A new approach to predicting and understanding cyanobacteria bloom toxicity

Cyanobacteria blooms pose a global threat, with toxin prediction remaining a critical challenge for effective management. However, a crucial often overlooked feature of toxic cyanobacteria blooms is the co-occurrence of strains producing different amounts of toxin, and the effect this has on bloom toxin yield. We propose that the overall toxin yield in cyanobacteria blooms is predominantly driven by the relative proportion of different strains in a population, rather than changes in toxin production. Future research needs to address strain dynamics within blooms to predict toxin amount. We are facing a crisis in ever expanding toxic freshwater cyanobacteria blooms. The ability to assess the risk of bloom toxicity, and hence develop effective management plans, is critical to ensure safe drinking water and a healthy environment. However, despite years of research, toxin yields in blooms cannot be effectively predicted (Busari et al. 2023; de Figueiredo 2024; Marrone et al. 2024; Rousso et al. 2020). We need to change our thinking and our approach to understanding toxicity in cyanobacteria blooms: this is both a challenge and an opportunity. Toxin production as cyanobacteria bloom leads to severe environmental, social, and economic problems. Cyanotoxins pollute freshwater ecosystems with impacts that include a reduction in biodiversity, killing fish, birds, and mammals, and significant financial costs to remove toxins from human drinking water. The main cyanotoxins produced by freshwater cyanobacteria are neurotoxins (e.g., saxitoxin and anatoxins) or hepatotoxins (e.g., microcystins and cylindrospermopsins): posing serious health effects on the nervous system, liver, and other major organs (Pearson et al. 2016). Large blooms have affected major lakes and drinking water sources such as Lake Erie, North America, and Lake Taihu, China, as well as smaller water reservoirs worldwide. Effective management and modeling rely on accurate quantification of toxin types and concentrations. Critically, bloom toxicity varies temporally and spatially, with different cyanobacterial species and strains producing different toxins or in different amounts. Spatially, blooms can be “patchy” in their toxicity, with “hot spots” of high toxin yield which can be a result of an accumulation of cells (e.g., scums) from wind action; higher toxin cell quotas; or a higher proportion of toxic strains (Puddick et al. 2016). In Lake Erie (Chaffin et al. 2021) microcystin concentrations were shown to vary by an order of magnitude (5–50 μg L−1), within a 5 km radius. There were also inconsistencies in ratios of chlorophyll a to microcystin, suggesting variation in the toxin cell quota, and the proportion of toxic to nontoxic cells, which ranged from 15.8% to 34% of the Microcystis population. In Lake Rotorua, New Zealand, spatial distribution of microcystin cell quotas, on two sampling times, varied 148- and 362-fold, with the increased cell quotas associated with high cell concentrations (Wood et al. 2021). In multiple reservoirs in France, spatial variation between reservoirs was attributed to strain differences in toxin cell quotas (0.01–3.4 pg cell−1), with toxic strains dominating the populations (70–100%) (Sabart et al. 2010). Comparison of multiple studies shows microcystin cell quota and yield can vary greatly between isolated strains and field samples (Fig. 1). Cyanobacteria species that bloom within the water column, rather than forming scums, such as Raphidiopsis spp, Aphanizomenon spp, and Umezakia ovalisporum, also have toxin yield variation across bloom samples (Fig. 2). In Lake Cobaki, Australia, a bloom of Raphidiopsis raciborskii had the highest CYN concentrations (101.4 μg L−1) in the hypolimnion, compared to surface waters (~ 10 μg L−1) (Everson et al. 2011). The higher concentrations did not correlate to higher cell concentrations, with authors suggesting either higher toxin production in the population, or decreased degradation. Comparison of R. raciborskii cell concentrations and toxin amount across 65 sites over 5.5 years found that cell counts did not reliably correlate with toxin concentrations (Veal et al. 2018). A study using field samples across Guangdong Province, China, found R. raciborskii populations comprised of both toxic and nontoxic strains (0.3–39% toxic strains), and suggested that higher nutrient levels promoted toxic strains (Lei et al. 2021). A mesocosm experiment found the proportion of toxic R. raciborskii strains increased with the addition of phosphorus, and phosphorus plus nitrate, within 5 days of the nutrient addition (Burford et al. 2014). The benthic toxic cyanobacterium Microcoleus (basionym Phormidium) is becoming widely recognized as a problematic bloom-forming cyanobacterium. Microcoleus forms extensive mats in cobble-bed rivers worldwide (Wood et al. 2020). It has been well-studied in New Zealand, where harmful blooms have increased rapidly over the past two decades, causing numerous dog deaths due to anatoxin poisoning. Studies showed that the benthic mats are a mixture of toxic