**2. Results**

A total of 20 cyanobacterial strains were successfully isolated, and grown from green biofilms collected from distinct locations in Guadeloupe, namely the Manche-à-Eau lagoon close to red mangrove trees (two stations ST1 and ST2 and five strains), the Marina Bas-du-Fort (one station ST4, one strain), and the Canal Des Rotours (three stations ST5, ST6, and ST7 and 14 strains, Figure 1 and Table 1). Biofilms were found either covering the sediment surface, attached to immersed roots of mangrove trees, or attached to sunken deadwood.

**Figure 1.** Google map showing the collection sites (ST: stations) located in Grande-Terre (ST4, ST5, ST6, and ST7) and in Basse-Terre (ST1 and ST2), the two main islands constituting Guadeloupe.

### *2.1. Phylogenetic and 16S rRNA Dissimilarity-Based Identification of Strains*

The 16S rRNA-encoding gene sequences from the 20 strains clustered within the three cyanobacterial orders Oscillatoriales, Synechococcales and Nostocales (ten, eight, and two strains, respectively, Figure 2). Strains were a ffiliated to hypothetical species and genera based on widely accepted 16S rRNA similarity cuto ff values for species and genus delimitation [7,10], respectively, and monophyly with members of these species and genera. A large-scale comparison of 6787 genomes from 22 prokaryotic phyla established that a 99% 16S rRNA similarity cuto ff value should be retained to delimit species [10]; and reference taxonomic works on Cyanobacteria recommend a 95% cuto ff of cyanobacterial genera delimitation [7].

Five of the ten Oscillatoriales strains clustered with sequences from uncultured *Oscillatoria* and a strain assigned to *Kamptonema formosum* (formely *Oscillatoria formosa*). Using the aforementioned criteria, these strains are congeneric, and can be assigned to three new hypothetical species represented by strain PMC 1075.18, strains PMC 1068.18 and 1076.18, and strains PMC 1050.18 and 1051.18, respectively (Table 1 and Table S1). Strains PMC 1056.18 and 1057.18 (0.8% dissimilarity) from Manche-à-Eau represent one hypothetical new species that displays 4% dissimilarity with the closest described genus, *Arthrospira*, and thus probably belong to this genus. Strains PMC 1071.18 (7.8% dissimilarity with *Symploca* sp. NAC 12/21/08-3), and PMC 1072.18 (5.1% dissimilarity with *Ramsaria avicennae* and *Coleofasciculus chthonoplastes*) represent two new species, each belonging to a new genus based on the 95% similarity threshold. Strain PMC 1092.19 displayed 2.2% dissimilarity with *Lyngbya* sp. ALCB114379 and thus likely represents a new species within this genus. Within the Synechococcales, five strains (PMC 1073.18, 1074.18, 1078.18, 1079.18, and 1080.18, all from Canal des Rotours) clustered with genus *Jaaginema* and represent a single new hypothetical species. Strain PMC 1066.18 was closely related (0.5% sequence dissimilarity) to a sequence from an uncultured *Nodosilinea* sp. CENA 322, isolated from leaves of the mangrove tree *Avicennia schaueriana* in Brazil [11]. Sequence from strain PMC 1064.18 was highly similar (0.5% dissimilarity) to sequences from two unpublished strains of *Limnothrix*. Strain PMC 1052.18 displayed at least 7.4% dissimilarity to all other sequences in databases, and represents a new species within a new undescribed genus. Finally, strains PMC 1069.18 and 1070.18 from Canal des Rotours were the only two Nostocales, clustering together and representing a single new hypothetical species belonging to the genus *Scytonema.*

