An efficient and affordable laboratory method to produce and sustain high concentrations of microcystins by Microcystis aeruginosa

Graphical abstract


Background
Microcystis is a genus of freshwater cyanobacteria that is frequently found in meso-to eutrophic waters. Their phylum originated some 3 billion years ago [1,2], and their photosynthetic activity prompted the oxidation of the Earth's atmosphere [3]. Throughout time they have successfully adapted to extreme environments through processes that have allowed them to flourish in the face of increasing human impact on aquatic ecosystems [4]. Today, Microcystis spp. are cosmopolitan and can pose risks to human health when freshwaters affected by blooms are used for drinking, irrigation, fishing, and recreational purposes [1,5]. The distribution, frequency and intensity of blooms have increased in response to climate change factors such as elevated temperatures, CO 2 levels, and eutrophication [6,7] which have increased risks to humans and aquatic biota. In a global analysis conducted by Harke et al. [8] on the present state of knowledge of Microcystis (e.g. geographic distribution, toxins, genomics, phylogeny, and ecology), blooms were reported in 108 countries, of which 79 confirmed the presence of the potent hepatotoxin microcystin.
Microcystins are cyclic heptapeptides capable of causing death to humans [9], animals [10], and aquatic invertebrates [11,12]. Upon ingestion by humans and animals, microcystins are transported to the liver where they cause toxicity by inhibiting protein phosphatases 1 and 2A [13]. The inhibition of these protein phosphatases can cause cytoskeletal degradation and breakdown of hepatocytes, causing blood to pool in the liver that is followed by hemorrhaging and organ failure [14]. Over 250 different microcystin congeners have been identified [15] whose toxicity is affected by the variation in molecular structure [16,17] and congener concentration within cyanobacterial cells is affected by changes in environmental factors including light intensity [18][19][20], pH [18], temperature [18,20], and nutrient concentrations [16,20,21]. Frequently detected and thoroughly studied among these congeners is microcystin-LR (CAS: 101043-37-2, C 49 H 74 N 10 O 12 ), which is regarded as one of the most toxic forms of microcystins [8]. In fact, the widespread occurrence of microcystin-LR led the World Health Organization to institute a guideline value of 1 mg L À1 in drinking water [22], which is in effect in most countries.
Increasing concern over impacts from cyanobacterial harmful algal blooms has highlighted the need for studies that aim to better understand the production and dynamics of algal toxins. Culture studies continue to lay the groundwork for characterizing the relationship between Microcystis cell growth and microcystin toxin production over short periods of time (typically between 10-30 days) [16,19,[23][24][25]. To support these studies, new culture techniques that can yield and sustain high concentrations of microcystins from Microcystis aeruginosa have not been explored and are of great interest. Currently, the costs to purchase microcystin analytical standards at >95 % purity can be prohibitive and may hinder researchers from designing large-scale and long-term exposure experiments. Moreover, recent work by Janssen [26] suggests that in addition to microcystins other bioactive metabolites (e.g. cyanopeptolins, aerucyclamides, aeruginosines, microginins) can also contribute to the overall toxicity of toxigenic strains of cyanobacteria found in the environment. Validating a new culture method for M. aeruginosa that retains microcystins and other bioactive metabolites, and demonstrating its advantage over existing techniques can, therefore, serve as a stepping-stone for researchers who would like to perform mechanistic studies, which ultimately will better define exposure-response relationships and improve our understanding of the risks posed to aquatic biota exposed to microcystins.
Here we describe a simple culture technique that circumvents the costs to purchase analytical standards and supports the production of high concentrations of microcystin-LR and its demethylated counterpart [D-Asp 3 ]-microcystin-LR produced by M. aeruginosa. To validate our method we: 1) measured cell concentrations using traditional approaches; 2) measured the concentrations of microcystins using a high throughput method based on liquid chromatography coupled to high resolution mass spectrometry that is rapid, precise, has low detection limits and can quantify each microcystin congener; 3) measured the concentrations of cell pigments (i.e. phycobiliproteins and chlorophyll a) using a spectrophotometric method in order to better understand any physiological changes occurring during growth of M. aeruginosa; and 4) calculated the rates and maximum concentration of microcystin congener production per M. aeruginosa cell (referred to as cell quotas). The significance of this method is that we were able to: 1) yield microcystins at concentrations that are environmentally relevant to freshwaters around the world (e.g. 1-300 mg L À1 ); 2) maintain these concentrations without resupplying fresh medium (further reducing costs); and 3) utilize rate equations to model the relationship between the environmental conditions in the cultures and changes occurring within the M. aeruginosa cells.
