Recognition of Overlapping Nucleotides by AraC and the Sigma Subunit of RNA Polymerase

Abstract

The Escherichia coli promoter p_BAD, under the control of the AraC protein, drives the expression of mRNA encoding the AraB, AraA, and AraD gene products of the arabinose operon. The binding site of AraC at p_BAD overlaps the RNA polymerase −35 recognition region by 4 bases, leaving 2 bases of the region not contacted by AraC. This overlap raises the question of whether AraC substitutes for the sigma subunit of RNA polymerase in recognition of the −35 region or whether both AraC and sigma make important contacts with the DNA in the −35 region. If sigma does not contact DNA near the −35 region, p_BAD activity should be independent of the identity of the bases in the hexamer region that are not contacted by AraC. We have examined this issue in the p_BAD promoter and in a second promoter where the AraC binding site overlaps the −35 region by only 2 bases. In both cases promoter activity is sensitive to changes in bases not contacted by AraC, showing that despite the overlap, sigma does read DNA in the −35 region. Since sigma and AraC are thus closely positioned at p_BAD, it is possible that AraC and sigma contact one another during transcription initiation. DNA migration retardation assays, however, showed that there exists only a slight degree of DNA binding cooperativity between AraC and sigma, thus suggesting either that the normal interactions between AraC and sigma are weak or that the presence of the entire RNA polymerase is necessary for significant interaction.

The sigma subunit of RNA polymerase (referred to here as sigma) is responsible for the binding of the holoenzyme to promoters during transcription initiation (2, 46). It does this by making sequence-specific contacts with bases in hexameric sequences centered at 10 and 35 bases upstream of the transcription start site on promoters (3, 13, 18, 32, 45, 50). At the −10 hexamer, sigma makes base-specific contacts with the nontemplate strand (23, 34, 41, 42). In addition to sigma-DNA interactions during initiation, protein-protein contacts also occur between transcriptional activators and subunits of RNA polymerase (1, 11, 14, 21, 22, 36, 43).

At many promoters, the recognition sequences of transcriptional activator proteins partly overlap the 6 bases of the −35 region that are contacted by the sigma subunit of RNA polymerase (4). In these cases, does the activator substitute for sigma in the recognition of the −35 region; do both proteins read the −35 region, necessitating overlapped reading by both proteins; or does sigma read an adjacent sequence?

On one hand, direct protein-protein contacts between sigma and upstream transcriptional activators seem to occur. At the λ p_RM promoter, the binding site of λcI overlaps the −35 region for sigma by 2 nucleotides, and genetic experiments suggest an interaction between the λcI protein and the −35 recognition motif of sigma 70 (25, 31). Recently, interactions between sigma and Ada, an AraC homologue from the XylS family of proteins, have been demonstrated genetically at the ada, alkA, and aidB promoters (27, 28). A direct sigma-Ada interaction at the ada and aidB promoters has also been revealed biochemically with DNA migration retardation assays similar to those presented in this paper (27). On the other hand, at the PhoB-dependent PpstS and the CRP-dependent P1gal promoters, where the activator binding site completely overlaps the −35 hexamer, it appears possible that the activator can substitute entirely for recognition by sigma in the −35 region (26).

We studied the ara promoter, p_BAD, which is under the control of two activators, CRP (29, 30) and AraC (12, 15) (Fig. 1). The binding of AraC to the I₁ and I₂ half-sites is stimulated by the presence of arabinose. When these sites are occupied by AraC, and if they overlap the −35 hexamer by 2 or 4 bases, transcription is actively initiated from p_BAD (39).

FIG. 1.

Polymerase-promoter and activator-promoter interactions at p_BAD. The ς⁷⁰ subunit of RNA polymerase contacts the −35 and −10 hexamers. Occupancy of the I₁ and I₂ half-sites by AraC activates transcription with the aid of the CRP protein, most likely utilizing the α-subunit–activator interactions as shown. The binding sites of ς⁷⁰ and AraC overlap by 4 bp at p_BAD. The nucleotides in the −35 hexamer that lie outside the region of overlap are shaded.

