Review of the occurrence, treatment technologies, and detection methods for saxitoxins in freshwaters

Harmful algal blooms (HABs) result from the excessive growth of cyanobacteria (also referred to as blue-green algae, although cyanobacteria are not true algae) in freshwater and seawater. HABs can adversely affect all levels of ecosystems, including but not limited to human life, fish, shellfish, marine mammals, and birds. Cyanotoxins can be classified as hepatotoxins (e.g., microcystins and cylindrospermopsin) and neurotoxins (e.g., saxitoxins and anatoxins). This review focuses on saxitoxins that are produced by several freshwater cyanobacterial species, such as Anabaena, Aphanizomenon, and Lyngbya.

The increasing occurrence of saxitoxins in freshwater reservoirs (Kaas & Henriksen 2000; Molica et al. 2002; Loftin et al. 2016; Grachev et al. 2018) and drinking water sources (AWWA 2016; Ohio EPA 2021), coupled with their high human toxicity (intraperitoneal LD₅₀ of 10 μg/kg in mice) (Wiberg & Stephenson 1960) present a pressing need for the development of strategies to monitor and mitigate these HABs. Toward this goal, it is essential for the U.S. EPA to establish a standard method for monitoring saxitoxins, as has been done for microcystins, cylindrospermopsin, and anatoxin-a (Shoemaker et al. 2015; U.S. EPA 2015c).

The detection methods of saxitoxins (Humpage et al. 2010; Rutkowska et al. 2019; Li & Persson 2021), treatment methods focused on the removal of saxitoxins (Silva et al. 2022), and the occurrence and fate of saxitoxins in freshwater (Christensen & Khan 2020) have been reviewed separately but there is no single comprehensive review of all the aforementioned elements, focused on implications for water treatment authorities. Hence, this review updates and provides a comprehensive review of the global occurrence of saxitoxins in freshwaters, and treatment technologies for removal of saxitoxins from drinking water. It also reviews the detection methods that can be applied for the monitoring of saxitoxins and their variants in water.

Saxitoxin, also referred to as the paralytic shellfish toxin (PST), is a neurotoxin produced by freshwater cyanobacteria. The more than 50 different variants of PSTs have a common backbone of a tetrahydropurine ring that can be substituted at the C11, N1, and C13 positions (Figure 1). The most common categories based on the functional group (R₄) at the C13 position are as follows in order of decreasing toxicity: carbamoyl (⁠⁠), decarbamoyl (⁠⁠), and N-sulfocarbamoyl (⁠⁠), (Genenah & Shimizu 1981; Shimizu et al. 1981; Oshima et al. 1989; Raposo et al. 2020). The most common variants of saxitoxin along with their toxicity equivalent factors (TEFs) are listed in Table 1.

Table 1

Variants of saxitoxin and their toxicity equivalent factor (TEF) (Raposo et al. 2020)

Group .	Toxin .	R1 .	R2 .	R3 .	R4 .	TEF .
Carbamoyl	STX	−H	−H	−H	−CO − NH2	1.0
	NeoSTX	−OH	−H	−H	−CO − NH2	1.0
	GTX1	−OH	−H		−CO − NH2	1.0
	GTX2	−H	−H		−CO − NH2	0.4
	GTX3	−H		−H	−CO − NH2	0.6
	GTX4	−OH		−H	−CO − NH2	0.7
Decarbamoyl	dcSTX	−H	−H	−H	−H	1.0
	dcNeoSTX	−OH	−H	−H	−H	0.4
	dcGTX1	−OH	−H		−H	NA
	dcGTX2	−H	−H		−H	0.2
	dcGTX3	−H		−H	−H	0.4
	dcGTX4	−OH		−H	−H	NA
N-sulfocarbamoyl	GTX5	−H	−H	−H		0.1
	GTX6	−OH	−H	−H		0.1
	C1	−H	−H			NA
	C2	−H		−H		0.1
	C3	−OH	−H			NA
	C4	−OH		−H		0.1

Group .	Toxin .	R1 .	R2 .	R3 .	R4 .	TEF .
Carbamoyl	STX	−H	−H	−H	−CO − NH2	1.0
	NeoSTX	−OH	−H	−H	−CO − NH2	1.0
	GTX1	−OH	−H		−CO − NH2	1.0
	GTX2	−H	−H		−CO − NH2	0.4
	GTX3	−H		−H	−CO − NH2	0.6
	GTX4	−OH		−H	−CO − NH2	0.7
Decarbamoyl	dcSTX	−H	−H	−H	−H	1.0
	dcNeoSTX	−OH	−H	−H	−H	0.4
	dcGTX1	−OH	−H		−H	NA
	dcGTX2	−H	−H		−H	0.2
	dcGTX3	−H		−H	−H	0.4
	dcGTX4	−OH		−H	−H	NA
N-sulfocarbamoyl	GTX5	−H	−H	−H		0.1
	GTX6	−OH	−H	−H		0.1
	C1	−H	−H			NA
	C2	−H		−H		0.1
	C3	−OH	−H			NA
	C4	−OH		−H		0.1

NA, not available.

