Tandem binding of six OmpR proteins to the ompF upstream regulatory sequence of Escherichia coli.

OmpR is a transcription factor in Escherichia coli whose function is modulated by phosphorylation in the presence of phosphorylated EnvZ, a transmembrane protein histidine kinase involved in osmosensing. Using a protein S-OmpR hybrid protein, we demonstrated that six OmpR molecules bind tandemly to the −100 to −39 sequence of ompF. This sequence consists of three 20-base pair units: F1, F2, and F3, each of which is bound by two OmpR proteins. Polymerase chain reaction selection of nine randomized base pairs within the F1 sequence revealed highly conserved C residues spaced 10 base pairs apart. Further mutational analysis of conserved bases indicated that two OmpR molecules bind tandemly to two direct repeats. Mobility shift assays showed that cooperative interactions play a role in enhancing binding of OmpR to lower affinity F2 and F3 sites. Activation and repression of ompF expression are thus regulated by a total of eight OmpR molecules, including two molecules that bind to a distal site (−380 to −361).

OmpR is a transcription factor in Escherichia coli whose function is modulated by phosphorylation in the presence of phosphorylated EnvZ, a transmembrane protein histidine kinase involved in osmosensing. Using a protein S-OmpR hybrid protein, we demonstrated that six OmpR molecules bind tandemly to the ؊100 to ؊39 sequence of ompF. This sequence consists of three 20base pair units: F1, F2, and F3, each of which is bound by two OmpR proteins. Polymerase chain reaction selection of nine randomized base pairs within the F1 sequence revealed highly conserved C residues spaced 10 base pairs apart. Further mutational analysis of conserved bases indicated that two OmpR molecules bind tandemly to two direct repeats. Mobility shift assays showed that cooperative interactions play a role in enhancing binding of OmpR to lower affinity F2 and F3 sites. Activation and repression of ompF expression are thus regulated by a total of eight OmpR molecules, including two molecules that bind to a distal site (؊380 to ؊361).
When exposed to changes in environmental osmolarity, Escherichia coli responds by altering the expression of OmpF and OmpC, two porin proteins present in the outer membrane (for review, see Csonka and Hanson (1991)). OmpF and OmpC each exist as homotrimeric proteins that span the outer membrane and serve as channels for the passive diffusion of small molecules between the external media and the periplasmic space (Nikaido and Vaara, 1987). Under low osmolarity conditions, OmpF is present in the outer membrane at significantly higher levels than OmpC. Upon a shift to high osmolarity, OmpF is no longer expressed, while the level of OmpC is increased (Kawaji et al., 1979).
Expression of the ompF and ompC genes is dependent upon two proteins, EnvZ and OmpR (Sarma and Reeves, 1977;Silhavy, 1981a, 1981b;Mizuno and Mizushima, 1987;Slauch et al., 1988;Slauch and Silhavy, 1989). These proteins belong to the family of highly conserved histidine kinase/response-regulator proteins that make up the two-component systems prevalent in prokaryotes (Nixon et al. (1986), Parkinson and Kofoid (1992), and for review, see Stock et al. (1989)). Such two-component systems have recently been identified in eukaryotes as well (Ota and Varshavsky, 1993;Chang et al., 1993). In each case, the histidine kinase is phosphorylated on a conserved histidine residue, using free ATP as the substrate. The response regulator then catalyzes the transfer of the phosphoryl group to a completely conserved aspartic acid residue, which accepts the phosphate from the histidine kinase. In many cases, the response regulator is a transcription factor with a DNA-binding domain.
EnvZ is phosphorylated at the histidine 243 residue to function as the substrate for OmpR phosphorylation (Kanamaru et al., 1990;Aiba et al., 1989a;Forst et al., 1989a;Igo and Silhavy, 1988). OmpR is phosphorylated at Asp 55 (Delgado et al., 1993;Kanamaru et al., 1990). EnvZ also acts as an OmpR-phosphate phosphatase (Igo et al., 1989;Aiba et al., 1989a). The level of phosphorylated OmpR present in the cell appears to be controlled through the regulation of EnvZ phosphatase activity (Yang and Inouye, 1991;Russo and Silhavy, 1991;Yang and Inouye, 1993;Jin and Inouye, 1993). At low osmolarity, EnvZ phosphatase activity maintains a low concentration of phosphorylated OmpR in the cytoplasm. Upon a shift to high osmolarity, the phosphatase activity of EnvZ is decreased, allowing an increase in the level of phosphorylated OmpR. Phosphorylation of OmpR enhances DNA-binding activity (Aiba et al., 1989b;Aiba and Mizuno, 1990).
OmpR binds to sequences upstream of the ompF and ompC promoters to regulate the transcription of these genes. In vivo and in vitro footprinting has demonstrated that OmpR binds to the Ϫ100 to Ϫ80 and Ϫ74 to Ϫ40 sequences of ompC (Maeda and Mizuno, 1990), and to the Ϫ100 to Ϫ40 Mizuno and Mizushima, 1986;Mizuno et al., 1988;Tsung et al., 1989) and Ϫ380 to Ϫ360 sequences of ompF (Rampersaud et al., 1994;Huang et al., 1994). Phosphorylated OmpR binds to the upstream regulatory region of ompC to activate ompC expression. However, in the case of ompF, OmpR acts as both an activator (during low osmolarity conditions) and repressor (during high osmolarity conditions) of transcription (Slauch and Silhavy, 1989;Forst et al., 1989b). Since low osmolarity conditions are correlated with low levels of phosphorylated OmpR, it would be expected that under these conditions, only high affinity sites are bound by OmpR. In contrast, under high osmolarity conditions when there is a higher level of phosphorylated OmpR in the cell, OmpR will bind to both high and low affinity sites. It is of interest to understand how OmpR carries out this dual function. Identifying high and low affinity sites will aid in identification of activator and repressor sites.
In the studies presented here, an OmpR fusion protein was designed and used for mobility shift assays to determine the exact number of OmpR molecules that bind to sequences upstream of the ompF promoter. Subsequently we were able to identify individual OmpR binding sites and use a random sequence selection/PCR 1 amplification method to identify critical base pairs within the OmpR binding site. Finally, mobility shift assays were used to demonstrate that cooperative interactions between OmpR proteins bound to the DNA increase the ability of OmpR to bind to low affinity sites. In this paper, we demonstrate that six OmpR molecules cooperatively bind to the Ϫ100 to Ϫ39 sequence of ompF. Taken with previously published data regarding the distal Ϫ380 to Ϫ361 binding site, a total of eight OmpR molecules are involved in the positive and negative regulation of ompF expression.

