Shotgun Metagenomic Sequencing to Assess Cyanobacterial Community Composition following Coagulation of Cyanobacterial Blooms

The excessive proliferation of cyanobacteria in surface waters is a widespread problem worldwide, leading to the contamination of drinking water sources. Short- and long-term solutions for managing cyanobacterial blooms are needed for drinking water supplies. The goal of this research was to investigate the cyanobacteria community composition using shotgun metagenomics in a short term, in situ mesocosm experiment of two lakes following their coagulation with ferric sulfate (Fe2(SO4)3) as an option for source water treatment. Among the nutrient paramenters, dissolved nitrogen was related to Microcystis in both Missisquoi Bay and Petit Lac St. François, while the presence of Synechococcus was related to total nitrogen, dissolved nitrogen, dissolved organic carbon, and dissolved phosphorus. Results from the shotgun metagenomic sequencing showed that Dolichospermum and Microcystis were the dominant genera in all of the mesocosms in the beginning of the sampling period in Missisquoi Bay and Petit Lac St. François, respectively. Potentially toxigenic genera such as Microcystis were correlated with intracellular microcystin concentrations. A principal component analysis showed that there was a change of the cyanobacterial composition at the genus level in the mesocosms after two days, which varied across the studied sites and sampling time. The cyanobacterial community richness and diversity did not change significantly after its coagulation by Fe2(SO4)3 in all of the mesocosms at either site. The use of Fe2(SO4)3 for an onsite source water treatment should consider its impact on cyanobacterial community structure and the reduction of toxin concentrations.


