Identification of critical residues in GlnB for its activation of NifA activity in the photosynthetic bacterium Rhodospirillum rubrum

Abstract

The P_II regulatory protein family is unusually widely distributed, being found in all three domains of life. Three P_II homologs called GlnB, GlnK, and GlnJ have been identified in the photosynthetic bacterium Rhodospirillum rubrum. These have roles in at least four distinct functions, one of which is activation of the nitrogen fixation-specific regulatory protein NifA. The activation of NifA requires only the covalently modified (uridylylated) form of GlnB. GlnK and GlnJ are not involved. However, the basis of specificity for different P_II homologs in different processes is poorly understood. We examined this specificity by altering GlnJ to support NifA activation. A small number of amino acid substitutions in GlnJ were important for this ability. Two (affecting residues 45 and 54) are in a loop called the T-loop, which contains the site of uridylylation and is believed to be very important for contacts with other proteins, but other critical residues lie in the C terminus (residues 95–97 and 109–112) and near the N terminus (residues 3–5 and 17). Because many of the residues important for P_II–NifA interaction lie far from the T-loop in the known x-ray crystal structures of P_II proteins, our results lead to the hypothesis that the T-loop of GlnB is flexible enough to come into proximity with both the C- and N-terminal regions of the protein to bind NifA. Finally, the results show that the level of P_II accumulation is also an important factor for NifA activation.

The members of the P_II protein family are among the most broadly distributed regulatory proteins and are found in all three domains of life. In Proteobacteria, this family is biologically important for the critical roles that it has been shown to play in all aspects of nitrogen regulation and the balancing of carbon and nitrogen utilization. Multiple P_II homologs (typically termed GlnB and GlnK) have been found in many bacteria and, whereas many possess similar functions, there are distinct functions as well (1).

The roles of P_II proteins have been characterized in some detail in Escherichia coli, Salmonella enterica, and Klebsiella pneumoniae, in which two P_II homologs, GlnB and GlnK, have been identified (1–3). Although glnB is constitutively expressed in the cell and glnK is only expressed under N-limiting conditions, these two P_II homologs have some similar properties (4, 5). Both proteins are subject to reversible uridylylation by the bifunctional uridylyltransferase/uridylyl-removing enzyme (UTase/UR, gene product of glnD), which is believed to be a primary sensor of intracellular nitrogen status, although curiously GlnK-UMP (the uridylylated form of Glnk) is only poorly deuridylylated by GlnD (6). Both proteins can also interact with NtrB to regulate NtrC activity, and with adenylyltransferase (ATase, the glnE product) to modify glutamine synthetase (GS) (4–6). However, they also display some distinct properties, such as the ability of GlnB-UMP, but not GlnK-UMP, to strongly stimulate the deadenylylation of GS-AMP (the adenylylated form of GS) (7). In K. pneumoniae, GlnK, but not GlnB, regulates the interaction of NifL and NifA to control the expression of the nif operons (8, 9).

The x-ray crystal structures of GlnB and GlnK from E. coli have been solved (10, 11). The monomer structures of GlnB and GlnK are very similar and contain two α-helices and six β-strands, connected by three loops. A largest loop (T-loop) stretches from residues 37 to 55 and contains the uridylylation site (Tyr-51). A small B-loop is between residues 82 and 88, and the C-loop is located at the C terminus from residues 102 to 105 (Fig. 1). In a native trimer form, the C-loop from one subunit lies near the T-loop and B-loop of another adjacent subunit, creating clefts (Fig. 2). These three clefts were believed to be important for P_II to interact with different targets (11, 12).

Fig. 1.

Alignment of the deduced amino acid sequences of R. rubrum GlnB, GlnJ, and GlnK and K. pneumoniae GlnB and GlnK. Conserved residues are shaded. Three loops are indicated by the lines. The location of two restriction enzyme sites (SacII and BglII) from the DNA sequence of glnB and glnJ are also indicated, and these sites were used to create GlnB/GlnJ chimeras. Asterisks indicate residues tested in GlnJ.

