S. elongatus P_II (Figure 2), as other P_II proteins, is a homotrimer of a polypeptide chain of 112 amino acids that exhibits the ferredoxin fold (βαβ)₂ followed by a beta hairpin (Xu et al., 2003). The trimer (Figure 2A) has a hemispheric body nucleated by three antiparallel oblique (relative to the three-fold axis) β-sheets, each one formed by the 4-stranded sheet (topology ↓β2↑β3↓β1↑β4) of a subunit (see for example subunit B in the central panel of Figure 2A) extended on its β4 end by the C-terminal hairpin (β5-β6) of an adjacent subunit (subunit A) and on the β2 end by the β2-β3 hairpin stem (the root of the T-loop, see below) of the other subunit of the trimer (subunit C). The three sheets become continuous on the flat face of the hemispheric body via their β2-β3 hairpins (Figure 2A). The subunit sheets encircle like a 3-sided pyramid the three-fold axis, filling the inner space between them with their side-chains. They are covered externally by 6 helices (two per subunit) that run parallel to the β strands, contributing to the rounded shape of the hemispheric trimer (Figure 2A, panel to the right) and to the outer part of its equatorial flat face. In the convex face, three crevices are formed at subunits junctions between adjacent β-sheets, over the β2-β3 hairpins (Figure 2B). These crevices host the sites for the allosteric effectors ATP/ADP [and in some species AMP (Palanca et al., 2014)] and 2-oxoglutarate (2OG) that endow P_II with its sensing roles (Kamberov et al., 1995; Zeth et al., 2014) (Figure 2B), the nucleotides reflecting the energy status (Fokina et al., 2011) and 2OG reflecting the abundance of carbon and, inversely, the nitrogen richness (see for example Muro-Pastor et al., 2001).

Very salient structural features of P_II are the long flexible T-loops (Figures 2A,C) formed by the 18 residues that tip the β2-β3 hairpin of each subunit (Xu et al., 2003). These loops are key elements (although not the exclusive ones, see Rajendran et al., 2011 and Schumacher et al., 2015) for P_II interaction with its targets (Conroy et al., 2007; Gruswitz et al., 2007; Llácer et al., 2007, 2010; Mizuno et al., 2007; Zhao et al., 2010b; Chellamuthu et al., 2014). By binding at the boundary between the T-loop and the P_II body, at the crevice formed between adjacent subunits, the adenine nucleotides and MgATP/2OG promote the adoption by the T-loop of different conformations (Figure 2C) (Fokina et al., 2010a; Truan et al., 2010; Maier et al., 2011; Zeth et al., 2014) that favor or disfavor P_II binding to a given P_II target.

The T-loop also is the target of regulatory post-translational modification (reviewed in Merrick, 2015), first recognized in the regulatory cascade of the GS of E. coli as uridylylation of Tyr51 (see Figure 2C, 3rd panel from the left) mediated by a glutamine-regulated bifunctional P_II uridylylating-deuridylylating enzyme, GlnD (Stadtman, 2001). Thus, in the enterobacterial GS regulating cascade P_II is uridylylated or deuridylylated depending on whether 2-OG is abundant and L-glutamine is low or the reverse. P_II-UMP activates the GS deadenylylating activity of ATase (Jiang et al., 2007), activating GS by decreasing its susceptibility to feed-back inhibition (Stadtman, 2001). This uridylylation (or in Actinobacteria adenylylation of Tyr51) occurs at least in proteobacteria and actinobacteria (Merrick, 2015), but it might be more widespread, since it has also been reported in an archaeon (Pedro-Roig et al., 2013). Structural studies with E. coli P_II (Palanca and Rubio, 2017) have excluded the stabilization of the T-loop into a fixed conformation by Tyr51 uridylylation, suggesting that the Tyr51-bound UMP physically interacts with the ATase. Although Tyr51 is conserved in cyanobacteria, it is not uridylylated. The T-loop serine 49 (Figure 2C, 1st, 2nd and 4th panels from the left) is phosphorylated in S. elongatus under conditions of nitrogen starvation by an unknown mechanism (Forchhammer and Tandeau de Marsac, 1994), whereas the phosphatase that dephosphorylates phosphoSer49 has been identified and proven to be 2OG-sensitive (Irmler et al., 1997).

FRET studies with engineered fluorescent S.elongatus P_II used as an ADP and ATP-sensitive probe (Lüddecke and Forchhammer, 2015) have challenged the claim (Radchenko et al., 2013) that P_II proteins have a very slow ATPase activity that would regulate P_II similarly as the signaling GTPases with bound GTP and GDP. Although this ATPase was reported as a 2OG-triggered switch that appeared an intrinsic trait of P_II proteins (Radchenko et al., 2013), the FRET experiments with S. elongatus P_II (Lüddecke and Forchhammer, 2015) appear to indicate that an endogenous ATPase is not a relevant mechanism for the transition of P_II into the ADP state.

