Evidence for a Second Interaction between the Regulatory Amino-terminal and Central Output Domains of the Response Regulator NtrC (Nitrogen Regulator I) in Escherichia coli*

Nitrogen limitation in Escherichia coli activates about 100 genes. Their expression requires the response regulator NtrC (also called nitrogen regulator I or NRI). Phosphorylation of the amino-terminal domain (NTD) of NtrC activates the neighboring central domain and leads to transcriptional activation from promoters that require σ54-containing RNA polymerase. The NTD has five β strands alternating with five α helices. Phosphorylation of aspartate 54 has been shown to reposition α helix 3 to β strand 5 (the “3445 face”) within the NTD. To further study the interactions between the amino-terminal and central domains, we isolated strains with alterations in the NTD that were able to grow on a poor nitrogen source in the absence of phosphorylation by the cognate sensor kinase. We identified strains with alterations located in the 3445 face and α helix 5. Both types of alterations stimulated central domain activities. The α helix 5 alterations differed from those in the 3445 face. They did not cause a large scale conformational change in the NTD, which is not necessary for transcriptional activation in these mutants. Yeast two-hybrid analysis indicated that substitutions in both α helix 5 and the 3445 face diminish the interaction between the NTD and the central domain. Our results suggest that α helix 5 of the NTD, in addition to the 3445 face, interacts with the central domain. We present a model of interdomain signal transduction that proposes different functions for α helix 5 and the 3445 face.

pounds, and degrade a few nitrogen-containing compounds (1,2). Intracellular glutamine is the primary effector of the Ntr response (3). With high glutamine (i.e. in nitrogen-rich media) NtrC (also called nitrogen regulator I or NR I ) is unmodified, which prevents expression of Ntr genes. In contrast, low glutamine (i.e. in nitrogen-limited media) leads to phosphorylation of NtrC and activation of Ntr genes (4 -6). For example, NtrC-P activates the glnALG operon, which specifies glutamine synthetase, NtrB (also called nitrogen regulator II or NR II ), and NtrC, respectively (6).
NtrC activates Ntr genes from enhancer sites, which often consist of two adjacent NtrC binding sites located up to 250 base pairs away from the transcription start site (7,8). Unphosphorylated NtrC in solution is a dimer, but DNA-bound NtrC-P forms either a hexamer or octamer (9 -11). The enhancer increases the local concentration of NtrC and serves as a template for oligomerization (12,13). From the enhancer, NtrC interacts with 54 -containing RNA polymerase and hydrolyzes ATP to provide the energy needed to catalyze open complex formation (5, 7, 14 -18).
NtrC is a multidomain response regulator that, like similar proteins, contains conserved receiver and output domains. NtrC also has a carboxyl-terminal domain required for dimerization and binding to DNA. The 124-residue amino-terminal domain (NTD) contains the site of phosphorylation, aspartate 54, and binds Mg 2ϩ , which is essential for phosphorylation and dephosphorylation (19,20). The phosphorylated form of NtrC has a half-life of around 4 min (21,22). In the absence of its cognate sensor kinase, small phosphate-containing molecules (e.g. acetyl phosphate) can phosphorylate NtrC (23). Some of these small phosphodonors may have physiologically relevant functions (23,24).
Structures of the unphosphorylated and phosphorylated receiver domains of NtrC have been determined (25)(26)(27)(28) (Fig. 1). Five parallel ␤ strands alternate with five ␣ helices (25). The solvent-exposed loops that join the units of secondary structure contain not only the site of phosphorylation (between ␤3 and ␣3) but also other conserved residues that are spatially located near the site of phosphorylation and comprise the phosphorylation active site: the ␤1-␣1, ␤4-␣4, and ␤5-␣5 loops. Upon phosphorylation, the region bounded by ␣ helix 3 and ␤ strand 5 is displaced with respect to the remainder of the NTD. This region has been termed the 3445 face (for ␣3, ␤/␣4, and ␤5). A major characteristic of this conformational change is the axial rotation of ␣ helix 4 by about 100°leading to a shift of two amino acids in the helix register and causing the hydrophobic surface of the helix, which is buried when the NTD is unphosphorylated, to become solvent-exposed. Furthermore the loop containing the site of phosphorylation and the loops flanking ␣4 appear to take on a stable conformation not seen in the inactive form of the domain (26). This phosphorylation-induced structural change is responsible for the altered interaction with the central domain that leads to activation of transcription (29,30).
All 54 -dependent activators have a domain homologous to the central domain of NtrC (31). This domain interacts with 54 and couples the energy derived from ATP hydrolysis to formation of open promoter complexes. The central domain of NtrC also contains a component of oligomerization that is required for transcriptional activation (32,33). Oligomerization leads to increased stability of binding to adjacent sites on DNA, which is referred to as cooperative binding. Phosphorylation of the NTD stimulates this oligomerization (13,(32)(33)(34), which, in turn, stimulates ATP hydrolysis. Phosphorylation does not affect activities of the carboxyl-terminal domain: dimerization and binding to a single site on DNA (13,34,35).
