Analysis of total microcystins and nodularins by oxidative cleavage of their ADMAdda, DMAdda, and Adda moieties

Microcystins (MCs) and nodularins (NODs) exhibit high structural variability, including modifications of the Adda (3S-amino-9S-methoxy-2S,6,8S-trimethyl-10-phenyldeca-4E,6E-dienoic acid) moiety. Variations include 9-O-desmethylAdda (DMAdda) and 9-O-acetylDMAdda (ADMAdda) which, unless targeted, may go undetected. Therefore, reference standards were prepared of [ADMAdda5]MCs and [DMAdda5]MCs, which were analyzed using multiple approaches. The cross-reactivities of the [DMAdda5]- and [ADMAdda5]MC standards were similar to that of MC-LR when analyzed with a protein phosphatase 2A (PP2A) inhibition assay, but were <0.25% when analyzed with an Adda enzyme-linked immunosorbent assay (ELISA). Oxidative cleavage experiments identified compounds that could be used in the analysis of total MCs/NODs in a similar fashion to the 2R-methyl-3S-methoxy-4-phenylbutanoic acid (MMPB) technique. Products from oxidative cleavage of both the 4,5- and 6,7-ene of Adda, DMAdda and ADMAdda were observed, and three oxidation products, one from each Adda variant, were chosen for analysis and applied to three field samples and a Nostoc culture. Results from the oxidative cleavage method for total Adda, DMAdda, and ADMAdda were similar to those from the Adda-ELISA, PP2A inhibition, and LC-MS/MS analyses, except for the Nostoc culture where the Adda-ELISA greatly underestimated microcystin levels. This oxidative cleavage method can be used for routine analysis of field samples and to assess the presence of the rarely reported, but toxic, DMAdda/ADMAdda-containing MCs and NODs.

One frequently employed method for the analysis of MCs and NODs is enzyme linked immunosorbent assay (ELISA). There are multiple ELISAs developed for the analysis of MCs/NODs, each exhibiting differential cross-reactivity to MC and NOD congeners depending on the antibody development approach used. For instance, ELISAs with antibodies raised against MC-LR have exhibited low-to-no cross-reactivity to non-arginine MCs [18], and to [ADMAdda 5 ]MCs even when arginine is present [11]. An improvement to congener cross-reactivity was achieved through the development of an ELISA with antibodies raised against Addahaptens [19]. The Adda-ELISA is commercially available and utilized in countries such as the USA to screen for MCs/NODs in ambient source and drinking water [20]. Results are directly actionable by utilities in some states, resulting in drinking water treatment plant and recreational beach closures [21]. The Adda-ELISA was chosen for monitoring as it is expected to react to MCs/NODs with approximately equal sensitivity, regardless of the remaining amino acid composition. However, although modifications to the Adda moiety could potentially alter the cross-reactivity, this has not been tested experimentally. One broadly specific ELISA developed using a multi-hapten approach, was shown to crossreact with crude extracts containing ADMAdda-and DMAddacontaining MCs, but this assay is not currently commercially available and the cross-reactivity was not measured quantitatively [22].
The identification of ADMAdda and DMAdda variants is currently limited to congener-specific methods (e.g. LC-MS/MS).  (3,4,5,6) were isolated in this study. Amino acid residue-numbers are shown in the circles, while atom numbering for each residue is shown in plain text starting from the carboxyl carbon.
However, standards for instrument calibration are not available and, unless targeted, some congeners might remain undetected. In order to facilitate analysis of these modified-Adda-variants, the approach utilized for total Adda determination via oxidative cleavage of Adda to MMPB (2R-methyl-3S-methoxy-4phenylbutanoic acid) could be applied. The originally-reported use of oxidation to cleave the Adda to measure total MCs was developed based on methodology for the analysis of unsaturated fatty acids [23,24]. The method preserves acyl ester bonds, while allowing for the quantitative determination of the oxidized products. Adda possesses olefinic bonds at C-4 and C-6, and oxidative cleavage of the 6,7-olefinic bond results in the formation of MMPB (Fig. 2). The MMPB approach has been used to quantitatively measure total Adda-containing MCs and NODs in water, benthic periphyton and animal tissues [25e28]. However, to date, there have been no reports of using oxidative cleavage for analysis of MCs containing modified Adda moieties.
