Benthic cyanobacterial mats of three distinct morphotypes were manually collected by SCUBA divers between 3 and 8 m depth in the Abrolhos Bank (16°40′, 19°40′S—39°10′, 37°20′W) (Figure 1). Sampling was performed at two different sites on pinnacle reef tops: the nearshore Pedra de Leste (PLES) of Parcel das Paredes reef and the offshore site of Parcel dos Abrolhos reef (PAB). Throughout the text samples, PLES and PAB will be referred to as PLES_BCM and PAB_BCM, respectively. The BCMs cover and long-term survey data are described in Teixeira et al. (2021). An aliquot of each sample was rinsed with filtered (GF/F) seawater and snap-frozen or fixed in 1% glutaraldehyde. Additional aliquots were transferred to vials containing f/2 medium (Guillard, 1975), prepared with 0.22 μm filtered Abrolhos seawater or filtered seawater, and transported to the laboratory. Single cyanobacterial filaments were isolated in a flow cytometer (MoFlo, Beckman Coulter, United States) (Fistarol et al., 2018) or by manual micromanipulation under an inverted microscope in case of wider filaments. Collected filaments were chosen to encompass distinct diameter diversities within conspicuous BCMs morphotypes in the Abrolhos Bank. Isolated filaments were transferred to f/2 medium (Guillard, 1975), prepared with 0.22 μm filtered Abrolhos seawater, and kept at 26°C with a 14:10-h light/photoperiod and ca. 20 μmol photons m^–2 s^–1. Established cultures were deposited in the Culture Collection of Microorganisms at UFRJ (CCMR).

For each BCMs sample, single wide and narrow filaments were selected from the tufts under an inverted microscope (Leica DMi8^®) and thoroughly rinsed with a sterile f/2 medium. The selected filaments were photographed with a digital camera for further morphometric analysis using Leica Application Suite X 3.4.1.17822 (Leica^®, Mannheim, Germany) software. We measured a minimum of 20 filaments of each morphotype from the same sample, whenever possible, using the AxioVs40 v4.8.2.0 imaging system (Carl Zeiss, Jena, Germany). Single filaments were transferred to ca. 1 μL of medium to 0.5 mL in PCR tubes and digested with 0.5 μL of proteinase K (23 mg ml^–1) at 37°C for 1 h, followed by 80°C for 5 min to denature the enzyme. These digested mixtures were further used directly in PCR to amplify the 16S rDNA gene. For cyanobacteria maintained in cultures, DNA was extracted from a cell pellet with 2% CTAB buffer (hexadecyl-trimethyl-ammonium bromide, pH 8.0), followed by chloroform/isoamyl alcohol (24:1, v/v) to take off organic compounds. DNA was precipitated with 95% ethanol and 0.3 M NaOAc, rinsed two times with cold 70% ethanol, and resuspended in 25 μL of TE buffer (10 mM Tris–HCl, 1 mM EDTA, pH 8.0). After DNA extraction, the integrity of the genetic material was checked on a 1% agarose gel. DNA quantification was done in a Nanodrop 1000 spectrophotometer (Thermo Scientific, United States).

Amplification of the 16S rDNA was performed according to Lane (1991) by nested PCR, using external universal 16S rDNA primers (forward 27F 5′-AGAGTTTGATCMTGGCTCAG-3′; reverse 1492R 5′-GGTTACCTTGTTACGACTT-3′) and subsequently internal cyanobacteria-specific primers (forward CYA359F 5′-GGGGAATYTTCCGCAATGGG-3′ and CYA106F 5′-CGGACGGGTGAGTAACGCGTGA-3′; reverse CYA781R 5′-GACTACTGGGGTATCTAATCCCWTT-3′), following previously described procedures (Nübel et al., 1997). Fragments using CYA106F were chosen to be amplified and confirmed by electrophoresis in agarose gel. The PCR products were purified with ExoSAP-IT (Thermo Scientific, United States). Sample sequencing was obtained by the SANGER method (Applied Biosystem 3500). Contigs were assembled with SeqMan Pro v.11.1, and electropherograms were visually checked for peak quality or reading errors and compared among them after alignments, and also to published sequences using the BLASTn tool in GenBank. In this study, 15 cyanobacterial sequences from Abrolhos were produced. A set of sequences from cultures with high similarity to these, sequences from type specimens, and sequences from closely related groups were retrieved to be used in phylogenetic reconstruction (49 sequences in total). Also, an external sequence from E. coli (NC000913) was chosen to root the topology. Sequences were aligned in ClustalW implemented in MEGA X (Kumar et al., 2018) with further manual refinement. Bayesian information criterion (BIC) determined the evolutionary substitution model with the model test, also implemented in MEGA X. Phylogenetic reconstructions were done by Bayesian and maximum likelihood (ML) methods. Mrbayes 3.2 (Ronquist et al., 2012) was used with 1 M of MCMC searching, discarding 10% as burn-in, and web-IQTREE (Trifinopoulos et al., 2016) with the UFBoot, ultrafast bootstrap approximation tool (Hoang et al., 2018), and the SH-aLRT test (Guindon et al., 2010). The Bayesian topology was edited in FigTree (Rambaut, 2018), and ML branch supports were added to the final figure.

