Chemical Diversity of Marine Filamentous Benthic Cyanobacteria

Abstract

Genomic and chemical analysis has revealed that numerous species of filamentous cyanobacteria harbor complex secondary metabolisms tailored to their particular ecological niche. The metabolomic analysis of strains and environmental samples from benthic cyanobacterial mats (BCMs) from coral reefs has the potential to expand the library of marine cyanobacteria-derived natural products. In this study, cyanobacterial strains were obtained from phytobenthos collected from coral reefs in Abrolhos, Brazil and Ishigaki, Japan. Phylogenetic analysis of isolates shows high similarity to previously described members of benthic mats and also suggests the geographic expansion of the Adonisia lineage. Chemical analysis by untargeted liquid chromatography-high resolution mass spectrometry and data processing via MZmine and FBMN-GNPS confirmed the presence of a wide diversity of secondary metabolites. In addition, similarity analysis applying the newly developed tool DBsimilarity indicated the broad coverage of various biosynthetic and chemical classes of compounds previously reported for cyanobacteria. This report is one of the first applications of untargeted metabolomics workflow and similarity network construction for groups of marine filamentous cyanobacteria isolated from benthic mats on corals reefs.

1. Introduction

Cyanobacteria are Gram-negative prokaryotes capable of performing oxidative photosynthesis and exert a large impact on ecoregions around the world [1]. Emerging as far back as around 2.4 billion years ago, they transformed the atmosphere by saturating the air with released oxygen and adapted to thrive in almost every ecological niche [2]. Cyanobacteria have endured changing environmental conditions over their long history by evolving mechanisms to cope with the difficulties of day-to-day life [3]. This was accomplished by responding to dynamic environmental conditions via genetic expansion and reprogramming to ensure survival. The fixation of atmospheric nitrogen by Nostoc species to enhance symbiotic interactions with plants and mosses [4,5,6] is one example of adaptation to low-nutrient environments that impacts global nitrogen cycling [7,8]. A standard cyanobacterial survival strategy is the precise regulation of metabolism and energy expenditure, to the point where various species can tolerate stresses such as extreme heat, salt and pH ranges, desiccation, high UV exposure, excess free metals concentrations, and nutrient deprivation [9]. The expansion of the cyanobacterial lineage through history is a success story of evolution and adaptation, as an eventual cyanobacterial endosymbiont was engulfed by a eukaryotic host one billion years ago to evolve into the chloroplasts of eukaryotic phytoplankton and terrestrial plants [10].

Having ~2.4 billion years to reprogram and reshuffle prokaryotic genomes for adaptation has resulted in the phylum displaying an astonishing range of characteristics and diversity at the cellular, chemical, and molecular levels. Both terrestrial and marine cyanobacteria flourish as planktonic unicellular species [11,12], while filamentous cyanobacteria evolved to a cosmopolitan multicellular life cycle consisting of complex morphologies [13] and sophisticated secondary metabolisms capable of producing toxins, anticancer, antibiotic, and antifungal compounds [14,15]. Cyanobacteria are found in soils, waterways, and marine environments and the chemical analysis of environmental samples and isolated laboratory cultures has generated a large library of cyanobacterial metabolites dating back over 75 years [16]. The prevalence of cyanobacteria around the planet suggests this reported collection may be but a fraction of the actual cyanobacteria-derived natural products present in the environment today.

