The four selected lagoons are public space for leisure and attract thousands of visitors weekly for recreational purposes. Visitors include families with children, cyclists, and animals, which can interact with these waters.

Bloom samples were collected from four different lagoons in São Paulo state, Brazil, during September 2010. ESALQ1 (22˚42'34.49''S, 47˚37'57.80''W) and ESALQ2 (22˚42'32.78''S, 47˚37'50.22''W) lagoons are located in the University of São Paulo, Piracicaba, São Paulo state, Brazil, with 1.4 m average depth and 161.0 m² surface area, and 2.0 m average depth and 1.5 m² surface area, respectively. ESALQ1 contains a fountain, which provides water aeration. Taquaral lagoon (22˚52'25.2''S, 47˚3'0.51''W) is located in Portugal Park, Campinas, São Paulo state, Brazil, with 2.8 m average depth and 146.5 m² surface area. Limeira lagoon (22˚30'56.06''S, 47˚23'55.94''W) is located in a sugarcane farm, Limeira, São Paulo state, Brazil, with 2.5 m average depth and 27.6 m² surface area.

All samples were collected in triplicate using 4 L amber glass flasks close to the border of the lagoons at a depth of 0 - 20 cm and stored at 4˚C for no longer than 24 h. Afterwards, 1 L of each sample was concentrated by centrifugation at 13,000 ×g for 15 min. at 4˚C. The supernatant was removed and the pellet stored at −20˚C. The pellet was dimidiate and used for the subsequent molecular and chemical analyses.

Total DNA was extracted with MoBio PowerSoil DNA Isolation Kit (MoBio Laboratories) according to the manufacturer’s instructions. The integrity and purity of the extracted DNA was verified by electrophoresis in 1% agarose gel stained with SYBR Green I (Molecular Probes) using Low Mass DNA Ladder (Invitrogen) as molecular marker. The gel was photodocumented in a Kodak Gel Logic 212 Imaging System (Molecular Imaging System Carestream Health). Afterwards, PCR of the 16S rDNA was performed in a Techne TC-412 thermocycler (Bibby Scientific) with the primers 27F1 and 1494 Rc under previously described conditions [12] .

The products of the 16S rDNA PCR were used as templates for a nested PCR with the primers CYA359F (with the addition of 40 bases GCclamp), CYA781Ra and CYA781Rb under previously described conditions [13] , which was verified by electrophoresis in agarose as mentioned above. The resulting amplicons were separated by DGGE in 6% polyacrilamide gel with a denaturing gradient of urea and formamide varying from concentrations of 20% to 60% according to [14] . Electrophoresis of the polyacrilamide gel was performed in Ingeny phor U2 (Ingeny) at 60˚C with a current of 100 V. After a 15-hour run, the gel was silver-stained [15] and photodocumented.

The 16S rDNA amplicons from the environmental samples were used as templates for a nested PCR with the primers CYA359F (without a GCclamp), CYA781Ra and CYA781Rb as described above. The fragments amplified by the nested PCR were cloned with the pGEM-T Easy Vector kit (Promega) and used to transform One Shot TOP 10 Eletrocomp E. coli cells (Invitrogen) according to the manufacturer’s instructions. Plasmids of successfully transformed cells were extracted by alkaline lysis [16] , verified by PCR and sequenced with DYEnamic ET Terminator Cycle Sequencing Kit (GE Healthcare) in ABI PRISM 3100 Genetic Analyser (Applied Biosystems). The resulting sequences were trimmed to remove vector and sequences with bases of phred quality values lower than 20 using Geneious Pro 4.7.5 [17] . The sequences were aligned with ClustalW [18] . A similarity matrix was produced from the alignment by the DNADIST software [19] with the JukesCantor evolutionary model. The sequences were grouped into Operational Taxonomic Units (OTU) using the software DOTUR [20] , in which they were binned using a cutoff of 95% of similarity at the nucleotide level based on generic concept [21] .

DOTUR was also used to generate rarefaction curves data, diversity (Shannon) and species richness (Chao1) indices for each library. To evaluate whether the populations were statistically distinct, ∫-Libshuff software was used [22] . Finally, one sequence from each OTU obtained was aligned together with closely related sequences retrieved from GenBank and used to generate a Neighbor-Joining phylogenetic tree with the Kimura-2 parameter and a bootstrap value of 1000 in MEGA5 [23] .

The presence of cyanobacterial genes involved in toxins and protease inhibitors biosynthesis in the studied lagoons was firstly evaluated by PCR amplification. The PCR for cyanobacterial toxin genes was conducted using the following primers: anatoxin―ALM_F/ALM_R [24] ; cylindrospermopsin―CYLATF/CYLATR [25] and cynsulF/cylnamlR [26] ; microcystin―mcyE_F2 and mcyE_R4 [27] and saxitoxin OCT-F/OCT-R [28] , according to the conditions originally detailed in the publications describing each of these primers pairs (Supplementary data: Table S1).