The first structurally solved P_II complex was the one of E. coli GlnK with the AmtB ammonia channel (Conroy et al., 2007; Gruswitz et al., 2007) (Figure 3A) formed under nitrogen richness conditions. This structure showed that AmtB was inhibited by GlnK because the extended T-loop fits the channel entry, with the insertion into the channel of a totally extended arginine emerging from the T-loop and blocking the channel space (Figure 2C, 3rd panel from the left, and Figure 3A, zoom). The ADP-bound and MgATP/2OG-bound structures of an Archeoglobus fulgidus GlnK protein (Maier et al., 2011) indicated that 2OG may prevent GlnK binding to the ammonia channel because of induced flexing outwards (relative to the 3-fold molecular axis) of the T-loops, preventing their topographical correspondence with the three holes of the trimer of ammonia channels (Figure 3B). Interestingly, the T-loops of MgATP/2OG-bound S. elongatus P_II (Figure 2C, leftmost panel) and A. fulgidus GlnK3 (Figure 3B) exhibited different flexed conformations (relative to the ADP-bound extended forms), and thus 2OG-binding by itself does not determine a single T-loop conformation, at least with different P_II proteins.

Yeast two hybrid approaches (Osanai et al., 2005) detected the interaction between P_II and the putative channel PamA (encoded by sll0985) of Synechocystis sp. PCC 6803 (from now on Synechocystis), but molecular detail on this protein is non-existent, and, therefore, it is uncertain whether such interaction might resemble the GlnK-AmtB interaction. PamA is not conserved in many cyanobacteria, and the most closely related putative protein of S. elongatus, the product of the Synpcc7942_0610 gene, failed to give interaction signal with S. elongatus P_II in yeast two hybrid assays (Castells, M.A., PhD Dissertation, Universidad de Alicante, 2010), despite the fact that the sequence identity with PamA concentrated in the C-terminal region, where P_II binds in Synechocystis (Osanai et al., 2005). In vitro studies with the recombinantly produced Synechococystis PamA and P_II showed that their interaction was lost in the presence of ATP and 2OG. Thus, similarly to the GlnK-AmtB and GlnK3-Amt complexes, the P_II-PamA complex is formed under conditions of nitrogen abundance. However, T-loop phosphorylation did not dissociate this complex (Osanai et al., 2005). The function of PamA is not known, but its deletion from Synechocystis changed the expression of some NtcA-dependent genes (Osanai et al., 2005) by unclarified mechanisms.

The complexes of P_II with the N-acetyl-L-glutamate kinase (NAGK) enzymes from S. elongatus and Arabidopsis thaliana presented a very different architecture with respect to the structure of the GlnK-AmtB complex of E. coli (Llácer et al., 2007; Mizuno et al., 2007) (Figures 4A,B). The P_II-NAGK complex is an activating complex in which the T-loops of P_II are flexed (Figure 2C, rightmost panel) and integrated into a hybrid (both proteins involved) β-sheet with NAGK, forming also a hybrid ion-pair network (Figure 4C; Llácer et al., 2007). Apparently this flexing from an extended conformation could occur in two steps (Fokina et al., 2010b). The initial step would be mediated by a smaller loop of P_II called the B-loop (Figures 2A, 4C). P_II binding of 2OG also favors the flexing of the T-loop (Figure 2C, leftmost panel) (Fokina et al., 2010a; Truan et al., 2010) although the resulting conformation appears inappropriate for interacting with NAGK. In addition, 2OG can also promote the disassembly of the P_II-NAGK complex because certain P_II residues like Arg9 that are involved in the binding of 2OG are also involved in the interaction with NAGK (and also with PipX, another target of P_II, see below). Therefore, 2OG, an indicator of low ammonia levels (Muro-Pastor et al., 2001), abolishes P_II-NAGK complex formation (Maheswaran et al., 2004) (Figure 1). In S. elongatus 2OG can also promote the disassembly of the P_II-NAGK complex by favoring the phosphorylation of Ser49 (Forchhammer and Tandeau de Marsac, 1994; Irmler et al., 1997), since the bound phosphate sterically prevents formation of the P_II-NAGK hybrid β-sheet (Llácer et al., 2007).

