Diversity Assessment of Toxic Cyanobacterial Blooms during Oxidation

Fresh-water sources of drinking water are experiencing toxic cyanobacterial blooms more frequently. Chemical oxidation is a common approach to treat cyanobacteria and their toxins. This study systematically investigates the bacterial/cyanobacterial community following chemical oxidation (Cl2, KMnO4, O3, H2O2) using high throughput sequencing. Raw water results from high throughput sequencing show that Proteobacteria, Actinobacteria, Cyanobacteria and Bacteroidetes were the most abundant phyla. Dolichospermum, Synechococcus, Microcystis and Nostoc were the most dominant genera. In terms of species, Dolichospermum sp.90 and Microcystis aeruginosa were the most abundant species at the beginning and end of the sampling, respectively. A comparison between the results of high throughput sequencing and taxonomic cell counts highlighted the robustness of high throughput sequencing to thoroughly reveal a wide diversity of bacterial and cyanobacterial communities. Principal component analysis of the oxidation samples results showed a progressive shift in the composition of bacterial/cyanobacterial communities following soft-chlorination with increasing common exposure units (CTs) (0–3.8 mg·min/L). Close cyanobacterial community composition (Dolichospermum dominant genus) was observed following low chlorine and mid-KMnO4 (287.7 mg·min/L) exposure. Our results showed that some toxin producing species may persist after oxidation whether they were dominant species or not. Relative persistence of Dolichospermum sp.90 was observed following soft-chlorination (0.2–0.6 mg/L) and permanganate (5 mg/L) oxidation with increasing oxidant exposure. Pre-oxidation using H2O2 (10 mg/L and one day contact time) caused a clear decrease in the relative abundance of all the taxa and some species including the toxin producing taxa. These observations suggest selectivity of H2O2 to provide an efficient barrier against toxin producing cyanobacteria entering a water treatment plant.


Cyanobacterial Bloom Characteristics Throughout Sampling
The cyanobacteria bloom samples were collected on 5 days of the bloom period (1, 13, 15, 21 and 29 August 2018) from Missisquoi Bay (Lake Champlain) close to the water intake of the drinking water treatment plant. Cyanobacterial bloom characteristics are presented in the Table 1. DOC (Dissolved Organic Carbon) and pH did not demonstrate considerable variation throughout the sampling period. Total cell counts and biovolumes follow the same decreasing and increasing trend during the sampling period. For both parameters, the highest values were found at the beginning of the sampling period (1 August), followed by a significant drop on 13 August, where cyanobacterial cell counts decreased from 3.3 × 10 5 to 7.8 × 10 4 cells/mL and the biovolumes from 30.6 mm 3 /L to 4.6 mm 3 /L. The second drop was observed between 15 August and 21 August, from 1.4 × 10 5 to 6.8 × 10 4 cells/mL for cell count and from 9.4 to 0.3 mm 3 /L for the biovolume. According to algal cell abundance descriptors (biovolume and cell counts) during the bloom period, the main peaks of cyanobacterial bloom occurred on 1 August and to a lower extent on 15 August. The observed cell count exceeds the alert level of the 6.5 × 10 4 cells/mL for drinking water treatment plants [38], except for the 29 August with 5.4 × 10 4 cells/mL.

Variation of the Cyanobacterial Bloom Composition
The diversity and community variation of bacterial and cyanobacterial communities during the bloom sampling period were studied using comparative metagenomics reads levels of phylum, order and genus. The number of reads for taxonomic data was normalized by relative abundance (Figure 1).
Analysis of relative abundance of the bacterial community at the phylum level, based on high throughput sequencing data, showed that Proteobacteria, Actinobacteria, Cyanobacteria, Bacteroidetes, Firmicutes and Verrucomicrobia were the six most abundant phyla throughout the sampling period ( Figure 1a). As expected, Proteobacteria was by far the most abundant phylum at the beginning of the sampling period (1 August) as has been observed by Pei et al. (2017) [36]. The high relative abundance of Proteobacteria, especially in the beginning of sampling and Bacteroidetes at the end of the sampling period, may indicate contamination of the sampling point with human/animal-associated fecal markers [39][40][41]. For the rest of the sampling dates (13,15 and 21 August), Proteobacteria remained the predominant phylum, but at a lower extent than what was observed in the first and last days of sampling. The cyanobacteria phylum accounts for 5 to 10% of total relative abundance assigned to the phylum level in all samples. The cyanobacteria relative abundance started at 5% of the phyla, followed by an increase in the middle of the sampling (13 August and 15 August) to 10%. By the end of August, the cyanobacterial contribution in the whole bacterial community decreased to 5%. At the order level, the cyanobacterial community was dominated by members of the Chroococcales, Nostocales and Oscillatoriales during the cyanobacterial bloom period (from 1 August and 29 August) ( Figure 1b). The relative abundance of the Nostocales and Chroococcales varied between the different dates and even within the same day (15 August). On the other hand, Oscillatoriales relative abundance remained steady (Figure 1b).
Analysis at the genus level showed that within the Chroococcales order, the predominant genera were Microcystis and Synechococcus. Synechococcus was the dominant genus within Chroococcales until 21 August of the sampling period, followed by Microcystis, which became the dominant genus on 29 August (Figure 1c). The predominant genera in the Oscillatoriales and Nostocales orders were Oscillatoria and Dolichospermum (formerly known as Anabaena), respectively (Figure 1c). Depending on the sampling date, Synechococcus, Microcystis and Dolichospermum were the predominant genera during the bloom period. Our observations are consistent with short/long term investigations of cyanobacterial bloom dynamics using taxonomic cell count and metagenomics [14,21,[42][43][44]. Microcystis and Dolichospermum share a spatio-temporal niche during blooms and often are dominant within the cyanobacterial community, but have distinct environmental preferences; for example, they have different responses to nutrients [45][46][47]. During this study, the abundant genus shifted from Nostocales members, dominated by Dolichospermum for the first three weeks, to the Microcystis genus by the end of the sampling period (29 August). The shift to Microcystis dominance can probably be attributed to species-specific associations between Microcystis spp. and associated bacteria referred to as the epibiont phenomenon. Some species within Proteobacteria, Bacteroidetes and other phyla can shelter in the Microcystis mucilage to avoid being grazed [48]. These species play essential roles in enhancing the environmental adaptation of Microcystis within the cyanobacterial bloom, like maintaining redox balance and coping with oxidative stress [48]. Our results are in accordance with the previously reported results that Proteobacteria, Bacteriodetes tend to dominate Toxins 2020, 12, 728 5 of 21 in the Microcystis mucilage (blooms and culture-dependent studies) [49,50]. Moreover, allelopathy may influence the successional dominance of Microcystis and Dolichospermum within aquatic systems, whereby organisms produce bioactive compounds (allelochemicals) in the environment to positively or negatively influence the growth of neighboring species [51]. The effects of these allelochemicals on cyanobacterial community distribution within the environment are connected with nutrient availability and environmental conditions [52][53][54]. Competition experiments between toxic Microcystis and Dolichospermum strains, based on the lab coculture-dependent method, showed that Microcystis significantly inhibited the growth of Dolichospermum, whereas the effects of Dolichospermum on Microcystis were minimal [53,55]. Dolichospermum biovolume and biomass were sharply reduced after exposure to Microcystis strains [56]. Further investigation is required to explain the interaction and succession of Microcystis and Dolichospermum in cyanobacterial blooms.
An interesting insight was gained considering within-day changes in bacterial community composition (morning-AM and afternoon-PM) for 15 August (Figure 1). Based on the relative abundance analysis, a shift in bacterial composition at the phylum level was observed between the two samples on 15 August. In the afternoon, Proteobacteria and Actinobacteria increased, while Bacteroidetes and Cyanobacteria decreased as compared to their initial composition at the start of the sampling in the morning. At the order level, the most abundant order shifted from Nostocales in the morning to Chroococcolas in the afternoon. At the genus level, the changes were more evident for Synechococcus (increase) and Dolichospermum (decrease). These within-day changes in bacterial composition reflect the variation in stratification and mixing patterns of the cyanobacterial community in Missisquoi bay, as shown by Ndong et al. (2014) Ndong, Bird [57]. These findings highlight the importance of timing the sample collection by considering the stratification variation in the diel cycle (morning, noon, afternoon).

