Tuning the in vitro sensing and signaling properties of cyanobacterial PII protein by mutation of key residues

PII proteins comprise an ancient superfamily of signal transduction proteins, widely distributed among all domains of life. In general, PII proteins measure and integrate the current carbon/nitrogen/energy status of the cell through interdependent binding of ATP, ADP and 2-oxogluterate. In response to effector molecule binding, PII proteins interact with various PII-receptors to tune central carbon- and nitrogen metabolism. In cyanobacteria, PII regulates, among others, the key enzyme for nitrogen-storage, N-acetyl-glutamate kinase (NAGK), and the co-activator of the global nitrogen-trascription factor NtcA, the PII-interacting protein-X (PipX). One of the remarkable PII variants from Synechococcus elongatus PCC 7942 that yielded mechanistic insights in PII-NAGK interaction, is the NAGK-superactivating variant I86N. Here we studied its interaction with PipX. Another critical residue is Lys58, forming a salt-bridge with 2-oxoglutarate in a PII-ATP-2-oxoglutarate complex. Here, we show that Lys58 of PII protein is a key residue for mediating PII interactions. The K58N mutation not only causes the loss of 2-oxogluterate binding but also strongly impairs binding of ADP, NAGK and PipX. Remarkably, the exchange of the nearby Leu56 to Lys in the K58N variant partially compensates for the loss of K58. This study demonstrates the potential of creating custom tailored PII variants to modulate metabolism.


Results
Lys58 is crucial for ligand binding. Formation of the PII-NAGK complex requires a compact bending of the protruding PII T-loop. The T-loop is inserted in the interdomain cleft of the NAGK subunits for anchoring PII on NAGK 11 . In this tightly folded conformation, the T-loop residue Glu44 forms a salt-bridge with the highly conserved residue Lys58 6,12 . The same Lys58 residue plays a key role for 2-OG binding to the PII-Mg 2+ -ATP complex, through formation of a salt-bridge with the C5-carboxyl-group of 2-OG 5,24 . When the 2-OG levels increase in the cell, the PII-NAGK complex must dissociate to allow PII-2-OG interaction 1 . Therefore, the intramolecular Lys58-Glu44 salt bridge must break and instead, the salt bridge between Lys58 and 2-OG forms, arresting the T-loop in a novel conformation that precludes NAGK interaction 5,12 (summarized in 1,25 ). To directly assess the importance of PII Lys58 for NAGK complex formation, we created a SePII (K58N) variant, where Lys58 was replaced by Asn, and therefore the Lys58-Glu44 salt bridge cannot form. We decided to mutate the conserved Lys58 residue to Asn because this replacement is found in the PII-like protein SbtB 3 (PDB: 5O3R), as revealed by structure-based sequence alignment between SePII (PDB: 2XUL) and Synechocystis SbtB 3 ( Supplementary  Fig. S1).
The binding properties of SePII (K58N) variant in comparison to the wild-type SePII (WT) protein were analyzed by isothermal titration calorimetry (ITC) (Fig. 1). As expected, the PII variant K58N lost the ability to bind 2-OG ( Supplementary Fig. S1), whereas ATP was still able to bind (Fig. 1A), albeit weaker than the SePII (WT) protein ( Table 1). The binding of ADP was even more strongly impaired (Fig. 1B), yielding only very weak isotherm signals. This agrees with the fact that the Lys58 residue is crucial for establishing a stable T-loop conformation of the ADP-complex through hydrogen-bonding interaction with the Gln39 side chain 7,23 . In the PII-ADP structures (PDBs: 4CNZ and 4C3K), the side chain of residue Leu56 also approaches the Gln39 side chain. We, therefore, asked the question whether replacement of Leu56 by Lys might be able to compensate the K58N mutation. To test this hypothesis, we created a double point mutation variant SePII (K58N/L56K). The SePII (K58N/L56K) variant was first tested for its effector molecule binding properties using ITC. This PII variant was still unable to bind 2-OG ( Supplementary Fig. S1), but the binding affinity toward ATP was enhanced in comparison to the SePII (K58N) variant (Fig. 1, compare C with A). Especially, the first biding site was occupied with very high affinity. The ADP binding events to the SePII (K58N/L56K) variant induced a strong isotherm, and in particular, the affinity for the second binding site was strongly enhanced (Table 1, Fig. 1, compare D with B). As can be deduced from the binding isotherms, the binding enthalpy for ATP and ADP binding to SePII (K58N/ L56K) variant was stronger than to SePII (K58N) variant (Fig. 1, compare A with C and B with D). These results indicate that a re-location of the Lys-residue by two amino acid positions does not rescue 2-OG binding but positively influences adenyl nucleotide binding, in agreement with the fact that Lys58 is a direct ligand to 2-OG 5,24 but is only indirectly involved in adenyl-nucleotide binding 7,23 through affecting the T-loop conformation.