In this work, four [ADMAdda 5 ]MC variants were extracted, purified and reference standards produced. Hydrolysis of [ADMAdda 5 ] MCs produced standards of two [DMAdda 5 ]MCs which were characterized by NMR spectroscopy and LC-MS. The standards were analyzed using a commercial PP2A inhibition assay and an Adda-ELISA to determine cross-reactivities. The MMPB method for oxidative cleavage and analysis of Adda-containing MCs/NODs was augmented to include ADMAdda-and DMAdda-containing MCs. This approach was applied to three field-collected samples and a culture of a Nostoc sp. to illustrate the potential of the method for monitoring for these rarely tested variants.
The residue was dissolved (20% CH 3 CN; 2 mL) and purified by semi-preparative high-performance liquid chromatography (HPLC) using a Thermo Separations Product P4000 Pump, with a UV 2000 Detector set to 238 nm and its output converted to a digital signal using an SN 4000 Controller. Details of the linear gradient conditions for all semi-preparative HPLC methods are shown in Table S1. Initial separation (Method 1 in Table S1) was achieved using a Luna C18 column (5 mm, 150 Â 10 mm, Phenomenex) and mobile phases A (0.01% TFA) and B (MeOH) at 2 mL min À1 Two major chromatographic peaks were collected (Fig. S1) eluting at 18.38 (3 and 4) and 19.47 (5 and 6) min. A portion of the first peak (18.38 min) was set aside for base hydrolysis (section 2.2.2), and the remainder was further separated by semi-preparative HPLC using the same column (Method 2 in Table S1) to give partial separation of 3 and 4 (Fig. S2). Final purification was achieved by semi-preparative HPLC on a NovaPak C18 column (4 mm, 4.6 Â 250 mm, Waters Corporation, Milford, MA, USA) (Method 4 in Table S1, Fig. S3) to give [ADMAdda 5 ]MC-LR (3) and [ADMAdda 5 ]MC-LR (4) with >95% purity. The peak containing 5 and 6 from the initial semi-preparative HPLC step was further purified (Method 3, Table S1  containing purified 3e6 were evaporated and the residue dissolved in water (1 mL) for characterization.

Purity and quantitation of [DMAdda 5 ]MCs and [ADMAdda 5 ] MCs
A Thermo Scientific Surveyor HPLC system coupled to a Surveyor photodiode array (PDA) detector and an LTQ XL Linear Ion Trap Mass Spectrometer were employed as previously described [27,30]. Briefly, analytical separations were achieved using a Kinetex C18 column (2.6 mm, 100 Å, 150 Â 2.1 mm; Phenomenex) with mobile phases of water (A) and 95% CH 3 CN (B), both containing 2 mM formic acid and 3.6 mM ammonium formate. The gradient (0.2 mL min À1 ) was: A held at 70% for 10 min, 70e65% A over 8 min, held 65% A for 2 min, 65e30% A over 4 min, 30e70% A over 2 min, and held at 70% A for 4 min. Purity was assessed using HPLCePDA (200e600 nm) of each standard at ca 10e20 mg mL À1 .
Quantitation was based on HPLCeUV (l ¼ 238 nm) peak areas relative to a CRM of MC-LR (9) as MCs purified in this work share identical UV chromophores to MC-LR. Identities were assigned through the comparison of LCeUVeMS n data (retention time, spectra) to previous work [7]. MS/MS scans were conducted using (approximately 200 mg each) were dispensed into vials and the solvent evaporated for NMR spectroscopy. Remaining solutions were portioned into 10 mg aliquots, the solvent evaporated, and stored at À20 C. A set of working stock solutions at 1.0 mg mL À1 in 10 mM phosphate buffer (pH 7) were maintained (À20 C) for experiments.  [27]) of the USA were screened for cyanobacterial dominance. Wet mounts were scanned using a Nikon TE200 inverted microscope equipped with phase-contrast optics at up to 400 Â . Samples (200e500 mL) were lyophilized, dried cells extracted, and fractionated by SPE as described above for the Nostoc sp. strain 152. The resultant eluates were evaporated, the residues reconstituted in 10% CH 3 CN (approx. 800 mg biomass mL À1 ), and further diluted (water) for analysis.