Our aim was to analyze the interaction between the NTD and the central domain of NtrC. We isolated and characterized mutants with alterations in the NTD that could activate nitrogen-regulated genes without NtrB, i.e. variants in which the NTD is apparently sending a phosphorylation-independent activation signal to the central domain. Altered residues were found not only in the 3445 face but also in ␣ helix 5. We show that the alterations in ␣ helix 5, unlike those in the 3445 face, do not affect the overall structure of the NTD. However, twohybrid interaction analysis indicated that alterations in helix 5 impair domain-domain interactions. We propose that ␣ helix 5, in addition to the 3445 face, interacts with the central domain and that the function of this interaction may be to inhibit phosphorylation-dependent oligomerization.
Enzyme Assays-Assays for glutamine synthetase and ␤-galactosidase were performed as described previously (37,38). Activity units are nmol of product min Ϫ1 mg of protein Ϫ1 .
The ATPase assay was based on measurement of ADP (39). A 100-l reaction mixture contained 50 mM Tris-HCl (pH 7.5), 100 mM KCl, 10 mM MgCl 2 , 1 mM dithiothreitol, 0.1 mM EDTA, 1.3 mM phosphoenolpyruvate, 35 units of lactate dehydrogenase (rabbit muscle), 50 units of pyruvate kinase (rabbit muscle), 0.2 mM NADH, and 500 nM NtrC. 200 nM NtrB was present to phosphorylate NtrC. When DNA was added, 10 nM supercoiled pVW7, which contains NtrC binding sites 1 and 2 upstream of the glnA promoter, was used (34). The reaction mixtures were incubated for 10 min at 37°C, and ATP (final concentration, 1 mM) was added to start the reaction. The change of NADH concentration was measured spectrophotometrically at 340 nm. Units of ATPase activity are mol of ATP min Ϫ1 mol of NtrC monomer Ϫ1 .
Mutagenesis and Mutant Isolation-The template used for mutagenesis of the DNA coding for the NTD of NtrC was pXY12, which is essentially pBLS1 (24) except for an ApaI site. This site was created by changing DNA for codons 135 and 136 from GGCCCA to GGGCCC, which introduces two silent mutations. PCR-based mutagenesis of the NTD of NtrC was accomplished as described previously except that 0.1 M MnCl 2 and 4 mM MgCl 2 were used (40). The PCR product was digested by restriction enzymes PflMI and ApaI (from codons 17 to 136) and subcloned back into pXY12. The mutagenized plasmid was transformed into strain SN24 and plated on glucose-arginine minimal medium. The colonies were inoculated in liquid overnight cultures, and the plasmids were purified and retransformed into SN24. The transformants were retested for their phenotype. The ntrC alleles were sequenced using the Sequenase kit from U.S. Biochemical Corp.
DNase I Footprinting-The method was basically as described pre-viously (32). A 32 P-labeled EcoRI-PstI fragment (153 bp) from pVW7 was used for binding of NtrC to sites 1 and 2. A 75-l reaction mixture contained 35 mM Tris acetate (pH 7.9), 70 mM K ϩ -acetate, 5 mM Mg 2ϩacetate, 19 mM NH 4 ϩ -acetate, 0.1 mg/ml bovine serum albumin, 0.7 mM dithiothreitol, 3 g of sonicated salmon sperm DNA, 5% glycerol, and 0.5 nM DNA. NtrC concentrations were as shown in the figures. NtrC and DNA were incubated for 10 min at 37°C, and then a predetermined dilution of DNase I was added to the reaction. The reaction was incubated for another 7 min and stopped by adding 37.5 l of a stop buffer containing 8 M NH 4 ϩ -acetate and 250 g/ml yeast tRNA. The mixture was extracted with phenol-chloroform and precipitated with ethanol. The samples were run on an 8% polyacrylamide gel containing 8 M urea. The protected bands were quantified using a PhosphorImager (Amersham Biosciences). The quantitation of occupancy has been described previously (32).
In Vitro Phosphorylation and Dephosphorylation-42.5 l of solution A (40 nM NtrB, 1ϫ transcription buffer (see below), 1 mg/ml bovine serum albumin, and 15 Ci of [␥-32 P]ATP) and 40 l of solution B (4 M of NtrC, 1ϫ transcription buffer) were separately preincubated at 37°C. A 2.5-l aliquot of solution A was mixed with an equal volume of a pH 6.8 stop buffer (10 mM EDTA, 2% SDS, 0.6% Tris-HCl, 10% glycerol, 0.1% dithiothreitol, 0.01% bromphenol blue) and put on ice. The remaining 40 l of solution A was then mixed with solution B. 10-l samples were taken at 0.5 and 2.5 min after mixing. At 3 min after mixing, 20 l of unlabeled ATP in 1ϫ transcription buffer was added to the reaction (giving a final unlabeled ATP concentration of 2.5 mM). Samples were taken at various times and loaded on a 10% polyacrylamide gel containing SDS without heating. The gel was placed in a PhosphorImager cassette, and the bands were quantified using a Phos-phorImager program (Amersham Biosciences).
