We have previously shown that Lrp and IHF bind simultaneously to gltBDF promoter (PgltB) DNA, yielding a band with intermediate gel shift mobility between the IHF-DNA and Lrp-DNA complexes (Fig. 3C in 14). Fig. 3A shows that the Lrp-binding region yields a subtly altered Lrp-dependent DNAse I footprint in the presence of IHF, though the IHF footprint is not obviously changed by the presence of Lrp.

RNA polymerase holoenzyme (RNAP) binds to PgltB in the absence of other proteins, as revealed by electrophoretic mobility shift of a DNA segment from -406 to +131 (relative to the transcription start; not shown). However DNAse I footprinting analysis of this binary complex revealed no obvious regions of protection (Fig. 3B, compare lanes 5 and 6). Since the absence of either Lrp or IHF leads to a >30-fold reduction in gltBDF transcription 514, we investigated whether these two proteins influence RNAP binding to PgltB. Mobility shift experiments were not informative as, under our conditions, the large shift due to RNAP alone was indistinguishable from shifts due to RNAP together with Lrp and/or IHF. However DNAse I protection analysis revealed that the combination of Lrp, IHF and RNAP results in a footprint in the promoter region, with strongly enhanced cleavage at the -19 position of the template strand and protection in the remainder of the region between the -10 and -35 hexamers (Fig. 3B, compare lane 7 to lanes 5 and 6). This protection extended ~30 bp upstream of the -35 hexamer, merging with the footprint of IHF, and some protection was visible even farther upstream between the IHF and Lrp binding sites. A combination of just Lrp and IHF does not result in any additional footprint aside from those of the individual proteins 14, and in particular the extended protection upstream of the -35 hexamer is not seen in the absence of RNAP (Fig. 3A).

We searched for proteins from an E. coli whole-cell extract that bound to immobilized, biotinylated PgltB DNA, using Ciphergen ProteinChips™ and mass spectrometry. This method had previously implicated IHF as a factor involved in gltBDF regulation, as we subsequently confirmed 14, but additional mass peaks representing other DNA binding proteins had not yet been investigated. The analysis suggested ArgR (regulator for the arginine regulon) and Crp (catabolite repressor protein) as potential additional regulators of gltBDF (Fig. 4). Several decades ago Prusiner et al. reported negative regulation of glutamate synthase expression by Crp 15. Specifically, glutamate synthase activity is doubled in a mutant lacking adenylate cyclase (cya), and reduced 1.5-fold by addition of exogenous cAMP or growth with glycerol as the carbon source (which increases cAMP levels) 15. However, it was not determined whether these results involved direct interactions between Crp and PgltB.

Our inspection of the gltBDF DNA sequence revealed a potential Crp-binding site between -76 and -55 (Fig. 5A, top two lines). We have not attempted to measure transcription of a gltB-lacZ fusion in isogenic crp and crp⁺ strains, to complement the earlier data of Prusiner et al. 15. This is because the crp mutant grows much more slowly than its isogenic partner 20, and comparisons would be complicated by the different growth rates 21. However we have taken three complementary approaches to demonstrate a direct effect of Crp-cAMP on glt transcription.

First, we carried out mobility shift assays using purified Crp and a gltBDF DNA fragment containing the region from -406 to +246. Crp bound the glt DNA with an apparent K_d of 35 nM in the presence of 20 μM cAMP (Fig. 5B). Replacing the region between -121 and -48 with heterologous DNA 14 resulted in loss of detectable binding (Fig. 5C). When cAMP was not added to the gel, but was present in the binding reaction, binding was still observed but the shifted bands were smeared, suggesting that the complexes were unstable under this condition (not shown).

Second, we carried out DNAse I footprint analyses, using purified Crp, cAMP, and PgltB DNA. A defined region of partial protection was seen between -71 and -52 on the non-template strand (Fig. 5E), and -72 to -57 on the template strand (not shown). Thus Crp-cAMP interacts directly with PgltB. The Crp binding is centered at -65.5, in the predicted binding sequence (-76 to -55), and between the -35 hexamer and the IHF binding site (Fig. 2A).