Intact MCs were quantitated as diluted aliquots in water (biomass concentrations of 0.01e1 mg mL À1 ) with internal standards added (d 7 -MC-LR and d 5 -MC-LF), and analyzed using targeted LC-MS/MS (section 2.6.2.2) for NOD-R and 21 MCs (Table S2). Standards used to calibrate the method included the six isolated in this work (1e6) and those listed in section 2.1 and Table S2. Transitions used to monitor the [ADMAdda 5 ]MCs (3e6), [DMAdda 5 ]MCs (1 and 2), and other MCs (7e11) sharing the same precursor ions are shown in Table 1. The remaining MRM transitions were as previously reported [31] (Table S2). Chromatographic procedures were as described in section 2.2.4. Xcalibur v 2.2 (Thermo Fisher Scientific, Waltham, MA, USA) was utilized for data processing via the internal standard method previously described [31].
The extract from the mid-west Microcystis bloom was analyzed by LC-HRMS/MS as previously described [32] except that the stepped collision energy used to acquire HRMS/MS spectra was 30 and 50 eV, with the same hardware and mobile phases as in section 2.6.4.

PP2A inhibition assay
The field sample extracts used for ELISA analysis were also analyzed using a protein phosphatase 2A (PP2A) kit (Eurofins Abraxis) at the same sample concentrations, except for the Nostoc sp. strain 152 extract, which was diluted to fit the calibration curve. A CRM of MC-LR (9) and each isolated variant (1e6) were diluted in water (0.25, 0.5, 1, 2.5 ng mL À1 ) and analyzed in duplicate following the manufacturer's directions (Fig. S5). The same plate reader used in ELISA analysis was utilized to obtain absorbances at 405 nm. SigmaPlot 12.5 (Systat Software Inc., San Jose, CA, USA) was used to calculate IC 50 values. MC-LR (3) (10 mg mL À1 ) were oxidized for characterization of their oxidation products. Standard curves (in duplicate) were also prepared by oxidation of 1, 3 and 9 at concentrations 1, 5, 10, 50, and 100 ng mL À1 . The remaining variants isolated in this study (2 and 4e6) were assessed with the oxidative cleavage method as 4-point curves (5, 10, 50, and 100 ng mL À1 ). Each field sample extract was diluted in water (biomass concentrations 0.2e1 mg mL À1 ) to achieve analyte responses within the range of the standard curves. Each extract was oxidized in triplicate with spikes (n ¼ 2). Spikes were prepared pre-oxidation using standards of MC-LR, [DMAdda 5 ] MC-LR and [ADMAdda 5 ]MC-LR at sufficient levels to double peak areas for quantitation (standard addition). Solutions of 1 M K 2 CO 3 , 0.25 M KMnO 4 and 0.25 M NaIO 4 were prepared in water. The oxidant was premixed just prior to addition, with each reaction containing 100 mL K 2 CO 3 , 200 mL KMnO 4 , 200 mL NaIO 4 , and the sample (diluted to 500 mL with water), for a final reaction volume of 1 mL. Final reagent concentrations were 50 mM KMnO 4 , 50 mM NaIO 4 and 100 mM K 2 CO 3 . After 1 h, 40% (w/v) NaHSO 3 (75e125 mL) was added until solutions were clear.
A time-course for oxidative cleavage of [DMAdda 5 ]MC-LR (1), [ADMAdda 5 ]MC-LR (3) and MC-LR (9) was conducted (in triplicate) to monitor the formation of targeted oxidation products at ambient temperature. A solution (500 mL) containing 250 ng of each standard in water was oxidized (addition of 500 mL oxidant, as above), and sub-sampled (150 mL, with addition of 50 mL 40% (w/v) NaHSO 3 ) after 5, 15, 30, 60 and 120 min. The solutions were fractionated by SPE, the residue from the eluate dissolved in water (375 mL), and filtered, as described above.

