Nodularia spumigena is a toxic, filamentous planktonic, heterocystous, gas-vacuolate cyanobacterium, which is representative of surface bloom-forming cyanobacteria in brackish waters. The strain N. spumigena CCY9414 was initially isolated from samples collected from surface water in the Bornholm Sea (Hayes and Barker, 1997; Stal et al., 2003; Voss et al., 2013). Before the experiment, Nodularia CCY9414 cells were pre-cultivated in P-replete medium for 10 days to induce exponential growth. The P-replete medium was composed as follows: 33% ASNIII and 67% BG11 medium mixture [as described by Hagemann et al. (2019)], with 0.02 g L^–1 K₂HPO₄ in the modified ASNIII (Rippka et al., 1979) and 0.00078 mg L^–1 K₂HPO₄ in BG11 medium modified with 20 mM TES buffer to pH 8 (Rippka et al., 1979). Omitting the inorganic N-source nitrate induced N₂-fixing conditions. Furthermore, the final NaCl concentration of the growth medium was set to 10.3 g L^–1, which is corresponding to the salt optimum of Nodularia CCY9414 (Möke et al., 2013). Sterile cell culture flasks (Greiner Bio-One GmbH, Frickenhausen, Germany) were used for both pre-cultivation and afterward for the experimental cultures. Each culture flask contained 10 mL of starting culture suspension and 90 mL of fresh medium and was manually mixed daily. Incubations were performed at a temperature of 19.5 – 20°C at a light/dark cycle of 16 h/8 h. During the light phase, cell suspensions were constantly exposed to 40 μmol photons m^–2 s^–1. Every 5 days the cells were transferred into fresh P-replete medium to prolong the exponential growth until the planned start of the experiment.

At the beginning of the experiment, all the pre-cultures were combined and mixed in a glass bottle for a total volume of 500 mL. Filaments were then separated from the medium by filtration through glass fiber filters (25 mm circle diameter; GE Healthcare, Chicago, IL, United States). Half of the biomass on the filters were washed off and inoculated into 1.2 L of fresh medium but without added inorganic phosphate (P-deplete), which was then divided into 12 culture flasks with 100 mL suspension each. These P-deplete cultures contained only traces of inorganic phosphate impurities from other chemicals in the medium and from the inoculation of the pre-experimental cultures. The final concentration of o-phosphate was less than 1 μM. The remaining half of the biomass was washed off and inoculated into 1.2 L of fresh P-replete medium equally divided in 12 culture flasks. Samples were taken by sacrificing three culture flasks of each P condition at the beginning of the experiment (d0), after 7 (d7) and 14 days (d14), and at the end of the experiment after 21 days (d21). The experiment was repeated two times independently.

For dry weight estimations, three 5 mL aliquots from each culture were filtered on pre-weighted glass fiber filters of 25 mm diameter (GE Healthcare, Chicago, IL, United States) and dried at 65°C overnight. The mean dry weight of these three filters was taken to obtain a representative estimate for each culture flask.

