Calm on the surface, dynamic on the inside. Molecular homeostasis of Anabaena sp. PCC 7120 nitrogen metabolism

Nitrogen sources are all converted into ammonium/ia as a first step of assimilation. It is reasonable to expect that molecular components involved in the transport of ammonium/ ia across biological membranes connect with the regulation of both nitrogen and central metabolism. We applied both genetic (i.e., Δ amt mutation) and environmental treatments to a target biological system, the cyanobacterium Anabaena sp PCC 7120. The aim was to both perturb nitrogen metabolism and induce multiple inner nitrogen states, respectively, followed by targeted quantification of key proteins, metabolites and enzyme activities. The absence of AMT transporters triggered a substantial whole-system response, affecting enzyme activities and quantity of proteins and metabolites, spanning nitrogen and carbon metabolisms. Moreover, the Δ amt strain displayed a molecular fingerprint indicating nitrogen deficiency even under nitrogen replete conditions. Contrasting with such dynamic adaptations was the striking near-complete lack of an externally measurable altered phenotype. We conclude that this species evolved a highly robust and adaptable molecular network to maintain homeostasis, resulting in substantial internal but minimal external perturbations. This analysis provides evidence for a potential role of AMT transporters in the regulatory/signalling network of nitrogen metabolism and the existence of a novel fourth regulatory mechanism controlling glutamine synthetase activity.