Fig. 2.

Location of critical residues on the R. rubrum GlnB trimer for interaction with NifA. The structure is based on the published structure of E. coli GlnB (kindly provided by D. Ollis, Australian National University, Canberra). Each monomer is shown in a different color. Residues 3–5 are labeled with purple, residue 17 is yellow, residue 45 is pink, residue 54 is blue, residues 95–97 are red, and residues 109–112 are green. Single T-loop (as well as Tyr-51), B-loop, and C-loop are indicated. The clefts referred to in the text lie between adjacent B, C, and T loops. Relative to A, B is rotated 60° toward the reader from the top of the figure.

Three P_II homologs have been identified in the photosynthetic, nitrogen-fixing bacterium Rhodospirillum rubrum and assigned the names GlnB, GlnK, and GlnJ, based on sequence similarities and other information (13). At least four different phenotypes are caused by different combinations of mutations affecting these homologs (13). The first involves the regulation of NifA activity. In R. rubrum, the transcription of nifA is not regulated, but NifA activity is tightly controlled in response to NH₄⁺. GlnB is required for the activation of NifA activity under NH₄⁺-limiting conditions, because nif expression is completely abolished in a glnB deletion mutant (14). A glnB mutant with a Y51F substitution, so that GlnB cannot be uridylylated, also showed low nitrogenase activity (14). These data suggest that in R. rubrum the activity of NifA is regulated positively by GlnB-UMP and that neither GlnK nor GlnJ can activate NifA.

A similar regulation of NifA activity by GlnB has been seen in Azospirillum brasilense, Rhodobacter capsulatus, and Herbaspirillum seropedicae (15–17). However, the mechanism for the activation of NifA activity (whether it is direct or indirect) is unknown in all cases. In A. brasilense and H. seropedicae, the N-terminal domain of NifA appears to have an inhibitory effect on NifA activity, because NifA activity is no longer inhibited by NH₄⁺ when this N-terminal domain is deleted (18, 19). The N-terminal domain also displayed an inhibitory effect on the DNA-binding activity of NifA in vitro (20). It has been suggested that GlnB might bind the N-terminal domain of NifA to prevent its interaction with the central catalytic domain (19). Consistent with this, the interaction between NifA and GlnB has been detected in R. capsulatus by the yeast two-hybrid system (21).

A second role of P_II homologs in R. rubrum is the posttranslational regulation of nitrogenase activity, where either GlnB or GlnJ is necessary for normal regulation (13). A third role is proper adenylylation of glutamine synthetase in response to NH₄⁺ addition, and any of these P_II homologs can support this function in vivo (13). A fourth role involves an interaction with an unknown factor that is necessary for normal growth (13). Strains lacking both glnB and glnJ, irrespective of the presence of glnK, grow poorly on both minimal and rich media.

These results indicate clear biochemical differences among the P_II homologs of R. rubrum, but the structural features that create these differences are unknown. Functional specificity of K. pneumoniae P_II homologs for NifA and NifL interaction has been analyzed, and substitution of residues 43 and 54 allows GlnB of K. pneumoniae to perform the “GlnK function” of interacting with NifL (22). However, this analysis was performed at an elevated level of expression of the altered GlnB that, as we show in this report, might alter the result. Recently, Atkinson et al. (23) reported that exchanges of their promoters could functionally convert GlnB to GlnK or vice versa in E. coli, suggesting that the differences between some P_II homologs are correlated with the timing of expression and level of protein in the cell. We undertook to define the basis of the specificity of the GlnB–NifA interaction in R. rubrum by converting GlnJ to serve the GlnB function at physiologically normal levels of protein accumulation.

Materials and Methods

Growth Conditions and Whole-Cell Nitrogenase Activity Assay. R. rubrum was grown in rich SMN (supplemented malate-NH₄⁺) medium or malate-glutamate medium (MG) as described (13). Whole-cell nitrogenase activity assay has been described (24).

