The Fate of the Cyanotoxin Dihydroanatoxin-a in Drinking Water Treatment Processes

Abstract

Only recently has the cyanotoxin dihydroanatoxin-a (dhATX-a) been detected more frequently in different surface waters, some of which are used for supplying drinking water. As data about the fate of dhATX-a in drinking water treatment processes are still scarce, the present study investigated the behavior of dhATX-a in different water treatment steps: slow sand filtration, flocculation, adsorption onto activated carbon, ozonation and chlorination. The almost complete removal (>95%) of dhATX-a was observed in sand columns simulating slow sand filtration without showing a long adaptation phase. The results further indicate that dhATX-a can be removed using powdered activated carbon at dosages of 50 mg/L with removal rates between 75 and 93% and also by using ozonation with dosages above 1 mg/L at a concentration of ca. 4.5 mg/L background organic carbon. In contrast, no elimination of dhATX-a was observed in flocculation and chlorination experiments.

1. Introduction

Cyanobacteria occur worldwide in surface waters and are a natural part of the phytoplankton community. Under eutrophic conditions, planktonic (i.e., suspended) cyanobacteria can proliferate massively, causing negative impacts within the ecosystem, especially due to the ability of some cyanobacteria to produce potent toxins [1]. These so-called cyanotoxins can be classified largely into hepatotoxins and neurotoxins, with intoxications being documented for wild and domestic animals, but rarely also for humans [2].

Anatoxins are one group of neurotoxins, with anatoxin-a (ATX-a) being the most studied congener so far [3]. Besides ATX-a, other congeners include homoanatoxin-a (HATX-a) as well as their dihydro and epoxy congeners [4]. While previous studies suggest that the dihydro and epoxy congeners are degradation products [5,6], more recent studies showed that dihydroanatoxin-a (dhATX-a) is produced by the same biosynthetical pathway as ATX-a [4]. Anatoxins are produced by representatives of the planktonic genera Dolichospermum and Aphanizomenon, and by benthic cyanobacteria such as Microcoleus, Phormidium, Tychonema, Kamptonema, Oscillatoria and Cylindrospermum [4]. Benthic cyanobacteria are attached to surfaces such as sediments and occur—in contrast to planktonic cyanobacteria—largely in clear, low-nutrient waterbodies. Here, they can form extended mats containing high contents of anatoxins, which have been the cause of dog intoxications in the past. Many poisoning episodes were attributed to the presence of ATX-a and/or HATX-a producing benthic cyanobacteria, but recently, benthic cyanobacteria producing mostly dhATX-a in high amounts were often reported to have caused dog fatalities, e.g., in Lake Tegel and in the Mandichosee reservoir, which otherwise have good water quality (Germany) [7]. As clear waterbodies are preferential bathing waters and the source of drinking water production, risk assessment and management are needed to protect human health from anatoxins, including the evaluation of treatment processes to remove anatoxins.

Several studies have evaluated treatment options for their potential to remove cyanotoxins such as microcystins (MC), cylindrospermopsin (CYN), ATX-a, and saxitoxins (PSP) [8,9]. Flocculation is known for its potential to remove particles and is thus efficient in removing cell-bound toxins, but dissolved constituents such as extracellular toxins only to limited extents. Slow sand filtration can also remove cell-bound toxins but additionally has the potential to eliminate extracellular toxins by biological transformation.

While the removal of cell-bound toxins via the retention of cells is independent of the type of toxin, the removal or transformation of extracellular toxins is toxin-specific. Extracellular toxins can occur through the liberation of primarily cell-bound toxins during cell lysis such as MC, through release from actively growing cells as shown for CYN and ATX-a, but also during treatment processes such as pre-ozonation due to the damage of cells [10,11]. Several processes are suitable for removing extracellular toxins. Activated carbon removes most cyanotoxins very effectively, and removals of ATX-a and HATX-a using powdered activated carbon (PAC) were reported in ultra-pure and natural water [12,13,14]. Further, extracellular cyanotoxins can be removed by oxidation with ozone, chlorine and potassium permanganate with varying efficiencies depending on the toxin [15]. For the oxidation of ATX-a, ozone is most effective followed by potassium permanganate, while the oxidation of ATX-a with chlorine failed even with high dosages. Among the different cyanotoxins, ATX-a adsorbs the best to sand [16], and also the biodegradation of ATX-a is usually quite fast under various conditions [17,18].