tonnes of fixed N per year (Rascio & La Rocca, 2013). Cyanobacteria are also involved in symbiotic associations, with reduced carbon delivered to cyanobacteria in order to sustain BNF (Backer et al., 2018). An example of such symbiotic associations is the aquatic fern Azolla caroliniana, which receives fixed N from a filamentous cyanobacterium (Anabaena azollae) hosted in the ovoid cavities of the plant's leaves (Lechno-Yossef & Nierzwicki-Bauer, 2005).
In their free-living form, planktonic Anabaena spp. make a significant contribution to the carbon and nitrogen economy of multiple ecosystems (Kellar & Goldman, 1979). Anabaena sp. PCC 7120 (henceforth 7120) is an isolated strain showing high genome sequence similarity with A. azollae and is commonly used as a model organism to investigate cyanobacterial N-fixation (Herrero, Stavans, & Flores, 2016).
As nitrogenases are oxygen-sensitive, photosynthesis-driven BNF calls for spatial and/or temporal separation between the metabolic pathway fuelling energy/carbon inputs (i.e., photosynthesis) and N-fixation ( Figure 1). Under diazotrophic conditions, 7120 differentiates 5-10% of its cells into specialized N-fixing heterocysts, following a highly F I G U R E 1 Schematic overview of major molecular players regulating N metabolism and the metabolic interaction between heterocysts (HC) and vegetative (VC) cells in Anabaena sp. PCC 7120, in diazotrophic conditions. Nitrogenase (N 2 ase) fixes one molecule of atmospheric N 2 into two molecules of NH 3 in heterocysts (Inomura, Bragg, & Follows, 2017), using reducing power (NADPH) from the catabolism of carbon -compounds (sucrose) photosynthesized in vegetative cells (Cumino, Marcozzi, Barreiro, & Salerno, 2007;Nürnberg et al., 2015) and energy (ATP) from the residual photosynthetic activity in heterocysts [i.e., cyclic electron flow around photosystem I (PSI) (Cardona & Magnuson, 2010)]. NH 3 is then assimilated through glutamine synthetase (GS) via the amidation of glutamate (Glu) to glutamine (Gln) (Forchhammer & Selim, 2019). GS activity is controlled through posttranslational inactivation (Bolay, Muro-Pastor, Florencio, & Klähn, 2018) by IF7A (Galmozzi, Saelices, Florencio, & Muro-Pastor, 2010). Glutamate dehydrogenase (GDH) marginally contributes to the assimilation flux of fixed N, catalysing the reversible conversion of 2-oxoglutarate (2-OG) to Glu (Meeks, Wolk, & Lockau, 1978). Subsequently, in vegetative cells (Martín-Figueroa, Navarro, & Florencio, 2000), glutamine oxoglutarate aminotransferase (GOGAT) catalyses the transfer of the amine group from Gln to 2-OG, generating two molecules of Glu. As N metabolism spans different cell types, a coordinated exchange of metabolites (i.e., sucrose, Gln, Glu and 2-OG) between vegetative cells and heterocysts via septal junctions [plasmodesmata (Mullineaux et al., 2008)] is required to maintain metabolic homeostasis. 2-OG is also a metabolic intermediate of the tricarboxylic acid (TCA) cycle [synthesized from isocitric acid (IA) by isocitrate dehydrogenase (IDH)], thus connecting N and C metabolism at a central point. Cyanophycin (CPG) is a polymer playing an important role in the distribution of N among the two cell types. In this figure, CPG was omitted and we refer the reader to Figure S6 in Data S1 for a complete description of its metabolism. N metabolism homeostasis is controlled by a molecular network, including the proteins NtcA, PipX and PII. When external N is available, PII is not phosphorylated and it sequesters PipX, preventing its biding to NtcA and consequently its activation. When N is limiting, PII is phosphorylated, freeing PipX, which ultimately binds and activates NtcA (Valladares et al., 2011). The red cross indicates the knockout (KO) mutant Δamt used in this work (Paz-Yepes, Merino-Puerto, . In this simplified scheme, we localized AMT transporters in the membrane of vegetative cells, according to Merino-Puerto, Mariscal, Mullineaux, Herrero, and Flores (2010)). Major molecular players targeted in this work are highlighted by a grey square or in bold, respectively for proteins and metabolites [Colour figure can be viewed at wileyonlinelibrary.com] regulated developmental pattern [i.e., a single heterocyst every 10-20 cells (Kumar, Mella-Herrera, & Golden, 2010)]. Heterocysts undergo a deep metabolic and structural remodelling to enable efficient N-fixation (Golden & Yoon, 1998). The oxygen-evolving photosystem II (PSII) is dismantled, carbon fixation is avoided, photorespiratory activity is increased during differentiation (Valladares, Maldener, Muro-Pastor, Flores, & Herrero, 2007), flavodiiron proteins catalyse the reduction of molecular oxygen to water (Ermakova et al., 2014) and cells are surrounded by a thicker cell envelope [through the deposition of two additional envelope layers, that is, an inner glycolipids and an outer polysaccharides layer (Nicolaisen, Hahn, & Schleiff, 2009)] than the vegetative cells, all contributing to the required microoxic environment for N-fixation activity (Kumar et al., 2010) (Figure 1).
Heterocysts and vegetative cells have complementary metabolism, with the former providing fixed nitrogen and the latter returning reduced carbon needed to sustain BNF (Malatinszky, Steuer, & Jones, 2017). This metabolic exchange and associated networks (summarized in Figure 1) are most likely carefully coordinated in order to ensure organism-level homeostasis (Mullineaux et al., 2008). The question is, how does this molecular coordination take place?
In many organisms, including 7120, N metabolism is orchestrated by a complex signalling network with the likely aim to balance the cellular C/N ratio (Forchhammer & Selim, 2019). N and C metabolisms are in fact tightly coupled as (a) the two elements are among the four most abundant in living organisms (i.e., oxygen, carbon, hydrogen and nitrogen, respectively), calling for coordination to avoid metabolic inefficiencies, and (b) N assimilation depends on the availability of C skeleton, with shortage or oversupply strongly affecting the metabolism of N (Zhang, Zhou, Burnap, & Peng, 2018). Therefore, a properly balanced N and C metabolism is necessary for optimal growth and different levels of regulation exist to control uptake and assimilation efficiencies of both chemical species. When the C source (i.e., CO 2 in case of phototrophic metabolism) is not limiting, the regulatory mechanisms controlling C/N balance depend on both the abundance and the nature of the N sources available to the cell. Although cyanobacteria can use multiple N sources, including NH 4 + , intracellularly they are all converted to NH 4 + , the most reduced and energetically favourable N source (Robinson, 2017). Ammonia translocation across biological membranes is actively driven by AMT transporters that belong to a family of permeases widely distributed in living organisms (Javelle et al., 2007), or through passive diffusion if the external pH pushes the equilibrium towards the uncharged form (NH 3 , ammonia).
In addition, AMT proteins are also known to be involved in the regulation of N metabolism, at least in some bacteria species (Arcondeguy, Jack, & Merrick, 2001). As an example, in the purple bacterium Rhodobacter capsulatus, AMT proteins are implicated in (a) the posttranslational regulation of nitrogenase, driven by a direct interaction between AmtB and GlnK (a homolog of PII) (Tremblay & Hallenbeck, 2008), and (b) the regulation of GS activity during the switch towards diazotrophic conditions (Yakunin & Hallenbeck, 2002). Also in cyanobacteria [e.g., Synechocystis sp. PCC 6803 (Watzer et al., 2019)] AMT proteins were found to directly interact with the PII protein to prevent intracellular accumulation of ammonium/ia. 7120 bears a gene cluster including three amt genes, namely amt4, amt1 and amtB (Paz-Yepes et al., 2008). In this work, we used a knock-out (KO) mutant of the whole gene cluster in 7120 [henceforth Δamt (Paz-Yepes et al., 2008)] with the aim to investigate how a N 2 -fixing cyanobacterium responds to perturbation of N-metabolism at a wholesystem level. Although several genes and proteins in 7120 have been individually studied previously Forchhammer & Selim, 2019), it is difficult to make over-arching conclusions on the regulatory system, also as cyanobacteria differ substantially relative to heterotrophic bacteria Reitzer, 2003). The aim of this work is also to enhance our understanding that contributes towards the practical goal of eventually  Figure S2 in Data S1), suggesting that 5 mM NH 4 + is not enough to support maximal growth (i.e., growth in NO 3 − in this experiment) over the whole experimental time frame, and that it likely runs out after $48 hr ( Figure 2A). The nitrogenase activity, measured after 96 hr, confirms that 5 mM NH 4 + runs out over the course of the experiment, triggering diazotrophic growth ( Figure 2B). Moreover, the lower nitrogenase activity with respect to N 2 conditions highlights that cells in NH4 + media, at the time of sampling, are in a transitory phase before reaching the maximal N fixation potential. Interestingly, in NH4 + media, the Δamt strain shows a higher nitrogenase activity per unit of chlorophyll (Chl) than the parental strain, although that does not result in an improvement in growth, suggesting possible compensatory modifications in the following metabolic steps (e.g., N assimilation). After 96 hr, GS activity, a (supposed) central player in N metabolism in this organism , is also affected by the mutation (as in the case of fully diazotrophic conditions [N 2 ] in which Δamt strain shows a greater N assimilation activity than the WT, see Figure 2C). Moreover, when both strains are grown in N 2 , GS activity is overall lower than in the other two growth conditions. The collective data indicated that the loss of AMT resulted in no phenotypic change, but that N-metabolism had adjusted, presumably to maintain homeostasis, raising the following question: how extensive was this adaption and what molecular players were involved?
In 7120, GS is regulated both transcriptionally and post-translationally, according to the C/N status of the cell .
The abundance of the protein is transcriptionally controlled and changes according to the N source(s) as observed in the WT strain, but not in Δamt ( Figure 3A).
The abundance of GS ( Figure 3A) does not reflect its measured activity ( Figure 2C). GS activity is known to be controlled by covalent binding of the inactivation factor IF7A, in response to the C/N balance of the cell (Galmozzi et al., 2010). As shown in Figure 3B, there is a statistically significant negative correlation between GS activity and the amount of its inactivation factor IF7A for both strains across dif-  Table S1 in Data S1), suggesting other molecular players also contribute to the regulation of GS activity in this organism. The absence of AMT transporters has an effect on the IF7A/GS activity relationship ( Figure 3B).
In N replete conditions (NO 3 − ), there is no difference between the two strains, whilst under the two other N deplete conditions, the deletion of amt results in a divergence between the two strains ( Figure 3B). The genetic and environmental treatments affect GS activity through a combined variation in both GS and IF7A abundance ( Figure 3A and Figure S3 in Data S1). In particular, it is worth noting that Δamt retains GS activity even if the amount of IF7A changes, supporting the idea that other regulatory players also may be involved. Based on the available data, we hypothesised that AMT transporters are directly or indirectly involved in the regulation of GS activity during the switch towards diazotrophic conditions in 7120, as already observed in the purple bacterium R. capsulatus (Yakunin & Hallenbeck, 2002).

