Diazotrophy modulates cyanobacteria stoichiometry through functional traits that determine bloom magnitude and toxin production

Harmful cyanobacterial blooms are an increasing threat to water quality. The interactions between two ecophysiological functional traits of cyanobacteria, diazotrophy (nitrogen [N]‐fixation) and N‐rich cyanotoxin synthesis, have never been examined in a stoichiometric explicit manner. We explored how a gradient of resource N : phosphorus (P) affects the biomass, N, P stoichiometry, light‐harvesting pigments, and cylindrospermopsin production in a N‐fixing cyanobacterium, Aphanizomenon. Low N : P Aphanizomenon cultures produced the same biomass as populations grown in high N : P cultures. The biomass accumulation determined by carbon, indicated low N : P Aphanizomenon cultures did not have a N‐fixation growth trade‐off, in contrast to some other diazotrophs that maintain stoichiometric N homeostasis at the expense of growth. However, N‐fixing Aphanizomenon populations produced less particulate cylindrospermopsin and had undetectable dissolved cylindrospermopsin compared to non‐N‐fixing populations. The pattern of low to high cyanotoxin cell quotas across an N : P gradient in the diazotrophic cylindrospermopsin producer is similar to the cyanotoxin cell quota response in nondiazotrophic cyanobacteria. We suggest that diazotrophic cyanobacteria may be characterized into two broad functional groups, the N‐storage‐strategists and the growth‐strategists, which use N‐fixation differently and may determine patterns of bloom magnitude and toxin production in nature.

production of two ammonia molecules from N 2 , but comes with an extreme demographic (Grover et al. 2022) and energetic (Fay 1992) cost, often resulting in a significant growth trade-off (Osburn et al. 2021). The evolutionary significance of cyanotoxin production remains elusive. The hypothesized fitness advantages of microcystins production include their role as chemical antioxidants in cells where high rates of photosynthesis create oxidative stress (Zilliges et al. 2011), and bound with ribulose-1,5-bisphosphate carboxylase/oxygenase (RubisCO) at the thylakoid membrane (Barchewitz et al. 2019). A hypothesized fitness advantage for cylindrospermopsin includes potential allelopathic properties that may increase nutrient acquisition (Bar-Yosef et al. 2010;Raven 2010). Both N-fixation and cyanotoxin production are sensitive to the external nutrient supply (Wagner et al. 2019;Brandenburg et al. 2020;Osburn et al. 2021) and recent theoretical models suggest cyanotoxin production may even be greater in diazotrophs growing in non-N-fixing conditions (Grover et al. 2020). However, the interactions between N-fixation and cyanotoxin quotas have never been examined empirically.
Cyanotoxins are synthesized using nonribosomal pathways by a large multifunctional enzyme complex (Pearson et al. 2010). The exact energetic and elemental costs of synthesizing the multifunctional enzyme complex are unknown, but it likely exceeds that of a cyanotoxin molecule. However, cyanotoxin quotas are often correlated to the resource N : P supply (Wagner et al. 2019;Brandenburg et al. 2020;Osburn et al. 2022). Like many secondary metabolites, cyanotoxins have a variety of elemental compositions, which can be categorized as either N-rich, or C-rich/N-poor relative to typical phytoplankton stoichiometric ratios such as the Redfield ratio (C : N 6.6). Using the theoretical frameworks of ecological stoichiometry (Sterner and Elser 2002) and stoichiometric ecotoxicology (Peace et al. 2021), it has been hypothesized that conditions causing N to be in excess compared to growth demand (i.e., P-limitation) will increase N-rich cyanotoxin production ). However, this hypothesis has not been examined in diazotrophic cyanobacteria, which may have minimal plasticity in their N composition because N-fixation maintains cellular N quotas at the cost of growth (Osburn et al. 2021).