Isolation of MOMAPH
Lyophilized Nostoc sp. strain 152 (75 mg) was vortex-mixed in 2.5 mL of oxidant (50 mM KMnO 4 , 50 mM NaIO 4 , 100 mM K 2 CO 3 ), allowed to react for 1 h, and the reaction stopped by dropwise addition of 40% (w/v) NaHSO 3 until the solution turned cloudy white. The sample was centrifuged and the supernatant retained. The pellet was resuspended (water; 1 mL), centrifuged, and the two supernatants pooled. The solution was applied to a Strata-X SPE column (200 mg), and the column was washed (water; 3 Â 1 mL), eluted (90% CH 3 CN; 5 mL), and the eluate evaporated to dryness. The residue was dissolved (10 mM phosphate buffer, pH 7; 2 mL) and injected onto a NovaPak C18 (4 mm, 4.6 Â 250 mm; Waters) column eluted isocratically with 35% CH 3 CN containing 2 mM formic acid and 3.6 mM ammonium formate (1 mL min À1 ) while monitoring the absorbance at 254 nm using the HPLC equipment described in Section 2.2.1. Fractions (1 mL each) were collected and analyzed by LC-MS/MS, which identified the compound as a single UV-absorbing peak eluting at ca 12 min. Fractions containing the compound were combined and the solvent evaporated. to equilibrate the column. Injection volume was 5 mL. The mass spectrometer was calibrated from m/z 74e1622 and m/z 69e1780 in positive and negative ionization modes, respectively, the spray voltage was 3.7 kV, the capillary temperature was 350 C, and the sheath and auxiliary gas flow rates were 25 and 8 units, respectively, with MS data acquired from 2 to 20 min. Mass spectral data were collected using full scan mode with alternating positive and negative scans with data collected from m/z 150e500 using the 60,000 resolution setting, an AGC target of 1 Â 10 6 and a max IT of 120 ms. Putative MOMAPH was further probed in a targeted manner in negative ionization mode using the PRM scan mode at m/z 291.1 with a ±0.5 m/z precursor isolation window, the 30,000 resolution setting, an AGC target of 1 Â 10 6 and a max IT of 100 ms, with a stepped collision energy of 15, 20 and 25 eV.
Chromatographically, [DMAdda 5 ]MCs 1 and 2 eluted far earlier than the Adda 5 -containing MC-LR (9), with [ADMAdda 5 ]MCs 3 and 4 eluting just after 9, followed by [D-Asp 3 ,ADMAdda 5 ]MC variants 5 and 6 (Fig. 3). The early elution of [DMAdda 5 ]MCs presents a benefit as well as possible pitfall. While their chromatographic behaviour provides additional qualification for compound identification when reference standards are not available, care should be taken to ensure that the detectors (e.g. MS, UV) are acquiring data and that peaks of interest do not co-elute with non-retained matrix. MS/MS fragmentation patterns also allow for some differentiation, as relative fragment ion intensities varied between Adda-, DMAddaand ADMAdda-containing MCs (Table 1). Diagnostic product ions for Arg 4 -containing ADMAdda-and DMAdda-MCs included m/z 627 and m/z 585, respectively, from [Arg 4 -ADMAdda/DMAdda 5 -Glu 6 þ H] þ . When Har 4 was present, diagnostic ions included m/z 641 for ADMAdda 5 -containing MCs (4, 6), but the corresponding dominant fragment ion with m/z 599 for [DMAdda 5 ]MC-LHar (2) was not unique as it was shared with Arg 4 -Adda 5 -containing MCs. The origin of the m/z 599 product ion from Arg 4 -containing ADMAdda-MCs has been reported as being [Arg 4 -ADMAdda 5 -Glu 6 e CO þ H] þ [9], which presents with the same m/z with [Arg 4 -Adda 5 -Glu 6 þ H] þ . LC-MS/MS generated curves exhibited good linearity (All R 2 ! 0.995) from 1 to 100 ng mL À1 for isolated standards (1e6) (Fig. S8). Stock solutions of each variant were monitored over the course of the study and were determined stable for at least one year stored between experiments at À20 C.

NMR spectroscopic analysis of [DMAdda 5 ]MCs
The identities of the semisynthetic [DMAdda 5 ]MCs 1 and 2 were confirmed through NMR spectroscopy ( Table 2) because no authentic standards were available for these compounds, and no published NMR data was available for 2. Detailed analyses of 1 H, DEPT, DEPTQ, COSY, DIPSI2, HSQC and HMBC NMR spectra recorded from CD 3 OH with ES, CW or combined ES and CW presaturation of the large H 2 O/HOD line at ca. 4.8 ppm and CW presaturation of residual CHD 2 OH lines. This, supported by higher resolution SEL-TOCSY, SHSQC and SHMBC spectra, established that 1 was the 9-O-desmethylAdda analogue of MC-LR and that 2 was an analogue of 1 containing Har instead of Arg at position-4. The 1 H and 13 C NMR chemical shifts, and 1 H 1 H coupling constants of 1 and 2, where resolved, are reported in Table 2. These assignments can be compared with those reported for MC-LR (9) [3]

and [ADMAdda 5 ] MC-LHar (4) [3], and with 1 H NMR data for MC-LHar (22) [34] and [DMAdda 5 ]MC-LR (1) [4], in CD 3 OD.
