The effects of salinity and N : P on N‐rich toxins by both an N‐fixing and non‐N‐fixing cyanobacteria

Freshwater ecosystems are experiencing increased salinization. Adaptive management of harmful algal blooms (HABs) contributes to eutrophication/salinization interactions through the hydrologic transport of blooms to coastal environments. We examined how nutrients and salinity interact to affect growth, elemental composition, and cyanotoxin production/release in two common HAB genera. Microcystis aeruginosa (non‐nitrogen [N]‐fixer and microcystin‐LR producer [MC‐LR]) and Aphanizomenon flos‐aquae (N‐fixer and cylindrospermopsin producer [CYN]) were grown in N : phosphorus (N : P) 4 and 50 (by atom) for 21 and 33 d, respectively, then dosed with a salinity gradient (0–10.5 g L−1). Both total MC‐LR and CYN were correlated with particulate N. We found Microcystis MC‐LR production and release was affected by salinity only in the N : P 50 treatment. However, Aphanizomenon CYN production and release was affected by salinity regardless of N availability. Our results highlight how cyanotoxin production and release across the freshwater–marine continuum are controlled by ecophysiological differences between N‐acquisition traits.

Freshwater salinization is a global threat to water quality due to a variety of anthropogenic causes (Kaushal et al. 2018). Model projections show that sea level is expected to rise due to melting ice caps (Nerem et al. 2018), and there is concern that encroaching saltwater will impact freshwater ecosystems. Salt encroachment is also occurring in inland lakes of northern latitudes where road salt is commonly applied for automobile safety. In a study of over 49,000 lakes (USA), 18% of lakes were at risk of negative effects due to salinization (Dugan et al. 2020). Increasingly saline lakes are at risk of shifting phytoplankton taxa from Chlorophyte dominance to Cyanophyte dominance (Ballot et al. 2009) or having higher proportions of filamentous cyanobacteria (Greco et al. 2021), thereby increasing the risk of toxic cyanobacterial blooms. However, the direct influence of salinization on the frequency, magnitude, and duration of harmful algal blooms (HABs) remains understudied.
Another emerging mechanism regarding the effect of salinization on HABs is the potential hydrologic transport of freshwater biological communities to saline coastal habitats. One management strategy to mitigate inland water HABs is to flush a bloom to downstream estuary or marine environments (Lundgren et al. 2013;Paerl et al. 2016;Grover et al. 2017). These flushing events move the problem downstream, with coastal areas experiencing exposure to freshwater cyanotoxins (Tatters et al. 2017) at levels that can be detected in shellfish (Miller et al. 2010). Exposing HABs to brackish or saline water could cause cell lysing, releasing cyanotoxins into the environment. A recent study revealed that HAB communities release cyanotoxins when exposed to salinities greater than 18 g L À1 (Rosen et al. 2018). Flushing HABs to coastal environments may cause extreme exposure events along the freshwater-marine continuum as toxic cyanobacteria lyse and release toxins into the water.
Cyanobacteria can be broadly categorized into two distinct functional groups: nitrogen (N)-fixers and non-N-fixers (Whitton 2012). Numerous genera within both groups have the capacity to synthesize N-rich toxins that are generally sensitive to the availability of N during growth and toxin production (Gobler et al. 2016;Huisman et al. 2018;Wagner et al. 2019). Microcystis is a colonial, non-N-fixing cyanobacteria capable of producing the hepatotoxin microcystin (MC), which exists as many different congeners (Díez-Quijada et al. 2019). Microcystis is capable of growth and cyanotoxin production in the presence of salt (Tonk et al. 2007). Aphanizomenon is an N-fixing (diazotrophic) filamentous cyanobacteria that can produce the hepatotoxin cylindrospermopsin (CYN). Salt stress above 2 g L À1 negatively affects Aphanizomenon growth in N-sufficient media (Rakko and Seppälä 2014). Nitrogenase, the enzyme responsible for N fixation, becomes active in N-limited conditions but is negatively affected by salt stress (Fernandes et al. 1993;Singh and Shrivastava 2017). Thus, the interactive roles of salinity and N availability may exert a strong influence on the potential for cyanobacteria to bloom and form toxins along the freshwatermarine continuum or in lakes along a salinity/N pollution gradient.
Here, we examine how two physiologically different cyanobacteria respond through growth, cyanotoxin production, and cellular release of cyanotoxins under differing N availability both before and after exposure to a gradient of salt. We hypothesized that (1) cyanobacteria growth would exhibit a subsidy-stress response (Odum et al. 1979) to salt exposure regardless of N availability, but (2) cyanotoxin production would exhibit a similar response only under high N availability with little or no cyanotoxin production in low N. We also hypothesized that (3) diazotrophic cyanobacteria growth and toxin production would decrease in response to N limitation but less so than nondiazotrophic cyanobacteria, and (4) that increasing salinity would increase the rate of toxin release into the surrounding water column.