Aphanizomenon and Raphidiopsis (formerly Cylindrospermopsis) are bloom-forming diazotrophic cyanobacteria that can produce the N-rich hepatotoxin, cylindrospermopsin (Huisman et al. 2018). Nutrient supply can alter the concentration and cell quotas of cylindrospermopsin; however, there is conflicting information on what nutrient regimes increase, decrease, or do not affect cylindrospermopsin production. For example, Raphidiopsis produced more cylindrospermopsin when grown on exclusively atmospheric N compared to combined N sources (Saker and Neilan 2001), whereas Aphanizomenon grown on atmospheric N had lower concentrations and cell quotas of cylindrospermopsin compared to high combined N conditions (Preußel et al. 2014;Osburn et al. 2022). Conflicting cylindrospermopsin responses also occur in P-limited conditions with Aphanizomenon increasing cylindrospermopsin concentrations and quotas (Preußel et al. 2014) or decreasing cylindrospermopsin concentrations (B acsi et al. 2006). Other studies have identified that cylindrospermopsin quotas are correlated with growth rate and produced constitutively (Davis et al. 2014). How cylindrospermopsin concentrations respond to different degrees of N-and P-sufficiency or -deficiency may be explained by phylogenetic diversity within this group. However, when multiple strains are examined within one experiment, the cylindrospermopsin response to nutrients is consistent (Davis et al. 2014;Preußel et al. 2014). A possible missing element from many cyanotoxin/nutrient stoichiometry studies is the quantification of macromolecules and elements related to the physiological changes that occur in nutrientstressed organisms.
Here, we examine N-fixation, N and P stoichiometry, light-harvesting pigments, and intra-and extracellular cylindrospermopsin concentrations in Aphanizomenon cultures grown over a large resource N : P gradient. Our overarching hypothesis was that Aphanizomenon functional traits (N-fixation, light-harvesting pigments, and cylindrospermopsin concentrations and quotas) are affected by the resource N : P conditions. We predicted that a growth Nfixation trade-off would occur in N-limiting conditions (low N : P), which results in less biomass produced due to the energetic costs of N-fixation. In addition, we predicted a positive relationship between resource N : P supply and light-harvesting pigments and cylindrospermopsin concentrations and cell quotas. Finally, we predicted that N-fixation would alter lightharvesting and cyanotoxin relationships through changes in N allocation, with N-fixing populations allocating less N to cylindrospermopsin quotas, and more N to light-harvesting pigments compared to non-N-fixing populations. We tested this hypothesis using batch culture Aphanizomenon flos-aquae (PCC 7905) experiments, which is known to be diazotrophic and produce cylindrospermopsin (Osburn et al. 2022).

Culture maintenance
The nonaxenic A. flos-aquae (PCC 7905) was obtained from the Pasteur Collection of Cyanobacteria in November of 2019 and maintained for 10 months at Baylor University. Stock Aphanizomenon cultures were grown in sterilized half strength (Â0.5) BG-11 media with 1.35 μg L À1 vitamin B 12 and a light intensity of 100 μmol m À2 s À1 on a 14 h : 10 h light : dark cycle at 26 C. Stock cultures were replaced monthly and used for seeding experimental treatments.

Experimental design
To examine how the resource supply of N : P affects the functional traits in Aphanizomenon, we manipulated the amount of nitrate-N (NO 3 -N) while maintaining the same P Wagner et al.
Nitrogen affects Aphanizomenon physiology concentration across all treatments using N-free BG-11 media.
Quadruplicates of the 11 different N : P treatments (by atom) were made by adding 5% of 1X N-free BG-11 media with 1.35 μg L À1 vitamin B 12 into a final volume of 850 mL. The resulting medium had a P concentration of 357 μg L À1 to which we varied the NO 3 -N concentration between 0.16 and 16 mg L À1 to generate a gradient of N : P conditions commonly found in lakes . We seeded the cultures with 0.2 mg L À1 carbon (C)-biomass of Aphanizomenon and placed them into an incubator with a light intensity of 140 μmol m À2 s À1 on a 14 h : 10 h light : dark cycle at 26 C. Every other day, cultures were removed from the incubator, shaken, and a 2 mL subsample was collected for quantification of in vivo chlorophyll a (Chl a) via fluorescence (RFU; Turner Designs; Fig. S1). After 37 d of growth, we collected samples for particulate C, N, P, phycobilin pigments (PBPs), Chl a, and cylindrospermopsin by filtering cultures onto 0.7 μm precombusted glass fiber filters. The remaining filtrate was saved for dissolved cylindrospermopsin, dissolved P, nitrate/ nitrite, and total dissolved N (TDN). The 0.7 μm glass fiber filters would capture the Aphanizomenon biomass as individual cells and filaments are much larger than 0.7 μm but would allow small particles including heterotrophic bacteria into our dissolved fraction. The small particles may have artificially increased the TDN concentrations. Finally, we preserved 5 mL from each culture with 0.2 mL of Lugol's iodine for cell counts.