The presence of a 9-OH group in the 9-O-desmethylAdda unit of 1, as opposed to a 9-OCH 3 group as in MC-LR, was revealed by the absence of 1 H [4] and 13 C NMR signals attributable to the presence of an Adda 9-OCH 3 group, and by the occurrence of the Adda C-9 signal of 1 at 78.2 ppm and its H-9 signal at 3.60 ppm, respectively, rather than at 88.3 ppm and 3.27 ppm, respectively, as reported by Namikoshi et al. [3] for MC-LR (9). The 2 J and 3 J couplings of the H-10 a (2.59 ppm) and H-10 b (2.82 ppm) signals of the desmethylAdda residue of 1 (dd, J ¼ 13.9, 8.6 Hz, and dd, J ¼ 13.9, 4.2 Hz, respectively), corresponded closely to those reported (to ± 0.5 Hz) for the equivalent protons of MC-LR [3]. These observations are consistent with the relative configuration of C-9 of the desmethylAdda of 1, and the stereochemical disposition of H-9 relative to the H-10 a and H-10 b methylene protons, being the same as those of the equivalent protons of MC-LR (9). Similarly, the 1 H NMR chemical shifts and coupling constants observed for the H-3 (4.57 ppm, m), H-4 (5.55 ppm, dd, 15.5, 9.0 Hz) and H-5 (6.25 ppm, d, 15.5 Hz) signals of the 9-O-desmethylAdda residue of 1 were essentially identical to those reported for the corresponding protons of MC-LR (9) [3]. Furthermore, all 1 H chemical shift and coupling constant assignments for the Adda 5 and Mdha 7 moieties of 1 in CD 3 OH were also nearly identical to those reported by Namikoshi et al. [4] for 1 in CD 3 OD. The D-Glu 6 H-4 b of 1, which occurred at 2.68 ppm as a ddd (Table 2) Correlations observed in COSY and in DIPSI2 experiments performed with mixing times of 80 and 160 ms verified that the foregoing proton signals assignments, and also those of protons associated with the other amino acid residue units of 1, were as reported in Table 2. Correlations observed in the ROESY NMR spectrum, and a series of higher resolution SELROESY spectra, of 1 verified that the diene portion of the DMAdda 5 residue was transsubstituted and had not been epimerized to a cisoid analogue [34]. Other than for the 1 H and 13 C NMR signals arising from the Har residue of 2, there was a close correspondence between the 1 H and 13 C assignments of 2 with those established for 1 ( Table 2). The H-2 signal of Har (4.35 ppm) exhibited a COSY correlation to the pair of non-equivalent H-3 protons of the Har residue at 1.52 and 2.02 ppm, respectively, while longer range correlations observed in DIPSI2 and in higher resolution 1D-SELTOCSYspectra performed with mixing times of 80 and 160 ms identified the resonances attributable to the H-3 a/b (1.32 and 1.37 ppm) and H-4 (1.54 ppm) methylene protons and the H-5 methine proton (3.13 ppm) of the Har residue of 2. The 1 H and 13 C shifts of 2 were correlated in an HSQC spectrum, and in the case of the Har C-4 signal at 29.38 ppm, a higher resolution SHSQC spectrum differentiated it from the D-Glu C-3 signal at 29.42 ppm.
ROESY correlations analogous to those observed for 1 were also observed for 2, indicating that, notwithstanding the presence of a Har residue in 2 compared to an Arg residue in 1, the preferred  Table 1. solution confirmation of the MC ring system of 2 was similar to that of 1 and comparable to those previously reported [35] for 9 and 12. This conclusion is also consistent with the finding that the 1 H and 13 C shifts of the amino acid residues present in 1 and 2, other than parts of their DMAdda 5 and Har 4 units, were very similar to those previously reported for MC-LR (9) [3]. NMR supporting data can be accessed in the SI file (Figs. S9eS31).