Benthic cyanobacterial mats (BCMs) contribute to coral reef communities via primary production and participation in nitrogen and sulfur cycling [17,18]. BCMs exist as periphytic biofilms but also develop into thick, horizontally spreading mats engulfing rubble and sediment structures in some situations. These mats consist of a consortium of microorganisms dominated by cyanobacteria and heterotrophic bacteria living in proximity with small dispersals of bacteriophages, diatoms, and fungi [19]. Benthic filamentous cyanobacteria will bloom in the appropriate conditions and out-compete other organisms, including reef-building corals [13,20], and effectively choke them out. There is also the deployment of allelochemicals secreted into the local environment to antagonize growth of healthy reef organisms [21,22]. The potential chemical profile of a cyanobacterial strain can be inferred by bioinformatic analysis of genome sequences and prediction of the biosynthetic gene clusters (BGCs) responsible for de novo synthesis of secondary metabolites [23,24]. Genomic analysis of Adonisia turfae [25], a filamentous cyanobacterium recently isolated from BCMs in Brazil, suggests a wealth of metabolic diversity rivaling some Moorena species [15,26]. Extracts from A. turfae inhibit growth of Symbiodinium sp., an ex-hospite endosymbiotic dinoflagellate associated with reef-building corals [27], suggesting the cyanobacterial component of BCMs may aid in the blooming process by suppressing the growth of native coral algal species through chemical mechanisms. BCM cyanobacteria therefore not only contribute to critical ecological functions but have the potential to influence coral reef population dynamics. Establishment and blooms of these cyanobacteria in vulnerable areas during times of stress undermine coral physiology and recovery after bleaching by rapid encroachment and chemical defenses to maintain competitive advantage [28].

In this study, cyanobacteria were collected from Iriomote-Ishigaki National Park in Japan and the Abrolhos Archipelago National Park in Brazil. The island of Ishigaki is located southwest of Okinawa Prefecture and hosts a diverse variety of life, from plants and fauna to the microscopic level [29,30]. The Abrolhos Archipelago is located on the northeastern Brazilian coast off the state of Bahia and has an extensive system of mapped coral reefs containing unique underwater flora and fauna [31,32]. Large areas of coral reefs in both the Ishigaki and Abrolhos nature reserves have or are currently undergoing coral bleaching and eutrophication due to overfishing, climate change and elevated seawater temperatures [33,34]. Bleaching events have a major impact on local coral reef population dynamics, leading to the encroachment of other species more suitably adapted to stressful conditions [35]. These microbial phase shifts, driven by cyanobacterial blooms, are a marker of coral reef decline and have been documented in locations around the planet [36].

Strain isolation studies can yield a large number of cultures but the chemical analysis of isolates on an individual basis prevents statistical comparison between samples. The ability to analyze huge datasets from groups of cyanobacteria with new metabolomics tools and methods can allow for faster prioritization of strains for identification of metabolites of interest. We therefore isolated cyanobacterial strains from the benthic reef samples, cultured them in laboratory conditions and subjected them to chemical analysis. Environmental and uncultured samples of Moorena were also analyzed as part of the dataset. It should be noted that the isolation and culturing of these benthic cyanobacteria under standard laboratory conditions implies that the results obtained within the study do not reproduce the production of secondary metabolites under natural conditions. However, instead of identifying metabolites produced in the wild by cyanobacterial BCM components, this method is an attempt to catalogue the chemical space of the metabolites encoded in the genomes of these organisms. Our aim was to perform a comprehensive chemical analysis of samples from these distinct ecosystems for determining the presence of compounds of interest, such as known toxins, therapeutical compounds, and new metabolites. We then applied an untargeted metabolomics workflow to establish similarity networks using custom databases in order to prioritize strains and molecules for further investigation. The results of this study expand the known chemical space of BCM-derived filamentous cyanobacterial secondary metabolism.

2. Materials and Methods

2.1. Collection of Samples from Abrolhos and Ishigaki

The harvesting of samples from the Abrolhos Bank, Bahia, Brazil, was previously described [22] and consisted of manual collection by SCUBA divers between 3 and 8 m depth in the area (16°40′, 19°40′ S–39°10′, 37°20′ W). Permission for sampling and studying these microorganisms is granted through a federal government license (SISBIO no. 65055-9, SisGen no. A8CD3DF). Four sites were sampled around Ishigaki Island, Okinawa, Japan. Samples were obtained using SCUBA diving at Nobaru (24°23.20′ N, 123°55.65′ E), Gareba (24°21.64′ N, 123°59.34′ E), Taketomi hydrothermal (24°20.18′ N, 124°06.14′ E), and Manta city point in June and July 2015. No permission was needed for water sample collections for cyanobacteria isolation during the Ishigaki sampling. BCM samples were preserved in Falcon screw cap tubes. Figure 1 shows a map with the sampling locations highlighted.

