Analysis of microcystins in alum water treatment sludges: holding times, temperatures, linearity of response, and sensitivity to pre-coagulation cell titers

Abstract

ELISA assays are a potential tool to screen for dissolved or cell bound microcystins in drinking water treatment sludges. In order to evaluate this potential more thoroughly, experiments were performed in alum sludges to: (1) evaluate the impacts of sample storage times, temperatures, and sludge composition on spiked microcystin-LR recovery by ELISA; (2) examine the linearity of ELISA responses to spiked microcystin-LR as a function of sludge composition; and (3) examine the sensitivity ELISA and LC/MS/MS to five different concentrations of microcystin-producing cyanobacteria entrained in sludges of two different compositions. During storage experiments, microcystin recovery efficiencies ranged from 85% to 125% across the range of 12 storage time and temperature combinations with recovery efficiencies in 7 of the 12 combinations falling into the 90% to 110% range. During the linearity experiments, linear models fit ELISA responses in all sludge compositions with R² values ≥ 0.95. During the sensitivity studies, simple freeze/thaw/centrifugation processing combined with ELISA or LC/MS/MS analyses resulted in detection of microcystins in thickened sludges derived from pre-coagulation cell suspensions of 10²–10⁶ cells/mL. In addition, the relationships between toxin concentrations in sludges and the original cell titers were linear regardless of analytical method.

Graphical Abstract:

Introduction

Characterizing cyanobacterial toxin concentrations in sludges is important for any conventional drinking water facility that treats bloom-impacted water. Because toxin-producing cyanobacteria propagate through treatment as particulates, they will accumulate and concentrate in the clarifier sludge along with other microorganisms, natural organic matter (NOM), and inorganic particles. Once entrained in the sludge, cyanobacteria have been shown to increase in number [1] and release toxins into solution [2–8]. These cells and their corresponding toxins, depending on how the sludge is managed, have the potential to enter the main flow stream of the treatment facility. As a result, drinking water treatment practitioners could benefit from a simple and inexpensive toxin quantification protocol that can be applied in sludges with a variety of compositions.

Increases in cyanobacterial cell numbers or toxin concentrations during sludge entrainment have been documented by several researchers. Jalili et al. [1] documented increases in the total number of cyanobacterial cells in stored polymer-based sludge in four of eight sampling events, with increases of up to 96% after 16 days of storage. Li et al. [2] demonstrated that extracellular cylindrospermopsin (CYL) concentrations increased by approximately 2 μg/L after six days of cell storage in polyaluminum ferric chloride (PAFC) sludges. Dreyfus et al [3], working with Al₂(SO₄)₃ (alum) sludges, measured microcystin (MC) and CYL concentrations in sludge supernatant after six days that were 2.2 and 2.8 times higher, respectively, than would have been predicted based on pre-coagulation intracellular toxin concentrations. Pestana et al. [4], also working with alum sludges, observed increases in extracellular MC concentrations after 20 days to levels that were twice as high as would have been predicted based on pre-coagulation intracellular toxin concentrations. In both of these cases, the authors attributed the observed increases to a combination of cell multiplication, toxin production, and toxin release while cyanobacterial cells were entrained in sludges. Xu et al [5], working with AlCl₃, FeCl₃ (ferric), and PAFC coagulants, showed that extracellular MC concentrations in sludges increased by approximately 10–40 μg/L over four to six days in storage. Li et al [6] observed a 4 μg/L increase in extracellular MC concentrations in (ferric) sludges after eight days of storage. Sun et al [7] measured 25 and 8 μg/L increases in polyaluminum chloride (PACL) and AlCl₃ sludges after 10 and 12 days of holding time, respectively. Ho et al. [9], working with alum sludges, observed extracellular toxin increases of up to 90 μg/L MC after 22 days and 25 μg/L extracellular CYL after 12 days. Finally, Ho et al. [10] reported increases in extracellular saxitoxins (STX) and CYL of 0.9–1 μg/L and 6–7 μg/L, respectively, after 7 days of hold time in alum and ferric sludges.

Of the nine studies discussed above, six used enzyme-linked immunosorbent assays (ELISA) for toxin quantification, two employed liquid chromatography coupled with a photo-diode array detector (LC-PDA), and one employed liquid chromatography coupled with tandem mass spectrometry (LC/MS/MS). Of these three analytical methods, the ELISA is financially and technically the most accessible and, for microcystins, can detect a broader range of congeners. Because of its lower costs and relative ease of use, ELISAs have the potential to be broadly adopted as a screening tool to identify sludge samples in need of follow-up analysis by liquid chromatography methods. The broader adoption of ELISAs for this purpose might be supported by the availability of basic analytical quality control data developed specifically for the assay. These data could include the impacts on toxin recovery efficiency of sludge sample holding times and temperatures, the linearity of assay response in sludges, and assay sensitivity to the titer of toxin-producing cells entrained in sludges. To the best of the authors’ knowledge, there are no publications presenting ELISA-specific data for these parameters in sludge matrices. In order to fill this gap, a study was undertaken with sludges of varying particulate and NOM concentrations to: (1) evaluate the impact of storage times and temperatures on microcystin-LR (MC-LR) recovery efficiency by ELISA; (2) examine the linearity of ELISA responses in sludges spiked with 0.5–4 μg/L MC-LR; (3) examine the ability of the ELISA and LC/MS/MS to detect MCs in sludges generated from waters spiked with 10²–10⁶ cells/mL of toxin producing Microcystis aeruginosa. Microcystins and cells of microcystin-producing M. aeruginosa were used for this study because microcystin is the most commonly detected cyanobacterial toxin world-wide [11].