Adda-ELISA and PP2A inhibition assay
The Adda-ELISA did not react to the purified [ADMAdda 5 ]MCs (1 and 2) or [DMAdda 5 ]MCs (3e6) at concentrations of 1 or 10 ng mL À1 . Therefore, higher concentrations were tested (50e1000 ng mL À1 ), with IC 50 values determined to be > 200 ng mL À1 as compared 0.49 ng mL À1 for MC-LR (9) (Table S3; Fig. 4 and S32), giving crossreactivities of under 0.25% relative to MC-LR. This is unsurprising given the assay design concept, as the Adda-ELISA was developed to recognize the unmodified Adda epitope [19]. Although some reports suggest that the Adda-ELISA responds to MC congeners containing modified Adda moieties [8], this work demonstrates that the crossreactivity is very low. Low cross-reactivity with [ADMAdda 5 ]MCs was also reported with anti-MC-LR polyclonal antibodies [11], indicating the need for alternative approaches to MC ELISA antibody development if MC congeners containing modified Adda moieties are to be quantified by immunoassay methods.

Oxidative cleavage experiments
The MCs purified in this work were used to develop an oxidative cleavage procedure to test for total MCs and NODs based on the MMPB approach for Adda-containing MCs [24,28]. Oxidative cleavage of [ADMAdda 5 ]MC-LR (3) and [DMAdda 5 ]MC-LR (1) (Fig. 1) produced compounds (Fig. 2) analogous to MMPB (from MC-LR (9)) that were initially observed in total ion LCeMS spectra (m/z 180e300) (Fig. S33) LCeMS/MS experiments were used to establish tentative structural identities and to verify that oxidation products were conserved across congeners containing the same type of Adda variant. Similar to the chromatographic behavior of intact MCs (elution of DMAdda-, followed by Adda-, and finally ADMAdda-containing congeners), the oxidized products followed the same order of retention, with the smaller molecules (MHPB, MMPB, MAPB) eluting approximately 1 min prior to their larger counterparts (MOMHPH, MOMMPH, MOMAPH) ( Fig. 5 and S34).
During the oxidative cleavage of the ADMAdda in 3, the expected smaller compound MAPB, formed through cleavage of the 6,7-ene, was initially targeted. However, although a peak with a pseudomolecular ion corresponding to MAPB (m/z 235 in negative mode) was detected after oxidation, a peak with the same retention time and product ion spectrum was also observed after oxidation of [DMAdda 5 ]MC-LR (1) (Fig. 5 and S35). Oxidation of the Addacontaining MC-LR (9) did not result in the formation of this compound, so m/z 235 was therefore considered to be unique to ADMAdda and DMAdda, but not Adda. Since the m/z 235 peak was not unique to ADMAdda, further investigations using MAPB as a diagnostic compound were abandoned. Rather, the compound formed through cleavage of 4,5-ene, and exhibiting [MÀH] À at m/z 291, was assessed as a unique conserved product from oxidative cleavage of ADMAdda-containing MCs.
The target used in analysis of oxidized ADMAdda, isolated and characterized in this work, was MOMAPH (possibly together with its corresponding enol), which was confirmed using both low-and high-resolution mass spectrometry. LC-MS/MS analyses in negative ionization of the peak with m/z 291 (MOMAPH) showed a facile  (Fig. S38). Positive ionization MS/MS spectra of [MþH] þ of MOMAPH at m/z 293 resulted in data-rich spectra (Figs. S39 and S40), which were analyzed together with the negative ionization MS/MS spectra (Fig. 6) in the structure elucidation. Purification of the oxidized product (Fig. S41) followed by LCeHRMS/MS analysis showed it to be composed of two isomeric forms, with m/z 291.1242 in negative and m/z 293.1380 in positive ionization modes, although the later-eluting isomer formed a prominent ammonium adduct ion in positive ionization mode (Fig. 7). LCeHRMS was consistent with a neutral elemental composition of C 16 H 20 O 5 for both peaks, and LCeHRMS/MS data ( Fig. 7) was also consistent with the proposed product ion identities (Fig. 6).
[ADMAdda 5 ]MCs purified in this work (3e6) were used to prepare 5-point standard curves ranging from 1 to 100 ng mL À1 and oxidized, and the resultant oxidation product, MOMAPH, was analyzed via LC-MS/MS (Fig. 8). The standards produced similar response curves exhibiting linear coefficients of determination (R 2 ) ! 0.992. The differences observed in MOMAPH formation may be attributed to variability introduced during quantification of the original standard (via HPLCeUV relative to MC-LR), variability in oxidation efficiency specific to analyte chemistry, or the competition of oxidant with the two closely located alkenes. Differences in molecular weight of the intact congeners varied less than 3% and could not have significantly contributed to observed differences in MOMAPH production.