Culturing of Microcystis aeruginosa CPCC 300 M. aeruginosa CPCC 300 was grown in a modified recipe of the Blue-Green-11 (BG-11) medium [27]. To prepare this medium, the BG-11 Trace Metals Solution (Table S1) was prepared by adding each of the components to approximately 900 mL of dH 2 O while continuously stirring, bringing the total volume to 1 L. Next, the BG-11 Vitamin B 12 + Biotin Solution (Table S2) was prepared by adding 0.05 mL of Vitamin B12 and 0.5 mL of Biotin to approximately 50 mL of dH 2 O [28,29]. Thiamine hydrochloride was not added to this solution. Lastly, the BG-11 liquid media (Table S3) was prepared by adding each of the components to approximately 900 mL of dH 2 O while continuously stirring, bringing the total volume to 1 L. The pH was then adjusted to 7.5 by adding drops from 10 mM of HCl as necessary. This procedure was repeated until a sufficient volume of BG-11 medium was prepared for the method validation, upon which contents were transferred to flasks and sterilized in an autoclave for 60 min at 121 C. After autoclaving, flasks were brought back to room temperature for a cool down period, after which 350 mL of BG-11 medium was inoculated with an initial cell concentration of 1. Fluorescent bulbs were 75 cm above the flasks (Fig. 1).

Experimental design
The experimental cultures were monitored for 46 d.  culturing M. aeruginosa and were therefore chosen for the method validation [20,30]. Each flask was inoculated with approximately 40 mL of M. aeruginosa CPCC 300 from the stock cultures, the lid was tightly covered with aluminum foil, and maintained under the same environmental conditions as outlined previously for the stock cultures. Approximately midpoint during the experiment (Day 21), lids for all flasks were loosened to allow for greater atmospheric exchange to prevent cultures from being stressed. Samples were collected until there remained approximately 50 % of the original volume in each flask. On each sample collection day, and from each flask, samples were withdrawn 2 h after the light phase had commenced to estimate cell, microcystins and cellular pigment concentrations. Samples were collected under a Microzone Inc. CLASS II Type A2 Biological Safety Cabinet (Ottawa, ON, Canada) to ensure sterilization was maintained while M. aeruginosa CPCC 300 was being handled. To ensure homogenization prior to sample collection, flasks were swirled for 10 s. Flasks were rearranged randomly daily to reduce effects caused by minor differences in photon irradiance (Fig. 1).

Measurement of cell concentrations by hemocytometer
To obtain an accurate estimate of cell concentrations, two 10 mL aliquots were collected and cell counts performed within an hour. Cell concentrations were determined by loading 10 mL on each grid of a Hausser Scientific Bright-Line TM Hemocytometer (Horsham, PA, United States) with Propper 1 rectangular hemocytometer cover glass (Long Island City, NY, United States) and observing it using an Olympus 1 B071 BH-2 Series System Microscope (Richmond Hill, ON, Canada) at 10Â magnification. The number of cells was recorded for each grid, and cell concentrations were calculated accordingly. When necessary, a dilution factor was applied. No intact dead cells were observed throughout the duration of the experiment (Fig. 1).

