Reactivity of Cyanobacteria Metabolites with Ozone: Multicompound Competition Kinetics

Cyanobacterial blooms occur at increasing frequency and intensity, notably in freshwater. This leads to the introduction of complex mixtures of their products, i.e., cyano-metabolites, to drinking water treatment plants. To assess the fate of cyano-metabolite mixtures during ozonation, a novel multicompound ozone (O3) competition kinetics method was developed. Sixteen competitors with known second-order rate constants for their reaction with O3 ranging between 1 and 108 M–1 s–1 were applied to cover a wide range of the O3 reactivity. The apparent second-order rate constants (kapp,O3) at pH 7 were simultaneously determined for 31 cyano-metabolites. kapp,O3 for olefin- and phenol-containing cyano-metabolites were consistent with their expected reactivity (0.4–1.7 × 106 M–1 s–1) while kapp,O3 for tryptophan- and thioether-containing cyano-metabolites were significantly higher than expected (3.4–7.3 × 107 M–1 s–1). Cyano-metabolites containing these moieties are predicted to be well abated during ozonation. For cyano-metabolites containing heterocycles, kapp,O3 varied from <102 to 5.0 × 103 M–1 s–1, giving first insights into the O3 reactivity of this class of compounds. Due to lower O3 reactivities, heterocycle- and aliphatic amine-containing cyano-metabolites may be only partially degraded by a direct O3 reaction near circumneutral pH. Hydroxyl radicals, which are formed during ozonation, may be more important for their abatement. This novel multicompound kinetic method allows a high-throughput screening of ozonation kinetics.


■ INTRODUCTION
Cyanobacteria are among the most ubiquitous organisms on the globe, comprising almost 2000 identified species living in freshwater, terrestrial, or marine environments. 1,2Some cyanobacteria can form dense blooms that can cause significant deterioration of the water quality, notably by increasing turbidity, depleting oxygen by decomposition of biomass after blooms subside, and releasing toxins. 3These blooms can occur in freshwater reservoirs that are resources for drinking water treatment plants and have been increasing in intensity and frequency worldwide in the last decades. 3,4Toxins can be part of a complex mixture of metabolites produced by cyanobacteria (i.e., cyano-metabolites) and can enter water treatment plants. 5If no appropriate treatment is in place, toxins may even end up in the finished drinking water, requiring temporary safety warnings to the population. 6,7Recognizing the human health concerns, the World Health Organization proposes chronic, lifetime, and acute short-term drinking water guideline values for cyanobacterial toxins. 8Therefore, water suppliers need to account for the presence of cyano-metabolites in raw water, anticipate the probable intensification of blooming events, and develop mitigation strategies at the source and/or appropriate treatment.
Ozone (O 3 ) is a well-established and widely applied oxidant for disinfection, micropollutant degradation, and mitigation of disinfection byproducts, which has many advantages over other chemical oxidants. 9,10,15It can be used to efficiently degrade known potent toxins such as microcystins. 11,12However, a large number and diversity of other cyano-metabolites exist, with 2425 currently known compounds listed in a shared database for cyano-metabolites (CyanoMetDB, version 02, 2023), for which ozonation kinetics have not been studied to date. 13,14Therefore, further investigations of the reactivity of O 3 with cyano-metabolites are needed to better assess its efficiency as a barrier against potential bloom-related toxins.Many cyano-metabolites contain functional groups such as olefins or phenols that have known reactivity with O 3 , 9,15 which can help predicting their abatement during ozonation.However, a given moiety can have significant variability in reactivity depending on the pH and substituents. 9One case in 20 min at 90 mbar, and 80 min at 20 mbar) and redissolved in ultrapure water.No loss of identified cyano-metabolites was observed during the evaporation to dryness (Figure S1b).This approach guaranteed the absence of MeOH during ozonation, which is a promotor for • OH formation during ozonation. 9yano-metabolite Analysis and Identification.Cyanometabolites were analyzed by HPLC (Dionex Ultimate 3000 RS pump, Thermo Fischer Scientific) with a Kinetex C18 column (2.6 μm, 2.1 × 100 mm, with SecurityGuard ULTRA precolumn, Phenomenex), coupled to a high-resolution tandem mass spectrometer (HRMS/MS, Exploris, Thermo-Fisher Scientific). 18,19Details for the HPLC and HRMS/MS methods are provided in Text S2.Elution was carried out using a gradient with MeOH and ultrapure water both acidified with 0.1% formic acid.HRMS/MS used electrospray ionization (ESI) with both positive and negative ionization modes at 3.5 and 2.5 kV, respectively.