In Vitro Transcription-The basic procedures have been described previously (42). In the first step, a 25-l mixture containing 1ϫ transcription buffer (50 mM Tris acetate, pH 8, 100 mM K ϩ -acetate, 8 mM Mg 2ϩ -acetate, 27 mM NH 4 ϩ -acetate, 3.5% polyethylene glycol, 1 mM dithiothreitol); either supercoiled plasmid pTH8, which carries the glnA promoter (43), or pAK10, which carries the ast promoter (8); 1 unit of RNase inhibitor (Eppendorf-5 Prime, Inc.); E. coli RNA core polymerase (Epicenter Technologies); and 54 protein were incubated for 5 min at 37°C. An NtrC variant (1 l) was then added. When NtrC was phosphorylated, 1 l of NtrB was added, and the mixture was incubated for 5 min at 37°C. To permit the formation of open complexes, 2 l of 50 mM ATP was added to the transcription buffer, and the mixture was incubated for 10 min. For the final step, a 7-l mixture containing 500 g of heparin; a 1.8 mM concentration of GTP, CTP, and UTP; and 10 Ci of [␣-32 P]UTP in transcription buffer was added into the reaction. The final concentrations of components were 10 nM DNA, 100 nM core RNA polymerase, 300 nM 54 , 100 nM NtrC for reactions with NtrB or 300 nM NtrC for reactions without NtrB, 100 nM NtrB (when added), 4 mM ATP, 500 M GTP, 500 M CTP, and 100 M UTP. After 10 min, the 25-l reaction was stopped by adding 25 l of a stop buffer containing 50 mM EDTA and 100 g/ml yeast tRNA. The sample was extracted with phenol, and 40 l was mixed with 15 l of a loading dye mixture containing 95% formamide, 20 mM EDTA, 0.05% bromphenol blue, and 0.05% xylene cyanol FF. The sample was heated at 90°C for 3 min, and 30 l was subjected to electrophoresis in a 5% polyacrylamide, 8 M urea gel.
Expression and Purification of Histidine-tagged NTD of NtrC for NMR-Plasmid pACH6 is a derivative of pRSET-A (New England Biolabs) that expresses the NTD of NtrC (residues 1-140) with a hexahistidine tag under the control of a T7 promoter. E. coli strain BL21 was transformed with pACH6 expressing wild-type or variant NTDs, and the proteins were overexpressed essentially as described previously (32). Cell pellets from overexpression were disrupted by sonication, and the purification protocol was as described for batch binding purification under native conditions (Qiagen "The QIAexpressionist" Handbook) with the following exception. To provide a higher stringency for binding histidine-tagged protein, the lysis and wash buffers contained 20 and 50 mM imidazole, respectively. Proteins were concentrated (Centricon system) and dialyzed into a 50 mM phosphate buffer, pH 6.75 that contained 150 mM NaCl.
Expression and Purification of Native NTD of NtrC for NMR-Plasmid pACH12 codes for the NTD of NtrC (residues 1-124) fused to the Saccharomyces cerevisiae vacuolar ATPase subunit (VMA) intein from pTYB11 under T7 control (Impact CN system, New England Biolabs). Autocatalytic cleavage at the junction site of NtrC and the intein allows for purification of the native 124-residue NTD. Overexpression was performed in E. coli strain ER2566 from New England Biolabs. Over-expression and purification were as described for the New England Biolabs Impact CN system.
NMR Spectroscopy-Proteins were purified as described above, and NMR spectroscopy was carried out on a 500-MHz Varian Unity-INOVA spectrometer with z axis gradient in the Department of Chemistry at The University of Texas at Dallas. 1 H-15 N heteronuclear single quantum correlation (HSQC) experiments used a sensitivity-enhanced pulse sequence (44) and were performed at 25°C with sweep widths of 6100 Hz for 1 H and 1400 Hz for 15 N. All HSQC experiments were collected with 64 complex points in the indirect dimension.
For the plasmids coding for the NTD of NtrC fused to the GAL4 DNA-binding domain, the plasmid pT7-NcoI-NT was used. pT7-NcoI-NT is the derivative of pCB5, which contains an ntrC gene with a deletion of the central domain coding region under the control of a phage T7 promoter (32). An NheI linker (CTAGCTAGCTAG), which contains a stop codon in all three reading frames, was inserted at an engineered SalI site at codon 142 of the glnG gene in pCB5. A unique NcoI site was also created at the translational start site of the glnG gene. A 0.8-kb NcoI-BamHI fragment of pT7-NcoI-NT, which codes for an NtrC with a deletion of the central domain, with stop codons at the end of the NTD was then ligated into the polylinker site of pAS1-CYH2, producing pAS1-CYH2-NTA. This plasmid contains the gene encoding a fusion of the DNA-binding domain of GAL4 to the NTD of NtrC. After construction, the fusion junctions were checked by sequencing.