LCeMS/MS experiments conducted on oxidatively cleaved Oxidation products formed through cleavage of 4,5-ene of the DMAdda moiety were observed as two chromatographic peaks eluting approximately 1 min after MHPB (Fig. 5). These two products may represent MOMHPH and its corresponding enol, or an additional isomer produced via ketoeenol tautomerism. Both isomers shared common product ions at m/z 119 (C 8 H 7 O À ) and 85 (C 5 H 9 O À ) (Fig. S44). The peak areas of both isomers were combined (integration of both peaks) for the time-course assessment and calibration curves, but ultimately, MHPB was used for analysis of field samples. Standard curves (Fig. 8) (2). In contrast, the opposite situation was observed for the formation of MOMHPH and MHPB from 1 and 2, indicating that differences in compound chemistry affect the cleavage reaction even with two very closely related alkenes. The LCeMS/MS analyses of oxidized MC-LR (1) confirmed the presence of MMPB, with a prominent product ion at m/z 131 (C 10 H 11 À ) from both ion trap and triple-quadrupole MS systems  [37]. Although MOMMPH might provide an alternative for the determination of total Adda MCs, the use of MMPB for quantitative analysis of total Adda-containing MCs has been well established [28]. Because it is also unknown whether MOMMPH exists in nature or could be produced by oxidation of other endogenous compounds, MMPB was used for routine sample analysis. However, monitoring of MOMMPH is warranted to determine its applicability as a quantitative metric for total Adda-containing MCs in future work. Oxidation products were not assessed for stability during longterm storage, but were stable in water or 5% methanol during short term storage ( 30 d; À20 C). The mechanism(s) driving the cleavage of the 4,5-ene vs. the 6,7-ene were not explored in this work. Rather, conserved products were chosen and applied to the analysis of field collections. However, since the two olefinic sites of oxidative cleavage are in close proximity (Fig. 2), it is possible that the oxidation conditions used in this study played a role in the relative oxidation product concentrations (Fig. 8). The KMnO 4 treatment of microcystins at neutral pH results in oxidation of both the 4,5-ene and 6,7-enes of the Adda to give a-hydroxyketones, which are further oxidized to produce cleavage products (e.g. carboxylic acid) [38]. Olefins oxidized with Lemieux reagent, similar to the oxidant used in this study, showed that KMnO 4 first converted olefins to hydroxyketones, which were rapidly cleaved by NaIO 4 , and products further oxidized by KMnO 4 [39]. The use of NaIO 4 is thought to allow the reaction to proceed with high specificity and at a faster rate than when using KMnO 4 alone [39]. While reaction conditions were not modified in this work to assess their effect on product formation, several parameters could easily be adjusted (e.g. pH, temperature) to affect reaction rates and yields.

Time course
Oxidative cleavage products reached !90% of their maximum concentrations within 30 min (Fig. 9). The 6,7-ene cleavage products (MHPB, MMPB) formed faster (75% in 5 min) than their counterparts from cleavage of the 4,5-ene (46e48% in 5 min). However, losses of MMPB were observed between 60 and 120 min, in accord with other studies that report degradation of MMPB over time [26]. Based on these observations, oxidation reaction times were limited to 60 min.

Sample analyses
The Nostoc sp. strain 152 culture and three lyophilized grab samples collected from the 'West Coast', 'Midwest' and 'East Coast' of the USA were extracted and analyzed by four different techniques (Table 3). A targeted LC-MS/MS method (21 MCs and NOD-R) was compared to the Adda-ELISA, PP2A inhibition assay, and the new oxidative cleavage procedure. The main congener from the west coast Dolichospermum-dominated bloom was MC-LR (9) to be present (Table 3, Fig. 10). LCeHRMS/MS analysis indicated that [DMAdda 5 ]MC-LA was present (Fig. S48), but this was not targeted for quantification in the LCeMS/MS analysis due to a lack of a suitable reference material. The sum of MCs by oxidative cleavage was 1840 mg g À1 , with 86% of these MCs accounted for in the targeted LCeMS/MS. Total MC/NOD concentrations by oxidative cleavage were similar to those obtained by PP2A inhibition (1500 mg g À1 MC-LR equivalents) and Adda-ELISA (1700 mg g À1 Adda-containing MCs). The targeted LCeMS/MS analysis of MCs in the bloom collected from the Poplar reservoir in the Chesapeake Bay revealed the predominance of [D-Leu 1 ]MC-LR together with lower levels of MC-LR, as reported previously [27]. Only two variants were present above method detection limits, but previous work showed contributions of at least 25 additional minor variants of MC that were not targeted in this work due to a lack of available calibration standards. Oxidative cleavage of the bloom material revealed the presence of DMAdda (10 mg g À1 ), which was likely primarily from contributions of [D-Leu 1 ,DMAdda 5 ]MC-LR previously reported in the sample [27] but not targeted in this study. If the 3% contribution of DMAdda relative to Adda followed the same pattern, the level of [DMAdda 5 ] MC-LR would have been 0.9 mg g À1 , which is below the method detection limit (1 mg g À1 ). The Adda-ELISA (350 mg g À1 ) and PP2A inhibition (400 mg g À1 ) analyses indicated that most of the MCs were accounted for in the targeted LCeMS/MS analysis (sum 386 mg g À1 ), with the total by oxidative cleavage (620 mg g À1 ) possibly representing some decomposed MCs (unrecognized by the  ELISA) (stored 7 y).