Plants and cyanobacteria stockpile ammonia as arginine, the protein amino acid with the largest nitrogen content (four atoms per arginine molecule). Arginine-rich proteins are very abundant in plant seeds (VanEtten et al., 1963). Cyanobacteria make non-ribosomally an arginine-rich amino acid polymer called cyanophycin (Oppermann-Sanio and Steinbüchel, 2002; Watzer and Forchhammer, 2018). The arginine stockpiling as arginine-rich macromolecules minimizes the osmotic effect while permitting rapid nitrogen mobilization for protein-building processes such as seed germination and cell multiplication. The selection of NAGK as the regulatory target stems from the fact that in many bacteria (including cyanobacteria) and in plants NAGK controls arginine synthesis via feed-back inhibition by L-arginine (Hoare and Hoare, 1966; Cunin et al., 1986; Lohmeier-Vogel et al., 2005; Beez et al., 2009). This inhibition must be overcome if large amounts of ammonia have to be stored as arginine (Llácer et al., 2008). Indeed, the P_II-NAGK complex exhibits decreased inhibition by arginine (Maheswaran et al., 2004; Llácer et al., 2008).

In arginine-sensitive NAGK (Figures 4D,E) the N-terminal α-helix of each subunit interacts with the same helix of an adjacent dimer, chaining three NAGK homodimers into a doughnut-shaped hexameric ring with three-fold symmetry and a central large hole (Ramón-Maiques et al., 2006). The NAGK reaction (phosphorylation of the γ-COOH of N-acetyl-L-glutamate, NAG, by ATP) occurs within each NAGK subunit. NAG and ATP sit over the C-edge of the central 8-stranded largely parallel β sheet of the N-terminal and C-terminal domains, respectively (Figure 4D; Ramón-Maiques et al., 2002). Catalysis requires the mutual approach of both domains of each subunit to allow the contact of the ATP terminal phosphate with the attacking NAG γ-COOH (Ramón-Maiques et al., 2002; Gil-Ortiz et al., 2003). Arginine, by binding in each subunit next to the N-terminal α-helix (Figure 4E), expands the hexameric ring hampering the contact of the reacting groups and preventing catalysis (Ramón-Maiques et al., 2006).

In the P_II-NAGK complex two P_II trimers sit on the three-fold axis of the complex, one on each side of the NAGK ring, making contacts with the inner circumference of this ring (Figures 4A,B). Each P_II subunit interacts via its T and B loops with each NAGK subunit (Figure 4C) gluing the two domains of this last subunit (Figure 4A). By restricting NAGK ring expansion (Figure 4C) even when arginine is bound, P_II renders NAGK highly active (Llácer et al., 2007; Mizuno et al., 2007). P_II does not compete physically with arginine for its sites on NAGK, simply these sites are widened in the P_II-NAGK complex (Llácer et al., 2007), resulting in decreased apparent affinity of NAGK for arginine (as reflected in the dependency of the NAGK activity on the arginine concentration). In addition, the hybrid P_II-NAGK ion pair network (Figure 4C) enhances the apparent affinity for NAG (assessed as the K_m or S_0.5 value of NAGK for NAG) of cyanobacterial NAGK (Maheswaran et al., 2004; Llácer et al., 2007). Overall, the NAGK bound to P_II exhibits decreased apparent affinity for arginine and increased activity, rendering NAGK much more active in the presence of arginine than when not bound to P_II (Llácer et al., 2008), something that is crucial for nitrogen storage as arginine.

NAGK appears to be a P_II target only in organisms performing oxygenic photosynthesis (cyanobacteria, algae, and plants, Burillo et al., 2004). P_II proteins from plants have lost the ability to bind ADP, while still binding ATP and 2OG (Lapina et al., 2018). In addition, except in Brassicae, the C-terminal part of plant P_II is extended to form two helical segments and a connecting loop (Q-loop; Figure 2B, rightmost panels), creating a novel glutamine site, resulting in glutamine-sensitivity of the P_II-NAGK interaction (Chellamuthu et al., 2014). This is not the case with cyanobacterial P_II, which binds both ADP and ATP and is glutamine-insensitive (Chellamuthu et al., 2014).