For the central domain fusion construct, the plasmid pXY15BNN was used. pXY15BNN is a derivative of pXY12 with three linkers inserted at different positions. An NheI linker (with stop codons in three reading frames) was inserted at a ScaI site within glnG gene (codon 387). An NcoI linker was inserted at an ApaI site (codon 136). A BamHI linker was inserted at an SspI site 500 bp downstream of the gene. Then the 1.3-kb NcoI-BamHI fragment of pXY15BNN was cloned into the polylinker site of pACTII to produce the final construct, pAC-TII-CEN, which contains the gene encoding a fusion of the GAL4 activation domain to the central domain of NtrC.
The plasmids were then transformed into yeast strain Y190 as described previously (46). The transformants were plated on synthetic dextrose medium (0.15% yeast nitrogen base without amino acids and ammonium sulfate, 0.5% ammonium sulfate, and 2% dextrose) supplemented with 0.15 mM adenine and 3 mM histidine. The colonies were then tested for growth on the same medium with 30 mM 3-amino-1,2,4triazole instead of histidine. The assay for ␤-galactosidase in yeast was performed as described previously (47).
The plasmids used for the positive controls, pACT4 and pSUG1, were kindly provided by Stephen Johnston (The University of Texas Southwestern Medical Center). ACT4 has previously been found to interact with SUG1. 2

Isolation of Arginine-utilizing Mutants
Phosphorylation of the NTD of NtrC controls central domainmediated oligomerization and ATP hydrolysis, which are required for transcriptional activation. To analyze the communication between the amino-terminal and central domains, we sought mutants with alterations in the NTD that could activate Ntr genes without NtrB. We used a previously described procedure: selection for derivatives of an ntrB mutant that could grow with arginine as sole nitrogen source (33). Arginine utilization requires high expression of the glnALG operon, which codes for the ammonia-assimilating glutamine synthetase, NtrB, and NtrC, respectively, and expression of the astCADBE operon, whose products degrade arginine. glnALG expression requires NtrC phosphorylation by either NtrB-P or acetyl phosphate. In contrast, astCADBE expression requires NtrC phosphorylation by both NtrB-P and acetyl phosphate (24,48). In other words, glnALG is expressed in a glnL mutant, but astCADBE is not.
We selected arginine utilizers from a strain with a deletion of ntrB and ntrC on the chromosome and ntrC expressed on a plasmid. After PCR mutagenesis of ntrC, we subcloned the NTD coding region of NtrC to ensure that the only alterations were within the NTD. We isolated 14 arginine-utilizing mutants with seven different substitutions: R56H, M75I, D86N, D109N, E110K, E116K, and E124K. One derivative had the double substitution D109N/E110K (Fig. 1). Arg-56 is located near the phosphorylation pocket at the carboxyl terminus of ␤3. Met-75 is in the loop between ␣3 and ␤4 on the edge of the ␤ sheet that is opposite the phosphorylation site. Asp-86 is within ␣4. The R56H, M75I, and D86N substitutions are within the 3445 face region that experiences a conformational change during phosphorylation. These alterations probably mimic the structural effects of phosphorylation to some extent as has been shown for the D86N variant (49,50). In contrast, the altered residues in the D109N, E110K, E116K, and E124K variants are not in the 3445 face, are surface-exposed (at least in the isolated NTD), and are not expected to affect the structure of the domain. (Subsequent NMR analysis, which is shown below, provides evidence that supports this expectation.) Therefore, we suspected that the altered residues might interact with the central domain, and we further analyzed these NtrC variants.

Ntr Gene Expression in a Strain with Constitutive NtrC Dephosphorylation
To quantify NtrB-independent gene expression in vivo, we assayed glutamine synthetase activity, a measure of glnA expression. This was done in a glnD mutant, which lacks uridylyltransferase. During nitrogen-limited growth, GlnD uridylylates P II , which results in net NtrC phosphorylation and Ntr gene expression. Without GlnD, unuridylylated P II interacts with NtrB, stimulates NtrC dephosphorylation, and prevents Ntr gene expression. However, NtrB-independent NtrC variants would be expected to activate Ntr genes in a glnD mutant. Consistent with this expectation, the mutant ntrC alleles resulted in 14 -29 times more glutamine synthetase than wildtype ntrC in a glnD background (Table I). These results suggest, but do not prove, phosphorylation-independent activation in vivo. It could be argued that acetyl phosphate phosphoryl-2 S. Johnston, personal communication. ates NtrC, and the P II -NtrB complex may not completely dephosphorylate the resulting NtrC-P. To address this issue, we purified and characterized several NtrC variants.

In Vitro Characterization of NtrC Variants
We purified and characterized the NtrC variants with alterations in ␣ helix 5, NtrC-E110K, NtrC-E116K, and NtrC-D109N/E110K (called NtrC-DE), and compared their properties with those of wild-type NtrC and the previously characterized NtrC-D86N, which has an alteration that affects the 3445 face (49).