| AMT transporters may be involved in the signalling and/or regulation of N metabolism in 7120
In order to test this hypothesis, the internal metabolic changes taking place in 7120 WT and Δamt during the transition from N replete to deplete conditions were followed by combining physiological data with targeted proteomic and metabolomic quantification of molecular players known to be involved in the regulation of N metabolism in F I G U R E 2 Growth of 7120 WT and Δamt strains with different N sources. The two strains were cultivated in different N sources for 96 hr. Growth (A) was monitored over the course of the whole experiment, while nitrogenase activity (B) and GS activity (C) were measured after 96 hr in such cultivation conditions. Data are indicated as average ± SD of six biological replicates. Statistically significant differences between WT (white bars) and Δamt (striped bars) are indicated with an asterisk, while the same alphabet letter indicates statistically significant differences for the same strain in different growth conditions (one-way ANOVA, p-value <.05) 7120 ( Figure 1). Both WT and Δamt strains were cultivated in BG11 0 supplemented with 5 mM NH 4 + . The concentration of ammonium/ia in the media and cell growth was monitored over time ( Figure 4A).
Four different time points were chosen to investigate the physiological and metabolic status of the cells [i.e., NR (N replete conditions), ND1, ND2 and ND4 (1, 2 and 4 days, respectively, after N depletion), F I G U R E 3 GS abundance and correlation between GS activity and IF7A amount for 7120 WT and Δamt strains, after 96 hr in the growth conditions of Figure 2a. In (A), GS abundance in the two strains is indicated after cultivation for 96 hr in the three different nitrogen sources tested in this work. In (B), the correlation between GS activity (i.e., y-axis, as indicated in Figure 2C) and IF7A abundance (i.e., x-axis) is indicated, for the same time point and different nitrogen sources tested in this work. IF7A quantification data are reported in Figure S3 in Data S1. Data are indicated as average ± SD of six biological replicates. Statistically significant differences between WT (white bars and black squares) and Δamt (striped bars and red circles) are indicated with an asterisk, while the same alphabet letter indicates statistically significant differences for the same strain in different growth conditions (one-way ANOVA, p-value <.05). In panel (B), the Pearson coefficient was calculated to assess the correlation between GS activity and IF7A abundance in both strains and the Student's t test was employed to assess the significance of the correlation, with the following results: WT, Pearson correlation −.72, p-value .02, Student's t test; Δamt, Pearson correlation −.55, p-value .001, Student's t test. The correlation analysis is detailed in Table S1 in Data S1 [Colour figure can be viewed at wileyonlinelibrary.com] F I G U R E 4 Growth, ammonium/ia consumption (A) and maximal photosynthetic efficiency [(Φ PSII ), (B)] monitoring for 7120 WT and Δamt strains in BG11 0 + 5 mM NH 4 + . In (a), black and red solid and dashed lines indicate growth and ammonium/ia concentration over time for 7120 WT and Δamt strains, respectively. Cultures were sampled at four time points over the course of the experiment [NR (Nitrogen Replete), ND1, ND2 and ND4, respectively 1, 2 and 4 days after Nitrogen Deprivation]. Data are indicated as average ± SD of six biological replicates. Statistically significant differences between WT (black squares and white bars) and Δamt (red circles and striped bars) are indicated with an asterisk, while the same alphabet letter indicates statistically significant differences for the same strain in different growth conditions (one-way ANOVA, p-value <.05) [Colour figure can be viewed at wileyonlinelibrary.com] corresponding to 24, 72, 96 and 144 hr from the start of the experiment]. Former studies in 7120 mainly focused on the first 24 hr after N deprivation (Galmozzi et al., 2010;Valladares et al., 2011), as heterocyst differentiation is expected to occur within such time frame (Valladares et al., 2011). Here, instead, we opted for an extended sampling protocol in order to complement the information already available in literature with the knowledge of the metabolic/proteomic adjustments happening over a longer time-scale.
The absence of the whole amt cluster does not affect the ammonium/ia consumption rate, indicating other uptake systems or the diffusion of ammonia are enough to sustain growth in 7120, both in the first 48 hr when ammonium is present in sufficient amount (i.e., mM concentration) and after 48 hr when the culture is depleted of measurable ammonium ( Figure 4A). This is in line with earlier studies indicating that AMT transporters are only expected to be involved in ammonium uptake at low pH (<7) and in limiting concentrations [i.e., μM, (Boussiba & Gibson, 1991;Paz-Yepes, Herrero, & Flores, 2007)]. It is in fact worth noting that we started the experiment providing a sufficient amount of ammonium/ia (i.e., mM concentration) to the cultures and for the first 48 hr we are thus not in condition to maximize the uptake activity of AMT transporters. The data also confirms that 5 mM NH 4 + is not enough to support maximal growth in 7120 over a longer 96 hr culture ( Figure 2A). Over the course of the experiment, both strains mostly showed a stable pigment content, suggesting the switch towards diazotrophic conditions does not unbalance the overall N status of the cell (Murton et al., 2017) (Table 1).
However, the Δamt strain shows a greater Chl/Car ratio, mainly achieved through the accumulation of a higher Chl content than the parental strain (Table 1). This effect on the pigment composition has also a consequence on the photosynthetic performances. The photosynthetic activity in the parental strain changes over the course of the experiment, whilst in Δamt it is more stable ( Figure 4A). Moreover, the mutant shows a higher photosynthetic efficiency than WT in N replete conditions, while the difference reverses both after 24 and 48 hr under N deprivation conditions and ultimately disappears after 96 hr ( Figure 4B). Given that the Δamt mutation does not trigger an altered growth phenotype, whilst there are substantial changes to both photosynthesis and N metabolism, we hypothesised that the deletion of the whole amt cluster is in fact triggering a whole cell metabolic response in order to maintain homeostasis.
In order to validate this unanswered hypothesis, we investigated the N metabolism of 7120 more closely, targeting the same proteins and enzymatic reactions as above, but this time measured in the same time intervals as indicated in Figure 4A. Both WT and Δamt strains activate N fixation ( Figure 5A) as a consequence of ammonium/ia deprivation ( Figure 4A). The Δamt strain shows a higher nitrogenase activity than the parental strain in both ND1 and ND2, indicating a faster response to N deprivation ( Figure 5A). Abundance of NifK and NifD, encoding for α and β subunits of nitrogenase, is higher in the mutant strain (Figure 5b,C, respectively), suggesting increased N fixation activity depends at least in part on a greater accumulation of the protein complex, given also the number of heterocysts over the course of the experiment is not affected by the mutation ( Figure S4 in Data S1). The increased N-fixation activity does not translate to a greater growth rate in the mutant, however, suggesting possible compensatory modifications in downstream steps of N metabolism.
Once atmospheric N is fixed into ammonium/ia, the latter is incorporated into amino acid metabolism. GS activity was strongly regulated in both strains also in this experiment, as observed before ( Figure 2B). In N replete conditions, Δamt showed greater N assimilation activity than the parental strain (i.e., NR in Figure 6A), likely to be the cause of the increased influx of N in the central metabolism, as indicated by the increased pigment content and photosynthetic activity observed in NR conditions (Table 1 and Figure 4B). Consequently, GS activity was influenced by the change in N metabolism, especially in the earlier samples. GS activity was at first reduced and then strongly increased (i.e., ND1 and ND2 in Figure 6A), before stabilising again to the same rate observed under N replete conditions (i.e., ND4 in Figure 6A). The trend in GS activity appeared to depend on the abundance of both GS and IF7A ( Figure 6B,C, respectively), which accumulate differentially in the two strains. It is worth noting that in this experiment, the negative correlation between GS activity and IF7A abundance is not significant in the parental strain, whilst there is a significant correlation in the mutant genetic background ( Figure 6D;  Figure 4A. Data are indicated as average ± SD of six biological replicates. Statistically significant differences between WT and Δamt are indicated with an asterisk, whilst the same alphabet letter indicates statistically significant differences for the same strain in different growth conditions (one-way ANOVA, p-value <.05).
compensatory molecular mechanisms controlling GS activity might exist in this species and that the latter might come into play, as a consequence of the N status of the cell. Moreover, the deletion of the whole amt cluster strongly affects the abundance of both GS and IF7A, which impact the overall regulation of GS activity, as the GS activity is similar across several different treatments even though the amount of IF7A varies ( Figure 6D). These results strengthen the hypothesis that AMT transporters might play a direct or indirect IF7Aindependent role on the GS activity in 7120.
GS is the major entry point of fixed N in the central metabolism of 7120 . Nevertheless, other enzymes control the availability of GS substrates, thus indirectly contributing to the regulation of N assimilation. These include GOGAT (responsible for the regeneration of Glu in the GS-GOGAT cycle), IDH (responsible for the synthesis of 2-OG, a substrate of GOGAT) and GDH (involved in the reversible conversion between Glu and 2-OG) Forchhammer & Selim, 2019;Martín-Figueroa et al., 2000). Under the tested experimental conditions, the deletion of the whole amt cluster had major consequences also on the abundance of such enzymes ( Figure S5 in Data S1). GOGAT accumulation followed the same trend in both strains over the course of the experiment (i.e., strong downregulation, as a consequence of N deprivation, Figure S5A in Data S1), while IDH and GDH displayed different trends in the two genetic backgrounds ( Figure S5B,C in Data S1, respectively). These observations support the notion that the absence of AMT transporters triggers both direct and indirect effects on N metabolism in 7120.