### *2.2. MS*/*MS Analysis and Annotation of Cyanobacterial Specialized Metabolites*

In an attempt to map the chemical diversity of the 20 cyanobacterial extracts, their LC-MS/MS data were preprocessed using MZmine 2 [12] and the obtained 2468 mass features were organized into a molecular network consisting of 156 clusters (two or more connected nodes of a graph, Figure 3). In order to visualize the distribution of the cyanobacterial metabolites across the 20 extracts, the whole molecular network was mapped at the genus identification level using a typical color tag (Figure 3). An examination of the network reveals that certain clusters are constrained to specific genera. This observation highlighted the distribution of closely related ye<sup>t</sup> different chemical structures in each genus (See supporting information for further details). Moreover, MS/MS data constituting the entire molecular network were searched against the GNPS spectral libraries [13] and yielded only 9 hits (triangle shapes on the network), including a nucleotide, diketopiperazines and phosphocholines (https://gnps.ucsd.edu/ProteoSAFe/result.jsp? task=9581427a15b7422d8bd2b3b4b086189e&view=view\_all\_annotations\_DB). To further expand the annotation coverage, we applied DEREPLICATOR, a recently developed dereplication algorithm that enables high-throughput peptide natural products (PNPs) identification from their tandem mass spectra [14–17]. Interestingly, this tool allows to putatively identify an unknown PNP from its known variants (for example, with a substitution, a modification or an adduct) through the so-called variable dereplication process (as opposed to the strict dereplication when a PNP is described in the database). This dereplication process allowed the annotation of 54 peptide spectrum matches (PSMs) (red ellipses on the network, Figure 3) analogues closely related to known ones (Table S2). Even though, no perfect match was generated through DEREPLICATOR algorithm, marine peptide natural products were proposed and therefore support the relevance of this tool for the exploration of peptide diversity. Furthermore, on the basis of the taxonomic annotation, some clusters were restricted to a single genus indicating the chemical uniqueness of the features within the whole map and, potential structural originality (Figure S1) [18].


**Table 1.** Cyanobacterial strains isolated from Guadeloupe habitats (see map on Figure 1). Strain ID corresponds to the reference number in the Paris Museum Collection (PMC) of cyanobacteria from which strains are available upon request. Affiliation is according to the 16SrRNA-based phylogenetic analysis displayed on

### *Mar. Drugs* **2020**, *18*, 16

**Figure 2.** Phylogenetic tree based on the maximum-likelihood analysis of the bacterial 16S rRNA-encoding gene. Sequences from this study are in bold and colored. Support values at nodes were obtained from 100 boostrap replicates. Scale bar represents estimated 5% sequence difference.

**Figure 3.** The molecular network obtained through the LC-MS/MS analysis of the 20 extracts of cyanobacterial strains collected in Guadeloupe. Peptides dereplicated via DEREPLICATOR are visualized in red ellipses. Nodes (round shape) are colored according to the mean precursor ion intensity per cyanobacterium genus (For further details See Figure S1).

### *2.3. Evaluation of the Antimicrobial Activity*

The evaluation of the antimicrobial activity was performed against six human pathogenic bacteria (*Escherichia coli*, *Klebsiella pneumoniae*, *Pseudomonas aeruginosa*, *Enterococcus faecalis*, *Staphylococcus aureus,* and *Bacillus cereus*) and four marine environmental pathogenic bacteria (*Vibrio alginolyticus*, *Vibrio anguillarum*, *Pseudoalteromonas atlantica,* and *Pseudoalteromonas distincta*). All the cyanobacterial strains extracts were tested in triplicates at a concentration of 100 μg/mL. Only positive results (i.e., inhibition above 50%) are presented in Table 2.

Most of the cyanobacterial strains showed moderate activity against *P. atlantica* with a mean growth inhibition of 20% to 60%; three *Jaaginema* strains PM 1078.18, 1079.18 and 1080.18 as well as *Oscillatoria n. sp. 2* PMC 1076.18 displayed highest growth inhibition properties.


**Table 2.** Summary of significant antibacterial activities among the 20 cyanobacterial strains expressed as percentage of growth inhibition against pathogenic human and marine environmental bacteria.

n.a.: not active at the concentration of 100 μg/mL (i.e., inhibition below 50%); Chloramphenicol was used as positive control in both human *E. coli* and marine environmental *P. atlantica* pathogenic bacteria. Values presented as the mean ± SEM (*n* = 3).

Interestingly, *Oscillatoria*/*Kamptonema* strain PMC 1051.18 was the only strain to reveal significant activity against *E. coli* with 100 % growth inhibition at 100 μg/mL, while its conspecific strain PMC 1050.18 did not show significant activity.