Third, we determined the effects of the membrane-permeable derivative dibutyryl-cAMP on expression of gltB-lacZ with and without alteration of the Crp binding sequence. The alterations are shown in Fig. 5A (bottom line), and replace all of the most highly-conserved nucleotides of known Crp-binding sequences (shaded) that were not already different in the native gltBDF promoter. Concentrations of dibutyryl-cAMP as high as 50 μM had no effect on the growth rate of the cultures (not shown), and we used 10 μM in the experiments described here. As a control, parental strain W3110 (with an intact lac operon) was grown in a defined glucose rich medium plus the inducer IPTG. As shown in Fig. 6A, dibutyryl-cAMP had the expected effect of increasing lacZ expression in the parental strain, due to relief of catabolite repression 22. In contrast, Fig. 6B shows that dibutyryl-cAMP reduced expression of lacZ fused to WT PgltB, while Fig. 6C shows that dibutyryl-cAMP had no effect on lacZ fused to PgltB having the altered Crp binding sequence. Similar results were obtained in glucose minimal medium (not shown).

As Crp most often acts as a transcription activator, we explored some possible explanations for its repressive effect on PgltB. One possibility, given the proximity of the Crp and IHF binding sites (Fig. 2A), and the requirement of IHF for activation by Lrp 14, is that Crp interferes with IHF binding. However mobility shift analysis indicates that Crp and IHF bind independently of one another (Fig. 5D), though the ternary and binary complexes could not be resolved from one another well enough to completely rule out limited negative (or positive) cooperativity in their binding.

A second possibility, which we did not test, is that the Crp-induced bend in the DNA 23 opposes that generated by IHF, preventing proper activation. A third possibility is that Crp interferes with RNAP binding. As noted above, RNAP alone fails to yield a clear protection pattern on PgltB DNA. Adding Crp-cAMP, far from interfering with RNAP binding, resulted in the appearance of a well-defined pattern of RNAP-dependent protection (Fig. 7). This pattern is, however, distinct from that seen when the coactivators Lrp and IHF are added in place of Crp (Fig. 3B). One particularly striking difference, as an example, is the DNase cleavage at position -19, that shows hypersensitivity in the presence of Lrp + IHF but not in the presence of Crp (compare lane 7 of Fig. 3B to lane 7 of Fig. 7).

The Ciphergen ProteinChip™ analysis referred to above yielded another E. coli polypeptide bound to PgltB DNA, having a molecular mass close to that of the regulatory protein ArgR (Fig. 4A). The involvement of ArgR in gltBDF regulation is reinforced by the following three observations. First, adding 7.5 mM L-arginine to MOPS-glucose medium nearly halved LacZ activity, in the gltB-lacZ fusion strain LP1000 (and its isogenic rph⁺ strain LP2020; see below) (Table 1). Second, deleting argR from strains LP1000 (LP1050) or LP2020 (LP2023) doubled LacZ activity. Third, adding arginine to the growth medium did not affect LacZ levels in the ΔargR strain LP1050.

We have evidence that ArgR mediates its regulatory effect on the gltBDF promoter via direct interaction. There are two good matches to the consensus ArgR binding site sequence 2425 upstream of the promoter-distal Lrp binding site (Figs. 2A and 8A). When gltBDF DNA containing the putative ArgR sites upstream of -270 was deleted, arginine had no effect on gltB-lacZ expression (strain LP1270, Table 1). We extended this observation by using purified ArgR in mobility shift and DNAse I footprint analyses. ArgR bound to gltB DNA (-406 to +246) with a K_d^app of 1.8 ± 0.7 nM (Fig. 8B), and to a 3'-shortened fragment (-406 to -128) with essentially the same affinity (K_d^app of 1.7 ± 0.06 nM; not shown). These data were fitted using the Hill equation as previously described 262728, and both binding curves were characterized by a Hill coefficient of 2.0, indicating that multiple ArgR molecules are binding cooperatively.

The sequence to which ArgR binds was determined by DNase I protection using gltB DNA from -462 to +161. ArgR protected a region upstream of the distal Lrp binding site (Figs. 2A and 8C), extending from -361 to -318 on the non-template strand. This encompasses the two ArgR consensus-matching sites, and is consistent with the Hill coefficient of 2.0. Similar results were obtained for the template strand (not shown; a longer fragment, extending from -575 to +131, gave lower resolution due to the position of the footprint on the fragment).

We considered two possibilities for the basis of gltBDF repression by ArgR. First, given the proximity of the ArgR and Lrp binding sites (Fig. 2A), we tested whether ArgR interfered with Lrp binding. Mobility shift assays with ArgR and Lrp were inconclusive in this respect, because adding the corepressor arginine to the gel prevented the Lrp-DNA complex from migrating far into the gel (not shown). However DNAse I footprint analyses suggest that Lrp and ArgR bind independently of one another (Fig. 9A).