At p_BAD, it is likely that the C-terminal domain of the α subunit of polymerase interacts both with CRP and with AraC (49). Two lines of reasoning suggest that AraC may also interact with the sigma subunit of RNA polymerase. First, the R596H mutation in the sigma subunit allows AraC to stimulate p_BAD to high levels in the absence of the normally required CRP (19). Second, although AraC can activate transcription from its position partially overlapping the −35 hexamer, it cannot activate (39) as CRP (14, 47) or OmpR (33) can when they are moved upstream by one or more helical turns.

We have examined whether sigma reads that part of the −35 region that lies outside the AraC-contacting region. If it does read this region, then AraC is not substituting for the contacts made by sigma in the region, and either sigma reads the −35 region as before, or it is only slightly displaced by the presence of AraC. We also analyzed sigma binding at the −35 hexamer at a second promoter where the AraC binding site overlaps the hexamer by only 2 bases.

Our results showed that sigma contacts the nonoverlapped bases of the −35 hexamer. Because of the close spatial placement of AraC and sigma on the promoter DNA, we then looked for an interaction between AraC and sigma that would reveal itself as cooperativity in the binding of AraC and sigma to DNA. To avoid the difficulties that would arise from the known interactions between AraC and the alpha subunit of RNA polymerase (49), we used purified sigma in the absence of the other RNA polymerase subunits. Also, to enhance the weak DNA binding affinity of sigma in the absence of core polymerase, we used a truncated variant of sigma (Δ133). This truncation rid the protein of the N-terminal acidic domain that interferes with binding of sigma to DNA (6, 7), and we were able to observe a slight cooperativity between AraC and sigma in binding to p_BAD.

MATERIALS AND METHODS

Strains and plasmids.

The plasmid used for the initial construction of the I₁-I*p_BAD mutants contained an I₁-I₂p_BAD–galK fusion in pES51 (20). The promoter region of the p7 plasmid, which carries an I₁-I₁–lacZ fusion (39), was replaced with the I₁-I*p_BAD promoter region, resulting in an I₁-I*p_BAD–lacZ fusion. Promoter activity was assayed in TR322 cells (araC⁺B⁺A⁺D⁺ galK Str^r) (only relevant markers are shown) (16).

The plasmid used for overexpression of the ς⁷⁰ variants was pQE30 (QIAgen), in which the rpoD gene is under the control of the T5 promoter (48). This was a kind gift from Alicia Dombroski. Protein was overexpressed in XL1 Blue cells (recA1 endA1 gyrA96 thi-1 hsdR17 supE44 relA1) from Stratagene.

Construction of mutant I₁-I*p_BAD templates.

Site-directed PCR mutagenesis was performed to modify the promoter-proximal araI site in p_BAD and to randomize the nonoverlapping bases (X) in the −35 box. The I* half-site (TAGCGGATCCATCCATA) contained the beginning sequence of the I₂ half-site (TAGCGGATCCTACCTGA) and the later sequence of I₁ (TAGCATTTTTATCCATA). The promoter region was amplified from pES51 with two oligo nucleotides, AAGATTAGCGGATCCATCCATAXXXXCTTTTTATCGCAA (containing the underlined I* araI half-site and the randomized nucleotides marked X) and ACTTAAACTAACCACTTGTG, in PCR buffer containing 50 mM KCl, 20 mM Tris-Cl (pH 8.3), 1.5 mM MgCl₂, 0.01% gelatin, 0.2 mM each deoxynucleoside triphosphate, 100 ng of each oligonucleotide, 1 ng of plasmid DNA as a template, and 5 U of Taq polymerase with 29 cycles of 94°C for 1 min, 40°C for 1 min, and 72°C for 1 min. The amplified fragment was treated with 5 μg of proteinase K/ml in 0.01 M Tris-Cl (pH 7.8), 5 mM EDTA, and 0.5% sodium dodecyl sulfate at 56°C for 30 min. The sample was extracted with an equal volume of phenol followed by ethanol precipitation, digestion with BamHI and HindIII endonucleases, and electrophoresis on a 0.8% agarose gel. The doubly digested fragment was purified from the agarose gel using the Geneclean II gel extraction kit from Bio 101 and cloned into the BamHI and HindIII sites of pES51 to obtain the I₁-I*–galK constructs. To transfer the mutant promoter region to a lacZ-containing plasmid, the promoter region of each I₁-I*–galK construct was cloned into the AseI and HindIII cloning sites of the p7 plasmid (39).

