Turf algae redefine the chemical landscape of temperate reefs, limiting kelp forest recovery

In temperate regions experiencing rapid ocean warming, kelp forests are being replaced by chemically-rich turf algae. Yet, the extent to which these turf algae alter the surrounding chemical environment or impact the rebound potential of kelp forests via allelopathy remains unknown. Here, we used underwater visual surveys, comprehensive chemical profiling, and laboratory experiments to reveal that turf algae release bioactive compounds into the water that fundamentally alter the reef “chemical landscape” and directly suppress kelp recruitment. Our study, therefore, reveals that allelopathy is critical in shaping modern kelp forest ecosystems and their resilience. Further, it demonstrates that reversing climate-driven state shifts will require not only curbing global carbon emissions, but also targeted local interventions that break harmful ecological feedback loops and foster recovery.

Methods
Quantifying seaweed community structure on the reef
Files

Algal_Biomass.csv

Kelp_Recruitment_Counts.csv

Turf_Cover.csv

At each site, SCUBA divers deployed a 40-meter transect on the reef, set perpendicular to shore and contouring the 5-7 m depth isobath (mean lower low water). Within replicate 1 m2 quadrats deployed at set intervals along the transect (n = 8 per site), we visually counted the number of juvenile kelp recruits. Then, in a portion (0.25 m2 area) of each quadrat, we estimated the abundance of turf algae (percent cover). To identify the species composition of the turf algae community (which can be difficult to discern underwater), we next harvested all kelps (in the 1 m2 area) and all low-lying seaweeds residing under the canopy, including bladed, foliose, and turf algae species (in the 0.25 m2 area) from a subset of quadrats (n = 4-6 per site). Collected seaweeds were kept cool on the boat and brought back to the lab within 6 hours, where they were sorted to species, spun (20 revolutions in a salad spinner), and weighed to estimate their biomass.

Dissolved organic matter and seaweed tissue collections
Files

DOM_Tissue_meta.csv

Water collection for nontargeted metabolomics

To characterize the metabolome of each site, in 2022, we collected water samples containing dissolved organic matter (DOM) and subjected them to non-targeted metabolomics (see below). We studied all six sites (see above) within a 4-week period to avoid conflating site with seasonal effects. Water samples (1L, n = 6 per site) were collected within the first 6 survey quadrats (see above), making them spatially linked. Water samples were taken from within the seaweed matrix (0.5 – 3 cm above the reef) using a custom designed benthic organic matter sampler (Fig. S7). These watertight cylinders – made from chemically inert HDPE and Teflon – were filled with air on the surface, and the inlets on each end were closed. Underwater, once on the benthos, both inlets were opened, with air escaping from the top, creating suction on the bottom to collect the DOM. Samples were brought to the surface, stored on ice, and transported to the lab to be immediately filtered and extracted. Additional samples were collected from the benthos in May 2022 (n = 36), and from mid-water stations (1-2 meters above the reef) in both May and August 2022 (n = 72); data from these samples were used to augment our chemical reference library (i.e., help increase feature annotations and maximize annotation propagation within feature-based molecular networks) but were not included in our study or core analyses because they were not linked in space or time with our questions of interest

Seaweed tissue collections for nontargeted metabolomics

To characterize the internal chemistry of seaweed and whether this chemistry is also found in the waterborne reef metabolome (i.e., exuded into the surrounding seawater), divers haphazardly collected and individually bagged the dominant seaweed species at each site. Care was taken to collect only clean thalli with few to no epiphytes. Seaweeds were brought aboard the boat, immediately rinsed with raw seawater, cleaned of any epiphytic organisms, placed in a precleaned and muffled 20 mL scintillation vial, and placed on dry ice to stop metabolic activity.

Generating MS/MS Data
Files

DOM_Tissue_MS_MS_Raw_Data.zip

Turf_Extracts_MS_MS_Raw_Data.zip

Sample preparation for UHPLC-MS/MS

Reef metabolome (i.e., DOM) extracts were re-dissolved in 100 µL methanol (LC-MS grade) and 1% formic acid (LC-MS grade). Two standards were created to account for instrument drift and batch effects – an internal positive control containing six synthesized compounds and a pooled standard containing 1 µL of 50 DOM samples combined into one. Seaweed tissue extracts were redissolved in methanol (LC-MS grade) and 1% formic acid (LC-MS grade) and diluted to 50 mg/mL. For seaweed chemical analysis, the same internal positive control was used, while a seaweed tissue pooled standard was created by combining 1 µL of 50 seaweed tissue extracts. Samples were randomized before being subjected to UHPLC-MS/MS.