Measurement of microcystins
To obtain measurements of microcystin concentrations, 5 mL was withdrawn from each flask at each time point and stored in 15 mL Corning 1 polypropylene centrifuge tubes at À80 C for later analysis. Microcystins were analyzed using on-line solid phase extraction coupled to liquid chromatography-quadrupole time-of-flight high resolution mass spectrometry (Waters Xevo G2-XS, Milford, MA, USA), optimized by Ortiz et al. [31]. In this targeted and non-targeted high throughput method, samples of M. aeruginosa CPCC 300 at each time point were measured and 12 microcystin variants (LR, YR, RR, HtyR, HilR, WR, LW, LA, LF, LY, Dha 7 -LR, and Dha 7 -RR) and anatoxin-A (all standards purchased from Enzo Life Sciences, Farmingdale, NY, USA) were quantified using nodularin as the internal standard. Using this high throughput method, batches of 50 samples can be prepared for instrumental analysis in less than 3 h. Detection limits were 0.05 mg L À1 with an expanded uncertainty ranging from 4 to 14 % for the different variants, which takes into account the uncertainty coming from the sample preparation, the instrument, and calibration standards. The microcystin cell quota (Q mcyst ), expressed as femtograms (fg) microcystin per cell, was calculated by dividing the concentration of each detectable microcystin congener (fg mL À1 ) by the cell concentration (cells mL À1 ) (Fig. 1).

Measurement of phycobiliproteins and chlorophyll a
For cellular pigment determinations, 5 mL was withdrawn for analysis of phycobiliproteins and chlorophyll a by ultraviolet-visible spectrophotometry. Phycobiliprotein and chlorophyll a content was quantified using a Molecular Devices LLC SPECTRAmax PLUS 384 Microplate Reader (San Jose, CA, United States) with SoftMax 1 Pro 6.3, Microplate Data Acquisition and Analysis Software (Sunnyvale, CA, United States). Optical density (OD) between 200-800 nm was captured, and the following absorbance maxima were recorded for each class of phycobiliproteins: allophycocyanin (l max = 650 nm, purple); phycocyanin (l max = 610-625 nm, blue); and, phycoerythrin (l max = 550-565 nm, red). The OD for chlorophyll a was measured at l max = 665 nm. Phycobiliprotein and chlorophyll a concentrations were calculated at 650 nm for allophycocyanins (AP), 620 nm for phycocyanins (PC), 565 nm for phycoerythrins (PE); and, 665 nm for chlorophyll a (chl a) using associated Eqs. (1),(2), and (3) described by Tandeau de Marsac and Houmard [30] and (4) described by Tailing and Driver [32] (Fig. 1).
The pigment cell quota (Q pigment ) -expressed as picogram (pg) pigment per cell-was calculated by dividing the concentration of each pigment (pg mL À1 ) by the cell concentration (cells mL À1 ).

Calculation of rates
First order rate kinetics provided by Orr et al. [25] were used to model the relationship between the environmental conditions in the cultures and the physiological changes in M. aeruginosa cells and production of microcystins (Fig. 1).
Using the first order rate constant (m), the initial concentration (C 0 ), the final concentration (C 1 ), and time intervals (t 0 and t 1 ), the rate of cell growth (m growth ) and rate of microcystin congener production (m mcyst ) were calculated using Eq.
The relationship between the rate of cell growth (m growth ) and rate of microcystin congener production (m mcyst ) were further investigated using Eqs. (6), (7), and (8) Eq. (6) declares the rate of microcystin congener production (m mcyst ) between t 0 and t 1 is slower than the rate of cell growth (m growth ), where 0.5 means no production is occurring and the existing microcystins are being diluted as cells continue to grow (i.e. cells are becoming less toxic).
Eq. (7) declares the rate of microcystin congener production (m mcyst ) between t 0 and t 1 is equal to the rate of cell growth (m growth ).
Eq. (8) declares the rate of microcystin congener production (m mcyst ) between t 0 and t 1 is faster than the rate of cell growth (m growth ) (i.e. cells are becoming more toxic).

Measurement of nutrients and metals
Samples of BG-11 medium were submitted to Ontario's Ministry of the Environment, Conservation and Parks (MECP) Laboratory Services Branch (Etobicoke, ON, Canada) for nutrient and metals analysis (Table S4).