Construction of Plasmids to Express R. rubrum glnB, glnJ, and glnK and K. pneumoniae glnB and glnK from R. rubrum glnB Promoter. To distinguish the genes from the two organisms throughout the rest of this report, we will prefix those from R. rubrum with Rr and those from K. pneumoniae with Kp. For generation of the P_glnB-Rr glnJ fusion, two primers were used to PCR amplify the R. rubrum glnB promoter region. Another set of primers was used to amplify the R. rubrum glnJ structural gene. After amplification, the two PCR products were recovered from the agarose gel with the QIAquick Gel Extraction Kit (Qiagen, Valencia, CA). Because of an overlapping sequence of one primer used in each PCR, these two PCR products could be annealed and amplified again with two primers from both ends containing EcoRV and HindIII sites, respectively, to generate a full-length of P_glnB-glnJ fusion, with P_glnB precisely positioned 5′ of the WT glnJ. The 1.1-kb PCR product was digested with EcoRV and HindIII, and then cloned into pSUP202, yielding pUX453. The insertion was sequenced to confirm the correct fusion. Other fusions, such as P_glnB-Rr glnK, P_glnB-Kp glnK, and P_glnB-Kp glnB were constructed similarly. These PCR fragments were cloned into pSUP202, yielding pUX454, pUX455, and pUX456, respectively. Rr glnB was also PCR-amplified and cloned into pSUP202, yielding pUX533.

To integrate these plasmids into the chromosome of the R. rubrum glnBJ mutant (UR808), EcoRV–HindIII fragments from these plasmids were cloned into pUX610, which contains a 4-kb PvuII–EcoRV fragment of the region 5′ of glnB cloned into pSUP202; this ensures that the single-copy integrants are at the normal glnB chromosomal location. aacC1, encoding gentamicin resistance (Gm^r) from pUCGM (25), was inserted into HindIII site, yielding pUX814, pUX815, pUX816, pUX817, and pUX818. These plasmids were transformed into E. coli S17-1, and then conjugated into R. rubrum glnBJ mutant (UR808) (13) by the method described (26). Sm^r Km^r Gm^r R. rubrum colonies were selected, and whole plasmid was integrated into the chromosome resulting from a single-crossover recombination event. UR808, rather than a glnBJK mutant, was used for technical reasons related to drug resistance. Because GlnK accumulates to very low levels (unpublished data) it is unlikely to perturb the results, and this notion is supported by the various controls.

To clone these genes into a broad host-range plasmid, EcoRV–HindIII fragments from these plasmids were cloned into pRK404, and then transferred into an R. rubrum glnBJK mutant (UR812) (13) by the method described (27).

Because of the spontaneous suppressor (sgn) mutations that occur in glnBJ or glnBJK backgrounds, we always used purified slow-growing UR808 or UR812 as the recipient in transconjugation. We used pRK404 or pUX610 as negative controls to estimate the frequency of sgn mutations after the mating, and it was <1%.

Various DNA Techniques. The MasterPure DNA Purification Kit (Epicentre Technologies, Madison, WI) was used for total DNA isolation from R. rubrum, with a few minor modifications as described (14). Other DNA manipulations were performed by standard methods. DNA sequences were determined with the Big Dye Terminator v. 3.1 Cycle Sequencing Kit (Applied Biosystems) and analyzed with software from DNASTAR (Madison, WI). The QuikChange method (Stratagene) was used to generate most single or double substitutions in the various glnJ mutants. Some multiple substitutions were also constructed by other PCR methods.

Results

The Requirement for GlnB in NifA Activation Reflects Properties of the Protein and Not a Different Level of Protein Accumulation. Because P_II homologs appear to function by protein–protein interactions with specific receptor proteins, the inability of one or another P_II homolog to serve a specific function might actually reflect low abundance in the cell rather than the inability to interact. Recent studies in E. coli showed that the different functions of GlnB and GlnK are correlated with the timing of gene expression and level of protein accumulation (23). We also noticed that all three P_II homologs accumulated at different levels in R. rubrum; GlnJ is the most abundant and GlnK is the least abundant in the cell (unpublished data).