Maintenance culture conditions
Microcystis aeruginosa (UTEX 2385) and Aphanizomenon flosaquae (PCC 7905) were maintained in laboratory conditions for over 24 and 4 months, respectively. Maintenance batch cultures were grown at 26 C at a light : dark cycle of 14 : 10 h (light intensity of 100 μmol m 2 s À1 ) in one-half strength (0.5Â) BG-11 media (Sigma) containing 1.35 μg L À1 vitamin B 12 .

Experimental design
We employed a factorial experimental design with three replicates, consisting of two N : P levels (N : P 4 and 50 by atom) and eight salinity treatments (0, 0.175, 0.35, 0.7, 1.4, 2.8, 5.6, and 10.5 g L À1 ) in 5% N-free BG-11 media with 1.35 μg L À1 vitamin B 12 . The N : P ratio was manipulated in the N-free BG-11 media by modifying N with nitrate-N (P = 357 μg L À1 ; nitrate-N = N : P 4: 640 μg L À1 , 50: 8060 μg L À1 ), and the salinity gradient was modified by adding Instant Ocean™ (dissolved inorganic phosphorus [DIP] = 0.79 μg L À1 , dissolved inorganic nitrogen [DIN] = 9.94 μg L À1 ; other micronutrients as well as DIP and DIN for each salinity treatment listed in Supplementary Table S1). The salinity treatments represent riverine to estuary salt conditions. To initiate experiments, 44.6 mg L À1 of Microcystis and 101 mg L À1 Aphanizomenon biomass as carbon (C) was added to experimental media, and cultures were grown at a 14 : 10 h light : dark cycle at a light intensity of 150 μmol m 2 s À1 and a temperature of 26 C. Prior to salt exposure, each population was grown until the N : P 4 reached stationary phase, resulting in pre-salt growth of 21 (Microcystis) and 33 (Aphanizomenon) days. The in vivo chlorophyll a values (Chl a; Turner Designs) were monitored during growth.

Sampling and analysis
Sampling for particulate nutrients and toxins occurred before salt addition on day 21 and 33 for Microcystis and Aphanizomenon, respectively, and again on days 1, 3, 5, 7, and 10 post salt addition (days 22,24,27,29,33 for Microcystis and days 34,36,38,40,43 for Aphanizomenon). Initial screening indicated that cyanotoxin differences were not seen until days 7 and 10 post-salt in the Microcystis cultures, and as such CYN samples were omitted on days 1 and 5 post-salt. Samples were collected for particulate C, N, and cyanotoxins on 0.7 μm GF/F Whatman filters that were combusted before use and the filtrate saved and stored at À20 C for dissolved toxin analysis. Filters for particulate C and N were stored at À20 C until analysis and filters for particulate cyanotoxins were lyophilized and stored at À80 C until analysis.
Particulate C (PC) and N (PN) filters were dried at 60 C for 24 h, and then analyzed simultaneously on an elemental analyzer as described by Wagner et al. (2019) (Thermo Finnigan FlashEA 1112, Thermo Fisher Scientific). Dissolved cyanotoxins were extracted and analyzed using an isotope dilution method coupled with LC-MS/MS (Agilent Technologies HPLC system with Agilent Technologies G6420 Triple Quadrupole MS) previously described in Haddad et al. (2019). were the sum of particulate and dissolved. The percent of PN allocated to total toxin was computed as follows (Eqs. 1A and 1B): where total MC-LR and CYN are the sum of the particulate and dissolved (mg L À1 ) and 0.141 and 0.169 are the %N in MC-LR and CYN, respectively. The %N of each toxin molecule was computed according to their molecular structures as reported in van de Waal et al. (2014).

Statistical analysis
Growth differences for each salinity treatment were the difference between in vivo Chl a values measured on days 43 and 31 for Microcystis and days 33 and 21 for Aphanizomenon. These data were analyzed with a one-way ANOVA using the "car" package in R (Fox and Weisberg 2019) and Bonferroni-corrected paired t-tests using the "stats" package (R Core Team 2021). Linear regression analysis between PN and total toxin as well as the percent PN allocated for both MC-LR TOT and CYN TOT was performed using the lm function in base R. A linear mixed model, using the lme4 package in R, was developed for C : N, total cyanotoxin, and dissolved cyanotoxin for each genus using day as a random effect and salinity, N : P, and genus as fixed effects (Bates et al. 2015). One-way ANOVAs were subsequently performed between salinity level by day for C : N, total cyanotoxin (μg L À1 ) and dissolved cyanotoxin (%) using the car package in R (Fox and Weisberg 2019). Bonferronicorrected paired t-tests were then performed for the previously described ANOVAs using the "stats" package in R (R Core Team 2021). A two-sample t-test was preformed to test for N : P influence on dissolved CYN. All analysis was done using R (R Core Team 2021). Data were made available in Dryad data repository https://doi.org/10.5061/dryad. 37pvmcvkh (Osburn et al. 2021).