Assays.

The promoter activity of the p_BAD promoter variants was quantitated in Escherichia coli TR322 cells (16) with either β-galactosidase or galactokinase levels. The cells were grown to an optical density at 600 nm of 0.6 in M10 minimal salts, 0.4% glycerol, 10 μg of vitamin B₁ per ml, 0.4% Casamino Acids, 1 mM MgSO₄, and 0.2% arabinose (44), 1 ml was withdrawn, and promoter activity was assayed for β-galactosidase, as described by Miller (37), or for galactokinase (10, 35).

Construction of promoter templates for the DNA migration retardation assay.

End-labeled DNA fragments were generated by PCR using two oligonucleotides such that the I₁-I* site was centrally located on the 100-bp product. PCR was performed using 100 ng of γ-³²P-end-labeled oligonucleotide (ATTTGCACGGCGTCACAC) at 10⁶ cpm/ng, 300 ng of unlabeled oligonucleotide (CGTTTCACTCCATCCAAA), and 10 ng of template plasmid with 0.4 U of Taq polymerase in PCR buffer for 29 cycles of 94°C for 1 min, 50°C for 1 min, and 72°C for 1 min.

The I₁-I₂p_BAD (Fig. 2a), I₁-I₂p_BAD consensus −10 (Fig. 2b), and I₁-I*p_BAD (Fig. 2c) variant promoter fragments used to test for sigma binding were generated by PCR. The I₁-I₂p_BAD bubble (Fig. 2d) was constructed by annealing two oligonucleotides. For the I₁-I₂p_BAD consensus −10 template, the TATAAT sequence at the −10 box was introduced into p_BAD by in vitro mutagenesis as described below before PCR amplification.

FIG. 2.

DNA template variants used to test sigma binding. The two AraC binding half-sites are shown with arrows (underlined in the sequences), and the −35 and −10 hexamers for sigma are shown with solid or broken lines (boldface in the sequences). (a) I₁-I₂p_BAD with the wild-type araI half-sites for AraC and the 4-nucleotide overlap at the −35 hexamer. The sequence of the −10 hexamer was not changed. 5′GCCCATAGCATTTTTATCCATAAGATT AGCGGATCCTACCTGACGCTTTTTATCGCAACTCTCTACTGTTTCTCC 3′. (b) I₁-I₂p_BAD consensus −10. The −10 hexamer in I₁-I₂p_BAD was changed to the consensus sequence TATAAT to generate a stronger ς⁷⁰ binding site. 5′G CCCATAGCATTTTTATCCATAAGATTAGCGGATCCTACCTGACGCTT TTTATCGCAACTCTCTATAATTTCTCC3′. (c) I₁-I*p_BAD containing the I* site in place of the I₂ araI half site in p_BAD. The sequence of the nonoverlapping nucleotides in the −35 hexamer was the same as the parental 5′GACG3′. The I* site is not as tight a binding site for AraC as I₁ but is tighter than the I₂ site. 5′GCCCATAGCATTTTTATCCATAAGATTAGCGGATCCATCCATAGAC GCTTTTTATCGCAACTCTCTACTGTTTCTCC3′. (d) I₁-I₂p_BAD bubble is the same as the wild-type I₁-I₂p_BAD with a heteroduplex mismatched stretch of DNA (lowercase letters) at the −10 region. 5′GCCCATAGCATTTTTATCCATAA GATTAGCGGATCCTACCTGACGCTTTTTATCGCAACTCTCTAgcacttctcc ATACCCG3′.

For the in vitro mutagenesis reaction, 50 ng of double-stranded-DNA template was mixed with 125 ng each of the two complementary oligonucleotides containing in 50 μl of 10 mM KCl, 6 mM (NH₄)₂SO₄, 20 mM Tris-Cl (pH 8.0), 2 mM MgCl₂, 0.1% Triton X-100, and 10 μg of nuclease-free bovine serum albumin (BSA)/ml. The extension reaction was performed with 2.5 U of Pfu polymerase with the following cycling parameters: 95°C for 30 s and then 18 cycles of 95°C for 30 s, 55°C for 1 min, and 68°C for 12 min per cycle. The reaction generated unmethylated complementary double-stranded DNA containing the desired mutation. Ten units of DpnI endonuclease was added for 1 h at 37°C to digest the original methylated DNA template present in the reaction. This is the site-directed mutagenesis technique of the QuikChange protocol of Stratagene.