To investigate the allelopathic effects of cyanobacterial chemicals on Symbiodinium sp. CCMR0100 (ITS2 type A4) (viz. Silva-Lima et al., 2015) isolated from the coral Mussismilia braziliensis (Verrill, 1868) in the same reef system, we exposed the photosymbionts to cell-free filtrates of cultured Moorena bouillonii (L. Hoffmann and Demoulin) Engene Tronholm (N = 4) and Adonisia turfae Walter et al. (N = 5), isolated from PAB_BCM. Strains were grown in an f/2 medium at 26°C, ca. 20 μE m^–2s^–1, and a 14:10-h light/dark photoperiod. Cyanobacterial cultures were grown for ca. 1 month, depending on the strain, until enough biomass was accumulated. The cells were then removed by filtration through glass microfiber filters (GF/F, 25 mm). Aliquots of 9 ml of filtrate from each cyanobacterial culture were distributed into quadruplicate 20 ml glass vials containing 9 ml of Symbiodinium sp. culture (harvested in the late exponential phase) in an f/2 medium. Controls were made by adding f/2 medium to the symbiont cultures instead of cyanobacteria culture filtrate. Two milliliters of fresh medium were added to each treatment and control vial to avoid nutrient limitation. Each experimental unit had an initial dinoflagellate concentration of ca. 6 × 10⁴ cells ml^–1. The experiment was run for 72 h at temperature and irradiance as described above. Cell counts were performed daily either by flow cytometry (BD Accuri C6) or by microscopy using Palmer–Maloney counting chambers. The allelopathic effect was assessed daily by measuring the difference in symbiont cell number between each treatment and its respective control. Negative effects resulted in significantly fewer cells in the treatment exposed to the filtrate than in the control. Positive effects resulted in more cells in the treatment flasks (i.e., greater cell number than the control).

To evaluate the allelopathic properties of secondary metabolites stored intracellularly in comparison with tests using naturally exuded allelochemicals (see previous section), we extracted a pooled biomass of PAB_BCM using organic solvents. Approximately 300 g (wet weight) of BCMs fronds was frozen in the field, transported to the laboratory, and freeze-dried. At that site, the BCMs is predominantly composed of M. bouillonii, A. turfae, Halomicronema sp., and Leptolyngbya sp. (Table 1). Exhaustive and successive organic extractions were done using ethyl acetate/methanol (1:1), yielding a homogeneous extract. Aliquots of this extract were solubilized in DMSO (dimethyl sulfoxide) and added to vials containing Symbiodinium sp. CCMR0100 to a final concentration proportional to the cyanobacteria biovolume used to prepare the extract, plus three dilutions by a factor of 10, that is, 10, 100, and 1,000 times of initial concentration. Symbiodinium cells were exposed to these extracts following the same experimental design as explained above for the cell-free filtrates. Cells were counted using an automated imaging system (FlowCam, Fluid Imaging Technologies, Inc.) before the experiment and after 48 h of incubation with the extracts to characterize the allelopathic effect of the cyanobacterial chemicals.