Cells for polyphosphate quantification were collected from 3 mL aliquots on polycarbonate filters of 0.22 μm pore size and 22.5 mm diameter (GE Healthcare, Chicago, IL, United States), placed on ice and stored at −80°C until further processing. Extraction and fluorometric quantification of polyphosphate was done as described by Martin and Van Mooy (2013). Briefly, reagents were obtained from the following providers: Proteinase K (BP1700) was obtained from Sigma-Aldrich Chemie GmbH (Schnelldorf, Germany); lysozyme (BP535), ADP (A2754), Ambion recombinant DNase (AM2235) and RNase cocktail (AM2286), and DAPI (4′,6-diamidino-2-phenylindole) were from Thermo Fischer Scientific (Langenselbold, Germany). Before the polyphosphate extraction, cells on filter were dried for 5 h at 35°C to estimate their dry weight to standardize polyphosphate content. Then, cells were scraped off the filters with a spatula and resuspended in 3 mL polyphosphate buffer (20 mM HEPES, 100 mM NaCl, 2 mM EDTA, and 2 mM MgCl₂). The dissolved polyphosphate extraction required eight freeze-thaw cycles (freeze at −20°C for 2 h and 30 min, thaw at 35°C for 40 min) for lysis of the cells. Subsequent boiling step and enzymatic digestion were performed as described in the method of Martin and Van Mooy (2013). To stain the dissolved polyphosphates 5 μL of DAPI (1 mg mL^–1) were added to each extract. The measurements were made on a Tecan Infinite 200 PRO multimode plate reader (Tecan Austria GmbH, Grödig, Austria) at an excitation wavelength of 415 nm and emission wavelength of 550 nm, with an integration time of 500 μs. Fluorescence emission spectra were acquired at 415 nm and emission from 230 nm to 850 nm in 1-nm increments and integrated for 20 μs. All bandwidths were 5 nm. The calibration curve was constructed from commercial polyphosphate with a chain length of 45 ± 5 residues (Sigma-Aldrich, S4379) in the range of 0.2–7.0 nmol polyphosphates.

During the second iteration of the experiment, aliquots of 10 mL were collected from each culture flask to measure the level of the o-phosphate (PO₄^3–) in the medium. Each aliquot was filtered through a glass fiber filter (25 mm circle diameter; GE Healthcare, Chicago, IL, United States), and o-phosphate and ammonium concentrations of the filtrate were measured colorimetrically according to Grasshoff et al. (2009) by means of a Seal Analytical QuAAtro constant flow analyzer (Seal Analytical GmbH, Norderstedt, Germany).

An aliquot of 1.5 mL was collected from each culture, fixed with 4% formaldehyde (v/v) and stored at −20°C until DAPI staining. After thawing, the filaments were harvested on polycarbonate membrane filters (0.88 μm pore size, 22.5 mm circle diameter, GE Healthcare, Chicago, IL, United States) and stained with 50 μl of DAPI solution (1 mg mL^–1, Thermo Fischer Scientific, Langenselbold, Germany). Staining was performed in the dark at room temperature for 2 min. Then, the dye was filtered off, the filter was rinsed with ultra-pure water and dried in the dark at room temperature for 5 min. Polyphosphates in the stained filaments were visualized under a fluorescence microscope (Axioskop 2 mot PLUS, Carl Zeiss, Jena, Germany) with the specific DAPI filter set (excitation: BP 390/22, beam Splitter: FT 420, emission: 460/50). Polyphosphate chains longer than 15 P-subunits formed a complex with the dye that increased its fluorescence and shifted its absorption spectrum from 456 nm – when only DNA is visualized via blue fluorescence emission – to 526 nm via bright yellow fluorescence of the accumulated polyphosphate granules (Tijssen et al., 1985; Ohtomo et al., 2008; Diaz and Ingall, 2010).

Filaments from aliquots of 30 mL of each culture were transferred to a 50 mL tube (Sarstedt, Nümbrecht, Germany) and immediately fixed for RNA extraction with 6 mL of a solution containing 95% (v/v) ethanol (molecular biology grade; Roth, Karlsruhe, Germany) and 5% (v/v) Roti-Aqua-Phenol (Roth, Karlsruhe, Germany). The fixed suspension was incubated in the dark at room temperature for 20 min to complete the collapse of gas vesicles before each sample was placed on ice. Then, cells were harvested by centrifugation at 12.000 g at 4°C for 8 min. The supernatant was quickly removed, and the collected pellets were flash frozen in liquid nitrogen until being processed for RNA isolation. The cell pellets were then suspended in 1 mL TRIzol reagent (Sigma-Aldrich, Steinheim am Albuch, Germany) and three spoons of acid-washed mixed glass beads of 425–600 μm (Sigma-Aldrich, Steinheim am Albuch, Germany) were added for bead beating for 3 × 30 s. Then, the suspension was heated at 65°C for 10 min and RNA was isolated as described by Steunou et al. (2006). RNA extracts were treated three times with two units of DNase I RNase-free (New England Biolabs, Frankfurt am Main, Germany) for 30 min as recommended by the manufacturer. DNase I was inactivated and removed with phenol/chloroform and total RNA was then precipitated with 3 M Na-Acetate pH 5.2 and 2.5 volumes of absolute Ethanol. RNA extracts were checked for DNA contamination by PCR using primers specific for the gene encoding Fe-superoxide dismutase subunit B (sodB forward: GACTCCTCTAAGGTGGGAATC; sodB reverse: CCCAGACATCCAAGGTTAAG) as was previously done by Kopf et al. (2015).