To our knowledge, no information on the sorption and oxidation of dhATX-a during common treatment processes is available. This work thus aims to evaluate the elimination of dhATX-a in individual drinking water treatment processes, including flocculation, adsorption, ozonation, sand filtration and chlorination. The experiments were conducted separately under controlled laboratory conditions.

2. Materials and Methods

2.1. Materials

All experiments were conducted with tap water from Berlin with a dissolved organic carbon (DOC) concentration of 4.7 mg/L, which is in the magnitude of regional surface waters. Test solutions with dhATX-a were prepared by diluting combined field samples with a final concentration of 8780 µg/L dhATX-a in Berlin tap water to obtain target dhATX-a concentrations of 10 µg/L. Prior to analyses, all samples were filtered with pre-rinsed membranes (0.45 µm pore size, PVDF membrane, Macherey-Nagel, Düren, Germany).

2.2. Flocculation Experiments

Batch tests were conducted with a flocculation system (Flocculator 2000, Kemira, Helsinky, Finland) consisting of six glass beakers (1 L) and programmable stirrers. Ferric chloride (FeCl₃·6H₂O) was dissolved in ultra-pure water to obtain a stock solution. Experiments were initiated after dosing stock suspension with 1 min rapid mixing (100 rpm), and the stirring speed was reduced to ca. 30 rpm for 15 min. After a settling time of 30 min, samples were taken from the supernatant of the beakers and filtered with a membrane (0.45 µm pore size).

2.3. Adsorption Experiments

Adsorption tests were performed with PAC obtained by grinding three different granular activated carbons: AS (AquaSorb 5000, Jacobi, Kalmar, Sweden) based on lignite, HCR (HCR + 1, Carbon Service and Consulting) based on bituminous carbon and HCC (Hydraffin CC, Donau Carbon) based on coconut husk. The powders were dried at 105 °C for 24 h and stored in a desiccator before the PAC was added to ultrapure water to obtain 2 g/L PAC stock suspensions. The stock suspensions were mixed on a stirrer to achieve a homogenous suspension during dosages into samples. Subsequently, up to five PAC dosages were applied in batch experiments to achieve 2, 5, 10, 20 and 50 mg/L. The highest dosage is needed for poorly adsorbing substances, while the lower dosages are sufficient for well-adsorbing organic micropollutants [19]. After dosing PAC, the flasks were placed on a horizontal shaker to ensure complete mixing. After 30 min and 48 h contact time, samples were filtered with a membrane (0.45 µm pore size) to separate the PAC. The adsorption equilibrium was thought to be reached after 48 h [20].

2.4. Ozonation Experiments

Ozonation tests were carried out in a bench scale system, as described by Altmann et al. [20] in more detail, by bubbling ozone-containing oxygen into a turbulently stirred glass ozone reactor (4 L volume). The gas flow and ozone concentrations in the in-gas and off-gas are continuously quantified to balance the ozone input and the ozone consumption. Samples were taken before and after treatment. The ozone consumption is automatically calculated by mass balances of the total ozone input and output.

2.5. Sand Filtration Experiments

Two parallel columns (transparent PVC-U, 30 mm inner diameter, 310 mm length) were filled with a supporting layer of 79 g gravel and sand (467 g) obtained from a groundwater infiltration basin in Berlin. The calculated bulk density of the sand after drying at 105 °C was 2.39 g/mL.

As schematically shown in Figure 1, the columns were fed in up-flow mode with a peristaltic pump at a volume flow of 530 mL/d corresponding with a hydraulic residence time of approximately 6 h. Directly after preparation, the columns were operated with degassed tap water for several days to fully absorb potential residual air bubbles within the pores (the artificially contaminated tap water was not degassed). The columns were then operated with non-chlorinated tap water from Berlin for 14 days to equilibrate and then 11 days with dhATX-a-containing tap water. The filtration parameters were chosen in accordance with the conditions found at the field sites [21].