| The absence of AMT transporters affects the master regulatory network of N metabolism
Given the impact of Δamt on both GS and nitrogenase, the question is how widespread the adjustments had rippled further into the cellular system? In order to address this question, we investigated the N metabolism more deeply, with an expanded number of protein quantification targets. The C/N balance of the cell is in fact also known to regulate the interaction between NctA, PipX and PII in 7120 (Forchhammer & Selim, 2019), which are expected to be the major molecular players controlling the metabolic remodelling in response to both the nature and availability of N source in the external environment, in 7120 ( Figure 1). The transcription factor NtcA, active mostly once bound to PipX, regulates the abundance of nitrogenase, GS and IF7A (Picossi, Flores, & Herrero, 2014). As Δamt both responds to N deprivation more quickly than the parental strain, by activating faster N-fixation ( Figure 5), and also shows major alterations in the regulation of N assimilation (greater GS activity), we wondered whether the mutation might have an effect on such master molecular regulators, which control both enzymatic steps in this species (i.e., NtcA, PII and PipX, Figure 1) and we thus quantified their abundance ( Figure 7).
Overall, Δamt accumulated more of the three proteins over the course of the experiment relative to the parental strain, suggesting the absence of AMT transporters has an extensive impact on the cellular system, involving the master regulatory network of N metabolism. This might explain the more rapid activation of Nfixation in response to N deprivation and also the effect on N assimilation, discussed above ( Figures 5 and 6). Moreover, while the abundance of the three proteins does not vary much over the course of the experiment in WT, PII does respond to N deprivation and is more abundant under NR conditions in Δamt ( Figure 7B).
This suggests PII might be degraded after induction of N deprivation, possibly as a consequence of a greater phosphorylation rate (see Figure 1 for the molecular mechanisms controlling PII/PipX interaction). These results strengthen the notion that AMT transporters might play a central role in the signalling/regulation of N metabolism in 7120.
F I G U R E 5 Nitrogenase activity (A), NifK (B) and NifD (C) abundance in both WT and Δamt strains, following N deprivation. Time points are those indicated in Figure 4A and correspond to NR (Nitrogen Replete), ND1, ND2 and ND4, respectively 1, 2 and 4 days after Nitrogen Deprivation. Data are indicated as average ± SD of six biological replicates. Statistically significant differences between WT (white bars) and Δamt (striped bars) are indicated with an asterisk, while the same alphabet letter indicates statistically significant differences for the same strain in different growth conditions (one-way ANOVA, p-value <.05)