2.2. Isolation, Purification and Maintenance of Strains

Samples of collected cyanobacterial biomass were washed with filtered seawater for debris removal. Fresh samples were analyzed under stereomicroscope after washing and dissected to isolate unicyanobacterial filaments. Each processed filament was incubated in F/2 media [37] supplemented with 0.3 mg L⁻¹ GeO₂ (Sigma-Aldrich, Saint Louis, MO, USA), and 30 mg L⁻¹ of cycloheximide (Sigma-Aldrich) to prevent unwanted growth of diatoms and other eukaryotes. Figure 2 shows images of collection, culture, and microscopy of some isolated strains. In order to validate the isolation of monocyanobacterial cultures, FACS (fluorescence-activated cell sorting) in a MoFlo (Dako-Cytomation) flow cytometer equipped with an electrostatic droplet deflection sorting device was applied. Individual colonies that lead to monocyanobacterial growth were barcoded (see Table S1, Supplementary Information) and maintained in F/2 medium [38]. The three environmental samples of Moorena spp. were frozen immediately after collection in Abrolhos (Figure 2A) and not subjected to further isolation and culturing. Therefore, it is understood that these samples likely contain other organisms that may be associated with the cyanobacterium and are non-axenic.

2.3. Identification and Phylogenetic Analysis

DNA extraction of samples was carried out according to previously described protocols [39]. Polymerase chain reaction (PCR) amplification of regions for some 16S rRNA genes was performed using forward primer CYA359F and an equimolar mix of the reverse primers CYA781R(a) and CYA781R(b) [40], while others were amplified using combinations of the 27F, 926F, 1093R, and 1492R primers [41,42,43]. The amplification protocol utilized GoTaq^® Green Master Mix (Promega, Madison, WI, USA) with the program: 94 °C for 5 min followed by 30 cycles of 94 °C for 1 min denaturation, 60 °C for 1 min for annealing, 72 °C for 1 min extension, with a final extension at 72 °C for 5 min. Primer sequences are provided in Table S2 (Supplementary Information). DNA sequencing was carried out according to BigDye^® Terminator v3.1 protocol (Thermo-Fisher, Waltham, MA, USA) via Sanger sequencing.

Recovered partial 16S rRNA gene sequences were used to probe the GenBank (https://blast.ncbi.nlm.nih.gov/, accessed on 13 September 2024) and CyanoSeq [44] databases. A phylogenetic tree with sequences coding for 16S rRNA was constructed using the maximum likelihood method and Tamura–Nei model [45]. The percentage of trees in which the associated taxa clustered is shown next to the branches (bootstrap with 1000 replicates) and values below 50% were omitted. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. This analysis involved 26 nucleotide sequences and there were a total of 1466 positions in the final dataset. Evolutionary analyses were conducted using MEGA version 11 [46].

2.4. Samples Preparation for Chemical Extraction

Cyanobacteria were cultivated in 500 mL Erlenmeyer flasks containing F/2 culture medium prepared with NaNO₃, NaH₂PO₄, trace metals, and vitamins and occasional mixing. Cultures were maintained at 20–40 μE/m²/s light intensity and a 16/8 light/dark cycle at room temperature [47]. Cyanobacterial biomass was harvested after 28 days of cultivation by centrifugation at 4500× g for 15 min at 4 °C. The supernatant was discarded and the biomass dried by lyophilization. For metabolite extraction, 50 mg of dried biomass was homogenized with zirconia beads in a 2 mL microtube containing 1.0 mL of dichloromethane-methanol (2:1, v/v) using a FastPrep-24 bead beater (MP Biomedicals, Santa Ana, CA, USA) for 1 min. Samples were centrifuged at 25,000× g for 15 min at 4 °C and the supernatant was then collected. This process was repeated twice, and the combined extracts were dried using SpeedVac. The samples were resuspended using 1 mL of MeOH and centrifuged prior to analysis to remove occasional precipitates.

Quality controls, including blanks and QC-pool samples, were included at each extraction step to ensure accuracy. A stepwise protocol is available online [47].