Particulate and dissolved elemental composition
Filters for particulate C, N, and P were stored at À20 C until analyzed. Filters for C and N were dried at 60 C for 24 h before analysis on an elemental analyzer (Thermo Fisher Flash-Smart NC Soil) as previously described in Wagner et al. (2019). Particulate P was digested by autoclaving filters in 3% (wt/vol) persulfate solution and quantified by colorimetry on a UV-Visible spectrophotometer at 885 nm using a 1 cm cuvette (APHA 1992).
Dissolved nitrate-N/nitrite-N and PO 4 -P (filtered through a 0.7 μm filter) were analyzed on a Lachat 8500 flow-injection auto-analyzer with an ASX-520 autosampler (Hach, CO.) using cadmium reduction to convert nitrate to nitrite and measured using standardized colorimetry methods (APHA 1992). TDN samples were oxidized using a basic persulfate solution 3% (wt/vol) and then buffered with boric acid (APHA 1992). We analyzed TDN samples on a Lachat 8500 flow-injection autoanalyzer with an ASX-520 autosampler (Hach, CO.).

Light-harvesting pigment analysis
Chl a and PBPs filters were stored at À20 C before analysis. We extracted Chl a in 90% acetone (90 : 10 acetone : water) for 24 h at À20 C and analyzed using a fluorometer with an acidchlorophyll module (Turner Designs). Concentrations of Chl a were calculated as described in EPA method 445.0 (Arar and Collins 1997).
PBPs were analyzed as described by Wang et al. (2021) with slight modifications. To each sample, we added 5 mL of 0.1 mol L À1 sodium phosphate buffer, pH 7, and extracted PBPs by performing two freeze-thaw cycles. After, we sonicated the samples for 7 min at 35 kHz (VWR ultrasonic cleaner). Samples were then stored at 4 C for 12 h and read on a UV-Visible spectrophotometer at 652, 615, and 562 nm to determine the concentrations of phycocyanin (PC), allophycocyanin (APC), and phycoerythrin (PE) on a 1 cm cuvette. Total PBPs were determined using the following equations: Total PBPs mg L À1 Nitrogen fixation rates Every method to determine N-fixation rates has advantages, disadvantages and assumptions associated with each assay. We used a mass balance approach to determine gross N-fixation rates over the whole 37-d experiment. This method was chosen over other common N-fixation rate assays because our system is closed and additional N into the particle form had to occur through N-fixation. In addition, by measuring different N pools, our method incorporated N 2 that was fixed and potentially "leaked" out of the cell. The mass balance approach presumes there is no denitrification, which is a reasonable assumption for our well mixed oxygenated cultures.
Gross N-fixation rates were calculated using the following equations: where the amount of particulate and dissolved organic N is subtracted from the initial nitrate-N concentration and converted into a rate by dividing by the length of the experiment (37 d) into hours.

Cylindrospermopsin analysis
Particulate cylindrospermopsin was lyophilized using a Virtis SP Scientific benchtop pro freeze dryer (SP Scientific) for 48 h and stored at À80 C until analyzed. An isotope dilution liquid chromatography triple quadrupole mass spectrometry method was used for cylindrospermopsin analysis as previously described (Haddad et al. 2019) and applied (Osburn et al. 2022). Samples were spiked with cylindrospermopsin-15 N 5 before extracting particulate cylindrospermopsin in 1 mL of 25 : 75 acetonitrile : water. Extractions were completed by sonicating samples for 5 min then centrifuging at 3500 RPM for 5 min. The supernatant was removed, pooled with supernatant from two additional extractions, and blown to dryness using nitrogen gas. Samples were then resuspended in 1 mL of 10 : 90 water : acetonitrile solution buffered with 5 mmol L À1 ammonium formate and 3.6 mmol L À1 formic acid (pH 3.7) and stored in an HPLC vial.
Particulate and dissolved cylindrospermopsin samples were separated on an Agilent 1260 HPLC using a Poroshell HILIC-Z column (2.1 Â 150 mm, 2.7 μm, 120 Å; Agilent Technologies) and analyzed on an Agilent 6420 triple quad using electron spray ionization in positive mode (Haddad et al. 2019). We monitored the precursor ion 416 m/z and product ions 194.1 and 176 m/z for cylindrospermopsin, and the precursor ion 421 m/z and product ions 197.1 and 176 m/z for the isotopically labeled cylindrospermopsin internal standard. The detection limit for cylindrospermopsin was 0.43 ng L À1 , with samples below detection limits reported as half the detection limit (0.215 ng L À1 ; Haddad et al. 2019).