Methods and materials

Water composition

Experiments were conducted with artificial water (AW) and water sampled from the Ohio River (ORW). The AW was based on a formula developed by Bajracharya et al [12] with the following modifications: NaN₃ was not included and NaNO₃ was substituted for NH₄NO₃. The remaining constituents followed the formula developed by Bajracharya, using the following ACS-grade reagents in de-ionized (DI) water: NaHCO₃ (241 mg/L), CaCl₂ · 2H₂O (20 mg/L), CaSO₄ · 2H₂O (240 mg/L), MgSO₄ · 7H₂0 (300 mg/L), KCl (5.0 mg/L), NaNO₃ (0.3 mg/L), FeCl₃ · 6H₂O (2.0 mg/L), and Na₂PO₄ (0.7 mg/L). The water was allowed to mix overnight prior to testing. The AW turbidity was adjusted to the target value by adding a suspension of acid washed diatomaceous earth (DE) (MP Biomedicals, Solon, OH, USA) in DI water. Lyophilized Ohio River water natural organic matter (ORW NOM)[13] was used to make a NOM solution, which was in turn used to adjust AW total organic carbon (TOC) concentrations. The lyophilized ORW NOM was reconstituted by adding 1.76 g lyophilized NOM to 175 mL milli-Q (MQ) water, mixing for 10 minutes, adjusting the pH to 10 with NaOH, mixing overnight, and then vacuum filtering through a 0.45 μm mixed cellulose membrane filter (Millipore, Bedford, MA, USA). The solution was stored in the dark at 2–5°C. The total organic carbon (TOC) concentration of the ORW NOM solution was 1.75 mg/mL.

Sludges for the holding time and linearity of response experiments were produced using three different AW variations and ORW. The three AW variations had turbidity and TOC values of 0.5 NTU and 0 mg/L, 2.6 NTU and 2.7 mg/L, and 2.4 NTU and 9.5 mg/L, respectively. The ORW had a turbidity of 120 NTU and a TOC concentration of 5.4 mg/L. For the three AWs, the optimum coagulant doses were determined to be 35, 35, and 80 mg/L, respectively. For ORW, the optimum dose was 40 mg/L. The whole cell sensitivity trials were carried out with one AW and a batch of ORW collected during a second sampling event. The AW turbidity was 2.6 NTU, the TOC was 2.7 mg/L, and the optimum coagulant dose was determined to be 35 mg/L. The second batch of ORW collected for the sensitivity experiments was determined to have a turbidity of 51 NTU, a TOC of 3.5 mg/L, and an optimum coagulant dose of 40 mg/L.

Jar testing parameters and coagulant dose determination

A six-position jar test stirring machine (Phipps & Bird, Richmond, VA, USA) was used throughout the study to produce sludges. While the jar tester was set to 30 rpm, DE suspensions and NOM solutions were added to achieve target turbidities and TOC concentrations. The mixing speed was then increased to 100 RPM for coagulant addition and left to mix for 90 s. For flocculation, the jar tester speed was set at 30, 20, and 10 rpm for 10 minutes at each speed. The tester was then turned off and the produced floc was left to settle for one hour prior to collection and thickening. The coagulant dosing solution consisted of 15 mg/mL Al₂(SO₄) • 14H₂O (GFS, Columbus OH) in DI water. Initial preparatory jar test trials were conducted for all water conditions to determine the optimum alum dose, defined as the minimum dose needed to reach a stable low plateau on a coagulant dose versus settled turbidity curve. Alum was chosen as the coagulant for this work because it is one of the most commonly used coagulants in North American treatment practice.

Culture, counting, and preparation of cyanobacterial cells

The Microcystis aeruginosa used in this study was an environmentally derived non axenic in-house strain that was propagated in BG-11 growth medium (Sigma, St. Louis, MO, USA) at 20°C in 250 mL De Long flasks (Bellco, Vineland NJ, USA). The toxin composition of these cells has been established as > 99% microcystin-LR (MC-LR) (> 99%) [14] Illumination was provided by muted fluorescent lamps (Phillips – Natural Light, Somerset NJ, USA) on 12-hour diurnal cycles at a light-phase illumination intensity totalling 5.4 μmol photons/m²x s across the 400–700 nm photosynthetically active radiation wavelength range (LI-190R, LI-COR, Lincoln NE, USA). Growth flasks were gently agitated by hand weekly.