End-labeled DNA templates for the in vitro DNA migration retardation assay were prepared by PCR amplification. For PCR, 100 ng of γ-³²P-end-labeled oligonucleotide at 10⁶ cpm/ng, 300 ng of unlabeled oligonucleotide(s), and 25 ng of template plasmid containing the required promoter were mixed in 100 μl of PCR buffer. The PCR cycle parameters were 95°C for 1 min, 55°C for 1 min, and 72°C for 1 min for 29 cycles. The oligonucleotides used to amplify the I₁-I₂p_BAD, I₁-I₂p_BAD consensus −10, and I₁-I*p_BAD templates were ATTTGCACGGCTCACAC and CGTTTCACTCCATCCAAA. The I₁-I₂p_BAD bubble DNA was prepared by hybridizing ACTTTGCTAGCCCATAGCATTTTTATCCATAAGAT TAGCGGATCCTACCTGACGCTTTTTATCGCAACTCTCTAgcacttctccATA CCCGTTTTTTTGG and CCAAAAAAACGGGTATcctcttcacgTAGAGAGTT GCGGATAAAAAGCGTCAGGTAGGTACCGCTATCTTATGGATAAAAA TGCTATGGGCTAGCAAAGT (the underlined sequences represent the AraC half-sites I₁ and I₂, the boldface letters represent the −35 sequence, and the lowercase letters show the bubble region around the −10 region). For this I₁-I₂p_BAD bubble, the two oligonucleotides were mixed in equimolar concentrations in 10 mM Tris-Cl (pH 8.0), 1 mM EDTA (pH 8.0), 5 mM MgCl₂, and 50 mM KCl, heated for 10 min at 94°C, and cooled slowly to room temperature over the course of an hour.

Purification of the sigma subunit.

The R596H mutation was introduced into the Δ133 sigma-encoding DNA template by in vitro site-directed mutagenesis (QuikChange). The hexahistidine tag-containing Δ133 and R596HΔ133 sigma variants were overexpressed, purified from inclusion bodies by using nickel columns under denaturing conditions, and renatured as described previously (9, 48).

DNA migration retardation assay.

The DNA migration retardation assay was used to measure dissociation rates of AraC from mutant I₁-I*p_BAD templates as previously described (17). AraC was bound to the mutant I₁-I*p_BAD templates in buffer containing 10 mM Tris-Cl (pH 7.4), 1 mM K-EDTA, 75 mM or 150 mM KCl (depending on the salt concentration required), 1 mM dithiothreitol, 5% glycerol, 50 mM arabinose, and 0.05% NP-40. The higher salt concentration was used when the binding reactions were performed in the presence of arabinose because AraC binds more tightly to DNA in the presence of its ligand and does not show any significant dissociation at lower salt concentrations. For the binding reaction, purified AraC was added so that just 100% of 1 ng (∼10⁴ cpm) of end-labeled DNA was bound. Binding of AraC to DNA was allowed to proceed for 10 min, after which an excess of a competitor containing four tandem I₁ half-sites was added. Aliquots were withdrawn at different time points and loaded onto a native 6% polyacrylamide gel cross-linked with 0.1% methylene-bisacrylamide. The samples were separated by electrophoresis at 150 V for 1.5 h in 100 mM Tris-acetate (pH 7.4) and 1 mM K-EDTA. A Molecular Dynamics PhosphorImager PC was used to quantitate bound versus free DNA, and dissociation rates were determined from a plot of the DNA fraction bound by AraC as a function of dissociation time by a least-squares fit.