Growth
Microcystis biomass (as in vivo Chl a) in the N : P 4 treatment reached a maximum on day 10 and remained nearly constant after salt addition with no significant differences between salinity treatments ( Fig. 1A; Supplementary Table S2) Table S2) increased throughout the experiment for salinities between 0 and 2.8 g L À1 and decreased in the 5.6 and 10.5 g L À1 salinity treatments after salt addition. Aphanizomenon biomass in the N : P 4 treatment ( Fig. 1C; Supplementary Table S2) temporarily plateaued on day 23 before decreasing in salinities greater than 1.4 g L À1 and increasing in salinities less than 1.4 g L À1 . However, the increased biomass in the low salinities was not different from the observed increase in zero salinity indicating that salt addition did not cause the increase in biomass. Aphanizomenon biomass in the N : P 50 treatment ( Fig. 1D; Supplementary Table S2) continuously increased until salt addition and continued to increase in salinity treatments between 0 and 0.35 g L À1 . However, Aphanizomenon biomass in the N : P 50 treatment decreased following salt addition greater than or equal to 0.7 g L À1 .

C : N ratios
The C : N stoichiometry of each genus responded uniquely to N : P and salinity (F 7,539 = 2.54, p = 0.014; Fig. 3). Prior to salt addition on day 21, Microcystis C : N in the N : P 4 treatment was greater than the C : N in the N : P 50 treatment (black lines with purple shading, Fig. 3A,B; F 1,47 = 460, p < 0.001). Microcystis C : N in low N : P conditions increased by day 31 regardless of salt addition; however, the 10.5 g L À1 salinity treatment resulted in lower C : N compared to the other salinities ( Fig. 3A; Supplementary Table S3A). Conversely, Microcystis C : N in high N : P treatment was less responsive to salt, but salinity treatments above 5.6 g L À1 resulted in significantly higher C : N compared to the lower salinities ( Fig. 3B; Supplementary Table S3B). We observed a small, but statistically significant, difference in Aphanizomenon C : N between N : P treatments prior to salt addition (black lines with purple shading; Fig. 3C,D; F 1,47 = 10.9, p = 0.002). Aphanizomenon C : N was greatest on day 10 between the 0.7 and 1.4 g L À1 salinity and lower at both high and low salinity in the low N : P treatment ( Fig. 3C; Supplementary  Table S3C). However, Aphanizomenon C : N was only found to be statistically different between the 2.8 and 10.5 g L À1 salinities in the N : P 50 treatments ( Fig. 3D; Supplementary  Table S3D).

Total cyanotoxin concentration
Total cyanotoxin production varied interactively among genera, N : P treatment, and salinity treatments (F 7,443 = 3.38, p = 0.002). MC-LR TOT was lower in the N : P 4 treatments compared to N : P 50 (Fig. 4A,B). Salinities between 2.8 and 5.6 g L À1 had significantly more MC-LR TOT than the 0 g L À1 treatment ( Fig. 4A; Supplementary  Table S4A). Conversely, MC-LR TOT was significantly higher in N : P 50 treatments exposed to the 0.35 g L À1 , and between the 1.4 and 5.6 g L À1 salinity treatments ( Fig. 4B; Supplementary Table S4B). However, MC-LR TOT in the 10.5 g L À1 salinity treatment was not different than the preexposure values. CYN TOT was more variable than MC-LR TOT before and after salt exposure (Fig. 4). Salinity did not affect the CYN TOT in N : P 4 treatments ( Fig. 4C; Supplementary  Table S4C) and was only greater than pre-exposure values in the 0.35 g L À1 salinity treatment and less than pre-exposure values in the 10.5 g L À1 salinity treatment ( Fig. 4D; Supplementary Table S4D).