Introduction
Cyanobacteria are among the most ancient organisms on Earth, and they are present in aquatic environments worldwide [1,2]. When intense blooms appear, cyanobacteria can produce large quantities of toxins. The increasing frequency and intensity of the toxic In PLSF, at T0, Microcystis was the predominant genus, accounting for 30-37% of total relative abundance that was documented on 24 July 2019, while the highest cyanobacterial relative abundances consisted of Dolichospermum (9-10.5%), Synechococcus (8-12%), and Microcystis (10-12.5%) on 05 August 2019 in all of the mesocosms ( Figure 1C,D). The relative abundance of the dominant genera remained stable in the control mesocosms at T48 on 26 July and 07 August in 2019 in PLSF. In the mesocosms that had the addition of 20 mgFe/L and 35 mgFe/L, the relative abundance of Microcystis decreased to 11.5% and 10%, respectively, at T48 on 26 July 2019 ( Figure 1C). This is coherent with the trends that were observed in the taxonomic cell count results on 26 July 2019 in our previous study [12]. On 07 August 2019, the reduction of 100% of the Dolichospermum in the taxonomic results (Table S1) may be the result of a notable increase in the relative abundance of Microcystis in the mesocosms that had 35 mgFe/L added to them ( Figure 1D).
The results of the microscopic taxonomic cell counts and shotgun metagenomics sequencing do not completely correspond with each other in this study (Figures 1, S1 and S2). Dolichospermum and Microcystis were detected by both of the approaches, but Aphanocapsa, Aphanothece, Coelosphaerium, Merismopedia, and Chroococcus were detected only by microscopic taxonomic cell counts. Moreover, Aphanizomenon was identified to be predominant by the taxonomic cell counts, while the metagenomics test found that Dolichospermum was the most dominant genus at the beginning of the sampling period in all of the mesocosms on 13 August 2019 ( Figures 1B and S2). These observations correspond with previously reported findings wherein the microscopic taxonomic cell counts and high-throughput sequencing do not completely match [6,8].
The limitations of microscopic and shotgun metagenomics approaches may provide an explanation for the differences in the structure of community composition that were In MB, Dolichospermum was the dominant genus (approx. 25-30% of the total relative abundance) and Microcystis, Nostoc, Fischerella, and Calothrix were the representative genera in all of the mesocosms at T0 on 08 August 2018 and 13 August 2019 ( Figure 1A,B, respectively). These results are in accordance with those of several previous studies which indicated that Dolichospermum was a cyanobacterial genus that was frequently found in the blooms in MB [3,6,13,14]. After two days, on 10 August 2018, there was no change in the relative abundance in any of the mesocosms in comparison with the initial composition ( Figure 1A). This observation corresponds with that of a previous report, thereby demonstrating the stability of the cyanobacterial community composition at the genus level after two days of oxidation with H 2 O 2 (20 mg/L) and CuSO 4 (2 mg/L) in the control mesocosms and the mesocosms with an oxidant in Petit Lac St. François [15].
However, a remarkable change in the cyanobacterial composition at the genus level was observed in both the control and the coagulated mesocosms at T48 on 15 August 2019 in MB ( Figure 1B). The most abundant genus shifted from being Dolichospermum to being Synechococcus (approx. 45% of the total relative abundance) and Microcystis (38.7% of the total relative abundance) in the control and the coagulant mesocosms, respectively. This shift can probably be attributed to the marked decrease of the Dolichospermum cell count (greater than 97% reduction) in all of the mesocosms (Table S1). The loss of Dolichospermum biomass in the control mesocosms from 13 August to 15 August in 2019 may be the result of a drop in pH from 8 to 6 which occurred (Tables S4 and S5), thereby potentially interfering with the cyanobacterial growth, in general [16]. This lower pH level could contribute to the dominance of Synechococcus in the control mesocosms because it can flourish at pH 6.5 [17,18]. However, we did not observe such an abundance of Synechococcus in the mesocosms at the doses of 20 mgFe/L and 35 mgFe/L, thereby implying that the Synechococcus cells were well captured by the coagulation process. Microcystis became the dominant genus, although both Dolichospermum and Microcystis were largely removed (>86%) in the coagulated mesocosms on 15 August 2019 (Table S1). This suggests that cell survival and re-growth might have occurred during the two days of the sludge storage process [8]. The competition between the Microcystis and Dolichospermum strains at the laboratory scale was investigated, and it was found that Microcystis significantly inhibited the growth of Dolichospermum [6].
In PLSF, at T0, Microcystis was the predominant genus, accounting for 30-37% of total relative abundance that was documented on 24 July 2019, while the highest cyanobacterial relative abundances consisted of Dolichospermum (9-10.5%), Synechococcus (8-12%), and Microcystis (10-12.5%) on 05 August 2019 in all of the mesocosms ( Figure 1C,D). The relative abundance of the dominant genera remained stable in the control mesocosms at T48 on 26 July and 07 August in 2019 in PLSF. In the mesocosms that had the addition of 20 mgFe/L and 35 mgFe/L, the relative abundance of Microcystis decreased to 11.5% and 10%, respectively, at T48 on 26 July 2019 ( Figure 1C). This is coherent with the trends that were observed in the taxonomic cell count results on 26 July 2019 in our previous study [12]. On 07 August 2019, the reduction of 100% of the Dolichospermum in the taxonomic results (Table S1) may be the result of a notable increase in the relative abundance of Microcystis in the mesocosms that had 35 mgFe/L added to them ( Figure 1D).
The results of the microscopic taxonomic cell counts and shotgun metagenomics sequencing do not completely correspond with each other in this study (Figures 1, S1 and S2). Dolichospermum and Microcystis were detected by both of the approaches, but Aphanocapsa, Aphanothece, Coelosphaerium, Merismopedia, and Chroococcus were detected only by microscopic taxonomic cell counts. Moreover, Aphanizomenon was identified to be predominant by the taxonomic cell counts, while the metagenomics test found that Dolichospermum was the most dominant genus at the beginning of the sampling period in all of the mesocosms on 13 August 2019 ( Figures 1B and S2). These observations correspond with previously reported findings wherein the microscopic taxonomic cell counts and high-throughput sequencing do not completely match [6,8].
