Distilled water was produced using a Barnstead/Thermolyne still (model no. A1013B). Surface water was obtained from Lake Rockwell, the primary drinking water reservoir of the City of Akron, Ohio. It was collected from a spigot at the inlet pipe of the plant and used within 24 hours of collection.

One mg of MC-LR was purchased from Cayman Chemical. Upon arrival, MC-LR was dissolved in 500 µL of 75% methanol and 25% distilled water solution [25] . The vial was sealed and gently shaken to mix the toxin with methanol and distilled water. STX standard solution (saxitoxin dihydrochloride) of 500 μL vials was purchased from Eurofins Abraxis (CRM-00-STX). STX standard arrived as a solution of 20.5 µg/g in aqueous HCl at a pH of 2.63. Each toxin was transferred from original vials using a micropipette and placed in separate screw-on cap vials for storage at −20˚C (freezer) until use.

Pahokee peat humic acid was purchased from the International Humic Substances Society (St. Paul MN, USA). The International Humic Substances Society determined there was 56.37% carbon with a carbon distribution of 5% carbonyl, 20% carboxyl, 47% aromatic, and 4%.

WL PAC is a blend of wood and lignite while BC PAC was a bituminous coal. Based on the source, the PAC contained primary micropores (d < 0.8 nm), secondary micropores (0.8 - 2 nm), mesopores (2 - 50 nm) or macropores (d > 50 nm) [20] .

The pH_PZC (i.e., pH where the surface charge of the PAC is zero) was determined for each PAC using the drift method [26] . Jars were filled with 0.01 M NaCl and adjusted to a particular pH between 5 to 12. After adjusting the pH to the required initial value (initial), each jar was dosed with 3 grams of PAC sample and mixed for three days. The pH measured at the end of three days was called pH (final). The graph of pH (initial) vs. pH (final) was plotted to determine the pH constant value (pH initial = pH final). The pore size and pH_PZC for each carbon used in the study are given in Table 1 .

A six-chambered Phipps and Bird (model 7790-900B) gang mixer was used to evaluate cyanotoxin removal with PAC. The mixing speed for rapid mix (coagulation) and flocculation were scaled to that used in the City of Akron’s drinking water treatment plant (DWTP). The lab-scale velocity gradient was 134 revolutions/min for 20 sec and 17 revolutions/min for 45 min for rapid mix and flocculation, respectively [27] .

The experiments were conducted using distilled water and raw surface water with MC-LR at the initial concentration of 0.3, 1.6, and 20 µg/L and STX at 0.3 and 1.6 µg/L. Cyanotoxin concentrations were based on the 2019 Ohio Environmental Protection Agency thresholds [5] . The experimental pH of distilled water was 6 and 9 and of raw water was 6, 8, and 9. The experiments were also conducted on the simultaneous presence of MC-LR and STX. Each PAC type tested used a dose of 15 mg/L, which was the same as that used by the City of Akron DWTP.

Raw water was filtered through 0.45 μm filter paper made of mixed cellulose ester (Nitrate and Acetate) purchased from Fisher Scientific. Raw water had an average of 1.7 meq/L alkalinity, 2.48 mg/L total organic carbon and pH of 7.4. Both distilled and raw water were stored at room temperature between 19˚C - 22˚C before experiments. Potassium diphosphate dibasic and potassium phosphate monobasic (Fischer Scientific) were added to water at 0.02 M and were used as a buffer for pH adjustment. Once the pH was adjusted, each jar was filled with 1 L of the adjusted distilled or raw water.

Fifteen mg of PAC dose was added to a beaker, then made into a slurry using 10 mL from the gang mixer jar. Prior to adding the PAC back to the gang mixer jar, the required cyanotoxin(s) was added and mixed at 200 revolutions/min for approximately one minute to properly mix toxins. A set of distilled water experiments was amended with 13 mg/L Pahokee peat humic acid (HA) to assess the impact of organic matter on treatment efficiency. After mixing of toxins in mixer, PAC slurry from small beakers was added to each of the corresponding non-control jars, and no PAC was added to the controls. Once the coagulation and flocculation mixing were over, the PAC was allowed for five minutes.