A second possible basis for repression of gltBDF by ArgR involves interference with the proposed Lrp-RNAP interaction. ArgR does in fact alter the protection pattern generated by the combination of Lrp, IHF, and RNAP (Fig. 9B, asterisks). In the absence of ArgR there is a limited region of RNAP-dependent hypersensitivity that overlaps the ArgR binding site; addition of ArgR and arginine alters this hypersensitivity pattern.

We considered the possibility that our measurements of gltBDF expression were being affected by a defect in the E. coli W3110 background. This commonly-used K-12 strain carries a frameshift mutation in rph, the structural gene for RNAsePH, that results in decreased pyrE (orotate phosphoribosyltransferase) expression and in turn leads to pyrimidine limitation during growth in minimal media 29. Pyrimidine biosynthesis is regulated by feedback inhibition, so it seemed plausible that W3110 would sense increased demand for glutamate (to provide needed glutamine precursors), and would thus have elevated gltBDF expression. Consistent with this possibility, adding 20 μg/ml of the pyrimidine uracil to strain LP1000 in minimal medium resulted in a 1.5 fold reduction of PgltB-lacZ activity (strain LP1000 is a W3110 derivative carrying a gltB-lacZ fusion in a Δlac background). We replaced the W3110 rph allele with the wild-type gene (see Methods), and the new strain was designated LP2002. Correcting the rph mutation roughly halved transcription from PgltB in strain LP2020 (strain LP2002 bearing Δlac and a gltB-lacZ fusion, see Table 2), in agreement with the uracil addition results. Strain LP2002 may be generally useful for studies of E. coli physiology.

IHF, a DNA-bending protein 30, is required for activation of gltBDF transcription by Lrp 14. IHF subtly affects the Lrp-dependent DNAse I footprint (Fig. 3A), suggesting that IHF alters the Lrp-DNA complex (or alters DNAse I access to that complex). In addition, IHF bending could bring the Lrp-DNA complex into activating contact with RNAP at PgltB (Fig. 2B). IHF plays an architectural role at several activated promoters, such as PilvP 31. We have not directly demonstrated looping, but two results support this view. First, no defined footprint for RNAP was seen at PgltB with either IHF or Lrp alone, only with both together (Fig. 3B and our unpublished results). Second, an earlier study demonstrated that a 5 bp (half helical turn) insertion between the proximal Lrp binding site and the promoter reduced gltB-lacZ expression 3-7-fold, while a 10 bp (full helical turn) insertion at the same location did not reduce fusion expression (Fig. 5B of Wiese et al. 26). Bending or looping models do not rule out the possibility of direct RNAP-IHF contacts occurring as well 32.

The flux through the GlnA/GltBD pathway can be quite high in minimal media. ATP is required for the GlnA/GltBD pathway and not for the GdhA pathway (Fig. 1), and governing the relative fluxes through these alternative pathways is probably one of the most critical regulatory problems faced by E. coli. We report direct regulation of PgltB by Crp, in agreement with earlier indirect evidence 15. A microarray-based global analysis of E. coli genes affected by Crp 33 did not identify the gltBDF operon, but this probably reflects the rich culture medium (LB) used in that work. Crp is broadly associated with the control of carbon and energy source utilization, usually activating genes for utilization of less-efficient carbon/energy sources (relative to glucose) 33. Competitions between gdhA and gdhA⁺ strains 34 reveal a clear advantage for the gdhA⁺ strain during glucose limitation, suggesting that Crp modulates the relative activities of the two alternative pathways for glutamate synthesis to favor the ATP-independent pathway during growth on suboptimal carbon sources 34. The reciprocal effects of carbon source and addition of cAMP on GdhA and GltBD activities are consistent with just this role for Crp 15. In addition, Crp-cAMP represses one of the promoters for the glutamine synthetase gene (glnAp2) 34, so both enzymes for the GlnA/GltBD pathway are controlled in parallel by Crp.