Cell counts
Preserved samples were loaded on a Palmer counting cell and counted using a compound microscope (Nikon Eclipse 80i) at Â400 magnification. A minimum of eight frames (volume 2.4 Â 10 À5 mL per frame) and 400 cells were counted to calculate the cell density. The cell density was converted to cells L À1 Resource nitrogen (mg L -1 )  and used to standardize cylindrospermopsin, particulate N, and pigment culture concentrations to cell quotas.

Statistical analysis
We checked all data for normality and heteroscedasticity before all analyses. Nitrogen-fixation rates and C : N ratios were compared among N : P resource treatments with a one-way ANOVA using the car package in R (R Core Team 2019). We separated the data into two subsets with N-fixation rates that were significantly different from zero (N-fixing populations) and N-fixation rates that were not significantly different from zero (non-N-fixing populations). The N-fixing populations represented resource N : P ratios between 1 and 12, while the non-N-fixing populations were resource N : P ratios between 16 and 100. All linear, power, and piecewise regressions were completed in SigmaPlot (version 14) using the dynamic fit chlorophyll a (μg L À1 ), (B) phycobilin pigments (PBP; mg L À1 ), the cell quotas of (C) chlorophyll a (pg cell À1 ) and, (D) phycobilin pigments (PBP; pg cell À1 ). Relationships between N cell quota (pg cell À1 ) and (E) chlorophyll a cell quota (pg cell À1 ) and (F) phycobilin pigments cell quota (PBP; pg cell À1 ). Open circles represent N-fixing populations while dark gray circles are non-N-fixing populations. Error bars are AE standard deviation for panels A-D.
mode. Models were chosen using a combination of the lowest AIC between different models, visual examination of model fit, and best adjusted r 2 . The p value was determined by comparing the regression model to the null model (no regression). We examined whether the linear regressions between the N-fixing and non-N-fixing populations had different slopes or intercepts using ordinary least squares regression in the program SMATR Warton et al. 2006). Finally, we calculated the 95% confidence intervals for the total dissolved cylindrospermopsin and percent dissolved cylindrospermopsin to determine which treatments have significantly more dissolved cylindrospermopsin than zero. All resource N : P ratios on the x-axis were logarithmically transformed for easier visualization of the low N : P populations; however, statistics were completed on nontransformed data.

Results
Regardless of resource N : P the Aphanizomenon cultures had the same biomass ( Fig. 1A; mean 15.4 AE 2.6 mg L À1 ; F 10, 42 = 0.824; p = 0.61) and particulate N concentration ( Fig. 1B; mean 2.8 AE 0.8 mg L À1 ; F 10, 42 = 2.09; p = 0.06). Cultures grown in resource N : P treatments between 1 and 20 had evidence of N-fixation, with the mean organic N concentration in the cultures being greater than what was added (Fig. 1B). Gross N-fixation rates differed among resource N : P treatments ( Fig. 1C; F 10, 42 = 21.95; p < 0.001). Gross N-fixation rates were the highest in resource N : P treatments between 1 and 12, which was $ 1.5-2.7 μg L À1 h À1 for the 37-d duration of the experiment (Fig. 1C). Resource N : P 16 and 20 treatments had a gross N-fixation rate of $ 0.5 μg L À1 h À1 for the 37-d duration. However, the N : P 16 and 20 treatments were not significantly different from the N : P 30-100 treatments, which had no evidence of N-fixation (Fig. 1C). The C : N stoichiometry in Aphanizomenon cultures ranged between 5.8 and 7.7 that differed by resource N : P ( Fig. 1D; F 10, 42 = 15.93; p < 0.001). Cultures grown in resource N : P treatments