Specificity and Regulation of Interaction between the P_II and AmtB₁ Proteins in Rhodospirillum rubrum

Abstract

The nitrogen regulatory protein P_II and the ammonia gas channel AmtB are both found in most prokaryotes. Interaction between these two proteins has been observed in several organisms and may regulate the activities of both proteins. The regulation of their interaction is only partially understood, and we show that in Rhodospirillum rubrum one P_II homolog, GlnJ, has higher affinity for an AmtB₁-containing membrane than the other two P_II homologs, GlnB and GlnK. This interaction strongly favors the nonuridylylated form of GlnJ and is disrupted by high levels of 2-ketoglutarate (2-KG) in the absence of ATP or low levels of 2-KG in the presence of ATP. ADP inhibits the destabilization of the GlnJ-AmtB₁ complex in the presence of ATP and 2-KG, supporting a role for P_II as an energy sensor measuring the ratio of ATP to ADP. In the presence of saturating levels of ATP, the estimated K_d of 2-KG for GlnJ bound to AmtB₁ is 340 μM, which is higher than that required for uridylylation of GlnJ in vitro, about 5 μM. This supports a model where multiple 2-KG and ATP molecules must bind a P_II trimer to stimulate release of P_II from AmtB₁, in contrast to the lower 2-KG requirement for productive uridylylation of P_II by GlnD.

The ammonium channel/rhesus (Amt/Rh) family of proteins is a widely distributed group of trimeric integral membrane proteins found in all domains of life that can function as gas channels of ammonia and perhaps carbon dioxide (13, 20, 29, 39, 43). A subset of this family, the AmtB proteins, is found in bacteria, archaea, some lower eukaryotes and plants. Homologs of amtB are often found in close proximity to genes encoding P_II homologs (46). P_II regulatory proteins are also found in most prokaryotes and some plant chloroplasts (3). P_II is a small, soluble, trimeric protein that regulates the functions of several other proteins involved in nitrogen metabolism (3, 4, 35, 36). AmtB appears to have two roles in the cell. The first function, to act as a channel for uncharged ammonia, has been explored physiologically, structurally, and computationally (28, 33, 44). The second function of AmtB is to interact with P_II and has only recently been described (6, 9, 21, 22, 53).

AmtB proteins have been shown to interact with homologs of P_II in several bacteria and in the archaeon Methanococcus jannaschii (8, 10, 17, 18, 44, 45, 47, 50, 53). The association of P_II with AmtB can physically block the ammonia gas channels of AmtB under conditions of nitrogen sufficiency in the cell. In addition, the AmtB-P_II complex is able to recruit at least one other protein to the membrane, dinitrogenase reductase-activating glycohydrolase (DRAG), in organisms capable of nitrogen fixation (22, 48). This membrane sequestration requires both an Amt protein and a P_II protein and results in the inability of DRAG to activate dinitrogenase reductase in vivo. Finally, AmtB is able to remove equimolar amounts of P_II from the cytoplasm, preventing P_II from interacting with at least some other proteins. Although membrane sequestration of P_II has been shown to be important in recovering from nitrogen starvation in Escherichia coli, this function of AmtB is still poorly understood (6).

P_II homologs sense several small-molecule pools. The best-studied P_II protein is GlnB from E. coli, and insights into P_II protein function from this homolog apply to all P_II proteins studied to date. Carbon status and perhaps energy status are sensed directly by the binding of 2-ketoglutarate (2-KG) and ATP to each subunit of the P_II trimer (26). The binding of one 2-KG molecule or no molecules to P_II signals carbon deficiency. At higher 2-KG concentrations, the binding of second and third 2-KG molecules to P_II signals carbon sufficiency (26, 36). The binding of ATP lowers the K_m of P_II for 2-KG, which could act as a signal of energy sufficiency in the cell; however, previous studies have suggested that ATP may always present at saturating levels in E. coli (23). In contrast, cellular nitrogen status is indirectly sensed by the reversible modification of P_II, rather than the direct binding of nitrogen-rich metabolites. In E. coli and other bacteria such as Rhodospirillum rubrum, the bifunctional, uridylyltransferase/uridylyl-removing GlnD protein carries out uridylylation of P_II. In E. coli this function is regulated by glutamine levels: low levels of glutamine, signaling nitrogen deficiency, lead to uridylylation of P_II (P_II-UMP) by GlnD, while high levels of glutamine, signaling nitrogen sufficiency, lead to deuridylylation of P_II-UMP by GlnD (3, 35). GlnD appears to sense 2-KG as well as glutamine in R. rubrum, where high levels of 2-KG stimulate uridylylation, and high levels of glutamine stimulate deuridylylation, of P_II by GlnD (25).

The specificity and regulation of interaction between P_II and AmtB proteins are not fully understood. There is evidence that both GlnB and GlnK of E. coli can interact with the sole AmtB species present in that organism in vivo, and it is suggested that GlnK may interact more strongly with AmtB than does GlnB (9). Likewise, both P_II homologs of Azospirillum brasilense appear able to associate with AmtB at the membrane in vivo (21). Only unmodified P_II has been found to interact with AmtB, and uridylylation of P_II has been shown to prevent the formation of the P_II-AmtB complex in E. coli, A. brasilense, and R. rubrum (11, 22, 53). Crystal structures of the E. coli complex reveal that the uridylyl group could sterically interfere with the surface of interaction between the two proteins (8, 17). However, not all P_II proteins are regulated by uridylylation. In Streptomyces coelicolor the same tyrosine residue uridylylated in enteric bacteria is instead adenylylated (19), and in cyanobacteria a conserved serine in P_II is phosphorylated in response to nitrogen limitation (16). In other organisms P_II might not be modified at all (49). Regardless of the decoration's identity, it is likely to inhibit interaction with an AmtB protein because of the location of the modified residue, but this hypothesis has not been tested.

