The OmpR Protein of Escherichia coli Binds to Sites in the ompF Promoter Region in a Hierarchical Manner Determined by its Degree of Phosphorylation*

a major membrane porin protein is differentially regulated by the OmpR protein. OmpR acts as a positive as well as a negative regulator of ompF expression by binding to DNA sequences in the ompF promoter region. The binding activity of OmpR is itself regulated by phosphorylation through the kinase protein EnvZ. Phosphorylation is believed to change the function of OmpR from an activator to a repressor molecule. of the -70 to -100 region is a high affinity site, while the -45 to -60 and -360 to -380 regions low affinity sites. that OmpR binding at the -360 to -380 region previous binding at downstream sequences, which is in- dicative of long range interactions between OmpR molecules. We intrepret our results in terms of a model for ompF regulation involving hierarchical binding

1 To whom all correspondence should be addressed. Tel.: 908-253-4115; Fax: 908-235-4559. and increase the relative amounts of OmpR-phosphate. High levels of OmpR-phosphate cause the ompC gene to become activated, while the ompF gene is repressed (7,12,18,19). This model is supported by studies correlating phosphorylated OmpR levels with outer membrane porin production (12,20). For example, pleiotropic envZ mutants such as em2473 and envZll produce very high levels of phosphorylated OmpR and have an OmpF-, OmpC' outer membrane porin pattern (12,18,20). With regard to the ompF gene, several studies have suggested that OmpR converts from an activator to a repressor molecule when its phosphorylated levels are elevated (7, 18). I n vitro and in vivo footprinting studies have demonstrated that OmpR binds to at least two regulatory sites in the upstream region of ompF (8, 21) (see Fig. 8). One region is between -70 to -100 (relative to the transcriptional start site). The -70 to -100 region is essential for ompF activation and contains three tandemly arranged 10-bp' elements (designated Fa, Fb, and Fc), which share sequence similarities with each other (see Figs. 8 and 9) (8,211. Another regulatory site, termed the Cd box, is between -42 and -52 and is a 10-bp sequence unrelated to the -70 to -100 region (see Figs. 8 and 9). The OmpR472 mutant protein is unable to bind t o the -42 to -52 region (9, 211, and strains harboring the ompR472 allele produce OmpF constitutively (4). Thus, the loss of binding at the -42 to -52 region correlates with the inability of OmpR472 to repress ompF expression and implicates this site in negative regulation (21,22). Another OmpR binding site located 360-380 bp upstream of the transcriptional start site has recently been identified (23). This region has been shown to be important for negative regulation of the ompF gene (24).
While the detailed mechanism of regulated ompF expression remains unclear, a correlation may exist between the extent of occupancy of binding sites in the ompF promoter and OmpR phosphorylation. That is, different levels of OmpR-phosphate may interact selectively with one region over another with the ultimate outcome affecting ompF expression. In this study we examined the phosphorylation requirement for OmpR-DNA interactions at the three binding sites in the ompF promoter. Our results indicate that OmpR binds sequentially to these sequences in a manner determined by the phosphorylated state of the protein. These results suggest a model of ompF regulation involving protein-protein interactions and DNA looping.

12560
Hierarchical Binding of OmpR to the ompF Promoter kilobase pair intervening sequence (ompF regulatory region) between Q S~S and the transcriptional start site for the ompF gene (see Fig. 9), and the 5' portion of the ompF structural gene. The 5' portion of the ompF gene was removed by exchanging SWQI-SQ~I fragments between pARO91 and plasmid pGR241 (8). The resulting plasmid pAR092 contained a BamHI site at the 3"region of the ompF regulatory region. Naturally occurring EcoRI, NruI, HaeII, A M , or SWQI sites were used along with the common BamHI site of pARO91 to clone variable amounts of the ompF promoter into pRS414, a promoterless lac2 vector (26).