Our data show that Crp-cAMP binding in vitro (Fig. 5B, E) is correlated with reduced transcription in vivo (Fig. 6), and that RNAP generates a protection pattern at PgltB in the presence of Crp-cAMP (Fig. 7, lane 7) different from the pattern yielded by RNAP in the presence of IHF + Lrp (Fig. 3B, lane 7). The Crp-cAMP repression thus appears to alter RNAP binding rather than blocking it. The Crp footprint is centered at -65.5 relative to the start of gltB transcription, close to the position from which Crp activates class I promoters 35 where Crp binds between the C-terminal domain of RpoA (αCTD) and region 4 of RpoD (σ⁷⁰). However optimal spacing from the proximal edge of the Crp binding site to that of the -35 hexamer is 13–16 bp in class I promoters, while in PgltB this spacing is 21 bp (Fig. 2A). Crp may thus repress PgltB by mispositioning or trapping RNAP. Repression has been seen due to mispositioning of RNAP via alternative binding sites 36, or trapping at the promoter 3738. A recent global analysis suggests that nearly a quarter of σ⁷⁰ (RpoD)-dependent promoters have bound RNAP that is "poised" but not transcribing 39. RNAP poised at PgltB would allow a rapid increase of GltBD levels in the face of falling ammonium concentrations, as the GdhA reaction becomes increasingly inefficient (Fig. 1).

As shown in Fig. 1, glutamate is a precursor of ornithine, which in turn is converted to arginine. Thus regulation of GltBD in response to arginine is an example of negative feedback regulation. The relatively modest effects are appropriate given the need for glutamate in many other pathways. A similar feedback regulation of glutamate biosynthesis by ArgR was recently observed in Pseudomonas aeruginosa 40. The mechanism by which ArgR represses PgltB is not clear, but our evidence is consistent with the possibility that ArgR interferes with formation of the final activation complex. As shown in Fig. 8, and illustrated in Fig. 2, ArgR binds far upstream of the promoter, near the Lrp-binding region. However ArgR does not appear to have a major effect on Lrp binding (Fig. 9A). Our working hypothesis is that ArgR allows pre-assembly of the Lrp-IHF-RNAP complex, but blocks a specific contact required for transcriptional activation. This would resemble the suggested role of Crp (see above) in reducing gltBD expression while allowing rapid return to full activation should the appropriate conditions change.

Lrp is not simply a signal of amino acid sufficiency. First, gltBDF is one of the operons for which Lrp regulation is relatively insensitive to the coregulator leucine 528. Second, Lrp levels (like those of the coactivator IHF 41) are sensitive to the alarmone ppGpp 42, which in turn is affected by sufficiency of both amino acids and other nutrients 43. Lrp levels vary between media, and (like those of the coactivator IHF 41) also through the growth cycle 42.

Much of what is known about regulation of GdhA, GlnA, and GltBD is consistent with the expected reciprocal pattern between the two pathways shown in Fig. 1. One would expect GdhA levels to parallel those of [NH₄⁺], while GlnA and GltBD levels would perhaps follow the inverse of [NH₄⁺] and parallel those of ATP. In fact, gdhA is repressed by Nac, with repression relieved in high [NH₄⁺], while neither glnA nor gltBD appear to respond substantially to Nac 174445. Similarly, transcription of both glnA and gltBD is repressed by Crp-cAMP while effects on gdhA are unclear (this work, 3446). Conversely, both GlnA activity (indirectly, via glnL) and gltBD transcription are boosted by Lrp, which has no obvious effect on gdhA 5628.

Modeling analysis 47 suggests that the major control of relative flux between the alternative pathways shown in Fig. 1 should involve regulation of GlnA activity (via adenylylation and deadenylylation 48). Lrp has major effects on GlnA expression and adenylylation status in response to ammonia levels 257. The effect of Lrp on GlnA expression and activity is indirect, probably involving the direct Lrp effect on GltBD expression that, in turn, modulates the intracellular ratios of α-ketoglutarate, glutamate, and glutamine 49. This model requires changes in gene expression levels for adaptation to sudden large changes in [NH₄⁺], with gltBD needing to be expressed above a threshold level. The 30-40-fold activation of PgltB by Lrp 28 is consistent with a major switch above or below this critical threshold expression level. In contrast, the effects of Crp and ArgR (and of leucine, in this case) are smaller, fine-tuning controls that may modulate the flow of glutamine out of the middle of the GlnA-GltBD pathway.

The strains, plasmids and PCR primers used in this study are listed in Table 2. All strains were derivatives of E. coli K-12 W3110 or LP2002 (in which the rph mutation of W3110 was corrected; see below). For β-galactosidase assays, strains were grown in glucose minimal MOPS medium 5051 containing ampicillin (20 μg/ml for chromosomal and 80 μg/ml for plasmid-borne resistance) or 7.5 mM L-arginine HCl (Sigma) where indicated. Cultures grown to test the effects of Crp-binding site mutations were grown in MOPS defined-rich medium (Teknova, Hollister, CA) containing 100 μg ampicillin per ml and 1 mM isopropyl-β-D-thiogalactoside (IPTG), in the presence or absence of 10 μM dibutyryl-cAMP (BioMol Research Labs, Plymouth Meeting, PA). Plates for most genetic work used agar containing "LB" medium 52.