Impact of Oxidation on Cyanobacterial Diversity
The impact of oxidation on cyanobacterial diversity was assessed by using the impact of increased CT (the terminology used in water treatment is CT, which represents the product of the oxidant concentration (C) and contact time (T) to inactivate microorganisms) on two different samples, first on 1 August, dominated by Dolichospermum and the second on 29 August dominated by the Microcystis genus. The oxidants considered Cl 2 , KMnO 4 , O 3 , H 2 O 2 differ widely in their mode of action and their persistence [58]. As a result, CT values may vary over three orders of magnitude from 0.1 mg·min/L Cl 2 and a maximum of 7035 mg·min/L for H 2 O 2 .
Using non-normalized data is a source of bias in statistical analyses of the effects of oxidants on the bacterial/cyanobacterial communities. Thus, the data were normalized prior to analysis. Using the common exposure unit (CT) is not representative to compare the effects of different oxidants simultaneously. A normalized oxidant exposure (relative CT) is used to compare different oxidants among each other (Relative CT for each oxidant; max CT = 1 and min CT = 0, the exposure points in between were calculated accordingly); for example, max CT obtained for H 2 O 2 , on 1 August, is 7035 mg·min/L and is considered as relative CT = 1, the control condition is considered as relative CT = 0.
The doses (concentration) and contact time (T) were selected using the product CT, which is the foundation of disinfection and oxidation in drinking water. The choice of CT also considered prior evidence showing differences in species' resistance to the oxidant exposure (CT) [6,10,12,[14][15][16]18,42,59,60]. However, all previously reported observations were based on taxonomic cell counts and not on high throughput sequencing. The applied dosages reflect drinking water industry practices using regulated treatment processes.

Cyanobacterial Composition
To assess the cyanobacterial bloom composition following oxidation using KMnO 4 , Cl 2 and H 2 O 2 , samples were taken on 1 August ( Figure 2) and 29 August ( Figure S1). Dolichospermum and Microcystis were the dominant genera on 1 August and 29 August, respectively.
The doses (concentration) and contact time (T) were selected using the product CT, which is the foundation of disinfection and oxidation in drinking water. The choice of CT also considered prior evidence showing differences in species' resistance to the oxidant exposure (CT) [6,10,12,[14][15][16]18,42,59,60]. However, all previously reported observations were based on taxonomic cell counts and not on high throughput sequencing. The applied dosages reflect drinking water industry practices using regulated treatment processes.