The limitations of microscopic and shotgun metagenomics approaches may provide an explanation for the differences in the structure of community composition that were found between these methods. Small-sized and morphologically similar species may escape their detection by microscopy [19,20]. The significant morphological deformation of cyanobacteria species by the coagulation or the oxidation process may also hinder the ability to identify these cells [6,8,21]. Shotgun metagenomics (high-throughput sequencing) is a powerful tool for microbial species or community detection and classification to observe the microbial dynamics through water treatment processes [8]. A wide variety of taxonomic and functional classifier pipelines have been developed and are available to make the analysis of metagenomic sequencing more accessible than they ever have been before. However, the accurate performance of these analyses may be hindered by the process of DNA extraction [20,22], frequent taxonomy database changes over time, and the number of false positive taxa predictions [23].   Figure S3 presents the PCA analysis of the cyanobacterial community composition following its coagulation using different doses which we performed in PLSF. PC1 and PC2 explain 54.0% and 41.6% of the variation in the cyanobacterial community composition among the samples, respectively. The cyanobacterial composition at T0 on 24 July 2019 and 05 August 2019 were clustered separately into two groups (blue and orange, respectively), thus implying that the cyanobacterial communities were distinct ( Figure S3A). At T0 of the study period, on 24 July 2019, in all of the mesocosms, Microcystis was the predominant genus, while on 05 August 2019, the cyanobacterial composition comprised mostly of Synechococcus, Dolichospermum, and Microcystis ( Figure  S3A,B). After 48 h, on 26 July 2019, the cyanobacterial community composition was similar to that of the initial community composition in the control mesocosms, whereas it experienced a substantial shift from Microcystis to Synechococcus and Dolichospermum in the mesocosms containing doses of 20 and 35 mgFe/L ( Figure S3B), thereby confirming the results that are shown in the bar graph ( Figure 1C,D). Similarly, after 48 h in the control mesocosms on 05 August 2019, the cyanobacterial composition was similar to that of the inital composition. A similar trend was observed in the mesocosms that had a dose of 20 mgFe/L, while the cyanobacterial composition changed in the mesocosms containing the dosage of 35 mgFe/L.
The PCA analysis was also utilized to evaluate the relationship between the cyanobacterial community and the cyanotoxins ( Figure S4). A relationship between the total intracellular microcystin (ΣMCs) and the intracellular toxin MC-LR with the genus Microcystis which was detected through metagenomics was found at T0 and T48 in all of the mesocosms. This is consistent with our previous observation using taxonomic cell counts where we found that the highest ΣMCs occured in the periods when Microcystis was dominant [12]. It also suggests that the potentially toxigenic genera that were correlated with the microcystin concentrations were Microcystis and Dolichospermum in a In MB, PC1 and PC2 explained 89.4% and 6.12%, respectively, of the variation in the cyanobacterial community composition among the samples ( Figure 2). The cyanobacterial composition at the start of the study period (T0) on 08 August 2018 and 13 August 2019 was clustered in the blue circle and the orange circle, respectively (Figure 2A), and comprised of mainly Dolichospermum, as shown in Figure 2B, which supports the result that are shown in the bar graph ( Figure 1A,B). Our result is in accordance with that of a previous paper which reported that the community composition did not vary considerably from year-to-year in Missisquoi Bay [14]. After 48 h, on 10 August 2018 and 15 August 2019, there were two different trends in the cyanobacterial composition in the mesocosms. On the one hand, in all of the mesocosms, the cyanobacterial composition clustered in the same blue circle, which is similar to the initial conditions that were observed on 10 August 2018 ( Figure 2A). On the other hand, the cyanobacterial composition displayed a progressive shift in both the control and coagulated mesocosms on 15 August 2019. Synechococcus became more abundant in the control mesocosms (Figure 2A,B), which is in agreement with the results that are shown in the bar graph ( Figure 1B). For the mesocosms with doses of 20 and 35 mgFe/L, the cyanobacterial community composition shifted mainly from being dominated by Dolichospermum to being dominated by Microcystis (Figure 2A,B). This result is in agreement with the recently reported results of the changes in the community structure at the genus level that were monitored in the samples that were retrieved from raw water and sludge supernatant after a coagulation process in a drinking water treatment plant [8]. Figure S3 presents the PCA analysis of the cyanobacterial community composition following its coagulation using different doses which we performed in PLSF. PC1 and PC2 explain 54.0% and 41.6% of the variation in the cyanobacterial community composition among the samples, respectively. The cyanobacterial composition at T0 on 24 July 2019 and 5 August 2019 were clustered separately into two groups (blue and orange, respectively), thus implying that the cyanobacterial communities were distinct ( Figure S3A). At T0 of the study period, on 24 July 2019, in all of the mesocosms, Microcystis was the predominant genus, while on 05 August 2019, the cyanobacterial composition comprised mostly of Synechococcus, Dolichospermum, and Microcystis ( Figure S3A,B). After 48 h, on 26 July 2019, the cyanobacterial community composition was similar to that of the initial community composition in the control mesocosms, whereas it experienced a substantial shift from Microcystis to Synechococcus and Dolichospermum in the mesocosms containing doses of 20 and 35 mgFe/L ( Figure S3B), thereby confirming the results that are shown in the bar graph ( Figure 1C,D). Similarly, after 48 h in the control mesocosms on 05 August 2019, the cyanobacterial composition was similar to that of the inital composition. A similar trend was observed in the mesocosms that had a dose of 20 mgFe/L, while the cyanobacterial composition changed in the mesocosms containing the dosage of 35 mgFe/L.
The PCA analysis was also utilized to evaluate the relationship between the cyanobacterial community and the cyanotoxins ( Figure S4). A relationship between the total intracellular microcystin (ΣMCs) and the intracellular toxin MC-LR with the genus Microcystis which was detected through metagenomics was found at T0 and T48 in all of the mesocosms. This is consistent with our previous observation using taxonomic cell counts where we found that the highest ΣMCs occured in the periods when Microcystis was dominant [12]. It also suggests that the potentially toxigenic genera that were correlated with the microcystin concentrations were Microcystis and Dolichospermum in a study of 22 lakes in southern Quebec [24]. In addition, a recent study observed that Fe 2 (SO 4 ) 3 can remove more than 93% of the Microcystis cells with a dose of 35 mgFe/L in a mesocosm experiment [12]. Therefore, the application of Fe 2 (SO 4 ) 3 may be considered to be an efficient onsite barrier against toxin-producing cyanobacteria when it is added to the source water before it enters a drinking water treatment plant.