Finally, small molecules are known to have diverse regulatory effects on the P_II-AmtB complex, depending on the organism studied. In E. coli purified AmtB-GlnK complexes disassociate only in the presence of both ATP and 2-KG, while in M. jannaschii purified GlnK1-Amt1 complexes disassociate in the presence of low levels of ATP (11, 50). These may reflect real physiological differences in regulation in different organisms or may be complicated by the buffers and detergents used to purify the protein complexes. To date, there has been no in vitro study of the regulation of this interaction using AmtB in its native state as an integral membrane protein.

There are three P_II homologs, GlnB, GlnJ, and GlnK, in R. rubrum that interact with up to five other proteins involved in nitrogen regulation (52, 53, 54). All three P_II homologs can regulate the activity of GlnE (ATase), which reversibly adenylylates glutamine synthetase (52). Expressed from their native promoters, GlnB and GlnJ, but not GlnK, regulate the NtrBC two-component regulatory system, leading to expression of genes involved in nitrogen metabolism under conditions of nitrogen deficiency (52). GlnB and GlnJ also indirectly regulate nitrogen fixation by controlling the activities of two enzymes with opposing functions, dinitrogenase reductase ADP-ribosyl transferase (DRAT) and DRAG, which reversibly modify dinitrogenase reductase (52). Finally, only uridylylated GlnB is able to interact with and activate NifA, which is a transcriptional activator of genes involved in nitrogen fixation (52). Thus, the three P_II homologs in R. rubrum are not functionally interchangeable. The genes encoding the two GlnK-type P_II proteins, glnJ and glnK, each have amtB homologs immediately downstream. However, only the amtB₁ gene downstream of glnJ plays important roles in the regulation of P_II in the cell (53). We therefore chose to examine whether specificity for a particular P_II existed in the R. rubrum AmtB₁ protein, as this could have substantial physiological consequences. Once such specificity was observed, we examined the role that small molecules have on the strength of interaction between a single P_II, GlnJ, and AmtB₁.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

The R. rubrum strains used in this study were UR2 (wild type), UR757 (ΔglnB3 glnK1::aacC1), and UR794 (amtB₁1::aacC1) (27, 40, 53). E. coli strains used in this study were UQ4037 (pUX754) and UQ4038 (pUX755). Similar to the construction of an R. rubrum GlnB overexpression vector (pUX753), a 0.4-kb fragment of R. rubrum glnK or glnJ was PCR amplified and cloned into pND706 at NdeI and EcoRI sites and then transformed into E. coli strain BK (ΔglnB Δmdl-glnK::Kan), yielding UQ4037 (pUX754) and UQ4038 (pUX755), respectively (32, 44). R. rubrum strains were grown in rich medium (SMN) or minimal malate glutamate medium (MG), as described previously (14, 30). To prepare larger volumes of cells for membrane harvesting, MG-grown cells were prepared by inoculating 100:1 from fully grown SMN cultures into MG medium and grown anaerobically under light to an optical density at 600 nm (OD₆₀₀) of about 0.9 (53). E. coli strains were grown in rich medium, Luria broth (LB) or LC (similar to LB medium but with NaCl reduced from 10 g/liter to 5 g/liter) (40).

Protein purification.

GlnJ and GlnK were purified using a protocol similar to that used previously to purify GlnB (54). UQ4037 (for GlnJ) and UQ4038 (for GlnK) were grown overnight in 10 ml of 2× LB and inoculated into 600 ml of 2× LB medium, both supplemented with 100 μg/ml ampicillin and 25 μg/ml kanamycin. Cultures were grown at 30°C to an OD₆₀₀ of 0.8 to 1.1 and heat shocked at 44°C for an additional 4 to 5 h. Cells were concentrated by centrifugation at 10,800 × g for 10 min, resuspended in 50 mM Tris, pH 7.5, and broken by French press. Cell lysate was centrifuged as before, and then protein in the supernatant was precipitated with 45% (final) ammonium sulfate, centrifuged as before, and resuspended in 50 mM Tris, pH 7.5. GlnJ and GlnK were purified to greater than 75% purity by eluting the protein from a Q-Sepharose column with a gradient of 100 to 400 mM NaCl. GlnJ was further purified to over 95% purity on a G-100 size exclusion column. GlnK was further purified to over 90% purity on a hydroxyapatite column after elution with 10 mM KPO₄ buffer, pH 7.5. Proteins were resuspended in 50 mM Tris, 100 mM NaCl, and 5% glycerol, pH 7.4.

Membrane preparation.

Cell harvesting, breakage by sonication, and separation of whole-cell clarified extract into cytoplasmic and membrane fractions using ultracentrifugation were performed as described previously (53). Separation of the P_II-AmtB₁-containing membrane into soluble and membrane-bound protein fractions was done in the following manner. Samples were centrifuged at 200,000 × g for 30 min at 4°C, and 80% of the total volume was removed as the supernatant fraction. The remaining supernatant was discarded, and the membrane pellet was resuspended in an equal volume of 50 mM NaPO₄ buffer, pH 6.8. The resuspended membrane was frozen at −80°C until use.