Cells were harvested during the late stationary phase and were centrifuged and rinsed three times in filtered AW or ORW, depending on the experiment. Centrifugation cycles were performed for 10 minutes at 1000 × G relative centrifugal force (RCF) (Thermo-Jouan, Waltham MA, USA). This relatively low RCF and short time interval were chosen in order to minimize stress on the cells during centrifugation. Following the final rinse, cells were resuspended in 20 mL of filtered AW or ORW. The cell titers of the stock suspensions were determined by flooding both wells on a hemacytometer (Hausser, Horsham PA) and counting at 400X magnification (Zeiss Axioskop, Zeiss, Thornwood NY). Counting quantified all cells that fluoresced red under green epi-fluorescent illumination using a 545 nm excitation/562 nm dichroic/570 nm emission optical filter set (Chroma, Bellows Falls VT). These estimates of cell titers were used to determine the volumes of spiking suspension needed to achieve target final concentrations during subsequent experiments.

Holding time and temperature experiments

MC-LR recoveries from four different sludges were evaluated at three different holding/ time/temperature combinations for each sludge: 24 hours at 3° C, 7 days at − 20° C, and 21 days at − 20° C. The 3° C and − 21°C temperatures were chosen because they are, respectively, common set points for general purpose laboratory refrigerators and freezers sold in North America. The 24-hour, 7-day, and 21-day holding times were chosen because they are, based on the authors’ personal experiences and communications, common holding times employed while analytical runs are being prepared and additional samples are being accumulated. Following the determination of the optimum alum dose, sludges for the holding time and linearity studies were produced by filling three to six 2 L glass beakers with 1.5 L of a given water, coagulating at the optimum dose, and then collecting the sludge after sedimentation. Following sedimentation, the sludges were collected with a 25 mL serological pipette from all jars and combined in a separatory funnel where the sludge was allowed to thicken for 30–45 minutes (Supplemental Figure S1). The thickened sludge was sampled from the bottom of the separatory funnel into a graduated cylinder for volume measurement. The sludge was then divided into seven 5 mL aliquots and added to each of seven glass centrifuge tubes (Corning, Tewksbury MA, USA) for each holding time and temperature combination for a total of 21 tubes. MC-LR spiking solution in methanol, derived from certified reference material (NRC Canada, Ottawa Ontario, Canada) was then added to each tube to yield a final toxin concentration of 1 μg/L. The MC-LR concentration of the spiking solution was adjusted to ensure a final methanol:water ratio < 1%. Samples were held at the indicated temperatures for the prescribed duration. At the appropriate time, seven replicate aliquots were pulled from storage, equilibrated to room temperature and then centrifuged for 15 minutes at 2900 × G. A 1 mL aliquot of supernatant was then removed from each tube and transferred to a 4 mL labelled glass vial (Fisher Scientific, Hampton NH) for ELISA analysis.

Linearity of response experiments

Linearity of response experiments were conducted with the same four waters used for the holding time/temperature trials. Thickened sludge was dispensed into a graduated cylinder and separated into 5 mL aliquots, which were transferred to each of ten glass centrifuge tubes (Corning, Tewksbury MA, USA). MC-LR spiking solution was then added to 2 tubes each to achieve final toxin concentrations of 0.5, 1, 2, 3, and 4 μg/L. This concentration progression was chosen because all samples could be run on ELISA and LC/MS/MS calibration curves without dilution. The spiking solution was the same concentration as that used for the holding time study and the final methanol:water ratio was ≤1%. Each concentration was prepared in duplicate. The samples were then centrifuged (2900 × G for 15 minutes) and 1 mL of supernatant was transferred to 4 mL glass vials (Fisher Scientific, Hampton, NH, USA) for analysis by ELISA. Linearity of response samples were analyzed the same day that they were spiked with MC-LR.

Whole cell sensitivity of response experiments

These experiments were conducted with two instead of four water types, as detailed in the water composition section. For each of the two water types, separate jar tests were run at each target titer of 10², 10³, 10⁴, 10⁵ or 10⁶ cells/mL, for a total of five jar tests per water. For a given water and target cell titer, each jar test was run with three jars. Following distribution of water to the jars, the tester speed was set to 30 rpm and each jar was spiked with the pre-counted suspension of cyanobacteria sufficient to achieve the final target titer set for that particular experiment. Initial duplicate extracellular (EX) and total (extra + intracellular; TOT) toxin samples were collected from each jar immediately after the cell suspension had dispersed in the jars. Following the initial sampling, the jar tester speed was increased to 100 rpm and identical coagulant doses were added to each jar. The flocculation and settling process described earlier was then followed to completion. Duplicate EX and TOT toxin samples were also collected from the supernatant in each jar after sedimentation was complete. All sludge was collected from each jar and combined into a single separatory funnel for thickening. The sludge collection and thickening procedures were the same as in the holding time/temperature and linearity experiments. Duplicate EX and TOT toxin samples were also collected from the thickened sludge. The TOT toxin samples underwent three freeze/thaw cycles between − 20/+20°C to lyse cyanobacterial cells. Following the final thaw, samples were allowed to reach room temperature, centrifuged (2900 × G for 15 minutes), and approximately 3 mL of supernatant was transferred to a 4 mL glass vial. This vial served as the sample source for both ELISA and LC/MS/MS analysis. The EX toxin samples were filtered through a 0.7 μm nominal pore size glass fibre filter (Whatman GF/F, GE-Whatman, Boston MA, USA) at the time of collection and the filtrate was transferred to glass vials for ELISA analysis.