2.5. LC-HRMS Analysis and Data Processing

Chemical profile analysis of the extracts was conducted using a Dionex LC UltiMate 3000 High-Performance Liquid Chromatography (HPLC) system coupled with a QExactive Plus hybrid high-resolution mass spectrometer (Quadrupole/Orbitrap) from Thermo Fisher Scientific. Ionization was achieved via electrospray ionization (ESI), alternating between positive and negative modes. The chromatographic separation was performed using a Waters Acquity Premier HSS T3 column (100 mm × 2.1 mm × 1.8 µm), with a gradient elution of solvent A (water with 0.1% formic acid) and solvent B (acetonitrile with 0.1% formic acid). The gradient started at 15% solvent B for 2.1 min, followed by a linear increase from 15% to 95% B from 2.1 to 9 min, kept isocratically at 95% B for 4 min (9.1–13 min). The column was re-equilibrated to 15% B from 13.1 to 16 min. The flow rate was set at 350 µL min⁻¹, with a 5 µL injection volume and column oven temperature of 40 °C. The ESI source parameters were spray gas at 50, desolvation gas at 10, spray voltage (ESI−/+) at −3500 V, capillary temperature at 300 °C, and source temperature at 380 °C. Data acquisition was performed using Full MS/dd-MS2 (Top 3) mode, with a monitored mass range of 150–2000 m/z.

Data obtained from the LC-HRMS analysis was processed using MZmine 3.90 [48]. MS-feature annotation was performed using the custom database match method based on the precursor ion m/z within MZmine with parameters specifically optimized for this dataset (detailed in Zenodo). The custom database was curated by combining two different databases of structures identified from cyanobacteria: CyanosNP2010-2023 (accepted at Natural Products Reports) and CyanoMetDB [16]. A mass tolerance of 0.001 m/z was applied to ensure precise matching, and each annotation was manually validated. Principal component analysis (PCA) was performed using the processed data from MZmine. The processed data files were also submitted to molecular networking analysis using the Feature-Based Molecular Networking (FBMN) [49] and Dereplicator+ [50] methods on the Global Natural Products Social Molecular Networking (GNPS) platform [51]. The parameters for molecular networking were also optimized for this dataset, as detailed in Zenodo.

2.6. Chemical Variability and Predicted Biological Activity

To assess chemical variability, the combined CyanosNP2010-2023 and CyanoMetDB databases were analyzed using the DBsimilarity method [52]. This approach facilitated the visualization of the chemical space expected to be found in cyanobacteria against the annotated compounds from the LC-HRMS analysis. Similarity networks were generated using Morgan fingerprints and Dice similarity coefficients (using a threshold of 0.85), providing a detailed overview of structural relationships among the compounds. Additionally, t-distributed stochastic neighbor embedding (t-SNE) plots were created to evaluate chemical compounds distribution using chemical descriptors generated with the Python library Mordred including cytotoxic activity reported on literature for cyanobacterial compounds in the databases. All raw chemical data is available at Zenodo.

3. Results

3.1. Phylogenetic Analysis of Isolated Cyanobacteria

Seventeen cyanobacterial strains were isolated from the samples collected in Abrolhos and Ishigaki and maintained under the codes CCMR0053, CCMR0055, CCMR0057, CCMR0059, CCMR0069, CCMR0071, CCMR0074, CCMR0076, CCMR0078, CCMR0280, CCMR0281, CCMR0282, CCMR0283, CCMR0284, CCMR0285, CCMR0286, and CCMR0287. Strains CCMR0053-0078 were isolated from the Ishigaki samples, while CCMR0280-0287 represents strains isolated from Abrolhos. Three environmental samples of Moorena spp. were frozen immediately after collection in Abrolhos and assigned codes A01, A02, and PAB. DNA sequencing and analysis of the 16S rRNA gene was successful for all strains except CCMR0057, CCMR0078, and CCMR0283. The 16S rRNA sequences were deposited to Genbank and assigned accession numbers (Table S1, Supplementary Information).