2.6. Chlorination Experiments

Chlorine concentrations in the sodium hypochlorite (NaClO) stock solution were quantified with the method proposed by Willson [22], and different hypochlorite dosages ranging between 0.1 and 1.0 mg/L were tested for a short contact time of 1 min and for longer contact times up to 30 min; after that, residual disinfectant was quenched by adding sodium thiosulfate solution. The dosages were chosen to represent typical disinfection processes [23]. All experiments were performed at ambient temperature (23 ± 2 °C).

2.7. Analyses

Concentrations of dhATX-a were quantified by high-performance liquid chromatography and with tandem mass spectrometry (HPLC-MS/MS) consisting of a 2900 Series HPLC (Agilent Technologies, Waldbronn, Germany) coupled to a 5500 QTrap mass spectrometer (AB Sciex, Framingham, MA, USA) with a turbo ion spray interface. An injection volume of 10 µL of the sample was separated on an Atlantis C18 column (2.1 mm, 150 mm, Waters, Eschborn, Germany) at 30 °C and a flow rate of 0.25 mL/min using a linear gradient of 0.1% formic acid (A) and 0.1% formic acid in methanol (B) [7]. The identification and quantification of dhATX-a were performed in the “multiple reaction monitoring” (MRM) mode using the characteristic transitions as described by [7]. Standards for calibration were purchased from Novakits (Nantes, France). The limit of detection for dhATX-a was 0.03 µg/L, and the limit of quantitation was 0.1 µg/L.

DOC was characterized by size exclusion chromatography with continuous UV and organic carbon detection (LC-OCD-UVD), as described in detail by Huber et al. [24]. The UV light absorption (UVA₂₅₄) was determined on a Perkin Elmer UV/VIS Spectrometer Lambda 12 UV-VIS photometer (Perkin Elmer, Germany) using quartz cuvettes (Suprasil, 10 mm, Hellma, Müllheim, Germany).

3. Results and Discussion

3.1. Flocculation

The residual dhATX-a concentrations after the addition of flocculant dosages of up to 25 mg/L in Figure 2 illustrate that flocculation with ferric chloride does not affect the dhATX-a concentrations. Even at a comparably high Fe³⁺ dosage of 25 mg/L, no significant removal was observed. Similar results were reported for ATX-a removal with flocculation [25]. Therefore, the flocculation step in many waterworks that treat surface waters does not represent a barrier towards dissolved dhATX-a at all. However, flocculation has been shown to remove largely cell-bound toxins such as MC and can thus be expected to also remove dhATX-containing cells [15].

3.2. Adsorption onto Activated Carbon

The removals of dhATX-a with different dosages of PAC AS are shown in Figure 3 for both a contact time of 30 min and equilibrium (48 h contact time). Even at equilibrium, a low removal was observed for comparably low dosages of up to 5 mg/L PAC. Significant reductions in dhATX-a concentrations were only observed when the PAC dose was increased to 20 mg/L for 48 h contact time. However, a certain percentage of dhATX-a remained even after treatment with a comparably high dosage of 50 mg/L PAC.

In contrast, the removal of the structurally related ATX-a by PAC showed higher removals of 60 and 90% with 5 mg/L and 11 mg/L PAC, respectively [26,27]. However, no information was given on the type of PAC, contact time and water matrix investigated, hampering the explanation for this observed difference between ATX-a and dhATX-a.

The adsorption behavior of dhATX-a is similar to that of the artificial sweetener acesulfame [19] that adsorbs to a much lower extent to conventional PAC compared to well-adsorbing anthropogenic organic micropollutants such as carbamazepine or diclofenac [28]. The results show that comparably high specific dosages of PAC AS (far more than 10 mg per mg DOC) are required to fully remove dhATX-a within short contact times of 30 min.

The comparison of the lignite-based PAC AS with the bituminous PAC HCR and the PAC HCC based on coconut husk indicates that PAC HCC with an elevated microporosity achieved greater and faster removals (Figure 4 (left)). The specific advantage of a microporous PAC for dhATX-a is in accordance with comparably higher specific removals of the comparably small molecule benzotriazole (119.13 g/mol) reported for a microporous PAC based on coconut shell compared to PAC based on lignite or bituminous coal [28]. The molecular mass of dhATX-a with 167.25 g/mol is lower than that of ubiquitous and well-investigated organic micropollutants such as carbamazepine, metoprolol or sulfamethoxazole, and thus dhATX-a might diffuse into the inner pores of the PAC HCC faster and more efficiently than competing adsorbates.