A follow-up set of experiments was conducted to examine the relationship between pre-coagulation cell titer and phycocyanin fluorescence. Cells were harvested, washed, and quantified using the same procedures outlined in the previous paragraph. Cells were then spiked into the same AW and ORW at the same target titers used for the previous study. The resulting in vivo phycocyanin fluorescence was measured as described in the analytical methods section.

Analytical methods

All materials used to contain, store, or transfer microcystin samples were glass. The only exceptions were the 25 mL serological pipettes used to transfer sludge from jar test beakers into a separatory funnel for thickening and the polypropylene pipet tips used to load the ELISA plates.

Cyanobacterial toxins in all three sets of experiments were quantified using the Microcystin/Nodularin – ADDA ELISA assay (Eurofins/Abraxis, Warminster PA, USA). The assay was run on 96-well plates according to the manufacturer’s protocol and the manufacturer’s calibration solutions were used to develop the calibration curves. The calibration curve extended from 0.15–5 μg/L. Each calibration, quality assurance, and unknown sample was run in duplicate wells. The absorbances for all plates were measured at 450 nm on a well plate reader (Spectramax Paradigm, Molecular Devices, San Jose CA). In addition to the manufacturer’s blank and calibration verification sample, the following extra quality assurance (QA) samples were included on each plate: a low-range calibration verification sample, a laboratory fortified blank (LFB), duplicate laboratory fortified matrix spikes (LFMS), and duplicate blanks. The extra microcystin-LR (MC-LR) quality control spikes were prepared in methanol using certified reference material (NRC Canada). The concentrations of the spiking solutions were adjusted so that the final methanol:water ratio in the samples loaded on to the plates was ≤1%.

For the whole cell sensitivity experiments, TOT MC-LR in the thickened sludges was also quantified by LC/MS/MS using the isotope dilution method. The internal standard was ¹⁵N₁₀ MC-LR (Cambridge Isotope Laboratories). All calibration, QA, and unknown samples were 90% aqueous/10% methanol. Calibration standards were prepared using certified reference material (NRC Canada, Ottawa Canada). The calibration curve extended from 0.01–5 μg/L. Analyses were performed on a Thermo Scientific Vanquish Flex liquid chromatography system coupled to two Thermo Scientific TSQ Quantis triple quadrupole mass spectrometers. Online solid phase extraction (SPE) was performed on a Hypersil Gold AQ 2 × 20 mm column with MQ + 0.1% formic acid at 1000 μL/minute. Chromatography was performed on a Waters Acuity UPLC HSS T3 1.8 μm 2.1 × 100 mm analytical column using a MQ:methanol:formic acid gradient that transitioned from 95:5:0.1–5:95:0.1 over 4 minutes at 500 μL/minute. The mass spectrometer was operated at a spray voltage of 3 kV, vaporizer and capillary temperature of 350°C, and sheath gas flow of 50 arbitrary units. Selected reaction monitoring used argon as the collision gas and MC-LR was measured as the parent M + H → 135 M/Z fragment corresponding to the ADDA moiety.

Turbidity was measured on a Hach TU5200 turbidimeter (Hach, Loveland, CO, USA). TOC was measured by combustion catalytic oxidation according to USEPA Method 415.3 [15]. Phycocyanin fluorescence of cell suspension was measured using a handheld fluorometer (Cyanofluor Turner Designs, San Jose, CA, USA). All phycocyanin fluorescence measurements were blank adjusted by subtracting the fluorescence of a filtered AW or ORW sample that did not contain cyanobacterial cells. All statistical analyses were carried out using Sigma-Plot software (Version 15.0, Inpixon, Palo Alto CA).

Results and discussion

Holding time and temperature experiments

The goal of the holding time/temperature experiments was to examine MC-LR recovery efficiencies at holding times that would plausibly be used in water treatment practice and at temperatures that can be maintained in commonly available cold storage equipment. The examined holding conditions were 24 hours at 2 °C, 7 days at − 20 °C, and 21 days at − 20 °C. The results of these experiments are summarized in Figure 1 and the experimental design groupings are summarized in Supplemental Table S1. Observed recovery efficiencies of the spiked MC-LR in all sludge composition and holding time/temperature combinations ranged from 85% to 125% with recoveries in 7 of the 12 combinations falling into the 90% to 110% range. To put these results into perspective, the US Environmental Protection Agency’s ELISA method for microcystins (EPA Method 546; [16]) specifies recovery efficiency limits of ± 30% for microcystins spiked into laboratory clean water (e.g. milli-Q) and ± 40% for microcystins spiked into sample matrices (e.g. ORW). All of the observed recovery efficiencies fell within the Method 546 limits for both laboratory water and sample matrices.

Figure 1.

Holding time and temperature results. For each sludge composition, holding time, and temperature combination, seven replicate samples were spiked with 1ug/L of MC-LR. Bars and error bars represent mean recovery efficiencies (n = 7) and 95% confidence intervals around the means, respectively.