Data evaluation and peak area extraction was performed with Skyline 22.2 (MacCoss Lab).The ion chromatograms were screened with a suspect list obtained from CyanoMetDB (Version 02, 2023), 14 searching for the , and [M + 2H] 2+ ions. 13,14Suspects were considered only if they fulfilled the following criteria: (i) exact mass < 4 ppm, (ii) isotope dot product > 0.9, and (iii) peak area ≥ 10 7 .Then, suspects were further considered if their retention times matched their expected polarities and if good MS 2 spectra were obtained.The final list of cyano-metabolites is provided in Table S4.Since no spectral libraries exist for most cyano-metabolites, in-silico fragmentation predictions were used to facilitate manual compound annotation of MS 2 spectra (MetFrag Web and Sirius 4.4). 22,23Suspected structures were manually evaluated, prioritizing fragmentation around the peptide bonds (details are given in Text S3).Only compounds identified with a confidence level of at least 3 were used (see Table S4). 24The MS 2 annotations and, when possible, head-to-tail MS 2 comparisons with standards or bioreagents are available in a separate data spreadsheet (SI2).
Ozonation Experiments.Kinetic studies were performed at room temperature (22 ± 1 °C), using cyano-metabolite mixtures extracted from 0.8 to 1.2 g biomass L −1 of Planktothrix or Microcystis, in the presence of competitors (at 0.4 μM each), phosphate buffer (4 mM, pH 7 or 8), and tert-butanol (80 mM).The dissolved organic carbon concentrations of the samples (excluding tert-butanol) were about 12−16 mgC L −1 .The cyano-metabolite and competitor mixture was split in 1 mL aliquots to which 1 mL of prediluted O 3 was added under vigorous stirring.O 3 was prediluted in pure water from the O 3 stock to prevent the spiking of samples with small volumes of highly concentrated O 3 , leading to strong concentration gradients and potentially undesired side reactions.The O 3 doses were in the range of 0.1−90 μM after dilution (0.01−7.9 mgO 3 /g biomass ).Experiments were performed in triplicate on three separate days.
Determination of Second-Order Rate Constants.Competition kinetics were applied to determine the k app,O3 of cyano-metabolites (k app,O3,cyanomet ): Instead of concentrations, peak areas can directly be used (as long as the response is linear) because only the relative abatement needs to be known.k app,O3,cyanomet can be derived

Environmental Science & Technology
from the linear regression slope of the ln of the cyanometabolite relative residual peak area as a function of the ln of the competitor relative residual peak areas.k app,O3,comp values are provided in Table 1 for pH 7. For competitors undergoing acid−base speciation, k app,O3,comp is pH-dependent and was calculated at a given pH by eq 2: with x being the number of species, k x the second-order rate constant for the reaction of a given species with O 3 (Table S3), and α x the molar fraction of the given species at a given pH.
Examples of correlation plots between relative abatements of cyano-metabolites and competitors are shown in Figure S4.Only data points with a relative abatement of cyano-metabolites and competitors between 10% and 90% were considered.There were two main reasons for this approach: (i) a wide range of O 3 doses was applied and each compound only reacted at a specific range of O 3 doses.For compounds with intermediate or low reactivity, many data points at negligible abatement would overly impact the linear regression, which was avoided by setting a minimum of 10% abatement.(ii) The 90% maximum abatement limit was set to avoid loss of linearity, prevent high leverage data points and eventual interference due to carryover between sample injections on the HPLC-HRMS/MS.k app,O3 for a given cyano-metabolite/competitor pair was only calculated if (i) there were 10 or more data points, (ii) the coefficient of determination (R 2 ) was higher than 0.9, (iii) the intercept was negligible (intercept < 10 × slope), and (iv) the    S4, Tables S11−S14, and Figure S5.