For DNase I footprinting a HaeII-Hind111 fragment from plasmid pAR092 was treated with Klenow fragment and deoxyribonucleotides. This was digested with PstI and cloned into Xbd, PstI sites of pGEM3Zf(+), in which the X b d site had been treated with Klenow fragment and deoxyribonucleotides. The resulting plasmid, pAR094, could be digested at flanking BamHI (adjacent to the HaeII site) or HindIII (adjacent to PstI) for radiolabeling with [a-:"P]dGTP or [CY-~*P]~ATP, respectively. Labeling at the HindIII site allowed footprinting of the -42 to -100 region, while labeling at the BamHI site allowed footprinting of the region between -360 to -380. A 314-bp XbaI-BamHI region (-193 to +121) from plasmid pKI0033 (8) was used for most ofthe DNA mobility shift experiments.
Mutations were created in the ompF promoter in pKI0033 using a site-directed mutagenesis kit (Amersham Corp.). Three base pairs were altered in the Cd box (-42 to -51), which changed the sequence from TGTAGCACTT to TGTCGACCTT. The mutation was moved into plasmid pAR092 and subsequently introduced into pRS414. Four base pairs were mutated in the Ff box (-360 to -3691, changing the sequence from ATTACATTGC to ATTGTCGTGC. This mutation was subcloned into plasmids pAR094 and pRS414. In another construction, a HindIII site was created in pKI0033 by making a T to G mutation at the -61 position. This clone, pAROO1, was digested with X b d and HindIII, which released a 130-bp ompF promoter fragment containing the region between -193 to -63 (F boxes alone). Digestion of pAROOl with the HindIII-BamHI fragment released a 184-bp ompF promoter fragment containing the region between -64 to +121 (Cd box).
E. coli strains AR134, AR135, and AR136, which express low copy numbers of plasmids with ColEl replication origins, were created from MC4100, MH1471, and MH760, respectively. Each strain contains a pcnB8O allele, which was introduced by P1 transduction and selecting for a tightly linked TnlO (27). The pcnB8O allele was introduced to minimize multicopy effects of the pBR322-based pRS414 vectors.
Partial Purification of OmpR from Various Strains-Plasmids pAT428 (28) and pHY1430 (29) were used to express the wild-type and the OmpR472 proteins, respectively. Both proteins were purified to homogeneity according to previously published procedures (11). En-vZ(C) is a cytoplasmic fragment of the membrane protein and was prepared according to published methods (11).
For partial purification of OmpR, overnight cultures of MC4100 or its derivative strains MH760, MH1471, or TK821 were used to inoculate 1 liter of Luna broth. The culture was monitored until the cells were in mid-log growth and then harvested by centrifugation. For procaine treatment, procaine was added to cells to a final concentration of 10 mM, and cells were harvested 1 h later. Cell pellets were suspended in 50 mM Tris, 5% glycerol, and 5 mM glycine betaine, pH 7.8 (FPLC buffer) and disrupted by passage through a French Press. Cellular debris was removed by two 10,000 x g centrifugations, and clarified lysates were centrifuged at 100,000 x g for 90 min in a Beckman 70Ti rotor. The resulting supernatant was treated with DNase I (1 pg) and RNase A (2 pg), filtered through a 0.22-pm filter, and diluted 2-fold with FPLC buffer. By using a 50-ml Superloop (Pharmacia LKB Biotechnology Inc.), approximately 25 mg of protein was applied to a DEAE-Fastflow column attached to an FPLC system (Pharmacia). After washing, the column was developed with a 0-100 m NaCl gradient in FPLC buffer.
Column fractions of 0.75 ml were collected in 1.5-ml microcentrifuge tubes containing 0.25 ml of 80% glycerol. Asmall aliquot was taken from each tube and assayed for OmpR DNA binding activity. The remaining was immediately frozen at -20 "C. The amounts of partially purified OmpR was estimated on the basis of comparisons with known amounts of pure OmpR in Western blots using anti-OmpR antibody.