Data analysis
To characterize changes in the concentration of cells, toxins and pigments, linear and nonlinear models were applied to all data (i.e. t 0 to t 46 ) and the Akaike Information Criterion (AICc) was used to determine the best fit. Based on the lowest AICc, a nonlinear regression exponential growth, single, 3 parameter equation (f = y 0 + ae bx ; α = 0.05) was chosen to describe changes in cell concentration (cells mL À1 ), microcystin congener concentration (mg L À1 ), pigment concentration (mg mL À1 ), microcystin cell quota (Q mcyst , fg cell À1 ), and pigment cell quota (Q pigment , pg cell À1 ) over time. Correlations between rates (m) and cell quotas (Q) were also calculated using Pearson's product moment correlation coefficient (α = 0.05). Data were divided into three groups (t 0 to t 11 , t 12 to t 21 , and t 25 to t 46 ) demarcated by distinct increases or decreases in the relative rates of cell growth (m growth ) and microcystin congener production (m mcyst ) between t 0 to t 46 to account for possible changes in cell physiology that may have occurred over the course of the method validation. Data was tested for normality using the Shapiro-Wilk's test (α = 0.05). When normality failed (p < 0.05), Kruskal-Wallis One Way Analysis of Variance (ANOVA) on Ranks (α = 0.05) was conducted between the three populations for the rates of cell growth (m growth ), microcystin congener production (m mcyst ), and pigment production (m pigment ). Linear regression was also performed to measure the extent that there was a linear relationship between the independent variable (m growth ) and dependent variables (m mcyst , m pigment ). When the Shapiro-Wilk's test failed (p < 0.05), a post hoc non-parametric pairwise multiple comparisons procedure was conducted using Dunn's method. Statistical analyses were performed using Sigma Stat (Version 4.0, Systat Software, San Jose, CA, US).

Method validation
Cell concentrations (cells mL À1 ) increased in all cultures from t 0 to t 11 , with growth rates slowly decreasing to near zero by t 11 . From t 12 to t 21 the average growth rate was 0.02 cells d À1 and there was a gradual decline in cell concentrations in all cultures. From t 25 to t 46 , growth rates increased, ranging between 0.03 and 0.17 cells d À1 . By t 46 the cultures had a fresh blue-green colour and maintained this colour for 6 months. After t 11 , some divergence between flasks was observed but the maximum difference between the greatest and smallest cell counts was <2 % (Fig. 2). Although changes in which flasks contained the greater number of cells differed between t 11 and t 46 , this had little effect on microcystin concentrations, which was significantly and positively correlated with cell concentrations ( r 2 = 0.970, p < 0.005). Collectively, this indicates consistent performance (i.e. growth and toxin production) of the cultures over the entire duration of the experiment.
Microcystin concentrations (mg L À1 ) also increased in all cultures from t 0 to t 11 ; however, unlike cell concentrations there was a briefer plateau (t 12 to t 14 ) followed by a short increase (t 15 to t 19 ) and then decrease (t 20 to t 21 ) in microcystin content (Fig. 3). Notwithstanding, the cultures experienced a recovery and increased after t 21 . Chemical analyses also confirmed that microcystins were intracellular throughout the duration of the experiment, and cell counts by hemocytometer did not detect dead cells, suggesting that despite the brief plateau observed between t 12 to t 19 M. aeruginosa cells were viable and intact. As for the predominance of microcystin congeners: from t 0 to t 11 the concentration of microcystin-LR was higher than [D-Asp 3 ]-microcystin-LR; from t 12 to t 21 both congeners were approximately equal; and, from t 25 to t 46 [D-Asp 3 ]-microcystin-LR exceeded microcystin-LR. Microcystin cell quotas (Q mcyst ) also exhibited comparable trends in the shift of predominance in microcystin congeners: from t 0 to t 11 the average ratio of Q mcyst-LR :Q mcyst-dmLR was 16:13 fg microcystin cell À1 ; from t 12 to t 21 the average ratio of Q mcyst-LR :Q mcyst-dmLR was 15:15 fg microcystin cell À1 ; and, from t 25 to t 46 the average ratio of Q mcyst-LR :Q mcyst-dmLR was 10:15 fg microcystin cell À1 (Fig. 4). Further, a significant negative correlation between microcystin-LR cell quotas (Q mcyst-LR ) and cell concentrations was found from t 0 to t 11 ( r 2 = 0.663, p = 0.0139), no significant correlation was found from t 12 to t 21 ( r 2 = 0.116, p = 0.455), and a significant negative correlation was found from t 25 to t 46 ( r 2 = 0.498, p = 0.0225). For correlations between [D-Asp 3 ]-microcystin-LR cell quotas (Q mcyst-dmLR ) and cell concentrations, a significant negative correlation was found from t 0 to t 11 ( r 2 = 0.490, p = 0.0535) but not from t 12 to t 21 ( r 2 = 0.100, p = 0.489) or t 25 to t 46 ( r 2 = 0.047, p = 0.549). This shows that while in the first half of the experiment Q mcyst-LR and Q mcyst-dmLR decreased as cell concentrations increased (i.e. cell growth outpaced toxin production), in the second half of the experiment Q mcyst-LR continued to exhibit a negative correlation with cell concentrations while Q mcyst-dmLR had no significant correlation with cell concentrations. The mean pigment cell quota (Q pigment ) experienced similar trends to the mean microcystin cell quota (Q mcyst ) throughout the experiment (data not shown). Among all pigments, chlorophyll a cell quota (Q chlorophylla ) showed a significant linear correlation with Q mcyst-LR ( r 2 = 0.350, p < 0.001) and Q mcyst-dmLR ( r 2 = 0.240, p = 0.003) from t 0 to t 46 .
Correlation of the rates of cell growth (m growth ), microcystins (m mcyst ) and cell quotas (Q mcyst ) of the cultures were determined for the three time periods: from t 0 to t 11 , t 12 to t 21 , and t 25 to t 46 , respectively ( Table 1). From t 0 to t 11 , there is no correlation between m growth and Q mcyst-LR ( r 2 = 0.001, p = 0.942) and Q mcyst-dmLR ( r 2 = 0.002, p = 0.926). However, from t 12 to t 21 there is a significant negative linear correlation between m growth and Q mcyst-LR ( r 2 = 0.676, p = 0.023) and Q mcyst-dmLR ( r 2 = 0.676, p = 0.023). As the cultures progressed into t 25 to t 46 , Q mcyst-LR resumed a very weak correlation with m growth ( r 2 = 0.032, p = 0.622) while Q mcyst-dmLR maintained a significant negative linear correlation m growth ( r 2 = 0.446, p = 0.035). The correlation between Q mcyst-LR and Q mcyst-dmLR also varies at each time point: from t 0 to t 11 there was a significant linear correlation ( r 2 = 0.929, p < 0.001); from t 12 to t 21 there was another significant linear correlation ( r 2 = 0.996, p < 0.001); and, from t 25 to t 46 there was a weak correlation ( r 2 = 0.204, p = 0.189). This indicates two findings: 1) as m growth decreased from t 12 to t 21 Q mcyst increased; and 2) the relationship between Q mcyst-LR and Q mcyst-dmLR becomes uncorrelated after t 21 when the lid for each culture flask was loosened to allow for greater atmospheric exchange.
Application of first order rate kinetics to assess the relationship between the production of microcystin-LR (m mcyst-LR ) and [D-Asp 3 ]-microcystin-LR (m mcyst-dmLR ) over cell growth (m growth ) was calculated from t 1 to t 46 (Table S5). Of interest is between t 12 to t 19 when M. aeruginosa cells were exhibiting symptoms of stress. As m mcyst-LR and m mcyst-dmLR increased at similar rates from t 12 to t 19 , [ ( F i g . _ 2 ) T D $ F I G ] microcystin-LR cell quota (Q mcyst-LR ) and [D-Asp 3 ]-microcystin-LR (Q mcyst-dmLR ) also increased. This suggests that during this time period M. aeruginosa cells were stressed and toxin production outpaced cell growth. However, after t 21 -when the lid for each culture flask was loosened to allow for greater atmospheric exchange-m mcyst-LR and m mcyst-dmLR began to stabilize at rates similar to cell growth. Furthermore, as the rate of microcystin production began to stabilize with cell growth, m mcyst-dmLR surpassed m mcyst-LR .