MOMAPH (from oxidation of ADMAdda) was not detected in any of the three oxidized planktonic bloom extracts, but was quantitatively measured in the Nostoc sp. strain 152 culture using the oxidative cleavage procedure (Table 3, Fig. 10). The sum of [ADMAdda 5 ]MCs by LCeMS/MS was 679 mg g À1 , 15% higher than the estimate using the oxidative cleavage method (588 mg g À1 total ADMAdda-MCs). The concentration of DMAdda by oxidative cleavage was 21 mg g À1 , almost double the total [DMAdda 5 ]MCs targeted by LCeMS/MS (13 mg g À1 ), although the basic conditions of the oxidation procedure (pH > 10) may have caused partial hydrolysis of the acetate group, which might account for this. Kinetic Table 3 Concentrations of MCs (mg g À1 dry weight) in crude extracts of bloom material and Nostoc sp. 152 culture using 4 techniques: targeted LC-MS/MS analysis, Adda-ELISA, PP2A inhibition assay, and total MCs/NODs by oxidative cleavage. Standard deviations for multiple extractions (when conducted) and the lowest achieved method detection limits (MDLs) are also shown. NOD-R and [D-Asp 3 ]MC-RR (21) were the only targeted analytes not detected.  5 ]MCs, the Adda-ELISA showed very low cross-reactivity with crude extracts Nostoc sp. strain 152. The MCs measured with the other three methods were >250 times the level measured by the Adda-ELISA. While the Adda-ELISA was not representative of the toxic MC content, the PP2A inhibition assay detected 1000 mg g À1 MC-LR equivalents, which was higher than the sum of MCs by oxidative cleavage (609 mg g À1 ) and the sum of MCs targeted by LCeMS/MS (692 mg g À1 ).
The results of samples analyzed in this work and others [16,40,41]

Conclusion
The comprehensive analysis of microcystins (MCs) and nodularins (NODs) is challenging due to the numerous structural variations that may be present in a given sample. Although broadspecificity analytical techniques are able to account for some modifications, methods targeting the Adda moiety failed to detect ADMAdda 5 -and DMAdda 5 -containing MCs. Furthermore, the lack of commercially available standards hampers targeted analysis approaches. This is problematic, as protein phosphatase inhibition assays (this work and [11]) and mouse bioassays [5,36] indicate the toxic potential of ADMAdda 5 -and DMAdda 5 -containing MCs to be similar to that of MC-LR (9). Therefore, the existing oxidative cleavage and analysis for Adda-containing MCs and NODs (i.e. the MMPB method) was augmented to include ADMAdda and DMAdda variants to achieve a comprehensive analysis of total MCs and NODs.
During the investigation of the chemical oxidation of MCs, it was determined there were two sites for the oxidative cleavage of Adda, ADMAdda, and DMAdda, leading to multiple oxidation products that could be targeted for analysis. This observation is important, as the competitive oxidation of the two olefinic sites could impact quantitation if not carefully calibrated using pre-oxidation standard addition with representative congeners (e.g. [DMAdda 5 ]MC-LR, [ADMAdda 5 ]MC-LR). Utilizing standard addition, the oxidative cleavage and analysis procedure applied to field samples was not only helpful in qualifying results, but was also successful in estimating the total Adda-, DMAdda-and ADMAdda-containing MCs in the samples. Data and interpretations were confirmed using LC-MS/MS, PP2A inhibition assay, and the Adda-ELISA.