PII variants K58N and K58N/L56K are still able to activate NAGK. Next, we wanted to investigate the ability of SePII variants (K58N and K58N/L56K) to interact and activate NAGK. The PII-based activation of NAGK was tested using a coupled enzyme assay 10 with recombinant NAGK proteins deriving from strains S. elongatus PCC7942 (SeNAGK) and Synechocystis sp. PCC 6803 (ScNAGK). The catalytic activity of the SeNAGK was determined with NAG (N-acetyl-glutamate) as a variable substrate in presence or absence of different PII protein variants (kinetic constants are listed in Table 2). Both PII variants were able to activate SeNAGK ( Fig. 2A), but weaker than SePII (WT). With ScNAGK, the SePII (K58N) variant was not able to activate ScNAGK ( Supplementary Fig. S2) whereas the SePII (K58N/L56K) variant was again able to partially activate ScNAGK ( Supplementary Fig. S2). This shows that interaction of S. elongatus PII with the non-cognate ScNAGK (from Synechocystis) is in principle less robust against variations in the amino acid sequence. However, re-location of Lys58 to position 56 helps PII to adopt a compensatory conformation, which allows productive interaction with the non-cognate NAGK partner.
Tight complex formation of PII-NAGK is required to relieve arginine feedback inhibition. The relief of NAGK from arginine feedback inhibition through PII-NAGK complex formation is the rate-limiting step for the metabolic switch of the arginine biosynthetic pathway 9,10,13 . To examine whether the different PII www.nature.com/scientificreports www.nature.com/scientificreports/ variants would relief NAGK from arginine inhibition, we assessed SeNAGK activity at fixed concentration of NAG (40 mM) in presence of different concentrations of arginine with and without different PII protein variants. Arginine feedback inhibition of non-complexed SeNAGK occurred with a half maximal inhibitory concentration (IC 50 ) of 11 µM (Fig. 2B). As expected, the arginine inhibitory effect on SeNAGK was strongly released in presence of SePII (WT) protein (Fig. 2B). Intriguingly, both SePII variants (K58N and K58N/L56K), which were able to activate SeNAGK, failed to markedly relieve SeNAGK from arginine feedback inhibition, although a subtle effect was observed (with IC 50 values of 14.2 or 15.7, as compared to 11 µM for SeNAGK alone) (Fig. 2B). Together, it appeared that the double point mutation variant activated SeNAGK more strongly and was slightly more efficient in relieving SeNAGK from arginine inhibition than the single point mutation (K58N) (Fig. 2B).
The catalytic activity of the SeNAGK with NAG as a variable substrate in the presence of 11 µM arginine (which corresponds to the IC 50 of SeNAGK alone) with or without different PII variants was determined to discern the two PII mutant variants (kinetic constants are listed in Table 2). In this assay, the SePII (K58N/L56K) variant was much more efficient than the SePII (K58N) variant to activate SeNAGK (Fig. 2C). Due to the low activity, the kinetic constants for the activation of SeNAGK by SePII (K58N) variant could not be calculated, as the data could not be fitted to Michaelis-Menten equation. Similar results were obtained using the ScNAGK enzyme ( Supplementary Fig. S2). Together these results reinforce that mutation of Leu56 to Lys56 in the SePII (K58N/L56K) variant partially compensates the function of Lys58 in stabilizing PII-NAGK complex, most likely by approaching again the Glu44 residue of the T-loop (PDBs: 2XBP and 2V5H).