Additional information
For over 60 years, researchers have attempted to decipher the conditions that enable the production and cessation of microcystins by freshwater cyanobacteria [23]. Numerous reports affirm the production of microcystins by M. aeruginosa is linked to physiological and environmental processes including cell division [25,33], light intensity and photoperiod [18][19][20]34], temperature [18,20], ultraviolet [35], and the availability of nutrients [16,20]. To advance further understanding, the need to optimize conditions for culturing M. aeruginosa to produce and sustain high concentrations of microcystins, while being cost-effective, is an important challenge facing researchers. In this experiment, the culture conditions adopted and strain of M. aeruginosa chosen  (196.94 mg L À1 ) for 46 d in the laboratory. Despite popular claims that metabolic end products will deteriorate the growth of cells in laboratory cultures, our results corroborate with those of Lyck [19] who found that this does not hold true for M. aeruginosa strains CYA 228 and PCC 7806. The strain selected for our method validation, M. aeruginosa CPCC 300, has shown it requires low maintenance and does not need a resupply of fresh medium in order to consistently produce microcystins at concentrations that are relevant both to freshwaters that regularly experience Microcystis blooms (i.e. 1-300 mg L À1 ) and the guideline value of 1 mg L À1 in drinking water instituted by the World Health Organization. In others experiments conducted by our laboratory, M. aeruginosa CPCC 300 has also shown the potential to: 1) produce microcystin-LR and [D-Asp 3 ]-microcystin-LR concentrations totaling as high as 1200 mg L À1 ; 2) maintain these concentrations for up to 6 months in stationary phase without resupply of fresh medium; and 3) remain stable for over 18 months when frozen at À80 C (R.S. Shahmohamadloo, University of Guelph, Guelph, ON, Canada, unpublished data). This information will be important for researchers who would like to stockpile fresh or frozen microcystins for future large-scale and long-term experiments.
While M. aeruginosa CPCC 300 has relatively high microcystin cell content with stable proportions of microcystin-LR and [D-Asp 3 ]-microcystin-LR [34], our findings indicate a decline from t 12 to t 21 and a shift in the proportion of congeners from t 25 to t 46 . Certain parameters can be ruled out to guide our postulation for this decline and shift in microcystins: light intensity, photoperiod, and temperature. These parameters were constant throughout the method validation and correspond with freshwaters conditions during a typical harmful algal bloom season [34]. Light limitation due to self-shading of M. aeruginosa cells was also considered, however given the cell concentrations increased markedly after t 21 it is unlikely this parameter caused a decrease in the concentration of microcystins. Another factor that can be ruled out is lack of ultraviolet light because the fluorescent bulbs were selected that include UV light and mimic the wavelength range that is needed for M. aeruginosa cell pigments (phycobiliproteins and chlorophyll a) to grow. Further, it is unlikely that nutrient limitation was a factor; if this were the case, we would have seen a continued decline in cell, microcystin, and pigment concentrations after t 21 . It is important to note that the experimental cultures were not resupplied with fresh medium, and subsamples were regularly taken out, thus causing a decrease in nutrients (e.g. nitrogen, phosphorus) over the course of the experiment. If nutrient limitation were a cause for the decline in microcystin concentrations seen from t 12 to t 21 , it should have continued to t 46 . Instead, the experimental cultures responded favourably to the perturbation on t 21 when atmospheric carbon exchange increased, and recovered, albeit not at the same rate of microcystin production (m mcyst ) but along similar patterns of growth. This points to the role of dissolved inorganic carbon (e.g. carbon dioxide, CO 2 ; carbonic acid, H 2 CO 3 ; bicarbonate, HCO 3 À ; and carbonate, CO 3 2-) as the parameter that likely caused the decline of microcystins from t 12 to t 21 .