Non-stranded rRNA-depleted libraries were prepared by the sequencing company LGC (LGC, Biosearch Technologies, Berlin, Germany) for 12 samples, collected at d7 and d14 for both P conditions from the second iteration of the experiment. Briefly, the total RNA was first depleted of rRNA using the Pan-Prokaryote riboPOOL kit (siTOOLs Biotech). The RNA was then converted into cDNA using the NEBNext RNA First Strand Synthesis and NEBNext RNA Second Strand Synthesis Modules (New England Biolabs). For the preparation of the indexed Illumina libraries the Encore Rapid DR Multiplex System 1–96 (NuGEN) was used. Libraries were amplified with 12 cycles and sequenced on Illumina NextSeq500/550.

Between 6.8 and 11.6 million single 75 bp reads were generated per sample. Adapter-clipped reads provided by the sequencing company were quality trimmed with BBDuk (Bushnell, 2014) using a sliding window approach with a window size of 4 bp and an average base quality of 15. Poly-G repeats longer than 10 bp were removed and reads shorter than 50 bp were discarded. Quality-trimmed reads were mapped against the reference genome of Nodularia CCY9414 (NCBI RefSeq accession: GCF_000340565.2) using the program bwa-mem (Li, 2013). Remaining hits to ribosomal RNA genes were excluded. Mapping results were further filtered to remove secondary and supplementary alignments, as well as alignments shorter than 50 bp and those with less than 95% sequence identity across the whole read to the reference. Read counts per gene were then calculated with featureCounts (Love et al., 2014) and converted to transcript percentages accounting for variable gene length. Operons were predicted with OperonMapper¹ (Taboada et al., 2018). To perform a functional enrichment analysis based on the KEGG pathway hierarchy, the reference genome of Nodularia CCY9414 was re-annotated against KEGG (release January 2022) using diamond blastp version 2.0.14.152 (Buchfink et al., 2021) in sensitive mode, retaining hits with an e-value below 1e-5 and a blast score ratio of more than 0.4. KEGG ortholog (KO) annotations were assigned based on the best hit. Genes without a KO assignment in the blastp search were attributed a KO number according to kofamscan version 1.3.0 (Aramaki et al., 2019) at an e-value of 0.01. Mapping of genes to KEGG pathways was performed based on KO assignments, excluding pathways from overview and structural maps and those exclusive to viruses or eukaryotic organisms. The transcriptomic reads and processed feature counts are accessible on the GEO database² with the following accession number: GSE213384.

All the statistical data analyses and visualization were performed in R (R Core Team, 2021) using the additional packages tidyverse (Wickham et al., 2019), DESeq2 (Love et al., 2014), vegan (Oksanen et al., 2020), car (Fox and Weisberg, 2019), emmeans (Lenth, 2022), multcomp (Hothorn et al., 2008), multcompView (Graves et al., 2019), lmerTest (Kuznetsova et al., 2017), hrbrthemes (Rudis, 2020), and RColorBrewer (Neuwirth, 2014). Post-processing of the figures was performed using the software Inkscape (Inkscape Project, 2020). Further details about the bioinformatic sequence processing and statistical data analysis are available on http://doi.io-warnemuende.de/10.12754/misc-2022-0005.