Because the 24-hour hold time at 2 °C was a unique combination within the experimental design (Supplemental Table S1), the impact of all four pre-coagulation water types, three AWs along with ORW, on MC-LR recovery efficiencies under these conditions was examined with a one-way analysis of variance (ANOVA). With an observed p = 0.757, the analysis did not indicate that pre-coagulation water quality was a statistically significant source of variation. The detailed results are presented in the Supplemental Materials.

The impacts of pre-coagulation water quality and seven – versus 21-day holding times at − 20 °C on MC-LR recovery efficiencies were evaluated with a two-way ANOVA combined with a Holm-Sidak multiple comparison of differences. The detailed ANOVA and comparison of differences results are presented in the Supplemental Materials and in Supplemental Tables S2 and S3. Pre-coagulation water quality and the interaction between pre-coagulation water quality and holding time were found to be statistically significant sources of variation, with p values of 0.003 and < 0.001, respectively. Within each of the four pre-coagulation water qualities, the differences between MC-LR recovery efficiencies as a function of hold time were statistically significant (p ≤ 0.005). For the seven-day holding time results, four of six differences in recovery efficiencies as a function of pre-coagulation water quality were statistically significant (p ≤ 0.033). For the 21-day holding time results, all differences between recovery efficiencies as a function of pre-coagulation water quality were statistically significant (p ≤ 0.017). Despite these significant differences, it is difficult to infer a physical explanation once the directions of the differences are considered, with no apparent trends as function of pre-coagulation water quality. For example, at − 20 °C, 21-day recovery efficiencies were lower than 7-day recovery efficiencies at pre-coagulation TOC values of 0 and 2.7 mgL. However, the opposite effect was observed in sludges derived from waters with pre-coagulation TOC values of 9.5 and 5.4 mg/L.

In light of these results, the primary practical conclusion is that the variability in recovery efficiencies was lower in samples stored overnight in the refrigerator versus samples stored for 7 and 21 days in the freezer. However, even the samples stored 7 and 21 days at − 20 °C exhibited recovery efficiencies that fell within widely published benchmarks such as EPA Method 546. Viewed in a broader context, the results of the holding time and temperature experiments indicated that treatment plant operators and chemists might have significant flexibility with respect to sample holding times. This flexibility in turn increases the probability of running full instead of partial ELISA plates, thus maximizing the price advantages associated with the method.

Linearity of response experiments

The results of the linearity of response experiments for all four waters are shown in Figure 2 and the associated regression statistics are summarized in Table 1. Based on visual analyses, regression R² values, and F-test values, a linear model is a satisfactory fit in all four cases. Variations were observed among the four estimated slope coefficients, with the largest to smallest values differing by more than a factor of 2. There was no discernible association between the water quality prior to coagulant addition and the value of the estimated slope coefficients.

Figure 2.

Linearity of response results. MC-LR was spiked into thickened sludge at 0.5, 1.0, 2.0, 3.0, and 4.0 μg/L. Data points and error bars represent the means of duplicate spiked samples and average deviations about the means, respectively. Each panel represents a different water quality prior to coagulant addition.

Table 1.

Summary of regression statistics for linearity of response experiments. MC-LR spiked into thickened sludge at 0.5, 1.0, 2.0, 3.0, and 4.0 μg/L.

Water quality prior to coagulant addition (alum dose)	R2	F-statistic (p value)	Slope coefficient (p value)
AW, TOC = 0 mg/L, turb. = 0.5 NTU (alum = 35 mg/L)	0.95	140 (2.4 × 10−6)	0.74 (2.4 × 10−6)
AW, TOC = 2.7 mg/L, turb. = 2.6 NTU* (alum = 35 mg/L)	0.99	480 (5.8 × 10−7)	1.6 (5.6 × 10−7)
AW, TOC = 9.5 mg/L, turb. = 2.4 NTU (alum = 80 mg/L)	0.96	200 (5.6 × 10−7)	0.85 (5.6 × 10−7)
ORW, TOC = 5.4 mg/L, turb. = 120 NTU* (alum = 40 mg/L)	0.95	120 (3.6 × 10−5)	1.5 (3.6 × 10−5)

^*Regression run on data points from 0.5–3.0 ug/L spikes because 4 ug/L spike results were > highest ELISA calibration standard.