DNA Binding and Footprinting-For DNA mobility shift experiments, 5 pg of EnvZ(C) was phosphorylated for 15 min at room temperature in a reaction consisting of 5 m ATP, 25 mM HEPES, 1 m~ MgCl,, 1 m CaCl,, and 50 mM KCl. The reaction was then transfemed to ice, and approximately 125 ng of OmpR was added to the chilled mixture. Following transfer of phosphate to OmpR, aliquots (representing 25 ng of OmpR) were added to a 15-pl DNA binding reaction containing labeled ompF DNA(about 5,000 cpm), 50 m~ " i s , 15% glycerol, min a t room temperature, and then samples were directly applied to a 5% nondenaturing polyacrylamide gel. The gel was dried and processed for autoradiography. For partially purified samples, 4-36 ng of OmpR protein was added directly to the DNA binding reaction.
For DNase I footprinting, various amounts of OmpR were phosphorylated as above and added to DNA binding reactions containing approximately 50,000 cpm of labeled probe. Approximately 2 ng of a freshly prepared dilution of DNase I (0.5 pg/ml) in 10 mM MgC1, was added, and complexes were digested for 1 min a t room temperature. The reaction was terminated by the addition of stop buffer (3 M sodium acetate, 2% SDS, 25 m EDTA, 50 pg/ml tRNA). Samples were extracted once with phenol:chloroform, ethanol-precipitated, washed with 70% ethanol, dissolved in sequencing dye, and applied to a 6% polyacrylamide sequencing gel containing 8 M urea.

DNA Shift Experiments with Phosphorylated OmpR-DNA
binding of phosphorylated and nonphosphorylated OmpR was examined using a 314-bp DNA fragment encompassing the region between -193 to +121 of the ompF promoter (XbaVBamHI small fragment of pKI0033). We first phosphorylated EnvZ(C) a t room temperature with ATP and then transferred phosphate from EnvZ(C) to OmpR on ice. The DNA binding activity of this newly phosphorylated OmpR was examined by mixing a portion of the reaction with the labeled promoter. The DNA binding reaction also contained an excess of EDTA to prevent any further phosphate transfer between EnvZ(C) and OmpR. Protein-DNA complexes were allowed to form, and complexes were resolved by nondenaturing electrophoresis. A representative autoradiograph of several experiments is shown in Fig. 1. Under these experimental conditions, EnvZ(C) alone or low concentrations of purified (nonphosphorylated) OmpR hardly formed protein-DNA complexes (Fig. 1, lanes 2 and 3). However, brief incubation of OmpR with EnvZ(C) and ATP produced two protein-DNA complexes shown in Fig. 1 as "a"-and "b"complexes (lanes 5-81. The a-complex migrated slower than the b-complex. As incubation of OmpR with EnvZ(C) and ATP was extended to increase the level of phospho-OmpR, more of the radiolabeled DNA appeared as the slower moving a-complex, and eventually all of the DNA were present as the a-complex (lane 8). These results demonstrate that in vitro, DNA binding of OmpR to the ompF promoter proceeds through two complexes, both of which require phosphorylated OmpR.
Competition experiments were performed to study the OmpR binding sites involved with the a-and b-complexes (Fig. 2). In these experiments, protein-DNA complexes were initially formed, and then an excess of unlabeled DNA fragment was added. The unlabeled DNA represented the 314-bp DNA promoter (lanes 3 and 41, 130-bp DNA fragments containing the -70 to -100 region (lanes 5 and 61, a 184-bp fragment containing the -40 to -60 region (lanes 7 and 81, or a 180-bp DNA fragment from a n unrelated gene (lane 9). Fig. 2 shows that both the 314-and 130-bp DNA fragments were effective competitors of both the a-and b-complexes, while the 184-bp DNA was a poor competitor. As both the 300-and 130-bp DNA fragments contained the -70 to -100 region, these results indicate that this region is involved with the formation of both the aand b-complexes.