### **3. Discussion and Conclusions**

### *3.1. Mangroves of Guadeloupe are a Source of Novel Cyanobacteria*

This study yielded 20 new cyanobacterial strains including ten Oscillatoriales, eight Synechococcales and two Nostocales [7]. Based on 16S rRNA phylogeny and similarity criteria, these possibly represent 12 new species within three new and seven already known genera (*Jaaginema*, *Scytonema*, *Oscillatoria*, *Nodosilinea, Lyngbya, Arthrospira,* and *Limnothrix*). Ascertaining the exact status of strains as new species or genera requires further characterization beyond simple 16S rRNA dissimilarity- and phylogeny-based criteria considered here. A polyphasic approach is required in order to offer a proper taxonomic description, that is beyond the scope of this paper [7,19]. However, this result based on the analysis of 16S rRNA-encoding genes alone, already emphasizes the high level of taxonomic novelty found in Guadeloupe coastal environments. Apart from a recent paper describing three cyanobacteria not closely related to strains described herein, namely *Oscillatoria* sp. clone gwada strain OG (displaying over 5 % 16S rRNA sequence dissimilarity with our *Oscillatoria*/*Kamptonema* sequences), '*Candidatus* Planktothricoides niger' strain OB and '*Candidatus Planktothricoides rosea*' strain OP, very little data is available regarding cyanobacterial taxonomic diversity in Guadeloupe [8]. Overall the level of novelty documented here reveals mostly untapped cyanobacterial diversity in Guadeloupe. Various authors have highlighted the potential of tropical ecosystems, in particular mangroves, as reservoirs of cyanobacterial diversity and bioactive molecules [11,20], and the predominance of orders Oscillatoriales and Synechococcales among recovered strains [21,22], so this novelty is not unexpected and warrants further investigations.

### *3.2. The Benthic Cyanobacterial Strains Reveal a High Level of Chemical Novelty*

The presence of 54 peptides, variants of known previously identified peptides, was documented using the DEREPLICATOR algorithm. These latter include eleven peptides previously isolated from yanobacteria (Table 3 and Figure S2) but also twelve known peptides from marine Sponges, three from Chordata, five from Fungi (including two marine fungi), and ten from Bacteria, whose one marine bacterium (Table S2). These results sugges<sup>t</sup> that some of the previously isolated peptides from marine organisms could have a symbiotic origin and could be in fact produced by cyanobacteria associated with these organisms.


**Table 3.** Eleven peptide natural products (PNPs) previously isolated from cyanobacteria (in the increasing order of *P* values) dereplicated by DEREPLICATOR.

\*: Peptide mass (in Da) found in our extracts and previously isolated from Cyanobacteria.

The comparison of ions observed in the chemical extracts of the 20 strains also revealed high heterogeneity in chemical composition despite that strains were grown under similar conditions (temperature, Z8 media, photoperiod). This indicates high levels of inter-strain variability.

The assessment of those cyanobacterial extracts using molecular networking and in silico annotation tool finally pointed out a high level of chemical novelty. Notably, this endeavor will allow prioritizing cyanobacterial strains for further chemical studies that will be pursued with full structure elucidation of each isolated compound through detailed NMR analyses.

### *3.3. Certain Cyanobacterial Strains Reveal Promising Antimicrobial Activities*

All strains inhibited the growth of *P. atlantica* to a certain extent. The most active strains, namely PMC 1078.18, 1079.18, and 1080.18, displaying more than 50% of inhibitory activity, belong to a single hypothetical new species of the genus *Jaaginema* and were isolated from the Canal des Rotours. The two other conspecific strains PMC 1073.18 and 1074.18 did not display a significant inhibition, suggesting that this property is limited to certain strains within this species. More interestingly, a single strain of *Oscillatoria*/*Kamptonema*, namely PMC 1051.18, isolated in the Manche à Eau lagoon, showed a very high inhibition against the human pathogenic bacterium strain *E. coli* with 100% of inhibition at the concentration of 100 μg/mL. This promising activity stimulates a deeper chemical study of this strain in order to isolate and identify the molecule(s) responsible for the detected activity. The antimicrobial activity assay will also be enlarged with assays against resistant pathogenic bacterial strains. Surprisingly, its closest relative in the phylogenetic tree, strain PMC 1050.18 did not display such an activity, again suggesting a strain-specific level of activity.