The widely used E. coli K-12 strain W3110 contains a frameshift mutation in rph associated with decreased levels of the pyrE product orotate phosphoribosyltransferase 29. This frameshift was corrected as follows. A 790 bp fragment of rph and downstream sequence was amplified from W3110 chromosomal DNA using the polymerase chain reaction (PCR) and the overlap extension procedure 53. The primers used were rph1-4 (Table 2). Primers rph2 and rph3 contain the additional base pair required to correct the frame shift in the rph gene (underlined italic in Table 2). The PCR product was cloned into E. coli strain XL-1Blue (Stratagene) using the pGEMT-easy vector (Promega). The insert was then excised with NotI and ligated into NotI-digested suicide vector pKO3 54. Electrocompetent W3110 cells were transformed with pKO3 carrying the modified rph gene, and integrants were selected at 43°C. Cells that had lost the plasmid were isolated by plating on LB agar containing 5% sucrose, which selects against the plasmid sacB gene 54. Colonies from LB-sucrose plates were replica plated onto unsupplemented LB and LB-chloramphenicol (20 μg/ml). Colonies that grew on LB plates but not on LB-chloramphenicol were screened for correction of the frame shift in the rph gene by amplifying the rph region by PCR and sequencing the product. The resulting rph⁺ strain was designated LP2002.

The lac operon was deleted from strain LP2002 using the overlap extension protocol 53 with primers lac6-9 (Table 2). The lac6 primer corresponds to nucleotides 814 to 842 within the coding region of lacI; the lac7 primer is complementary to the 5' region of primer lac8; the 3' end of lac8 bridges the deletion region (see below), and lac9 corresponds to nucleotides 1019 to 1044 of the coding region of cynX gene downstream of the lacZYA operon. The final amplified product has a deletion starting immediately upstream of the -10 region of the lacZYA operon and extending to the 3' end of the lacA gene, leaving the last 19 codons of lacA intact. This product was introduced into the chromosome of LP2002 using the suicide vector pKO3 54. The ΔlacZYA recombinants were identified as white colonies on LB-Xgal plates, and the deletion was verified by amplifying the chromosomal lac region and sequencing the fragment. The resultant strain was designated LP2010. The gltB-lacZ transcriptional fusion from LP1000 1455 was introduced into LP2010 via P1 transduction 55 and designated LP2020.

A portion of argR was deleted from strains LP1000 and LP2020 as follows. The argR gene from strain PS2209 (W3110 Δlac-169 from F. C. Neidhardt) was amplified from chromosomal DNA by PCR, using the primers arg1 and arg2 (Table 2). The amplified product was ligated into the pGEMTeasy vector (Promega). The argR fragment was excised from the vector with NotI, and the region of argR coding for amino acids 15–90 were deleted in-frame by removing a DraI/EcoRV fragment from the argR gene and religating the blunt ends. This fragment was ligated into suicide vector pKO3 54 and introduced into strains LP1000 and LP2020. Prospective mutants were screened by amplifying the argR region from their genomic DNA; the amplified fragment was shorter in the deletion mutants. The ΔargR strains were designated LP1050 (LP1000 derivative) and LP2023 (LP2020 derivative).

Surface-enhanced laser desorption and ionization (SELDI) ProteinChip technology (Ciphergen Biosystems, currently distributed by BioRad, Hercules, CA) was used to identify proteins binding to the upstream region of the gltBDF operon. In this approach, biotin-labeled gltB DNA (from -406 to +246) was attached to streptavidin-coated chips and incubated with centrifugally-cleared whole-cell extracts. The masses of proteins that bound to the DNA fragments were determined using SELDI mass spectroscopy. This binding experiment was carried out under nonstringent conditions so as to detect relatively weak binding interactions. E. coli cells (strain W3110) were grown in glucose minimal MOPS medium and opened using glass beads in a BeadBeater (BioSpec Products, Bartlesville, OK). PS-1 (carbonyl diimidazole) ProteinChips were coated with streptavidin (Pierce, Immunopure™) at 0.2 microgram/spot, washed, and the unbound sites were blocked with 1 M ethanolamine (pH 8.0). The DNA, generated by PCR using biotinylated primers, was first bound to the streptavidin on the chip, and then incubated with cell extract in 20 mM Tris-acetate (pH 8.0), 0.1 mM EDTA, 0.1 mM dithiothreitol, 250 mM NaCl, 4 mM Mg acetate, 12.5% glycerol (v:v), and 200 μg/ml of salmon sperm DNA (Amersham Pharmacia Biotech, N.J.). The chips were washed with 20 mM Tris-acetate containing 0.1% Triton X-100. The matrix EAM-1 (Ciphergen) was added to the chip prior to SELDI mass analysis.