between 1 and 4 had significantly higher C : N compared to resource N : P 8 and between N : P 20 and N : P 100 treatments (Fig. 1D). The concentrations and cell quotas of Chl a and PBPs were sensitive to resource N : P (Fig. 2). Chl a concentration increased linearly at a rate of 44.33 μg L À1 from N : P 1 until N : P 17.9. Above N : P 17.9 Chl a concentration decreased at a rate of À4.89 μg L À1 (r 2 = 0.73; Fig. 2A). Similarly, the concentration of PBPs increased linearly at a rate of   0.125 mg L À1 from N : P 1 until N : P 18.6. Above N : P 18.6 the rate of increase in PBPs slowed to 0.009 mg L À1 (r 2 = 0.81; Fig. 2B). A piecewise regression also best explained the Chl a cell quota and PBPs cell quota responses to experimental N : P. Chl a cell quotas increased at a rate of 0.015 pg cell À1 from N : P 1 to N : P 18, but decreased at a rate of À0.004 pg cell À1 when N : P was greater than 18 (Fig. 2C). PBP cell quotas increased at a rate of 0.035 pg cell À1 from N : P 1 to N : P 23 PBPs and decreased at a rate of À0.005 pg cell À1 at N : P greater than 23 (Fig. 2D). We examined how cell quotas of Chl a and PBPs correlate to N cell quotas because light-harvesting molecules contain various amounts of N (Fig. 2E,F). Chl a cell quota in the non-N-fixing populations (N : P 16 to 100) increased at a rate of 0.389 pg cell À1 with increasing N cell quota. However, N-fixing Aphanizomenon cultures (N : P 1 to N : P 12) had no relationship between Chl a and N cell quotas (Fig. 2E). Both the non-N-fixing and N-fixing populations had positive relationships between PBPs cell quotas and N cell quotas (Fig. 2F).
Aphanizomenon grown in non-N-fixing conditions allocated more PBPs per N cell quota (slope 1.11) compared to the N-fixing conditions (slope = 0.499; t 2 = 6.38; p = 0.018; Fig. 2F). Resource N : P affected the cylindrospermopsin concentration and cell quotas (Fig. 3). Total cylindrospermopsin concentrations ranged between 16.7 and 96.6 μg L À1 in a piecewise relationship across the resource N : P gradient (Fig. 3A). The rate of increase from N : P 1 to N : P 20 was 2.04 μg L À1 and above N : P 20, the rate of increase with increasing resource N : P was 0.34 μg L À1 cylindrospermopsin (r 2 = 0.85; Fig. 3A). Total cylindrospermopsin cell quotas responded in a piecewise regression with resource N : P (Fig. 3B). The rate of cylindrospermopsin cell quotas increased from N : P 1 to 20 was 0.879 fg cell À1 , and above N : P 20, the rate of increase was 0.0005 fg cell À1 (r 2 = 0.47; Fig. 3B). Particulate concentrations of cylindrospermopsin, that were normalized to volume of the culture filtered, ranged between 16.7 and 59.5 μg L À1 in a piecewise relationship across the N : P gradient (Fig. 3C). The rate of increase in cylindrospermopsin concentration from N : P 1 to N : P 19.6 was 1.65 μg L À1 , and above N : P 19.6, the rate of increase was 0.046 μg L À1 (r 2 = 0.89; Fig. 3C). Particulate cylindrospermopsin cell quotas responded in a piecewise relationship with resource N : P (Fig. 3D). The rate of cylindrospermopsin cell quotas increased from N : P 1 to 20 was 0.59 fg cell À1 . Above N : P 20, the particulate cylindrospermopsin cell quotas decreased with increasing resource N at a rate of À0.078 fg cell À1 (r 2 = 0.25; Fig. 3D). Dissolved cylindrospermopsin concentrations ranged between 0 and 41 μg L À1 (Fig. 3E), representing 0% to 78.2% of the total cylindrospermopsin concentrations (Fig. 3F). Both dissolved cylindrospermopsin concentration and percent dissolved cylindrospermopsin were statistically greater than zero in the resource N : P treatments above 30, while all treatments below N : P 30 were not statistically different from zero (Fig. 3E,F).