Half-life of Phosphorylation-An unusually stable phosphorylation can explain the mutant phenotypes. To test this possibility, we determined the half-life of phosphorylation. Variant and wild-type proteins were incubated with NtrB and [␥-32 P]ATP for 10 min, which is sufficient to achieve steadystate phosphorylation (not shown). All proteins had virtually identical steady-state levels of phosphorylation (not shown). An excess of cold ATP was then added, and loss of label was used to assess phosphorylation stability. The phosphorylation halflives for wild-type NtrC, NtrC-D86N, NtrC-E110K, NtrC-E116K, and NtrC-DE were 7.3, 6.0, 4.9, 8.5, and 6.0 min, respectively. Therefore, an unusually stable phosphorylation cannot account for the mutant phenotype.
Oligomerization-Phosphorylation of the NTD stimulates central domain-mediated oligomerization (13,32,34). We tested whether the variants have enhanced phosphorylationindependent oligomerization. We assessed oligomerization indirectly by determining the binding of NtrC with and without phosphorylation to site 2 of the glnAp2 promoter in the presence of site 1. Binding to adjacent sites is cooperative, and phosphorylation enhances the cooperative binding. Unphosphorylated NtrC-D86N, NtrC-E116K, and NtrC-DE required 2-fold less NtrC for 50% occupancy of site 2 than that of unphosphorylated wild-type NtrC ( Fig. 2A). In contrast, unphosphorylated NtrC-E110K bound as well as wild-type NtrC. The change in binding may not seem dramatic, but then phosphorylation only enhances binding of wild-type NtrC 2-fold (Fig. 2,  A and B). All the variants have phosphorylation-enhanced cooperative binding (Fig. 2B).
It has previously been shown that phosphorylation does not affect binding to a single site, and we confirmed this result (not shown). Since the site of the alteration in the variants is not in the DNA-binding carboxyl-terminal domain, we conclude that the changes in binding are an indication of enhanced oligomerization.
ATPase Activity-Phosphorylation stimulates oligomerization, which in turn stimulates ATP hydrolysis, which is required for transcriptional activation (15,51,52). We confirmed that wild-type NtrC had a low intrinsic (phosphorylation-independent) ATPase activity, which phosphorylation stimulated about 10-fold (Fig. 3). NtrB-independent variants would be expected to have enhanced phosphorylation-independent ATPase activity. This is the case for two variants, NtrC-D86N and NtrC-DE, which had about 7-8-fold higher intrinsic ATPase activity than that of wild-type NtrC (Fig. 3). Phosphorylation stimulated ATPase only 2-fold for these variants, but the stimulated activity was higher than that for wild-type NtrC-P. The other two variants, NtrC-E110K and NtrC-E116K, had essentially wild-type ATPase activity with phosphorylation (Fig. 3).
Transcription of glnA with Purified Components-The mutants were selected based on utilization of arginine in a strain lacking NtrB. The NtrC variants might activate transcription without phosphorylation, but it is also possible that these NtrC variants require phosphorylation by acetyl phosphate. To distinguish between these possibilities, we assessed transcription with purified components. We used templates containing the glnA and ast promoters. Phosphorylation-independent transcription from both promoters was observed for NtrC-D86N, NtrC-E110K, and NtrC-DE but not for NtrC-E116K (Fig. 4).

NMR Analysis of NtrC Variants
To assess the extent of structural changes brought about by the alterations in ␣ helix 5 residues, we collected 1 H-15 N HSQC NMR spectra of purified wild-type and variant NTDs. The NTDs were expressed and purified as fragments containing NtrC residues 1-140 fused to amino-terminal histidine tags. Fig. 5 shows the HSQC spectrum for each ␣5 variant overlaid on the spectrum of the wild-type NTD. All spectra exhibited narrow and well dispersed cross-peaks characteristic of well folded proteins. In addition, the vast majority of cross-peaks in the variant spectra coincided with cross-peaks in the wild-type spectrum indicating that the overall fold of the variant domains were very similar to the wild-type domain. An NMR structure has been determined for a shorter NTD fragment containing residues 1-124 without a histidine tag (25,26). Except for the presence of additional peaks, the HSQC spectrum for our wildtype NTD fragment was very similar to that published for the smaller NTD fragment. As a further check we also generated a wild-type fragment corresponding to NtrC residues 1-124 without a histidine tag. Cross-peaks in the HSQC spectrum of this shorter fragment either overlapped or showed only minor shifts in position from cross-peaks in the HSQC spectrum of the longer wild-type fragment, and they closely matched published chemical shift positions for the shorter wild-type domain (25) (data not shown). These results indicated that the spectra and structure of residues 1-124 in our wild-type NTD were not significantly altered by the presence of the additional residues 125-140 and the histidine tag.