To probe the basis for the specificity of NifA activation for GlnB, we wished to express the various P_II homologs equivalently, at both modest and high levels, and examine their ability to activate NifA. This was done by constructing the fusion of R. rubrum glnK or glnJ with the glnB promoter (see Materials and Methods) and introducing them into R. rubrum strains on either a 10-copy-per-cell pRK404 plasmid or on a pSUP202-derivative suicide plasmid pUX610, such that the construct integrated in a single copy in the chromosome. The recipient strains either lacked glnB and glnJ or genes for all three homologs. glnB and glnK of K. pneumoniae were also analyzed in R. rubrum in the same way. Because of antigenic differences, it was impossible to precisely quantitate accumulation of these homologs or the chimeras described below with the several antibodies we possess. However, such analyses gave no evidence of dramatically different accumulation of any of these proteins.

The ability of these homologs to activate NifA is shown in Table 1 by the level of nitrogenase activity in strains derepressed in MG (minimal malate-glutamate) medium. All of these P_II homologs were expressed from the R. rubrum glnB promoter (P_glnB) on a multicopy plasmid pRK404 in a glnBJK mutant background (UR812). As a positive control, UR1035 (expressing glnB of R. rubrum) showed reduced nitrogenase activity relative to that seen at normal expression levels in WT (UR2) (Table 1). The basis for this effect on nitrogenase activity is unclear and is presumably an indirect effect of overexpression of glnB. UR1031, UR1032, and UR1033 showed no nitrogenase activity, indicating that Rr GlnJ, Rr GlnK, and Kp GlnK are unable to activate NifA, even though they are expressed at the same level as R. rubrum GlnB. UR1034 (expressing Kp glnB) showed a low nitrogenase activity. This result shows that the inability of Rr GlnJ or GlnK to activate NifA is a result of structural differences and not a matter of differences in timing or level of expression as has been reported in E. coli (23).

Table 1. Nitrogenase activity in R. rubrum glnBJK mutants containing P_glnB-Rr glnB, Rr glnJ, Rr glnK, Kp glnK, or Kp glnB on pRK404.

Strain	Gene(s) on pRK404	Nitrogenase activity*
UR2 (WT)	None	800
UR1035	PglnB-Rr glnB	300
UR1031	PglnB-Rr glnJ	<10
UR1032	PglnB-Rr glnK	<10
UR1033	PglnB-Kp glnK	<10
UR1034	PglnB-Kp glnB	70

^*Each unit of nitrogenase activity is expressed as nmol of ethylene produced per h per ml of cells at an optical density at 600 nm of 1, rounded to the nearest 10 units. Each activity value is from at least five replicate assays from different individually grown cultures. The SD is between 5 and 15%

As seen previously, R. rubrum strains lacking both glnB and glnJ grow slowly on all tested media and give rise to faster-growing suppressor mutants (13). It appears that a normal GlnK level is insufficient to support normal growth when the cell lacks both GlnB and GlnJ. In contrast, all strains with glnB, glnJ, or glnK of R. rubrum or glnK of K. pneumoniae on the multicopy plasmid grew only slightly less well than did WT and gave stable colonies. Only UR1034 (expressing Kp glnB) grew significantly slower than other strains, but it grew faster than a strain lacking any P_II homolog and also formed stable colonies. The ability of UR1032 (expressing Rr glnK from the glnB promoter) to grow well in SMN (rich) and MG media indicates that the failure of Rr GlnK to provide normal growth in a glnB glnJ strain is apparently due to poor expression of glnK from its normal promoter. The structural requirements for interaction with the unknown protein that underlies this phenotype are therefore not very restrictive.