Next, we determined the response of the various PII-NAGK complexes to 2-OG in presence of 11 µM arginine, the IC 50 of free SeNAGK. As shown in Fig. (2D), the addition of 2-OG had a negligible influence on both variants of PII proteins confirming the inability of Lys58 mutant to bind 2-OG effectively (Fig. 2D). Since in presence of 11 µM arginine the activity of NAGK is only weakly inhibited in presence of SePII (WT), the effect of 2-OG is modest. However, in presence of 25 µM arginine, the complex of SeNAGK-SePII (WT) showed the typical inhibitory effect upon 2-OG addition 6,10 ( Supplementary Fig. S2). Form the previous enzyme assays, it turned out that both PII variants were still able to interact with NAGK, but this interaction seemed to be impaired to different extent, with the (K58N/L56K) variant partially compensating the impairment of the K58N variant. To further confirm this assumption, we first tried to isolate PII-NAGK complexes of different variants of PII (K58N) and (K58N/L56K) using analytical gel-filtration coupled with multi angle light scattering (MALS) (Fig. 3A). SePII (WT) protein forms a stable complex with SeNAGK with a molar-mass of 275.4 kDa, corresponding to two-PII trimers (each trimer of ≈40.8 kDa) sandwiching one hexameric SeNAGK (≈194 kDa) 12 . In this experimental setting, both proteins eluted together from the gel-filtration column with an apparent mass according to MALS analysis of 255.9 kDa (Fig. 3A), which agrees almost with the theoretical calculated mass of NAGK-PII complex. However, with both variants, we were unable to detect any indications of a stable SeNAGK-PII complex, since SeNAGK eluted as free hexamer with an apparent mass of 193.8 kDa (Fig. 3A). In agreement, the corresponding fractions as analyzed by SDS-PAGE contained SeNAGK without any traces of PII (Fig. 3B). This experiment confirmed that the physical interaction between SeNAGK and different variants of PII protein is weak and the weak complexes dissociate during the gel-filtration.
Additionally, we assessed whether binding of SePII (K58N) and (K58N/L56K) variants to SeNAGK could be measured by surface plasmon resonance (SPR) spectroscopy (Fig. 3C). The His-tagged SeNAGK was immobilized on a Ni-NTA sensor chip and SePII variants (K58N) and (K58N/L56K), as well as SePII (WT) protein, were injected sequentially. No response signal could be detected in binding assays between SeNAGK and the two PII variants (K58N) or (K58N/L56K), implying that the interaction between these SePII variants and SeNAGK is too weak to be detected by SPR in comparison to SePII (WT), which forms a strong complex with SeNAGK (Fig. 3C).

K58N mutation influences negatively PII-PipX complex formation. Another major interacting
partner of the PII signaling network in cyanobacteria is PipX 25 . In the absence of 2-OG, three PipX monomers can bind to one PII trimer 15 . The crystal structure of the PII-PipX complex revealed that the T-loop of PII acts as an antenna that attracts the PipX monomers 15,26 . Unlike the bent conformation of the T-loop in NAGK-PII complex, the T-loop in the PipX-PII complex is in an extended conformation 15 , resembling PII in its ADP-complex 7 . PII in the Mg 2+ -ATP-2-OG complex is unable to bind PipX 1,25 . Since the PII (K58N) mutation impaired ADP binding (Fig. 1), we assumed that the K58N variant may affect PII-PipX complex formation. We analyzed the formation of PII-PipX complexes by SPR spectroscopy, using an indirect assay, as described previously 8,16 . When PipX is incubated with PII prior the injection on the sensor chip, PII-PipX complex formation increases the binding of His 6 -PipX to the sensor chip due to mass increase and PipX trimerization (Fig. 4A). In absence of the effector molecules, the PII (K58N) variant was not able to increase PipX binding above the background level www.nature.com/scientificreports www.nature.com/scientificreports/ (binding of His 6 -PipX to the sensor in absence of PII), while the PII (K58N/L56K) was able to partially restore the binding of His 6 -PipX to the sensor chip, although not as efficient as PII (WT) (Fig. 4A). In presence of 3 mM ADP, the PII (K58N) variant regained the ability to at least partially interact with PipX (Fig. 4B). Again, the PII (K58N/L56K) variant showed a stronger interaction with PipX than the PII (K58N) variant but still weaker than PII (WT) (Fig. 4B). To compare the effect of PII variants on the dissociation rate of the complexes from the sensor chip (which is a good indicator of the efficiency of PII-PipX interaction), the dissociation curves were normalized to the RUs at the end of the association phase (taken as 100%) (Fig. 4C). In presence of ADP, the PipX-PII (K58N/L56K) variant complex dissociated slowly, in comparable way to the PipX-PII (WT) complex, while the PipX-PII (K58N) complex dissociated faster (Fig. 4C), confirming again that L56K could compensate the loss of Lys58. In presence of 1 mM ATP, PII (WT) bound to PipX weaker than in presence of ADP, in agreement with previous reports 8, 16 , while both PII variants (K58N) and (K58N/L56K) were not able to efficiently interact with PipX (Fig. 4D).