| Metabolic remodelling as a consequence of amt deletion
Under diazotrophic conditions, coordinated metabolic interaction between vegetative cells and heterocysts is seminal for optimal growth. The absence of AMT transporters triggered an extensive remodelling at the protein level in 7120, spanning both cell types, given some of the proteins investigated in this work are known to be exclusively expressed in one of the two cell types [e.g., GOGAT in vegetative cells (Martín-Figueroa et al., 2000)]. We therefore wondered whether the same also happens at the metabolite pool level.
Cyanobacteria evolved the ability to store assimilated N in the form of cyanophycin granule polypeptide (CPG), possibly acting as a buffer to naturally varying N-fixation due to fluctuations in N supply and day/night cycles (Watzer & Forchhammer, 2018). It is however worth noting that the amount of CPG in Anabaena is expected to reach at F I G U R E 6 Regulation of N assimilation during the switch towards N deprivation in both 7120 WT and Δamt strains. (A) GS activity; (B) GS abundance; (C) IF7A abundance; (D) correlation between GS activity and IF7A abundance. Time points are those indicated in Figure 4A and correspond to NR (Nitrogen Replete), ND1, ND2 and ND4, respectively 1, 2 and 4 days after Nitrogen Deprivation. Data are indicated as average ± SD of six biological replicates. Statistically significant differences between WT (white bars and black squares) and Δamt (striped bars and red circles) are indicated with an asterisk, while the same alphabet letter indicates statistically significant differences for the same strain in different growth conditions (one-way ANOVA, p-value <.05). In panel (D), the Pearson coefficient was calculated to assess the correlation between GS activity and IF7A abundance in both strains and the Student's t test was employed to assess the significance of the correlation, with the following results: WT, Pearson correlation −.77, p-value .13, Student's t test; Δamt, Pearson correlation −.63, p-value .03, Student's t test.
The correlation analysis is detailed in Table S1 in Data S1 [Colour figure can be viewed at wileyonlinelibrary.com] best 10% of the dry weight of the cell, whilst proteins accumulate up to 60% of the cell dry weight (Simon, 1973). Therefore, the major sink of nitrogen in Anabaena is proteins, rather than CPG. Nevertheless, in heterocystous filamentous cyanobacteria, CPG accumulates at the contact sites between heterocysts and adjacent vegetative cells and is expected to regulate the transfer of fixed N from the former to the latter (Burnat, Herrero, & Flores, 2014), thus influencing metabolite exchange between the two cell types (see Figure S6 in Data S1 for a schematic overview of CPG metabolism in 7120). In order to investigate whether the metabolites exchange between the two cell types was also affected by the mutation, we studied potential alterations in the CPG metabolism ( Figure 8).
Under the tested experimental conditions, both strains accumulated CPG in NR conditions as expected (Forchhammer & Watzer, 2016) and they showed the same CPG content ( Figure 8A), suggesting the deletion of AMT transporters did not perturb N storage to the point of affecting CPG accumulation. Nevertheless, the abundance of three out of four major enzymes controlling CPG metabolism ( Figure S6 in Data S1) is significantly different between the two strains, with cyanophycin synthetase (CphA1), cyanophycin synthetase 2 (CphA2) and isoaspartyl dipeptidase (ISO) showing increased accumulation in the mutant with respect to the parental strain in NR ( Figure 8B,D,E). We therefore hypothesized that the overall CGP metabolic flux is accelerated in the mutant under NR conditions, possibly enabling a faster response to environmental changes.
When cells experience N deprivation, the cyanophycin content decreases, presumably as it is rapidly used as N source (Forchhammer & Watzer, 2016) (i.e., after 1 day of N deprivation CPG is fully consumed, Figure 8A). Nevertheless, while CPG starts building up again in the parental strain after 2 days of N deprivation, a constant consumption trend is observed over the course of the experiment in the mutant ( Figure 8A). This suggests that N fixation in the WT exceeds metabolic needs and a fraction of the assimilated N is thus stored as CPG, while in the mutant this trend is disrupted. Out of the four major enzymatic steps controlling CPG metabolism in 7120, CphA1, the major enzyme controlling CPG biosynthesis (Forchhammer & Watzer, 2016), is most affected by N deprivation ( Figure 8B-E), with the mutant showing a faster reduction in its abundance than the parental strain ( Figure 8B). It is also worth noting that CphA2, a truncated version of CphA1 catalysing the direct recycling of the β-aspartyl-arginine dipeptide into CPG [ Figure S6 in Data S1 (Forchhammer & Watzer, 2016)], is more abundant in the mutant strain in all time points ( Figure 8D), strengthening the notion that CPG metabolism is accelerated in Δamt.
The regulation of CPG accumulation in N fixing cyanobacteria is mediated by PII which in turn regulates N-acetyl-N-glutamate kinase (NAGK) activity (Forchhammer & Selim, 2019). NAGK catalyses the conversion of N-acetyl-L-glutamate to N-acetyl-L-glutamyl-phosphate, which is further converted to ornithine, from where Arg, the end-product of the pathway, is derived. Therefore, CPG biosynthesis directly follows the concentration of free Arg in the cell, as a consequence of feedback inhibition of NAGK (Watzer et al., 2015). In our experimental conditions, the free Arg concentration indeed strongly decreased upon N deprivation in both strains ( Figure 9A,B), confirming the strong reduction in CPG content upon N deprivation, observed before ( Figure 8A).
Overall, both strains display a reduction in the whole pool of free amino acids as a consequence of N deprivation ( Figure 9A,B), likely suggesting a faster turnover upon N-fixing conditions. Nevertheless, upon N deprivation, the Δamt strain also shows substantial remodelling of the free amino acid pools ( Figure 9B), relative to the parental strain ( Figure 9A). Major amino acids affected by the mutation are Lys and Asn, which display a stronger reduction in the mutant upon N deprivation, followed by Thr, Trp, Arg, Gly, Ala and Asp ( Figure 9B), which instead show minor but still relevant alterations.
These results indicate a comprehensive impact on the amino acid metabolism in 7120, as a consequence of the mutation. It is also worth noting the pool of acetyl-lysine is differentially regulated in the F I G U R E 7 Abundance of the three major molecular players regulating N metabolism in 7120. (A). NtcA; (B). PII; (C). PipX. Time points are those indicated in Figure 4A and correspond to NR (Nitrogen Replete), ND1, ND2 and ND4, respectively 1, 2 and 4 days after Nitrogen Deprivation. Data are indicated as average ± SD of six biological replicates. Statistically significant differences between WT (white bars) and Δamt (striped bars) are indicated with an asterisk, while the same alphabet letter indicates statistically significant differences for the same strain in different growth conditions (one-way ANOVA, p-value <.05) F I G U R E 8 Legend on next page. two strains upon N deprivation ( Figure 9A,B), suggesting a comprehensively different regulation of whole central metabolism (Christensen et al., 2019;Nakayasu et al., 2017).
Amino acids are substrates for the synthesis of several molecular players, important for the homeostatic control of the cell. Among them, glutathione (Cameron & Pakrasi, 2010) and ophthalmic acid (Ito, Tokoro, Hori, Hemmi, & Yoshimura, 2018) belong to a robust antioxidant buffering system which plays an important role in protecting against reactive oxygen species (ROS) generated as by-product of photosynthetic metabolism (Narainsamy et al., 2016).
Interestingly, while no major differences between the two strains were observed for ophthalmic acid upon N deprivation, the content of reduced glutathione (GSH) increased in WT ( Figure 9C) and it decreased in Δamt ( Figure 9D), suggesting the mutant suffers from redox stress upon N deprivation.

| Gln, Glu and 2-OG pools
The metabolic pool concentration of several key metabolites in N metabolism, Gln, Glu and 2-OG [ Figure 1 (Böhme, 1998;Martín-Figueroa et al., 2000;Picossi et al., 2005)], was also affected by the amt mutation. The pool of free Glu increased in both strains over the course of the experiment ( Figure 10A,B), while the concentration of Gln dropped substantially in the wild-type upon N deprivation ( Figure 10A). Interestingly, the concentration of Gln is severalfold lower in Δamt under N replete conditions and also drops upon elimination of assimilable N ( Figure 10B). Hence, the Gln/Glu ratio in the mutant is indicative of a partially deprived N metabolic status even in the presence of assimilable N ( Figure S7 in Data S1), potentially affecting also the metabolic exchange between the two cell types (no difference in the number of heterocysts was observed between the two strains over the course of the experiment, see Figure S4 in Data S1). It is however worth noting that our data describes the abundance and activity of specific proteins/metabolites and enzymes, respectively, and therefore do not provide information regarding metabolic N flux. In order to investigate the latter, metabolic flux analyses would be needed and in future studies this might enable our hypothesis of a partially deprived N metabolic status in the mutant to be confirmed. Similarly, the mutant also has a slightly lower 2-OG content than the parental strain under NR conditions ( Figure 10C), and consequently a higher 2-OG/Gln ratio ( Figure S8 in Data S1), in line with the hypothesis that 2-OG is indicative of metabolic N availability (M. I. Muro-Pastor, Reyes, & Florencio, 2001). The difference between the two strains is admittedly small, at only around 18%but then, the decrease in 2-OG following N deprivation is only about 40%, so even this small decrease could represent a change in N status.