To evaluate the susceptibility of PAB_BCM to consumption and the potential influence of BCMs presence on consumer–prey interactions, we performed susceptibility assays of macroalgae and cyanobacteria in the field. We hypothesized that reef fish would refrain from consuming PAB_BCM filaments due to their high toxicity and/or low nutritional value. A small piece (5–10 cm) of fresh macroalgae and cyanobacteria was fixed to a 20-cm length 3-strand polypropylene rope with both ends fastened to a fishing weight, according to conventional methods (Hay et al., 1988), with modifications. Halymenia sp. was selected as a palatable macroalgae reference to evaluate BCMs consumption since it was readily consumed in previous tests (the authors, unpublished). Assays recorded grazing in three different treatments: a mix of macroalgae and BCMs, BCMs only, and macroalgae. A GoPRO camera was fastened to the bottom, and all treatments were placed within the camera’s field of view. The experimental design was conceived after running pilot studies in which we repeatedly observed no consumption of stranded BCMs, and full consumption of Halymenia when offered separately to consumers. Due to such strong contrast, we added the mixed treatment to evaluate the grazing behavior toward Halymenia baits in face of unpalatable BCMs of similar color covering the thalli. We performed three sequential trials, using a single replicate for each treatment in every trial (n = 3) and recorded each one for 3 h, independent of the total consumption of any given treatment. The consumer species, number of bites, and time before consumption of the total biomass offered in a single bait were identified and recorded in each video.

Samples of PAB_BCM (i.e., Moorena-dominated tufts, see Table 1) were snap-frozen and taken to the laboratory. The toxicity was assessed by exposing Artemia salina nauplii within 24 h after eclosion to crude aqueous extracts of BCMs biomass. The BCMs biomass was harvested on 25-mm fiberglass filters (GF/C), moistened with seawater, and preweighted. Filters containing the biomass were immediately weighted to calculate the wet weight of the cyanobacteria samples. Filters were then transferred to 15-ml polystyrene vials filled with 8 ml of filtered seawater. Cyanobacteria cells were disrupted by three cycles of sonication (1 min each with pulses of 1 s, full power, in an ice bath) interspersed with freeze–thaw at −18°C. Then, the homogenates were filtered on 25-mm fiberglass filters (GF/C) to remove cell debris, 9 ml of AMF was added to 3 ml of the cell-free filtrates, and 6 ml of AMF was added into six 1:1 dilutions of the homogenate. Dilution series were prepared to achieve six concentrations from a max mean (n = 11) concentration of 26, 29 mg f.w. eq. ml^–1. Aliquots of 2 ml from each filtrate dilution were added to a 5-mL well plate in triplicates. Each well contained 2 ml of filtered seawater and 20 A. salina nauplii, which were observed after 48 h and scored as alive (motile) or dead (non-motile).

The specific growth rates of Symbiodinium cultures in incubations were calculated based on the cell concentrations, according to Equation 1:

Where N₁ and N₀ are the cell concentrations at t₁ and t₀, and t is the time, in days, between sampling occasions.

Allelopathy data were analyzed with ANOVA with growth rate as response variable and treatment as fixed factor, followed by Tukey’s post hoc test. Validity of equal variance, normality, and independence assumptions on the error terms was confirmed using the Shapiro–Wilk test and visual inspection of the plotted residuals. Susceptibility data were analyzed with the Kruskal–Wallis non-parametric test to check for differences between treatments with Halymenia sp. Toxicity was measured as LC₅₀-48 h (median lethal concentration with 95% confidence intervals after 48 h of exposure) calculated by the Trimmed Spearman–Karber method (Hamilton et al., 1977). All statistical analyses were performed in the R environment (R Core Team, 2016).

The BCMs were predominantly composed of M. bouillonii or O. erythroflocculosa, non-ramified cyanobacteria species enclosed in exopolysaccharide sheaths and exhibit discoid cells arranged in wide trichomes, corresponding to the Oscillatoriales order (Komárek and Anagnostidis, 2005). In addition, Adonisia turfae Walter et al., Leptolyngbya sp., and Halomicronema sp., were also found in the BCMs. However, they belong to the Synechococcales order, which has thin, non-ramified filaments, with cells longer than wide.

The BCMs from PAB_BCM presented different macroscopic morphologies and cyanobacteria taxonomic compositions than those collected from PLES_BCM. In PAB, the BCMs form fasciculate tufts of flame-like appearance, dark-red contours, and a pale inner region (Figure 2A), or as long, pink-brown fasciculate tufts (Figure 2C). These BCMs were predominantly composed of M. bouillonii (Figures 2D,E) and a few entangled filaments of A. turfae (Figure 2H) and Halomicronema sp. (Figure 2I). In BCMs from PLES, O. erythroflocculosa was the predominant taxon (Figure 2F) with Leptolyngbya sp. and A. turfae (Figures 2G,H), as relatively less abundant taxa, forming dark pink fasciculate, horizontally-spreading mats (Figure 2B and Table 1).