Membrane extract from UR757 (glnB glnK) without bound P_II was prepared by incubating membranes for 1 h at 30°C in the presence of 10 mM 2-KG and 3 mM ATP, followed by two rounds of ultracentrifugation at 200,000 × g for 30 min at 4°C. All supernatant was removed, and the pellet was resuspended in an equal volume of 50 mM NaPO₄ buffer, pH 6.8, after each centrifugation to remove 2-KG and ATP from the final membrane preparation. No GlnJ was present in the membrane extract following such treatment, as determined by Western blotting using anti-GlnJ antibodies.

P_II disassociation and reassociation assays.

Both disassociation and reassociation reactions took place in reaction buffer with a final concentration of 50 mM NaPO₄ buffer, pH 6.8, 100 mM NaCl, 25 mM MgCl₂, 1 mM MnCl₂, and 1 mM dithiothreitol, adapted from previous reports (9, 24, 54). ATP, ADP, and AMP-PNP [adenylyl 5′-(β,γ-imido)triphosphate tetralithium salt], at 95% or greater purity (Sigma-Aldrich, St. Louis, MO), and 2-KG were added to reactions. Membrane extract (250 μl), normalized to an OD₅₈₁ of 0.14, was added to 250 μl of 2× reaction buffer, and the mixture was mixed and incubated for 1 h at 30°C. Afterwards, the reaction mixtures were immediately placed on ice and then transferred to ultracentrifugation tubes for separation of soluble proteins from membranes and insoluble proteins by ultracentrifugation as described above. Incubation of reaction mixtures on ice prevented further changes in association between GlnJ and AmtB₁ (data not shown).

Immunoblotting of R. rubrum GlnB, GlnJ, and GlnK.

Trichloroacetic acid precipitation was used to extract and concentrate protein from all protein samples (51). Proteins samples were loaded on low-cross-linker tricine sodium dodecyl sulfate-polyacrylamide gel electrophoresis gels as described previously (41, 53). Proteins were electrophoretically transferred to a nitrocellulose membrane and probed with polyclonal antibody against R. rubrum GlnJ or GlnB. Visualization was performed with either enhanced chemiluminescence detection reagent (Amersham Biosciences, Piscataway, NJ) and Kodak film for protein detection or with SuperSignal West Dura extended-duration substrate (Pierce, Rockford, IL) for quantitation using a Storm 860 phosphorimager (Molecular Dynamics, Sunnyvale, CA).

Software.

Storm scanner control, version 5.03 (Amersham Biosciences), was used to operate the Storm 860 phosphorimager, and ImageQuant, version 5.0, build 050 (Molecular Dynamics), was used to quantitate bands from Western blots. Data were then exported to Excel 2000 (Microsoft Corporation, Redmond, WA) and graphed and analyzed in KaleidaGraph, version 4.02 (Synergy Software, Reading, PA). Global adjustments to contrast and brightness of Western blots were performed using Photoshop CS2 (Adobe Systems Incorporated, San Jose, CA).

RESULTS

The P_II homolog GlnJ associates with AmtB₁-containing membranes in vivo.

Previously, we observed that some P_II was sequestered at the membrane in R. rubrum under conditions of nitrogen sufficiency and that membrane localization of P_II was absolutely dependent on the presence of AmtB₁ (53). However, the identity of the sequestered P_II was unclear. It has been predicted that the P_II homolog whose encoding gene is in close proximity to amtB will interact with AmtB (46). This hypothesis suggests that, since glnJ is in an operon with amtB₁ in R. rubrum, the proteins expressed by these genes are likely to interact. To test this possibility, we used a glnB glnK strain (UR757) in which GlnJ is the only P_II homolog produced. This strain also produced the AmtB homolog AmtB₁, but not the other AmtB homolog in R. rubrum, AmtB₂, since the glnK::aacC1 insertion is polar on the downstream amtB₂ gene. We grew this strain in nitrogen-limiting MG medium to early exponential phase anaerobically in the presence of light; the glnJ amtB₁ operon is highly expressed under these culture conditions. However, GlnJ accumulates in a uridylylated form under nitrogen-limiting conditions. Therefore, the culture was then shifted to darkness for 20 min, which leads to partial deuridylylation of GlnJ. The cells were then separated into cytoplasmic and membrane fractions and analyzed by Western blotting. Equal volumes of each fraction, concentrated to the same degree, were loaded onto the tricine sodium dodecyl sulfate-polyacrylamide gel electrophoresis gel so that the relative amount of P_II protein in cell extracts was measured.

As expected, in Fig. 1 both an upper uridylylated GlnJ band and a lower nonuridylylated GlnJ band are visible in whole-cell extracts of the glnB glnK mutant UR757. Only nonuridylylated GlnJ is present in the membrane extract, while uridylylated GlnJ predominates in the cytoplasmic extract. This is in agreement with our earlier observation that nonuridylylated P_II in wild-type extracts preferentially interacts with an AmtB₁-containing membrane compared to uridylylated P_II, as well as prior observations for E. coli by others (9, 53). The proportion of total P_II in membrane extract versus cytoplasmic extract was also similar to the proportions we found in wild-type extracts. It is also consistent with recent structural model predictions that indicate that an uridylyl group on Tyr51 would sterically hinder interaction between P_II and AmtB in E. coli (8, 17).

FIG. 1.