View Large

Saxitoxins are one of the most potent cyanotoxins, with the lowest LD₅₀ value in mice (as shown in Table 2). The LD₅₀ of saxitoxin in humans ranges from 1 to 4 mg depending on the age and physical condition of the human (Suarez-Isla 2015). Cases of human exposure to saxitoxins have been documented through the ingestion of contaminated shellfish. Symptoms can include gastrointestinal disorders, and oral and facial paresthesias, which can appear within 0.5–2 h of ingestion, and in severe cases if left untreated, symptoms can lead to respiratory arrest and cardiovascular failure (Suarez-Isla 2015; Coleman et al. 2018). Recovery is possible through immediate symptomatic treatment such as respiratory support and diuretics, which can help in the elimination of the saxitoxins from the body through urination. There has been no report indicating any long-term effects of saxitoxin exposure. However, there is no antidote that has been approved for the treatment of saxitoxin poisoning in humans (Suarez-Isla 2015; Coleman et al. 2018). The four major cyanotoxins – microcystins, cylindrospermopsin, anatoxin-a, and saxitoxins – have been listed on the Contaminant Candidate List 4 (CCL4) and List 5 (CCL5) by the U.S. EPA. However, the U.S. EPA has only established drinking water health advisories for microcystins and cylindrospermopsin (U.S. EPA 2015a, 2015b).

Saxitoxins are hydrophilic polar compounds that are basic due to the presence of guanidinium and hydroxyl groups. The guanidinium group at C7, C8, C9 has a pKa of 8.22 and that at C1, C2, C3 has a pKa of 11.28 (Rogers & Rapoport 1980; Strichartz 1984; Schantz 1986; Hall et al. 1990). At a neutral pH, the saxitoxin molecule carries a bivalent positive charge which, upon an increase in pH, decreases as the C8 guanidinium group donates a hydronium ion (Shimizu et al. 1981; Strichartz 1984). Even with a low octanol-water partitioning coefficient (K_ow) of <0.001, saxitoxins can bioaccumulate in fish and animals through the intestine caused by the alkaline environment (pH > 8.22), which results in the deprotonation of the guanidinium group of saxitoxins at the C7, C8, C9 positions, causing the molecule to become nonpolar and hence more prone to diffusion across the lipid bilayer (Llewellyn 2006).

Saxitoxins are very unstable in nature and are likely to undergo chemical transformation to the other toxic variants of saxitoxin (Jones & Negri 1997; Negri et al. 1997). Saxitoxins are resistant to biodegradation in the surface water (Tang et al. 2012) but have shown to be susceptible to bacterial degradation when exposed to bacteria isolated from the digestive tracts of blue mussels with >90% degradation of saxitoxins within 3 days (Donovan et al. 2008).

Saxitoxins are produced by freshwater cyanobacteria such as Anabaena, Cylindrospermopsis, Aphanizomenon, Planktothrix, and Lyngbya, and by eukaryotic dinoflagellates in marine environments. Fifteen freshwater species of cyanobacteria have been identified as saxitoxin producers (Christensen & Khan 2020). Aphanizomenon flos-aquae was the first freshwater cyanobacterium identified as a saxitoxin producer in the late 1960s (Onodera et al. 1997; Jackim & Gentile 2021). A review by Christensen & Khan (2020) provides a comprehensive list of cyanobacterial species producing saxitoxins. Table 3 provides a summary of the global occurrence of cyanobacterial species that produced saxitoxins in freshwaters.

Table 3

Global occurrence of saxitoxin and its variants along with the associated cyanobacteria in freshwaters