Environmental Science & Technology
slope was between 0.1 and 10 (i.e., if the k app,O3 for the cyanometabolite and the k app,O3 for the competitor were within 1 order of magnitude).The statistical parameters of each successful linear regression are provided in Tables S5−S10.For a given cyano-metabolite/competitor pair, the standard error (SE) of the cyano-metabolite k app (SE cyanomet/comp ) was the result of the propagation of the standard error on the slope (SE slope ) and on the k app,O3 of the competitor (SE comp ): The uncertainty in k app,O3,comp given in the literature was used for SE comp .If no error was provided in the literature for the k app,O3 of the competitor, a 20% relative error was set (most k app,O3,comp values have relative errors of 20% or less).For competitors with acid−base speciation, the error of k app,O3 in the literature was replaced by the error induced from ±0.05 pH variation, if the latter was greater.This was the case for roxithromycin, tramadol, and triclosan at pH 7.
For a given cyano-metabolite, several k app,O3 could be obtained from several competitors (e.g., anabaenopeptin A correlated with trimethoprim, carbamazepine, tylosin, sulfamethoxazole and dibromomethylparaben, each giving a k app,O3 ).In addition, the same cyano-metabolite could be monitored at different ionizations (given in Table S4) in HRMS/MS (e.g., the [M + H] + and [M − H] − ions of anabaenopeptin A were simultaneously monitored).Different ionizations did not significantly impact the k app,O3 (<20% difference, k app,O3 for individual ionizations are not shown).All the k app,O3 determined for a given cyano-metabolite were averaged and the standard error was either the standard deviation of all the k app,O3 or the highest competitor-specific k app,O3 standard error, whichever was the highest.

■ RESULTS AND DISCUSSION
Identification of Cyano-metabolites in Microcystis and Planktothrix Extracts.A total of 31 cyano-metabolites with sufficient signal intensity for kinetic studies were identified from the selected strains: 1 aeruginosin, 5 anabaenopeptins, 7 cyanopeptolins, 6 cyclamides, 10 microcystins, and 2 unclassified cyano-metabolites. 32Figure 1 shows the structures of representative cyano-metabolites for each class.Highlighted moieties represent the parts of the molecule that can change for other identified variants of the same class.The full cyano-metabolite list with their structures is provided in the SI (Table S4, Tables S11−S14, and Figure S5).
Aeruginosin-group-608 is the only noncyclic cyano-metabolite identified in this study.It refers to one of the three known stereoisomers of the shown structure (Figure 1).Aeruginosins are tetrapeptides notably characterized by a lactic acid derivative that contains a phenol, which is expected to be a major reactive site for O 3 . 33yclamides, represented by aerucyclamide C in Figure 1, are cyclic hexapeptides characterized by heterocyclic groups such as thiazole, oxazole, thiazoline, and oxazoline that have unknown reactivities toward O 3 .In our study, oxazole was determined to be the main reactive site in aerucyclamide C (see explanation in "Determination of Second-Order Rate Constants for the Reactions of Cyano-metabolites with Ozone").One cyclamide, aerucyclamide D, contains a methionine, which is expected to be a major reactive site (Table S13). 34nabaenopeptins are characterized by a cyclic pentapeptide containing a lysine.The lysine's α-amine branches out to form a urea bond with the N-terminal of a sixth amino acid outside of the cycle (an arginine in the case of anabaenopeptin B, Figure 1).All the anabaenopeptins identified in this study contain a homotyrosine and eventually a tyrosine, which are expected to be the major reactive sites for O 3 . 34

Environmental Science & Technology
Cyanopeptolins are characterized by a cyclic hexadepsipeptide containing a 3-amino-6-methoxy-2-piperidone and a threonine that forms an ester bond (Figure 1).In addition, the N-terminus of the threonine branches out to two supplementary amino acids.The main O 3 attack site in cyanopeptolin D is expected to be the tertiary amine on the side chain of the lysine derivative.In other cyanopeptolins, the lysine derivative is replaced by other amino acids such as a normal lysine or a tyrosine, which are also expected to be major reactive sites (Table S12). 