Dry weight and polyphosphate concentrations were analyzed in a generalized linear mixed model to assess the effect of sampling time point and P conditions with experiment iteration as random factor. To meet the assumption of normality, polyphosphate concentrations were square-root transformed. Outlier observations with a Cook’s distance of more than 4 divided by sample size (Cook and Weisberg, 1984; Williams, 1987) were removed from the analysis. Post hoc tests, i.e., pairwise comparisons between experimental conditions and sampling time points, were implemented in emmeans (Lenth, 2022). Differential gene expression was assessed using DESeq2 (Love et al., 2014) between P-replete and P-deplete conditions at each sampling time point and between d7 and d14 in each P treatment. Genes were detected as differentially expressed at a Benjamini–Hochberg adjusted p-value of 0.1 and an absolute log2-fold change of at least 1. Functional enrichment analysis was conducted for each pathway using the proportion of genes per pathway of the total number of genes in the genome in a X² goodness-of-fit analysis. The aim was to assess if the number of differentially expressed genes per pathway was higher than expected by chance. Significance was assessed at a family-wise error rate of 0.05 after Benjamini–Hochberg adjustment of p-values to account for multiple testing. Cases with expected frequencies below 1 were marked as potentially unreliable in the results.

All experiments were performed under conditions that are prevailing in the Baltic Sea during Nodularia spp. summer blooms. Dry weight (DW) measurements were performed for each culture in both conditions (P-deplete and P-replete) to evaluate growth throughout the two independent experiments. The increase in biomass was faster in P-replete than in P-deplete conditions (Figure 1A). However, even the cyanobacteria in the P-deplete medium were able to almost triple their biomass throughout the 3 weeks’ incubation time, whereas P-replete conditions permitted a fourfold increase.

In addition, o-phosphate and ammonium concentrations were measured weekly. In media of P-deplete cultures, the amount of o-phosphate was always near the detection limit with values between 0.9 and 0.2 μM (Figure 1B). In the P-replete medium, o-phosphate concentrations were approximately 170 μM and dropped during the first 7 days of the experiment to about 50% of the initial amounts. Almost all o-phosphate was consumed by the cells in the P-replete conditions after 14 days, i.e., at that time point similarly low o-phosphate levels were detected in P-deplete and P-replete cultures. In contrast to o-phosphate, similar levels of ammonium, released from the N₂-fixing filaments, were found in P-replete and P-deplete cultures during the entire experimental period. Its amount was always approximately 100 μM and did not differ significantly with time between the two experimental treatments (Supplementary Figure 1).

Polyphosphates were quantified in filaments at all sampling time points in P-replete and P-deplete cultures in two independent experiments (Figure 1C). At time point 0, cultures of both P conditions started with a relatively high internal polyphosphate amount of approximately 2.5 μmol g^–1 DW. This amount continuously decreased in filaments grown under P-replete conditions, although the rate of decrease differed between the two experiment iterations. Especially in the second iteration of the experiment, the polyphosphate content decreased to less than 50% already after 7 days, when still substantial amounts of o-phosphate were available for the cells (Figure 1B). In contrast, filaments grown under P-deplete conditions did not significantly alter their internal polyphosphate pool during the 3-week time period, despite the observed divergent measurements between the experiment iterations at day 7 (Figure 1C). In addition to the chemical quantification of polyphosphate, we used DAPI-staining to obtain a qualitative picture of polyphosphate accumulation in different cell types of the Nodularia CCY9414 filaments. Stained filaments from P-replete and P-deplete cultures showed yellow inclusions at each time point in vegetative cells and in heterocysts (Supplementary Figure 2).

RNA was isolated and sequenced from Nodularia CCY9414 cells cultivated for 7 and 14 days (d7 and d14) under P-replete and P-deplete conditions, respectively, during the second experiment iteration. The RNA-seq approach enabled the differential expression analysis of 4752 non-ribosomal genes (Supplementary Data 1). Thereby, a larger number of genes showed increased compared to decreased expression in the three comparisons: day 7 P-replete relative to day 7 P-deplete conditions, day 14 P-replete relative to day 14 P-deplete conditions, and day 14 P-replete relative to day 7 P-replete conditions.