A composite sample was collected from three locations in the jar using a sterile pipette. For STX, a 9 mL sample was mixed with 1 mL of 10× diluent for preservation. As MC-LR does not need a preservative, 10 cm³ of sample was collected from each jar and transferred to separate sample vials. Samples with expected concentrations above the analytical method’s detection limit were diluted with distilled water to yield a concentration within the detection interval of the equipment. Once sample collection was done, each sample bottle was gently shaken and placed in the refrigerator until analysis. The pH of all jars was checked after experiments to ensure the pH did not change during the experiment. Cyanotoxins were analyzed at an EPA certified using ELISA (Enzyme-linked immunosorbent assay). The upper detection limit of ELISA is 5 µg/L for MC-LR and 0.4 µg/L for STX.

Control vs treatments for all conditions were compared using a 1-sample t-test. ANOVA, Tukey, and 2-sample t-test comparisons were run to compare treatment efficiencies based on pH, concentration, toxin type, and type of water (natural and distilled). Any two conditions were compared using a 2-sample t-test. Multiple analyses of variance (ANOVA) and Tukey were used to compare three or more conditions. An equal variance was assumed. The mass of toxin removed (Δ = Control − Treatment) was used for comparing conditions. Statistical analysis was done using Minitab. Statistically significant and/or different data was represented in p < 0.05 and insignificant and/or similar in p > 0.05.

Control (no PAC) vs treatment (PAC) runs were performed with WL, and BC when 0.3 µg/L STX was the sole toxin in distilled and raw water at pH 6, 8, and 9 ( Figure 1 ). The removal was significant for all varying PAC types and water pH with STX (p < 0.05), except with WL at pH 6 (p > 0.05). For 0.3 µg/L STX in distilled water ( Figure 1(a) ) at pH 6, average removal was 15% and 21% with WL and BC, respectively (p > 0.05). At pH 9 distilled water ( Figure 1(a) ), 0.3 µg/L STX removed was 77%, and 49% with WL, and BC, respectively and the PAC performance trend was WL > BC (p < 0.05). For both PAC types, more 0.3 µg/L STX was removed when the water was at pH 9 than pH 6 (p < 0.05). STX has a neutral charge at pH 9 and its positive charge increases (+1 to +2) from pH 6 to 8 [28] . As the same charges repel each other, the adsorption efficiency of positively charged STX to positively charged PAC will decrease as pH lowers.

When 0.3 µg/L STX was added to raw water ( Figure 1(b) ) at pH 6, 8, and 9, WL removed 65%, 62%, and 29%, respectively (p < 0.05). Removal at pH 9 was lower (p < 0.05) than when pH was 6 and 8 (p > 0.05). Rorar et al. [27] reported similar removal of 31%, 30%, and 13% for pH 6, 7, and 9, respectively with a PAC made of a proprietary blend (PB) of wood, coconut, and coal for Akron source water. At lower pH, more of the NOM present in surface water adsorbed to PAC surface, thereby blocking the adsorption sites. This was more pronounced when STX was present at low concentration. At pH 8, WL removed 60% of STX compared to 46% with BC (p < 0.05). Although all PAC types may have the same effect on STX based on the pH_PZC and surface charge, WL performed better given the contribution of micropores from multiple sources. Shi et al. [28] reported the highest to lowest micropores for PAC sources as lignite coal > wood-based > bituminous coal and concluded removal efficiency was wood > bituminous coal > lignite coal. As lignite coal was the only PAC with the same charge as STX at experimental pH levels, it had the lowest efficiency showing the effect of charges on adsorption. Hence, pore size and charge are equally important in STX adsorption.