The shape of the non-linear ELISA calibration curve combined with between-plate variability may have contributed to the observed variations in the slopes of the sludge dose–response curves. The analyses for each of the sludge samples to assess linearity were run on a single ELISA plate, and the four calibration curves for these plates are shown in Figure 3. These curves relate MC concentrations (x-axis) to blank-normalized absorbances (y-axis), where the magnitude of the normalized absorbance is inversely proportional to the MC concentration. The curves display a shape that is characteristic of calibration curves for the ELISA kits used in this study, with a relatively steep slope at concentrations ≤ 1 μg/L and transitioning to shallower slopes at concentrations > 1 μg/L. The curves also exhibit between-plate variability, a situation that has been observed in other kits completed in this laboratory (Supplemental Figure S2) and remarked on by non-related researchers using the same kits [17]. One impact of between-plate variability is variation in the estimated MC concentration for a given normalized absorbance. This variability can be magnified at lower relative absorbances, which correspond to higher MC concentrations and the shallow slope regions of the calibration curves. For the four calibration curves developed in the sludge dose–response experiments (Figure 3), the MC concentrations interpolated from normalized absorbances at 0.6 and 0.3 ranged from 0.4–0.6 and 1.7–3.7 μg/L, respectively. The resulting coefficients of variation (mean/std. dev) for MC concentrations calculated at normalized absorbances of 0.6 and 0.3 were 0.15 and 0.27, respectively. A similar progression was observed for the curves sampled from other work in this laboratory (Supplemental Figure S2), where MC concentrations ranged from 0.5–0.6 and 2.5-and 4.5 μg/L at normalized absorbances of 0.6 and 0.3, respectively; CVs increased from 0.080–0.24 at the respective normalized absorbances.

Figure 3.

Impacts of ELISA calibration curve non-linearity and between-plate variability on calculated MC concentrations during the sludge linearity of response experiments. Each curve represents the calibration curve for one ELISA plate run during one of the four linearity experiments. Variations in calculated MC concentrations are summarized by their coefficients of variation (CV) where CV = std. dev / mean. The CVs were calculated at blank-normalized absorbances of 0.6 and 0.3.

On the basis of these observations, it can be plausibly argued that variabilities in sludge dose–response curves may be at least partially attributable to differences between calibration curves developed during MC analyses. In the final assessment, the distribution of calibration curve shapes and ensuing MC concentration variabilities indicated the desirability for follow-up analyses by LC/MS/MS in situations where high degrees of accuracy and precision are required. However, the fact that the ELISA demonstrated linear responses across widely differing sludge matrices and across a near order of magnitude concentration range, indicates that the method holds promise as a tool to screen not only for the presence of MC in sludges but also broad changes in concentration. The MC-LR concentration range evaluated during these tests was representative of what might be seen in sludges during earlier stages of bloom development, when cell counts are low. As a result, the experiments described here need to be repeated across higher concentration ranges and with microcystins other than MC-LR in order to evaluate the full potential of the method.

Whole cell sensitivity of response experiments

A set of experiments was conducted to examine the relationships between spiked pre-coagulation titers of toxin-producing cyanobacterial cells and microcystin concentrations by ELISA and LC/MS/MS in thickened post-coagulation sludges. The experiments were run with the moderate NOM low turbidity AW (TOC = 2.6 mg/L, Turbidity = 2.6 NTU, alum dose = 40 mg/L) and ORW (TOC = 3.5 mg/L, turbidity = 51 NTU, alum dose = 40 mg/L). Sludges were produced in each water with starting cell titers of 10², 10³, 10⁴, 10⁵, and 10⁶ cells/mL, for a total of five different sludges per water type. The settled sludges were composed of floc, inorganic particulates, adsorbed and precipitated NOM, and entrained cyanobacterial cells. The results of these experiments are summarized in Figures 4 and 5 and Tables 2 and 3. The EX and TOT MC-LR concentrations prior to coagulant addition (rows 1 and 2 of Tables 2 and 3) indicate that the bulk of the toxin was contained inside the cells and was not leaking into solution. Note also that EX toxin was not detectable in any of the settled waters and only a remnant of TOT toxin was detected after settling in ORW at the highest initial cell titer. These results imply that effectively all of the measurable pre-coagulation toxin burden had transferred from solution or suspension to the thickened sludge.

Figure 4.

Sensitivity experiment results – artificial water. TOC = 2.7 mg/L, Turb. = 2.6 NTU, Alum dose = 40 mg/L. Toxin producing cyanobacterial cells were spiked into artificial water at five different concentrations in five separate jar tests. Phycocyanin fluorescence was measured in uncoagulated water. The coagulation sludge was collected, thickened, freeze/thaw processed, and analyzed for extra + intracellular microcystins by ELISA and extra + intracellular microcystin-LR by LC/MS/MS.

Figure 5.

Sensitivity experiment results – Ohio River water. TOC = 3.5 mg/L, Turbidity = 51 NTU, Alum dose = 40 mg/L. Toxin producing cyanobacterial cells were spiked into Ohio River water at five different concentrations in five separate jar tests. Phycocyanin fluorescence was measured in spiked but uncoagulated water. The coagulation sludge was collected, thickened, freeze/thaw processed, and analyzed for extra + intracellular microcystins by ELISA and extra + intracellular microcystin-LR by LC/MS/MS.

Table 2.

Microcystin concentrations in experiments correlating pre-coagulation cell titers in AW (TOC = 2.7 mg/L, turb. = 2.6 mg/L, Alum dose = 40 mg/L) with microcystin concentrations in thickened sludge.