Cyanobacterial Composition
To assess the cyanobacterial bloom composition following oxidation using KMnO4, Cl2 and H2O2, samples were taken on 1 August ( Figure 2) and 29 August ( Figure S1). Dolichospermum and Microcystis were the dominant genera on 1 August and 29 August, respectively.  Figure 2 shows the dissimilarity among groups of bacterial communities following oxidation at different exposures (relative CT) using Principal Component Analysis (PCA) on 1 August. Samples that appear more closely together within a PCA are assumed to be more similar in bacterial and cyanobacterial composition ( Figure 2). For the bacterial community, principal axis one and principal axis two for PCA represent 51.5 and 31.4% of the variation among the samples, respectively. Highrelative CT exerted clustering of two groups, the first one includes samples after Cl2 and KMnO4 oxidation and the second one encompasses KMnO4 and H2O2 oxidation. Chlorinated samples showed a clear progressive shift (as the relative CT increased) in the bacterial community.
On the other hand, the KMnO4 and H2O2 induce large shifts as relative CT increases. Bacterial and cyanobacterial composition similarity following the different oxidation on 29 August (Microcystis) is presented in Figure S1. Like the 1 August result, a large shift in the bacterial community composition following H2O2 oxidation was observed for 29 August. On the other hand, KMnO4 results showed similar bacterial composition on 29 August.
For the cyanobacterial community (1 August), principal axis one, and principal axis two for PCA represent 74.4 and 19.8% of the variation among the samples, respectively (Figure 2b). A clear trend in the cyanobacterial composition variation is observed following chlorination. Cyanobacterial composition after exposure to Cl2 (low relative CT < 0.05), KMnO4 (low and high relative CT), H2O2 (low relative CT = 0.08) oxidation clustered in the same group with the control condition. Highrelative (CT = 1) using H2O2 and (mid-relative = 0.53) KMnO4 are grouped in a distinct second cluster,  Figure 2 shows the dissimilarity among groups of bacterial communities following oxidation at different exposures (relative CT) using Principal Component Analysis (PCA) on 1 August. Samples that appear more closely together within a PCA are assumed to be more similar in bacterial and cyanobacterial composition ( Figure 2). For the bacterial community, principal axis one and principal axis two for PCA represent 51.5 and 31.4% of the variation among the samples, respectively. High-relative CT exerted clustering of two groups, the first one includes samples after Cl 2 and KMnO 4 oxidation and the second one encompasses KMnO 4 and H 2 O 2 oxidation. Chlorinated samples showed a clear progressive shift (as the relative CT increased) in the bacterial community.
On the other hand, the KMnO 4 and H 2 O 2 induce large shifts as relative CT increases. Bacterial and cyanobacterial composition similarity following the different oxidation on 29 August (Microcystis) is presented in Figure S1. Like the 1 August result, a large shift in the bacterial community composition following H 2 O 2 oxidation was observed for 29 August. On the other hand, KMnO 4 results showed similar bacterial composition on 29 August.
For the cyanobacterial community (1 August), principal axis one, and principal axis two for PCA represent 74.4 and 19.8% of the variation among the samples, respectively (Figure 2b). A clear trend in the cyanobacterial composition variation is observed following chlorination. Cyanobacterial composition after exposure to Cl 2 (low relative CT < 0.05), KMnO 4 (low and high relative CT), H 2 O 2 (low relative CT = 0.08) oxidation clustered in the same group with the control condition. High-relative (CT = 1) using H 2 O 2 and (mid-relative = 0.53) KMnO 4 are grouped in a distinct second cluster, revealing similar cyanobacterial assemblages. This is in contrast with cyanobacterial/bacterial community trends following chlorination that display a progressive shift. Differences between the observed trends for the three oxidants were expected because of different mechanisms of actions and kinetics, persistence and selectivity. Cell count based studies show progressive shifts following oxidation [12,14]. Moreover, selective cyanobacteria oxidation has been demonstrated in the lab [11,60], in the field [37,61], and in drinking water treatment plants [8,62].
Our results enlighten the different oxidation impacts on the bacterial community between the Cl 2 and H 2 O 2 oxidation. For 29 August, which was dominated by the Microcystis genus, KMnO 4 results revealed similar cyanobacterial composition. However, a large variation in cyanobacterial composition following H 2 O 2 was observed (e.g., the samples from the 1 August, Figure 3). community trends following chlorination that display a progressive shift. Differences between the observed trends for the three oxidants were expected because of different mechanisms of actions and kinetics, persistence and selectivity. Cell count based studies show progressive shifts following oxidation [12,14]. Moreover, selective cyanobacteria oxidation has been demonstrated in the lab [11,60], in the field [37,61], and in drinking water treatment plants [8,62].
Our results enlighten the different oxidation impacts on the bacterial community between the Cl2 and H2O2 oxidation. For 29 August, which was dominated by the Microcystis genus, KMnO4 results revealed similar cyanobacterial composition. However, a large variation in cyanobacterial composition following H2O2 was observed (e.g., the samples from the 1 August, Figure 3). Redundancy analysis (RDA) was used to explore the correlation between the relative abundance of nine dominant cyanobacterial genera (60% of the cyanobacterial community) observed in 1 August or 29 August samples and the different oxidant exposures ( Figure 3). The relationship between the CT and diversity indexes (Shannon and Chao1) varies between the two dates. On 1 August, both diversity indices increase with CT, while on 29 August only the Richness index (Chao1) increased with CT. In general, the diversity of the cyanobacterial community increased due to oxidation of the dominated genus, as discussed further in detail for each oxidant.
For 1 August (Dolichospermum dominant), RDA analysis establishes an inverse correlation between the Dolichospermum genus and a wide range of chlorine exposures with relative CT > 0.04. Chlorination appears to have a lesser impact on the Dolichospermum genus than the Microcystis, possibly because of its abundance. On the other hand, no such relationship is observed between Dolichospermum and KMnO4 exposure (relative CT = 1) and high H2O2 exposure (relative CT = 1), suggesting that these oxidants at these exposure levels had a negative impact on the Dolichospermum genus persistence within the community. Furthermore, for KMnO4 at relative CT > 0.53, Microcystis, Synechococcus, and Leptolyngbya are less impacted by oxidation. Figure 3b shows the effect of the different oxidants on the cyanobacterial community of 29 August, when Microcystis genus is dominant. An immediate shift from the control is observed for all oxidants and for all relative CT. Unlike 1 August, the Shannon diversity is no longer correlated with Dolichospermum. These differences could reflect the morphological differences between the dominant genera (Microcystis vs. Dolichospermum) as a unicellular aggregate that are less resistant to oxidation as compared to filaments [63]. Chlorine and KMnO4 exposures had a low impact on Microcystis and Dolichospermum. On the other hand, any H2O2 exposure causes a reduction in Dolichospermum and Microcystis, more Redundancy analysis (RDA) was used to explore the correlation between the relative abundance of nine dominant cyanobacterial genera (60% of the cyanobacterial community) observed in 1 August or 29 August samples and the different oxidant exposures (Figure 3). The relationship between the CT and diversity indexes (Shannon and Chao1) varies between the two dates. On 1 August, both diversity indices increase with CT, while on 29 August only the Richness index (Chao1) increased with CT. In general, the diversity of the cyanobacterial community increased due to oxidation of the dominated genus, as discussed further in detail for each oxidant.
For 1 August (Dolichospermum dominant), RDA analysis establishes an inverse correlation between the Dolichospermum genus and a wide range of chlorine exposures with relative CT > 0.04. Chlorination appears to have a lesser impact on the Dolichospermum genus than the Microcystis, possibly because of its abundance. On the other hand, no such relationship is observed between Dolichospermum and KMnO 4 exposure (relative CT = 1) and high H 2 O 2 exposure (relative CT = 1), suggesting that these oxidants at these exposure levels had a negative impact on the Dolichospermum genus persistence within the community. Furthermore, for KMnO 4 at relative CT > 0.53, Microcystis, Synechococcus, and Leptolyngbya are less impacted by oxidation. Figure 3b shows the effect of the different oxidants on the cyanobacterial community of 29 August, when Microcystis genus is dominant. An immediate shift from the control is observed for all oxidants and for all relative CT. Unlike 1 August, the Shannon diversity is no longer correlated with Dolichospermum. These differences could reflect the morphological differences between the dominant genera (Microcystis vs. Dolichospermum) as a unicellular aggregate that are less resistant to oxidation as compared to filaments [63]. Chlorine and KMnO 4 exposures had a low impact on Microcystis and Dolichospermum. On the other hand, any H 2 O 2 exposure causes a reduction in Dolichospermum and Microcystis, more so on 29 August. Removal of Microcystis using H 2 O 2 was shown by Lusty and Gobler. (2020) and is in accordance with our observations of diverging correlation between Chao1 and Microcystis [37]. Figure S2 shows the cyanobacterial composition following oxidation at genus level. Results show no significant variation in the relative abundance of Dolichospermum and Microcystis following Cl 2 , KMnO 4 oxidation. Thus, no relative persistence to oxidation was observed at the genus level. On the other hand, high H 2 O 2 exposure caused a decline in both Dolichospermum and Microcystis genus, which demonstrates the effect of high H 2 O 2 exposure on cyanobacterial removal.