The proteases inhibitors synthetases were investigated using the primers for aeruginosin aerA_F/aerB_R and cyanopeptolin mcnC_F/mcnE_R [29] . The amplification of these genes was verified by electrophoresis in 1% agarose gel by comparison with 100 bp DNA ladder (Invitrogen).

Water samples were centrifuged at 9000 × g at 22˚C for 5 min. The supernatant was discharged and 5 mL of methanol and glass beads of 3 mm diameter were added to the pellets, following vortex agitation for 1 min. The samples were concentrated by centrifugation at 9000 × g for 5 min. and the supernatant was collected. Extracts were then dried at 55˚C and subjected to a solid-phase extraction on a C₁₈ reversed-phase cartridge (Phenomenex^® 10 g), using gradient in H₂O in methanol. Fractions were evaporated, diluted in a solution of methanol: water:formic acid (50%:50%:0.1%) and analysed using a LC/MS Agilent 6410 Triple Quadrupole equipped with Agilent 1200 Series Binary Pump SL pumping system and Agilent 1200 Autosampler (Agilent Technologies Inc. Santa Clara, CA, USA). All the parameters for data analyses were performed according to [30] . The mass spectrometry analysis was performed via direct infusion of the extracts into the ESI source of the mass spectrometer. Molecular ions were scanned in the range of 100 to 1800 Da. Cone voltage was set to 50 V and collision energy of 60 eV was used.

DGGE analysis was performed to represent the diversity of cyanobacteria present in the lagoons in order to direct the subsequent analyses. DGGE banding patterns showed low variation among the communities from the different water samples and suggested that there was a predominance of few OTUs in these locations (data not shown). Limeira, Taquaral and ESALQ2 lagoons showed nearly the same banding patterns. On the other hand, ESALQ1 lagoon presented the most different band profile.

The analyses of sequences obtained in clone libraries resulted in a total of 233 clones, distributed as follows: 59 from ESALQ1, 57 from ESALQ2, 57 from Limeira and 60 from Taquaral lagoons (Table 1). In total, nine different known cyanobacterial genera belonging to four different orders were identified in clone libraries when the sequences were compared with those available in the NCBI GenBank database.

Taquaral lagoon presented low richness and diversity values, revealed by the Chao1 and Shannon indices (Figure 1(a), Figure 1(b)), confirming the low cyanobacterial diversity observed in the DGGE methodology. Also, the sequences obtained for this sample match with only one known cyanobacterial genus, 49 clones out of 60 are related to Microcystis sequences and the remaining clones correspond to uncultured cyanobacteria (Table 1). Limeira and ESALQ2 water samples presented clones of three (Brasilonema, Microcystis and Pseudanabaena) and four (Anabaena, Microcystis, Nostoc and Pseudanabaena) cyanobacterial strains from different genera, respectively (Table 1). In both cases, the predominantly observed genus was Microcystis (46 and 51 clones, respectively). Only few OTUs were not possible to be affiliated to known cyanobacterial genera in those cases. Lagoons ESALQ2, Limeira and Taquaral had 89%, 82% and 81%, respectively, of their sequences related to the genus Microcystis. The sample that showed the highest richness and diversity of cyanobacteria species was ESALQ1 lagoon (Figure 1(a), Figure 1(b)), presenting five different genera (Anabaena, Cylindrospermopsis, Limnococcus, Synechococcus and Woronichinia) (Table 1). In comparison to other lagoons, Limeira lagoon showed a high number of uncultured cyanobacteria. Interestingly, no sequences matching Microcystis was observed in this lagoon. The estimated sample coverage (ESC) ranged from 81% to 95% for all lagoons (Table 1). Hence, it was possible to consider that the sampled environments were well represented even presenting a low number of clones. This assumption was confirmed by the rarefaction curves calculated for the clone libraries, which were close to reaching the asymptote for all samples, thus indicating that the probability of finding new OTUs was very low, especially in ESALQ2 and Limeira samples (Figure 2). The libraries reciprocal coverage evaluated by ∫-Libshuff showed that the lagoons differ statistically from each other, with the exception of Taquaral and ESALQ2, which showed a p value of 0.4914, indicating an overlapping of sequences in these samples.

¹No. of clones of each genus found per lagoon; ²Operational taxonomic units; ³Estimated sample coverage.