Year .	Location .	Cyanobacteria .	Variants of saxitoxin .	Concentration range .	Method of detection .	Reference .
1990–1993	Murray-Darling basin, New South Wales and South Australia	Anabaena circinalis	STX, GTX1-6, dcGTX2, dcGTX3, and C1-2	STX equiv.	HPLC and Fast Atom Bombardment (FAB)-MS	Humpage et al. (1994), Baker & Humpage (1994)
1993	Guntersville reservoir, Alabama, USA	Lyngbya wollei	dcSTX, dcGTX2, dcGTX3	STX equiv.	HPLC-FLD	Onodera et al. (1997), Carmichael et al. (1997)
1994	Farm dam near Forbes, New South Wales, Australia	Anabaena circinalis	C1-2, dcGTX2, dcGTX3, GTX2-5, STX, dcSTX	A. circinalis	HPLC-FLD	Negri et al. (1995)
1994	96 freshwater ponds and lakes, Denmark	Anabaena lemmermannii (dominant species)	STX, GTX1-5, dcSTX, neoSTX	STX equiv.	HPLC-FLD	Kaas & Henriksen (2000)
1994–1996	2 reservoirs in State of São Paulo, Brazil	Cylindrospermopsis racibroskii	STX, neoSTX	Not available	HPLC-FLD and HPLC-ESI-MS	Lagos et al. (1999)
1996	Montargil reservoir in Portugal	Aphanizomenon flos-aquae, Microcystis aeruginosa	STX, neoSTX, dcSTX, GTX5-6	Not available	HPLC-FLD and LC/MS	Pereira et al. (2000)
1996	Crestuma-Lever reservoir in Portugal	Aphanizomenon flos-aquae, Microcystis aeruginosa	GTX1, GTX3-4	STX equiv.	HPLC-FLD	Ferreira et al. (2001)
1997	Lake Varese, Italy	Planktothrix sp.	STX	Not available	HPLC-FLD and LC/MS	Pomati et al. (2000)
2000	Armano Ribeiro Gonçalves reservoir and Pataxó channel, Brazil	Cylindrospermopsis raciborskii	STX, GTX, C1-2		HPLC-FLD	Costa et al. (2006)
2002–2003	Finnish freshwater in south-eastern and central Finland	Anabaena lemmermannii	STX		HPLC-FLD and LC/MS	Rapala et al. (2005)
2005–2008	Recreational area of Champs-sur-Marne at Paris, France	Aphanizomenon gracile, Aphanizomenon flos-aquae	STX, neoSTX	equiv. STX	HILIC-MS	Ledreux et al. (2010)
2006	Lakes and reservoirs in Missouri, Iowa, Kansas, and Minnesota (USA)	Anabaena, Aphanizomenon, Planktothrix	N/Aa		ELISA	Graham et al. (2010)
2008–2009	Lake Pamvotis, Greece	Aphanizomenon flos-aqua	N/Aa		ELISA	Gkelis & Zaoutsos (2014), Gkelis et al. (2014)
2009	Lake Atitlan, Guatemala	Lyngbya	N/Aa		ELISA	Rejmánková et al. (2011)
2009	Arctic freshwaters in northern Baffin Island	Scytonema cf. crispum, Lyngbya wollei	N/Aa	Not available	ELISA	Kleinteich et al. (2013)
2009–2011	Reservoirs in Rio Grande do Norte, Brazil	Cylindrospermopsis raciborskii, Planktothrix agardhii, Aphanizomenon gracile, Anabaena circinalis	N/Aa		ELISA	Fonseca et al. (2015)
2010	Lake Baikal, Russia	Anabaena lemmermannii	N/Aa		ELISA	Belykh et al. (2015)
2010	19 lakes and reservoirs in Czech Republic	Anabaena sp., Aphanizomenon sp.	N/Aa		ELISA	Jančula et al. (2014)
2011	Drinking water pretreatment reservoir and lakes in a recreational reserve of The Groynes in South Island, New Zealand	Scytonema cf. crispum	STX, neoSTX, GTX1-5, dcSTX, dcGTX2-3	dry weight	HPLC-FLD	Smith et al. (2011)
2014	The Vistonis lake, Greece	Aphanizomenon favaloroi	STX, neoSTX	phytoplankton dry weight for STX and neoSTX, respectively	HILIC-MS/MS	Moustaka-Gouni et al. (2017)
2014–2015	Lakes Gauštvinis, Jieznas, and Širvys in Lithuania	Aphanizomenon gracile	STX		HILIC-MS/MS	Karosienė et al. (2020)
2016	Karla reservoir, Greece	Aphanizomenon favaloroi, Cylindrospermopsis raciborskii	N/Aa		ELISA	Papadimitriou et al. (2018)
2017	Irkutsk reservoir, Russia	Anabaena lemmermannii	STX		HPLC-MS and ELISA	Grachev et al. (2018)
2017–2018	Peri Lake, Brazil	Cylindrospermopsis raciborskii	STX, neoSTX, dcSTX, GTX1-5	STX equiv.	HPLC-FLD	Ramos et al. (2021)
2018	5 lakes in northern Zealand, Denmark	Anabaena lemmermannii	STX, GTX, dcSTX, neoSTX, dcneoSTX		HPLC-FLD	Podduturi et al. (2021)
2019	Archbold Water Treatment Plant Reservoir, Ohio, USA	N/A	N/Aa		ELISA	Ohio EPA (n.d.)
2019–2020	Lake Taihu, China	Dolichospermum, Apanizomenon, and Oscillatoria	N/Aa	Mean of in May and in November	ELISA	Li et al. (2022)
2020–2021	Rawal Lake, Pakistan	N/A	N/Aa	Not available	ELISA	Batool et al. (2024)
2022	St. Joseph Lake, Ohio, USA	N/A	N/Aa		ELISA	Ohio EPA (n.d.)
2022	Burr Oak Reservoir, Ohio, USA	N/A	N/Aa		ELISA	Ohio EPA (n.d.)
2023	St. Joseph Lake, Ohio, USA	N/A	N/Aa		ELISA	Ohio EPA (n.d.)
2023	Burr Oak Reservoir, Ohio, USA	N/A	N/Aa		ELISA	Ohio EPA (n.d.)