and nontoxic strains (0–30.3% toxic; Kelly et al. 2018), with fine-scale variability of toxin content (Wood et al. 2010). Additionally, changes in the anatoxin content in M. autumnalis-dominated mats were primarily related to the dominance of toxic strains, with variability in toxin correlated to the proportion of toxic strains (Thomson-Laing et al. 2020). Among the toxic strains, there were 100-fold differences in the cell quota of anatoxin (Wood et al. 2012). These data indicate that both the relative abundance of anatoxin and non-anatoxin-producing genotypes, and variation in anatoxin-producing capability, can influence the overall toxin concentration of benthic Microcoleus mat samples. This leads to a wide temporal and spatial variation in risks from exposure to toxins (Heath et al. 2011; Kelly et al. 2018). To discover the drivers of bloom toxicity dynamics, researchers have studied the biosynthesis of cyanobacterial toxic metabolites, and the abiotic and/or biotic factors that affect toxin production rate within cells, rather than the differences between strains. The abiotic factors have focused on temperature, light, salinity, and nutrients, with biotic factors including microbiome composition and exposure to predators, such as Daphnia (Pearson et al. 2016). These factors represent the major influences on bloom development and persistence. Laboratory studies have focused on measuring changes in toxin cell quota and the rate of toxin production in monoclonal cultured strains, often using only one or two strains. These studies have found that environmental stressors and culture growth phases may have transient effects on the toxin production rate, with rate changes typically between two- and fourfold (Hellweger et al. 2022; Kramer et al. 2022; Lei et al. 2021; Lyck 2004; Martin et al. 2020; Silveira and Odebrecht 2021; Yang et al. 2016; Zhang et al. 2021). These two- to fourfold differences can often be attributed to differences in cell size, which halve or double during the cell cycle (Orr et al. 2018). In fact, under the abiotic and biotic conditions tested in laboratory experiments, on a per cell basis, toxins are frequently constitutively produced, that is, toxin cell quotas within a strain have low variability (e.g., Jacinavicius et al. 2019; Orr et al. 2018; Pierangelini et al. 2015; Willis et al. 2015). The majority of research investigating factors affecting toxin production in cyanobacteria has been on microcystin production by Microcystis, and cylindrospermopsin and saxitoxin by Raphidiopsis. Changes in microcystin quota by Microcystis have been dependably linked to both higher light and lower temperatures, that is, short-term oxidative stress (Zilliges et al. 2011). The microcystin cell quotas of Microcystis NIES-843 increased under lower temperature (18°C compared to 22–30°C) from 35 to 241 fg cell−1, that is, an approximately six- to sevenfold change (Peng et al. 2018), while the total microcystin cell quota of Microcystis PCC 7806 increased approximately twofold under lower temperatures (Martin et al. 2020; Roy et al. 2023; Stark et al. 2023). Importantly, there was a time-lag of 3 d between the change in environmental conditions and microcystin cell quota in continuous culture experiments (Martin et al. 2020). This time-lag indicates that the change in microcystin cell content (including production, turn-over, and destruction) is, at least partially, linked to cell-division. In comparison, the toxin cell quotas of a range of Microcystis strains have been shown to vary between 0.005 and 160 fg cell−1 (Lyck and Christoffersen 2003; Mowe et al. 2015), with surface scums having cell quotas of microcystin up to 30 pg cell−1 (Puddick et al. 2016). Cylindrospermopsin production by R. raciborskii has generally been found to be constitutively expressed (reviewed by Burford et al. 2016, Pearson et al. 2016), and more recent studies have failed to convincingly show changes in cell quota under different experimental conditions. The cell quota of cylindrospermopsin in R. raciborskii strains can vary greatly from 0.02 to 25 fg cell−1 (Nor et al. 2019; Willis et al. 2015). On the other hand, saxitoxin production by R. raciborskii strains in South America may vary up to approximately five- to sixfold in response to light, temperature, water hardness and salinity (Carneiro et al. 2013; Mesquita et al. 2019). Crucially, the changes in toxin production and cell quota found in strains under laboratory conditions are insignificant when compared with the variability in toxin cell quota found between strains of the same species. For example, toxin cell quotas of co-occurring strains of R. raciborskii varied by three- to sixfold, Microcoleus by 100-fold, and Microcystis aeruginosa by 362-fold, with cell quotas ranging from ~ 0.28 to 4.2 pg cell−1 (Willis et al. 2015, 2016; Wood and Puddick 2017; Wood et al. 2021). Additionally, the variation in cell quotas is not yet explained by genetic differences or cellular mechanisms. Variation in the arrangement of the toxin gene cluster is found between cylindrospermopsin-producing species (Cullen et al. 2025); however, within species