34,35icrocystins (abbreviated as MC) are cyclic heptapeptides with a characteristic Adda moiety (3-amino-9-methoxy-2,6,8trimethyl-10-phenyl-deca-4,6-dienoic acid) (Figure 1).Adda contains conjugated olefins that are a major reactive site for O 3 . 12,31One microcystin, [Mdha-GSH7]MC-LR, contains an additional thioether that is expected to be more reactive than the conjugated olefins (Figure S5). 34wo additional cyclic cyano-metabolites, identified in Planktothrix, are planktocyclin and piricyclamide ILGEGEGW-NYNP+prenyl (Figure S5).They contain a methionine and tryptophan, respectively, which are expected to be the main reactive sites. 34alidation of Competitors.First, the abatement of competitors and their previously published k app,O3 values were evaluated in the presence of cyano-metabolite mixtures extracted from the two selected cyanobacteria strains.The abatement of the selected competitors as a function of the O 3 dose in the presence of the cyano-metabolite mixture from Microcystis at pH 7 is shown in Figure 2a.The abatement of the competitors was generally consistent with their reactivity: a higher k app,O3 led to a higher abatement for a specific ozone dose.However, a few exceptions were observed: dibromomethylparaben (DMP, k app,O3 = 8.4 × 10 7 M −1 s −1 ) should be more reactive than triclosan (TRI) and vancomycin (VM, not shown in Table 1) (3.8 × 10 7 and 1.2 × 10 6 M −1 s −1 , respectively). 17,29,30However, Figure 2a shows that DMP (open diamonds) required higher or similar O 3 doses than TRI (reverse blue triangles) or VM (reverse red triangles) for an equivalent abatement.Likewise, carbofuran (CBF, blue circles) required higher O 3 doses than benzafibrate (BZF, red circles) to reach an equivalent abatement, while their reported k app,O3 values are comparable (6.2 × 10 2 and 5.9 × 10 2 M −1 s −1 , respectively). 25,26o evaluate whether the observed discrepancies were significant, competitors were systematically compared to each other, using eq 1.The abatement of a given competitor was correlated with the other competitors with similar reactivities, i.e., leading to regression slopes between 0.1 and 10, from which apparent second-order rate constants were calculated (k measured ). Figure 3a shows for each competitor the ratios between k measured in the Microcystis extract at pH 7 and the previously published apparent second-order rate constant (k literature , shown in Table 1).For example, for penicillin G (PG), three symbols are shown, which represent three k measured /k literature ratios using tramadol (TRA, open circle), BZF (filled red circle), and ciprofloxacin (CIP, filled blue diamond) as competitors.The k measured /k literature ratios for PG were between 0.7 and 1.4, which is in the range of variations for second-order rate constants from different studies.Overall, the majority of k measured /k literature ratios were between 0.5 and 2, which is an acceptable variation range for competition kinetics (Figure 3a).However, using carbofuran (CBF), VM, or DMP as competitors consistently led to ratios beyond a 2-fold difference, suggesting that their k literature could be inaccurate, as discussed in more detail in the following.For diazepam (DZP), alachlor (ALA), and picloram (PCL) no k measured /k literature values were determined as they were not significantly degraded at the applied O 3 doses (Figure 2a).
Carbofuran (CBF).The k measured values of CBF with acetylsulfamethoxazole (ASMX) or BZF as competitors were 2 to 3-fold lower than its k literature (black and red filled circles on the CBF line, Figure 3a).Conversely, the k measured values of ASMX and BZF with CBF as competitor were 2 to 3-fold higher than their k literature (blue filled circles, Figure 3a).This discrepancy was attributed to the determination method of k O3,CBF .k O3,CBF (6.2 × 10 2 M −1 s −1 ) was previously measured by monitoring the O 3 decrease in excess of CBF. 25 However, the O 3 :CBF reaction stoichiometry was determined in our study to be 3:1 (see Figure S6), implying that k O3,CBF related to CBF abatement has to be divided by a stoichiometric factor of 3, resulting in a corrected k O3,CBF of 2.1 × 10 2 M −1 s −1 .Using  1).Panels (a) and (b) show the k measured /k literature ratios using unmodified and adjusted k literature , respectively (only k literature of CBF and DMP were adjusted; see explanation in the text).The vertical lines correspond to the limits for the acceptable k measured /k literature range, set between 0.5 and 2. For the abbreviations of the competitors see Table 1 (except VM which stands for vancomycin).