Effect of Oxidation on Cyanobacterial Community Richness and Diversity
The diversity of the cyanobacterial composition following oxidation (1 August) at the species level is presented in Figure 4 (top 25 most abundant species). Dolichospermum sp.90 was the dominant species in control conditions, and its relative abundance increases following chlorination and decreases after KMnO 4 and H 2 O 2 . Although no trends in relative abundance were seen at the genus level ( Figure S2), a similar trend can be seen for the three species present Dolichospermum sp.90, Dolichospermum cylindrica and, Dolichospermum sp. PCC7108. Regardless of the oxidant Dolichospermum sp.90 remains the dominant species. Following the KMnO 4 and H 2 O 2 oxidation, Dolichospermum sp.90 was still the dominant species; its relative abundance declined compared to the control. The relative abundance of Microcystis aeruginosa is not impacted by chlorine but increases with KMnO 4 and H 2 O 2 exposure, confirming selective removal of Microcystis showed by Lusty and Gobler. (2020) [37].
in accordance with our observations of diverging correlation between Chao1 and Microcystis [37]. Figure S2 shows the cyanobacterial composition following oxidation at genus level. Results show no significant variation in the relative abundance of Dolichospermum and Microcystis following Cl2, KMnO4 oxidation. Thus, no relative persistence to oxidation was observed at the genus level. On the other hand, high H2O2 exposure caused a decline in both Dolichospermum and Microcystis genus, which demonstrates the effect of high H2O2 exposure on cyanobacterial removal.

Effect of Oxidation on Cyanobacterial Community Richness and Diversity
The diversity of the cyanobacterial composition following oxidation (1 August) at the species level is presented in Figure 4 (top 25 most abundant species). Dolichospermum sp.90 was the dominant species in control conditions, and its relative abundance increases following chlorination and decreases after KMnO4 and H2O2. Although no trends in relative abundance were seen at the genus level ( Figure S2), a similar trend can be seen for the three species present Dolichospermum sp.90, Dolichospermum cylindrica and, Dolichospermum sp. PCC7108. Regardless of the oxidant Dolichospermum sp.90 remains the dominant species. Following the KMnO4 and H2O2 oxidation, Dolichospermum sp.90 was still the dominant species; its relative abundance declined compared to the control. The relative abundance of Microcystis aeruginosa is not impacted by chlorine but increases with KMnO4 and H2O2 exposure, confirming selective removal of Microcystis showed by Lusty and Gobler. (2020) [37].  In the 29 August samples, Microcystis aeruginosa was the most abundant species ( Figure S3); its relative abundance increased after chlorination and KMnO 4 oxidation (as the chlorine exposure increased) and decreased after H 2 O 2 oxidation as compared to the control condition. Despite the very low relative abundance of Dolichospermum sp.90, Dolichospermum cylindrica and Dolichospermum sp. PCC7108, similar trends were observed on 1 August (when Dolichospermum was abundant).
The community richness and diversity indices for each treatment for 1 August and 29 August samples are illustrated in Figure 5 and Figure S4. Shannon and Chao1 show a small decline following Toxins 2020, 12, 728 9 of 21 chlorination in comparison with the control, while they increase slightly following KMnO 4 exposure. The total cell numbers following KMnO 4 decreased by up to 63% for high KMnO 4 exposure ( Figure S7). A remarkable decrease in richness is observed at high relative CT of H 2 O 2 , while the diversity (Shannon index) increases. The decline in the richness index could be the result of some less abundant species no longer identified. Indeed, total cell counts following H 2 O 2 relative CT = 1, decreased by more than 50% ( Figure S6). The same trends in the alpha diversity measurements are observed in the last week of the sampling (29 August), where the Microcysits were the most abundant genus ( Figure S4).
In the 29 August samples, Microcystis aeruginosa was the most abundant species ( Figure S3); its relative abundance increased after chlorination and KMnO4 oxidation (as the chlorine exposure increased) and decreased after H2O2 oxidation as compared to the control condition. Despite the very low relative abundance of Dolichospermum sp.90, Dolichospermum cylindrica and Dolichospermum sp. PCC7108, similar trends were observed on 1 August (when Dolichospermum was abundant).
The community richness and diversity indices for each treatment for 1 August and 29 August samples are illustrated in Figure 5 and Figure S4. Shannon and Chao1 show a small decline following chlorination in comparison with the control, while they increase slightly following KMnO4 exposure. The total cell numbers following KMnO4 decreased by up to 63% for high KMnO4 exposure ( Figure  S7). A remarkable decrease in richness is observed at high relative CT of H2O2, while the diversity (Shannon index) increases. The decline in the richness index could be the result of some less abundant species no longer identified. Indeed, total cell counts following H2O2 relative CT = 1, decreased by more than 50% ( Figure S6). The same trends in the alpha diversity measurements are observed in the last week of the sampling (29 August), where the Microcysits were the most abundant genus ( Figure  S4).