| DISCUSSION
Biological systems can be fuelled by multiple N sources, but they are all converted to ammonium/ia before assimilation, as the latter is the and this function is expected to be conserved also in cyanobacteria (Boussiba & Gibson, 1991).
In this work we exploited a KO mutant of the whole amt cluster  Figure 4A). Data are indicated as average ± SD of six biological replicates. Statistically significant differences between WT (white bars) and Δamt (striped bars) are indicated with an asterisk, whilst the same alphabet letter indicates statistically significant differences for the same strain in different growth conditions (oneway ANOVA, p-value <.05). See Figure S6 in Data S1 for an overview of CPG metabolism. It is worth noting that Isoaspartyl dipeptidase involved in CPG metabolism has recently been named ladC (Flores, Arévalo, & Burnat, 2019) ammonium/ia provided by the media feed. However, that would substantially limit opportunities for the number of replicates and treatments.
We cultivated both 7120 WT and Δamt strains in different N regimes (i.e., different N sources and N replete/deplete conditions).
The underlying idea was to trigger different inner N states and investigate them through physiological, proteomic and metabolomic analyses. Upon N deprivation, Anabaena sp. PCC 7120 differentiates a fraction of its cells into heterocysts to enable efficient N fixation (Golden & Yoon, 1998;Kumar et al., 2010), posing several limitations to integrated studies such as this one. One of them is the need to process the samples as a mixture of the two cell types, in order to avoid metabolic changes that are inevitable consequence of physical separation (Ermakova et al., 2014). This necessary choice forgoes discrimination of the metabolic status of the two cell types. Nevertheless, some proteins and metabolites are unique to one of the two cell types (Martín-Figueroa et al., 2000), enabling to partially overcome such limitations. Moreover, we did not observe any difference in the ratio of heterocysts to vegetative cells in response to any of the genetic or environmental treatments investigated in this work ( Figure S4 in Data S1).

| Lack of AMT transporters triggers a substantial response at the whole cell level, but does not induce any visible phenotypic change
In the constant laboratory conditions tested in this work, we observed AMT transporters are not essential to support growth of 7120, regardless of the N source used to sustain the central metabolism ( Figures 2A and 4A), as previously reported in the same organism  Figure S3 in Data S1 and Figure 6C) and (b) changes in the abundance of GDH, GOGAT and IDH, which re-generate its substrates ( Figure S5 in Data S1). It is worth noting that some of these observations correlate with what has already been observed in other organisms, such as the photosynthetic purple bacterium R. capsulatus, in which the absence of the AMTB transporter influenced both nitrogenase and GS activity, thus strengthening the hypothesis that such proteins might share the same role in the regulation of N metabolism, even in distinct bacteria species (Yakunin & Hallenbeck, 2002). Moreover, we also observed that the pool of free amino acids ( Figure 9A,B), redox markers ( Figure 9C,D) and metabolites with a key role in N metabolism orchestration (i.e., Gln, Glu and 2-OG, Figure 10) is affected by the mutation. The Δamt mutant also displays substantial changes in photosynthetic performances and pigment content with respect to the parental strain ( Figure 4B and Table 1).
It is worth noting that the biochemical changes observed in the mutant do not translate in any phenotypic difference with respect to the parental strain (i.e., growth is unaffected, Figures 2A and 4A), highlighting the strong robustness of the biological system under investigation. The latter most likely depends on its ability to undergo this very substantial homeostatic adjustment at the whole cell level, as a consequence of both genetic (i.e., Δamt) and environmental treatments (i.e., different N regimes).

| Lack of AMT transporters induces metabolic adaptation spanning both C and N metabolism, with a potential impact on the metabolites exchange between heterocysts and vegetative cells
As in many biological systems, N and C metabolism is expected to be tightly coupled (Zhang et al., 2018) also in 7120, thereby maintaining a properly balanced C/N ratio even when exposed to external perturbations (Forchhammer & Selim, 2019).  Figure 4a. Glu (white bars) and Gln (grey bars) content is split in two distinct panels for WT (A) and Δamt (B). (C) 2-OG content in WT (white bars) and Δamt (striped bars). Results come from the same amount of biomass for both strains and for different growth conditions. Data are indicated as average ± SD of six biological replicates. Statistically significant differences between WT and Δamt for each metabolite at a specific time point are indicated with an asterisk, whilst the same alphabet letter indicates statistically significant differences for the same strain and metabolite, in different growth conditions (one-way ANOVA, p-value <.05) sources and abundance) treatments were used to perturb both such parameters and investigate the response of 7120 at a whole cell level.
The mutant displayed substantial changes in the abundance of several metabolites (Figure 9), including Gln, Glu and 2-OG. Among them, it is worth noting the difference in the free pool of acetyl-lysine ( Figure 9A,B) which suggests the overall central metabolism regulation might be comprehensively affected as a consequence of the mutation (Christensen et al., 2019;Nakayasu et al., 2017). The observed differences in the free pool of acetyl-lysine might reflect either changes in total protein acetylation, or else changes in the turnover rates of acetylated proteins, with a potential regulatory role in both photosynthesis and carbon metabolism, as suggested in Synechocystis sp. PCC 6803 (Mo et al., 2015). Further confirmation in the lab is however needed to clarify the regulatory role of protein acetylation in 7120.  Figure 10C) and this correlates with the observed increase in the pool of free Glu ( Figure 10A,B), as cyanobacteria lack 2-OG dehydrogenases and therefore 2-OG is mainly used for the biosynthesis of Glu or other Glu-derived compounds (Herrero et al., 2001). It is worth noting that this trend is not affected by the mutation (Figure 10C), which instead induces a reduction in the content of 2-OG in NH 4 + -replete conditions ( Figure 10C).
In the literature, the concentration of 2-OG is reported to increase upon nitrogen deprivation (Laurent et al., 2005 (Huergo & Dixon, 2015). In Anabaenea sp. PCC 7120, the concentration of 2-OG increases to promote heterocyst differentiation (Li, Laurent, Konde, Bé du, & Zhang, 2003), upon N deprivation. However, heterocyst differentiation happens within the first 6-8 hr of N deprivation, by the generation of proto-heterocysts that turn into functional heterocysts within 24 hr (Harish & Seth, 2020;A. M. Muro-Pastor & Hess, 2012). It is thus reasonable to expect that the increase in the concentration of 2-OG which triggers heterocyst differentiation occurs mainly within the first few hours upon nitrogen deprivation.
Experimental work performed so far in Anabaenea sp. PCC 7120 focused mainly/only on the initial few hours (i.e., from few minutes to 2-4 hr) after nitrogen deprivation. This might explain why the concentration of 2-OG has been reported to increase upon nitrogen deprivation (Laurent et al., 2005;M. I. Muro-Pastor et al., 2001). In Laurent et al. (2005), the concentration of 2-OG peaks after 1-2 hr upon removal of combined nitrogen from the cultivation medium and after that the concentration returns to the initial value. Our work differentiates from the available literature as we intentionally designed a longer-term experiment with a first data point collected 24 hr after N deprivation. An increase in the concentration of 2-OG was therefore not observed as we skipped the initial differentiation phase of vegetative cells into heterocysts and the latter, in our first data point, are already fully differentiated ( Figure S4 in Data S1). The present study was primarily designed to provide novel information on how this species behaves at a whole-system level upon protracted nitrogen deprivation.
Taken together, our data suggests that the absence of AMT transporters results in a metabolic adjustment in response to environmental treatments (i.e., different N regimes), and although not investigated in the present study, this is likely to affect also the flux of metabolites between heterocysts and vegetative cells, as also suggested by the observed differences in CPG metabolism (Figure 8). The reduction of CPG in the mutant raises the following question: where is the excess N if we have not observed any altered growth phenotype nor a difference in the uptake of N from the external environment? Even if not measured in this work, we hypothesise that any changes in N availability most likely will affect the total protein pool, as it is the major sink for N in Anabaena (Simon, 1973). This hypothesis is supported by the general observation that the concentration of most protein targets in the mutant strain increased (e.g., increased accumulation of nitrogenase complex as reported in Figure 5B,C).