UHPLC-MS/MS

We subjected our water and seaweed tissue samples to UHPLC-MS/MS using previously developed methods (64). For chromatographic separation, we used a C18 core-shell column (Kinetex, 150 × 2.1 mm, 1.8 µm particle size, 100 A pore size, Phenomenex, Torrance, USA) with a flowrate of 0.5 mL/minute (Solvent A: H2O + 0.1% formic acid (FA), Solvent B: Acetonitrile (ACN) + 0.1% FA). After injection, the samples were eluted with a linear gradient from 0 to 0.5 minutes, 5% B, 0.5 to 8 minutes, 5 to 50% B, 8 to 10 minutes, 50 to 99% B, followed by a 3-minute washout phase at 99% B and 3-minute re-equilibration phase at 5% B.

Electrospray ionization (ESI) parameters were set as follows: Gas flows were 50 L/minute for sheath, 12 L/minute for auxiliary, and 1 L/minute for sweep. The auxiliary gas temperature was 400°C. The spray voltage was set to 3.5kv with the inlet capillary at 250°C. Additionally, a 50 V S-lens was applied. For the full scan (MS1) acquisition, the scan range was 150–1,500 m/z with a resolution at m/z 200 (Rm/z 200) of 120,000 with one micro-scan, and the scan polarity was set in positive mode. Automated gain control (AGC) was set to 1.0E6 with a maximum ion injection time of 100 milliseconds. MS/MS spectra were recorded in data-dependent acquisition (DDA) mode (65). In addition to MS1 survey a maximum of 5 MS/MS scans of the most abundant ions per duty cycle were measured with Rm/z 200 of 15,000 with one micro-scan. Automatic gain control targets were set to 5.0E5 with a minimum 10% C-trap filling for MS/MS. MS/MS precursor selection windows were set to m/z 1. The normalized collision energy was increased from 25 to 35 to 45%, with z = 1 as the default charge state. An apex trigger was applied to MS/MS experiments with 2-15 seconds from their first occurrence. Dynamic exclusion was set to 5 seconds. Ions with unassigned charge states were excluded from DDA and isotope peaks.

UHPLC-MS/MS: extracts of turf algae exudates

To explore potential bioactive compounds in the extracts of waterborne exudates from turf algae (which caused significant kelp gametophyte mortality), we subjected each extract to UHPLC-MS/MS using the same methods as described above. Post-processing followed the same procedures using MZmine3, SIRIUS, and GNPS (Datefile S2).

Generating the GNPS Quantification table using mzmine3
Files

DOM_Tissue_GNPS_quant.csv

Post-processing of UHPLC-MS/MS data (MZmine3 and SIRIUS)

Before post-processing, the 6 compounds that made up the internal positive control were assessed to account for mass to charge (m/z) and retention time (RT) shifts (Figs. S8-S9). Thermo.raw datasets were converted to .mzXML in centroid mode using MSConvert (66). Centroided data were processed in batch mode with MZmine3 for feature extraction, characterization, and quantification (32). Noise levels for MS1 and MS2 mass detection were 2.0E5 and 1.0E3, respectively. Ion chromatograms were built with a minimum group size of 3, group intensity threshold of 5.0E5, minimum peak height of 1.0E6, and relative mass tolerance of 3 ppm. Chromatographic deconvolution was performed using a local minimum resolver, with a chromatographic threshold of 80%, minimum peak height 1.5E6, minimum ratio of peak top/edge 1.5, and peak duration between 0.01 - 5 minutes. For isotope peak grouping, mass and retention time tolerances were set to 3 ppm and 0.1 minutes, respectively, with a maximum charge of 2. Extracted ion chromatograms were aligned using the join aligner with the same mass and retention time tolerances as above. Only extracted chromatograms that contained 2 isotope peaks and occurred in 5 samples were considered. The peak list was further refined using a duplicate peak filter with a mass and retention time tolerance of 3 ppm and 0.1 minutes, respectively. Finally, the gap-filling function was used with intensity tolerance at 20%, mass tolerance of 3 ppm, and retention time tolerance of 0.1 minutes. The MZmine3 output quantification table and .mgf files were used to create a feature-based molecular network (FBMN) in GNPS (31).

Generating the formula identification and canopus compound predictions from SIRIUS.
Files

Formula_identifications.csv

Canopus_Predictions.csv

Canopus_formula_summary

Post-processing of using SIRIUS

SIRIUS (5.6.3) was used to annotate and predict molecule identities using the tandem mass spectrometry data (35). Using the SIRIUS module, molecular formulas were computed by matching experimental with predicted isotopic patterns from the fragmentation trees analysis of MS2. Parameters for SIRIUS were as follows: Instrument: Orbitrap, MS/MS ppm: 5, Isotope scorer: ignore, Candidates stored:10, Min candidates per Ion: 10, Databases used: no selections, Possible Ionizations: Pos, Tree timeout:0, Compound timeout:0, Use heuristics above m/z: 300, Use heuristics only above m/z: 650. Furthermore, in silico structure annotations were obtained with CSI:FingerID (33), using Bio Database, while class annotations were obtained with CANOPUS with the NPClassifier ontology (36).