Vancomycin (VM).The k measured values of VM determined with trimethoprim (TMP), tylosin (TYL), sulfamethoxazole (SMX), and TRI as competitors were 3 to 7-fold higher than its k literature (symbols on the VM line, Figure 3a).Conversely, the k measured values of TMP, TYL, SMX, and TRI determined with VM as competitor were 2 to 8-fold lower than their k literature (red filled triangles, Figure 3a).VM contains a resorcinol, a phenol, and amine groups that serve as secondary reaction sites.Due to the complexity of VM (six pK a values and three activated aromatic rings) no species-specific secondorder rate constants could previously be determined and only k app,O3 values were available, which may have caused such a discrepancy. 17Because of this complexity, k O3,VM was not reassessed, omitted from Figure 3b, and not considered further in this study.
Dibromomethylparaben (DMP).The k measured values of DMP determined with TMP, CBZ, TYL SMX, and TRI as competitors were more than 10-fold lower than the reported literature value (symbols on the DMP line, Figure 3a).Conversely, the k measured values of TMP, CBZ, TYL, SMX, and TRI determined with DMP as competitor were 10 to 25-fold higher than the reported values (open diamonds, Figure 3a).k O3,DMP of the deprotonated DMP (8.4 × 10 7 M −1 s −1 ) has previously been determined by competition kinetics with indigotrisulfonate as competitor at pH ≥ 7. 30 The k O3 of indigotrisulfonate has been determined with 1,3,5-trimethoxybenzene as competitor and the k O3 of the latter with buten-3ol as competitor. 36This cascade for determinations of second- The k app,O3 values were determined by competition kinetics with the indicated competitors.For abbreviations of competitors, see Table 1.b Cyanometabolites for which an isomer with the same MS 2 fragmentation was found within 1 min of retention time.k app,O3 for the two isomers were within ±20%.c The k app,O3 of these cyano-metabolites is to be taken with caution.Cyanopeptolin C is a potential product, yet minor, of cyanopeptolin D ozonation and cyanopeptolin B a potential product, yet minor, of cyanopeptolin C ozonation (see explanation in text and in Text S5).d Whenever a cyano-metabolite was present in the two strains, two k app,O3 are reported.

Environmental Science & Technology
order rate constants might have led to accumulated errors.In the present study, k O3,DMP was redetermined using cinnamic acid and phenol as competitors, for which directly measured k O3 are available. 31,33A corrected k O3,DMP = (4.3± 0.3) × 10 6 M −1 s −1 was measured for the deprotonated DMP, 19-fold lower than the previous literature value (details are provided in Text S4).The k O3 for protonated DMP was not re-evaluated as it is expected to be negligible in the studied pH range (pK a = 4.7).Using the corrected k O3,DMP , the k measured /k literature ratios involving DMP improved from 0.04−27 to 0.7−1.8(Figure 3b).
Validation in Planktothrix Mixture and at pH 8.After correcting the apparent k O3,CBF and k O3,DMP values and excluding VM, all the k measured were within a factor of 2 of the k literature at pH 7 and therefore the selected competitors were suitable for this study.A similar exercise was done in the presence of the cyano-metabolite mixture from Planktothrix at pH 7 and 8, but only for competitors with a k app,O3 > 10 4 M −1 s −1 (see Figure S7).Again, all the k measured values were within a factor of 2 of the k literature values, excluding VM, further validating the competitor k app,O3 , especially for the pHdependent competitors ROX, TMP, and TYL.For these three competitors, due to pK a values between 7.1 and 9.2 and k O3,neutral > k O3,protonated (see Table S3), their k app,O3 values increase 1.7 to 9.5-fold when increasing the pH from 7 to 8.This pH effect can be a supplementary source of error when determining k app,O3 due to added uncertainties on pH control, pK a , and model fitting.The k app,O3 of TRI is also pHdependent, but it was too reactive to be correlated with any of the other competitors at pH 8.
Determination of Second-Order Rate Constants for the Reactions of Cyano-metabolites with Ozone.The abatement of representative cyano-metabolites as a function of the specific O 3 dose is shown in Figure 2b.The abatement of each cyano-metabolite was correlated to the abatement of the competitors with similar reactivity.Examples of abatement correlation plots used for the determination of k app,O3 of cyanometabolites are given in Figure S4.A summary of the averaged k app,O3 of cyano-metabolites at pH 7 determined with multiple competitors is provided in Table 2, classified by their main reactive moieties.Competitor-specific k app,O3 values are provided for all cyano-metabolites in Table S16.Overall, of the 31 metabolites, three did not react with O 3 at a measurable rate (k app,O3 < 10 2 M −1 s −1 ), six metabolites reacted slowly/ moderately with k app,O3 ≤ 5.0 × 10 3 M −1 s −1 , 19 metabolites reacted readily with k app,O3 ranging between 3.6 × 10 5 and 1.9 × 10 6 M −1 s −1 , and three metabolites reacted rapidly with k app,O3 ranging between 3.4 and 7.3 × 10 7 M −1 s −1 .The structure−reactivity dependency for the reaction of these metabolites with O 3 is discussed in the following.