Cyanobacterial Community Assessment Following Oxidation; Longitudinal Study
The induced changes of cyanobacterial composition structure following the oxidation (Cl2, KMnO4, O3 and, H2O2) are assessed separately. The analysis was performed at the genus and species level.

Cyanobacterial Community Assessment Following Oxidation; Longitudinal Study
The induced changes of cyanobacterial composition structure following the oxidation (Cl 2 , KMnO 4 , O 3 and, H 2 O 2 ) are assessed separately. The analysis was performed at the genus and species level.

Chlorination (Cl 2 )
The chlorination experiments were conducted on 1 August (Figure 6a), and 29 August (Figure 6b). In the first chlorination trial (1 August), the abundant genera were Dolichospermum and Nostoc, representing approximatively 20% and 10% of the cyanobacterial community, respectively (Figure 6a). Figure 6 shows a limited effect of chlorination on the relative abundance of all cyanobacteria genera except Synechococcus. In addition, taxonomic cell counts show decrease (up to 30%) in total cyanobacteria cell counts for the trial on 1 August and limited variation for the 29 August trial (15% variation) ( Figure S5). In terms of the species, Dolichospermum sp.90 and Dolichospermum cylindrica were dominant for the 1 August trial. For the 29 August trial, Microcystis aeruginosa was the dominant species, followed by Dolichospermum sp.90 ( Figure S6). In all trials, as chlorination exposure increased, the relative abundance of the abundant species, either Dolichospermum sp.90 or Microcystis aeruginosa, increased slightly. Moreover, the relative abundance of Dolichospermum sp.90 increased as Microcystis aeruginosa did on 29 August. Chlorination results show that Dolichospermum species and Microcystis aeruginosa are relatively more persistent than the other species.
6a). Figure 6 shows a limited effect of chlorination on the relative abundance of all cyanobacteria genera except Synechococcus. In addition, taxonomic cell counts show decrease (up to 30%) in total cyanobacteria cell counts for the trial on 1 August and limited variation for the 29 August trial (15% variation) ( Figure S5). In terms of the species, Dolichospermum sp.90 and Dolichospermum cylindrica were dominant for the 1 August trial. For the 29 August trial, Microcystis aeruginosa was the dominant species, followed by Dolichospermum sp.90 ( Figure S6). In all trials, as chlorination exposure increased, the relative abundance of the abundant species, either Dolichospermum sp.90 or Microcystis aeruginosa, increased slightly. Moreover, the relative abundance of Dolichospermum sp.90 increased as Microcystis aeruginosa did on 29 August. Chlorination results show that Dolichospermum species and Microcystis aeruginosa are relatively more persistent than the other species.

Potassium Permanganate (KMnO4)
The relative abundance of the different genera following oxidation using permanganate shows limited variation for both KMnO4 tests (Figure 7). Total cell counts decreased (up to 57%) in the first trial (1 August) and remained stable in the second KMnO4 trial (less than 1% variation) (29 August). Total cell counts following the first KMnO4 trial showed a decrease at 278 mg·min/L ( Figure S7). The relative abundance of the Dolichospermum sp.90 increased slightly, whether it was the abundant species or not, suggesting the relative persistence of the Dolichospermum sp.90 during KMnO4 oxidation ( Figure S8).

Potassium Permanganate (KMnO 4 )
The relative abundance of the different genera following oxidation using permanganate shows limited variation for both KMnO 4 tests (Figure 7). Total cell counts decreased (up to 57%) in the first trial (1 August) and remained stable in the second KMnO 4 trial (less than 1% variation) (29 August). Total cell counts following the first KMnO 4 trial showed a decrease at 278 mg·min/L ( Figure S7). The relative abundance of the Dolichospermum sp.90 increased slightly, whether it was the abundant species or not, suggesting the relative persistence of the Dolichospermum sp.90 during KMnO 4 oxidation ( Figure S8).

Ozonation (O3)
The first and second ozonation trials were performed on 15 August (Dolichospermum most abundant genus) and 21 August (Synechococcus most abundant genus). Cyanobacterial community results following ozonation in the second trial (at genus level- Figure 8) showed a decline of the relative abundance of Synechococcus, followed by an increase for Microcystis, as compared to the control condition. However, no significant variation was observed for the relative abundance of the different genera in the first ozonation trials. Furthermore, total cyanobacteria cell counts revealed no significant change for both ozonation tests (up to 15%) ( Figure S9). In the control condition of the 15 August ozonation trial, Microcystis aeruginosa was the dominant species, followed by Dolichospermum sp.90 ( Figure S10). Dolichospermum sp.90 was not the dominant species, but it remained intact following ozonation. Although Cyanobioum gracil and Synechococcus sp. were the dominant species in the 21 August trial control, Microcystis aeruginosa became the dominant species following ozonation. At the low dosage applied, only secondary oxidation radical by-products are likely to react with cyanobacterial cells. Under these soft ozonation conditions, unicellular Cyanobium gracil and Synechococcus sp. cells were more susceptible than Microcystis aeruginosa.