The dosage dependent abatement of UVA₂₅₄ indicates that the PAC AS and HCR achieve greater removals of background organic matter and thus are more amenable to adsorption competition. Due to the microporosity, PAC HCC provides some specific advantages for dhATX-a over other adsorbing water constituents. The preferred dhATX-a adsorption is also reflected in the relation between UVA₂₅₄ abatement and the removal of various target compounds reported by Zietzschmann et al. [29].

For the control of water treatments during potential releases of dhATX-a from cyanobacteria, a UVA₂₅₄ abatement of 40% indicates a dhATX-a removal of 80% for the PAC HCC, while for the other two PACs, UVA₂₅₄ abatements of 50% indicate 70% dhATX-a removals. This surrogate indication might be helpful, since HPLC-MS/MS analyses are much more time consuming and the results are not available in real-time. However, the DOC composition and related UVA254 during blooms might differ from the DOC in the test water of this study.

The corresponding changes in LC-OCD chromatograms in Figure 5 (left) revealed that the lowest residual DOC concentration after a dosage of 50 mg/L PAC and 30 min contact time was achieved with the PAC AS. Since adsorbed DOC constituents occupy adsorption capacity, high DOC removals are not beneficial but contribute to enhanced adsorption competition with dhATX-a. All constituents were similarly adsorbed with the highest relative removals of compounds detected between 62 and 65 min, of which [24] assigned to low molecular weight acids.

The LC-UVD chromatograms in Figure 5 (right) show lower relative concentrations after adsorptive treatment since DOC constituents with UV light absorbing moieties typically adsorb to a greater extent.

3.3. Ozonation

An ozone consumption of only 0.2 mg/L achieved 13% dhATX-a elimination, with 2.0 mg/L obtaining complete dhATX-a removal as shown in Figure 6. An ozone consumption of 2.0 mg/L at a DOC concentration of 4.7 mg/L corresponds to a specific ozone consumption of 0.43 mg ozone per mg DOC and thus dhATX-a belongs to the group of very easily oxidizable organic micropollutants such as carbamazepine. Comparably, the structurally related ATX-a was eliminated to more than 90% by ozone concentrations around 2 mg/L, with a higher efficiency at pH 8 and 9 [14,15]. However, in organic-rich waters, ozone consumption above 2 mg/L is required to achieve an elimination >90% [14].

The respective abatement of UVA₂₅₄ shown in Figure 6 indicates that increasing consumptions of ozone further oxidizes water constituents and thus dhATX-a is selectively oxidized. The relation of UVA₂₅₄ abatement and ozone consumption corresponds well with the previous data reported for wastewater treatment plant effluents [20].

The UVA₂₅₄ abatement was reported as surrogate parameter for micropollutant removal, which provides reliable predictions in real-time and is thus much faster than the time-consuming analyses by HPLC-MS/MS. Figure 6 shows that a UVA₂₅₄ abatement of 20% indicates a dhATX-a removal of 50% while full dhATX-a elimination can be expected for approximately 35% UVA₂₅₄ abatement. Thus, ozone consumption in a flow-through system can be easily adjusted to a UVA₂₅₄ abatement (or ΔUVA₂₅₄ [30]) of 30% to fully remove dhATX-a.

The LC-OCD and LC-UVD chromatograms of the influent and exemplary effluents displayed in Figure 7 show that DOC was only slightly removed while UVA₂₅₄ was abated to a much greater extent. Humic substances typically detected between ca. 45 and 57 min showed a slight shift towards longer detections times, probably due to a slight decrease in molecular weight. The small shoulder detected between 62 and 65 min was not eliminated by ozonation, while adsorption onto PAC revealed a significant removal (Figure 5 (left)). Thus, there is a significant ozone consumption by DOC but the DOC is not mineralized.

The LC-UVD chromatogram in Figure 7 (right) shows a clear UVA₂₅₄ abatement in all fractions of the chromatogram. Thus, the UV light-absorbing aromatic moieties and double bonds in all DOC size fractions are oxidized.