When STX concentration was increased to 1.6 µg/L in distilled water at pH 6 ( Figure 2(a) ), WL removed 40% (p < 0.05) and BC removed 26% (p < 0.05). Both treatment efficiencies were higher for 1.6 µg/L than 0.3 µg/L (p < 0.01). Pavagadhi et al. [10] also found that removal increased with an initial concentration of toxin. For 1.6 µg/L STX added to distilled water at pH 9, most of the data was above the detection limit of ELISA (samples could not be reanalyzed). For WL the control was >0.4 µg/L and treatments were 0.216 µg/L, 0.3 µg/L, and >0.4 µg/L while BC had both control and treatments concentrations above 0.4 µg/L. Though the analysis does not provide accurate results, it could still be concluded from the overall data that WL removed more STX than BC. This result also agrees with 0.3 µg/L STX at pH 9 and 1.6 µg/L at pH 6 in distilled water. However, the best-performing pH for 1.6 µg/L STX cannot be concluded due to insufficient data.

WL removed 16%, 32%, and 80% of 1.6 µg/L STX in raw water ( Figure 2(b) ) at pH 6 (p < 0.05), 8, (p > 0.05) and 9 (p < 0.05), respectively. When raw water was at pH 8, WL removed a slightly STX than 32% of 1.6 µg/L STX with BC (p < 0.05). STX removal was higher at pH 9 than at pH 6 (p < 0.05). Shi et al. [28] and Rorar et al. [27] concluded STX removal increased with an increase in pH. This could be due to the decrease in similar charges of STX and PAC at higher pH.

WL removed more of 1.6 µg/L STX than 0.3 µg/L STX (p < 0.05) when the raw water pH was raised to 9. In contrast, the removal was similar for both concentrations when water was at pH 6 and 8 (p > 0.05). STX removal with BC at pH 8 was more effective when concentration was higher (p < 0.05), which aligns with the results of Pavagadhi et al. [10] . Removal of 1.6 µg/L STX was similar for distilled and raw water at pH 6. NOM can block adsorption sites reducing the removal of targeted contaminants [12] [15] [29] [30] . This could be a combination of competition of pore size and attraction of opposite charges. STX removal exhibited a mixed response to the presence of NOM.

MC-LR removal at 0.3 µg/L, 1.6 µg/L, and 20 µg/L was statistically significant for treatments using WL and BC regardless of pH levels (p < 0.05). At pH 6 ( Figure 3(a) ), 59% of the 0.3 µg/L MC-LR was removed with WL which was higher than 42% with BC (p < 0.05). However, for 1.6 µg/L MC-LR, WL and BC yielded similar treatment efficiencies (p > 0.05). At pH 9 ( Figure 3(b) ), 72% of the 0.3 µg/L MC-LR was removed with WL and 21% with BC; whereas for 1.6 µg/L MC-LR, WL removed 86% and BC removed 33%. At pH 9 ( Figure 3(b) ), both 0.3 and 1.6 µg/L had higher MC-LR reduction with WL than BC (p < 0.05). Villars et al. [31] and Koshigoe et al. [32] also reported that a PAC wood-coal blend with higher mesopores would remove more MC-LR than bituminous coal PAC.

For 0.3 µg/L MC-LR, WL showed similar removal at both pH values (p > 0.05), whereas BC showed a higher reduction at pH 6 than at pH 9 (p < 0.05). Comparing pH values for 1.6 µg/LMC-LR, more MC-LR was removed with WL when in water with pH 6 than pH 9 (p < 0.05). BC yielded similar MC-LR removal at both pH levels (p > 0.05). WL showed a similar removal of 50% and 52% at pH 6 and 9 (p > 0.05), respectively, for 20 µg/L MC-LR ( Figure 3 ). This was attributed to the charge of MC-LR at pH 6 and 9 which may have a similar affinity towards PAC without any influence from organic matter and ions. The increase in MC-LR adsorption with decreasing pH was consistent with the results of Huang et al. (2007). For the removal of 20 µg/L MC-LR with BC, at pH 6 and 9, the concentration of control was above the detection limit of ELISA.