Generating molecular feature identification using GNPS 
Files

Library_hits.csv

Unique_Library_hits.csv

DOM_Annotations_318.csv

Bioactive_Compounds_from_turf_extracts.csv

GNPS Jobs for both the DOM and tissue samples, along with the waterborne/surface extracts, can be found at:

DOM/Tissue

(https://gnps.ucsd.edu/ProteoSAFe/status.jsp?task=bee692af684e4097b83bde2f8e12ab10)

Turf Extracts

https://gnps2.org/status?task=5843f1ba09ce4157ab8f854d8e860fe5

Collecting seaweed, generating extracts, and conducting (turf seaweed vs gametophyte) allelopathy trials.
Allelopathy trials: effects of the metabolomes from kelp-dominated and turf-dominated reefs

To quantify the effects of the reef metabolome on kelp recruitment, we combined DOM collected from each study site with kelp gametophytes at natural concentrations (1.1-1.8 µg/mL) and assessed the effects of DOM on gametophyte survival, relative to controls. For each site, we combined 200 µL from each of the 6 DOM extracts (see Water collection for nontargeted metabolomics) into a pooled, site-level extract (to account for within-site variation). Ecologically relevant concentrations (µg/mL) were calculated by dividing the total volume of water collected by the dried extract mass. DOM extracts (i.e., treatments) were resuspended in methanol, and a 0.25 µL aliquot was combined with ~20 gametophytes in a single well of a 96-well flat bottom plate, along with 300 µL of pre-made sterile seawater (i.e., Instant Ocean, MilliQ water, and nutrients f/2 which had been autoclaved to prevent microbial contamination). We used sterile seawater to avoid confounding factors (such as bacterial activity, pathogens, or variation in nutrient concentrations in raw seawater) that could have obscured our ability to directly test and quantify the effects of allelopathy in this ecosystem. Controls – SPE extracts of 1L of water (LC-MS grade) that were filtered at the same time as DOM samples – were resuspended in methanol, and a 0.25 µL aliquot was combined with ~20 gametophytes and 300 µL pre-made sterile seawater (n = 10 replicate wells per treatment and control). Assays were kept in a 10°C incubator in the dark for 96 hours, at which time they were scored (see below). Clonal male gametophytes were cultivated at the Aquaculture Research Institute (ARI) at the University of Maine’s Darling Marine Center.

Collecting turf algae for species-specific allelopathy trials

We collected the five most abundant species of turf algae (Polysiphonia stricta, Dasysiphonia japonica, Ceramium spp., Vertebrata fucoides, and Melanothamnus harveyi) for use in allelopathy trials (see below). These species comprised 83% of the total turf algae biomass we have observed across seasons and years in this ecosystem (2021-2023; unpublished data), indicating that they consistently underpin the turf algal community in this system. Collections occurred in August 2023 to coincide seasonally with DOM sampling, making our evaluation of the chemical effects of these turf algae on kelp gametophytes spatially and temporally linked to the metabolome trials (see above). For collections, seaweeds free of epiphytic organisms were haphazardly collected, kept in fresh seawater in the dark, and returned to the lab for extraction.

Generating waterborne metabolite extracts from turf algae, for allelopathy trials

To obtain waterborne turf algae exudates, 35.2 g of a given turf algae species was placed in a glass container with 1L of artificial seawater (Instant Ocean and MilliQ water) and incubated at 12°C for 1 hour. To calculate this ecologically relevant concentration (grams of algae/volume seawater), we first calculated the average total biomass of turf algae at our turf-dominated sites (2021-2023) and standardized this value to 100% cover. We then estimated the active space of waterborne chemistry on the reef – in terms of volume per m2 – to be 1m x 1m x 0.015m (15L), resulting in a concentration of 528 g algae/15L (or 35.2 g/L). The conditioned water was then filtered through a GF/C filter (1.2 µm, 120 mm diameter) and split into two 500 mL portions to limit saturating the SPE cartridge. The filtrates were then extracted via two 0.2 g bed mass SPE PPL cartridges (as described above). Both cartridges were eluted with 2 mL methanol into a 20 mL scintillation vial and dried. The two extracts were resuspended in 2 mL of methanol, combined, and transferred to a pre-weighed 2 mL HPLC vial and dried under vacuum.