Tryptophan and Thioether.Tryptophan-and thioethercontaining cyano-metabolites were the most reactive toward O 3 with k app,O3 at pH 7 in the range 3.4−7.3× 10 7 M −1 s −1 , excluding [Mdha-GSH7]MC-LR (Table 2).These k app,O3 were an order of magnitude higher than reported k app,O3 values for free tryptophan (7.0 × 10 6 M −1 s −1 ) and methionine (4.0 × 10 6 M −1 s −1 ). 34In the previous study, k app,O3 values for free tryptophan and methionine were measured with histidine or 3hexenoic acid as competitors for which k app,O3 are significantly lower (1.9 × 10 5 and 2.4 × 10 5 M −1 s −1 , respectively, at pH 7). 34This large difference in k app,O3 is not ideal for competition kinetics and may lead to high errors.Therefore, the k O3 values of these two amino acids need to be redetermined in future studies.The k app,O3 obtained for [Mdha-GSH7]MC-LR (1.9 × 10 6 M −1 s −1 ) was much lower than those obtained for aerucyclamide D and planktocyclin (Table 2).In [Mdha-GSH7]MC-LR, the thioether is attached to the microcystin ring on one side and to the rest of the glutathione on the other side (Figure S5).This position may cause a decrease in O 3 reactivity of the thioether in [Mdha-GSH7]MC-LR.Large variations in the reactivity of thioethers have previously been observed for pharmaceuticals. 17The measured k app,O3 for [Mdha-GSH7]MC-LR was similar to that of other microcystins (Table 2).It is therefore uncertain whether the main attack site for O 3 is the thioether or the olefins.Further investigation of the effect of complex substituents on the thioether reactivity is required.
Amines.Three cyano-metabolites contain amines: cyanopeptolin B, cyanopeptolin C, and cyanopeptolin D. These three cyanopeptolins have the same structure, apart from the amine-containing moieties, which are expected to be the main reactive sites (Table S12).Cyanopeptolin B contains a primary amine on the side chain of a lysine.Cyanopeptolins C and D contain lysine derivatives, in which the side chain amine is substituted by one and two methyls, respectively.The reactivity order of these three cyanopeptolins at pH 7 was cyanopeptolin D (tertiary amine) > cyanopeptolin C (secondary amine) > cyanopeptolin B (primary amine), consistent with the ozonation literature on amines (Table 2). 35However, the k app,O3 of the tertiary amine-containing cyanopeptolin D ((2.0 ± 0.5) × 10 3 M −1 s −1 ) was an order of magnitude higher than the k app,O3 of triethylamine reported in the literature ((2.2 ± 0.1) × 10 2 M −1 s −1 ). 35This difference may be, in part, due to a lower pK a for cyanopeptolin D compared with that of triethylamine.The predicted pK a of cyanopeptolin D was 0.5 pH-unit lower than the predicted pK a of triethylamine (pK a predicted by ChemAxon software, shown in Table S17).As the neutral amine is the reactive form with O 3 , a lower pK a leads to a higher reactivity at pH 7. Another difference between cyanopeptolins and small model compounds was the extent of the decrease in reactivity between the tertiary and secondary amine.The k app,O3 of cyanopeptolin D (tertiary amine) was 20-fold higher than the k app,O3 of cyanopeptolin C (secondary amine) (Table 2).By comparison, the literature value of the k app,O3 of triethylamine (tertiary amine) is only 1.7 times higher than the k app,O3 of diethylamine (secondary amine). 35Again, pK a may play a role.The predicted pK a for cyanopeptolin C (secondary amine) was Environmental Science & Technology 1.2 pH-units higher than that for cyanopeptolin D (tertiary amine), while the predicted pK a for diethylamine was only 0.4 pH-units higher than that for triethylamine (Table S17).An increased pK a for secondary amines compared to tertiary amines can explain, in part, their lower reactivity at pH 7.However, the k app,O3 values of cyanopeptolin C and B need to be taken with caution.A possible yet minor product of tertiary amine ozonation is the corresponding secondary amine (5% from trimethylamine). 35Similarly, the ozonation of a secondary amine can lead to minor yields of the corresponding primary amine (8% from diethylamine). 35Cyanopeptolin C can therefore be formed from cyanopeptolin D ozonation, and cyanopeptolin B can be formed from cyanopeptolin C ozonation.Although no increase was observed for cyanopeptolin C and B (Figure S8), it could have been masked by their degradation during ozonation, leading to underestimated k app,O3 and to the large k app,O3 difference between cyanopeptolins D and C.However, it was estimated that no more than 10% of the initial cyanopeptolin C should be formed from cyanopeptolin D, suggesting that its k app,O3 was not significantly underestimated (see details in Text S5).