When the same expression constructs were placed on the single-copy plasmids and integrated at the glnB chromosomal location in UR808 (glnBJ mutant), as described in Materials and Methods. As a control, UR1054 (expressing Rr glnB) showed high nitrogenase activity, similar to that seen in WT UR2 (Table 1 and 2), so that a low level of GlnK in the cell has little effect on the GlnB–NifA interaction. This finding indicates that decreasing nitrogenase activity in UR1035 (Table 1) is caused by excess GlnB in the cell. However, no nitrogenase activity was seen in strains expressing other P_II homologs, even in UR1058 (expressing Kp glnB) (Table 2). We interpret this to mean that Kp GlnB has a poor affinity for NifA that can be partially overcome by elevated accumulation levels. This implies that levels of expression are critical in the analysis of P_II homolog specificity. We have more evidence below to support this hypothesis.

Table 2. Nitrogenase activity in R. rubrum ΔglnBJ mutants containing P_glnB-Rr glnB, Rr glnJ, Rr glnK, Kp glnK, or Kp glnB integrated into the chromosome.

Strain	Gene(s) integrated into the chromosome	Nitrogenase activity
UR1054	PglnB-Rr glnB	800
UR1055	PglnB-Rr glnJ	<10
UR1056	PglnB-Rr glnK	<10
UR1057	PglnB-Kp glnK	<10
UR1058	PglnB-Kp glnB	<10

As in the case with multicopy plasmids, all strains expressing R. rubrum P_II homologs grew well in SMN and MG media, suggesting that Rr GlnK is able to support normal growth when expressed at an elevated (GlnB-like) level. Unlike the phenomenon seen with multicopy plasmids, UR1058 (expressing Kp glnB) also grew well on SMN plates, indicating that the slow-growth phenotype seen in UR1034 was caused by overexpression of Kp GlnB in the cell.

Construction and Analysis of GlnB/GlnJ Chimeras to Determine Critical Regions in GlnB for the Interaction with NifA. As mentioned in the Introduction, the T-loop has been believed to be the most important region in P_II for its interaction with different targets. Substitutions altering this loop in enteric bacteria affect the interaction of GlnB with GlnD, ATase, NifL, and NtrB (12, 22, 28). Partial deletion of this T-loop (Δ47–53) abolished interaction with all tested targets (12), and this altered P_II also showed no interaction with NtrB in the yeast two-hybrid system (29). However, as shown in Fig. 1, there are relatively few differences in the T-loop region between GlnB of R. rubrum and those homologs that fail to activate NifA. We therefore considered it likely that regions outside of the T-loop might also be critical for this activation, and tested this notion by creating chimeras of GlnB and GlnJ of R. rubrum. Two restriction sites, SacII and BglII, cut the genes approximately into thirds, allowing the creation of six glnB/glnJ chimeras, representing all possible combinations of these portions from the two genes: we term the products of these chimeras GlnBBJ, GlnJJB, GlnBJJ, GlnJBB, GlnBJB, and GlnJBJ, where the letter order indicates the source of each third of the chimeric gene. All of these chimeras are expressed from Rr glnB promoter. The genes for these chimeras were integrated at the glnB chromosomal location in UR808 (glnBJ mutant), as described in Materials and Methods. None of these chimeras were proficient for NifA activation (Table 3), indicating that one or more residues in each third of GlnB are important for NifA activation.

Table 3. Nitrogenase activity in R. rubrum glnBJ mutants containing glnB/J chimeras or altered glnJ integrated into the chromosome.