Year .	Location .	Cyanobacteria .	Variants of saxitoxin .	Concentration range .	Method of detection .	Reference .
1990–1993	Murray-Darling basin, New South Wales and South Australia	Anabaena circinalis	STX, GTX1-6, dcGTX2, dcGTX3, and C1-2	STX equiv.	HPLC and Fast Atom Bombardment (FAB)-MS	Humpage et al. (1994), Baker & Humpage (1994)
1993	Guntersville reservoir, Alabama, USA	Lyngbya wollei	dcSTX, dcGTX2, dcGTX3	STX equiv.	HPLC-FLD	Onodera et al. (1997), Carmichael et al. (1997)
1994	Farm dam near Forbes, New South Wales, Australia	Anabaena circinalis	C1-2, dcGTX2, dcGTX3, GTX2-5, STX, dcSTX	A. circinalis	HPLC-FLD	Negri et al. (1995)
1994	96 freshwater ponds and lakes, Denmark	Anabaena lemmermannii (dominant species)	STX, GTX1-5, dcSTX, neoSTX	STX equiv.	HPLC-FLD	Kaas & Henriksen (2000)
1994–1996	2 reservoirs in State of São Paulo, Brazil	Cylindrospermopsis racibroskii	STX, neoSTX	Not available	HPLC-FLD and HPLC-ESI-MS	Lagos et al. (1999)
1996	Montargil reservoir in Portugal	Aphanizomenon flos-aquae, Microcystis aeruginosa	STX, neoSTX, dcSTX, GTX5-6	Not available	HPLC-FLD and LC/MS	Pereira et al. (2000)
1996	Crestuma-Lever reservoir in Portugal	Aphanizomenon flos-aquae, Microcystis aeruginosa	GTX1, GTX3-4	STX equiv.	HPLC-FLD	Ferreira et al. (2001)
1997	Lake Varese, Italy	Planktothrix sp.	STX	Not available	HPLC-FLD and LC/MS	Pomati et al. (2000)
2000	Armano Ribeiro Gonçalves reservoir and Pataxó channel, Brazil	Cylindrospermopsis raciborskii	STX, GTX, C1-2		HPLC-FLD	Costa et al. (2006)
2002–2003	Finnish freshwater in south-eastern and central Finland	Anabaena lemmermannii	STX		HPLC-FLD and LC/MS	Rapala et al. (2005)
2005–2008	Recreational area of Champs-sur-Marne at Paris, France	Aphanizomenon gracile, Aphanizomenon flos-aquae	STX, neoSTX	equiv. STX	HILIC-MS	Ledreux et al. (2010)
2006	Lakes and reservoirs in Missouri, Iowa, Kansas, and Minnesota (USA)	Anabaena, Aphanizomenon, Planktothrix	N/Aa		ELISA	Graham et al. (2010)
2008–2009	Lake Pamvotis, Greece	Aphanizomenon flos-aqua	N/Aa		ELISA	Gkelis & Zaoutsos (2014), Gkelis et al. (2014)
2009	Lake Atitlan, Guatemala	Lyngbya	N/Aa		ELISA	Rejmánková et al. (2011)
2009	Arctic freshwaters in northern Baffin Island	Scytonema cf. crispum, Lyngbya wollei	N/Aa	Not available	ELISA	Kleinteich et al. (2013)
2009–2011	Reservoirs in Rio Grande do Norte, Brazil	Cylindrospermopsis raciborskii, Planktothrix agardhii, Aphanizomenon gracile, Anabaena circinalis	N/Aa		ELISA	Fonseca et al. (2015)
2010	Lake Baikal, Russia	Anabaena lemmermannii	N/Aa		ELISA	Belykh et al. (2015)
2010	19 lakes and reservoirs in Czech Republic	Anabaena sp., Aphanizomenon sp.	N/Aa		ELISA	Jančula et al. (2014)
2011	Drinking water pretreatment reservoir and lakes in a recreational reserve of The Groynes in South Island, New Zealand	Scytonema cf. crispum	STX, neoSTX, GTX1-5, dcSTX, dcGTX2-3	dry weight	HPLC-FLD	Smith et al. (2011)
2014	The Vistonis lake, Greece	Aphanizomenon favaloroi	STX, neoSTX	phytoplankton dry weight for STX and neoSTX, respectively	HILIC-MS/MS	Moustaka-Gouni et al. (2017)
2014–2015	Lakes Gauštvinis, Jieznas, and Širvys in Lithuania	Aphanizomenon gracile	STX		HILIC-MS/MS	Karosienė et al. (2020)
2016	Karla reservoir, Greece	Aphanizomenon favaloroi, Cylindrospermopsis raciborskii	N/Aa		ELISA	Papadimitriou et al. (2018)
2017	Irkutsk reservoir, Russia	Anabaena lemmermannii	STX		HPLC-MS and ELISA	Grachev et al. (2018)
2017–2018	Peri Lake, Brazil	Cylindrospermopsis raciborskii	STX, neoSTX, dcSTX, GTX1-5	STX equiv.	HPLC-FLD	Ramos et al. (2021)
2018	5 lakes in northern Zealand, Denmark	Anabaena lemmermannii	STX, GTX, dcSTX, neoSTX, dcneoSTX		HPLC-FLD	Podduturi et al. (2021)
2019	Archbold Water Treatment Plant Reservoir, Ohio, USA	N/A	N/Aa		ELISA	Ohio EPA (n.d.)
2019–2020	Lake Taihu, China	Dolichospermum, Apanizomenon, and Oscillatoria	N/Aa	Mean of in May and in November	ELISA	Li et al. (2022)
2020–2021	Rawal Lake, Pakistan	N/A	N/Aa	Not available	ELISA	Batool et al. (2024)
2022	St. Joseph Lake, Ohio, USA	N/A	N/Aa		ELISA	Ohio EPA (n.d.)
2022	Burr Oak Reservoir, Ohio, USA	N/A	N/Aa		ELISA	Ohio EPA (n.d.)
2023	St. Joseph Lake, Ohio, USA	N/A	N/Aa		ELISA	Ohio EPA (n.d.)
2023	Burr Oak Reservoir, Ohio, USA	N/A	N/Aa		ELISA	Ohio EPA (n.d.)

HPLC, high-performance liquid chromatography; FLD, fluorescence detector.

Note:^aMeasured in terms of total saxitoxins.

View Large

Saxitoxins were found to be present in 8% of the 1,118 globally reported instances of cyanotoxins, with the highest percentage of saxitoxins, i.e., 21%, occurring in Australia and New Zealand (Svirčev et al. 2019). A national study conducted by the U.S. EPA in 2007 of 1,161 lakes and reservoirs found saxitoxin to be present in 7.7% of the samples with a mean concentration of ⁠; however, saxitoxin producers were present at a high percentage, i.e., 79% of the samples (Loftin et al. 2016). Ohio EPA required public water systems based in Ohio to monitor for saxitoxin and their data are displayed on a monitoring map which is updated regularly (Ohio EPA n.d.); it is worth mentioning that Ohio is one of the few states in the US that monitors for saxitoxin. Saxitoxin concentrations as high as 1.35 μg/L have been observed in the intakes of water treatment plants in Ohio in recent years.