Since cylindrospermopsin is a N-rich molecule, we examined the cell quota correlations between cylindrospermopsin, total and particulate, to N cell quotas and pigment quotas (Fig. 4). In general, we found stronger relationships between particulate cylindrospermopsin quotas with N and pigments quotas compared to the relationships using total cylindrospermopsin quotas (Fig. 4). We identified positive linear relationships in the non-N-fixing populations between N cell quotas and total and particulate cylindrospermopsin cell quotas (Fig. 4A,B). The rate of increase was 20.24 and 16.34 fg cell À1 per N cell quota for total and particulate cylindrospermopsin cell quotas, respectively. There was a positive linear relationship for only the N-fixing populations between total cylindrospermopsin and Chl a cell quota (Fig. 4C). For both the N-fixing and non-N fixing populations, a positive linear regression between Chl a and particulate cylindrospermopsin cell quotas were observed (Fig. 4D). The non-N-fixing and N-fixing populations exhibited the same rate of change between Chl a cell quota and cylindrospermopsin cell quota (t 2 = 0.151; p = 0.73); however, the non-N-fixing Aphanizomenon had elevated cylindrospermopsin cell quota compared to the N-fixing Aphanizomenon (t 2 = 13.92; p < 0.001) (Fig. 4D). We found that total cylindrospermopsin and particulate cylindrospermopsin cell quotas were positively correlated to PBP cell quotas for both the non-N-fixing and N-fixing populations (Fig. 4E,F). The non-N-fixing and N-fixing Aphanizomenon exhibited similar rates of change between PBP and cylindrospermopsin cell quotas (total cylindrospermopsin slope t 2 = 0.05; p = 0.83; particulate cylindrospermopsin slope t 2 = 0.50; p = 0.47); however, the non-N-fixing Aphanizomenon had elevated cylindrospermopsin cell quota compared with N-fixing Aphanizomenon (t 2 = 27.87; p < 0.001 for total cylindrospermopsin; t 2 = 7.88; p = 0.005) (Fig. 4E,F).
Dissolved cylindrospermopsin concentrations were correlated with the dissolved nitrate-N and nitrate-N : phosphate-P concentrations but unrelated to phosphorus-P concentrations (Fig. 5). In cultures actively fixing N, dissolved nitrate-N was below detection limits and Aphanizomenon produced little dissolved cylindrospermopsin (Fig. 5A). However, the non-N-fixing Aphanizomenon populations had a positive linear relationship where increasing nitrate-N resulted in higher dissolved cylindrospermopsin (Fig. 5A). Similarly, N-fixing Aphanizomenon cultures had phosphate-P remaining in the media that ranged between 6.9 and 127 μg L À1 , whereas the non-N-fixing treatments were almost exclusively below detection limits (Fig. 5B). Most of the non-N-fixing Aphanizomenon cultures with measurable dissolved cylindrospermopsin had phosphate-P concentrations below detection limits (Fig. 5B). The dissolved N:P for the N-fixing Aphanizomenon populations ranged between 0.007 and 8; whereas, the non-N-fixing Aphanizomenon populations had a N : P range between 38 and 115,000 (Fig. 5C). We identified a significant power relationship for the non-N-fixing Aphanizomenon cultures between media N : P and the dissolved cylindrospermopsin concentration (Fig. 5C), implying more P-poor and N-rich conditions (higher N : P) increase dissolved cylindrospermopsin concentrations.

Discussion
In our experiments, A. flos-aquae exhibited flexible C : N stoichiometry and no growth trade-off due to N-fixation. Other genera such as Dolichospermum exhibit stoichiometric N homeostasis but at the expense of biomass production when N-fixation is their primary N source (Grover et al. 2020(Grover et al. , 2022Osburn et al. 2021). This trade-off may represent trait variation in how diverse diazotrophs have evolved competing strategies for managing N deficiency. Thus, our study suggests that it may be possible to fit some heterocystous cyanobacteria strains across a broad spectrum that ranges from two functional extremes: (1) the N-storage-strategist and (2) the growthstrategist. However, it is unknown if this stoichiometric-growth trade-off is a functional trait that is specific to different cyanobacteria strains, species, or genera, or if these seemingly disparate responses are a function of environmental variability alone. If more support is identified that the N-fixation growth trade-off is a trait difference, species that use N-fixation to maintain a strict C : N homeostasis at the expense of population growth are the N-storage-strategists, and species that use N-fixation to fuel population growth by maximizing C : N plasticity are the growth-strategists. The N-fixation trait physiology may extend from the r-(opportunistic) or K-selected (gleaner) growth strategies (Grover 1990;Papanikolopoulou et al. 2018). Diazotrophs that are fast growers (opportunistic) may have the N-storage-strategy because they cannot maintain growth in diazotrophic conditions, while the gleaners have slower growth, but can maintain growth in diazotrophic conditions and thus have a growth strategy.