Phylogenetic inference (Figure 3) grouped the 16S rDNA sequences from broad filaments into two well-supported clades (maximum likelihood, UF-bootstrap percentage/Bayesian analyses, posterior probability), Moorena (97/0.99) and Okeania (99/1.00). The Moorena clade included the isolates CCMR0174, PAB1, 2, 3, 4, and 5, identified as M. bouillonii. The sequences of PLES1 and PLES2, identified as O. erythroflocculosa, clustered together in the Okeania clade (98/0.96). The seven sequences from thin filament isolates were distributed within two clades, one represented by Halomicronema sp. (92/0.95), including the strain CCMR0085, and the other (−/0.81) composed of the remaining six isolates, identified as A. turfae (CCMR0080, CCMR0081, CCMR0082, CCMR0083, and CCMR0086) and Leptolyngbya sp. (CCMR0087).

The 16S rDNA sequences presented a considerable similarity to others already published, such as PLES1 and PLES2, whose sequences showed 100% similarity with O. erythroflocculosa KC986930 (Engene et al., 2013b). The sequences of PAB1, 2, 3, 4, 5, and CCMR0174 were 99–100% similar to M. bouillonii GU111927 (Engene et al., 2012). Regarding the thin filaments, sequences of strains CCMR0080, CCMR0081, CCMR0082, CCMR0083, and CCMR0086 previously identified as A. turfae had 100% similarity to Leptolyngbya sp. JX481735 (Konstantinou et al., 2019). The sequence of the strain CCMR0085 presented 99% similarity to Halomicronema sp. AB257773 (Yamazaki et al., 2008), which was found in the skeleton of the reef-building coral Goniastrea aspera Verrill, 1866. The sequence of the strain CCMR0087 was 99% similar to Leptolyngbya sp. EU743968, collected on the coral Orbicella annularis Ellis and Solander, 1786, which is one of the species responsible for causing Black band disease (BBD) (Gantar et al., 2009).

Growth rates of the ex hospite Symbiodinium sp. exposed to cell-free filtrates (i.e., exudates) of the cultured M. bouillonii and A. turfae were significantly lower than those of corresponding controls (Supplementary Table 1 and Figure 4). Cell-free filtrates from cultures of the remaining taxa did not cause significant effects on Symbiodinium. Organic extracts of PAB_BCM significantly reduced the growth rates of Symbiodinium sp. at natural concentrations after a 48-h incubation (Supplementary Table 1 and Figure 5).

There was no BCMs consumption by fish in susceptibility field assays performed in PAB (Table 2, Figure 6, and Supplementary Video). Treatments with only macroalgae were consumed first, before the Halymenia pieces in mixed treatments. All BCMs in the experiment remained untouched until the end of the trial. BCMs present in mixed treatments was avoided during delicate targeting of the macroalgae thalli. Halymenia sp. in macroalgae-only and mixed treatments were consumed in all trials compared with no-consumption of BCMs-only treatments (Table 2). There was no difference in the total number of bites between macroalgae treatments (chi-square = 0.2, df = 1, p > 0.05). The angelfish Pomacanthus paru and the surgeon fish Acanthurus bahianus were the primary consumers. They had distinct strategies to avoid BCMs while grazing the macroalgae in mixed treatments. Pomacanthus paru have a protruding feeding apparatus that allowed them to break the bait and selectively suck the pieces of interest floating in the water. The surgeon fishes have a smaller mouth and repeatedly bite the Halymenia pieces, taking out small fragments. Other eventual consumers were the juvenile of the parrotfish Sparisoma axillare, the yellowtail snapper Ocyurus chrysurus, and the damselfish Stegastes fuscus, which consumed Halymenia only in the mixed treatment.

Three out of 11 PAB_BCM samples caused more than 50% mortality to A. salina nauplii (Supplementary Figure 1). The LC₅₀-48 h for these three samples was 9.8 mg w.w. eq. ml^–1 (95% CI: 8.3–11.6 mg w.w. eq. ml^–1), 9.3 mg w.w. eq. ml^–1 (95% CI: 6.8–12.6 mg w.w. eq. ml^–1), and 3.2 mg w.w. eq. ml^–1 (95% CI: 2.5–4.1 mg w.w. eq. mL^–1). The other eight samples had LC₅₀-48 h values higher than the maximum concentration used in tests (i.e., > 13.7–32.6 mg w.w. eq. ml^–1).