The T-loop of the PII (I86N) variant allows PipX interaction. Previously, we reported a NAGK
hyper-activating variant PII (I86N) 6,22 . The I86N substitution causes the T-loop of PII to adopt a compact conformation through formation of hydrogen bond between the backbone oxygen of Thr43 and the amido group of Asn86 resulting in a contraction of the T-loop 6 . As consequence, the PII (I86N) variant binds constitutively to NAGK. However, the interaction of PII (I86N) variant with PipX has not been analyzed before. Here, we used the same indirect SPR assay to determine PII-PipX complex formation by determining binding of the complex to the Ni-NTA sensor surface. When the PII (I86N) variant or PII (WT) was pre-incubated with PipX in the absence of effector molecules, the PII (I86N) variant promoted a stronger binding of PipX to the Ni-NTA sensor chip surface and the complex dissociated slower than with PII (WT) (Fig. 5A). To quantitatively compare the effect of the two PII proteins on the dissociation of the complexes from the sensor chip, the dissociation curves were normalized to the RUs at the end of the association phase (taken as 100%) (Fig. 5B). The percent RUs remaining bound to the www.nature.com/scientificreports www.nature.com/scientificreports/ chip after 400 s of dissociation were then taken as a proxy for PII-PipX interaction and used to quantify the effect of different effector molecules (Fig. 5B-E).
Different effector molecules were tested on PII-PipX complex stability. As expected, a strong positive effect of ADP on the interaction of PII (WT) with PipX, was obtained 8 (Fig. 5D,E). By contrast, binding of the PII (I86N) variant to PipX was negatively affected by ADP (Fig. 5C-E). ATP showed for the PII (WT) protein a slightly lower stability of the complex than in the ADP-complexed state whereas the PII (I86N) variant interacted stronger with PipX in the ATP state than with ADP ( Fig. 5C-E). As expected, 2-OG in presence of ATP impaired PipX-PII (WT) complex formation 8,16 (Fig. 5D). An inhibitory effect of ATP and 2-OG on the PII-PipX complex was also visible with the PII (I86N) variant, however not as strong as with PII (WT) protein (Fig. 5C-E). This agrees with the decreased affinity of the PII (I86N) variant towards 2-OG 6 . Taken together, these data demonstrated that the www.nature.com/scientificreports www.nature.com/scientificreports/ PII (I86N) variant is very efficient in complex formation with PipX and that this binding does not require positive stimulation by ADP as is the case with PII (WT). It appears that the T-loop of this variant is not permanently fixed in the bent conformation but is still flexible enough to adopt the extended conformation for PipX binding, highlighting the defining role of PII target proteins for the actual T-loop conformation.

Discussion
Previous structural analysis of PII proteins in various complexed states showed that Lys58 plays an important role in the function of PII 5,7,23,24 . However, the Lys58 residue is not involved in direct interactions with ADP or ATP, but seems to be important to stabilize the position of the B-loop via a hydrogen bond to residue Gly87 (PDBs: 2XUL, 3MHY, and 2XBP). Binding of ADP induces conformational changes within the surface exposed T-loop enabling Lys58 to interact with Gln39 at the base of the T-loop 23 . Thereby, Lys58 anchors the T-loop in the ADP bound conformation via hydrogen bond to Gln39 (PDB: 4CNZ) 7,23 . Further, it anchors also the T-loop in the bent conformation of the PII-NAGK complex via a salt-bridge to Glu44 (PDBs: 2V5H and 2XBP) 6,12,[23][24][25] .
Remarkably, binding of 2-OG is coordinated mainly by the highly conserved residues of PII proteins Lys58 and Gln39 5,24 . When 2-OG breaks the interaction between PII and NAGK, the Lys58-Glu44 interaction in PII is replaced by a new salt bridge interaction between Lys58 and 2-OG (PDBs: 2XUL and 3MHY). Elegantly, the C5 carboxyl group of 2-OG quite precisely replaces the Glu44 carboxyl group, such that Lys58 does not need to change its orientation. Therefore, we expected the K58N variant to be defective in both 2-OG binding and in anchoring the T-loop in the NAGK-bound conformation.