P_II has also been shown to interact in plants (Feria-Bourrellier et al., 2010) and bacteria, including cyanobacteria (Rodrigues et al., 2014; Gerhardt et al., 2015; Hauf et al., 2016), with the biotin carboxyl carrier protein (BCCP) of the enzyme acetyl coenzyme A carboxylase (AcCoA carboxylase) (Figure 1), although this complex has not been characterized structurally. BCCP is the component that hosts the covalently bound biotin that shuttles between the biotin carboxylase component and the transferase component of AcCoA carboxylase (Rubio, 1986). P_II-BCCP complex formation tunes down AcCoA utilization and thus subsequent fatty acid metabolism (Feria-Bourrellier et al., 2010; Gerhardt et al., 2015; Hauf et al., 2016), promoting uses of AcCoA for different purposes than the synthesis of fatty acids, and therefore linking P_II to AcCoA and fatty acid metabolism. For interaction, P_II has to be in the ATP-bound and 2OG-free form (Figure 1) (Gerhardt et al., 2015; Hauf et al., 2016), which are conditions at which P_II also binds to NAGK (Llácer et al., 2008). Therefore, there could be in vivo simultaneous activation of NAGK and inhibition of AcCoA carboxylase by P_II. Mutational evidence suggests the involvement of the T-loop in this interaction with AcCoA carboxylase (Hauf et al., 2016), in principle excluding P_II-NAGK-BCCP ternary complex formation and raising the possibility of competition between NAGK and BCCP for P_II.

The classical example of interaction of P_II with an enzymatic target was with the ATase of E. coli (see introductory section), with which uridylylated or deuridylylated GlnB (GlnB is one of the two P_II proteins of E. coli) can interact (Stadtman, 2001). We will not deal with this enzyme here because the P_II/ATase/GS cascade of enterobacteria does not appear to have general occurrence, for example in cyanobacteria, and also because we have only partial information on the structure of the ATase (Xu et al., 2004, 2010) and no direct information on the structure of the GlnB-ATase complex, although a model for such complex has been proposed (Palanca and Rubio, 2017).

A yeast two hybrid search for proteins interacting with P_II in S. elongatus identified (Burillo et al., 2004), in addition to NAGK, a small novel protein (89 amino acids) that was named PipX (P_II-interacting protein X). This protein was identified later in a search (Espinosa et al., 2006) for proteins interacting with NtcA, the global nitrogen regulator of cyanobacteria (Vega-Palas et al., 1992). PipX binding to P_II occurs under conditions of ammonia abundance (Figure 1), the same conditions prevailing for P_II-NAGK complex formation (Espinosa et al., 2006). NAGK-PipX competition for P_II was revealed in NAGK assays that showed that PipX decreased P_II-activation and increased arginine inhibition of NAGK (Llácer et al., 2010), excluding NAGK-P_II-PipX ternary complex formation. 2OG binding to P_II disassembles the P_II-PipX complex (Espinosa et al., 2006; Llácer et al., 2010), leaving PipX free to interact with NtcA (Figure 1).

The crystal structures of P_II-PipX complexes of S. elongatus (Llácer et al., 2010) and of Anabaena sp. PCC7120 (Zhao et al., 2010b) provided the first structural information on PipX (Figure 5A), revealing that it is formed by a compact body folded as a Tudor-like domain (a horseshoe-curved β-sheet sandwich) (Lu and Wang, 2013), followed by two C-terminal helices. In the P_II-PipX complex (Figure 5B) the three PipX molecules are enclosed in a cage formed between the flat face of the hemispheric P_II trimer and its three fully extended T-loops (see Figure 2C, 2nd panel from the left) emerging perpendicularly to the P_II flat face at its edge. The shape and orientation of these T-loops is very different relative to the P_II bound to NAGK, Figure 5C). In turn, the caged Tudor-like domains form a homotrimer over the P_II flat surface (Figure 5B, bottom), with the PipX self-interaction detected in yeast three-hybrid assays using P_II as bridging protein (Llácer et al., 2010). Tudor-like domains characteristically interact with RNA polymerase (Steiner et al., 2002; Deaconescu et al., 2006; Shaw et al., 2008), suggesting that PipX could have some role in gene expression that would be blunted by sequestration of these domains in the P_II cage.