Sigma or variants were diluted in 10 mM Tris-Cl (pH 8.0), 10 mM KCl, 10 mM β-mercaptoethanol, 1 mM EDTA, 0.1% (vol/vol) Triton X-100, 0.4 μg of BSA/ml, and 5% glycerol. Binding reactions were performed in 25 mM Tris-acetate (pH 7.4), 14 mM KCl, 0.1 mM EDTA, 1 mM dithiothreitol, 0.03% Triton X-100, 100 μg of BSA/ml, and 5% glycerol. To look for cooperative DNA binding between AraC and sigma, sufficient AraC was added so that ∼100% of 1 ng (∼10⁴ cpm) of γ-³²P-end-labeled DNA would be bound. After 10 min, sigma protein was added for 20 min before electrophoresis of the sample to separate free DNA and the various protein-bound species.

RESULTS

Do Sigma and AraC both contact the −35 region?

Since the I₂ binding half-site of AraC overlaps the promoter −35 recognition region by 4 bp, it is conceivable that the sigma subunit does not contact the −35 region at all and that AraC assumes the role normally taken by part of sigma. One way to determine whether sigma makes DNA contacts in the −35 region is to vary the sequence of that part of the −35 region that is not contacted by AraC. If promoter activity is insensitive to such sequence changes, we could reasonably infer that sigma does not contact DNA in the region.

Altering the two bases of the natural ara p_BAD promoter −35 region (Fig. 3) that are not part of I₂ from CG→TT and CG→CC decreased promoter activity to 90 and 50%, respectively, of that of the parental sequence. These results suggest that sigma does read the sequence of the two bases. In the P22 ant promoter however, Moyle and coworkers found that C and G are equivalent at position −30 and that a C-to-T change at position −31 reduces activity to 10% (38). Most likely the difference between the modest change to 90% activity in our system and the dramatic change to 10% in the ant promoter results from the very different contexts in which the sequence changes occur. In the ara system, AraC and CRP are required for normal activation of RNA polymerase, whereas in the ant system, no auxiliary activators are needed by RNA polymerase.

FIG. 3.

Sequences of I₁-I₂p_BAD promoter (top) and I₁-I*p_BAD promoter (bottom). Oligonucleotide-directed PCR mutagenesis was used to randomize the nonoverlapping bases (marked X and shaded) in the −35 hexamer. wt, wild type.

To increase the number of bases in the −35 region that are not contained within the promoter-proximal AraC binding site, we designed a promoter variant, I₁-I*p_BAD (Fig. 3), in which the AraC binding site has been moved upstream by 2 bases. The I* site contains the beginning sequence of the I₂ site and the later sequence of the I₁ site. This promoter is still AraC dependent and is 2.3 times as active as the wild-type p_BAD promoter. Making such a change in the promoter permits four bases in the −35 region to be altered without affecting AraC binding. We chose to alter these four nucleotides in two ways—by directed and by random mutagenesis. If, despite its requirement for AraC and CRP for full stimulation, and despite the results obtained from the 2-base overlap, p_BAD possesses the same promoter sequence dependence as “bare” promoters like P22 ant, then a change to taGACA should have a particularly dramatic stimulatory effect because of increased homology to the consensus −35 hexamer. Table 1 shows that the activity of taGACA was not significantly different from that of the parental sequence, taGACG. The activities of most of the entries shown in Table 1 that resulted from random mutagenesis, however, were strongly dependent upon the sequence of the −35 region outside the I* half-site, thereby indicating that sigma does contact the four bases.

TABLE 1.

Properties of the I₁-I^* variant promoters

−35 sequence	Relative promoter activitya	AraC dissociation half-time (min)b
		−Ara	+Ara
Parental taGACG	1.00	25	110
taGACA	1.00	22	110
taGCCT	0.35	25
taTGAT	0.30	26
taTGGT	0.30	25	110
taGGTT	0.20	26
taCAAC	0.20
taCAAT	0.20	22	110
taATTA	0.20	23
taTGTC	<0.05	28
taTTTG	<0.05	28	110
taATTG	<0.05	26
taTTGT	<0.05	26	100
taAGGT	<0.05	28
taAAGT	<0.05	28
taATTT	<0.05	26
taTTCT	<0.05	25

^aRelative in vivo promoter activities were quantitated from β-galactosidase assays performed in the presence of arabinose on exponentially growing TR322 cells containing the mutant promoters in the p7 plasmid. The parental I₁-I*p_BAD promoter activity of 16,000 U was assigned a relative value of 1. 

