ATPase as a switch in P_II signal transduction

The signaling protein P_II, highly conserved in archaea, bacteria, and plants, is an example of beauty in biology (1). P_II is relatively small, a homotrimer with each subunit of just 12–13 kDa; yet it binds to several effectors and exercises its regulatory roles by interacting with its partners that can be enzymes, transcriptional factors, or membrane transport proteins. Moreover, P_II can be modified posttranslationally, which may take the form of phosphorylation, uridylylation, or adenylylation. Depending on how many of the three monomers of P_II bind to an effector or are modified, P_II is given a large array of conformational variations that allow it to signal various metabolic states of the cells and fine-tune the metabolic activities. The primary role of P_II is in the control of nitrogen metabolism; it was first reported in 1969 (2) and has been extensively studied since then. In a review published in 2013, 47 structures of the P_II proteins were recorded, representing 19 proteins from 16 organisms spreading among the three domains of life (1). Thus, one may think that we know more or less all the details about the biochemical basis of P_II function. Now, Radchenko et al. report in PNAS that P_II has a weak but detectable activity of ATPase (3), and their studies clarify several puzzling observations previously reported on the function of P_II.

Because of the importance of nitrogen metabolism, bacteria harbor complex regulatory mechanisms that allow them to adapt to the changes of nitrogen supply in their environment. These mechanisms can operate at different levels, from transcriptional control to protein–protein interaction, or transport of nitrogen compounds across the cell membrane. P_II can be involved in any of these regulatory levels in different organisms so far studied (1). Three subgroups of P_II proteins are defined based on phylogenetic studies: GlnB, GlnK, and NifI. With exceptions, glnB is linked to glnA encoding the glutamine synthetase or nadE encoding NAD⁺ synthetase; glnK is cotranscribed with amtB encoding ammonium transporter; nifI is associated with the structural genes encoding the nitrogenase complex. All P_II proteins so far crystallized possess a well-conserved 3D structure. The compact, cylindrical, core structure of P_II is formed by an antiparallel β-sheet of four β-strands with two α-helices at the lateral surface. A variable loop, the T-loop, extends from each core, and forms an antenna-like structure (Fig. 1). The T-loop contains the posttranslational modification site and can adopt either a flexible conformation or a rigid one. Both the binding of the allosteric effectors and posttranslational modification affect the conformation of the T-loop, which acts as a plastic site to recognize various protein partners. The binding sites for the three allosteric effectors, ATP/ADP and 2-oxoglutarate (2-OG), are highly conserved among the P_II proteins. Three nucleotide-binding sites are located in the clefts at the interface between two monomers, and 2-OG binds at the vicinity of the nucleotide-binding site. ATP and 2-OG binding occur cooperatively and require the presence of Mg²⁺ ions (4).

Fig. 1.

P_II is a homotrimer, each subunit with a cylindrical core structure and antenna-like T-loop (in blue) extending out from the core. (A) Under nitrogen limitation, Mg-ATP and 2-OG (red dot) bind at proximity at the cleft between each subunit; the presence of 2-OG inhibits the hydrolysis of ATP, and the T-loops adopt a flexible conformation. (B) When nitrogen supply increases, 2-OG is consumed for ammonium assimilation and its level drops; Mg-ATP is hydrolyzed to ADP and the T-loops become more rigid. ADP inhibits ATP binding under low 2-OG levels. (C) High 2-OG and ATP concentrations favor the exchange of ADP to ATP, and P_II is reloaded with Mg-ATP and 2-OG.

2-OG is a key intermediate of the Krebs cycle and a signal of nitrogen status (1, 5, 6). P_II has been considered as a signaling protein that integrates the status of carbon and nitrogen availability, as well as the energy charge. Although the fact that P_II acts as a sensor of 2-OG is generally well accepted, the role of binding of ATP/ADP is much less obvious (1). ATP and ADP compete for binding to P_II, leading to the proposition that P_II may be a sensor of energy charge. This hypothesis has yet to be confirmed in vivo, and raises several questions. First, a large number of P_II proteins from different organisms have been crystallized, some of them with ADP bound on them and others with ATP. In some cases, ADP was even found on P_II when ATP was initially added during the experimental process (1). Second, at least in Escherichia coli cells, the concentration of ATP is 6- to 10-fold higher than that of ADP (7, 8); thus, it is difficult to imagine that ATP drops to a level where ADP competes efficiently for P_II binding. Finally, the energy charge [(ATP + 0.5× ADP)/(ATP+ADP+AMP)] is relatively stable and goes around 0.9 in E. coli; cells become nonviable when it goes down to less than 0.5 (8). Thus, the significance of the binding of the two nucleotides on P_II remains unclear.