Previous studies showed the level of activity of several NtrC variants to correlate with the degree of NTD conformational change as detected by the degree of change in amide cross-peak position for certain residues in 1 H-15 N HSQC spectra (49,50). These changes reflected a conformational change in the 3445 face and in certain active site residues such as Asp-10, Asp-11, and Asp-54. We wished to determine whether our ␣ helix 5 variants exhibited similar conformational changes by analyzing the degree of shift in cross-peak position for these specific residues. Spectra of the larger, 140-residue wild-type and variant NTDs showed excessive line broadening at NTD concentrations greater than ϳ0.5 mM, indicative of aggregation and The strain used to measure growth on glucose-arginine minimal medium was SN24. The symbols are as follows: Ϫ, no growth; Ϫ/ϩ, very slight growth; ϩ/Ϫ, some growth; ϩ, good growth; ϩϩ, very good growth.
b The strain for glutamine synthetase measurements was SPS1. Glutamine synthetase activity (nmol min Ϫ1 mg of total protein Ϫ1 ) is the average of three determinations Ϯ S.D. precluding facile peak assignments by standard heteronuclear NMR experiments. However, the cross-peaks for many of these residues could be assigned in our HSQC spectra by comparing our wild-type and D86N variant spectra (Fig. 5B) with published chemical shifts and spectra for the shorter wild-type and D86N variant domains (25,49). Assignments were made for cross-peaks that had chemical shifts very similar to the published values, that were also present in our short wild-type NTD spectrum, and that exhibited the same shift (relative magnitude and direction in both the 1 H and 15 N dimensions) in the D86N variant as observed in the published spectra of wildtype and D86N variant NTDs (49). These latter shifts in the D86N variant were strikingly similar to those published for the shorter D86N variant (compare Fig. 5A with spectra in Ref. 49).
In this manner we assigned 15 cross-peaks that shifted in the HSQC spectrum of our D86N variant (Fig. 5A).
When overlaid with the wild-type spectrum, the HSQC spectra of ␣ helix 5 variants E110K, E116K, and D109N/E110K showed less shifting of assigned cross-peaks than observed in the D86N variant (Fig. 5). This was true for active site residues (Asp-10, Asp-11, and the phosphorylation residue Asp-54) and for residues sensitive to the 3445 face conformational change (Gly-36, Ile-69, His-73, Met-81, Thr-82, Asp-86, Leu-87, Asp-88, Ser-92, and Leu-102) (Fig. 5). Quantitation showed the magnitudes of changes in 1 H and 15 N chemical shift to be consistently lower in the ␣ helix 5 variants compared with the D86N variant suggesting a lower degree of conformational change in the active site and the 3445 face (Fig. 6). This was despite some of these variants exhibiting the same level of activity as observed in the D86N variant ( Fig. 3 and Table I). The ␣ helix 5 variants did show some shifted peaks that were not shifted in the D86N variant. Many of these cross-peaks were tentatively assigned by their distinct published chemical shifts and/or by their likelihood of being the point of variation (Fig. 5, unboxed cross-peaks). For the D109N/E110K variant these peaks largely coincided with residues at or near the points of variation (Asp-109 and Glu-110) in the ␣ helix 5 (e.g. Asp-107, Ile-108, Glu-110, and Val-112). Thus, the data strongly suggest that the alterations in ␣ helix 5 affect surface residues of the NTD and do not significantly distort the nonphosphorylated conformation of the NTD. Instead we suspect that the phosphorylation-independent activities of the NTD with alterations in ␣ helix 5 result from altered interdomain interactions.

The Interaction between the Amino-terminal and Central Domains: Two-hybrid Analysis
We evaluated a number of methods to detect an interaction between the amino-terminal and central domains. A direct biochemical assay was not successful. We could purify the NTD, but an NtrC variant with a deletion of the NTD rapidly aggregated (32). We also tried, unsuccessfully, to detect this interaction in vivo by a trans-complementation test. We placed two compatible plasmids in an ntrC mutant: one coded for a polypeptide containing just the NTD, and the other specified an NtrC with a deletion of the NTD. The latter protein repressed the minor glnAp1 promoter of the glnA-ntrBC operon (assessed by measuring glutamine synthetase) but not as well as fulllength NtrC (32). The presence of a gene coding for NTD in trans did not affect this binding. We knew that the NTD fragment was stable in vivo because its presence in an ntrC ϩ strain prevented the utilization of arginine as a nitrogen source. We could demonstrate that this phenotype resulted from titration of NtrB-dependent phosphorylation since a D54N substitution at the phosphorylation site of the amino-terminal fragment reversed the defect in arginine utilization. Thus, we did not We could detect a specific interdomain interaction with the yeast two-hybrid system (Table II). We fused the wild-type amino-terminal and central domains to the Gal4 DNA-binding and activation domains, respectively. When together, the two fusion proteins produced a high level of ␤-galactosidase activity (Table II), which implies an interaction. Three control experiments suggested that the interaction is specific. First, neither fusion protein alone activated lacZ expression. Second, neither fusion protein interacted with a second protein from an unrelated system. Finally, two proteins that are known to interact, but not with the NtrC domains, demonstrated specificity. We were unable to determine whether the interaction was polar, i.e. whether the same result would be observed with the reverse fusions, because strains with the fusion of the NTD to the GAL4 activation domain grew unexpectedly slow. We also utilized a second reporter system for the two-hybrid system that measures expression of the HIS3 gene. Resistance to 3-aminotriazole, a measure of HIS3 expression, further suggested an interaction between the amino-terminal and central domains (data not shown).