Cellular localization of R. rubrum GlnJ (and GlnB) in cell extracts. Shown are Western blots of whole-cell, cytoplasmic, and membrane extracts of photosynthetically grown R. rubrum UR2 (wild type) and UR757 (glnB glnK) with antibodies against R. rubrum GlnB and GlnJ. Whole-cell extracts (W), cytoplasmic extracts (C), and membrane extracts (M) were loaded as indicated. (Upper two panels) Western blots of GlnJ in extracts of UR2 (top) and UR757 (bottom) extracts from cells 20 min after a shift from light to darkness. UR2 extracts contain all three P_II proteins, while UR757 extracts contain only GlnJ. Both P_II-UMP (upper band) and P_II (lower band) are present in whole-cell extract, while only unmodified P_II is present in membrane extract. (Lower two panels) Western blots of GlnJ (top) and GlnB (bottom) in UR2 extracts 1 h after a shift from light to darkness. Deuridylylation of P_II is almost complete. The majority of GlnJ signal is found in membrane extract, while more GlnB signal is present in cytoplasmic extract than in membrane extract. Both GlnB and GlnJ antibodies cross-react with other P_II proteins but are more sensitive to the P_II homolog they were raised against.

Next, we sought to determine whether GlnJ is the primary P_II homolog found associated with the membrane in wild-type R. rubrum. Although GlnJ associated strongly with the membrane in vivo in the absence of GlnB and GlnK, there may be a mixture of P_II homologs at the membrane in wild-type cells. We grew UR2 (wild-type) cells photosynthetically but moved cells to darkness for 1 h before harvesting, allowing nearly complete deuridylylation of P_II. We then blotted equal volumes of cell extracts against anti-GlnB or anti-GlnJ antibodies (Fig. 1). Although there is substantial cross-reactivity between antibodies to different P_II homologs, anti-GlnB antibodies detected GlnB about 10-fold more strongly than GlnJ, while anti-GlnJ antibodies detected GlnJ about 3-fold better than GlnB (data not shown). When anti-GlnJ antibodies are used, the majority of P_II signal is found in membrane extract, suggesting that most of the GlnJ is found at the membrane under these conditions. In contrast, with anti-GlnB antibodies there is slightly more P_II signal in the cytoplasmic extract than in membrane extract, suggesting that GlnB may be present in both cytoplasmic and membrane extracts. Thus, both GlnB and GlnJ can associate with AmtB₁ in the membrane under these conditions but GlnJ appears to be preferentially bound. To avoid complications from a potentially mixed P_II population, the following experiments on P_II-AmtB interaction used the glnB glnK strain as a source of purified AmtB₁-containing membrane, where GlnJ is the only P_II present in these extracts.

ATP and 2-KG affect the stability of GlnJ interaction with the AmtB₁-containing membrane.

Proposed models of P_II interaction with AmtB suggest that all three sites of uridylylation, the Tyr51 residue of each monomer in the P_II trimer, would be inaccessible for modification by GlnD when GlnJ is bound to AmtB₁, and recent crystal structures of P_II-AmtB complexes support this hypothesis (2, 8, 17). Thus, although P_II uridylylation could prevent association between P_II and AmtB, the uridylylation process should not be able to cause complex disassociation. It seemed reasonable that small-molecule binding, not uridylylation, would be the primary regulator of interaction between these two proteins. This idea was supported by results obtained using purified E. coli GlnK-AmtB complexes, where 3.5 mM ATP and 1 mM 2-KG led to complex disassociation (11). However, these results were obtained with AmtB dissolved in detergent rather than in its native membrane environment. We therefore studied this issue in the physiologically relevant in vivo environment of purified membrane extracts containing AmtB₁ and bound GlnJ; membrane extracts were prepared by sonication and therefore consisted primarily of inside-out vesicles exposing the GlnJ-AmtB₁ complex to solvent (42).

As seen in Fig. 2, when purified membrane extract from an R. rubrum glnB glnK mutant was incubated in buffer lacking exogenous ATP or 2-KG, almost all GlnJ remained associated with the membrane. In the presence of 3 mM ATP and 10 mM 2-KG, all GlnJ was released from the membrane and was found solely in the supernatant. To clarify whether ATP or 2-KG affected the stability of the GlnJ-AmtB₁ complex, either 3 mM ATP or 10 mM 2-KG was added separately. While 3 mM ATP alone had no effect, 10 mM 2-KG alone was able to efficiently remove GlnJ from the membrane. Increasing ATP levels up to 15 mM ATP had very little effect on the stability of GlnJ binding (data not shown). In contrast, in the absence of ATP lower levels of 2-KG (1 mM) caused only a small amount of GlnJ to disassociate (Fig. 2).

FIG. 2.

Effects of small molecules on the stability of GlnJ-AmtB₁ interaction in vitro. Shown are Western blots of GlnJ on soluble and membrane-bound fractions of R. rubrum UR757 membrane extract. Small molecules were added to membrane extract containing AmtB₁-bound GlnJ at the concentrations listed, as described for the P_II association and disassociation assays in Materials and Methods. Soluble and membrane-bound proteins were separated by ultracentrifugation and concentrated by trichloroacetic acid precipitation to equal levels compared to the original membrane extract. Proteins from the supernatant (S) were loaded in the left lane of each panel, and membrane-associated proteins (M) were loaded in the right lane. High levels of 2-KG (10 mM) are sufficient to disrupt GlnJ-AmtB₁ interaction, while low levels of 2-KG (1 mM) are not, except in the presence of ATP. Substrate analogues ADP and oxaloacetic acid (OAA) do not substitute for ATP and 2-KG, respectively.