The phylogenetic analysis shows the taxonomic proximity among these strains and type species based on 16S rRNA partial sequences (Figure 3). Analyzing from top to bottom, the strains CCMR0280, CCMR0281, CCMR0282, and CCMR0284 appear closely related to each other, as well as Adonisia turfae CCMR0081, a recently described species of nitrogen-fixing filamentous cyanobacteria also collected in Abrolhos [15]. The CCMR0074 strain from Ishigaki is similar at a species level to CCMR0081, suggesting Adonisia spp. may be found in BCMs in regions outside of Brazil. A. turfae CCMR0081 was initially described as a Leptolyngbya sp. but appears to be more closely related to the Leptothoe spp., a genus of marine sponge-associated cyanobacteria [53,54]. This group includes CCMR0286 and CCMR0287, which are very similar on a species level but distinct from the A. turfae-like strains. CCMR0059 closely groups with Leptothoe spongobia TAU-MAC 1015, a producer of the toxin microcystin-LR [54]. CCMR0285 grouped closely to Halomicronema spp., while CCMR0053 is relatively more distant. Previous reports show that Halomicronema metazoicum, long thought to be a marine sponge symbiont, can be isolated as a free-living culture producing bioactive compounds toxic to protozoa [55].

In the middle of the tree, a distinct branch formed by CCMR0055 is grouped closer to Metis spp., another genus of sponge-associated filamentous cyanobacteria [53]. The three Moorena strains A01, A02, and PAB grouped with type strain Moorena producens NAC8-47, with the A01 strain differing from the other two isolates at the species level. Moorena spp. are a large contributor to the library of cyanobacterial secondary metabolites as well as environmental irritants [56] and have been an area of intense interest in natural products research [26]. Towards the bottom of the tree, the highly related CCMR0069 and CCMR0071, along with CCMR0076, group with Geitlerinema sp. BBD, a cyanobacteria first identified as a component of black bands in diseased coral [57,58]. The non-sequenced strains CCMR0057, CCMR0078, and CCMR0283 are absent from the tree. This analysis places the newly isolated strains from both Abrolhos and Ishigaki near reef cyanobacterial lineages known for biosynthesis of interesting secondary metabolites.

3.2. Chemical Analysis and Raw Data Processing

All the cyanobacteria strains, except the Moorena spp. A01, A02, and PAB, were cultured and underwent chemical extraction for obtaining the secondary metabolites. For Moorena, the environmentally harvested, non-axenic clumps were extracted without previous strain purification or lab culturing and therefore likely contain metabolites of associated bacteria and other organisms. We developed a protocol for optimizing the extraction of a range of compounds to assess the chemical diversity in cyanobacteria. Strains and isolates were analyzed as single samples, duplicates, or triplicates, as follows: CCMR0053, CCMR0055, CCMR0057, CCMR0078, CCMR0280, and A01 as single samples; CCMR0059, CCMR0069, CCMR0071, CCMR0074, CCMR0076, CCMR0281, CCMR0282, CCMR0283, CCMR0285, CCMR0286, A02 and PAB as duplicates; and CCMR0284 and CCMR0287 as triplicates, resulting in 36 cyanobacterial samples (excluding quality control and blank samples).

These 36 secondary metabolite samples, 5 from environmental and 31 from cultured cyanobacteria, were analyzed by LC-HRMS. On average, 2000 MS features for each sample were generated with MZmine (available in Zenodo). For data processing, we first created a custom curated database using the public database CyanoMetDB, which contained 2010 cyanobacterial metabolites and 99 structurally related compounds in 2021 [16] and was updated in 2023 [59], and the newly published CyanosNP2010-2023, which contains 900 metabolites whose structures were validated by NMR. Excluding redundant compounds, 2619 compounds are listed in our custom database. The software MZmine facilitated data preprocessing, including noise removal, peak detection, and alignment. Processed data files were used for PCA and next processing in GNPS.

The PCA was used to simplify the large dataset generated from LC-HRMS analysis, aiding visualization of sample distribution. The PCA score plot of the LC-HRMS data illustrates the variability among the samples studied, providing a preliminary exploratory analysis that helps prioritize strains for further investigation. Among the five Moorena secondary metabolite samples, four exhibited high variability compared to other cyanobacterial samples (Figure 4A), suggesting that Moorena samples are the most chemically distinct. A second replicate of A02 appears ungrouped and can be seen in the expanded view of the PCA score plot (Figure 4B). These results are consistent with the environmental context, where Moorena often coexists with other associated species in clumps. Moorena species are also known for their potential to produce a higher diversity of secondary metabolites, given the large number of biosynthetic gene clusters (BGCs) in their genomes [26], making this genus a particularly promising target for further exploration [60,61].