We then used this system to quantify the interaction between the mutant NTDs and the wild-type central domain (Table II). NTD variants with the E116K substitution and the D109N/E110K double substitution reduced ␤-galactosidase 4 -10-fold. To rule out the possibility that the mutations reduced the stability of the fusion protein, we assayed the levels of the wild-type amino-terminal fusion protein and the mutant fusion protein with an immunological assay and showed that they were similar (results not shown). The levels of the central domain fusion protein were also the same in the various constructs. These results imply that the various substitutions reduce the interaction between the amino-terminal and central domains. We also determined whether the other NTD variants with substitutions in the 3445 face had a similar effect on the interdomain interaction. The D54E and D86N substitutions each reduced ␤-galactosidase activity 10fold. NtrC-D54E has previously been shown to result in a constitutive phenotype (41,53). We conclude that the twohybrid system can detect altered interactions between the amino-terminal and central domains and that substitutions in either ␣ helix 5 or the 3445 face alter the interaction between the two domains. The boxed cross-peaks were assigned by having the same spectrum positions and exhibiting the same shifts in the D86N variant as reported previously (49). Unboxed peaks were assigned by having distinctive chemical shifts highly similar to values reported previously (25,49) and/or by their likelihood of being the point of variation.

DISCUSSION
Interdomain Interactions and ␣ Helix 5-Kustu and colleagues (26,29,30,49) have analyzed phosphorylation-independent NtrC variants and the structure of the NTD with and without phosphorylation. They proposed a mechanism of interdomain signal transduction in which phosphorylation and several alterations that mimic the effect of phosphorylation (i.e. those that activate transcription without phosphorylation) change the position of the ␣3-␤4-␣4-␤5 region (the 3445 face) relative to the phosphorylation site and the remainder of the NTD (49). This conformational change exposes a hydrophobic surface that appears to interact with the central domain and stimulate its activities (26,30,49).
Our results suggest that interdomain signal transduction is more complex and that the 3445 face is not the only region of the NTD that interacts with the central domain. We used a previously described procedure to isolate strains with alterations in the NTD of NtrC that stimulated central domain activities without phosphorylation (33). This previous selection used Salmonella typhimurium NtrC and resulted in substitutions in three sites in the NTD: D86N, A89T, and V115I (33,49). We isolated 14 E. coli mutants that had seven different alterations: R56H, M75I, D86N, D109N, E110K, E116K, and E124K. One of our variants had two substitutions, D109N and E110K. The number of mutants with alterations in ␣ helix 5 was unexpected. The isolation of these mutants may simply have been a function of isolating more mutants or perhaps is an indication of the substantial organismic differences in the activation mechanism of the ast operon, which was the basis for the mutant isolation (8,54).
The alterations in ␣ helix 5 mimicked many of the functional consequences of phosphorylation. The purified variants generally, but not always, exhibited phosphorylation-independent oligomerization, ATP hydrolysis, and transcription from the promoters of the glnALG and astCADBE operons. The mutant containing the variant with the double D109N/E110K substi-FIG. 6. Chemical shift changes for conformationally sensitive residues in the amino-terminal domains of the D86N and ␣ helix 5 variants. The amide chemical shifts for residues Asp-10, Asp-11, Gly-36, Asp-54, Asp-88, and Met-81 are sensitive to changes in NTD structure with the magnitude of change from the unactivated wild-type state correlating with the degree of structural change (49). A and B compare the chemical shift changes of these residues in the D86N variant with those in the ␣ helix 5 variants as determined from the HSQC spectra in Fig.  5. A shows chemical shift differences in the 1 H dimension, and B shows differences in the 15 N dimension (for D86N the chemical shift differences at position 86 are off scale). The positions of the residues within the secondary structure elements of the NTD are shown diagrammatically above A. DE, D109N/E110K. tution grew the fastest with arginine as a nitrogen source, and all assayed properties were phosphorylation-independent. In contrast, the mutants with either the E116K or E110K substitution grew slower with arginine, and not all of the properties of the purified protein were phosphorylation-independent. The purified E110K variant activated transcription without phosphorylation, yet oligomerization and ATPase activity were not phosphorylation-independent. This is perplexing since oligomerization and ATPase activity are essential for transcriptional activation, which means that these properties must be phosphorylation-independent, at least in vivo. The purified E116K variant had phosphorylation-independent oligomerization but not phosphorylation-independent ATPase activity or transcriptional activation. Cells with this variant grew slower with arginine as a nitrogen source than did cells with the E110K or D109N/E110K variants (Table I). Furthermore the two-hybrid interaction assay suggests that the NTD with the E116K substitution still retained some of its ability to interact with the central domain. NtrC-E116K was simply the weakest phosphorylation-independent variant. A likely explanation for the properties of purified NtrC-E116K and NtrC-E110K is that the central domain exists in inactive and active conformations and that the active conformation in the two singly altered proteins is less stable in vitro than that with the doubly altered protein. In any case, the conclusions of this study can be based solely on the properties of the double substitution variant, NtrC-D109N/E110K, for which all of its properties are phosphorylation-independent.