The binding of ATP and 2-KG to P_II is known to be cooperative in E. coli, and this is likely to be true in R. rubrum as well (26, 36). Given this, we next tested a combination of 3 mM ATP and 1 mM 2-KG. If the presence of ATP increased the affinity of GlnJ for 2-KG, we would expect to see larger amounts of GlnJ released from the membrane than we see when 2-KG is added in the absence of ATP. Supporting this notion, addition of both ATP and 2-KG resulted in significant release of GlnJ from the membrane, while either ATP or 2-KG alone at the same concentration had little or no effect (Fig. 2). To confirm that this was a specific effect of ATP and 2-KG, substrate analogues ADP and oxaloacetic acid were added at similar concentrations. However, neither molecule was able to act as a substitute, indicating that the ability to disrupt the GlnJ-AmtB₁ complex is likely to be specific to ATP and 2-KG. The addition of up to 10 mM NH₄Cl did not significantly affect these results, suggesting there is no direct sensing of ammonia by the P_II-AmtB₁ complex (data not shown). These results suggest that 2-KG binding to P_II may directly cause disruption of GlnJ-AmtB₁ interactions in R. rubrum, while ATP binding to P_II may act indirectly by lowering the amount of 2-KG required for destabilization.

The ratio of ATP to ADP affects the stability of the GlnJ-AmtB₁ complex.

Recent crystal structures show ADP bound to P_II in the E. coli P_II-AmtB protein complex (17). Although the absolute level of ATP is thought to be the most important indicator of cellular energy status sensed by P_II, this result suggested that ADP levels could also be an important factor in controlling the interaction between GlnJ and AmtB₁ in R. rubrum. Specifically, we hypothesized that ADP might be able to compete with ATP for binding sites on P_II, thus lowering the effective concentration of available ATP. This would lead to an increase in the proportion of GlnJ bound to AmtB₁ in our in vitro assay using an isolated R. rubrum membrane. If this were true, it would suggest that P_II homologs may be able to sense the ratio of ATP to ADP as well as the absolute level of ATP in the cell as a measure of energy status.

We tested the stability of GlnJ binding to the AmtB₁-containing membrane as in the previous section (Fig. 3A). In the presence of 500 μM 2-KG alone, there was little disassociation of GlnJ from the membrane compared to the 0 μM 2-KG control sample. As expected, a large fraction of GlnJ disassociated when 500 μM ATP was added to the reaction mixture in the presence of 500 μM 2-KG. We then increased ADP from 0 mM to 2.5 mM while keeping 2-KG and ATP levels constant and assayed the amount of GlnJ found at the membrane. There was no difference in levels of GlnJ association with the membrane when 250 μM ADP was added, but when 500 μM, 1 mM, or 2.5 mM ADP was added, an increasing amount of GlnJ was bound to the AmtB₁-containing membrane. Finally, the presence of 2.5 mM ADP had no effect on GlnJ membrane association in the absence of ATP. This demonstrates that ADP is able to interfere with the ability of ATP to disrupt GlnJ-AmtB₁ interaction in the presence of 2-KG, presumably by direct competition for the three primary nucleotide binding sites on the P_II trimer.

FIG. 3.

Effects of ADP/ATP ratio or AMP-PNP on GlnJ-AmtB₁ complex stability in vitro. Shown are Western blots developed with antibody against GlnJ on soluble (S) and membrane-bound (M) fractions of R. rubrum UR757 membrane extract. Small molecules added to each reaction are listed. Samples were treated as described for Fig. 2. (A) ADP stabilizes the GlnJ-AmtB₁ complex in the presence of ATP and 2-KG, with higher levels of ADP providing more stabilization. (B) AMP-PNP is able to substitute for ATP in destabilizing GlnJ-AmtB₁ interaction in the presence of 2-KG, indicating that ATP hydrolysis is not required for disassociation of the complex. Controls for the 0.5 mM 2-KG, 0.5 mM AMP-PMP condition are shown in panel A and were performed at the same time as reactions shown in panel B.

Finally, we examined whether ATP hydrolysis was necessary for the release of GlnJ from AmtB₁ by adding equal amounts of either ATP or a nonhydrolyzable analog, AMP-PNP, to isolated membranes. We chose to test a high-2-KG condition, where a relatively small amount of ATP would be required for release of GlnJ, and a low-2-KG condition, where a large amount of ATP would cause release of GlnJ. After separation of soluble proteins from membrane-bound proteins, we compared the proportions of GlnJ remaining at the membrane under both conditions. As seen in Fig. 3B, AMP-PNP appears to be an effective if imperfect substitute for ATP in the release of GlnJ from AmtB₁-containing membranes. With high 2-KG (2 mM) present, almost all GlnJ was released when 100 μM of either ATP or AMP-PNP was added in the presence of 2 mM 2-KG (Fig. 3B). Under the low-2-KG condition (500 μM), ATP again caused nearly complete disassociation of GlnJ (Fig. 3A) while AMP-PNP caused partial disassociation (Fig. 3B), perhaps due to a slightly lower binding affinity of GlnJ for AMP-PNP compared to ATP, but this was not explored further. We conclude that, even if ATP hydrolysis by P_II can occur, it does not appear to be required for disruption of the GlnJ-AmtB₁ complex in R. rubrum.

GlnB and GlnK bind the AmtB₁-containing membrane less efficiently than does GlnJ in vitro.