Effect of Coagulation on Cyanobacterial Community Richness and Diversity
To assess the effect of coagulation on the cyanobacterial community richness and diversity, the species richness and Shannon indices were utilized (Figure 3). The richness index is used for estimating the number of different species that are present in an environment, and the Shannon index is used for measuring the homogeneity in the abundance of the different species in a sample [25,26].
In MB, the species richness and Shannon indices did not significantly change in any of the mesocosms after 48 h of their exposure (Table S8). The average species richness index decreased in the control mesocosms and in the mesocosms containing 20 mgFe/L, while an opposite trend was observed in the mesocosms containing 35 mgFe/L at T48 (Figure 3). This could be attributed to the reduction in the dominant species with a high abundance, such as Dolichospermum spp. and Microcystis spp., and this could potentially increase the genus richness [14,27]. Furthermore, a prior study demonstrated that a dose of 35 mgFe/L was more efficient than a dose of 20 mgFe/L was in reducing the total cell counts and certain dominant genera such as Dolichospermum, Microcystis, and Aphanizomenon [12]. The Shannon index dropped slightly following the administration of 20 mgFe/L, while it remained almost stable in the control and mesocosms with 35 mgFe/L (Figure 3).
Similarly, there was no significant change in the species richness and the Shannon indices between T0 and T48 in all of the mesocosms in PLSF (Table S8). The decrease in the species richness index was observed in the coagulated mesocosms, while it increased slightly in the control mesocosms ( Figure 3). The decline in the species richness index could be the result of the presence of a genus that could no longer be identified [6]. The Shannon index at T0 (approx. 4) showed that there was a slight change in the diversity profile at T48 (around 4.2) in the control and the 20 mgFe/L mesocosms. In contrast, the Shannon index decreased in the mesocosms with 35 mgFe/L when it was compared with that of the control.