Interestingly, either N-storage-strategist or growthstrategists are compatible with the stoichiometric control of cyanotoxin production. For example, N deficiency stimulated N-fixation but decreased light-harvesting pigment synthesis and cylindrospermopsin production in A. flos-aquae studied here. Conversely, other studies have shown that Dolichospermum flos-aquae maintained C : N stoichiometric homeostasis when actively fixing N by increasing its light-harvesting pigment cell-quota (Wang et al. 2021), and numerous studies have now demonstrated the critical link between low C : N ratios and N-rich toxin production in cyanobacteria Wagner et al. 2019). However, to our knowledge no studies have examined a diazotrophic N-storage-strategists that produces cyanotoxins. Regardless, species identity and their respective trait variations may influence how N-fixation fuels the magnitude of cyanobacterial blooms, but stoichiometry and growth trade-offs influence the magnitude of toxins production when diazotrophic blooms occur.
Our study is the first to explicitly test if resource stoichiometry controls the N-rich cyanotoxin production in a diazotroph. Previous studies have reported that cylindrospermopsin production is constitutive in Raphidiopsis species and correlated with growth rate (Willis et al. 2015). Other studies found Raphidiopsis cylindrospermopsin production is high in N-fixing conditions and low when cultures were grown on ammonium (Saker and Neilan 2001). Alternatively, Aphanizomenon cultures grown with, and without N, increased cylindrospermopsin production with N added to the growth medium (Preußel et al. 2014). Furthermore, other studies have indicated that Raphidiopsis strains increase the dissolved cylindrospermopsin concentration as the culture enters the stationary phase (Dyble et al. 2006;Davis et al. 2014). We found cylindrospermopsin concentrations increased with more N in the media and without P-limitation, with a slow growth rate of 0.03-0.05 d À1 (Fig. S2) and a stable C : P at $ 100 regardless of resource N : P (Fig. S3). Increasing cyanotoxin concentrations and quotas in the absence of P-limitation with excess N was also observed in Microcystis (Wagner et al. 2019). The dissolved cylindrospermopsin concentration forced the positive relationship between N : P and cylindrospermopsin concentration because particulate cylindrospermopsin quotas decreased slightly above N : P 20. This suggests cylindrospermopsin is exported out of the cell at higher rates with higher resource N : P, while cultures were still in log phase (Fig. S1). A possible explanation for the conflicting results is the use of different taxa and strains. Future experiments should examine multiple genera grown over a range of N : P to explore if production of cylindrospermopsin (by cell quota and biovolume), other cylindrospermopsins, and cyanotoxins follow the stoichiometric prediction, regardless of the cyanobacteria genus or strain.
Functions of cyanotoxins remain elusive regardless of the toxin class or specific cyanotoxin, which routinely vary in terms of biological activities. Microcystins are hypothesized to function in light capturing, associate with RubisCO at the thylakoid membrane, and/or have antioxidant qualities (Zilliges et al. 2011;Omidi et al. 2017;Barchewitz et al. 2019). Cylindrospermopsin concentrations are affected by the light intensity, with higher intensities increasing cylindrospermopsin concentrations (Dyble et al. 2006). In addition to changing light intensity, light-harvesting molecules are also affected by the nutrient status of phytoplankton, with both N-and P-limitation decreasing chlorophyll and PBP cell quotas (Forchhammer and Schwarz 2019;Duan et al. 2021;Wang et al. 2021). Light-harvesting proteins, particularly the phycobilisome that contains the PBPs, can comprise over 60% of the soluble proteins in cyanobacteria (Stadnichuk et al. 2015), making these light-harvesting compounds a large sink of cellular N, even though the pigment N is only 9.53% of the overall PBP compound mass. Since PBPs can act as a N-storage pool (Wang et al. 2021), we expected a relationship between cellular N, PBPs, and cylindrospermopsin quotas. We found positive relationships between cylindrospermopsin concentrations and quotas in both the N-fixing and non-N-fixing conditions with light-harvesting pigments, which implies that cylindrospermopsin is allocated as a proportion of light-harvesting proteins. However, in N-fixing treatments, cellular N and cylindrospermopsin quotas were not correlated, suggesting fixed atmospheric N was allocated to processes besides cylindrospermopsin production. A possible explanation for the lack of correlation between cylindrospermopsin and N cell quotas in N-fixing treatments is that a lag may exist in which cyanophycin (Li et al. 2001) is utilized for N storage before it is incorporated into other biomolecules.