Although we had shown that GlnJ is capable of interacting with AmtB₁ both in vivo and in vitro, it was unclear whether other P_II homologs in R. rubrum could bind AmtB₁ as strongly as GlnJ could. If the GlnJ-AmtB₁ complex could be reconstituted in vitro, we would be able to test the ability of GlnB or GlnK to also complex with AmtB₁ in vitro. We took advantage of our observation that addition of high levels of ATP and 2-KG caused complete release of GlnJ from isolated membranes to prepare AmtB₁-containing membranes stripped of GlnJ. Both GlnJ and small molecules were subsequently washed away with repeated ultracentrifugation steps, producing AmtB₁-containing membrane extract that we expected would be able to rebind GlnJ.

We first tested the ability of washed membranes to rebind GlnJ. Washed membranes treated with a twofold excess of purified GlnJ protein, compared to the original membrane extract, were able to bind approximately 60% of the GlnJ found in extract before treatment with ATP and 2-KG (data not shown). The decrease in GlnJ binding capacity of washed membranes was probably due to membrane loss during the washing and resuspension steps or imperfect resuspension of the membrane after ultracentrifugation, which might mask some AmtB₁ inside membrane particles. However, the important observation remained that a membrane stripped of GlnJ could rebind GlnJ in our buffer conditions. This allowed us to test the ability of the AmtB₁-containing membrane to individually bind each of the three P_II homologs in R. rubrum.

Data shown in Fig. 4 are from representative experiments, all using the same preparation of membrane extract from wild-type cells; different preparations differed slightly in GlnJ binding capacity but showed the same trends. Different amounts (0.21 μg to 10 μg) of R. rubrum GlnJ were added to washed membranes, corresponding to a final P_II concentration of about 11 to 540 nM. The maximal amount of added GlnJ, 10 μg, was 2.5-fold more than the amount of GlnJ found in the isolated membrane and about 4-fold more than the maximal amount of rebound GlnJ. The amounts of GlnJ recovered from soluble and membrane-bound fractions were quantitated by Western blotting. The amount of GlnJ associated with the membrane increased as more GlnJ was added to the membrane extract but approached saturation of the membrane's binding capacity at about 2.5 μg total GlnJ, a concentration of about 135 nM GlnJ (Fig. 4A).

FIG. 4.

In vitro comparison of specificities between AmtB₁ and the P_II homologs. Shown is a quantitative analysis of P_II rebinding to the AmtB₁-containing membrane. P_II remaining in solution (circles) and in the membrane pellet (squares) was measured by quantitative Western blotting. (A) Increasing amounts of GlnB are added to a fixed amount of AmtB₁-containing membrane stripped of P_II. (B) GlnJ titration. (C) GlnK titration. GlnB antibodies were used to probe GlnB, while GlnJ antibodies were used to probe GlnJ and GlnK. Data points are from a representative experiment using the same preparation of membrane extract. Protein was concentrated or diluted such that final amounts of P_II loaded fell within the linear range of protein as determined by using pure protein standards. Repetitions of several different data points suggest an error of less than ±10%.

When either purified GlnB or GlnK was added to membrane extract, the results were very different (Fig. 4B and C). Very little GlnB or GlnK was associated with the membrane, though in each case, the detected amount was higher than that in membrane samples from a strain lacking AmtB₁ (data not shown). At higher levels of added GlnB or GlnK there was no associated increase in membrane binding, unlike the GlnJ results. This suggests that, while both GlnB and GlnK are able to associate specifically with AmtB₁, this interaction is less stable than the GlnJ-AmtB₁ interaction and is lost during our sample processing. This supports the notion that in wild-type R. rubrum AmtB₁ primarily interacts with GlnJ.

ATP and 2-KG act synergistically to promote release of GlnJ from the AmtB₁-containing membrane.

The strong, relatively stable interaction between GlnJ and AmtB₁ in membrane extracts allowed us to further explore the effects of small molecules on the stability of this complex. Specifically, we examined a broad range of both ATP and 2-KG concentrations to estimate the disassociation constants for each small molecule in the presence or absence of the other. After treatment with small molecules, Western blotting quantitated the proportions of GlnJ found associated with the membrane and free in solution. These values were then normalized to the maximal proportion of GlnJ found at the membrane under any condition, as a small amount of GlnJ, about 5%, disassociated from the membrane during treatment under any condition.

We first examined the effects of 2-KG on the stability of the GlnJ-AmtB₁ complex (Fig. 5A). In the absence of ATP, levels of 2-KG under 1 mM had no significant effect on complex stability. Increasing levels of 2-KG resulted in more release of GlnJ; however, even 15 mM 2-KG failed to cause complete disassociation of GlnJ. The K_d of GlnJ-AmtB₁ for 2-KG in the absence of ATP was calculated to be 5,000 ± 430 μM by curve fitting a standard disassociation equation to the data with the software program KaleidaGraph. Next, we examined the effects of 2-KG in the presence of 3 mM ATP. This concentration of ATP was chosen as a physiologically relevant level that may be present under conditions of energy sufficiency (34). ATP dramatically lowered the levels of 2-KG required for destabilization of the GlnJ-AmtB₁ complex. Furthermore, complete disassociation of GlnJ was seen only in the presence of ATP. The K_d of the GlnJ-AmtB₁ complex for 2-KG in the presence of 3 mM ATP was approximately 340 ± 30 μM. This synergistic effect of 2-KG and ATP is also observed in the model of cooperative binding found for E. coli and Synechococcus sp. P_II proteins (15, 26, 35). R. rubrum P_II also demonstrates cooperative binding of ATP and 2-KG in in vitro modification experiments using R. rubrum GlnD and GlnB proteins (data not shown).