| Are AMT transporters an integral part of the regulatory/signalling network of N metabolism?
In our experiments, we repeatedly observed substantial changes in GS activity, as a consequence of the mutation. The mutant in fact displays an increased GS activity in NH 4 + replete and also under prolonged N deplete conditions ( Figures 2C and 6A), likely in response to changes in the abundance of both the GS protein itself and its posttranslational regulator IF7A (Figures 3 and 6). Moreover, we observed that GS activity is often retained even if IF7A abundance varies, potentially suggesting an additional molecular player(s) might also be involved in its regulation. Taken together, our results suggest AMT transporters may play a direct or indirect IF7A-independent role on GS activity in 7120, thus calling for further scientific efforts in order to fill potential gaps in the regulatory/signalling network of N metabolism.
The changes observed in this work are widespread at a whole cell level, also affecting the master molecular players which orchestrate N metabolism (i.e., NtcA, PII and PipX, Figure 7). It is worth noting that the concentration of the three master regulators do not correlate with the observed GS activity ( Figure 6A). For example, NtcA is expected to have multiple transcriptional targets (Picossi et al., 2014) which might also inversely control GS activity. Interestingly, the mutant displays an increased abundance of PII in N replete conditions, with respect to the parental strain ( Figure 7B). PII belongs to one of the most widely distributed families of signal transduction proteins in nature, involved in various aspects of N metabolism and regulation of C/N homeostasis (Arcondeguy et al., 2001;Forcada-Nadal, Llácer, Contreras, Marco-Marín, & Rubio, 2018;Forchhammer, 2004Forchhammer, , 2008. Among them, in heterotrophic bacteria and in archaea, PII proteins of the subfamily GlnK directly interact with AMT transporters to regulate their activity, typically reducing their uptake rate in N excess conditions to prevent intracellular over-accumulation of ammonium (Arcondeguy et al., 2001). Several studies suggest the PII protein binds AMT transporters also in cyanobacteria [i.e., Synechocystis sp. PCC 6803 (Watzer et al., 2019)] and in purple bacteria [e.g., R. capsulatus (Tremblay & Hallenbeck, 2008)]. In the latter organism, AMT proteins have been implicated in the posttranslational regulation of nitrogenase, driven by a direct interaction between AMTB and GlnK (a homolog of PII) (Tremblay & Hallenbeck, 2008). Our observation that there is greater PII abundance in the Δamt strain supports the hypothesis that PII might retain this function also in 7120. This is one likely explanation for why the disruption of AMT transporters influences the regulatory/signalling network of N metabolism in 7120.
Such a scenario would most likely involve a global mechanism by which 7120 filaments sense nitrogen sufficiency according to the external availability of fixed N. Studies in Anabaena variabilis ATCC 29413 suggest that filaments can differentiate between the availability of external and internal N, through a mechanism operating at whole-cellular system level (Thiel & Pratte, 2001). Do AMT transporters belong to such a global sensing mechanism in this species?
Further experimental work will be needed to clarify this hypothesis.
We were surprised to find that the molecular fingerprint of Δamt cells displays symptoms of N -deficiency relative to the parental strain, also under NH 4 + replete conditions, and that observations at both the protein and metabolite level support this hypothesis. Among them: (a) the faster activation of nitrogenase upon N deprivation ( Figure 5A), (b) the increased GS activity ( Figure 6A), (c) the strong reduction in the Gln/Glu ratio ( Figure S7 in Data S1), (d) the slightly reduced 2-OG content ( Figure 10C) and (e) the increased 2-OG/Gln ratio ( Figure S8 in Data S1). A possible explanation is that the mutant experiences a limitation in the free pool of internal ammonium/ia. The latter might explain why the free Gln pool is lower under NR conditions, as the lack of substrate is likely to lower flux through the GS reaction. This might also explain the observed increase in GS activity as a compensatory mechanism. In contrast, in the parental strain, the greater Gln concentration likely drives a Gln-feedback loop, possibly controlled by a homolog of the Gln-sensitive riboswitch recently described in Synechocystis sp. PCC 6803 . Further work is needed to confirm the existence of a similar molecular mechanism also in 7120.
In contrast to other indicators of N-deficiency, the greater photosynthetic activity ( Figure 4B) and Chl/Car ratio (Table 1)  A potential role for AMT transporters might be to prevent leakage of ammonium from the nitrogenase reaction under diazotrophic conditions. It should be noted that only a fraction of our experiment (i.e., the first 48 hr) was performed with measurable ammonium in the media, while the majority of the time points were collected in N deprivation ( Figure 4A). Therefore, in most of the experiment, the only available ammonium/ia in the culture is expected to be that lost from internally generated metabolism and diffusion of uncharged ammonia into the media.

| CONCLUSIONS
We investigated how a mutant of Anabaena sp. PCC 7120, missing the whole amt cluster, responds to environmental treatments affecting the inner N status of the cells. The whole cell system responds with substantial internal perturbations embracing both N and C metabolisms. Moreover, the absence of AMT transporters leaves a molecular fingerprint suggesting N-deficiency, which does not lead to any externally measurable phenotypic effect. We thus hypothesise that 7120 evolved a robust regulatory/signalling molecular network to maintain N metabolism homeostasis. The observed changes involve both proteins and metabolites, highlighting a pleiotropic effect of the mutation. We also provided evidence of perturbations to nitrogenase and GS activity, as well as to master regulators orchestrating N metab-

| Cyanobacteria strains and growth conditions
Strains of Anabaena sp. PCC 7120 used in this work are summarized in Table 2  Growth was monitored through OD 750 in 96-wells polystyrene plates with a multimode spectrophotometer (Tecan Infinite M200 Pro). Linear correlation between OD 750 and biomass dry weight was confirmed for both strains and the growth ranges measured in this work. Biomass dry weight was measured gravimetrically as previously reported in Perin et al. (2015). Specific growth rate was calculated by the slope of different growth phases for growth curves plotted in logarithmic scale. Cyanobacteria cultures were also monitored through microscopy analysis. A novel high-throughput heterocyst detection method has also been set up (see Supplementary Materials and Methods and Figure S1 in Data S1).