The production of saxitoxins in freshwaters is largely influenced by environmental factors, such as temperature, light intensity, conductivity, water hardness, and nutrient presence (Castro et al. 2004; Carneiro et al. 2009, 2013; Burford et al. 2016). High nitrogen to phosphorus ratios have been shown to result in at least a threefold increase in the production of saxitoxin by Alexandrium tamarense (Granéli & Flynn 2006). The positive correlation of saxitoxin production with higher TN (total nitrogen):TP (total phosphorus) ratio was also reported by Moraes et al. (2021) while on the other hand, phosphorus limitation was reported to have increased the saxitoxin concentration in both Alexandrium sp. and R. raciborskii species, suggesting saxitoxin production to be a form of survival strategy (Moraes et al. 2021). Global warming has resulted in increased drought periods, especially in semi-arid regions like those in Northeast Brazil. A study conducted by Carneiro et al. (2013) revealed that this prolonged drought period, which causes intense evaporation which in turn leads to high salt concentrations in water bodies, resulted in higher saxitoxin production for a certain period. A positive correlation between conductivity and saxitoxin concentrations was also found in Peri Coastal Lake (Brentano et al. 2016). Pomati et al. (2004) produced evidence that suggested either STX metabolism or the toxin itself could be linked to maintaining homeostasis of the cyanobacterial cell under alkaline conditions, hence explaining the production of saxitoxins in higher salt concentrations. Temperature was also shown to affect the production of saxitoxins in dinoflagellates such that cells exposed to cold stress conditions resulted in higher production of saxitoxins (Kim et al. 2021). Higher production of saxitoxin occurred at high light intensity (⁠⁠) combined with a temperature of 25 °C as compared with a combination of high light intensity with a temperature of 30 °C (Mesquita et al. 2019). Turbidity had a negative correlation with saxitoxin concentrations (Moraes et al. 2021).

The removal of cyanotoxins, including saxitoxins, largely depends on whether the cyanotoxin is intracellular or extracellular. For removal of intracellular toxins, the intact cyanobacterial cells need to be removed, which is usually achieved by the conventional water treatment process such as coagulation and sand filtration and it is crucial to remove the cells from the solids of the sedimentation basin or from filters to avoid lysing of the cells. The use of disinfectant prior to the application of conventional treatment processes can result in lysing of the cyanobacterial cells. The release of intracellular toxins, due to cell lysis or death, to form extracellular toxins, which are dissolved in water, poses a challenge for water treatment plants. The various technologies that are employed for the removal of extracellular toxins include physical removal with activated carbon or membrane filtration, chemical inactivation using oxidants such as chlorine, potassium permanganate, ozone, or UV light, and biological inactivation through biological activity. The removal of extracellular saxitoxins by the different treatment technologies is discussed in this section and is summarized in Table 4.

Saxitoxins with a deprotonated C8 guanidinium group are susceptible to oxidation, which explains their degradation even by hypochlorite at pH > 8 (Rogers & Rapoport 1980; Newcombe & Nicholson 2002; Nicholson et al. 2003). A pH of >8.5 along with a residual free chlorine concentration of 0.5 mg/L was required for >90% removal of saxitoxins (Newcombe & Nicholson 2002; Nicholson et al. 2003). When tested with post-filtration phase water for the drinking water treatment conditions of the City of Akron (i.e., 3.2 mg/L NaClO with a contact time of 118 min), the removal of 0.3 μg/L saxitoxin was higher at pH 9 (61%) as compared with at pH 6 (13%). This was compared with the City of Alliance drinking water treatment conditions (i.e., 1.8 mg/L NaClO with a contact time of 235 min), which only removed 36% saxitoxin at pH 9. The presence of microcystin-LR at a concentration of 1.6 μg/L increased the removal of saxitoxin by ∼2 fold under the City of Alliance treatment conditions (Garcia et al. 2024). The majority of saxitoxin variants were resistant to batch ozone treatment within a pH range of 7–8 at a residual ozone concentration of 0.5 mg/L for 10 min (Rositano et al. 2001; Newcombe & Nicholson 2002; Orr et al. 2004). Total and extracellular saxitoxin concentrations remained unchanged after treatment with potassium permanganate (0.5 mg/L) in natural waters at pH ≈ 8 (Ho et al. 2009). STX and GTX2/3 were susceptible to photolysis at a pH of 8 and not at a pH of 6, and the study also suggested that hydroxyl radical was not a significant contributor in the degradation, implying that the toxins did not undergo hydrolysis or direct photolysis (Kurtz 2021).

The only study that evaluated catalytic wet peroxide oxidation (CWPO) of saxitoxin showed that only 60% removal of STX occurred after 5 h of reaction as compared with 100% removal of microcystin-RR in 1.5 h reaction time (Munoz et al. 2019).