Heterocycles.Four cyano-metabolites contained heterocycles that likely served as primary O 3 reaction sites.At pH 7, k app,O3 values for aerucyclamide A and B were much lower (<10 2 M −1 s −1 ) compared to aerucyclamide C and microcyclamides 7806A and 7806B (4.2−5.0 × 10 3 M −1 s −1 ) (Table 2).Comparing the structures of these four cyano-metabolites can help identify the reactive moieties.Oxazole was only present in the more reactive aerucyclamide C and microcyclamide 7806A and 7806B while thiazole, oxazoline, and thiazoline were present in at least one of the less reactive aerucyclamide A and B (Table S13).Altogether, this suggests that the moieties responsible for the reactivity of aerucyclamide C and microcyclamide 7806A and 7806B are oxazole.The k app,O3 of the other heterocycles are likely <10 2 M −1 s −1 and need to be further studied.
pH Effect on Reaction Kinetics.The k app,O3 of cyanometabolites present in the Planktothrix extract were also measured at pH 8 (Table S18).For the k app,O3 of olefincontaining cyano-metabolites, a <20% difference was observed between pH 7 and 8, consistent with their pH-independent reactivities (Tables 2 and S18).Conversely, the k app,O3 for phenol-containing cyano-metabolites increased by a factor ranging between 7 ± 2 and 9 ± 2 when increasing the pH from 7 to 8 (using k app,O3 determined with TYL, which is the only competitor that correlated at both pH values), due to the shift from protonated to deprotonated phenol moieties (Tables S16  and S18).This increase is close to the 10-fold increase of k app,O3 expected for phenolic compounds when increasing the pH from 7 to 8. 33 For tryptophan-and thioether-containing cyano-metabolites, large discrepancies between the two competitors used at pH 8, DMP and TRI, were observed.The k app,O3 values determined with TRI were 6-fold higher than those determined with DMP (Table S18).When validating competitors, TRI was too reactive to be correlated to any other competitor at pH 8 while DMP was validated by four other competitors (see the section Validation of Competitors and Figure S7b).Hence, the k app,O3 determined with TRI at pH 8 may be overestimated.The k app,O3 of the thioether-containing planktocyclin was comparable at pH 7 and 8 when measured with DMP ((2.8 ± 0.4) × 10 7 and (1.8 ± 0.2) × 10 7 M −1 s −1 , respectively), consistent with the expected pH-independence of thioether reactivity.For the tryptophan-containing piricyclamide ILGEGEGWNYNP +prenyl, only TRI could be correlated at pH 7 and 8; therefore the pH effect on its k app,O3 cannot be discussed.