Generating surface-bound metabolite extracts from turf algae, for allelopathy trials

To generate turf algae surface extracts, we first determined an appropriate ratio of hexane:dichloromethane (LC-MS grade) that would extract surface-bound molecules – but not lyse the cell walls of turf algae – via dipping each species in increasing concentrations of DCM (0%, 2%, 4%, 6%, or 8%) for 30 seconds. The resultant turf algae were stained with Evans Blue, and cell lysis was quantified using a Leica DMi8 microscope with 640nm light for excitation and a 700nm emission filter to visualize the Evans Blue fluorescence using a low-powered (5x) objective lens and a Leica DFC9000GT sCMOS camera. Images were captured using LASX software. If cell lysis occurred, algae would take up the cell impermeable stain. Once appropriate ratios were determined, surface-bound molecules were extracted by dipping whole thalli (n = 5 for each species) in 60 mL of a hexane / DCM mixture (percentages based on the above tests) for 30 seconds with agitation. Each thallus was then dried to remove excess water, weighed, and spread flat for imaging to determine surface area. Extracts were dried via rotary evaporation. To remove residual salt carryover, we partitioned each extract using 5 mL ethyl acetate (LC-MS grade) and 5 mL water (LC-MS grade). After partitioning, we discarded the water fraction while the non-polar ethyl acetate fraction was saved and dried into a pre-weighed 20 mL scintillation vial. To incorporate any intraspecific variation between thalli, all 5 replicates per species were pooled (by combining all 5 individual extracts and drying down the pooled extract) and pooled extracts were used in the assays (see below). Ecologically relevant concentrations (mg/cm2) were calculated by dividing the dried extract mass by the 2-D surface area of the associated alga (determined via ImageJ).

Allelopathy trials: effects of waterborne metabolites from turf algae

To test the lethality of waterborne exudates from turf algae on kelp recruitment, we subjected kelp gametophytes to the extracts of exudates from each of the 5 most abundant turf species (at ecologically relevant concentrations, 0.4-2.3 µg/mL). Extracts were resuspended in methanol; for each replicate, a 0.25 µL aliquot of extract was combined with ~20 gametophytes and 300 µL of sterile seawater (consisting of Instant Ocean, MilliQ water, and nutrients f/2) in a single well of a 96-well flat bottom plate. Control extracts – SPE cartridges that eluted 1000 mL of LC-MS grade water and were then extracted and dried with the same methods as above – were resuspended in methanol. For each control replicate, a 0.25 µL aliquot of extract was combined with ~20 gametophytes and 300 µL of pre-made sterile seawater (consisting of Instant Ocean, MilliQ water, and nutrients f/2) (n = 10 wells per treatment and control). Gametophytes were kept in a 10°C incubator in the dark for 96 hours, at which time they were scored (see below).

Allelopathy trials: effects of surface-bound metabolites from turf algae

To test the effects of surface-bound molecules from turf algae on kelp gametophytes, we painted surface extracts from turf algae on the bottoms of wells within 96-well flat bottom plates using hexane (LC-MS grade) (0.5-3.8 µg/mL). After complete evaporation of the hexane, ~20 gametophytes and 300 µL of pre-made sterile seawater (consisting of Instant Ocean, MilliQ water, and nutrients f/2) were added to each well (n = 10 replicate wells per treatment). A control (hexane only) was also employed (n = 10 replicate wells). Gametophytes were kept in a 10°C incubator in the dark for 96 hours, at which time they were scored (see below).

Scoring and statistical analysis of kelp gametophyte mortality

After 96 hours, each well was stained with CellTox Green – a cell impermeant DNA stain – to assess gametophyte mortality under epifluorescence. Live kelp gametophytes exhibit no fluorescence with Celltox Green staining, while dead gametophytes become permeable to the dye, which binds DNA and induces bright fluorescence. Following staining, the number of living and dead gametophytes in each well was immediately scored using a Leica DMi8 microscope with 470nm light for excitation and a 590 emission filter to visualize the CellTox fluorescence in dead gametophytes using a low-powered (20x) objective lens and a Leica DFC9000GT sCMOS camera. Given uncertainty regarding how fast allelochemicals degrade in nature, we limited our assay duration to 96 hours. The duration of allelochemical exposure to gametophytes was thus short, relative to what they likely experience in nature (where nearby turf algae would be exuding waterborne metabolites frequently, over weeks to months of interaction time). Our results are, therefore, likely to be conservative.

To test for differences in kelp gametophyte mortality between a given treatment and its paired control, we employed a test for equality of proportions with continuity correction. We also employed a generalized linear mixed effects model using the glmmTMB package (66) within the program R to assess whether treatments were confounded by gametophyte starting concentration. Survival was unaffected by starting concentration; thus, the tests for equality of proportions with continuity correction are presented.