Strain	Gene product from integrated gene	Nitrogenase activity
UR1054	GlnB (WT)	800
UR1059	GlnBBJ	<20
UR1060	GlnJJB	<20
UR1061	GlnBJJ	<20
UR1062	GlnJBB	<20
UR1063	GlnBJB	<20
UR1064	GlnJBJ	<20
UR1102	GlnJBB-V45L	100
UR1104	GlnJJB-V45L N54D	60
UR1178	GlnJJB-R17K V45L N54D	210
UR1214	GlnJJB-F3–I5 V45L N54D	500
UR1179	GlnJBB-R17K V45L	180
UR1213	GlnJBB-F3–I5 V45L	520
UR1371	GlnJJB-F3–I5 N54D	100
UR1372	GlnJJB-R17K N54D	300
UR1070	GlnBJB-N54D	600
UR1106	GlnBBJ-E109–L112	250
UR1107	GlnBJJ-N54D E109–L112	110
UR1182	GlnBJJ-N54D D95-G97 E109–L112	540
UR1373	GlnBJJ-N54D D95–G97	20
UR1367	GlnJJJ-F3-I5 V45L N54D D95–G97 E109–L112	400
UR1368	GlnJJJ-R17K V45L N54D D95–G97 E109–L112	300

Many more combination of substitutions were created and analyzed than those shown in Tables 3 and 4.

However, different results were seen when these chimeras were cloned into pRK404 and then transferred into R. rubrum glnBJK mutant (UR812) (Table 4). Two chimeras, GlnBBJ (UR1036) and GlnJBB (UR1039), were able to activate NifA. A low nitrogenase activity was also seen in GlnBJB (UR1040) and GlnJBJ (UR1041), indicating that levels of expression affect the analysis. At higher protein levels, the central region of GlnB combined with either end of GlnB is sufficient for NifA activation, although at a normal protein level, all three regions of GlnB are necessary.

Table 4. Nitrogenase activity in R. rubrum glnBJK mutants containing glnB/J chimeras or altered glnJ on multicopy plasmids.

Strain	Gene product from gene on pRK404	Nitrogenase activity
UR1035	GlnB (WT)	300
UR1036	GlnBBJ	600
UR1037	GlnJJB	<20
UR1038	GlnBJJ	<20
UR1039	GlnJBB	400
UR1040	GlnBJB	50
UR1041	GlnJBJ	100
UR1045	GlnJBJ-Q42H	60
UR1046	GlnJBJ-V45L	500
UR1049	GlnJBJ-G21–I25	100
UR1047	GlnBJB-N54D	500
UR1051	GlnBJB-L70–V74	60
UR1052	GlnBJB-T76–T80	60
UR1339	GlnJJB-N54D	500
UR1044	GlnJJJ-N54D	150
UR1087	GlnJJJ-V45L N54D	200
UR1048	GlnJBJ-E109–L112	200
UR1334	GlnJJJ-N54 E109–L112	630
UR1363	GlnJJJ-F3–I5 N54D	430
UR1364	GlnJJJ-R17K N54D	350
UR1340	GlnJJJ-R17K V45L N54D D95–G97 E109–L112	660
UR1376	GlnJJJ-F3–I5 V45L N54D D95–G97 E109–L112	630

Determination of Specific Residues in GlnB That Are Critical for the Activation of NifA. For technical ease, we initially studied chimeras expressed from multicopy plasmids for testing NifA activation. Because the GlnJJB, GlnBJJ, GlnBJB, and GlnJBJ chimeras activate NifA poorly (Table 4), the use of these chimeras made the functional dissection of each section much more tractable. For example, we started with the GlnJBJ chimera and created a number of single and multiple substitutions in either first or last part of the “GlnJ” to make it more similar to the normal GlnB sequence. Three single or multiple substitutions were made in first part of the GlnJ (Fig. 1): the single Q42H substitution and a multiple substitution between residues 21 and 25 (G21–I25) provided little effect on the NifA activation (UR1045 and UR1049 in Table 4), whereas the single V45L substitution provided high nitrogenase activity (UR1046 in Table 4) (5-fold increase by comparison with UR1041).

A similar approach was used with the GlnBJB chimera, where again three single or multiple substitutions were tested in the central region of GlnJ. As shown in Table 4, the N54D substitution (UR1047) provided a high nitrogenase activity, whereas other multiple substitutions (L70–V74 and T76–T80) had little effect. However, a single N54D substitution, or the combination of V54L and N54D, in GlnJ supported only moderate nitrogenase activity, indicating that other residues are important for NifA activation.