Two studies have previously investigated the effects of inorganic carbon concentrations on the relative abundance of microcystin congeners [36,37]. Sandrini et al. [36] investigated the response of M. aeruginosa strain PCC 7806 to rising levels of CO 2 , and how the production of microcystins and cell physiology is altered in the presence of elevated CO 2 partial pressure (pCO 2 ). Many cyanobacteria including Microcystis have evolved highly advanced CO 2 -concentrating mechanisms that include CO 2 and bicarbonate uptake systems [3,36]. These mechanisms are important for Microcystis to grow and spread rapidly in supersaturated lakes that contain high levels of dissolved CO 2 , sometimes exceeding equilibrium with the atmosphere [38]. To test this hypothesis, Sobek et al. [38] cultured M. aeruginosa strain PCC 7806 in controlled laboratory chemostats that were changed from low pCO 2 (200 ppm) to high pCO 2 (1450 ppm) conditions that are representative of supersaturated lakes. Results confirmed Table 1 Correlation analyses between rates of cell growth (m g ) and microcystins (m mcyst ) and cell quotas (Q mcyst  that M. aeruginosa cells benefited strongly from increased dissolved CO 2 . Elevated CO 2 levels alleviated M. aeruginosa cells from inorganic carbon limitation, which led to a higher population abundance, increased buoyancy in the cells, and about a 2.5-fold increase in microcystin cell quota (Q mcyst ).
Another study used a similar set-up but instead provided nitrogen-limited conditions and exposed three strains of M. aeruginosa to low pCO 2 (400 ppm) and high pCO 2 (1200 ppm); Liu et al. [37] hypothesized that elevated pCO 2 would increase the cellular C:N ratios in M. aeruginosa, and that a higher C:N ratio would promote the production of the more toxic microcystin congeners. Results confirmed that microcystin congeners with higher C:N ratios (e.g. microcystin-LW, microcystin-LF) increased with elevated pCO 2 , and vice versa (e.g. microcystin-RR, microcystin-LR). Liu et al. [37] suggest that Microcystis may preferentially produce more toxic microcystin congeners in elevated pCO 2 under nitrogen-limited conditions. Our results corroborate these findings. In fact, when viewed from t 0 to t 46 , the microcystin cell quotas for microcystin-LR (Q mcyst-LR ) and [D-Asp 3 ]-microcystin-LR (Q mcyst-dmLR ) were highest between t 12 to t 21 when the C:N ratio was assumed to be lower than between t 0 and t 11 . This further confirms that nitrogen was not a limiting nutrient in our experiment and did not inhibit the production of microcystins.
Although it is well documented in the literature that CO 2 can influence the production of microcystins by M. aeruginosa, as far as we can ascertain no studies have evaluated the role of carbon in relation to congener production. While serendipitous, our results demonstrate that carbon limitation can influence the production of [D-Asp 3 ]-microcystin-LR over microcystin-LR when other relevant environmental parameters are constant (Fig. 4). Our findings contrast with the concluding remarks of Orr et al. [25] who suggest that microcystin production is not triggered by environmental stressors. Environmental stressors are believed to drive changes in Microcystis bloom toxicity by affecting the rates of cell growth (m growth ) and microcystin cell quotas (Q mcyst ), but do not trigger any effect on the pathway in which microcystin congeners are produced [25]. We provide evidence that under carbonlimited conditions the following can occur: as m growth decreases, Q mcyst increases ( r 2 = 0.676, p = 0.023); the relationship between Q mcyst-LR and Q mcyst-dmLR will be tightly correlated in carbon-stressed situations ( r 2 = 0.996, p < 0.001), however after this stress is relieved and carbon is no longer limited, Q mcyst-LR and Q mcyst-dmLR will be uncorrelated (r 2 = 0.204, p = 0.189); and, the average ratio of Q mcyst-LR : Q mcyst-dmLR can shift in dominance depending on the extent to which atmospheric carbon exchange is limited, which in our case showed a shift from 16:13 fg microcystin cell À1 (t 0 to t 11 ) to 15:15 fg microcystin cell À1 (t 12 to t 21 ) to 10:15 fg microcystin cell À1 (t 25 to t 46 ). However, given the known abundance of CO 2 in freshwaters, field data is a necessary point for comparison to assess the relevance of our discovering a shift in microcystin congener production. To substantiate our discovery, we received water samples from the Great Lakes and inland waters across Ontario, Canada, during harmful algal blooms season from 2016 to 2018. We postulated that if microcystin-LR and [D-Asp 3 ]microcystin-LR were present in samples, the ratio of microcystin-LR would be greater because atmospheric carbon exchange was abundant. Results confirmed the presence of both microcystin congeners across the samples received, and found the concentration of microcystin-LR was higher than [D-Asp 3 ]-microcystin-LR after quantification by on-line solid phase extraction coupled to liquid chromatography-quadrupole time-of-flight high resolution mass spectrometry (R.S. Shahmohamadloo, University of Guelph, Guelph, ON, Canada, unpublished data). Our field results highlight the need to maintain culture conditions in order that M. aeruginosa can produce microcystin-LR and [D-Asp 3 ]microcystin-LR at ratios comparable to those found in the environment. To ensure uninterrupted growth of M. aeruginosa and high concentrations of microcystins in the laboratory, we recommend experimental cultures flasks be sufficiently ventilated. Cell culture flasks with vented caps -filled no more than 50 % of the flask volume to allow for sufficient air exchange-are an excellent and costeffective approach to maintaining cell growth and producing microcystins at a range between 300 to 1200 mg L À1 , as has been the case in our laboratory (R.S. Shahmohamadloo, University of Guelph, Guelph, ON, Canada, unpublished data). Furthermore, cell culture flasks with vented caps provide greater stability (i.e. little to no stress) and reduce the variability in production of microcystins among replicates when compared to Erlenmeyer flasks, as was the case in our experiment.