^bThe dissociation half-times of AraC from mutant I₁-I*p_BAD DNA templates in 75 mM KCl (no arabinose [−Ara]) and 150 mM KCl (plus arabinose) were measured using the DNA migration retardation assay as described in Materials and Methods and in the legend to Fig. 4. 

Two potential factors could invalidate the conclusion that I₁-I*p_BAD activity is dependent upon the identity of the −35 region nucleotides outside the I* half-site. First, introduction of the altered nucleotides might have inadvertently altered nucleotides elsewhere in the plasmid as well, for example, within the β-galactosidase gene. To verify that the decreased levels of β-galactosidase activity we observed with some of the variants were indeed due to changes in the −35 region of the promoter and not due to extraneous mutations elsewhere on the plasmids, we changed the four randomized bases in three mutant templates back to the parental sequence by oligonucleotide-directed site-specific mutagenesis. These changes returned the β-galactosidase levels to those observed for the parental sequence, indicating that the plasmid carried no additional relevant mutations on the mutant templates.

A second possibility is that AraC binding actually is sensitive to DNA sequence outside I₁ and I*. This possibility was excluded by measurement of the dissociation rates of AraC from the I₁-I*p_BAD templates using the in vitro DNA migration retardation assay (typical data is shown in Fig. 4). Identical dissociation rates were obtained for AraC from all the I₁-I*p_BAD templates (Table 1), indicating that the reduction in promoter activity from these templates was unlikely to be due to altered AraC binding at the −35 region. The results suggest that altered sigma binding is the cause of the reduction.

FIG. 4.

In vitro DNA migration retardation assays to measure dissociation rates of AraC from the I₁-I*p_BAD templates. The two gels show the bound and free DNA at each time point (lanes 1 to 6) as well as fully bound DNA (lane 7) and free DNA (lane 8). The ratio of bound to total DNA was quantitated and plotted as a function of time. + Ara, with arabinose; − Ara, without arabinose.

Does sigma directly interact with AraC?

On one hand, the partial interdigitation of the AraC and RNA polymerase sigma subunit binding sites on DNA suggests that the two proteins could be located very close to each other and hence might have critical interactions with one another. On the other hand, the fact that the binding site of AraC can be moved 2 bases upstream without strongly affecting promoter activity, as in I₁-I₁p_BAD (39) and I₁-I*p_BAD, suggests that perhaps AraC and sigma do not make specific contacts with each other. To test if AraC and sigma do interact with one another, we looked for cooperativity in their binding to DNA.

Purified sigma factor does not detectably bind to promoters by itself, but truncation of its acidic N-terminal domain reveals a weak promoter binding specificity (6, 7, 48). Therefore in looking for cooperativity between AraC and sigma factor in binding at AraC-activated promoters, we used a sigma variant with its N-terminal 133 amino acids deleted. The binding of sigma was examined on I₁-I₂p_BAD and parental I₁-I*p_BAD templates, but no binding was observed on either template in the presence or absence of bound AraC protein (see Materials and Methods for a description of the DNA templates). Changing the conserved arginine at position 596 to a histidine in the sigma subunit enables RNA polymerase to be active on p_BAD in the absence of CRP (19). Possibly the increased activation results from a sigma-AraC interaction, either an interaction where none existed before or a stronger interaction. Therefore, we introduced the R596H mutation into the Δ133 sigma variant. We were still unable to observe sigma binding to either the I₁-I₂p_BAD or parental I₁-I*p_BAD DNA in the presence or absence of AraC. To create a stronger sigma binding site on p_BAD, we changed the −10 hexamer to the consensus −10 sequence (see Materials and Methods), but still no binding was observed for either the Δ133 or the R596HΔ133 sigma variant on I₁-I₂p_BAD consensus −10.

In a further effort to increase sigma binding, we used a template that mimics the DNA present in the open complex (RP_o) during transcription initiation (6, 8). Such a bubble sequence provides a significant advantage for sigma binding, as shown by the preference of RNA polymerase holoenzyme for binding to premelted sequences (5). We used a heteroduplex mismatch bubble-containing template, I₁-I₂p_BAD bubble, that contained the AraC binding sites, I₁ and I₂, with a mismatch region spanning the −10 region (see Materials and Methods). With such DNA, we observed some AraC-dependent DNA binding by the Δ133 and R596HΔ133 sigma variants (Fig. 5). Using another bubble template with binding sites for AraC and the consensus −10 region on the nontemplate strand of the bubble (34) did not enhance binding by sigma in the presence or absence of AraC. We note that the AraC-dependent binding by the truncated sigma protein in all these experiments was not completely reliable, and occasional experiments failed to demonstrate any cooperativity in the binding of AraC and sigma to DNA.