In the 3′ third of the gene, replacement of residues 109, 110, and 112 (E109–L112) in the GlnJBJ chimera provided modest nitrogenase activity (UR1048 in Table 4). Surprisingly, higher nitrogenase activity was seen in GlnJ (GlnJJJ) with the combination of N54D and E109–L112 substitutions (UR1334) than in GlnJBJ E109–L112 (UR1048). This finding suggests that some residues in the middle of GlnB actually inhibit NifA activation, and this issue will be discussed later.

We also examined the functionality of a number of these substitutions in a single copy when integrated into the chromosome of a glnBJ mutant background (UR808). Because all chimeras were unable to activate NifA when they were expressed from a single-copy gene, we used chimeras with only a single third of GlnJ for mutagenesis. For example, we started with the GlnJBB chimera and created substitutions in the “GlnJ-third.” As shown in Table 3, the V45L substitution in GlnJBB supports some nitrogenase activity, but G21–I25 and Q42H do not (data not shown), indicates that, unlike multicopy (overexpression) conditions, other residues in the N-terminal region are also important for the NifA interaction. Two other substitutions, R17K and F3–I5, were added to a V45L and N54D of GlnJJB (UR1178 and UR1214) and dramatically increased nitrogenase activity, compared with UR1104 (Table 3). Without V45L, F3–I5 substitutions (UR1371) showed poorer activity, indicating that V45L is important the activity of this F3–I5 substitution. Thus, although either F3–I5 or R17K substitutions, when combined with N54D, support good nitrogenase activity in multicopy (UR1363 and UR1364 in Table 4), the F3–I5 V45L combination is optimal in single copy.

Similar to the result seen with multicopy plasmids, the N54D substitution is the most important one in the central region of GlnB (UR1070 in Table 3); L70–V74 and T76–T80, for example, had little effect (data not shown). This finding is consistent with the previous report of the importance of this residue in T-loop in the specificity of GlnK for the regulation of NifA activity in K. pneumoniae (22).

In contrast to the multicopy case, the E109–L112 substitution only supported a moderate nitrogenase activity, even with N54D substitution (UR1106 and UR1107 in Table 3). This finding suggests that other residues in C-terminal regions are important for the NifA interaction; therefore, another multiple substitution, D95–G97, was constructed. Together with N54D and E109–L112, this substitution supported a high nitrogenase activity in the GlnBJJ chimera (UR1182). However, without E109–L112, D95–G97 was not sufficient for NifA activation (UR1373). These results indicate that the combination of D95–G97 and E109–L112 in this region is sufficient to convert GlnJ to GlnB for the interaction with NifA.

Two combinations of these substitutions from each region were constructed: GlnJJJ-F3–I5 V45L N54D D95–G97 E109–L112 and GlnJJJ-R17K V45L N54D D95–G97 E109–L112. In single copy, these two mutants (UR1367 and UR1368) showed a good nitrogenase activity (≈50% of that seen with WT glnB) (Table 3), indicating that the critical residues for NifA activation have been identified. These glnJ mutants also showed a high nitrogenase activity in multicopy plasmid (UR1340 and UR1376 in Table 4).

Discussion

It should not be surprising that multiple changes are necessary to convert GlnJ to GlnB function. Indeed, it is a bit surprising that so few residues appear to serve as the basis for this specificity. The results are striking when the known structures of P_II homologs are considered (Fig. 2). Based on a variety of evidence, it is a very reasonable hypothesis that P_II homologs exert their various effects through specific protein–protein interactions, so one would reasonably expect that P_II residues critical for a given interaction would lie on a single surface of the protein trimer. Examination of the positions of the residues defined in this report shows that this is apparently not the case. As shown in Fig. 2, residues 45 and 54 lie in the T-loop, which is very far from the important residues in both C terminus and N terminus. The site of uridylylation, which is important for many activities, including the GlnB–NifA interaction, is at the far end of the T-loop, and therefore far from much of the rest of the protein (Fig. 2). This positioning would not be a concern if all critical interaction surfaces lay in the T-loop, but the present data suggest that this is not the case for the NifA interaction. How then can this distant uridylylation affect interactions with other parts of GlnB?