Ozonation (O 3 )
The first and second ozonation trials were performed on 15 August (Dolichospermum most abundant genus) and 21 August (Synechococcus most abundant genus). Cyanobacterial community results following ozonation in the second trial (at genus level- Figure 8) showed a decline of the relative abundance of Synechococcus, followed by an increase for Microcystis, as compared to the control condition. However, no significant variation was observed for the relative abundance of the different genera in the first ozonation trials. Furthermore, total cyanobacteria cell counts revealed no significant change for both ozonation tests (up to 15%) ( Figure S9). In the control condition of the 15 August ozonation trial, Microcystis aeruginosa was the dominant species, followed by Dolichospermum sp.90 ( Figure S10). Dolichospermum sp.90 was not the dominant species, but it remained intact following ozonation. Although Cyanobioum gracil and Synechococcus sp. were the dominant species in the 21 August trial control, Microcystis aeruginosa became the dominant species following ozonation. At the low dosage applied, only secondary oxidation radical by-products are likely to react with cyanobacterial cells. Under these soft ozonation conditions, unicellular Cyanobium gracil and Synechococcus sp. cells were more susceptible than Microcystis aeruginosa. Toxins 2020, 12, x FOR PEER REVIEW 12 of 21

Hydrogen Peroxide (H2O2)
The Dolichospermum/Dolichospermum sp.90 and Microcystis/Microcystis aeruginosa were the most abundant genus/species on 1 August and 29 August H2O2 oxidation, respectively. The relative abundance of the Dolichospermum declined following H2O2 exposure of CT = 7035 mg·min/L. Microcystis relative abundance decreases by more than 10% at 8442 mg·min/L exposure of H2O2 on 29 August experiment (Figure 9). At the same H2O2 exposure, Dolichospermum decreased by 5%. Total cyanobacteria cell counts declined following H2O2 oxidation for both dates: 52% decrease on 1 August and 49% decrease on 29 August ( Figure S11). The relative abundance of the dominant species (Dolichospermum sp.90 1 August and Microcystis aeruginosa 29 August) decreases after H2O2 oxidation. The relative abundance of Microcystis aeruginosa in the H2O2 trial on 1 August increased after high H2O2 exposure. Microcystis aeruginosa was susceptible to H2O2 oxidation when abundant (as the relative abundance decreased), but it persists as an abundant species. Our results are in accordance with Lusty and Gobler. (2020) [37]. Figure S12 unveils higher susceptibility of Dolichospermum species to high H2O2 exposure as compared to Microcystis aeruginosa. Our results show the effect of high H2O2 exposure on the cyanobacteria species, which is in accordance with previous studies [61,64].

Hydrogen Peroxide (H 2 O 2 )
The Dolichospermum/Dolichospermum sp.90 and Microcystis/Microcystis aeruginosa were the most abundant genus/species on 1 August and 29 August H 2 O 2 oxidation, respectively. The relative abundance of the Dolichospermum declined following H 2 O 2 exposure of CT = 7035 mg·min/L. Microcystis relative abundance decreases by more than 10% at 8442 mg·min/L exposure of H 2 O 2 on 29 August experiment (Figure 9). At the same H 2 O 2 exposure, Dolichospermum decreased by 5%. Total cyanobacteria cell counts declined following H 2 O 2 oxidation for both dates: 52% decrease on 1 August and 49% decrease on 29 August ( Figure S11). The relative abundance of the dominant species (Dolichospermum sp.90 1 August and Microcystis aeruginosa 29 August) decreases after H 2 O 2 oxidation. The relative abundance of Microcystis aeruginosa in the H 2 O 2 trial on 1 August increased after high H 2 O 2 exposure. Microcystis aeruginosa was susceptible to H 2 O 2 oxidation when abundant (as the relative abundance decreased), but it persists as an abundant species. Our results are in accordance with Lusty and Gobler. (2020) [37]. Figure S12 unveils higher susceptibility of Dolichospermum species to high H 2 O 2 exposure as compared to Microcystis aeruginosa. Our results show the effect of high H 2 O 2 exposure on the cyanobacteria species, which is in accordance with previous studies [61,64]. Toxins 2020, 12, x FOR PEER REVIEW 13 of 21 Cyanobacterial composition analysis at species level exhibits relative persistence of different Dolichospermum species and Microcystis aeruginosa when it is abundant following soft-chlorination, soft ozonation and permanganate oxidation. This result is in accordance with the previous studies, which were mainly based on the cell count and the lab-cultured species [10,12,14,58,59]. Our results highlight the ability of hydrogen peroxide to decrease the relative abundance of different taxa, including the toxin-producing taxa of interest. H2O2 selectivity provides an efficient barrier against toxic cyanobacteria entering a drinking water treatment plant.

Comparison of the Microscopic Cell Count vs. High Throughput Sequencing Results
The taxonomic cell count results (genus) from the chlorination on 1 August and ozonation on 15 August are presented in Figure S13. The observed genera from microscopic cell counts do not completely match with high throughput sequencing results as Aphanocapsa and Aphanothece were only reported by taxonomic cell count. Misclassification at the genus level is less common than for species level; microscopic cell counts showed that Dolichospermum spiroids, Aphanocapsa delicatissma, Aphanotheche clathrate brevis and Aphanocapsa holistca were the most abundant species following chlorination and ozonation. However, high throughput sequencing did not identify them as an abundant species. The abundant species from high throughput sequencing were Microcystis aeruginosa, Dolichospermum Sp.90 and, Dolichospermum cylindrica ( Figures S6 and S10). Despite the potential of the microscopic cell count to provide absolute quantitative data such as cells/mL and biovolume, it has some drawbacks. Cyanobacterial composition analysis at species level exhibits relative persistence of different Dolichospermum species and Microcystis aeruginosa when it is abundant following soft-chlorination, soft ozonation and permanganate oxidation. This result is in accordance with the previous studies, which were mainly based on the cell count and the lab-cultured species [10,12,14,58,59]. Our results highlight the ability of hydrogen peroxide to decrease the relative abundance of different taxa, including the toxin-producing taxa of interest. H 2 O 2 selectivity provides an efficient barrier against toxic cyanobacteria entering a drinking water treatment plant.