Cylindrospermopsin is hypothesized to be an allelopathic chemical that causes other phytoplankton to produce alkaline phosphatase, thus making organic P available for uptake (Bar-Yosef et al. 2010;Raven 2010;Lu et al. 2021). Some Aphanizomenon strains grow poorly on organic P, implying the extracellular alkaline phosphatase excreted from Aphanizomenon is not able to support optimal growth (Vahtera et al. 2007). If cylindrospermopsin causes phytoplankton/heterotrophic bacteria to increase extracellular alkaline phosphatase, it could increase the available P to the ecosystem. Adding dissolved cylindrospermopsin to phytoplankton in both P-sufficient and P-deficient conditions increased extracellular alkaline phosphatase activity above the no cylindrospermopsin added control (Lu et al. 2021). Once organic P is hydrolyzed, Raphidiopsis can outcompete other phytoplankton by having a higher P uptake rate (Isv anovics et al. 2000). Our results indirectly support this hypothesis in that the dissolved cylindrospermopsin concentrations increased with higher N : P ratios and when P was undetectable in the media. However, our experimental design cannot separate the N or P effects because dissolved cylindrospermopsin increased with nitrate-N concentrations. We recommend exploring this extracellular cylindrospermopsin mechanism further with different phytoplankton and monitoring P-uptake rates in cylindrospermopsin producers.
Our results also have important implications for inferences about N-fixation as an ecosystem process. The importance of N in fueling cyanobacterial blooms and accelerated eutrophication has been widely debated for more than a decade (Schindler et al. 2008;Paerl et al. 2016). The growth-strategist categorization of A. flos-aquae described here provide a biological mechanism that may explain the lack of an N-fixation growth trade-off observed in some ecosystems. For example, in the experimental lakes area, Lake 227 was fertilized with N and P for 20 years from 1969 to 1989, then fertilized with only P since 1990 (Schindler et al. 2008;Schindler 2012;Higgins et al. 2018). Phytoplankton biomass during the N and P fertilization and P only fertilization were indistinguishable, with the P only fertilization dominated by Aphanizomenon (Schindler et al. 2008). If the genus Aphanizomenon is indeed more broadly evolved as a N-fixing growth-strategist, the maintenance of biomass during P fertilization alone may be a function of Aphanizomenon consistent biomass production regardless of N-supply, and their ability to increase C : N stoichiometry, potentially increasing the plasticity in their N cell quotas. A different response may have been observed if Dolichospermum had been the dominant N-fixer. Other evidence supports this hypothesis. For example, Aphanizomenon has a slower growth rate (DeNobel et al. 1998) and high nitrogenase activity (Carlton and Paerl 1989) compared to Dolichospermum, which may allow some Aphanizomenon strains to have a higher N-fixation efficiency (i.e., produce the same biomass as non-N-fixing cultures); thus decreasing the Nfixation growth trade-off. Future studies examining trade-offs in diazotrophic cyanobacteria are needed in freshwater ecosystems that will reveal mechanisms of how N-fixation effects aquatic ecosystem functioning and processes (Marcarelli et al. 2022).
Our results highlight how changing the resource N : P can affect the potential risks of cyanobacteria toxin production and associated interdependence on diazotrophic cyanobacteria and their functional traits. Historically, lake and water managers have focused on controlling P inputs while leaving N inputs largely unmitigated, which has caused an increase in the N : P of lakes dominated by wastewater inputs (Tong et al. 2020). While eutrophic lakes tend to experience stoichiometric N deficits relative to P , higher microcystin concentrations occur in high total N lakes (Yuan et al. 2014) under relatively balanced N : P (Guildford and Hecky 2000). Our results indicate, if an Aphanizomenon bloom occurred in a eutrophic or hypereutrophic lake with a low N : P, we would expect low cylindrospermopsin concentrations with only a small proportion of dissolved cylindrospermopsin. Whereas, if a bloom were to occur in a eutrophic lake with a N : P higher than 30, we would expect high cylindrospermopsin concentrations with a higher proportion of dissolved cylindrospermopsin. Our results suggest that cylindrospermopsin and potentially other toxin concentrations and quotas are in part regulated by freely available N and P; therefore, to fully understand cylindrospermopsin risks to public health and the environment, dissolved nutrients need to be monitored along with toxin levels during surveillance activities.

Data availability statement
All data used in this manuscript are available in the Dryad repository doi: 10.5061/dryad.zcrjdfnh4.