There are a couple of different possibilities to explain this paradox: (i) The NifA contact surface might actually be rather large and effectively wrap around P_II thus contacting the residues we have identified. Although this cannot be dismissed, we note that it appears likely that only the N-terminal domain of NifA is involved in this interaction (19), making this notion rather unattractive. (ii) Alternatively, the T-loop of P_II might itself fold back toward the rest of the protein, thus presenting a rather more compact surface for interaction with other proteins. Such a possibility is at least weakly supported by the observation that the T-loop is highly flexible and, indeed, remains disordered in E. coli GlnK structural analyses for this reason (30).

The relatively large surface of GlnB that is apparently involved in NifA interaction has another implication: because GlnB appears to interact with multiple proteins in the cell, it seems highly likely that there is significant overlap in these different interaction surfaces. This is not to say that many of the residues critical for one interaction are necessarily critical for others, but rather, that it will be challenging to disentangle one interaction without perturbing the others. Indeed, we assume that this is the reason for the curious inhibition of growth that was seen when we overexpressed certain variants (such as Kp glnB); we assumed that we were causing an inappropriately complete interaction of GlnB with some other protein in the cell.

We found the curious result that certain GlnJ residues gave higher nitrogenase activity than did the GlnB residue in a specific chimera, as if the GlnB residue was causing an inhibition of the interaction (UR1048 and UR1334 in Table 4). We imagine several possibilities for this effect. First, it is certain that some regions of GlnB interact with other regions that are separated in the primary sequence. Indeed, salt bridges between the N and C-terminal regions of GlnB have been noted (30). It is therefore likely that some chimeras have disrupted these interactions and might therefore indirectly perturb chimera function. We have not analyzed this in detail here because we do not know the structures of either protein from R. rubrum, nor can we predict with certainty the nature of the structure that actually interacts with NifA. Secondly, although the interaction of GlnB with NifA must be of biologically optimal affinity, it is not likely to be of the highest possible affinity. In other words, the ensemble of interactions over the entire surface of interaction must be appropriate, but individual residues might provide either positive or negative effects. By this hypothesis, certain GlnB residues might actually antagonize the interaction with NifA. A clear example of this phenomenon is with CAP (catabolite activator protein) and its interaction with RNA polymerase, where elimination of certain interfering residues on CAP actually improves interaction (31). Finally, as already suggested, we already know that other proteins interact with GlnB and GlnJ. It is therefore possible that some minor effects are actually indirect and reflect altered interactions with these other proteins, either causing a competition for the chimeric protein or having other physiological effects on the cell. It is also possible that some substitutions interfere with trimerization in some chimera contexts. In fact, with these various possibilities, it is a bit surprising that the analysis was as straightforward as it was.

The results with chimeras expressed at different levels demonstrate the critical importance of controlling this issue in any such analysis. The fact that the level of proteins and timing of expression of homologs in some other organisms is physiologically critical is consistent with this notion. However, this does not preclude an additional level of specificity in different interactions, as we have clearly demonstrated in this particular case.

Finally, there is growing evidence that an additional level of regulation of P_II homologs might be through interactions with the membrane-associated AmtB protein. AmtB has been thought to be an ammonium transporter, but it is also reported to facilitate diffusion of NH₃ gas across the cytoplasmic membrane (32). However, recent studies suggest that AmtB might regulate P_II levels in the cytoplasm by sequestering it on the membrane. In E. coli, membrane-associated AmtB can interact with GlnB and GlnK in response to nitrogen status (33, 34). Similarly, AmtB of R. capsulatus is necessary for the regulation of nitrogenase activity, consistent with a sensing function (35). Such interactions with AmtB require yet another interaction surface on P_II homologs, and differences in this surface would provide another basis for discrimination between different homologs.