Morphological and genetic analyses of cyanobacteria in the BCMs unveiled two species, M. bouillonii and O. erythroflocculosa, hitherto unknown to the Abrolhos reefs and the tropical Southwestern Atlantic (SWA). Species belonging to both genera have been reported in Florida, the Caribbean, and the Pacific regions (Engene et al., 2012). Okeania erythroflocculosa has been detected as an abundant component of the BCMs worldwide (Engene et al., 2012, 2013a). Despite the high gene abundance of Oscillatoriales cyanobacteria previously reported for Abrolhos cyanobacterial mats (Walter et al., 2016), the species have not been identified to date. The species-level identification of cyanobacteria with thin filaments requires diacritic features that are difficult to observe. Their genotypic diversity can be high within the different phenotypes, characterizing them as cryptotaxa (Engene et al., 2010; Caires et al., 2018). Here, the BCMs cryptotaxa were identified as A. turfae, previously reported (Walter et al., 2020), and two strains closely related to Leptolyngbya sp. and Halomicronema sp. The 16S rDNA sequences of the two latter cyanobacteria showed great similarities with the sequences found in the skeleton of scleractinian corals (Yamazaki et al., 2008; Gantar et al., 2009).

Laboratory experiments evidenced allelopathic effects of chemical substances from cultured strains of cyanobacteria (M. bouillonii and A. turfae), expressed by inhibiting the growth of the ex-hospite Symbiodinium sp. isolated from M. braziliensis. This negative effect on Symbiodinium was also evidenced in incubations with organic extracts from BCMs collected in Abrolhos. Similar incubation bioassays reported a decrease in the number of Chlorella vulgaris cells after exposure to freshwater cyanobacteria exudates (Leão et al., 2009). Growth inhibition potential was also reported by the metabolites hapalindole from Fischerella sp. and calothrixin A from Calothrix sp., which affected growth and inhibited the genetic replication of prokaryotic and eukaryotic cells in bioassays (Doan et al., 2000). Also, exudates from the Baltic Sea cyanobacteria Nodularia spumigena and Anabaena sp. exhibited antagonistic effects on cryptophytes (Suikkanen et al., 2004). Cyanobacterin, a metabolite from Scytonema hofmanni, and fischerelin A, from Fischerella muscicula, were reported to affect the growth of photoautotrophs by inhibition of photosystem II (Gleason, 1990). Cyanobacteria extracts also exerted allelopathic effects in the field, manifested through a decreased pigment concentration of endosymbiotic dinoflagellates and the consequent bleaching of the Porites coral (Puyana et al., 2019).

The decrease in growth rate of Symbiodinium sp. cultures supports previous evidence of an antagonistic relationship between BCMs and reef builders in Abrolhos (Ribeiro et al., 2018). The effects shown here are likely involved in interactions at the BCMs-coral boundary, which accumulate and transfer metabolites (Wangpraseurt et al., 2012). The increase in cyanobacterial biomass observed during blooms could result in thicker diffusion boundary layers that enhance transfer mechanisms and have antagonistic effects on coral photosynthesis and respiration (Jorissen et al., 2016). Such enhanced mass transfer through boundary layers and tissues (Patterson, 1992) could expose Symbiodinium to combined stress due to molecules produced by heterotrophs under anoxic conditions within the BCMs, and exuded allelochemicals diffused during blooms (Brocke et al., 2015). An increase in BCMs cover would also be a threat to coral recruitment success (Ritson-Williams et al., 2020). Additionally, allelopathic effects during episodic cyanobacteria blooms can be further aggravated after acute disturbances such as heatwaves (Teixeira et al., 2019) when both the host and photosymbionts are physiologically impaired due to oxidative stress (Oakley and Davy, 2018).