Altogether these findings support that activities are limited to certain strains of a given cyanobacterial species. Inter-strain variability is most often attributed to the existence of certain genes or pathways that are found in some, but not all strains of a given species [23]. This could explain some of the chemical and activity differences observed herein between closely related strains. Alternatively, recent studies on cyanobacterial strains maintained in culture have revealed the existence of an overlooked associated cyanosphere, i.e., a cohort of other microorganisms (mostly bacteria) that is co-isolated and co-cultured with the cyanobacterium [24]. In this case, as observed in other organisms that co-exist with a microbiota, this cyanosphere certainly interacts and influences the

cyanobacterial strains physiology, possibly being in obligate symbiosis, a consequence being the reported difficulty to obtain axenic cultures in Cyanobacteria [24]. These interactions might result in strain-specific differences in chemical composition and activities, either by modulating cyanobacterial gene expression, or because compounds and activities are in fact due to other members of the cyanosphere, and not the cyanobacterium itself. This may explain why different strains within a single species and cultured under similar conditions can display very different compounds and properties, and emphasizes the necessity to account for this heterogeneity by investigating various strains in parallel within each species.

In conclusion, easy-to-sample coastal areas in Guadeloupe harbor an untapped diversity of benthic cyanobacteria that probably represent novel lineages and display a diversity of potentially novel molecules, some of which have promising antimicrobial properties. The taxonomic affiliation of the strains has to be further investigated using polyphasic approaches (that include morphological, ultrastructural and molecular analyses). The different types of activities of the isolated peptides have also to be further explored. It certainly emphasizes the need to further investigate the different habitats in Guadeloupe, in particular mangroves, and more generally tropical coastal habitats, as environmental conditions have a major impact on cyanobacterial diversity [20]. To understand the significance of these Cyanobacteria in these ecosystems, and the possible roles of antibacterial compounds in nature, work is under way to address the actual abundance and functions of isolated strains in biofilms in natura using additional approaches including metagenomics, metabolomics and metatranscriptomics. This study paves the way for further promising investigations on benthic cyanobacteria from tropical mangroves.

### **4. Materials and Methods**

### *4.1. Sampling and Strain Isolation*

Green biofilms, presumed to contain abundant cyanobacteria, were sampled in July 2018 from distinct locations in Guadeloupe, namely two stations located in the Manche-à-Eau lagoon close to red mangrove trees (*Rhizophora mangle,* [25], five strains), the Marina Bas-du-Fort (one strain), and the Canal Des Rotours, a 6-km long canal build between 1826 and 1830 connecting the city of Morne-à-l'eau to the sea on three stations, representing a transition between the coastal mangrove (stations 5 and 6 with six and three strains, respectively) and freshwater (station 7 with five strains, Table 1). In accordance with Article 17, paragraph 2, of the Nagoya Protocol on Access and Benefit-sharing, a sampling permit was issued and published (https://absch.cbd.int/database/ABSCH-IRCC-FR-246959). Biofilms, either benthic or attached to submerged tree roots or branches (periphyton) were sampled by snorkeling. Back to the lab, biofilms were examined under a binocular and individual cyanobacterial morphotypes were transferred to plates containing solid Z8 medium [26] containing 0, 20, and 35 g/<sup>L</sup> salt (Instant Ocean, Aquarium Systems, France). Isolation in liquid Z8 medium was also attempted.

### *4.2. Strain Cultivation and Biomass Production*

Back to the laboratory, surviving non-axenic strains were stabilized. Strains were registered in the Paris Museum Collection (PMC) under labels PMC 1050.18 to PMC 1092.19 (Table 1). They are maintained in liquid medium and are available upon request (https://www.mnhn.fr/fr/collections/ensembles-collections/ressources-biologiques-cellulesvivantes-cryoconservees/microalgues-cyanobacteries). Biomass was produced for two months in increasing volumes of liquid Z8 media (25 ± 1 ◦C; 15 μmol/m<sup>2</sup>/s white light; 16 h light: 8 h dark) without salt (PMC1069.18, 1070.18, 1071.18, 1072.18, 1073.18, 1092.18) or with 20 g/<sup>L</sup> salt (other strains).