Adsorption using granulated activated carbon (GAC) with an empty bed contact time of 15 min completely removed all the high potency saxitoxins, i.e., STX, dcSTX, and GTX2-3 (Orr et al. 2004). GAC with a greater amount of mesopores favor higher adsorption of saxitoxins (Silva Buarque et al. 2015). The Langmuir isotherm is best known for describing the kinetics of the adsorption of saxitoxin onto GAC (Capelo-Neto & Buarque 2016). During adsorption by powdered activated carbon (PAC), an increase in the carbon dose and contact time were correlated with higher removal efficiency of saxitoxins (Ho et al. 2009; Shi et al. 2012). In addition, a pH > 8 resulted in a higher removal rate of saxitoxins (Shi et al. 2012; Rorar et al. 2023; Walke 2023). Absorption was highest at a pH of 10.7 when the saxitoxin molecule was in its neutral form, suggesting that dispersive and H-bonding interactions played a dominant role during the adsorption of saxitoxin and the effect of NOM on the adsorption reduced as pH increased (Shi et al. 2012). The presence of negatively charged microcystin-LR also increased the adsorption of saxitoxin at a higher pH of 7–9 due to the electrostatic forces of the larger molecule microcystin-LR, which help attract the positively charged saxitoxin to the surface of PAC (Rorar et al. 2023; Walke 2023). Higher initial concentrations of saxitoxin were shown to be related to higher removal percentages using PAC (Rorar et al. 2023). In a review on adsorption processes for the removal of saxitoxin, the key parameters that governed adsorption were: (i) the type, chemical structure, pore size, and surface area of the adsorbent, (ii) molecular weight of the variant of saxitoxin, (iii) pH of the water during the adsorption process, and (iv) the organic matter present in the sample (Silva et al. 2022). The same study also ranked the efficiency of adsorbents that have been studied in the past for saxitoxin in the following order: wood-based PAC > bituminous PAC > lignite PAC > coconut shell GAC > chitin > oyster shell powder.

When the removal of saxitoxin and its variants was investigated on two types of nanofiltration membranes, NF-90 and NF-270, the NF-90 membrane showed higher removal in comparison with the NF-270. The higher hydrophobicity coupled with a smaller average pore diameter favored higher removal and prevented a decline in the permeate flux (Coral et al. 2011). The comparison of rejection of saxitoxins by NF-270 with two hydrophobic (HYDRACoRe-10 and HYDRACoRe-50) and another hydrophilic membrane (TRISEP 8040-SBNF-TSA) showed that NF-270 had the highest rejection of >90% over the period of 48 h as compared with the other membranes (Lebad et al. 2024).

Ohio EPA monitoring data shows the detection of saxitoxins at concentrations below the drinking water guideline, i.e., 1.6 μg/L in finished drinking water at some water treatment plants (Ohio EPA n.d.). The data also shows that water treatment plants employing activated carbon at the start of the water treatment process have shown to be capable of removing approximately 80% of saxitoxins based on their intake and finished water concentrations.

Ingestion of toxic shellfish has been the primary route of exposure to saxitoxins in humans, resulting in the focus of most studies being on detecting saxitoxin in shellfish samples. The mouse bioassay (MBA) was the earliest biological method that was developed and standardized for the detection of saxitoxins as indicated in Figure 3 (Sommer & Meyer 1937). However, the limit of detection (LOD) of the method is 40 μg STX/100 g shellfish, equivalent to a concentration of 200 μg/L saxitoxin in water, which is well beyond WHO's 3 μg/L drinking water guideline (World Health Organization 2019). In addition, the method's use of live animals combined with its lack of sensitivity, which is essential to detect low concentrations of saxitoxin, and inconsistent results, resulted in the need for alternative methods.

Saxitoxins inherently lack fluorescence, which previously limited their detection by analytical methods like gas chromatography and spectrometry. However, Bates & Rapoport (1975) developed a technique that involved alkaline hydrogen peroxide oxidation of saxitoxin to yield fluorescent byproducts, which formed a basis for further advancement in the development of analytical methods for saxitoxin detection. The incorporation of this fluorometric method in a post-column reaction system that involved the separation of saxitoxins by high-pressure liquid chromatography (HPLC) was first brought about by Sullivan (Sullivan & Wekell 1984). However, this method was only limited to detecting GTX 1-6, STX, and neoSTX, and was not capable of separating N-sulfocarbomyl-11-hydroxysulfate toxins (C1–C4). This was resolved in Oshima's study involving post-column derivatization liquid chromatography (LC) that could detect as many as 15 variants of saxitoxin (Oshima et al. 1989; Oshima 1995).

While Oshima's method used periodic acid for oxidizing the saxitoxins after chromatographic injection, Lawrence set up a method to oxidize the toxins with peroxide prior to injection (Lawrence et al. 1995). This precolumn oxidation of saxitoxins, also known as the ‘Lawrence method’ (Lawrence et al. 2005), was later adopted as the AOAC Official Method 2005.06. Another discovery of using electrospray ionization-mass spectrometry (ESI-MS) to detect saxitoxins proved beneficial as it meant that the direct detection of the toxins was possible without oxidation. Also, since saxitoxin is basic, it provides strong []⁺ ions that can be effectively detected by ESI-MS (Quilliam et al. 1989).