3.4. Slow Sand Filtration

On the first day of starting the experiment with dhATX-a-containing tap water as the influent, dhATX-a was fully eliminated in both sand columns, as shown in Figure 8. This elimination might be due to the fact that the sand probably had no contact with dhATX-a before and thus might have provided adsorption sites for it. The slightly decreasing dhATX-a elimination after one day probably indicates the beginning of an adsorption breakthrough since the adsorption capacity of sand is limited.

The microorganisms on the sand have been previously shown to be capable of transforming a number of organic micropollutants [31] but were not conditioned to dhATX-a transformation. However, the increasing dhATX-a elimination between day 1 and 8 indicates a quick adaptation of microorganisms to dhATX-a transformation that compensates for the decreasing adsorption. After eight days of operation, dhATX-a is again fully removed, most probably due to biological transformation. For ATX-a, one study also observed an adaptation phase of only 4 days, while longer adaptation phases of around two weeks up to even one year were reported for other cyanotoxins such as MC and CYN [15,32]. However, it has been shown for CYN and MC that adaptation phases can be reduced by the re-addition of the toxin [15,32]. To enable the reliable removal of a filter, regular contact with toxins is emphasized.

Despite the relatively short residence time of 6 h, full biological transformation was observed after a short adaptation time. Therefore, bank filtration and artificial groundwater infiltration with typically much longer residence times most probably represent safe barriers towards dhATX-a.

3.5. Chlorination

The chlorination experiments with dosages between 0.1 and 1.0 mg/L and contact times up to 30 min showed no substantial effect on the dhATX-a concentrations, as shown in Figure 9. Other studies on the similar compound ATX-a reported similar results due to slow reaction kinetics or the non-reactivity of chlorine with ATX-a [33,34].

The toxin dhATX-a does not substantially react with hypochlorite at relevant dosages or within relevant contact times. Thus, chlorination as a last step directly before drinking water enters the distribution network is not capable of reducing dhATX-a.

3.6. Comparison

The investigated treatment processes and elimination, ranging from no removals with flocculation and chlorination to full eliminations for ozonation and biological filtration, are summarized in

Table 1. A summary of elimination, dosages, approximate treatment times and the limitations of the different treatment processes.

Treatment	Elimination [%]	Dosage [mg/L]	Time	Limitations
flocculation	0	up to 50	ca. 20 min	ineffective for dissolved toxins
adsorption onto PAC	75–93	20–50	>30 min	PAC separation needed, competition with DOC
ozonation	>95	2 mg/L consumption	ca. 10 min	ozone consumption by DOC, oxidation products
slow sand filtration	>95	no	<6 h	adaptation time required
chlorination	0	up to 1		ineffective

, together with respective dosages, treatment durations and limitations. The processes may be combined efficiently to reduce contact times and efficiencies, for example, combining a dosage of PAC with deep bed filtration [35]. The residence time of 6 h in the sand columns was sufficient for full elimination. However, much shorter contact times might have been sufficient, but further investigations are needed to elucidate the minimum contact time. Chlorination is typically applied for disinfection at the end of treatment processes, and residual chlorine over long contact times is intended to extend its disinfecting capabilities within the drinking water distribution system.

The eliminations obtained with both adsorption onto PAC and ozonation are typically influenced by natural organic matter (DOC) that either occupies adsorption sites on the activated carbon or consumes ozone [20].

4. Conclusions

The removal of dhATX-a in waterworks and effective treatment strategies are needed when blooms of cyanobacteria cause occurrences of dhATX. For the first time, this study provides insights into the effectiveness of flocculation, adsorption onto activated carbon, ozonation, biologically active sand filtration and chlorination for removing dhATX-a from drinking water. The results revealed the incomplete adsorptive removal of dhATX-a with comparably high PAC dosages of 50 mg/L and 30 min contact time and the slight advantages of microporous activated carbons. Ozonation completely removed dhATX-a at an ozone consumption of only 2 mg/L, corresponding with a specific ozone consumption below 0.5 mg/L ozone per mg/L DOC. Biologically active sand filtration showed promising performance for dhATX-a elimination after a short adaptation of the microorganisms. Both flocculation and chlorination did not show any removals of dhATX-a, even at high dosages, and thus do not contribute to the multibarrier principal for the specific target compound dhATX-a.