Complex formation of PII with NAGK follows a two-step mechanism 6 : first, an encounter complex involving the B-loop of PII is formed. This involves a contact between Arg233 of NAGK and Glu85 of PII. This interaction breaks a salt bridge between Glu85 and Arg47 in the T-loop of PII, which is now free and can adopt a bended structure. The T-loop bended confirmation is stabilized by a new salt bridge between Lys58 and T-loop residue Glu44. In the second step, the bended T-loop deeply inserts into the NAGK clefts to form the tight complex. Apparently, the mutation of Lys58 still allows bending of the T-loop and insertion into the NAGK clefts (with the help of the other hydrophobic interactions between Ile229, Ile253, and Ala257 of NAGK and Phe11 and Thr83 of PII) 12 , however, the overall stability of the complex is weakened and feedback inhibition by arginine is almost not affected (Fig. 2B). Therefore, we conclude that a tight complex between PII and NAGK is required to relief NAGK from arginine feedback inhibition, whereas the weak complex is sufficient to activate the NAGK in absence of arginine ( Fig. 2A).
The in vitro assays confirmed that Lys58 is a key residue for proper signaling function of PII via affecting the sensory properties of ATP/ADP/2-OG as well as the interactions with NAGK and PipX. The PII (K58N) variant is still able to bind ATP but lost the ability to bind 2-OG and influence ADP binding negatively. Consistent with our data, previous results demonstrated that a PII (K58M) variant was not able to bind to 2-OG 5 . The Lys58 residue is not involved in direct interactions with adenyl nucleotides (ATP and ADP) but rather stabilizes the T-and B-loops via hydrogen bonds to Gln39 and Gly89, respectively. Therefore, we assume that the defect in anchoring the T-and B-loops in the PII (K58N) variant is most likely the cause for changing the affinities of ATP and ADP in comparison to SePII (WT) ( Table 1). The effect of the Lys58 mutation in the destabilization of PII T-loop is probably the dominating effect and explains, why the Lys58 mutation impairs the NAGK and PipX interactions, which are mediated by the T-loop of PII. Intriguingly, the addition of a compensatory mutation to K58N by replacement of the nearby Leu56 to Lys partially restores the ability of this PII variant to interact with NAGK and PipX. This compensatory effect indicates that the Leu56 to Lys mutation might restore a productive T-loop bent conformation, possibly by approaching the Glu44 residue of the T-loop (model from PDBs: 2XBP and 2V5H).
In cyanobacteria, the regulation of nitrogen metabolism mainly depends on interaction network composed of the signal-transduction protein PII, the transcription factor NtcA and its co-activator PipX (PII interacting protein X) 25 . The crystal structure of PII (I86N) 6 (PDB: 2XBP) showed an almost identical backbone as PII (WT). However, the T-loop adopts a compact conformation, which is a structural mimic of PII in NAGK complex 6 . Hence, the PII (I86N) variant strongly activates NAGK in vitro and in vivo, causing huge accumulation of arginine and cyanophycin 6,22 . Therefore, we assumed that this variant was unable to interact with other PII interaction partners, which would influence the NtcA regulon through impaired PipX interaction. To our surprise, we detected strong interaction between the PII (I86N) variant and PipX. SPR spectroscopy confirmed that the PII (I86N) maintains the ability to bind PipX. In absence of effector molecules, the PII (I86N) variant showed even stronger interaction with PipX than PII (WT). Furthermore, the PII (I86N) variant did not require positive stimulation by ADP to form a stable complex as it is the case of PII (WT). It was previously shown that the PII (I86N) variant did not respond to 2-OG in complex with NAGK 6 . Nevertheless, an inhibitory effect of ATP and 2-OG on PII (I86N)-PipX complex was observed, although not as strong as in the case of the PII (WT) protein. Altogether, the interaction of PipX with PII (I86N) is different than with PII (WT), and complex stability is generally higher compared to PII (WT). Due to the stronger sequestration of PipX by PII (I86N), we speculate that PII (I86N) might tune down PipX mediated activation of NtcA in vivo. This could explain the delayed nitrogen starvation response observed in the Synechocystis strain BW86 harboring the PII (I86N) variant 22 . As a future perspective, a transcriptomic analysis would be informative to characterize the in vivo influence of PII (I86N) on the NtcA regulon, in response to nitrogen limitation conditions. Altogether, our study provides new insights into the signaling function of PII proteins and sheds some lights on the plasticity of the PII core-structure. This is, for example, demonstrated by the partial compensation of the Lys58 mutation, by the Leu56 to Lys mutation indicating the plasticity of the PII body. This flexibility of PII body could be used in protein engineering to design new proteins or sensors with new desired characters. Notably, the PII variant (I86N) is able to bind citrate as a new effector molecule 27 , and was used to engineer a cyanobacterial strain that produces huge amounts of the biopolymer cyanophycin 22 . In other studies, the trimeric architecture of PII proteins was used to engineer 2-OG 28,29 or ATP 30 sensors. Since the PII variants (K58N) and (K58N/L56K) no