Practical Implications.This study demonstrates the applicability of the multicompound approach for the screening of k app,O3 for many cyano-metabolites and with a minimal set of experiments.As toxins and bioactive metabolites from cyanobacteria are usually not commercially available and can only be extracted from the producing bacteria as mixtures, multicompound competition kinetics are the most efficient way to gain knowledge on the efficiency of their abatement during ozonation.Multicompound competition kinetics are a rapid screening tool to determine the kinetics of O 3 for a wide range of compounds at once.It is important to note that possible interferences associated with the complexity of the matrix and the measurement method exist: (1) the formation of a target compound from the oxidation of another (un)known compound, (2) the variation of the signal inherent to mass spectrometry, and (3) the effect of the matrix on the ionization of analytes.Strict regression criteria can help prevent such problems (high R 2 , negligible intercept, and restricted abatement ranges).In addition, the use of multiple competitors per target compound provides more resilience, preventing a strong bias when determining the k app,O3 .To this end, it was shown that a number of competitor compounds had incorrect k app,O3 , which could be problematic if they would have been used as single competitors.Therefore, a cross check of secondorder rate constants is recommended for competition kinetics experiments.A limitation for the determination of secondorder rate constants was the lack of competitors with k app,O3 > ∼10 7 M −1 s −1 , which was evident for the determination of k app,O3 of thioether-and tryptophan-containing cyano-metabolites.Therefore, further studies are needed to determine second-order rate constants at the higher end of the O 3 reactivity, even though for compounds with such high reactivities, a complete abatement is already expected at low specific O 3 doses.Overall the proposed novel approach has many advantages as a screening tool; however, if k app,O3 needs to be determined with high precision, conventional methods focusing on individual compounds should be favored, and if possible, direct methods and not competition kinetics should be applied.
The majority of the 31 cyano-metabolites identified in Microcystis and Planktothrix had k app,O3 ≥ 10 5 M −1 s −1 , indicating they should be degraded by specific O 3 doses typically applied in drinking water treatment. 12Without reactive moieties such as tryptophan, thioether, phenol, and olefin, other cyano-metabolites showed significantly lower reactivity (k app,O3 ≤ 10 3 M −1 s −1 ), indicating they might only be partially degraded by O 3 .In this case, the oxidation by the secondarily formed • OH may enhance their abatement significantly. 37Because of the high molecular weights of the detected cyano-metabolites and a low selectivity of • OH, it can be assumed that the second-order rate constants for the reaction of cyano-metabolites with • OH are close to diffusion control (10 10 M −1 s −1 ). 38,39For example, MC-LR and aerucyclamide A have k app,OH values of 1.1 × 10 10 and 6.4 × 10 9 M −1 s −1 , respectively, at pH 7. 12,40 Furthermore, the multicompound competition kinetics allowed us to gain insight into structural moieties that have been underrepresented in the ozonation literature thus far.This is notably exemplified in this study by the high reactivity of tryptophan-and thioethercontaining cyano-metabolites compared to the literature,

Environmental Science & Technology
revealing a lack of knowledge on the O 3 reactivity of these moieties.In addition, preliminary insights were obtained on the reactivity of O 3 with the heterocycles oxazole, thiazole, oxazoline, and thiazoline.

Figure 1 .
Figure 1.Representative cyano-metabolites detected in Microcystis aeruginosa and Planktothrix rubescens cultures.Highlighted moieties represent the parts of the molecules that can vary in the other cyano-metabolites of the same class identified in this study.Circles indicate the main attack sites of the O 3 .The full lists of cyano-metabolites and their structures are provided in Table S4, Tables S11−S14, and Figure S5.

Figure 2 .
Figure 2. Simultaneous abatement of (a) the selected competitors and (b) representative cyano-metabolites from 0.6 g biomass L −1 of Microcystis as a function of the specific O 3 dose at pH 7 (2 mM phosphate) and 22 °C and in the presence of tert-butanol (40 mM).For the abbreviations of the competitors, seeTable 1 (except VM which stands for vancomycin).

Figure 3 .
Figure 3. Competitor evaluation during the ozonation of a Microcystis extract (0.6 g biomass L −1 ) at pH 7 (2 mM phosphate) and 22 °C and in the presence of tert-butanol (40 mM).The evaluation was done by calculating the ratios between the k app,O3 determined by pairs of competitors (k measured ) and the k app,O3 from the literature (k literature , see Table1).Panels (a) and (b) show the k measured /k literature ratios using unmodified and adjusted k literature , respectively (only k literature of CBF and DMP were adjusted; see explanation in the text).The vertical lines correspond to the limits for the acceptable k measured /k literature range, set between 0.5 and 2. For the abbreviations of the competitors see Table1(except VM which stands for vancomycin).

Table 1 .
List of 16 Selected Competitors with Their k app,O3 Values at pH 7 a

Table 1 (
except VM which stands for vancomycin).