Failure to allow for sufficient atmospheric carbon exchange within culture flasks can increase the overall toxicity of M. aeruginosa CPCC 300. In one study, Shimizu et al. [39] assessed the cytotoxicities of 16 microcystin congeners on primary cultured rat hepatocytes by measuring cellular adenosine triphosphate content, followed by calculating the inhibitory concentration at 50 % (IC 50 ). Results showed [D-Asp 3 ]-microcystin-LR was approximately four times more cytotoxic (IC 50 = 0.217 mg mL À1 ) than microcystin-LR (IC 50 = 0.800 mg mL À1 ) to rat hepatocytes. Although lethal dose toxicity information is currently only available for microcystin-LR (LD 50 = 50 mg kg -1 intraperitoneal injection in mice, [37]), the potential exists for [D-Asp 3 ]-microcystin-LR to inflict with greater effect liver and kidney damage, tumour promotion, gastroenteritis, reduced DNA repair and reproductive toxicity in mammals and humans [3]. If this is the case, strains of Microcystis that are similar to M. aeruginosa CPCC 300 may have the ability to favour the production of [D-Asp 3 ]-microcystin-LR in low pCO 2 (<400 ppm) and microcystin-LR in high pCO 2 (>1200 ppm) environments, respectively, which has implications for the overall toxicity of microcystins in experiments conducted in the laboratory and field.

Conclusion
This method presents a cost-effective culture technique that can yield and sustain high concentrations of microcystin-LR (120.75 mg L À1 ) and [D-Asp 3 ]-microcystin-LR (196.94 mg L À1 ) produced by M. aeruginosa CPCC 300 when relevant environmental conditions are controlled in the laboratory. In the face of adversity from t 12 to t 21 , we discovered M. aeruginosa CPCC 300 are resilient in carbon-limited situations and may respond to stress by shifting the ratio of microcystin congeners. Future work is needed to explore the impact of nutrient modifications on the profile of microcystin congeners produced by cyanobacterial species, using both laboratory cultures and field samples. Given the variability in toxicity of microcystin congeners, a greater understanding of which and how environmental variables can regulate the abundance of microcystins will help predict the periods of greatest risk to users of freshwaters who may be exposed to these toxins through cyanobacterial harmful algal blooms [16,24,25]. We recommend the use of M. aeruginosa CPCC 300 as a reliable strain that can yield microcystins at concentrations that are environmentally relevant to freshwaters (e.g. 1-300 mg L À1 ). Adopting this culture technique and strain of M. aeruginosa can support the work of future researchers who require a continual supply of microcystin congeners to perform toxicological studies on aquatic biota at concentrations relevant to freshwater ecosystems.
Author's contributions R.S.S., R.C., D.G.P. and P.K.S. conceived and designed the experiment. R.S.S. collected the data. R.S.S., X.O., C.H. analyzed the data. R.S.S. wrote the paper. All authors read, amended, and approved the final manuscript.