The higher molecular weight of the a-complex is due to further OmpR interactions at the -40 to -60 region (see below). Nevertheless, at a n 8-fold molar excess, this region caused a minor change in the ratios of a-to b-complexes but otherwise did not significantly compete either the a-or b-complex. These results may indicate that the -40 to -60 region is a low affinity site that requires previous occupancy of the high affinity -70 to -100 region before it can be bound. This aspect is developed in further studies (see below).
Analysis of Partially Pure Forms of OmpR-"0 correlate the a and b protein-DNAcomplexes with various porin phenotypes, we examined the DNA binding activities of partially purified OmpR from ompR472 and envZ473 strains that have OmpF+/ OmpC-and OmpF-/OmpC+ outer membrane porin phenotypes, respectively (5). OmpR was also prepared from procainetreated MC4100 cells to stimulate phosphorylation of wild-type OmpR by wild-type EnvZ. Procaine was used in place of high osmolarity to avoid the growth inhibitory effects of high osmolarity and to limit the contribution of cross-talk pathways that  2 , 5, 8, and 12). or 32 ng (lanes 3 , 6, 9, and 13) of OmpR were tested for their relative DNA binding profile using the 314-bp XbaI-BarnHI ompF promoter fragment. Lune 10 shows the DNA binding activity of approximately 4 ng of OmpR prepared from strain MH1471. Protein DNA complexes are designated as a, b, b', or c. The b'-complex OmpR472 protein.
appears to be a variant of the b-complex and was only observed for the are alternative routes of OmpR phosphorylation (30)(31)(32).
As OmpR-phosphate is usually lost during complete purification, we attempted a rapid but partial purification to preserve its in vivo phosphorylated state. This involved applying cell extracts to a DEAE-Fastflow column, which was attached to a n FPLC chromatography system (Pharmacia). Column fractions eluting from the DEAE matrix between 70 and 80 mM NaCl were immediately frozen at -20 "C as 35% glycerol mixtures. In general, we could partially purify OmpR within a few hours after preparing cell extracts with the preparations remaining active for DNA binding for a t least a week.
DNA binding of the OmpR preparations was tested by mobility shift experiments using the 314-bp promoter fragment (Fig. 3). OmpR from untreated or procaine-treated cells produced protein-DNA complexes identical to those obtained with in vitro phosphorylated OmpR (Fig. 3, lanes 1 4 , compare with Fig. 1, lanes 4-81. OmpR472 preparations formed a complex designated the %'"-complex, which migrated slightly slower than the b-complex (Fig. 3, compare lanes 4-6 with lanes 7-9). On the other hand, OmpR from MH1471 cells produced only the a-complex even at low levels of protein (Fig. 3, lanes 10-13). None of these partially purified forms of OmpR required in vitro phosphorylation by EnvZ for their DNA binding activities, suggesting that they were phosphorylated before purification. Thus the relative DNA binding activities observed by in vitro phosphorylation of OmpR are comparable with those obtained from in vivo phosphorylated forms of the protein.
Previous studies have shown that the OmpR472 mutant binds to the -70 to -100 region but does not bind to the -40 to -60 region (9, 21). Given these results, the appearance of the b'-complex by our OmpR472 preparation would be consistent with an interaction occurring at the -70 to -100 region. Due to technical problems we could not confirm this result by DNase I footprinting using OmpR472 in these preparations (although we verified this by footprinting studies using purified OmpR472 protein (Fig. 5)). Therefore we tested these preparations for DNA binding to an ompF promoter, which had several base changes in the -40 to -60 region.