In the structure of the Anabaena P_II-PipX complex, the two C-terminal helices of each PipX molecule lie one along the other in antiparallel orientation (“flexed”), being exposed between two adjacent T-loops in transversal orientation relative to these loops (Zhao et al., 2010b). Recent structural NMR data on isolated PipX showed that when PipX is alone (that is, not bound to a partner) the C-terminal helices are “flexed” (Figure 5A) (Forcada-Nadal et al., 2017). As shown below, the C-terminal helices of PipX in the NtcA-PipX complex are also flexed (Llácer et al., 2010). However, in the S. elongatus P_II-PipX complex only one PipX molecule presents the “flexed” conformation, whereas in the other two PipX molecules the C-terminal helix is “extended,” not contacting the previous helix and emerging centrifugally outwards from the complex, between two T-loops (Figure 5B) (Llácer et al., 2010). P_II binding might facilitate the extension of the PipX C-terminal helix, endowing the P_II-PipX complex with a novel surface and novel potentialities for interaction with other components. These novel potentialities were substantiated recently by the identification, in yeast three-hybrid searches (Labella et al., 2016), of interactions of PipX in the P_II-PipX complex with the homodimeric transcription factor PlmA (see proposal for the architecture of this complex in Figure 1 bottom left; Labella et al., 2016). Interactions were not observed in yeast two-hybrid assays between PipX or P_II and PlmA. Residues involved in three-hybrid interactions, mapped by site-directed mutagenesis, are largely localized in the C-terminal helix of PipX. PlmA belongs to the GntR super-family of transcriptional regulators, but is unique to cyanobacteria (Lee et al., 2003; Hoskisson and Rigali, 2009; Labella et al., 2016). Little is known about PlmA functions other that it is involved in plasmid maintenance in Anabaena sp. strain PCC7120 (Lee et al., 2003), in photosystem stoichiometry in Synechocystis sp. PCC6803 (Fujimori et al., 2005), in regulation of the highly conserved cyanobacterial sRNA YFR2 in marine picocyanobacteria (Lambrecht et al., 2018), and that it is reduced by thioredoxin, without altering its dimeric nature in Synechocystis sp. PCC6803 (Kujirai et al., 2018). The P_II-PipX-PlmA ternary complex suggests that PipX can influence gene expression regulation via PlmA, although the PlmA regulon remains to be defined.

When ammonia becomes scarce the increasing 2OG levels should determine the disassembly of the P_II-NAGK, P_II-BCCP, and P_II-PipX complexes (Figure 1). These same conditions promote the binding of PipX (see below) to the transcriptional regulator NtcA (Figure 1), an exclusive cyanobacterial factor of universal presence in this phylogenetic group (Vega-Palas et al., 1992; Herrero et al., 2001; Körner et al., 2003). The determination of the structures of NtcA from S. elongatus (Figures 6A,B) (Llácer et al., 2010) and from Anabaena sp. PCC7120 (Zhao et al., 2010a) confirmed the sequence-based inference (Vega-Palas et al., 1992) that NtcA is a homodimeric transcriptional regulators of the family of CRP (the cAMP-regulated transcriptional regulator of E. coli) (McKay and Steitz, 1981; Weber and Steitz, 1987). Similarly to CRP, NtcA has a C-terminal DNA binding domain of the helix-turn-helix type. In CRP, the DNA binding helices of its two C-terminal domains are inserted in two adjacent turns of the major groove of DNA that host the imperfectly palindromic target DNA sequence (called here the CRP box) (McKay and Steitz, 1981; Weber and Steitz, 1987). The consensus DNA sequence to which NtcA binds (consensus NtcA box) is quite similar to the consensus CRP box (Berg and von Hippel, 1988; Luque et al., 1994; Jiang et al., 2000; Herrero et al., 2001; Omagari et al., 2008), and thus NtcA and CRP are expected to bind in similar ways to their target DNA sequences.

In vitro studies revealed that 2OG is an NtcA activator (Tanigawa et al., 2002; Vázquez-Bermúdez et al., 2002), increasing NtcA affinity for its target sequences. As in the case of cAMP for CRP, 2OG binds to the NtcA regulatory domain. This domain is responsible for the NtcA dimeric nature (Figure 6A) (Llácer et al., 2010; Zhao et al., 2010a). The regulatory domain of NtcA is highly similar to the corresponding domain of CRP (Llácer et al., 2010). The main differences reflect the changes in the characteristics of the site for the allosteric effector that enable the accommodation of 2OG instead of cAMP. Each 2OG molecule interacts in NtcA with the two (one per subunit) long interfacial helices that form the molecular backbone, crossing the molecule in its longer dimension, linking in each subunit both domains (Figures 6A,C) (Llácer et al., 2010; Zhao et al., 2010a). 2OG interactions with both interfacial helices favor a twist of one helix relative to the other, dragging the DNA binding domains and helices to apparently appropriate positions and interhelical distance for binding in two adjacent turns of the major groove of DNA where the NtcA box should be found (Figures 6A,B, and inset therein), although the experimental structure of DNA-bound NtcA should be determined to corroborate this proposals.

Although NtcA and CRP boxes are quite similar, plasmon resonance experiments (Forcada-Nadal et al., 2014) revealed that CRP exhibits complete selectivity and specificity for the CRP box, with absolute dependency on the presence of cAMP. In contrast, NtcA had less strict selectivity, since it still could bind to its promoters in the absence of 2OG, although with reduced affinity, and it could also bind to the the CRP promoter tested. Nevertheless, it is unlikely that NtcA could bind in vivo to the CRP boxes of cyanobacteria in those species where CRP is also present, given the much higher affinities for the CRP sites of cyanobacterial CRP and the relative concentrations in the cell of both transcriptional regulators (Forcada-Nadal et al., 2014).