When the sequences of all clone libraries were grouped together, a total of 53 OTUs was found, with 27 singletons among them (13 from ESALQ1, 4 from Taquaral, 6 from ESALQ2 and 4 from Limeira). No OTU was observed in common for all the lagoons. OTU 1 was the most abundant in number of sequences (59), including 21 from ESALQ2, 21 from Limeira and 17 from Taquaral. The representative sequence of this OTU grouped with Microcystis sequences and with others OTUs in the phylogenetic tree; OTUs 2, 3 and 8 also presented sequences from those three lagoons, but in lower numbers; nine OTUs (4, 6, 9, 11, 13, 14, 17, 18 and 26) were observed only in ESALQ1; three OTUs (12, 19 and 20) were observed only in Taquaral lagoon; four (16, 21, 23 and 24) were observed in Limeira lagoon; one (25) was observed in ESALQ2 lagoon; OTUs 5, 15 and 22 were represented by sequences from Limeira and ESALQ2 lagoons; OTU 7 was represented by sequences of Taquaral and ESALQ1 lagoons; OTU 10 was represented by sequences from Taquaral and ESALQ2 lagoons.

The phylogenetic tree was constructed using a representative sequence of each 53 OTUs. Among these OTUs, 30 grouped with sequences from genera identified as representative of the order Chroococcales (Gloeocapsa, Limnococcus, Microcystis, Snowella and Woronichinia); five grouped with genera from the order Nostocales (Brasilonema, Cylindrospermopsis and Nostoc); five grouped with genera from the order Synechococcales (Synechococcus and Prochlorococcus); five with genera from Pseudanabaenales order (Leptolyngbya and Pseudanabaena); three with genera from the order Oscillatoriales (Halospirulina and Planktothrichoides); and five OTUs did not group with cyanobacteria from any known genus (Figure 3).

The PCR-amplification of genes encoding sequences related to proteases inhibitors synthetase biosynthesis resulted in positive results for aeruginosin in samples from Limeira, Taquaral and ESALQ2 lagoons and for cyanopeptolin in Limeira and ESALQ2 lagoons (Table 2).

Genes involved in microcystin production were observed in Limeira, Taquaral and ESALQ2 lagoons, and with saxitoxin only in ESALQ1 lagoon. It was not possible to verify amplification for anatoxin and cylindrospermopsin to the evaluated samples (Table 2).

Chemical analysis performed by LC-MS/MS confirmed the presence of aeruginosin in Limeira, Taquaral and ESALQ2 samples; and cyanopeptolin and microcystin in Limeira samples (Table 2). Anatoxin, saxitoxin and cylindrospermopsin were not detected in any sample. Microcystin-LR was identified in the sample of Limeira lagoon and showed two variants, m/z 981 and 995 (Supplementary data: Figure S1 and Figure S2). The fragmentation of the toxins was performed and confirmed by the presence of the ion m/z 135, corresponding to the unusual Adda group, the molecule responsible for the toxicity of microcystin. The fragmentation of ion at m/z 981 observed in Limeira sample indicated that the molecule is the variant (D-Asp³) microcystin-LR due to the presence of the ions m/z 70, 86, 135, 213 and 375, corresponding to Arg, Leu (immonium ion), Adda, Glu + Mdha + H and Adda + Glu + Mdha, respectively [31] . In Figure 4 was showed the generalized structure of microcystin-LR.

Fragmentations of aeruginosins showed a strong signal of the ion at m/z 140, corresponding to the Choi immonium-(acid 2-carboxy-6-hydroxyoctahydroindole), which is indicative of aeruginosin. The fragmentation of two aeruginosins at m/z 561 and 593 (Supplementary data: Figure S3 and Figure S4), ions at m/z 309, 291 and 266 were detected, indicating the presence of Choi-Argininal. Fragments associated with Hpla (lactic

^*mcy―microcystin; sxt―saxitoxin; aer―aeruginosin; mcn―cyanopeptolin; ana―anatoxin and homoanatoxin; cyr―cylindrospermopsin.

hydroxyphenyl) showed an increase in mass corresponding to the level of chlorination in the molecule: aeruginosin m/z 593 and 601 (Supplementary data: Figure S5 and Figure S6) referred to the molecule without chlorine (ion m/z 250); aeruginosin m/z 601, 635 and 669 (Supplementary data: Figures S7-S9) presented one chlorine (ion m/z 284); aeruginosin m/z 627 (Supplementary data: Figure S10) contained two chlorines (ion m/z 318). Only aeruginosin m/z 593 has been previously reported [32] . In Figure 5 was showed the generalized structure of aeruginosin.

In the fragmentation spectra of cyanopeptolins they all showed a strong signal of the ion m/z 150 (MTyr immonium), which is indicative of cyanopeptolin. The occurrence of the fragment m/z 115 (Ahp) also supports the identification of cyanopeptolins (Supplementary data: Figures S11-S15). Fragments with molar masses of 127 and 198 Da appear when chlorination of Tyr occurs [33] , which was not observed in any sample. The cyanopeptolins detected showed m/z 505, 959 [34] , 969, 973 [35] , and 987 [36] . In Figure 6 was showed the generalized structure of cyanopeptolin.