At day 7, 112 genes were up-regulated and 87 down-regulated comparing P-deplete and P-replete cultures, respectively, whereas these numbers increased after 14 days to 344 up-regulated and 231 down-regulated genes (Figure 2). Among them, 77 genes were commonly up-regulated in P-deplete cultures at the two time points and 40 remained down-regulated. Furthermore, the gene expression changes at day 14 compared to day 7 in P-replete cultures resembled the changes observed in P-deplete versus P-replete conditions when compared at the same sampling time point. Overall, 28 of the 88 up-regulated genes at day 14 in P-replete cultures were also up-regulated at day 7 in P-deplete Nodularia CCY9414 filaments (Figure 2). A similar relative overlap is also found between down-regulated genes in day 14 P-replete relative to day 7 P-deplete (each compared to day 7 P-replete) cultures. These relations indicate that during the long-term growth under initially P-replete conditions many P-limitation-induced genes became up-regulated, which is consistent with the complete consumption of o-phosphate between 7 and 14 days in these cultures (Figure 1B).

Specific functional pathways were over-represented among the differentially expressed genes under specific P conditions (Figure 3). Especially at day 7, a disproportionally large number of genes up-regulated under P-deplete conditions were recruited from ABC transporters, among them 17 transporters for different P sources, and many genes for phosphonate and phosphinate metabolism. In addition, there were hints that P starvation had marked influence on the overall cell metabolism, e.g., many genes encoding regulatory proteins from two-component systems or enzymes involved in the metabolism of secondary metabolites as well as the oxidative pentose-phosphate (OPP) pathway were higher expressed, whereas genes for photosynthetic complexes or enzymes of the Calvin-Benson cycle appeared down-regulated at day 7 (Supplementary Data 1). After 14 days of growth without P, many more genes became stimulated. In addition to the previously mentioned P-associated genes, genes for proteins involved in energy metabolism, such as photosynthesis, oxidative phosphorylation and carbon fixation, were among the functional groups enriched in low P-stimulated genes after 14 days (Figure 2). Many genes encoding P limitation-related transport or regulatory proteins also became higher expressed when we compared the gene expression of P-replete cultures at day 14 to day 7, when all available o-phosphate had been consumed from the medium (Figure 1B).

In the initial genome annotation study, the authors provided a comprehensive overview of putatively P-associated genes in Nodularia CCY9414 (Voss et al., 2013). To confirm their involvement in the acclimation to P-limitation in our experimental set-up, the relative expression of all these genes was reviewed in the RNA-seq data (Supplementary Table 3 and Supplementary Data 1). This comparison permitted an experimentally supported annotation of the specific P-related stimulon in Nodularia CCY9414 (Table 1).

Approximately half of the previously suggested P-associated genes (Voss et al., 2013) were among the differentially expressed genes identified in our experiment, comprising genes which exhibited the strongest responses including transport systems for different inorganic and organic P-sources (Supplementary Figure 3). Among them were two Pst systems (organized in operons 1909 and 3425) for o-phosphate that were concordantly stimulated but to a different extent. The operon 1909 that encodes for proteins similar to the high affinity but low velocity Pst2 system in Synechocystis sp. PCC 6803 (Pitt et al., 2010) is about three times higher expressed than the genes for the lower affinity but higher velocity Pst1 system. In addition, a large chromosomal region was found on which many P-regulated genes were situated. It comprises the genes nsp7450 to nsp7590 that encode at least two phosphonate transport systems in the operons 515 and 513, which were induced after 7 as well as 14 days of P-limitation (Table 1 and Figure 4). Interestingly, a third annotated phosphonate uptake system encoded by the operon 1262 was weakly down-regulated after 7-day growth under P-deplete conditions (Table 1). Moreover, the phosphite ABC transporter encoded by the operon 2311 was also stimulated after 7 and 14 days of P-limitation. In addition, many genes encoding alkaline or acid phosphatases, phytases and related proteins known to be able to release o-phosphate from different organic P-sources belonged to the most strongly induced genes. Among them, the alkaline phosphatase NSP_RS05785 was predicted with high confidence by TargetP³ to possess a transit peptide at the N-terminus, which indicates that this enzyme is likely released from the cell to acquire extracellular P (Table 1).