To overcome the challenges posed by LC/MS, an alternate separation method was sought out that would allow for the simultaneous detection of all saxitoxin variants in a single analysis while also providing high sensitivity (Dell'Aversano et al. 2004). With HILIC-MS (Buszewski & Noga 2012), the stationary phase is a polar compound, e.g., amide, which is capable of easily absorbing polar solvents like water. The polarity of the stationary phase increases with a layering of polar solvents, which attracts polar analytes like saxitoxin. The retention of the analyte is also dependent on the polarity of the mobile phase. Hence, the gradient is designed such that the mobile phase is composed of lower polarity solvents like acetonitrile in the start to facilitate retention, with a gradual increase in high polar solvents like water to enable elution of the analyte from the column. The use of formate buffers is necessary to achieve a good peak shape and also decrease the retention time (Quilliam et al. 2001). A summary of HILIC-MS methodologies for the detection and quantification of saxitoxins is provided in Table 5.

Dell'Aversano et al. (2004) were the first to test the suitability of the HILIC-MS detection method on field cyanobacterial cell samples containing Anabaena circinalis and Cylindrospermopsis raciborskii. In another study, Johnson et al. (2009) extracted and quantified STX and neoSTX from human urine. This study was further implemented with online-solid phase extraction before LC/MS/MS (Bragg et al. 2015). The authors claim that using the online method reduced the time required for sample preparation (1 h online versus 3 h offline).

While Dell'Aversano et al. (2004, 2005) were the first to develop the HILIC-MS method for quantification of saxitoxins, their method run time was long with STX being detected at 20.3 min. Hence, Halme et al. (2012) developed a method for the fast and quantitative analysis of STX, achieving a retention time of 6.5 min. The developed method was also verified by application on freeze-dried Alexandrium Ostenfeldii samples.

In another study, neoSTX was detected in human plasma using HILIC-MS. Sample cleanup was accomplished using cation exchange SPE, a modification of the SPE method developed by Johnson et al. (2009) and Peake et al. (2016). A novel extraction method using a combination of silica and strong cation exchange SPE was developed to extract saxitoxins from food and water (Jansson & Åstot 2015). To ensure accurate and sensitive measurement of saxitoxins from urine samples, polyamide was used as a HILIC SPE material for cleanup of samples, with recoveries ranging between 90 and 120%, prior to detection using HILIC-MS (Xu et al. 2018).

Haddad et al. (2019) used a zwitterionic HILIC column for the separation of saxitoxin following its extraction from freshwater samples using carbon-based SPE with a recovery of 53% (Haddad et al. 2019). Vo Duy et al. (2022) optimized an on-line enrichment method using hydrophilic–lipophilic balance-based adsorbents coupled with HILIC-MS. They also evaluated the adsorptive losses of saxitoxin to be maximum using glass autosampler vials and identified polypropylene as the preferred material to avoid time-dependent adsorptive losses of saxitoxin while analyzing aqueous samples. In another study, the concentrations of saxitoxins were quantified in seawater (Bosch-Orea et al. 2021).

The higher sensitivity of detection is always of peak interest for researchers, which in turn increases the demand for research in extraction methods. While SPE and liquid–liquid extraction (LLE) provide the highest selectivity and sensitivity, microscale sample preparation is a green technique with the potential to be applied for the detection of saxitoxins (Ishak et al. 2023).

ELISA is a biochemical assay that uses antibodies based on the target analyte. The concentration of toxins is measured based on a colorimetric reaction. Since the assay is designed to detect the antibody on the plate, the signal is inversely proportional to the amount of toxins present in the sample. ELISA is a popular alternative to LC/MS due to its sensitivity, rapidity, and ease-of-use and is compared to HILIC-MS in Table 6. The evolution of assays developed for the detection of saxitoxins has been summarized in previous reviews (Usleber et al. 2001; Humpage et al. 2010; Cusick & Sayler 2013). Since this detection method has been researched extensively over the years, it has led to the commercialization of several ELISA kits (Li & Persson 2021). The LOD of the commonly used 96-well ELISA plate by Abraxis is 0.015 μg/L for STX present in water samples without prior sample preparation. ELISA is utilized by the Ohio EPA, using the Ohio EPA Total Saxitoxin by ELISA analytical methodology (Ohio EPA 2016) to conduct surveillance for the presence of saxitoxin, whereas repeat sampling in response to the detection of saxitoxin in finished water is measured using LC/MS/MS (Heather Raymond 2018).

Table 6

Comparison between ELISA and HILIC-MS methods for detecting saxitoxins from water

Comparison factor .	HILIC-MS .	ELISA .
Sample preparation	Samples need to be cleaned up using SPE and reconstituted into the mobile phase solvents of the HILIC-MS system.	No sample preparation is necessary. Dilutions can be made as the range falls beyond the linear detection range.
Typical cost	$325/sample cost includes: • Standard curve • QC measurements • Additional cost for sample preparation not included in the above cost.	$535/plate Kit includes: • Ability to perform 96 measurements including standards and control.
Time (if analysis is conducted in-house)	Can perform 3–6 samples per hour. The addition of blank samples between high concentration samples to prevent carryover increases the total time.	Approximately 2 h per plate (10–96 samples)
LOD	Varies depending on the method and enrichment by solid phase extraction (0.72 ng/L–0.5 μg/L)
Limitations	High cost of analysis Sample matrix needs to be changed to match the mobile phase solvents of the LC system. Requires a trained and skilled personnel in LC/MS/MS to perform analysis. Total time of analysis depends on the method and can exceed several hours.	High cross-reactivity of dcSTX and GTX2&3 (23–29%) LOD can be affected if the above toxins are present in the water sample. Not capable of quantifying the individual variants of saxitoxins. Requires access to a plate reader to measure the absorbance of the samples.
Advantages	Depending on the availability of standards can quantify the variants of saxitoxin. High accuracy in measurement Matrix effects can be reduced by sample preparation techniques.	A quick method to check the presence of saxitoxins in samples. Does not need any prior skills or knowledge of the method to perform the measurement. Kits are readily available for purchase.