A second PCA, excluding the environmental Moorena samples (Figure S1), was carried out to assess whether phylogenetically related strains would also show greater metabolic similarity. In this new dataset, strains with relative higher variability, such as CCMR0069, CCMR0074, and CCMR0076, stood out, with both replicates of each strain positioned far from the grouped samples. This suggests a relatively consistent result, although a higher number of replicates would be recommended for stronger statistical confidence. Phylogenetic analysis grouped some strains based on their 16S rRNA genetic relationships and we aimed to verify if the strains would group in a similar manner based on chemical profiles. The 16S rRNA sequence of CCMR0076 was only 97% similar to sequences in the database, and this strain was the closest phylogenetically to Moorena among the analyzed strains (Figure 3). Similarly, CCMR0069 was placed near CCMR0076 in the phylogenetic tree. For CCMR0074 (isolated from Ishigaki), one replicate was more distant from the grouped samples than the other, yet both were relatively separate in the PCA (Figure S1). This strain is phylogenetically like the four Abrolhos strains CCMR0280, CCMR0281, CCMR0282, and CCMR0284 (Figure 3), but metabolically it was more distant from its Brazilian counterparts. Although no definitive conclusions could be drawn about the metabolic and phylogenetic correlations for the first three highlighted strains (CCMR0069, CCMR0074, CCMR0076), our results indicate that these cyanobacteria in laboratory cultures exhibit more variation in metabolite profiles and should be prioritized for future studies.

CCMR0057 and one replicate of CCMR0282 were positioned apart from the grouped samples. However, since the other replicate of CCMR0282 clustered within the grouped samples, we considered this result unreliable and excluded it from further discussion. CCMR0057 appears to be a promising strain from a metabolic perspective; however, the lack of phylogenetic information and the absence of replicates in the chemical analysis prevent confident conclusions. Although some chemical variability exists among the other samples, they are less pronounced compared to those previously discussed. Even within the phylogenetic group containing the four Brazilian strains related to Adonisia turfae, which would be expected to be chemically similar—and largely are—some variability is still evident. This group contributed eight of the 31 samples analyzed, almost one quarter of samples, which probably resulted in the chemical redundancy and low metabolic variability shown in the PCA. Still, for being related to a newly described strain, these strains deserve further investigation. The remaining samples in this dataset will not be discussed further, as they are relatively similar and clustered together in the PCA analysis. All the PCA files are available in Zenodo.

3.3. GNPS Data Processing

Next, the MZmine processed data files of each sample were collectively analyzed in GNPS. The purpose of this comprehensive analysis was to demonstrate the cyanobacterial chemical space covered by the cyanobacteria studied here. Data processing in GNPS enabled the visualization and annotation of the chemical space, identifying structurally related compounds and clustering them into molecular families. The molecular networks calculated by the FBMN method do highlight the complexity of the chemical space obtained from these cyanobacteria (Figure S2). The data were filtered to retain only compounds annotated by the Dereplicator+ method and through MS1 precursor mass comparison with the combined database of expected compounds. This resulted in molecular networks that show the interconnections between different compounds based on their mass spectral similarities and structural characteristics. Accurate annotations could not be made with a satisfactory level of confidence, though (Figure S3). All annotations presented were made using MS1-level data reaching MSI level 3 [62] and their chemical classes could be assumed by their fragmentation patterns, which were yielded by the Dereplicator+ method. A high resolution FBMN annotated figure is available in Zenodo.

3.4. DBsimilarity and t-SNE Plots

To evaluate chemical variability, the combined databases of chemical structures identified and/or isolated from cyanobacteria were compared to the annotated compounds from our samples using the DBsimilarity method (Figure 5 and Figure S4). This enabled the visualization of the chemical space typically associated with cyanobacteria (from the databases) and the placement of the chemical space of the samples from our studied cyanobacteria within it. By generating similarity networks, additional insights into the structural relationships among the compounds were gained, revealing the complexity and connections within the chemical space.