Although the alterations in ␣ helix 5 generally mimicked the functional consequences of phosphorylation, they did not mimic the structural consequences of phosphorylation. They did not cause the structural change in the 3445 face observed with phosphorylation, and they did not alter the overall structure of the NTD. The location of these substitutions on the surface of the NTD suggests that this outcome is expected (see Fig. 1). Despite the absence of conformational alterations, two-hybrid analysis suggests that alterations in ␣ helix 5 diminish the interaction between the amino-terminal and central domains. We conclude that ␣ helix 5 directly interacts with the central domain and that the function of this interaction differs from that of the interaction with the 3445 face. This is not a surprising conclusion since ␣ helix 5 of other response regulators participates in interdomain interactions, which implies that the interdomain interaction of ␣ helix 5 is not a peculiarity of NtrC. Multidomain DNA-binding response regulators are grouped into OmpR, NarL, and NtrC subfamilies (28). Response regulators can also be single domain pro-teins, such as CheY, which interacts with a flagellar motor protein. Removal of the NTD, more specifically ␣ helix 5, of rhizobial DctD (a member of the NtrC subfamily) is sufficient to activate transcription from a DctD-dependent promoter (55,56). ␣ helix 5 has also been implicated in the interdomain interface of two members of the OmpR subfamily, in interdomain communication in CheB, and in the interaction between CheY and FliM (57)(58)(59)(60)(61).
Signal Transduction in NtrC-Any model of signal transduction in NtrC must be placed within the sequence of events that leads to transcriptional activation. Phosphorylation of the NTD stimulates central domain-dependent oligomerization, resulting in formation of either a hexamer or octamer (9,10,(62)(63)(64). The regulatory interdomain interaction appears to be between the NTD of one polypeptide and the central domain of another (30). Oligomerization in turn stimulates ATP hydrolysis, which drives formation of an open promoter complex.
This established sequence suggests that a primary function of the NTD is to control oligomerization. This raises the questions of how the NTD controls oligomerization and which region of the NTD is involved. The answer requires knowledge of the oligomerization determinants and the structure of the central domain, which are currently not known. Without this knowledge, two general possibilities can be considered: either the 3445 face stabilizes an oligomerization-competent conformation of the central domain (the stabilization hypothesis) or ␣ helix 5 of the NTD blocks the oligomerization determinants of the central domain (the blocking hypothesis). The stabilization hypothesis is consistent with the observation that phosphorylation results in movement of ␣ helix 4 when the NTD is studied without the central domain but impairment of this movement in the intact protein. This result was interpreted to suggest that phosphorylation of the NTD results in an interaction between ␣ helix 4 and the central domain (29,30). The stabilization hypothesis also explains why loss of the NTD impairs NtrC activity (32,33). However, this loss of activity is difficult to interpret unambiguously since the truncated NtrC is not stable (32). The blocking hypothesis is consistent with the observation that loss of the NTD activates DctD, an NtrC-like activator (55). (Unlike a similarly truncated NtrC, the truncated DctD is stable.) The blocking hypothesis more readily explains how some ␣ helix 5 alterations in the absence of changes to the 3445 face are sufficient for oligomerization and activation of the central domain. In other words, the blocking hypothesis accounts for the important observation that movement of the activated 3445 face is not necessary for oligomerization and transcriptional activation. However, the blocking hypothesis by itself does not account for the role of phosphorylation in signal transduction, which requires the 3445 face.
The blocking and stabilization hypotheses are not mutually exclusive. We propose a hybrid blocking/stabilization model in which ␣ helix 5 blocks oligomerization, and the interaction between the activated 3445 face and the central domain overcomes this block. A simple possibility is that ␣ helix 5 and the 3445 face stabilize alternate conformations of the central domain. This hybrid blocking/stabilization model not only is consistent with all the available information but also explains the seemingly anomalous observation that alterations that mimic the structural effects of phosphorylation appear to diminish the interaction with the central domain (Table II). We suggest that the interaction between the central domain and the activated 3445 face is weak and undetectable by the two-hybrid analysis, and movement of the 3445 face interferes with the stronger interaction between the central domain and ␣ helix 5.
Concluding Comments-We have presented evidence that ␣ helix 5 of the NTD contributes to interactions with the central