	Target cell titer prior to coagulant addition (#/mL)
Sample point	106	105	104	103	102
EX prior to coagulant addition (Mean, n = 3, of all jars ± σn-1)	0.52 ± 0.1	0.18 ± 0.1	<0.15	<0.15	<0.15
TOT prior to coagulant addition (Mean, n = 3, of all jars ± σn-1)	33 ± 7	3.2 ± 0.28	0.57 ± 0.21	<0.15	<0.15
EX settled (Mean, n = 3, of all jars)	<0.15	<0.15	<0.15	<0.15	<0.15
TOT settled (Mean, n = 3, of all jars)	<0.15	<0.15	<0.15	<0.15	<0.15
EX thickened Sludge	36	2.1	0.20	<0.15	<0.15
TOT thickened Sludge	1500	170	32	3.3	0.16
TOT thickened sludge (by LC/MS/MS)	330	36	5.2	1.1	0.063

All microcystin concentrations in μg/L. All microcystins quantified by ELISA unless indicated otherwise. Quantification by LC/MS/MS for microcystin-LR only. EX = extracellular toxin concentration. TOT = extra + intracellular toxin concentration.

Table 3.

Microcystin concentration in experiments correlating pre-coagulation cell titers in Ohio River water (TOC = 3.5 mg/L, turb. = 51 NTU, alum dose = 40 mg/L) with microcystin concentrations in thickened sludge.

Ohio River Water Sample Point and Time	Target cell titer prior to coagulant addition (#/mL)
	106	105	104	103	102
EX prior to coagulant addition (Mean, n = 3, of all jars ± σn-1)	0.63 ± 0.3	<0.15	<0.15	<0.15	<0.15
TOT prior to coagulant addition (Mean, n = 3, of all jars ± σn-1)	39 ± 9.7	4.3 ± 0.35	0.59 ± 0.24	<0.15	<0.15
EX settled (Mean, n = 3, of all jars)	<0.15	<0.15	<0.15	<0.15	<0.15
TOT settled (Mean, n = 3, of all jars)	0.24	<0.15	<0.15	<0.15	<0.15
EX thickened Sludge	4.1	0.68	<0.15	<0.15	<0.15
TOT thickened Sludge	1970	322	25	4.2	0.65
TOT thickened sludge (by LC/MS/MS)	560	120	9.1	1.7	0.14

All microcystin concentrations in μg/L. All microcystins quantified by ELISA unless indicated otherwise. Quantification by LC/MS/MS for microcystin-LR only. EX = extracellular toxin concentration. TOT = extra + intracellular toxin concentration.

TOT MCs were detected at concentrations greater than the lowest ELISA and LC/MS/MS calibration standards (0.15 and 0.01 μg/L, respectively) in freeze/thaw processed thickened sludges at all pre-coagulation cell titers in both waters (Tables 2 and 3). The relationships between TOT MC by ELISA and TOT MC by LC/MS/MS in thickened sludges versus the pre-coagulation cell titers were linear across multiple orders of magnitude for both waters. The observed linearity during the whole cell experiments implies a high likelihood that controlled toxin spiking experiments across concentration ranges larger than those described in the previous section would also yield linear responses.

The observed concentration differences between ELISA and LC/MS/MS were due to the fact that the two methods represent different approaches to the quantification of cyanobacterial toxins [18]. LC/MS/MS methods are typically designed to be compound specific, selecting for mass transitions and retention times which are characteristic of a given chemical. ELISA methods rely on antigen–antibody interactions (which may be broadly occurring for entire classes of chemicals, e.g. microcystin congeners) and generation of a colorimetric indicator proportional in some manner to the concentration of antigen or antigens present in the sample. Both assays are susceptible to interferences to the analytical response, frequently in the form of ion suppression/enhancement in mass spectrometric methods, and interferences in the antibody response from matrix interferences either positively or negatively for ELISA. ELISA is also potentially able to produce responses to degraded or modified target compounds. The overall effect of sample matrix on ELISA response ratios can be seen in Figure 2, which shows the variation in response versus concentration in different water quality samples. Taken together, it isn’t unusual for the absolute magnitude of the response/concentration measured in a sample matrix to not be identical between the two measurement methods, particularly for a matrix as complex as sludge as in this case. The relative responses, on the other hand, should trend similarly, and this is seen in the data in Tables 2 and 3 and Figures 4 and 5 for the thickened sludge results by ELISA and LC/MS/MS.

The Spearman rank order correlation coefficients between sludge MC concentrations measured by the two analytical methods were equal to 1 (p < 0.05) in both waters. It is noteworthy that TOT MCs were detected in all thickened sludges, even at pre-coagulation cell titers of 10² and 10³ cells/mL; titers which were too low to yield detectable TOT MC concentrations prior to coagulant addition (Tables 2 and 3). The ability to detect toxins in sludges originating from even the lowest pre-coagulation cell titers is most likely due to the fact that coagulation and sludge thickening serve as particulate concentration mechanisms that act on cyanobacterial cells as well as on non-biological particulates. The observed concentration effect implies that regular screening for toxins in thickened sludge can provide indication of bloom formation even when toxin concentrations in untreated water are still too low to detect.