Effect of Coagulation on Cyanobacterial Community Richness and Diversity
To assess the effect of coagulation on the cyanobacterial community richness and diversity, the species richness and Shannon indices were utilized (Figure 3). The richness index is used for estimating the number of different species that are present in an environment, and the Shannon index is used for measuring the homogeneity in the abundance of the different species in a sample [25,26]. In MB, the species richness and Shannon indices did not significantly change in any of the mesocosms after 48 h of their exposure (Table S8). The average species richness index decreased in the control mesocosms and in the mesocosms containing 20 mgFe/L, while an opposite trend was observed in the mesocosms containing 35 mgFe/L at T48 (Figure 3). This could be attributed to the reduction in the dominant species with a high abundance, such as Dolichospermum spp. and Microcystis spp., and this could potentially increase the genus richness [14,27]. Furthermore, a prior study demonstrated that a dose of 35 mgFe/L was more efficient than a dose of 20 mgFe/L was in reducing the total cell counts and certain dominant genera such as Dolichospermum, Microcystis, and Aphanizomenon [12]. The Shannon index dropped slightly following the administration of

Impact of Environmental Conditions on Cyanobacterial Community Composition in the Mesocosms before and after Coagulation
To assess the impact of the environmental conditions on the cyanobacterial communities in MB and PLSF. A redundancy analysis (RDA) was used to evaluate the relationship between the environmental parameters (Tables S3, S4, S5, S6 and S7) and the cyanobacterial communities. The total nitrogen (TN), dissolved nitrogen (DN), dissolved phosphorus (DP), and dissolved organic carbon (DOC) exerted significant effects (ρ < 0.05) on the community profiles in MB (Figure 4).
In MB, in 2018 and 2019, PC1 and PC2 represented 89.4% and 6.12% of the variation, respectively (Figure 4). A relationship between the environmental conditions and the cyanobacterial community composition was not observed in any of the mesocosms from 8-10 August 2018. The result corresponds with that of a previous study which reported that these nutrients (TN, DN, TP, and DOC) did not affect the cyanobacterial community in the supernatant water after coagulation [8]. In contrast, TN, DN, DOC, and DP had a strong impact on Synechococcus in the control mesocosms, while DN contributed to the presence of Microcystis in the coagulated mesocosms at T48 on 15 August 2019 in MB.
In PLSF, PC1 and PC2 represent 54% and 41.6% of the variation, respectively ( Figure S5). A clear association was observed between the DN and Microcystis in the control mesocosms on 24-26 July 2019, in the mesocosms with 20 and 35 mgFe/L at T0 on 24 July 2019, and in the mesocosms with 35 mgFe/L on 07 August 2019, where Microcystis was the dominant genus. In contrast, the DN was negatively related to Dolichospermum at both of the sites. Therefore, the persistence of Dolichospermum in the mesocosms could be the result of its environmental resistance [8]. Our result is in accordance with those of previous studies, reporting that the DN may lead to the dominance of Microcystis [28][29][30].

Impact of Environmental Conditions on Cyanobacterial Community Composition in the Mesocosms before and after Coagulation
To assess the impact of the environmental conditions on the cyanobacterial communities in MB and PLSF. A redundancy analysis (RDA) was used to evaluate the relationship between the environmental parameters (Table S3, S4, S5, S6, and S7) and the cyanobacterial communities. The total nitrogen (TN), dissolved nitrogen (DN), dissolved phosphorus (DP), and dissolved organic carbon (DOC) exerted significant effects ( < 0.05) on the community profiles in MB (Figure 4). In MB, in 2018 and 2019, PC1 and PC2 represented 89.4% and 6.12% of the variation, respectively (Figure 4). A relationship between the environmental conditions and the cyanobacterial community composition was not observed in any of the mesocosms from 08-10 August 2018. The result corresponds with that of a previous study which reported that these nutrients (TN, DN, TP, and DOC) did not affect the cyanobacterial community in the supernatant water after coagulation [8]. In contrast, TN, DN, DOC, and DP had a

•
The shotgun metagenomic sequencing method is a more robust method that is used to identify cyanobacterial communities when it is compared with the microscopic methods.

•
In the beginning of the sampling period, Dolichospermum was the predomiant genus in Missisquoi Bay, while Microcystis was a common genus in all of the mesocosms in Petit Lac St. François.

•
The change in the cyanobacterial composition at the genus level in the mesocosms after two days varied across the studied sites and over the sampling time. Therefore, the choice to use Fe 2 (SO 4 ) 3 as an onsite source-control treatment should be made while considering its impact on the cyanobacterial community structure. • Synechococcus may be satisfactorily removed in coagulated mesocosms.

•
The intracellular microcystin concentrations were strongly associated with the presence of Microcystis.

•
The cyanobacterial community richness and diversity did not change significantly after the coagulation by Fe 2 (SO 4 ) 3 in any of the mesocosms at either of the sites.

•
The dissolved nitrogen content was related to Microcystis in both Missisquoi Bay and Petit Lac St. François, while Synechococcus was influenced by the total nitrogen, dissolved nitrogen, dissolved organic carbon, and dissolved phosphorus contents.

•
The source water coagulation process is a potential option for removing toxin-producing cyanobacterial species and intracellular toxins from the water column.