While the structures of 2OG-bound NtcA of S. elongatus (Llácer et al., 2010) and of Anabaena (Zhao et al., 2010a) are virtually identical, the reported structures of “inactive” NtcA of Anabaena without 2OG (Zhao et al., 2010a) and of S. elongatus (Llácer et al., 2010) differed quite importantly in the positioning of the DNA binding domains (Figure 1, “Inactive forms” under “NtcA”), although in both cases the DNA binding helices were misplaced for properly accommodating the NtcA box of DNA, raising the question of whether these structural differences are species-specific or whether “inactive” NtcA can be in a multiplicity of conformations.

Soon after PipX was found to interact with NtcA (Espinosa et al., 2006), it was also shown to activate in vivo transcription of NtcA-dependent promoters under conditions of low nitrogen availability (Espinosa et al., 2006, 2007). Direct binding studies with the isolated molecules proved that PipX binding to NtcA requires 2OG (Espinosa et al., 2006). Nevertheless, as PipX was not totally essential for transcription of NtcA-dependent promoters (Espinosa et al., 2007; Camargo et al., 2014), it was concluded (1) that PipX was a coactivator of 2OG-activated NtcA-mediated transcription (Espinosa et al., 2006; Llácer et al., 2010); and (2) that the degree of activation by PipX depended on the specific NtcA-dependent promoter (Espinosa et al., 2007; Forcada-Nadal et al., 2014). Detailed plasmon resonance studies (Forcada-Nadal et al., 2014) using sensorchip-bound DNA confirmed for three Synechocystis promoters that PipX binding to promoter-bound NtcA has an absolute requirement for 2OG, since no PipX binding was observed when NtcA was bound to the DNA in absence of 2OG. In these studies PipX increased about one order of magnitude the apparent affinity of NtcA for 2OG. In other in vitro experiments with four NtcA-dependent promoters of Anabaena sp. PCC 7120, PipX was also found to positively affect NtcA binding to its DNA sites (Camargo et al., 2014). The induction by PipX of increased NtcA affinity for 2OG and for its promoters could account for the PipX-triggered enhancement of NtcA-dependent transcription.

The crystal structure of the NtcA-PipX complex of S. elongatus (Figure 7) (Llácer et al., 2010) corresponded to one “active” NtcA dimer with one molecule of each 2OG and PipX bound to each subunit. PipX is inserted via its Tudor-like domain (Figure 7A), filling a crater-like cavity formed over each NtcA subunit (Figure 7B) largely over one regulatory domain, being limited between the DNA binding domain and the long interfacial helix of the same subunit, and the regulatory domain of the other subunit. PipX extensively interacts with the entire crater, with nearly 1200 Ų of NtcA surface covered by each PipX molecule, of which 65% belongs to one subunit (40%, 15% and 10% belonging to the DNA-binding domain, the interfacial helix and the regulatory domain, respectively) and 35% belongs to the regulatory domain (including the interfacial helix) of the other subunit, gluing together the elements of half of the NtcA dimer in its active conformation, stabilizing this conformation (Llácer et al., 2010). This conformation is the one that binds 2OG and that should have high affinity for the DNA, thus explaining the requirement of 2OG for PipX binding and the increased affinities of NtcA for 2OG and DNA when PipX was bound to NtcA (Forcada-Nadal et al., 2014). Since a similar crater-like cavity exists in other transcription factors of the CRP family including CRP (see for example McKay and Steitz, 1981 or Weber and Steitz, 1987) it would be conceivable that PipX-mimicking proteins could exist for these other transcriptional regulators of the CRP family, although PipX cannot do such a role since it does not bind to CRP (Forcada-Nadal et al., 2014). Furthermore, a large set of highly specific contacts (Llácer et al., 2010) ensure the specificity of the binding of PipX to NtcA.