Dissolved cyanotoxin concentration
Unlike total cyanotoxins, there was no significant interaction among genera, N : P treatment, and salinity treatments for dissolved cyanotoxin (%; F 7,436 = 1.38, p = 0.21). However, there was a significant interaction between salinity and N : P treatment was found for MC-LR DISS (Fig. 5A,B; F 7,260 = 6.92, p ≤ 0.001). MC-LR DISS was always less than 20% of MC-LR TOT in the N : P 4 treatment and was not different in the pre-and post-salt exposure ( Fig. 5A; Supplementary  Table S5A). MC-LR DISS in the N : P 50 treatment tended to increase with increasing salt exposure and was greatest when exposed to 5.6 and 10.5 g L À1 salinity ( Fig. 5B; Supplementary  Table S5B). CYN DISS was much more variable and tended to be greater than MC-LR DISS , but exhibited no significant interaction between N : P treatment and salinity (Fig. 5C,D; F 7,173 = 1.16, p = 0.33). There were no differences in CYN DISS due to N : P treatment throughout the experiment (t = 0.49, df = 190, p = 0.62; Fig. 5). However, there were few significant differences in CYN DISS due to salinity treatments throughout the experiment for both the N : P 4 treatment (F 1,94 = 6.61, p = 0.012) and the N : P 50 (F 1,94 = 11.43, p = 0.001; Fig. 5

Discussion
The magnitude and duration of cyanobacterial blooms and toxin production can be influenced by N and micronutrient availability . In this study, we demonstrated that growth and toxin production by a diazotrophic and nondiazotrophic cyanobacteria were highly variable, but that N availability and salinity interactively affected this variation in multiple ways. Both genera displayed positive correlations between biomass N content and total cyanotoxins, supporting previous studies (van de Waal et al. 2014;Wagner et al. 2019). Although neither taxa exhibited a growth response to salt subsidy, both exhibited a stress response at high N availability and Aphanizomenon experienced a stress response to salt at low N. Interestingly, both taxa exhibited a subsidy-stress response in cyanotoxin production across salinity exposures at high N availability, but not at low N availability. These results indicate the potential for different physiological costs and benefits in toxin-producing cyanobacterial blooms across gradients of N availability and salinity in nature and provide realistic risk boundaries for observing toxic blooms in ecosystems across these gradients.

Growth and toxin responses to salt
Both Microcystis and Aphanizomenon populations experienced growth limitation at low N availability (Fig. 1A,C) but only Aphanizomenon biomass changed following salt exposure at low N. Although Aphanizomenon biomass increased in low salinity treatments following salt exposure, biomass also slightly increased in controls, suggesting that the salt additions did not provide a nutrient resource subsidy for a physiological response to N limitation. Neither Microcystis nor Aphanizomenon growth plateaued at high N availability and growth rate following salt exposure was not different between the controls and low salinity treatments. Thus, there was no salt-subsidy effect at high N availability. Other studies have shown that Microcystis strains can persist (Sellner et al. 1988; Aphanizomenon cultures grown in N : P 4 (C) and N : P 50 (D). Black line represents the average from pre-salt additions (Microcystis = day 21, Aphanizomenon = day 33) and the purple shaded area represents the standard deviation. Black circles represent 10 d post-salt addition with standard deviation lines (Microcystis = day 31, Aphanizomenon = day 43). Orr et al. 2004;Miller et al. 2010) and produce toxins in salinities as high as 17.5 g L À1 (Tonk et al. 2007). We hypothesized that a resource subsidy response to low salt exposure would occur in both diazotrophic and nondiazotrophic cyanobacteria based on micronutrient-stimulated growth in other studies (Facey et al. 2019;Wagner et al. 2021). However, we did not intentionally reduce micronutrient concentrations in our original media like these other studies, which could explain the lack of any subsidy response to growth. Although we found no evidence that salt subsidized growth, we did observe a salt subsidy response in MC-LR TOT production by Microcystis below 10.5 g L À1 salinity and in CYN TOT by Aphanizomenon at 0.35 g L À1 , which supports findings from other recent studies (Georges des Aulnois et al. 2020). Low salt stress conditions may cause an incomplete inhibition of photosystem II as seen in higher salt conditions (Allakhverdiev and Murata 2008), leading to higher reactive oxygen species. A potential hypothesis for cyanotoxin production, particularly MC, is that cyanotoxins can act as antioxidants (Zilliges et al. 2011), benefiting cyanobacteria exposed to low salt concentrations. The literature for this response in cyanobacteria is not extensive, and we suggest further investigation to determine the mechanism.
Salt exposure decreased biomass in both taxa at high N availability and in Aphanizomenon at low N availability. Microcystis was negatively impacted by salt concentrations above 5.6 g L À1 ; however, MC-LR DISS tended to increase in salt treatments above 0.7 g L À1 indicating a possible osmatic imbalance at concentrations below those inhibiting growth. Nitrogenlimited Microcystis increases carbohydrate content that assists in osmoregulation when exposed to salt stress (Li et al. 2021). Other gram-negative bacteria entering stationary phase show cellular membrane changes that promote resistance to environmental stressors (Huisman et al. 1996 and references within). Aphanizomenon was sensitive to salt stress regardless of N availability, supporting previous findings (Rakko and Seppälä 2014). Aphanizomenon is diazotrophic and nitrogenase is sensitive to salt stress at high salinities (Severin et al. 2012), which can reduce growth in low N environments (Brutemark et al. 2015). Rapala et al. (1993) reported that Aphanizomenon produces more toxin while actively fixing N, but our results did not support this pattern. Additionally, CYN DISS can increase when cells enter stationary phase and lyse (Davis et al. 2014). In our experiment, growth slowed in N : P 4 conditions with higher CYN DISS compared to N : P 50 cultures, regardless of salt stress (Fig. 5C,D).