However, many genes that were previously suggested to be part of the P-stimulon (Voss et al., 2013) did not show pronounced phosphate-related gene expression changes, such as metallophosphoesterases, haloacid dehalogenase-like hydrolases and the majority of the arsenate-related genes with a few exceptions (Supplementary Table 3 and Supplementary Data 1). Furthermore, most genes annotated to be involved in polyphosphate synthesis and mobilization did not clearly change the expression upon P-limitation. Only the ppk gene, encoding the polyphosphate synthesizing kinase, became significantly higher expressed after 14 days of growth under P-deplete conditions, which was consistent with the largely stably maintained polyphosphate pool in these filaments (Figure 1C). Finally, despite the induction of many transporters for different P-sources, the expression of the related transcriptional regulators, i.e., the two-component regulatory system PhoB (SphR) and PhoR (SphS), were not significantly changed (Supplementary Table 3 and Supplementary Data 1). Only one LysR-type transcriptional regulator that is found downstream of the operon 2311 encoding the phosphite-specific ABC transporter showed significantly enhanced expression under P-limiting conditions (Table 1).

In addition to the above-mentioned genes encoding proteins that were likely directly involved in the acclimation to P limitation, many more genes appeared to be directly or indirectly P-regulated in Nodularia CCY9414. Amongst them, many genes encode hypothetical proteins with no functional annotation, i.e., at the time point day 7 this group comprised approximately 25% all up-regulated genes under P-deplete conditions (Supplementary Data 1). However, several functionally annotated and physiologically important genes were also strongly induced in P-starved cells (Table 2). One example is represented by three genes (new locus tags NSP_RS14115, NSP_RS14120, and NSP_RS14125) encoding proteins involved in the synthesis of a CTB family bacteriocin, which were induced after 7 and 14 days of growth in the P-deplete medium. The genome of Nodularia CCY9414 harbored a second cluster of genes (NSP_RS06270, NSP_RS06275, NSP_RS06280, NSP_RS06285, and NSP_RS06290) for bacteriocins, which was not differentially expressed under different P conditions (Supplementary Data 1). Furthermore, several genes associated with bloom formation of Nodularia CCY9414 appeared to be up-regulated under P starvation. Amongst them, several of the encoded proteins were related to N₂ fixation, such as the Mo-dependent nitrogenase C-terminal domain-containing protein (NSP_RS06835), the nitrogen fixation protein NifX (NSP_RS18010), the nitrogenase-stabilizing/protective protein NifW (NSP_RS18025), which were higher expressed at day 7 in N-deplete filaments (Table 2). Additional bloom-related genes became induced after 14 days of growth under P-deplete conditions. Notably, a cluster of genes encoding different gas vesicle constituents, such as GvpA and GvpC (operon 1062: NSP_RS06890, NSP_RS06895, NSP_RS06900, as well as NSP_RS06910), or a non-ribosomal peptide synthetase operon 3225 (NSP_RS21715, NSP_RS21720, and NSP_RS21725) were among them. It further seemed that after 2 weeks of P starvation, the iron homeostasis in Nodularia CCY9414 was affected, because genes for iron uptake systems, such as the TonB receptor and the iron-siderophore ABC transporter substrate-binding protein (NSP_RS05344 and NSP_RS05350), as well as iron ABC transporter permeases (NSP_RS05360 and NSP_RS05365) became up-regulated (Table 2). Finally, the sugar-phosphate metabolism appeared to be altered due to strong P-limitation. Especially genes for enzymes involved in the OPP pathway, including the entrance enzyme glucose 6-phosphate dehydrogenase (zwf, NSP_RS22185), were found to be up-regulated after 7 and 14 days of P starvation.