Comparison factor .	HILIC-MS .	ELISA .
Sample preparation	Samples need to be cleaned up using SPE and reconstituted into the mobile phase solvents of the HILIC-MS system.	No sample preparation is necessary. Dilutions can be made as the range falls beyond the linear detection range.
Typical cost	$325/sample cost includes: • Standard curve • QC measurements • Additional cost for sample preparation not included in the above cost.	$535/plate Kit includes: • Ability to perform 96 measurements including standards and control.
Time (if analysis is conducted in-house)	Can perform 3–6 samples per hour. The addition of blank samples between high concentration samples to prevent carryover increases the total time.	Approximately 2 h per plate (10–96 samples)
LOD	Varies depending on the method and enrichment by solid phase extraction (0.72 ng/L–0.5 μg/L)
Limitations	High cost of analysis Sample matrix needs to be changed to match the mobile phase solvents of the LC system. Requires a trained and skilled personnel in LC/MS/MS to perform analysis. Total time of analysis depends on the method and can exceed several hours.	High cross-reactivity of dcSTX and GTX2&3 (23–29%) LOD can be affected if the above toxins are present in the water sample. Not capable of quantifying the individual variants of saxitoxins. Requires access to a plate reader to measure the absorbance of the samples.
Advantages	Depending on the availability of standards can quantify the variants of saxitoxin. High accuracy in measurement Matrix effects can be reduced by sample preparation techniques.	A quick method to check the presence of saxitoxins in samples. Does not need any prior skills or knowledge of the method to perform the measurement. Kits are readily available for purchase.

View Large

Although the HILIC-MS and ELISA methods are sensitive and reliable methods for the detection of saxitoxin, they cannot be applied in the field for the rapid measurement of saxitoxins. Hence, research has taken a new direction in the detection of saxitoxin, which involves the use of biosensors. Biosensors are equipped with bioreceptors and signal processing units, that allow for the measurement of signal changes that are caused by the interaction between the bioreceptor and the target material (Conroy et al. 2009). Bioreceptors in the form of antibodies, aptamers, and nanomaterials are designed, creating two broad categories of biosensors: electrochemical and optical biosensors, which have been applied for the detection of saxitoxins from freshwater and seawater (Park et al. 2022).

Saxitoxins, possessing neurotoxicity, are the most potent of all cyanotoxins with over 50 different variants. The most common and well-known route of exposure to humans has been through ingestion of toxic shellfish, giving saxitoxin the popular terminology of paralytic shellfish toxins (PSTs). Saxitoxins were predominantly produced in marine environments by dinoflagellates but are now increasingly being detected in freshwaters, produced by cyanobacteria. Cyanobacterial growth is largely related to nutrient pollution of freshwater bodies and is exacerbated by climate change. Through the knowledge presented in this review, it is evident that climate change is not only responsible for increasing the occurrence of saxitoxin-producing cyanobacteria in freshwaters but is also a promoter of saxitoxin production. Given that reliable analytical methods exist for saxitoxin quantification, in addition to the inevitable effects of climate change and the expected increase in saxitoxins in the future, the U.S. EPA has sufficient reason to establish a drinking water health advisory for saxitoxins which will thereby necessitate its monitoring, hence protecting our drinking water.

Most treatment technologies reviewed in this article have only been studied at the laboratory scale and there is enough evidence to suggest that some treatment methods employed by water treatment plants are incapable of completely removing the toxicity of the saxitoxin molecule based on the Ohio EPA monitoring database; hence, water treatment plants rely on multiple barriers of treatment (Ohio EPA n.d.). With a rise in HABs in temperate regions, there are increased chances of cyanotoxins entering the source water of many drinking water treatment plants, posing a challenge for their simultaneous removal. Microcystins, the chemistry of which is different from saxitoxins, have been the focus of oxidative treatment for the majority of water treatment plants, which cannot be applied for the treatment of saxitoxins. Hence, it is crucial to develop a treatment technology which can reliably remove all cyanotoxins.

The treatment of saxitoxins comes with another challenge posed by the polar saxitoxins, i.e., its monitoring in water. Unlike microcystins, saxitoxins lack inherent fluorescence and are unable to be retained on reverse phase columns, which makes it challenging to detect saxitoxins by chromatography or spectrometry without chemical derivatization/oxidation. The additional step of chemical derivatization/oxidation reduces signal efficiency and selectivity of the variants of saxitoxin. The use of HILIC-MS for the detection of saxitoxins has exceeded all other detection methods in terms of sensitivity, selectivity, and time efficiency. However, the cost of instrumentation is very high and requires highly skilled personnel for the operation of these instruments, which creates a major limitation for water treatment operators. ELISA is a preferred alternative for the detection of saxitoxins due to its ease-of-use, sensitivity, and reduced cost but is incapable of quantifying the variants of saxitoxin. The structure of saxitoxin makes the development of sample pretreatment techniques, like SPE, just as difficult as analytical methods used for the detection of saxitoxins. However, since saxitoxins are present at low concentrations in surface waters, it is essential to develop concentration methods with a good recovery that can lead to their successful quantification. Hence, it is evident that there is still much progress to be made in the monitoring and detection of saxitoxins from water which can serve the purpose of regulation in water treatment plants, to protect public health.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.