Radchenko et al. (3) use GlnK from E. coli as a model to test the ATPase activity of P_II. The target of GlnK in E. coli is AmtB, a transmembrane ammonium channel (9). Under nitrogen limitation, the level of 2-OG remains relatively high; under such conditions, GlnK is cytoplasmically located, urydylylated at the T-loop, and expected to contain ATP, Mg²⁺, and 2-OG. When nitrogen-limited cells get a sudden ammonium supply, the level of 2-OG drops rapidly and P_II becomes deurydylylated; the T-loop adopts a rigid conformation protruding into the AmtB channel, forming a complex in a 1:1 stoichiometry to block further ammonium entry (9). The idea that P_II may have an intrinsic ATPase activity was first proposed by Gruswitz et al. in their publication about the AmtB-GlnK complex (9). ATP enhances the binding of GlnK to AmtB, but the complex incubated initially with ATP contained ADP instead. Based on the proximity of the ADP terminal phosphate to side chains of several positively charged residues (Arg and Lys) and a highly-coordinated buried water in GlnK, Gruswitz et al. suggested that this nucleotide binding pocket may serve to catalyze ATP hydrolysis. It took a few years for a laboratory to test this hypothesis experimentally, and one of the reasons may be the weak activity of ATP hydrolysis of GlnK. Under standard experimental procedures—that is, using a mixture of cold ATP molecules with radioactively labeled ones—ATP hydrolysis was hardly seen. However, this procedure was possible when GlnK was incubated with hot ATP molecules only (3). Radchenko et al. offer two lines of evidence to rule out the possibility of a contaminating activity from cell extract. First, this low ATP hydrolysis activity is no longer present in several GlnK mutants, each replacing K90, R101, or R105, to an Ala residue. These three positively charged residues have been previously suggested by Gruswitz et al. as being possibly involved in ATP hydrolysis (9). All three mutants displayed a much weaker ATPase activity compared with the wild-type. Second, the physiological effector, 2-OG, inhibits ATP hydrolysis of GlnK, and this inhibitory effect requires Mg²⁺ ions. The involvement of Mg²⁺ can be explained by the role of this ion in coordinating ATP and 2-OG binding in P_II proteins (4); each 2-OG binds to Mg²⁺ ion, which is in turn coordinated by three phosphate groups of ATP. As expected, those mutant variants, K90A, R101A, and R105A, although displaying a weak activity of ATP hydrolysis compared with the wild-type, still retain the ability to respond to the presence of 2-OG. In contrast, mutation of either residue Q39 or K58, both involved in 2-OG binding, showed either little ATP hydrolysis or had a weak ATPase activity that is no longer influenced by the presence of 2-OG (3). The ATP hydrolysis activity of GlnK in E. coli can further be extended to GlnZ, a GlnK homolog in Azospirillum brazilense, and the P_II protein from the plant Arabidopsis thaliana, suggesting that it is a general phenomenon among P_II proteins (3).

The ATPase activity of P_II acts as a switch, much as many small GTPases, in driving protein conformational change in signal transduction.

What is the major insight into the signaling mechanism of P_II? Under nitrogen limitation, 2-OG levels are high in the cells, and that favors the binding of Mg-ATP on each of the three pockets of P_II; the T-loop adopts a flexible conformation (Fig. 1). When the 2-OG level drops as nitrogen supply increases, less 2-OG occupancy triggers Mg-ATP hydrolysis into ADP, and this activity acts to translate the drop of 2-OG signal into a change in the T-loop conformation, which becomes rigid and extended. Thus, the ATPase activity of P_II acts as a switch, much as many small GTPases, in driving protein conformational change in signal transduction. Although the change of the P_II form with 2-OG and Mg-ATP to that with ADP can be well described based on the extensive structural information of P_II and the recently described data by Radchenko et al. (3), how ADP is kicked out of P_II by ATP is much less obvious to imagine, because ADP and ATP compete for binding to P_II (1). We suggest the following scenario: under low 2-OG level, P_II hydrolyzes ATP into ADP, which occupies the nucleotide-binding pocket and inhibits ATP binding (1, 3); thus, as long as the 2-OG levels remain relatively low, ADP-bound P_II should be favored over ATP-bound one, or the ATP-bound form cannot be stabilized (Fig. 1). However, in E. coli cells, the ATP concentration is 6- to 10-times higher than that of ADP (7, 8), which should help P_II to reload ATP when 2-OG level increases as the later molecule stabilizes Mg-ATP at its binding pocket (Fig. 1). In essence, the P_II function is to sense the level of 2-OG, a signal of nitrogen availability, and ATP hydrolysis serves as a switch for conformational change of P_II as 2-OG level varies.

Although the data of Radchenko et al. (3) simplify our view on the function of P_II in the control of nitrogen metabolism, the finding raises more questions, and some debate would even be expected. Why is the activity of ATP hydrolysis by GlnK so low, and how to explain the rapid response of P_II to nitrogen availability in view of such a low activity? Is there an activator for the P_II ATPase activity? How to confirm this finding in vivo? Obviously, the more we understand on the function of P_II, the more questions are raised. This aspect may be why P_II is so fascinating and has been studied ever since its discovery in 1969 (2).