As shown in Fig. 4, the wild-type OmpR preparation formed both the a-and b-complexes with the wild-type promoter but only produced the b-complex with the DNA having the dis-  (lanes 2 and 6 ) , MH760 cells (lanes 3 and 7), and corresponding fractions from an ompR::TnlO strain TK821 (lanes 4 and 8) were used to form complexes with the XbaI-BamHI ompF promoter fragment (lanes [1][2][3][4] or with an XbaI-BamHI ompF promoter fragment having mutations in the -42 to -51 region (lanes 5-8). Protein DNA complexes are designated as in Fig. 3. The wild-type and mutant promoters in the absence of any protein are shown in lanes 1 and 5, respectively.
rupted C box region. The OmpR472 preparation formed the b'-complex with either the wild-type or mutant promoter region. Therefore, these experiments, along with the competition results (Fig. 2), demonstrate that the b-and b'-complexes result from interactions at the -70 to -100 region because ( a ) they are not affected by mutations at the -42 to -52 region and ( b ) the complexes can be competed with DNA containing the -70 to -100 region. The a-complex can also be competed with DNA containing the -70 to -100 region, but its formation was dramatically affected by the C box mutant. Therefore we conclude that the a-complex arises from interactions at both the -70 to -100 and -40 to -60 regions.
Another protein-DNA complex designated as a "c-complex" was also present in the OmpR preparations (Fig. 3, lanes 1-9;   Fig. 4, lanes 2, 3, 6, and 7). The c-band also appeared when complexes formed between in vitro phosphorylated OmpR and DNA were subjected to limited proteolysis (data not shown). Based on these latter results, it appears that our preparations either contained a proteolyzed form of OmpR still capable of DNA binding or were becoming degraded after complex formation.

DNase Z Footprinting of the Upstream Region of the ompF
Promoter-An OmpR binding site has recently been found 360 to 380 bp upstream to the ompF transcriptional start site (23). While it is believed to act as a repressor site to prevent ompF transcription, the involvement of phosphorylated OmpR is not understood. To address this question, we carried out DNase I footprinting experiments in which this upstream site was carried within a 470-bp HaeII-PstI ompF promoter fragment of pAR094. This DNA encompasses the region from -470 to +1 and thus contains the -40 to -60 and the -70 to -100 regions as well as the -360 to -380 region. For footprinting experiments, pure wild-type OmpR or OmpR472 protein were phosphorylated in vitro using wild-type EnvZ(C) to form the protein-DNA complexes. Fig. 5A shows a DNase I footprint of the -40 to -100 region using the ompF promoter fragment labeled at the downstream HindIII site. A broad band of protection was seen in the -40 to -100 region with 150 ng of OmpR. Such protection has been reported previously (9, 21). DNase I footprinting using the same ompF promoter fragment labeled at the upstream BamHI The mobility shift experiments described in the previous section showed that the mutant OmpR472 protein was unable to produce the slower migrating a-complex. This was further tested by DNase I footprinting of the ompF promoter using purified OmpR472 phosphorylated in vitro. As with wild-type OmpR, increasing levels of phosphorylated OmpR472 produced increased protection of the -60 to -100 region (Fig. 6A). However, protection of the -60 to -100 region required 400 ng of OmpR472, as compared with 150 ng of wild-type OmpR required for the same protection. In addition, OmpR472 was unable to protect the -40 to -50 region except at 1 pg of OmpR472.

IO, 1 mg. Lanes IO of panels A and B represent protection by OmpR
For wild-type OmpR, protection of the -40 to -50 region was observed at 150 ng of protein. Protection of the -360 to -380 region of o m p F also required greater amounts of OmpR472 as compared with wild-type OmpR. Protection of this region occurred with 300 ng of wild-type OmpR. In contrast, 800 ng of OmpR472 was required for the same level of protection.
To determine whether the -360 to -380 region could be bound by OmpR in the absence of the downstream sequences, a fragment of the o m p F promoter containing the sequence between -465 to -195 was used for DNase I footprinting as above.
No protection was observed when as much as 1 pg of OmpR was used (data not shown). Furthermore, 4-point mutations were created a t -363 to -366. DNase I footprints using this promoter and phosphorylated OmpR showed a loss of protection by OmpR to the entire -360 to -380 region (data not shown).