For all concentrations of MC-LR added to City of Akron source water ( Figure 4 ), removal was significant regardless of pH and PAC source (p < 0.05). 0.3 µg/L MC-LR removal was 78%, 64%, and 84% ( Figure 4(a) ) with WL at pH 6, 8, and 9, respectively (p < 0.05). MC-LR removal was higher at pH 9 than at pH 6 (p < 0.05). Liu et al. [33] found that more Suwanee River fulvic acid adsorbed onto clays at a pH of 5 than 7. If NOM adsorbed to a higher extent at the lower pH it could interfere with the subsequent adsorption of MC-LR. This was the opposite of what was observed in distilled water. The trend exhibited in distilled water was similar to Bajracharya et al. [34] reported that MC-LR removal increased with pH with wood-coal blend PAC and that NOM reduced adsorption. NOM can also compete for adsorption sites and block pores based on its size [12] , thus, reducing MC-LR adsorption at pH 6 than at pH 9. For 1.6 µg/L MC-LR, WL removed 75%, 62%, and 84% of MC-LR for pH 6, 8, and 9, respectively. WL showed no difference in removal based on pH (p > 0.05). This could be due to the concentration gradient effect for higher concentrations regardless of pH [10] . The removal of 60% of 20 µg/L MC-LR with WL at pH 8 ( Figure 4 ), which was substantially higher than 12% obtained by BC (p < 0.05). The same trend was observed regardless of concentration, where MC-LR removal was higher with WL than 21% of BC (p < 0.05) for both 0.3 and 1.6 µg/L concentration at pH 8 ( Figure 4(b) ). Villars et al. [31] and Liu et al. [33] also found that more PAC with higher mesopore content results in higher MC-LR adsorption. For both distilled and raw water, the mass of MC-LR adsorbed with WL increased with an increase in concentration (p < 0.05). However, BC showed similar removal at all concentrations (p > 0.05), except at pH 6 of 0.3 and 1.6 µg/L in distilled water (p < 0.05). This was attributed to the limited mesopores of BC.

WL performed better than BC for most experiments of individual toxins; therefore, it was the PAC source used when both MC-LR and STX were present. Figure 5(a) shows the removal of 1.6 µg/L MC-LR and 0.3 µg/L STX in the presence of NOM with WL in raw water (pH 6 and 8) and distilled water at pH 8. Both toxins had significant removal for all conditions tested (p < 0.05). At pH 8, distilled water removed a higher amount of MC-LR (0.82 µg) than STX (0.25 µg), (p < 0.05). This was the same trend for raw water, which removed higher MC-LR at 65% than STX at 49% (p < 0.05). For the same condition at pH 8 ( Figure 5(a) ), distilled water and raw water removed 16% (p < 0.05) and 4% (p > 0.05) NOM, respectively. MC-LR depicting a lower percent removal than STX in distilled water could be due to i) direct competition from NOM [35] , ii) pore blockage by NOM, iii) repulsion due to the same charges of MC-LR and NOM [12] [30] , and iv) positively charged STX in the absence of competition may show affinity towards negatively charged NOM and MC-LR, thus increasing the its overall removal when distilled water. In raw water at pH 6, MC-LR removal was higher (62%) than STX (31%, p < 0.05). Raw water with a pH 8 exhibited a similar treatment trend of more MC-LR removed than STX (p < 0.05). Shi et al. [28] and Bajracharya et al. [34] obtained the same trend of increased MC-LR and STX removal with an increase in pH. For toxin concentrations (MC-LR > STX), Rorar et al. [27] reported that with an increase in pH, MC-LR removal increased for treatment with PB, in contrast to STX which had a similar or increased reduction in raw water. NOM removal ( Figure 5(a) ) was higher in raw water at pH 6 (19%) than at pH 8 (p < 0.05). Increased opposite charges of NOM and PAC at pH 6 can increase the adsorption of NOM. This can either reduce the removal of toxins by competition with sites or enhance adsorption by increasing surface hydrophobicity [33] . When STX and MC-LR were present simultaneously, they both showed a reduced removal in raw water compared to distilled water spiked with HA. However, this was opposite to the results when the toxins were present alone. In the absence of other components from raw water, STX may partially neutralize the effect of repulsion of MC-LR to NOM, which may increase NOM’s adsorption resulting in the blocking of more pores for toxins.