Comparison of the Microscopic Cell Count vs. High Throughput Sequencing Results
The taxonomic cell count results (genus) from the chlorination on 1 August and ozonation on 15 August are presented in Figure S13. The observed genera from microscopic cell counts do not completely match with high throughput sequencing results as Aphanocapsa and Aphanothece were only reported by taxonomic cell count. Misclassification at the genus level is less common than for species level; microscopic cell counts showed that Dolichospermum spiroids, Aphanocapsa delicatissma, Aphanotheche clathrate brevis and Aphanocapsa holistca were the most abundant species following chlorination and ozonation. However, high throughput sequencing did not identify them as an abundant species. The abundant species from high throughput sequencing were Microcystis aeruginosa, Dolichospermum Sp.90 and, Dolichospermum cylindrica ( Figures S6 and S10). Despite the potential of the microscopic cell count to provide absolute quantitative data such as cells/mL and biovolume, it has some drawbacks.
The differences in community composition structure retrieved from the microscopic taxonomic cell counts and high throughput sequencing could be the result of the limitations inherent to these methods. For high throughput sequencing, these limitations include incomplete DNA sequencing libraries or using different libraries to identify genus and species. In the case of microscopic taxonomic cell counts, several sources of uncertainty and error have been identified [65,66]. The morphological similarity among cyanobacteria taxa may lead to overlooking or misidentifying cyanobacteria under the microscope, especially low abundant species [5,28]. Few genomes might be available for some taxa identified by taxonomic cell count (e.g., Aphanocapsa and Aphanothece), which may result in low relative abundance of these taxa using high throughput sequencing. Additionally, oxidation may further hinder the ability to identify cells because of its impact on the cell structure. Oxidation at higher dosages can cause significant morphological deformation of the cyanobacteria species, especially for H 2 O 2 and KMnO 4 in our study [63]. Flow cytometry conducted using the method described by Moradinejad et al. (2019) showed partial membrane damage (up to 80%) under the soft-oxidation conditions performed in this study (data not shown) [63].
This study is the first to provide insights into the impact of different pre-oxidants (doses and contact time) on the diversity and cyanobacterial community composition. The experimental design focused on CTs at which the shifts occurred, in order to provide actionable results for water utilities. Additional investigation would be beneficial to extend our observations to other water bodies. In addition, using assembly/binning method is suggested for future studies to provide a more accurate view of cyanobacterial genus/species.

Conclusions
A comparison of the microscopic vs. high throughput sequencing results demonstrates the ability and robustness of high throughput sequencing to fully reveal a wide diversity of cyanobacterial communities in response to oxidant stress.
Results from longitudinal sampling over a bloom period of 4 weeks by high throughput sequencing highlight quick composition and abundance shifts in the cyanobacterial communities that occur within a day. High throughput sequencing revealed clearer shifts during a bloom from an initial dominance of Dolichospermum/Dolichospermum Sp.90 toward a late summer dominance by Microcystis/ Microcystis aeruginosa.
Overall, pre-oxidation caused deeper changes in the diversity of whole bacterial communities, especially proteobacteria, than was observed for the cyanobacterial community. Such changes should be considered when assessing the impact of using oxidants for onsite source control.
Depending on the oxidants used, alpha diversity indexes (Shannon and Chao1) showed that oxidation resulted in different structural composition shifts within bacterial and cyanobacterial communities.
Soft-chlorination using dosages 0.2 and 0.6 mg/L that were typically used for pre-oxidation caused a progressive shift in the bacterial and cyanobacterial communities with increasing CTs (0-3.8 mg·min/L). The results using KMnO 4 (5 mg/L) and H 2 O 2 (10 mg/L) induced larger and distinct shifts in the structural composition of bacterial and cyanobacterial communities.
Regardless of the significant differences in community distribution shifts caused by different oxidants, some toxin producing species could persist after oxidation whether they were dominant species or not.
Soft-chlorination results revealed that Dolichospermum sp.90 was relatively persistent with increasing CTs whether it was the dominant species or not. Relative persistence of Dolichospermum sp.90 was also observed within KMnO 4 oxidation with increasing KMnO 4 exposure, regardless of the dominant species.
Only pre-oxidation with H 2 O 2 (10 mg/L) caused a clear decrease in the relative abundance of all taxa and some species including the toxin-producing taxa of interest. As such, H 2 O 2 would provide an effective first barrier against toxin producing cyanobacteria entering the drinking water treatment plant.
Selection of the most effective pre-oxidant for drinking water purposes should be made considering its impact on the cyanobacterial community structure/diversity, prevention of disinfection by-product formation and other drivers such as color and the presence of taste odor compounds.

Sampling Site Description
The oxidation tests were conducted using natural bloom samples. Cyanobacterial bloom samples were collected and transferred to the laboratory from the Bedford water treatment plant intake (Missisquoi Bay-Lake Champlain) in southern Quebec, Canada (45 • 02 22.0" N 73 • 04 40.5" W). Sampling was performed during the bloom season from 1 August to 29 August 2018. Several cyanobacterial blooms have been documented for the Lake Champlain in previous years [14,21,42,43].