| Pigments content and photosynthetic efficiency
Pigments from intact cells grown in 6-wells plates were extracted using a 1:1 biomass to solvent ratio of 100% methanol, at 4 C in the dark for at least 20 min (Sinetova, Červený, Zavřel, & Nedbal, 2012).
Photosynthetic efficiency was assessed measuring in vivo chlorophyll fluorescence of intact cells using an AquaPen-C AP 110-C (PSI, Photon Systems Instruments, Czech Republic). Photosystem II (PSII) functionality was assessed as PSII maximum quantum yield (Φ PSII ), according to Maxwell and Johnson (2000)).

| Ammonium/ia quantification
Ammonium/ia quantification was performed using an adaptation of the method described by Willis, Montgomery, and Allen (1996) 5.4 | Enzymatic activities

Preparation of cell-free total proteins extracts
Strains of Anabaena sp. PCC 7120 grown in 6-wells plates were harvested by centrifugation and the supernatant discarded. Cell pellets were transferred to 2 ml polypropylene tubes and disrupted twice using −20 C. Cell extracts were collected through centrifugation for 10 s, 9000 g, 4 C. Cell extracts were centrifuged twice for 10 min, 21,000 g, 4 C to eliminate cell debris and insoluble proteins. Total proteins concentration in the cell-free lysate was assessed through DC protein assay (BIO RAD), following the manufacturer's manual, using 96-wells plates against a BSA (bovine serum albumin) standard curve.

| Nitrogenase
Nitrogenase activity was determined with an acetylene reduction assay under oxic culture conditions. Cells grown in 6-well plates were incubated in 2-ml gas chromatography glass vials under an atmosphere of 10% acetylene in air. Vials were incubated for 3 hr in the same original growth conditions (i.e., shaking, light, 30 C) and the quantity of ethylene in the headspace was determined by gas chromatography (7820A GC system, Agilent Technologies). The nitrogenase activity is expressed as % conversion of added acetylene into ethylene, per hour and per μg of Chl.

| Cyanophycin extraction
The same amount of biomass (OD 750 = 0.3) of different Anabaena sp. PCC 7120 strains grown in 6-wells plates was harvested by centrifugation and the supernatant discarded. Cyanophycin was extracted from the cell pellets following the protocol detailed in (Watzer et al., 2015), with some modifications as follows. Briefly, cell pellet was resuspended in 1 ml 100% acetone and incubated in a shaker for and samples incubated for 40 min, 4 C. Samples were then centrifuged for 17 min, 21,000 g, 4 C and pelleted cyanophycin polymers were resuspended in 500 μl 0.01 M HCl for quantification.

| Cyanophycin quantification
Cyanophycin is a polymer of arginine and aspartate (Forchhammer & Watzer, 2016) and the quantification of arginine released by cyanophycin granules can be used as proxy for cyanophycin content determination (Burnat et al., 2014). Arginine quantification was performed through a modified colorimetric Sakaguchi method, according to (Messineo, 1966 Typically 40 μl injections were used for the analysis. The MS was configured with an Ion Drive Turbo V source; Gases 1 and 2 were set to 40 and 60, respectively; the source temperature to 500 C and the ion spray voltage to 5,500 V. MS, configured with high mass enabled, was The identity of candidate peptides was then confirmed by EPI scans. Background proteome of Anabaena sp. PCC 7120 (http://genome. kazusa.or.jp/cyanobase) was used to check for uniqueness of target peptides. Typically, 3-5 transitions per peptide were used. The final method includes 1-4 peptides per protein for unique identification and quantification. Signature peptides for all the proteins investigated in this work are listed in Table 3.
Protein quantification was performed accounting for the intensities of all transitions peaks for all the peptides belonging to a specific protein. The resulted peak area was normalized to the peak intensity of the GluFib peptide standard.

| Sample preparation
Strains of Anabaena sp. PCC 7120 grown in 6-wells plates were harvested by fast filtration, modifying the protocol from (Eisenhut et al., 2008). Briefly, cells were fast filtered in the light without any subsequent washing step, using a vacuum filtration system (0.45 μm pore size nitrocellulose filter, 47 mm diameter [Sigma Aldrich]), using stainless-steel stand and funnel (Sartorius). Filters were then transferred to 50 ml tubes and immediately frozen in liquid nitrogen and stored at −80 C until metabolites extraction. Time between harvesting and metabolic inactivation by freezing was <10 sec. Deep frozen cells were scraped off the nitrocellulose filters using 80% cold methanol (−20 C). Cells in cold methanol were transferred to 2 ml polypropylene tubes and metabolite extraction was carried out with a TissueLyser II (Qiagen), using tube holders pre-cooled at −20 C and adding the same volume of acid-washed glass beads (Sigma-Aldrich), for 5 min at 30 Hz. Metabolites were collected after centrifugation for 10 min, 21,000 g, 4 C. Metabolite extraction was repeated twice, the extracts pooled together and centrifuged again to separate cell debris and other immiscible products. Metabolic extracts were then dried by vacuum centrifugation overnight and stored at −80 C until use.

| Sample derivatization and mass spectrometry
Dried metabolic extracts were reconstituted in 300 μl of water, using L-phenylalanine-d 5 as internal standard (final concentration 2 μg/ml), then vortexed and centrifuged at 16,000 g for 10 min. Quality control (QC) samples were prepared by mixing 10 μl of the supernatant of each sample, with the analytical batch including 10% QC samples.

Amino acids quantification
6-aminoquinolyl-N-hydroxysuccinimidyl carbamate was used for derivatization (AccQTag derivatization) according to the manufacturer's manual (AccQTag, Waters Corp). Briefly, 70 μl of borate buffer (pH 8.6) was added to 10 μl sample, followed by the addition of 20 μl AccQTag reagent (in acetonitrile). Samples were then vortexed and heated at 55 C for 10 min. 5 μl of each sample were analysed by HPLC-electrospray ionisation/MS-MS using a Shimadzu UFLC XR/AB SCIEX Triple Quad 5,500 system, running in multiple reaction monitoring (MRM) via positive ionization mode.
The LC-MS method here exploited is based on the one previously described by Gray et al. (2017)), with some modifications, as detailed below.  Table S3 in Data S1.

Data processing
The raw LC-MS data were analysed using Skyline [MacCoss Lab (K. J. Adams et al., 2020)]. External dilution curves were used to determine the range for linear response.

Chemicals and reagents
Metabolite standards (

| Statistical analysis
Descriptive statistical analysis was applied for all the data presented in this work. The mathematical correlation between GS activity and IF7A abundance was assessed through Pearson correlation in Microsoft Excel. Statistical significance was assessed by one-way analysis of variance (One-way ANOVA) and Student's t test using OriginPro 2018b (v. 9.55) (http://www.originlab.com/). Samples size was at least >4 for all the measurements collected in this work.