Figure 5(b) illustrates the removal of cyanotoxins when present simultaneously at different concentrations. The removal of both toxins was significant under all conditions (p < 0.05). Forty-eight percent of 0.3 µg/L MC-LR and 66% of 0.3 µg/LSTX were removed with WL for raw water at pH 8. Statistically, the mass of MC-LR and STX removed were similar (p > 0.05). The presence of the same concentration gave an equal opportunity for both toxins to adsorb. A low concentration of MC-LR would potentially block fewer pores, leaving adsorption sites for STX. For this condition ( Figure 5(b) ), the NOM (26%) removed was highest among other concentrations of simultaneous toxins at pH 8 (p < 0.05). Lower concentrations of both toxins may have left some pores available where NOM could readily be adsorbed [9] . The other reason could be the low concentration of MC-LR, which may have shown less repulsion towards NOM, thus increasing its removal. STX may have shown an affinity towards NOM and MC-LR due to the opposite charge. Thus, NOM can both help and hinder toxin adsorption [33] [36] .

Removal was efficient for both toxins (61% STX and 72% MC-LR) and had a concentration of 1.6 µg/L in pH 8 raw water (p < 0.05). WL simultaneously removed 7% (p < 0.05) of NOM for this condition. A higher negative charge on MC-LR and its larger size may have led to more adsorption than STX. When less MC-LR was present, more STX was removed. The removal of 0.3 µg/L MC-LR and 1.6 µg/L STX with WL at pH 8 was 63% (p < 0.05) and 52% (p < 0.05), respectively. Although the treatment percentage appears less for STX, more mass was removed (0.83 µg STX vs. 0.19 µg MC-LR, p < 0.05). NOM removal was 4% for the same sample (p > 0.05).

Higher concentrations of toxin having higher removal even in the presence of another toxin was attributed to the concentration gradient effect. Among all the combinations of MC-LR and STX concentration, removal of 0.3 µg/L STX was higher when present alone or with a lower concentration of MC-LR (p < 0.05). However, for 1.6 µg/L STX, an increase in MC-LR concentration promoted the removal of STX (p < 0.05), which was also true for 0.3 µg/L STX in distilled water. For all cases, STX removal was effective in the presence of MC-LR. This was consistent with Rorar et al. [27] , which revealed that MC-LR did not inhibit the removal of STX. Similarly, MC-LR removal was effective in the presence of STX.

The extent of MC and STX removal that can be achieved will depend on the sorption mechanism(s) associated with PAC removal which include, but are not limited to, electrostatic interactions, surface complexation, π-π interactions, pore-filling, and diffusion [37] - [40] . The dominant mechanism will depend on the specific cyanotoxin present, water chemistry, and PAC source used. Some studies have stated that PAC made from coal and wood-based materials are the most effective for MC removal due to higher mesopore volumes [41] . For instance, Jaszczyszyn et al. [42] reported that the coal-based PAC used in their study had only half the mesopores as the wood-based PAC. Changes in electrostatic interactions have also been found to increase cyanotoxin removal [40] . Huang and Lenhart [43] reported that the electrostatic interactions between MC-LR’s functional groups and wood-based PAC increased MC removal. Park et al. [44] studied the adsorption of MC-LR with different sorbents and reported a two-stage process: fast surface adsorption and gradual adsorption as MC diffused into mesopores. Regardless of the adsorption mechanism, as mentioned earlier the presence of NOM can both hinder and help adsorption. Liu et al. [33] found that if fulvic acid-based NOM adsorbed to PAC surface, it decreased the sorption and removal of MC-LR while Bajracharya et al. [34] found algal-derived NOM to be more inhibitive than fulvic acid-NOM.