DNase I footprinting was also performed with a promoter containing the 3-point mutations in the Cd box (Fig. 7). Foot- printing of this mutant promoter showed that only the -58 to -100 region was protected by phosphorylated OmpR. Because the amount of OmpR needed for protection of this region was similar to that observed for the wild-type promoter, the loss of binding to the Cd box does not affect the affinity of OmpR for the -58 to -100 region. Similarly, the -360 to -380 region was also protected in the mutated promoter, although the protection was not as great as that seen in the wild-type o m p F promoter. Thus, binding to the -360 to -380 region is also not dependent upon binding to the Cd box, From these results as well a s those described above, OmpR binding to the -360 to -380 region is most likely to be dependent upon binding to the -60 to -100 sequence.
Within the -360 to -380 region, there are two 10-base pair sequences (shown as Fe and Ff in Fig. 8) that have similarities to the sequences previously assigned as F boxes (see Fig. 8).

Hierarchical Binding of OmpR to the ompF Promoter
Although we consider sequences of the F boxes to be high affinity binding sites, the low affinity of OmpR for the -360 to -380 region could possibly be due to the presence of only two 10-bp sequences rather than the 3 tandem sequences observed in the -70 to -100 region. Promoter Activity of ompF-lac2 Plasmids-To correlate the in vitro DNA binding and DNase1 footprinting results, ompF-lac2 fusions were created and analyzed on MacConkey plates in various ompB pcnB80 strains. As shown in Fig. 9, ompF-lac2 plasmids produced phenotypes that depended on ( a ) the nature of the ompB allele contained in the strain and ( b ) the amount of ompF promoter DNA contained on plasmids. Three clones, containing between 0.5 and 1.12 kilobase pairs of DNA upstream to the ompF structural gene, were lacZin an envZ473 strain but la&+ in both wild-type and ompR472 strains. These clones carry the repressor region between -360 and -380 in the ompF gene and thus must actively be preventing lac2 expression in the MH1471 strain. Clones that do not carry the upstream repressor site were la&+ in an MH1471 strain, indicating little repression of lacZ. Our results with plasmids are consistent with studies where ompF-lac2 fusions carried on bacteriophage were used to study ompF repression (18).
The ompF promoter mutant having base changes in the -42 to -52 region was also examined as an ompF-ZacZ fusion. This construction still became l a din the envZ473 strain, AR135. This would indicate that the mutation does not disrupt either ompF expression or repression. To examine this further we compared P-galactosidase levels in liquid cultures for the AR134 strain transformed with plasmids pAR0102 (wt) and pAR0105 (mutated Cd). Both plasmids produced comparable P-galactosidase levels at low osmolarity (approximately 100 unitdml) and in both cases their levels were equally reduced at high osmolarity conditions (approximately 45 units/ml). These results suggest that the lack of DNA binding to the -42 to -52 region alone may not be sufficient to interfere with repression. In a related experiment, the 4-point mutations in the Ff box were also analyzed as an ompF-lac2 fusion (pAR0108). When AR135 cells were transformed with pAR0108, cells became pink due to a higher expression of the ompF-lac2 fusion. This would indicate that mutations in the Ff box partially disrupt negative regulation of ompF expression.

DISCUSSION
OmpR binds to three regions in the ompF promoter for regulated porin gene expression (9, 21). One of these, the -70 to -100 region, is essential for ompF activation, while the -42 to -52 and -360 to -380 regions represent binding sites from which OmpR facilitates negative regulation (9, 17,211. It is not clear how OmpR controls ompF expression from these sites, but the DNA binding affinity for them is affected by the phosphorylation state of OmpR. In turn, the phospho-OmpR levels are well correlated with phenotypic changes in ompF expression.
In the present analysis, mobility shift assays were used to evaluate protein-DNA complexes formed by phosphorylated OmpR. Our results indicate that OmpR binding to the above target sites occurred in an ordered fashion that required increasing amounts of OmpR-phosphate. Occupancy occurred initially at the activator site between -70 to -100 and proceeded to the -42 to -52 region. DNase I assays further showed that binding of the -42 to -100 region is followed by binding at the -360 to -380 region. Protein-DNAcomplexes formed by in vitro phosphorylated OmpR were very similar to those formed by in vivo phosphorylated forms of OmpR.