In all cases, protein purity was assessed from single (or, in the case of IHF, double) Coomassie-stained bands when μg amounts were loaded onto SDS polyacrylamide gels. Lrp was purified from strain JWD3 and isolated as previously described 28. The genes argR, crp, himA and himD (the latter two coding for IHF subunits) were amplified from strain W3110 chromosomal DNA. The argR gene was amplified using primers arg3 and arg4 (Table 2), cloned into the expression vector pET23b (Novagen) at the NdeI/XhoI sites creating pLP1060, and introduced into the expression host BL21(DE3) (Novagen) which carries an inducible gene for T7 RNAP. The protein was purified from the resulting strain LP1060, as described by Sunnerhagen et al. 56.

The crp gene was amplified using primers crp5 and crp6 (Table 2), ligated into the NdeI/XhoI sites of expression vector pET29a (Novagen) to create pLP1070, and introduced into strain BL21(DE3) resulting in strain LP1070. The overexpressed protein was purified using a nickel affinity column (Pharmacia) since the wild type Crp (without a his-tag) binds nickel columns with moderate affinity 57.

The genes himA and himD were amplified from W3110 chromosomal DNA using primers himA1 and himA2 and primers himD1 and himD2 (Table 2). The amplified products were ligated in tandem into the expression vector pET28a (Novagen) at the NcoI/BamHI sites to create pLP1080, and introduced into strain BL21(DE3). The strain was designated LP1080. IHF with an N-terminal histidine tag on HimA was purified using a 3 ml Hi-Trap Heparin Sepharose column from Pharmacia 58.

The promoter region upstream of gltB was amplified from E. coli chromosomal DNA using primers gltP1-2 (Table 2), and the resulting PCR product was digested with BamHI and SalI. This product was ligated into pBH403, a derivative of pKK232-8 with a promoterless lacZ gene between two bidirectional transcription terminators, to generate pPM2005 (Table 2). This process was repeated with a mutagenic primer pair (gltcrp1-2, Table 2) to generate pPM2006, which has alterations within the Crp-binding site in the glt promoter. Plasmids were sequence confirmed and used to transform the ΔlacZ W3110 derivative PS2209.

Cultures were grown to exponential phase in the indicated media. Samples were taken at 30-min intervals throughout the growth period. Levels of β-galactosidase were determined by o-nitrophenyl-β-D-galactoside (ONPG) hydrolysis as described by Platko et al. 59. β-galactosidase levels were plotted against culture absorbance, and points were fitted via linear regression. The resulting slope yields the β-galactosidase acitivity.

RNA polymerase holoenzyme was purchased from Epicentre Biotechnologies (Madison, WI). Assays were carried out as described in Paul et al. 14 except where indicated. In electrophoretic mobility shift experiments involving ArgR, the reaction mixture and the polyacrylamide gel both contained 5 mM L-arginine, and the running buffer (1 × TBE [90 mM Tris borate, pH 8.3, 0.2 mM EDTA]) contained 1 mM L-arginine. Crp-DNA binding was carried out in a buffer modified from Seoh and Tai 60 containing 10 mM Tris-HCl (pH 8.0), 50 mM KCl, 1.0 mM EDTA, 0.1 mM DTT, 50 μg/ml BSA, 10 μg/ml poly (dI-dC):poly(dI-dC) from Pharmacia and 12.5% glycerol. cAMP (20 mM, Sigma) was added to the reaction buffer, the gel and the running buffer (0.5 × TBE). The buffer for DNase I protection assays (10 mM Tris-HCl, pH 8.0, 5 mM MgCl₂, 1 mM CaCl₂, 2 mM dithiothreitol, 50 mg of bovine serum albumin/ml and 2 μg of poly(dI-dC):poly(dI-dC)/ml) contained 5 mM L-arginine or 20 mM cAMP for reactions involving ArgR and Crp respectively.