Cytotoxins, including neurotoxins and antibiotics, were annotated, indicating the relevance of studying these strains. Many well-known classes of cyanobacteria compounds were clustered (Figure S4), validating the DBsimilarity as a reliable method for grouping structurally related compounds. Most annotated compounds have a peptidic origin, as well as the compounds in the cyanobacteria databases. The first cluster of the similarity network includes the cyclamides, cyanopeptolins, and anabaenopeptides, and a huge group of other cyclodepsipeptides and cyclic peptides that are centrally located in the cluster. All those classes were annotated in our cyanobacteria samples (Figure S4). This cluster (no. 1) also contains the hassalidins and microviridins, but they were not annotated in our samples. In the second big cluster (no. 2), aeruginosins, microginins, microcolins, and other analogs were grouped. Compact cluster 3 includes the variants of microcystins and cluster 4 contains chlorinated acyl amides, including taveuniamides, columbamides, malyngamides, and other aliphatic compounds. Monoterpene indole alkaloids such as hapalindoles and ambiguines were grouped in cluster 5. Cyclophanes (no. 6), aplysiatoxins and oscillatoxins (no. 7), cryptophycins (no. 8), and saxitoxin and its analogues (no. 9) were grouped in different clusters. In cluster 10, there are hybrid polyketide macrolides, such as leptolyngbyolides, scytophycins, and other more diverse analogs. Other compounds, such as mycosporines (no. 11), tumonoic acids (no. 12), cyclodextrins (no. 13), bartolosides (no. 14), doscadenamides (no. 15), biselyngbyolides (no. 16), lyngbyatoxins (no. 17), and jamaicamides (no. 18), were in smaller clusters (Figure S4). Some analogs of the neurotoxin anatoxin A were annotated, while the cytotoxin cylindrospermopsin was absent. Compounds not belonging to the aforementioned classes may also be present in those highlighted clusters because of structural similarity. The complete similarity network with the list of compounds from databases and the annotated ones are available in Zenodo.

In addition, a t-SNE plot was generated using chemical descriptors based on physical and chemical properties calculated using the Python library Mordred [63]. This visualization technique further confirmed the broad distribution of the compounds annotated in this study within the expected chemical space comprising the cyanobacteria compounds. The t-SNE plot demonstrated the diversity and coverage of the dataset, aligning with the expected structural variety of cyanobacterial metabolites (Figure 6A). These findings support the robustness of the dataset and its potential for identifying novel secondary metabolites. Most of the dark chemical matter is shown in the molecular networks, with diverse structural and functional properties. By incorporating information about the cytotoxic activity of compounds from the combined databases as a categorical variable (presence or absence of cytotoxicity, in this case), compounds of interest can be highlighted. Additionally, similar compounds, i.e., closely spaced in the t-SNE plot, annotated in this study might exhibit similar biological responses (Figure 6B). This plot proximity and the potential biological significance will guide the prioritization of samples for in-depth analysis of each cyanobacterial strain of this study.

4. Discussion

This study provides insights into the secondary metabolisms present in strains of cyanobacteria isolated from BCMs located in reef systems in the Southwestern Atlantic and North Pacific. In addition, environmental samples of Moorena species were also analyzed. Our research aimed to not only produce a more complete picture of the cyanobacterial components of these mats, but also to apply new metabolomic data visualization tools to prioritize strains of interest. This study successfully isolated and characterized cyanobacterial strains from benthic mats in two distinct reef systems, Abrolhos (Brazil) and Ishigaki (Japan), offering valuable insights into their phylogenetic relationships and secondary metabolite profiles. The combination of phylogenetic analysis using 16S rRNA gene sequences and application of an LC-HRMS data analysis and data processing workflow provides information about the diversity and potential biotechnological relevance of these cyanobacteria. Establishing a molecular network, aided by the visualization tool DBsimilarity, produced interactive similarity networks that grouped several classes of known cyanobacterial metabolites.

Importantly, it appears that isolation and cultivation of many of these strains in the laboratory did not result in a repression of secondary metabolism, as we collected a rich chemical dataset for further analysis. The production of these secondary metabolites under lab conditions is not reflective of the biosynthesis that occurs in the natural environment but instead indicates the metabolic potential encoded into the genomes of these cyanobacteria.

