples were dried, re-suspended in 200 µl of 50–150 ng deuterated standard mix dissolved in MQ water (Table S2), vortexed, and aliquoted into pre-combusted analysis vials containing inserts. A pooled sample was created for instrument quality control and to make a matrix-matched standard curve. To make the 11-point matrix-matched standard curve, 50 µl of pooled sample was combined with MQ water and various concentrations (ranging from 0.5 to 1000 ng ml–1) of metabolite mixes prepared in MQ water. Instrument conditions and data processing were identical to the methods outlined above. The targeted raw and .mzML files can be accessed on the MetaboLights database under accession MTBLS3286.
To generate the sequencing library for microbial community analysis, DNA was extracted from filters using the PowerBiofilm DNA extraction kit following the standard protocol. Seven DNA extraction controls were created to test for contamination. Amplification was conducted with the 515F-Y [62] and 806R-B [63] primers and sequencing occurred using Illumina MiSeq 2×250 bp, with additional details recorded in Supplementary Methods.
Trends in microbial abundances of Prochlorococcus, Synechococcus, picoeukaryotes, and unpigmented cells (~heterotrophic bacteria and archaea) yielded via flow cytometry were inspected using line graphs and standard error was calculated across replicate samples for each time point. Microbial community analysis was completed using the methods outlined above for the exudate uptake experiments including general compositional analysis as well as specific ASV enrichment analysis using corncob. An average of 232 ASVs (standard deviation ± 135) per sample were recovered. The data is available on NCBI Sequence Read Archive (SRA) under BioProject PRJNA739882. Cell abundances can be accessed via BCO-DMO under the project entitled “Signature exometabolomes of Caribbean corals and influences on reef picoplankton” (dataset accession 865159).
Source: Ecology - nature.com