Chemicals and Reagents
A free chlorine stock of 2000 mg/L was freshly prepared from sodium hypochlorite (5.25%) on the day of the experiment. Free chlorine residual was measured using an N,N-diethyl-p-phenylenediamine (DPD) colorimetric method based on Standard Methods (SM) 4500-Cl G [67]. Samples were dosed at room temperature (22 • C). A stock of sodium thiosulfate 3000 mg/L was used to quench the chlorine samples at a dose of 1.1 mg/L per 1 mg/L chlorine.
A potassium permanganate (KMnO 4 ) stock solution of 5000 mg/L was prepared by dissolving KMnO 4 crystals into the ultrapure water. A DPD colorimetric method ratio of 0.891 KMnO 4 /Cl 2 SM 4500-Cl G was used to determine the KMnO 4 residual [67]. An amount of 1.2 mg/L sodium thiosulfate per 1 mg/L KMnO 4 was used to quench further oxidation with KMnO 4 .
A bench-scale ozone generator (details in [14]) was used to prepare an ozone stock solution 50-60 mg/L. SM 4500-O 3 was used to determine the stock and ozone residual concentration [14]. An ozone stock solution (50-60 mg/L) was prepared with gaseous ozone using a bench-scale ozone generator (additional details in [14]). Ozone stock concentration and residual ozone in water samples were measured using SM 4500-O 3 [14]. An amount of 1.6 mg/L sodium thiosulfate per 1 mg/L ozone was used to quench the ozonation samples.
Stabilized hydrogen peroxide (30%, Sigma Aldrich, MO, USA) was used to prepare the Hydrogen peroxide (H 2 O 2 ) stock (10 g/L). A colorimetric test kit (Chemetrics K-5510, Midlands, VA, USA) was performed to measure the hydrogen peroxide residual. Moreover, 1.2mg/L sodium thiosulfate per 1 mg/L H 2 O 2 was used to quench further oxidation with H 2 O 2 .
The selected oxidant doses and contact times are presented in Table 2. The doses (concentration) and contact time have been selected based on the common pre-oxidation doses used in drinking water treatment intake.
Oxidant exposures, concentration vs. contact time, (CT) were calculated using Equation (1): where k decay (min −1 ) is the first-order decay rate, t (min) is the exposure time, and C 0 (mg/L) is the initial concentration of oxidant at time zero. The selection of oxidant dose was based on the commonly used pre-oxidation doses in the operation of the drinking water treatment plant.

DNA Extraction, Metagenomic Preparation, Sequencing and Bioinformatic Analysis
Total nucleic acid was extracted from the frozen filters (after filtration, samples were quickly transferred to −80 • C before DNA extraction) using an RNeasy PowerWater Kit (Qiagen Group, Germantwon, MD, USA) with modification. Before the extraction, 200 µL of nuclease-free water and 5 µL of TATAA Universal DNA spike II (TATAA Biocenter AB) were added to the filters to evaluate DNA extraction yields using RT-qPCR. RNeasy PowerWater Isolation kit solution PM1 was used to lyse the cells along with Dithiothreitol (DTT), which prevents disulfide bonds forming residues of proteins. A total volume of 60 µL nuclease-free water provided with the kit was used to elute the total nucleic acid, of which 30 µL of DNA (with minimum 1 ng of DNA) extracts were stored at −20 • C. DNA was subsequently purified with the Zymo Kit (Zymo Research, Irvine, CA, USA) according to the manufacturer's instructions. Each DNA sample was resuspended in 60 µL of nuclease-free water and quantified with a Qubit v.2.0 fluorometer (Life Technologies, Burlington, ON, Canada). A volume of 30 µL DNA was sent for pyrosequencing (Roche 454 FLX instrumentation with Titanium chemistry) to the Genome Quebec.
An Illumina NovaSeq 6000 platform using S4 flow cells was applied to sequence DNA libraries. A home-made bioinformatic pipeline was used for further analysis of Paired-end raw reads of 150 base pairs (bp) as follows. First, raw reads trimming quality was performed using the SolexaQA v3.1.7.1, default parameters [68]. Further analyses were carried out on the trimmed reads shorter than 75 nt. An in-house script, based on the screening of identical leading 20 bp, was used to remove artificial duplicates. Gene fragments were predicted using FragGeneScan-Plus v3.0 based on the trimmed high-quality reads [69]. Then, predicted fragments of protein were clustered at 90% similarity level using cd-hit v4.8.1 [70]. A diamond engine was used for similarity search on the M5nr database based on a representative of each cluster (https://github.com/MG-RAST/myM5NR). Best hits (minimal e-value of 1 × 10 −5 ) combined with the last common ancestor approach were used to assess the taxonomic affiliation of protein fragments. It should be mentioned that the annotation process uses a read-mapping process of small gene fragments of encoding proteins on a large database of proteins. One gene fragments encoding protein in the database could match with multiple species or strains.

DOC and Cyanobacteria Cell Count
Pre-rinsed 0.45 µm membrane filters (Supor 45 µm, 47 m, PES PALL, Port Washington, NY, USA) and carbon-free glass vials were used for Dissolved Organic Carbon (DOC) samples. DOC measurements were performed via a 5310 total organic carbon analyzer (Sievers Analytical Instruments, Boulder, CO, USA). An inverted microscope with 20× magnification was used to perform cell count samples preserved with Lugol's Iodine [71,72].

Statistical Analysis
All analyses were performed using a custom bioinformatics pipeline implemented in R (v.3.6,2, RStudio, Inc., Boston, MA, USA), phyloseq (V.1.28.0) to visualize the community composition at phylum (all bacteria reads), order, and genus (cyanobacteria reads) [73]. The twenty-five most abundant cyanobacteria species were visualized using pheatmap (v.1.0.12) [74]. Then, the alpha diversity metrics were estimated using phyloseq's estimate richness function (Shannon and Chao1). Taxonomic data were normalized by the centred log-ratio transformation using easy CODA (v.0.31.1) [75]. The beta-diversity was analyzed using the vegan package (v.2.5-6), where the similarity matrices were calculated based on the Euclidean distance [76]. The homogeneity of variances of normalized data related to each oxidant was analyzed before building the model. A Redundancy Analysis (RDA) constrained ordination to each oxidant applied to the cyanobacteria genus and tested by the permutation test (>95% significance).

Supplementary Materials:
The following are available online at http://www.mdpi.com/2072-6651/12/11/728/s1. Figure S1: Principal components analysis (PCA) of the normalized relative abundance of comparative metagenomics reads in 29 August 2018 sample. Data are plotted following the genus-level classification (a) PCA analysis of bacterial community following oxidation using different CT (b) PCA of the cyanobacterial community following oxidation using different CT. Figure