FIG. 5.

2-KG and ATP disrupt GlnJ-AmtB₁ interaction in a dose-dependent fashion. Shown is a quantitative analysis of GlnJ binding to the AmtB₁-containing membrane at increasing ATP or 2-KG concentrations. (A) Various amounts of ATP were added to membrane extract isolated from UR757 while 2-KG levels were kept constant at 0 μM or 500 μM. (B) 2-KG was added in various amounts to membrane extract while ATP levels were kept constant at 0 mM or 3 mM. (C) The AmtB₁-containing membrane was stripped of GlnJ by incubation at high ATP and 2-KG and resuspended in buffer lacking ATP and 2-KG. Various amounts of 2-KG and either 0 mM or 3 mM ATP were added before adding 2 μg GlnJ. The amounts of GlnJ in solution and at the membrane were quantified by Western blotting. Error bars represent the standard deviations of multiple independent samples; data points without error bars are based on a single sample.

Next, we titrated increasing amounts of ATP in the absence of 2-KG or in the presence of 500 μM 2-KG, a level of 2-KG that caused only a small amount of GlnJ release in the absence of ATP (Fig. 5B). In the absence of 2-KG, over 90% of GlnJ remained associated with the membrane even at 15 mM ATP, the highest level of ATP tested. This indicates that either ATP alone is unable to interact with the GlnJ-AmtB₁ complex or ATP binding does not significantly affect the stability of the complex. However, in the presence of 500 μM 2-KG, addition of ATP promoted complex disassociation. Levels of ATP below 100 μM had little effect on stability, and the K_d of the GlnJ-AmtB₁ complex for ATP in the presence of 500 μM 2-KG was 630 ± 65 μM. However, this is not a saturating level of 2-KG; for example, our data in Fig. 3B indicate that, in the presence of 2 mM 2-KG and absence of ADP, even as little as 100 μM ATP was sufficient to release most GlnJ from AmtB₁.

Finally, we tested the effect of small molecules on the ability of the GlnJ-AmtB₁ complex to reform in vitro. The results, shown in Fig. 5C, are almost identical to those in Fig. 5B. This was expected, as both reaction mixtures were incubated for an extended period of time and should have reached equilibrium between bound and unbound GlnJ in the samples. This is also a confirmation that the reconstituted P_II-AmtB₁ complexes behave similarly to those isolated from R. rubrum.

DISCUSSION

Most prokaryotes have one or more P_II homologs, as well as at least one AmtB homolog, yet the question of specificity between these two families of proteins has not been examined in any depth. Our results indicate that in R. rubrum the P_II homolog GlnJ, which is transcribed from the glnJ amtB₁ operon, interacts more strongly with AmtB₁ than do the other two P_II homologs present, GlnB and GlnK. This is direct evidence supporting the notion that, if the genes encoding the P_II and AmtB proteins are in close proximity to each other, the proteins can interact in vivo (46). However, both GlnB and GlnK are able to interact with AmtB₁ in the absence of GlnJ in R. rubrum. This verifies our previous experiments that indirectly suggested that interaction between GlnB and AmtB₁ also occurs in vivo and that this interaction can have physiological significance (52, 53). These results are consistent with our previous observations that in R. rubrum AmtB₁ can interact with all three P_II homologs in vivo, but with various degrees of specificity (53).

Why would there be specificity in interaction between P_II and AmtB proteins? A known difference between P_II homologs in R. rubrum is that only GlnB is able to activate NifA, which activates transcription of genes involved in nitrogen fixation; GlnJ and GlnK are unable to activate NifA. In addition, while the glnJ amtB₁ operon is expressed only under nitrogen-limiting conditions, glnB is expressed constitutively, although expression is also upregulated under nitrogen-limiting conditions (3, 54). We have found that, when R. rubrum experiences an increase in nitrogen levels in the cell, GlnJ may be the preferred P_II homolog sequestered to the membrane. This would leave a relatively large proportion of GlnB free in the cytoplasm, which might then reactivate NifA upon subsequent decrease in nitrogen levels. Therefore, one possible reason for specificity between P_II and AmtB is to regulate which functions the remaining free P_II is capable of performing. Additionally, if the level of 2-KG required for uridylylation of GlnB by R. rubrum GlnD is less than that required for disassociation of GlnJ from AmtB₁, then GlnJ would be modified after GlnB in vivo; this would again allow a specific uridylylated P_II to function despite the presence of multiple P_II proteins in the cell. However, the minimal level of 2-KG required for efficient modification of R. rubrum P_II in vivo is not yet known (25).

In organisms such as E. coli, the two P_II homologs GlnB and GlnK appear to be functionally interchangeable when expressed at equal levels (4). In such a case, P_II binding to AmtB still has the effective consequence of transiently and reversibly sequestering the majority of cytoplasmic P_II to the membrane. This would enable the cell to more rapidly recover from a transient increase in nitrogen status by quickly increasing the concentration of active P_II in the cell (6). While GlnK is preferentially bound to AmtB at the membrane over GlnB, there is no clear requirement for this specificity of interaction between GlnK and AmtB in E. coli (9).