The mortality of brine shrimp after exposure to aqueous extracts of Abrolhos BCMs at natural cyanobacterial densities unveils the potential toxic properties of these benthic structures. Variable toxicity within the same BCMs that we observed (i.e., 3 out of 11 samples caused >50% mortality) should not imply a lack of chemical defense by cyanobacteria, as spatial and seasonal variability of chemical defenses against herbivores is expected due to metabolic responses to changes in the environment (Puglisi et al., 2019). The high spatial variability of Abrolhos reefs likely contributes to a context dependency of BCMs toxicity (Ribeiro et al., 2018; Teixeira et al., 2021). The large biomass of morbid sea slugs, Stilocheylus striatus, observed during cyanobacterial blooms in Abrolhos reefs (Ribeiro et al., 2017) is in line with the toxic potential of Abrolhos BCMs observed here. This opistobranch is a specialist consumer of BCMs and has been shown to sequester toxic compounds from the cyanobacteria, becoming unpalatable to predators (Pennings and Paul, 1993).

Our results indicate that O. erythroflocculosa, M. bouillonii, Leptolyngbya sp., and A. turfae are potential producers of toxic compounds within Abrolhos BCMs. Moorena, Okeania, and Lyngbya species have been reported to produce more than 40% of the cyanobacteria secondary metabolites described to date (Engene et al., 2011; Dittmann et al., 2015), many of which mediate interactions with grazers (Thacker et al., 1997) and competitors (Kuffner et al., 2006; Titlyanov and Titlyanova, 2009). The Leptolyngbya consortium produces cyanotoxins and is associated with BBD in corals (Gantar et al., 2009).

Field assays of susceptibility to herbivory showed that BCMs remained untouched by consumers, either in treatments with BCMs alone or mixed with Halymenia sp., a more palatable macroalgae. Even though the genus Halymenia produces ecologically relevant secondary metabolites (Malik et al., 2020), the presence of these types of chemicals did not seem to influence algal consumption by surgeon fish elsewhere (Wylie and Paul, 1988). Among consumers, both A. bahianus and S. axillare are detritivorous but consume macroalgae and have a scraper feeding mode. These species display high foraging plasticity and can explore different food items, including cyanobacteria, and also calcareous and sheet-like macroalgae (Ferreira and Gonçalves, 2006; Mendes et al., 2018). The angelfish P. paru has a varied, omnivorous diet, encompassing sessile invertebrates and several groups of macroalgae (Batista et al., 2012). These species readily consumed the macroalgae treatments in every trial, with just a few bites given their relatively large size and protruding feeding apparatus. Although reef fish may consume benthic cyanobacteria (Ferreira and Gonçalves, 2006; Cissell et al., 2019), our exploratory results suggest avoidance of BCMs by reef fishes in PAB. The limited sample size of our feeding trials requires caution in interpreting such results and making generalizations of this pattern to BCMs in general. Given that a significant discrepancy between feeding selectivity and nutrient content of cyanobacteria was reported for reef herbivores (Mendes et al., 2018), BCMs consumption and deterrence seem to be a trade-off between its composition, growth form, and chemistry (Cissell and McCoy, 2021; Ford et al., 2021). The bottom-up effects on reef herbivores remain elusive as contrasting evidence has recently emerged (Cissell et al., 2019; Ford et al., 2021). Consumers with different feeding strategies and nutritional demands may then target high nutritious cyanobacteria (Clements and Choat, 2018), and also associated non-cyanobacterial components (Cissell and McCoy, 2021), or avoid less nutritious (Nagarkar et al., 2004) and toxic cyanobacteria (Ford et al., 2021).

Phase shifts from coral dominated to opportunistic BCMs taxa have been reported worldwide (de Bakker et al., 2017; Ford et al., 2018), and such expansion disturbs reef trophic dynamics (Ford et al., 2021) by reducing energy flow and ultimately promotes a trophic simplification of the ecosystem (Ullah et al., 2018). This trophic disruption brings reef communities to alternative unstable states (Teixeira et al., 2021), because cyanobacteria are highly efficient in nutrient uptake and grow fast (den Haan et al., 2016; Ford et al., 2018), boosting their capacity to outcompete other benthic species and change community structure (e.g., de Bakker et al., 2017; Ribeiro et al., 2018; Teixeira et al., 2021). In the Abrolhos reef complex, BCMs can be structured by chemically defended species of cyanobacteria that display allelopathic, grazing deterrent, and toxic properties. These ecologically significant, albeit antagonistic, interactions with reef organisms could promote deleterious cascading effects on the ecosystem and contribute to the homogenization of reef communities. The disturbance of such trophic relations is probably amplified during episodic blooms of benthic cyanobacteria.