This included the OmpR472 protein as well as OmpR prepared from a putative phosphatase-defective envZ mutant, envZ473.
Progressive binding of regulatory sequences by OmpR may be fundamentally important to the mechanism of ompF regulation. This idea is incorporated into the following model for ompF regulation. When OmpR-phosphate levels are low (such as at low osmolarity conditions) OmpR binds to the activator site at the -70 to -100 region to stimulate ompF expression. Elevation of OmpR-phosphate levels by high osmolarity conditions (12), envZ phosphatase mutants (201, or the presence of local anesthetics (32) facilitates occupancy at the -42 to -52 region with eventual binding at the -360 to -380 region. Filling of the -360 to -380 region then causes ompF repression. This mechanism fits well with the reported changes in phospho-OmpR levels measured in vivo.
OmpR binds to the -70 to -100 region in the absence of other DNA sequences (33). In contrast, binding at the -42 to -52 and  -360 to -380 does not appear to occur without the -70 to -100 region (this study). We interpret these results as an indication of high and low affinity binding sites although it is not yet clear whether the high affinity site is defined more by its sequence or by the fact that there are three such binding sites juxtaposed to each other (8,9,21). In any event it appears that binding to the -70 to -100 region occurs before binding at other sites and thus may be a prerequisite for subsequent interactions on the DNA.
The -42 to -52 region has been correlated with negative regulation of o m p F , because an OmpR472 protein fails to bind to this sequence and strains harboring this allele produce OmpR constitutively. While our studies confirm that OmpR472 does not bind the Cd box, we also show that OmpR472 binds poorly to the -360 to -380 sequence. Considering that the Cd mutant promoter was still capable of repressing ompF-lac2 expression, the Cd box is not absolutely required for o m p F repression. On the other hand, sequences upstream of -100 have been shown to be essential for o m p F repression, and point mutation of the Ff box results in a decrease in ompF-lac2 repression. Thus, the lack of repression by OmpR472 may be more readily explained on the basis of a lost interaction at the -360 to -380 sequence.
The observation that binding at the -360 to -380 region required downstream regulatory sequences suggests that OmpR molecules bound at this region are stabilized by OmpR molecules bound at other sites. Furthermore, binding to this low affinity site did not occur at the expense of the other low affinity -42 to -52 region. If the -70 to -100 region acts as a high affinity site then OmpR molecules bound to this region may be capable of multiple interactions with those bound at the low affinity regions. We envision that stabilization occurs through OmpR-OmpR interactions, because previous studies indicate that OmpR can form stable multimers upon phosphorylation (34). In this regard the o m p F promoter has been shown to be inherently "bent" (35) and to have binding sites for integration host factor, which affects curvature of the o m p F promoter (23, 36). OmpR-OmpR interactions might be greatly enhanced by DNA topology and factors that bring the bound OmpR molecules closer to each other. A putative DNA looping has been proposed to explain how the distantly located -360 t o -380 region could regulate o m p F expression. Evidence for long range interactions, together with the recognition of DNA bending at the o m p F promoter, are consistent with this loop model. DNA looping is important for regulated transcription of a number of prokaryotic systems and has been well characterized for the uruCBAD operon (reviewed in Refs. 37 and 38). In this system, the uru02 site, 200 base pairs upstream of the promoter, is important for self-repression by the AraC protein.
Helical twist experiments show that looping provides interactions between the uru02 site and the u r d 1 site, which is adjacent to the promoter to effect repression. Deletion of the uruOl and u r d sites decreases the ability ofAraC to bind to the aru02 site, indicating cooperative interaction. Such a model is certainly implicated by our studies, and experiments are currently underway to evaluate this mechanism for the regulation of o m p F expression.