and particularly strains within a population, the synteny appears conserved (Willis et al. 2018, 2022). Therefore, overall, strain variability in toxin cell quota is far higher than changes in the toxin production of a strain. This critical knowledge means that a rethink of our experimental approaches to improve prediction and monitoring of toxic blooms is needed. These laboratory findings are highly relevant to the field observations, which show that there can be significant spatial and temporal variability of toxin yield within and between blooms (Veal et al. 2018; Wood and Puddick 2017, Wood et al. 2021). As a result, there may be a weak correlation between cell densities of toxic species and toxin concentrations in field samples (John et al. 2024; Lei et al. 2021; Pacheco et al. 2016). Studies of blooms find multiple co-occurring strains, each with a different toxin cell quota and physiology (Burford et al. 2014; Halary et al. 2023; Jackrel et al. 2019; Willis et al. 2016). These physiological differences result in different growth responses of strains to environmental conditions (Dick et al. 2021; Xiao et al. 2020). Thus, fluctuations in environmental drivers lead to changing proportions of different strains in a population, which in turn lead to changes in toxin yields temporally or spatially across a bloom (Fig. 3; Burford et al. 2014; Kiledal et al. 2024; Li et al. 2017; Park et al. 2018; Ryderheim and Kiørboe 2024). Until now, sequencing and bioinformatic technology has not been able to differentiate strain diversity within a population, thus it has not been readily possibly to discern strain diversity from field samples or in laboratory studies, except in the case of comparing toxic and nontoxic strains. Competition experiments, with toxic and nontoxic strains of Microcystis and Raphidiopsis have demonstrated dynamics and shifts in strain proportions under different environmental conditions (Jiang et al. 2024; Kardinaal et al. 2007; Lei et al. 2015, 2021; Park et al. 2018; Suominen et al. 2016; Wang et al. 2019). In all studies, changes in nutrients (nitrogen and phosphorus), temperature, light, and CO2 affected the proportion of toxic to nontoxic strains. Population shifts are not always consistent between studies, further highlighting strain variation. Kiledal et al. (2024) used strain specific markers to identify the presence of multiple toxic strains in Lake Erie, with each sampling time having a different proportion of strains, demonstrating the potential of characterizing population shifts between toxic strains. Importantly, the potential magnitude of changes in bloom toxin yield, based on the proportions of different strains, is much greater than changes in bloom toxin yield based on changes in the toxin quota within a strain. To further understand the drivers of toxin yield in cyanobacteria blooms we need to focus on species intraspecific diversity and population dynamics between strains. In experimental design and analysis, we need to be particularly mindful that interpretation of data on cyanobacterial toxins can be confounded by the use of either per cell quotas, biovolumes or volumes, or dry weight, and is affected by cell division rates, cell viability and lysis, and by the proportion of intracellular vs. extracellular toxins. Standardizing the way toxin data is expressed is of utmost importance, as each method can lead to a different interpretation of results (Pierangelini et al. 2015; Martin et al. 2020). For example, using R. raciborskii CS-506, Pierangelini et al. (2015) found a statistically significant difference in the toxin cell quota between low and high light conditions; however, when cell size was taken into account and toxin was determined per biovolume, there was no statistical differences between the conditions. For Microcystis, cell size also changes with conditions, and in Martin et al. (2020) the toxin quota is expressed as per biovolume, demonstrating that larger cells resulted in an increase in microcystin quota under lower temperatures. In addition, it is also important to consider the rate of production, and its relationship to the rate of cell division: as explained in Orr et al. (2018). This is a critical factor in determining constitutive production. Studying the role of environmental conditions in driving changes in strain proportions in cyanobacterial populations is critical, as is attaining a greater understanding of the role that cyanotoxins play in cyanobacterial biology. These future studies should improve our understanding of the features that allow cyanobacteria to be so successful and persist across many environments, increase our ability to accurately predict blooms, and discover new methods for the reduction of toxic blooms. Embracing this new approach will stimulate a new era of cyanobacterial research. Anusuya Willis and Michele A. Burford contributed equally to this article. Anusuya Willis and Michele A. Burford thank Seqwater for supplying the monitoring data. There is no funding associated with this paper. None declared. Data are available in the CSIRO DAP repository at: https://doi.org/10.25919/z3b1-hg49.