Scaling to ecosystems
Although salinization in riparian ecosystems have been demonstrated to create a growth subsidy (Entrekin et al. 2019), less is known about the potential subsidy effect in freshwaters. Increases in extreme rain events and droughts can change the salinity of inland lakes (Atkinson and Mabe 2006; Liu and Bao 2020; Dugan and Rock 2021). The Cl À concentration in suburban/urban streams with roadways can approach~5 g L in the winter months (Kaushal et al. 2005), which is 20 times the proposed limit for water quality (Lewis 1999). However, salt pollution derived from road de-icers, sea-level rise, or climate-driven hydrologic extremes are likely to create salinity gradients across historically freshwater ecosystems at landscape scales. Low salt stress in freshwater lakes can influence phytoplankton community dynamics, favoring cyanobacteria over other phytoplankton (Chakraborty et al. 2011). If mild increases in salinity also increase cyanotoxin production, ecosystems even moderately impacted by salt pollution could experience an increase the frequency and magnitude of toxin-producing cyanobacteria. Cyanobacteria blooms in areas where anthropogenic salt stress is more acute (Cirés and Ballot 2016) could be at greatest risk. Our study indicated that relatively N-rich lakes could experience increased cyanotoxins and increased extracellular export of these toxins. Other studies have indicated that Aphanizomenon blooms can rapidly release 99% of their CYN to the water column, as seen from a survey of 21 German lakes (Rücker et al. 2007). CYN has the potential to bioaccumulate in freshwater fish and invertebrates (Scarlett et al. 2020) and may be further enhanced in salt stressed conditions when compared to MC, due to its slow degradation rate (Chiswell et al. 1999).
Increasing salinity has been shown to decrease colonial cyanobacteria while increasing picocyanobacteria in estuaries (Gonz alez del Río et al. 2007). However, sea level rise may increase salt intrusion and result in new patterns. For example, filamentous cyanobacteria (Planktothrix) have been mostly found in the riverine area of an estuaries with low salinity (Muylaert et al. 2009). Slow salinization may select for salt tolerant toxic species (Melero-Jiménez et al. 2019), and our results suggest that for toxin-producing cyanobacteria this may result in a greater proportion of toxins per cell or a more rapid release of toxins from cells. Although aquatic toxicity and bioaccumulation data for cyanotoxins are becoming more prominent (Ibelings and Havens 2008;Gibble et al. 2016;Lovin et al. 2019;Scarlett et al. 2020;Camacho-Muñoz et al. 2021;Mehinto et al. 2021), there remains a critical demand for continued research across the freshwater-marine continuum and in landscapes at risk of nutrient and salt pollution.

Implications
Our study indicates that exposure to salinization may increase the risk of cyanotoxins released from cyanobacteria blooms in aquatic ecosystems. However, our study design specifically addresses the emerging management strategy of mitigating toxic cyanobacterial blooms by flushing them to downstream environments (Lundgren et al. 2013;Paerl et al. 2016;Grover et al. 2017). For example, recent blooms in Lake Okeechobee, Florida, have been flushed to the Indian River Lagoon and other downstream estuaries (Rosen et al. 2018). Salinity in the Indian River Lagoon can range from~3 to 34 g L À1 depending on location from the river inflow to the ocean (Indian River Lagoon Observatory data, accessed 13 August 2021; http://fau.loboviz.com/). The experiments presented in this study were conducted between 0 and 10.5 g L À1 , indicating the potential for a 2X increase in total MC-LR in a Microcystis bloom before cells lyse at approximately 0.35 g L À1 . Thus, flushing blooms to coastal environments does not ensure that bloom biomass or toxin concentration will decrease in magnitude before mixing with ocean water, and the process of this mixing may magnify the harmful effects of cyanobacterial blooms.