The elements of the Tudor-like domain that interact with NtcA are largely the same that interact with the flat surface of the hemispheric body of P_II (many of them mediated by the upper layer of the Tudor-like β-sandwich, particularly strands β1 and β2), predicting total incompatibility for the simultaneous involvement of PipX in the NtcA and P_II complexes (Llácer et al., 2010). While the Tudor-like domain monopolizes the contacts of PipX with NtcA, the C-terminal helices of PipX do not participate in these contacts and remain flexed, as in isolated PipX (Forcada-Nadal et al., 2017), protruding away from the complex (Figure 7A). The coactivation functions of PipX for NtcA-mediated transcription could also involve these helices. However, in vitro experiments (Llácer et al., 2010; Camargo et al., 2014) and modeling (based on CRP) of DNA binding by the NtcA-PipX complex (Figure 7A and inset therein) (Llácer et al., 2010) did not support the idea of PipX binding to DNA. Alternatively, these helices could interact with RNA polymerase, particularly given the location, in the homologous CRP-DNA complex, of the binding site for the α-subunit of the C-terminal domain (αCTD) of RNA polymerase (Figure 7C; and discussed in Llácer et al., 2010). Further structures of P_II-PipX bound to DNA alone or with at least some elements of the polymerase are needed to clarify these issues.

The gene encoding P_II could not be deleted in S. elongatus unless the pipX gene was previously inactivated (Espinosa et al., 2009). Further studies led to the conclusion that decreasing the P_II/PipX ratio results in lethality in S. elongatus, indicating that PipX sequestration into P_II-PipX complexes is crucial for survival and implicating both proteins in the regulation of essential processes (Espinosa et al., 2010, 2018; Laichoubi et al., 2012). The ability of P_II to prevent the toxicity of PipX suggests that P_II acts as a PipX sink even under conditions in which the affinity for NtcA would be highest, supporting the idea that not all PipX effects are related to its role as NtcA co-activator. Mutational studies (Espinosa et al., 2009, 2010) and massive transcriptomic studies of S. elongatus mutants centered on PipX (Espinosa et al., 2014) also support the multifunctionality of PipX, stressing the need for additional studies, including the determination of PlmA functions and the search for further potential PipX-interacting proteins (Figure 1, broken yellow arrow).

Massive proteomic studies (Guerreiro et al., 2014) have estimated the number of chains of each protein of the P_II-PipX network (Figure 1) in S. elongatus cells. The values obtained (Table 1) are corroborated by those obtained in focused western blot studies for some of these macromolecules (Table 1) (Labella et al., 2016, 2017). These quantitative data give an opportunity to evaluate the possible frequency of the different complexes and macromolecules of the P_II-PipX-NtcA network (Figure 1) in one or another form (schematized in Figure 8). Of all the proteins mentioned here until now, P_II is by far the most abundant in terms of polypeptide chains (Table 1). In comparison, the sum of all the chains of other known P_II-binding proteins represents no more than 20% of the P_II chains. Among these molecules is PipX, which only represents ~10% of all the P_II chains. This indicates that P_II has the potential to sequester all the PipX that is present in the cell (Figure 8). In turn, PlmA could be fully trapped in the P_II-PipX-PlmA complex if this complex has the 1:1:1 stoichiometry proposed for it (Figure 1) (Labella et al., 2016), since the number of PlmA chains only represent about 10% of the number of PipX chains. Thus, about 10 and 1% of the P_II trimer could be as the respective P_II-PipX and P_II-PipX-PlmA complexes under nitrogen abundance conditions. In contrast, with nitrogen starvation all of the NtcA could be bound to PipX (Figure 8), given the ~five-fold excess of PipX chains over NtcA chains (Table 1). Thus, assuming that under conditions at which 2OG and ATP reach high levels P_II is totally unable to bind PipX, ~80% of the PipX molecules could be free to interact with additional protein partners.

These inferences are consistent with the K_D values for the P_II-NAGK (Llácer et al., 2007) and P_II-PipX complexes in the absence of 2OG (Llácer et al., 2010) and for the PipX-NtcA complex at high 2OG and ATP (Forcada-Nadal et al., 2014) (~0.08, 7 and 0.09 μM, respectively). For the estimated cellular levels of the different components (Table 1), assuming a cell volume of 10⁻¹² ml, virtually all the NAGK and ~95% of the PipX could be P_II-bound in the absence of 2OG, and ~98% of the NtcA could be PipX-complexed in the presence of 2OG. However, the impacts of varying concentrations of 2OG on the disassembly or assembly of the complexes most likely differ for the various complexes. For example, a two-order of magnitude increase in the K_D value for the P_II-NAGK complex due to 2OG binding might have much less impact (a 7% decrease in the amount of NAGK bound to P_II would be estimated from the mere total protein levels and K_D value) than a two-order of magnitude increase in the K_D for the P_II-PipX complex (an 80% decrease in the amount of PipX bound estimated similarly). These estimations are very crude, since they do not take in consideration that in S. elongatus P_II phosphorylation prevents NAGK binding (Heinrich et al., 2004), and that this phosphorylation is greatly influenced by the abundance of ammonia (Forchhammer and Tandeau de Marsac, 1994). Furthermore, we have not considered in these estimates the influence of the ATP concentrations, recently shown to decrease in vivo in S. elongatus upon nitrogen starvation (Doello et al., 2018). Therefore, the situation is much more complex than would be expected from the mere consideration of the abundances of the different proteins and of the K_D values for the non-phosphorylated form of P_II. Nevertheless, it appears desirable to estimate the influence of different 2OG levels on K_D values as an important element to take into consideration in future attempts to model the concentrations of the different complexes of P_II and PipX under intermediate conditions of nitrogen richness.