Study Sites Description
Missisquoi Bay (MB) and Petit Lac St. François (PLSF) were selected as study sites because of the frequent presence of cyanobacterial blooms [30,31]. MB is located in southern Quebec (Canada; 45 • 02 23 N; 73 • 04 41 W). It is shallow and eutrophic, with a surface area of 77.5 km 2 , an average depth of 2.8 m, and its groundwater and runoff sources are from predominantly agricultural lands via its tributary rivers [32,33]. MB's tributaries have high turbidity and excessive concentrations of nitrogen and phosphorus [32,34]. PLSF is located in south-western Quebec (Canada; 45 • 32 16 N; 72 • 02 11 W; max depth (Z) = 1.8 m). This lake is relatively small, shallow, with half of its watershed being occupied by agricultural land use, and there being high concentrations of phosphorus, resulting in frequent cyanobacterial blooms [30].

Description of the Mesocosm Experiments
The mesocosms were designed according to Wood's work in 2012 [35]. A mixture of raw lake water and cyanobacterial bloom scum was used to ensure that there were homogeneous initial conditions within the mesocosms. Sets of duplicate mesocosms (n = 6) were then filled with that mixed water of 20 L and were left unamended to serve as a control, or amended with two doses of the ferric sulfate coagulant: 20 mgFe/L and 35 mgFe/L. The samples were mixed immediately after the addition of the coagulant [12]. Seven mesocosm experiments were initiated with the appearance of cyanobacterial blooms (Tables S1 and S2).

Preparation and Application of Chemical Coagulant
Standard liquid grade Fe 2 (SO 4 ) 3 with 12.2% Fe content was purchased from Kemira Water Solution, Inc. (Varennes, QC, Canada). Two different doses of Fe 2 (SO 4 ) 3 coagulant (20 mgFe/L and 35 mgFe/L) were selected based on a previous study [12].

Sampling and Filtration Procedure
At the beginning (T0) (before adding Fe 2 (SO 4 ) 3 ) and after two days (T48) of the sampling period, the water in the mesocosms was collected in 1 L sterilized high-density polyethylene (HPDE) bottles. For phytoplankton enumeration, the samples were taken in 40 mL glass vial and preserved in Lugol's iodine solution. At the sampling site, a 50 mL sample was immediately filtered through a 0.22 µm filter (Millipore Sigma, Oakville, ON, Canada) in sterile conditions that were created with a burner stand. The filters were immediately stored with dry ice during their transfer to the lab where they were frozen at −80 • C for a metagenomic analysis. Samples of 120 mL were filtered on hydrophilic polypropylene; 0.45 µm filters (PALL, Mississauga, ON) were used for the toxins. The samples for the total organic carbon (TOC), total nitrogen (TN), and total phosphorus (TP) were aliquoted without being filtrated. Dissolved organic carbon (DOC) samples were filtered using a 0.45 µm filter membrane (PALL, Mississauga, ON, Canada). A 0.45 µm filter membrane (Millipore Sigma, Oakville, ON, Canada) was used for subsamples of dissolved nitrogen (DN), dissolved phosphorus (DP), and soluble reactive phosphorus (SRP). Genomic samples were taken in triplicate while nutrient samples were taken in duplicate. More detail are available in [12].
Samples for taxonomic cell counts were stored at room temperature in the dark. TN/TP/TOC/DOC samples were stored at 4 • C. Toxin and DN/DP/SRP samples were kept at −25 • C until the analysis was conducted.

Taxonomic Cell Counts
Taxonomic cell counts were performanced by using an inverted microscope with a Sedgwick-Rafter counting chamber at 40X magnification (Leica Microsystems GmbH, Wetzlar, Germany) according to [13,36,37].

Nutrient Analysis
Standard methods 353.2 and 350.1 were used for analyzing nitrogen [38]. TOC and DOC were measured based on method 415.1 [39]. Standard method numbers 365.3 and 365.1 [40] were used for phosphorus and phosphate analysis.

Toxin Analysis
The total microcystins samples were analyzed via a Lemieux oxidation step and an online solid phase extraction (SPE) and salting which were coupled with ultrahigh performance liquid chromatography tandem mass spectrometry (UHPLC-MS/MS). The accuracy was in the range of 93-100%, and it was deemed satisfactory if it was in this range, and the limit of detection was 0.5 ng/L. More details can be found in [12].