A set of follow-up experiments were conducted to examine the relationship between pre-coagulation cell titers and the phycocyanin fluorescence response of commercially available instrumentation for detecting cyanobacteria in suspension. These experiments were conducted in the same waters and at the same cell titers that were used for the cell titer versus thickened sludge toxin concentration experiments. The pre-coagulation cell titer versus fluorescence results are plotted in Figures 4 and 5 along with the pre-coagulation cell titer versus thickened sludge TOT MC concentrations. For the lower turbidity AW, fluorescence measurement was able to detect cyanobacteria down to the lowest pre-coagulation titer of 10² cells/mL, the same concentration that yielded the lowest detectable TOT MC-LR measurement in thickened sludge. For the higher turbidity ORW, cyanobacteria were detected by fluorescence down to a pre-coagulation titer of 10³ cells/mL, an order of magnitude higher than for the lower turbidity water. The decreased sensitivity of fluorescence measurement in the higher turbidity water was probably due to increased particulate interference with the excitation light, emission light, or both.

Overall, the fluorescence and toxin concentration results indicate that the two methods can serve complementary roles in the operation of a treatment facility that is contending with HABs-impacted waters. The fluorescence instrumentation lends itself to the real-time detection of cyanobacteria in plant influents at low cell titers, especially in lower turbidity waters. However, these concentrations may be too low to permit the detection of toxins at the same sample point by ELISA. To address this information gap, concurrent analyses of toxin concentrations in freeze/thaw treated sludge could provide insight into whether or not the cyanobacteria entering the treatment facility are toxin producers. Both methods, if employed together, could start generating actionable information at approximately the same early stage of the bloom season, when cyanobacterial biovolume concentrations are well below the lowest vigilance threshold recommended by the World Health Organization (WHO) for recreational waters, thereby allowing ample time for drinking water providers to prepare for higher concentrations. The WHO recommends the initiation of vigilance at cyanobacterial biovolume concentrations of 1–4 mm³/L [19]. Using conversion factors from Padisák et al. [20], a cell titer of 10² M. aeruginosa cells/mL, the lowest used in the current study, is equivalent to biovolume concentrations of 0.0034–0.011 mm³/L, so several orders of magnitude below the WHO thresholds (see sample calculation in the Supplemental Materials)

The results obtained during this study demonstrate that the ELISA assay combined with a simple sample processing protocol can, within the parameters explored for this work, generate acceptable responses and recoveries across a range of toxin concentrations, sludge matrices, and sample storage conditions. These are useful insights in light of the work referenced in the Introduction. Investigators in those studies demonstrated that cyanobacterial cells entrained in sludge can proliferate [1] and release toxins [2–8]. Given that cyanobacterial blooms and the concomitant potential for cell entrainment in sludges are widespread, there exists a need for precise, accurate, sensitive, and economically implementable methods for toxin detection in sludges. With these methods in hand, treatment plant managers can screen for early season bloom activity, optimize the timing of sludge withdrawals from sedimentation basins, and characterize toxin concentrations in sludges destined for land application or sewer disposal.

Conclusion

The results quantitatively support the proposition that an ELISA assay can be an effective and economical screening tool for the analyses of microcystins in alum drinking water treatment sludges. The holding time and temperature data indicated that acceptable recovery efficiencies of microcystin-LR were possible from thickened alum sludges derived from waters with broad ranges of initial TOCs and turbidities. These recoveries were observed following storage in the refrigerator (4° C) and freezer (−20° C) at holding times ranging from 24 hours to 21 days, implying that microcystin-laden alum sludge samples do not require expensive − 80° C storage over the investigated time intervals. The ability to store samples confidently and inexpensively eases the assembly of full ELISA plates, thus minimizing the cost of performing the assay.

The ELISA assay also exhibited a linear response over a near order of magnitude concentration range for MC-LR spiked into thickened alum sludges. These sludges were derived from the same source waters used for the holding time and temperature experiments. These linear responses implied that the ELISA assay can reliably detect changes in alum sludge toxin concentrations, an important consideration early in the bloom season when treatment plant managers are looking for clues that cyanobacterial biomass levels are increasing.

Finally, experiments with suspensions of toxin-producing M. aeruginosa demonstrated that the ELISA assay in conjunction with simple freeze/thaw/centrifugation processing detected the presence of cell-derived microcystins in thickened sludge produced from waters with cell titers as low as 10²/mL. This combination of sample processing and assay also demonstrated a linear response over a 4 order of magnitude pre-coagulation cell concentration range, from 10² to 10⁶ cells/mL. This linear response was confirmed by analyzing processed sludge samples with LC/MS/MS. The broader implication of the whole cell results is that a technically and economically accessible processing and analytical method can help provide early notice of changes in the concentration of toxin-producing cyanobacteria in the treatment plant influent.

While the ELISA results presented here are encouraging, they don’t eliminate the need for follow-up analyses by LC/MS/MS in situations that require greater levels of accuracy and precision. The data also indicate multiple potential areas for follow-up research. These include evaluating recoveries of different microcystin congeners, along with cylindrospermopsins and anatoxins; examining the whole-cell processing/assay response when environmental instead of cultured cyanobacteria are coagulated; and evaluating the impact on recovery efficiency of powdered activated carbon entrained in sludges.

Data availability statement

All data in this manuscript will be made publicly available through EPA Science Inventory (https://cfpub.epa.gov/si/)