The small-molecule requirements we observed for release of GlnJ from AmtB₁ in R. rubrum are similar to those observed for E. coli: both ATP and 2-KG are required for GlnJ release when only 1 mM of 2-KG is added to high levels of ATP (11). However, 2-KG alone led to complex disassociation at millimolar concentrations, and high ATP levels neither stabilized nor destabilized the complex, whereas ATP was required for complex stability in E. coli. Higher levels of 2-KG were not tested in E. coli, so it is unclear if 2-KG alone would have effects on complex stability in this organism (11). In contrast, ATP alone was sufficient to disrupt the P_II-AmtB complex in Bacillus subtilis and the presence of 2-KG had almost no effect on complex stability (18). Low levels of ATP (100 μM) in the absence of 2-KG were also sufficient to disrupt a purified GlnK1-Amt1 complex from M. jannaschii (50).

Both allosteric effectors of P_II, ATP and 2-KG, affected the stability of the GlnJ-AmtB₁ complex in R. rubrum. Overall, the K_d values obtained for both ATP and 2-KG were significantly higher than values others have observed for binding of one ATP or one 2-KG molecule in E. coli GlnB (24). In vitro experiments with purified R. rubrum P_II homologs and purified E. coli GlnD suggest that the K_d of R. rubrum P_II for 2-KG is similar to that of E. coli (our unpublished results). The large increase in measured binding constants for 2-KG and ATP could be attributed to two possibilities: either multiple molecules of 2-KG and ATP must bind the GlnJ-AmtB₁ complex for it to disassociate or the K_d of GlnJ for 2-KG is dramatically altered when GlnJ is bound to AmtB₁. Our results favor the former model, as the calculated binding constants were similar whether GlnJ was already bound to the AmtB₁-containing membrane or whether it was added to a membrane stripped of P_II. However, while we predict that the binding of one 2-KG molecule has little effect on GlnJ-AmtB₁ complex stability, it is still unclear whether only two or all three binding sites for 2-KG on GlnJ must be filled to allow release of GlnJ from AmtB₁. It has been suggested that an intermediate state of 2-KG binding has physiological relevance in E. coli, where P_II with three ATP molecules and one 2-KG molecule bound is likely to interact with NtrB (23, 26, 35). In addition, the ability of high concentrations of 2-KG and ATP to prevent interaction between P_II and AmtB proteins is the simplest explanation for the recent observation for Rhodobacter capsulatus that fully unmodified GlnK fails to be sequestered in the membrane in a glnB mutant despite the presence of AmtB protein (47).

Our results indicate that competition between ADP and ATP for binding sites on P_II can be a relevant factor in regulating interaction between P_II and AmtB. Other proteins have also been shown to be regulated by the ratio of ATP to ADP such as SpoIIAB in Bacillus subtilis, where one of two potential protein targets is bound according to which nucleotide is in the majority (1). In our system, high levels of ADP are able to stabilize the GlnJ-AmtB₁ complex in the presence of ATP and 2-KG, which would otherwise destabilize the complex. Sensing the ATP to ADP ratio allows yet another means to exert subtle control over the degree of response to energy depletion and nitrogen deficiency. ADP has also recently been found to influence the ability of E. coli P_II to interact with receptors GlnE and NtrB and the signal transduction protein GlnD; in each case, ADP appears to signal energy limitation and limits the degree of response by P_II to nitrogen limitation (P. Jiang and A. J. Ninfa, personal communication).

Although effects on the stability of the AmtB₁-GlnJ complex of R. rubrum were seen only when the ratio of ADP to ATP was greater than 1 in our experiments, these data still support the idea of energy charge being an important regulatory input of P_II protein function. Adenylate energy charge is determined by formula ([ATP] + 1/2[ADP])/([ATP] + [ADP] + [AMP]) and has commonly been estimated to be near 0.85 during logarithmic growth in bacteria, implying there is much less ADP than ATP in these cells (7). However, many reports of energy charge give a ratio of ADP to ATP that is significantly higher. Specifically, in R. rubrum under conditions identical to those used during collection of cells for determining the localization of P_II to the membrane, the estimated energy charge is only about 0.66 and the ratio of ADP to ATP is approximately 1 within the limits of experimental error (38). Also, adenylate energy charge has been measured at significantly lower levels in both log and stationary-phase growth conditions in multiple bacterial species, where the ratio of ADP to ATP is as much as 3 (5, 12, 31, 37). Given that nonlogarithmic growth conditions are more likely to be relevant in the wild, we predict that energy sensing by P_II is an important function of this protein.

Uridylylation of GlnJ in R. rubrum provides an additional level of regulation of interaction between GlnJ and AmtB₁, above and beyond small-molecule concentrations. Only nonuridylylated P_II interacts with AmtB, and recent crystal structures show that modification of P_II would sterically block interaction. Therefore, there are two effective concentrations of P_II in the cell: total P_II and that fraction of P_II that is unmodified and therefore available for interaction with AmtB₁. Even if the concentrations of small molecules in the cell are permissive of interaction between P_II and AmtB, there necessarily is a delay while P_II modifications are removed before interaction can take place.

In summary, we have demonstrated that AmtB₁ of R. rubrum shows some specificity for binding the P_II homolog GlnJ over the other two P_II homologs present, GlnB and GlnK, and that only nonuridylylated GlnJ interacts strongly with AmtB₁. The binding of ATP and 2-KG to GlnJ specifically disrupts the interaction between GlnJ and AmtB₁, but ATP hydrolysis is not required for this process. The level of 2-KG required suggests that multiple 2-KG molecules may need to bind GlnJ for complex disassociation in R. rubrum. Finally, ADP stabilizes the GlnJ-AmtB₁ complex in the presence of ATP and 2-KG, suggesting that the ratio of ADP to ATP may be sensed by P_II as a measure of cellular energy status for at least some P_II functions.