FIG. 5.

DNA migration retardation assay showing weak cooperativity in the binding of AraC and ς⁷⁰ to I₁-I₂p_BAD bubble DNA. The ingredients present (+) are indicated, and except in lanes 7 and 8, where the concentrations of R596HΔ133 ς⁷⁰ were 190 nM and 130 nM, respectively, the concentrations used were 420 pM AraC, 60 nM Δ133 ς⁷⁰, and 380 nM R596HΔ133 ς⁷⁰.

DISCUSSION

Our experiments yield the following conclusions: AraC and the sigma subunit of RNA polymerase both make contacts with DNA in the −35 region of the p_BAD promoter, and interactions between AraC and a truncated form of sigma can be observed in their binding in vitro to DNA, but these interactions are not strong.

The −35 hexamer of p_BAD shares homology of four bases with the consensus −35 sequence, making it a potentially tight binding site for the sigma subunit of RNA polymerase (Fig. 3). On the other hand, because the polymerase-proximal half-site of AraC overlaps the −35 region by 4 bp, it is possible that AraC substitutes for the role taken by the domain of sigma that normally contacts the −35 region. In the experiments reported here, we found that p_BAD activity is strongly dependent on the identities of the bases of the −35 hexamer that are not contacted by AraC. We presume, then, that these bases are contacted by sigma. Consequently, we further presume that either the remaining bases of the −35 region are also contacted by the sigma subunit or sigma is displaced and reads the bases immediately adjacent to the AraC binding site.

It is possible that AraC and the sigma subunit sequentially contact the common four bases in their partially overlapping binding sites at p_BAD. We note, however, that simultaneous contact of these bases by alpha helices is also geometrically possible (Fig. 6). Our detection of cooperativity, albeit weak, in the binding of sigma and AraC to the DNA indicates a direct interaction between the two, suggesting simultaneous DNA binding. We must add that our in vitro binding studies utilized AraC and the sigma subunit alone but that the normal binding involves AraC and the RNA polymerase holoenzyme. The additional subunits of RNA polymerase could also interact with AraC or alter the structure of sigma and alter the interaction.

FIG. 6.

Model showing simultaneous recognition of four overlapping bases by two α-helices fitted into the major groove visualized from the top (left), front (middle), and side looking down one DNA major groove directly at one α-helix (right).

We first used a truncated variant of sigma (Δ133) that lacked the N-terminal acidic domain to enhance the weak DNA binding affinity of sigma in the absence of core polymerase. Because we observed no binding by the truncated sigma factor and no binding cooperativity between AraC and sigma, we then tried DNA templates that should have higher affinity than double-stranded DNA. Ultimately, we did observe binding cooperativity between AraC and truncated sigma, but this required the use of a DNA template possessing a single-stranded region in the −10 region. We are aware of only one other experiment to examine by direct biochemical means an interaction between an upstream activator and sigma factor (27). This work used the Ada protein, double-stranded DNA, and intact sigma factor. As significant binding cooperativity was observed with the Ada and sigma proteins at the ada and aidB promoters, it is possible that the Ada protein interacts significantly more strongly with sigma than does AraC. Alternatively, it is possible that in removing the N-terminal portion of sigma to enhance its DNA binding abilities, we also removed important regions for the AraC-sigma protein interaction.

Promoter recognition at p_BAD by sigma is intriguing because, while any deviation from the parental sequence causes a reduction in promoter activity, we could see no correlation between specific sequence changes and promoter activity. Similarly, at the pmelR promoter, positions 3 to 6 of the −35 region lie outside the CRP binding site and play an important role in activation by sigma. Some −35 hexamer sequences at pmelR are more tolerant of substitutions than others, and mutations that change nonconsensus bases to consensus do not necessarily increase promoter activity (40). Similar observations have been noted with the melAB promoter (24) and the P22 ant promoter (38).