DNA Extraction, Metagenomics Preparation and Bioinformatics Analysis
The sample filters were kept in 5 mL Falcon ® tubes (VWR, Radnor, PA, USA) at −80 • C and were used for DNA extraction. Total nucleic acid was extracted using this filter by RNeasy ® PowerWater ® Kit (Qiagen Group, Germantown, MD, USA). Two hundred µL of nuclease-free water and 5 µL of TATAA Universal DNA spike II (TATAA Biocenter AB, Gothenburg, Sweden) were filtered to quantify extraction yields. Dithiothreitol (DTT) was injected into RNeasy ® PowerWater ® Kit solution PM1 buffer to prevent the formation of disulfide bonds protein residuals. Total nucleic acid was eluted with 60 µL nuclease-free water, and 30 µL was used for the DNA extraction. After resuspending it in 60 µL of nuclease-free water, each DNA sample was quantified using a Qubit ® v.2.0 fluorometer (Life Technologies, Burlington, ON, Canada). To perform the pyrosequencing (Roche 454 FLX instrumentation with Titanium chemistry), 30 µL of DNA was sent to the Génome Québec Innovation Centre for sequencing.
An Illumina NovaSeq 6000 S4 (Illumina Inc., San Diego, CA, USA) was used for sequencing DNA libraries. A homemade bioinformatic pipeline was applied for the analysis of paired-end raw reads of 150 base pairs (bp). The raw reads were trimmed to ensure their quality, which was firstly performed using SolexaQA v3.1.7.1 using the default parameters [41]. Next, the trimmed reads that were shorter than 75 nt were analysed. Artificial duplicates were removed using an in-house script based on the screening of identical reads of 20 bp. FragGeneScan-Plus v3.0 was used to predict gene fragments based on the trimmed high-quality reads [42]. Then, predicted fragments of protein were clustered at 90% similarity using cd-hit v4.8.1 [43]. More details are available in [6].

Statistical Analysis
Statistical analysis was performed using R statistical software (R Core Team, 2020). Taxanomic data were normalized by the centred log-ratio transformation using easyCODA (v.0.34.3) [44]. The alpha diversity metrics (Richness and Shannon) were analyzed using vegan package (v.2.6-2) [45], and the similarly matrices were calculated based on the Euclidean distance. The differences between the groups in alpha diversity were determined using Kruskal-Wallis test. Redundancy analysis (RDA) was performed to evaluate the impact of environmental conditions on cyanobacterial community in MB and PLSF, and this was conducted with the vegan package (https://cran.r-project.org/package=vegan, assessed on 10 June 2022).
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/toxins14100688/s1. Figure Figure S4: Principal components analysis (PCA) of the normalized relative abundance of cyanobacteria community composition in control, 20 mgFe/L, and 35 mgFe/L mesocosms with respect to intra-and extracellular total microcystins, MC-LR in Missisquoi Bay and Petit Lac St. François at T0 (left panel) and T48 (right panel). Figure S5: Principal components analysis (PCA) of the normalized relative abundance of cyanobacteria community composition in control, 20 mgFe/L and 35 mgFe/L mesocosms with respect to environmental conditions in Petit Lac St. François (PLSF). Table S1: Removal effectiveness (%) of taxonomic cell counts of individual cyanobacterial genera after 48 h (Mean ± Standard deviation) in Missisquoi Bay (MB) and Petit Lac St. François (PLSF). Table S2: Removal effectiveness (%) of taxonomic cell counts of individual cyanobacterial genera after 48 h on 8-10 August 2018, in Missisquoi Bay (Mean ± standard deviation), Table S3: Mean ± standard deviation value for environmental conditions of the sampled water in the control, mesocosms on 8-10 August 2018, in Missisquoi Bay (n = 2). Table S4: Environmental conditions of lake water samples in control mesocosms (n = 6) at T0 (mean ± standard deviation). Table S5: Environmental conditions of lake water samples in control mesocosms at T48 (mean ± standard deviation). Table S6: Environmental conditions of lake water samples in mesocosms with dose of 20 mgFe/L at T48 (mean ± standard deviation). Table  S7: Environmental conditions of lake water samples in mesocosms with dose of 35 mgFe/L at T48 (mean ± standard deviation). Table S8