4.1. Phylogenetic Relationships and Biogeography

The phylogenetic analysis revealed interesting biogeographic patterns among the isolated strains. For example, the close relationship between the Adonisia-like strains from both Abrolhos and Ishigaki suggests a broader distribution of this cyanobacterial genus than previously reported. This observation aligns with earlier studies that identified cyanobacterial taxa as widely dispersed across marine environments, likely facilitated by ocean currents and similar ecological niches [64]. The close phylogenetic grouping of strains such as CCMR0074 from Ishigaki with Adonisia turfae strains from Abrolhos suggests that this genus, previously only known from the South Atlantic, may also inhabit Asian coral reefs. This highlights the potential for underexplored marine cyanobacterial diversity in benthic mats across different regions of the world’s oceans. Furthermore, the phylogenetic proximity of certain cyanobacterial strains to genera known for bioactive compound production, such as Leptothoe and Moorena, underscores the importance of these cyanobacteria as sources of secondary metabolites. The distinct clustering of Moorena spp. from Abrolhos in both the phylogenetic tree and the chemical analysis suggests that these strains may produce unique compounds, justifying further in-depth studies on their biosynthetic potential.

4.2. Chemical Diversity and Metabolic Profiling

Chemical profiling of the isolated cyanobacteria revealed a rich diversity of secondary metabolites, with several strains producing compounds that cluster with well-known cytotoxic and bioactive cyanobacterial metabolites. LC-HRMS analysis on this number of samples generates a huge amount of data and the application of PCA provides a clear visual representation of the chemical variability present within the cyanobacteria samples. Notably, strains such as CCMR0069, CCMR0074, and CCMR0076 exhibited significant chemical variability, indicating that these strains may harbor unique biosynthetic gene clusters responsible for novel metabolites and/or that they are more amenable for expressing the chemical potential under unnatural lab conditions. Moreover, the correlation between phylogenetic relationships and metabolite profiles remains inconclusive.

The clear differentiation of samples from Moorena A01, A02, and PAB, which exhibited more distinct chemical profiles compared to the other cyanobacterial isolates, suggests a higher degree of metabolic versatility within this genus. This is consistent with previous reports on the ecological role of Moorena spp. in producing diverse metabolites, likely due to their larger genome size and higher number of biosynthetic gene clusters [26], as well as their intimate association with other microorganisms [65].

Additionally, the molecular network analysis in GNPS revealed complex secondary metabolite space occupied by these cyanobacteria. However, many compounds could not be annotated with high confidence, suggesting that the cyanobacteria studied here may produce novel metabolites or structural analogs of known compounds that were not identified in these analyses. The DBsimilarity method proved very useful for grouping similar compounds based on their chemical structures, with annotations highlighting several classes of well-known metabolites, such as microcystins, aeruginosins, anabaenopeptins, and others [66]. The identification of saxitoxin analogs in some samples is particularly noteworthy, as this neurotoxin class has significant environmental and public health implications [67]. This finding is particularly important given the role of cyanobacteria in producing pharmacologically active compounds, such as toxins, antibiotics, and anticancer agents [2,65].

Despite the promising results, there are limitations to this study. The lack of multiple replicates for certain strains and the incomplete annotation of some chemical compounds may hinder the full understanding of their biosynthetic capabilities. Future studies in our research group are focused on expanding the number of replicates and improving metabolite annotation through additional analytical techniques, such as nuclear magnetic resonance (NMR) and bioactivity assays, to validate the chemical structures, assess their biological functions and contribute to FAIR (Findable, Accessible, Interoperable and Reusable) principles.

4.3. Implications for Future Research and Biotechnological Potential

The findings from this study highlight the biotechnological potential of cyanobacteria isolated from marine benthic mats. The high diversity of secondary metabolites, combined with the phylogenetic novelty of certain strains, suggests that these cyanobacteria could serve as valuable resources for drug discovery and other industrial applications. The presence of cytotoxins and antibiotics in the metabolomic profiles indicates their potential for the development of pharmaceuticals as well as their environmental impact and importance. Further exploration of the genetic regulation of secondary metabolism in cyanobacteria, as well as studies on environmental factors that may influence metabolite production and their environmental roles, are among our next goals.