Recently, a novel protein has been identified as belonging to the PipX regulatory network (Figure 1, yellow 3D arrow projecting upwards from the plane of the network). In this case no direct protein-protein interaction with PipX has been shown (Labella et al., 2017). This protein (PipY), is the product of the downstream gene in the bicistronic pipXY operon. The regulatory influence of PipX on PipY was originally detected in functional, gene expression and mutational studies (Labella et al., 2017). More recently it has been concluded (Cantos et al., 2018) that PipX enhances pipY expression in cis, preventing operon polarity, a function that might implicate additional interactions of PipX with the transcription and translation machineries, by analogy with the action of NusG paralogues, which are proteins bearing, as PipX, Tudor-like domains. It has been proposed (Cantos et al., 2018) that the cis-acting function of PipX might be a sophisticated strategy for keeping the appropriate PipX-PipY stoichiometry.

PipY is an intriguing pyridoxal phosphate-containing protein that is folded as a modified TIM-barrel (Figure 1). The PipY structure (Tremiño et al., 2017) gives full structural backing to unsuccessful experimental attempts to show enzymatic activity of PipY and its orthologs (Ito et al., 2013). Because of these negative findings, and given the pleotropic effects of the inactivation of the PipY orthologs in microorganisms and humans, it has been concluded that these proteins have as yet unclarified roles in PLP homeostasis (Ito et al., 2013; Darin et al., 2016; Prunetti et al., 2016; Labella et al., 2017; Plecko et al., 2017). Interestingly, these proteins are widespread across phyla (Prunetti et al., 2016) and the deficiency of the human ortholog causes vitamin B₆-dependent epilepsy (Darin et al., 2016; Plecko et al., 2017; Tremiño et al., 2018), providing an excellent example that investigations of cyanobacterial regulatory systems like the one summarized here can have far-reaching consequences spanning up to the realm of human and animal pathology. If any lesson can be inferred from all of the above, is that the investigation on P_II and particularly on PipX proteins require further efforts.

The rich P_II regulatory network summarized in Figure 1 of this review even for a unicellular microorganism with a single type of P_II protein, attests the importance of P_II and of its regulatory processes. This importance, possibly underrecognized until now, is highlighted, for example, by the very wide distribution of P_II proteins among microorganisms and plants. Furthermore, in the many organisms with several genes for P_II proteins the levels of complexity expected from P_II regulatory networks may be much greater than the one presented here. Each one of these paralogous P_II proteins may command a regulatory network, and it would be unlikely that these networks would not be interconnected into a large meshwork that will require the instruments of systems biology to be fully understood.

PipX also deserves deeper attention than received until now. Massive transcriptomics studies (Espinosa et al., 2014) have ascribed to this protein a paramount regulatory role in S. elongatus. For full understanding of this role further searches for PipX-interacting or functional partners like PipY appear desirable, with detailed investigation of the molecular mechanisms of the physical or the functional interactions. PlmA merits particular attention to try to characterize the roles of the P_II-PipX-PlmA complex. PipY and its orthologs deserve similar attention, to try to define molecularly their PLP homeostatic functions, a need that is made more urgent by the role in pathology of the human ortholog of PipY. In addition to all of this, the structural evidence reviewed here makes conceivable that adaptor proteins capable of stabilizing active conformations of other transcriptional regulators of the CRP family could exist outside cyanobacteria, mimicking the PipX role. In summary, there are many important questions to be addressed arising from the field reviewed here, some within cyanobacteria, but others concerning whether the mechanisms and complexes exemplified here could have a parallel in other bacterial or even plant species. Clearly more investigations on P_II and its partners in other phylogenetic groups using the approaches and experimental instruments used to uncover the cyanobacterial P_II regulatory network would appear highly desirable.

AF-N, JLL, AC, CM-M, and VR reviewed the literature and their own previous work and contributed to the discussions for writing this review. The main writer was VR, but all the authors contributed to the writing of the manuscript. AF-N and CM-M prepared the figures, with key inputs from VR.