EXPERIMENTAL PROCEDURES
Plasmids and Oligonucleotides-pPR010 is a pT7-7 derivative that contains the fusion gene that encodes the PrS 2 -OmpR protein. The wild-type ompR gene was obtained from the plasmid pTB0201, which harbors the ompB locus. PCR was used to create a fragment containing the wild-type ompR gene with a NdeI site located at the 5Ј-end where the first methionine of OmpR is encoded. The 3Ј-PCR primer annealed to 3Ј-end of ompR and provided a BamHI site. The PCR fragment was digested with NdeI and BamHI and ligated into the T7 expression vector pET7-7, which had been linearized by a BamHI/NdeI digest, to create the pT70mpR plasmid. To create the PrS 2 -OmpR fusion gene, plasmid pUCtps N was used. This is a pUC9 derivative that contains a fragment encoding the amino-terminal half of protein S. This gene was obtained through PCR amplification of the protein S gene such that NdeI sites were created at the 5Ј-end where the first methionine is encoded and at the 3Ј-end where arginine 92 is encoded. The PCR product was sequenced to confirm that no mutations had been created by PCR. Subcloning this NdeI fragment into the pT7OmpR plasmid digested with NdeI allowed the ligation of two tandem proteinS PCR products (encoding protein S amino acids 1-92 followed by a histidine residue introduced by the 3Ј-PCR primer) to the 5Ј-end of the ompR gene. The plasmid was named pPR010.
The construct pCF001 contains the upstream regulatory regions of both the ompC and ompF genes, cloned into pGEM3Zf(ϩ) (Promega). The ompF upstream regulatory sequence was obtained from the plasmid pAR094 (Rampersaud et al., 1994). The ompC regulatory sequence was obtained from the plasmid pGR241 (Norioka et al., 1986). The fused sequence was created by a two-step PCR reaction. To amplify the ompC sequence, a 5Ј-PCR primer (AGGAATTCATTTCGCCATTCC) was annealed to the Ϫ199 to Ϫ187 sequence of ompC and also created an EcoRI site. The 3Ј-PCR primer (AACTGGATCCAATCCAACAC-GAATATGA) was annealed to the Ϫ44 to Ϫ27 sequence of the ompC gene and created a BamHI site. To amplify the ompF sequence, a 5Ј-PCR primer (GATTGGATCCAGTTCCTTAAATTTTACT) was annealed to the Ϫ112 to Ϫ95 sequence of ompF and created a BamHI site. The corresponding 3Ј-PCR primer (TCAAGCTTGCATGC) was annealed to the Ϫ2 to ϩ12 sequence of ompF in pAR094, where a HindIII site is present. The two PCR products that encoded the ompF and ompC upstream regulatory sequences were then used in a second PCR reaction such that the sequences at the 3Ј-end of ompC sequence annealed to the 5Ј-end of the ompF sequence during the PCR reaction to fuse the two sequences. Thus, the final product contained the ompC Ϫ199 to Ϫ27 sequence upstream of the Ϫ112 to ϩ1 sequence of ompF. A BamHI site is present between the two sequences, an EcoRI site is at the 5Ј-end, and a HindIII site is at the 3Ј-end. The correct sequence was confirmed by DNA sequencing.
The 20-base pair OmpR binding sites were synthesized on an Applied Biosystems DNA synthesizer. Complimentary strands were annealed to create double-stranded DNA sequences for mobility shift assays.
Mobility Shift Assays-DNA fragments containing the ompF regulatory sequences were labeled by first digesting the pCF001 plasmid with HindIII followed by incubation with Klenow in the presence of [␣-32 P]dATP. The DNA was then digested with BamHI, and the labeled fragment was gel-purified. Synthetic DNA molecules were labeled using T4 kinase and [␥-32 P]ATP.
Binding reactions were carried out in 50 mM Tris-HCl, pH 7.5, 50 mM KCl, 5 mM CaCl 2 , 5% glycerol, 0.0025% Nonidet P-40, and 200 ng of poly(dI-dC). First, OmpR was phosphorylated in vitro using ATP and the purified catalytic cytoplasmic domain of EnvZ (amino acids Arg 180 -Gly 450 ), referred to as EnvZ(C) (Delgado et al., 1993). OmpR was incubated with 7 nM EnvZ(C) in the binding buffer with 60 mM ATP at 25°C for 10 min. Appropriate amounts of purified OmpR or PrS 2 -OmpR were then added and allowed to incubate for another 15 min at 25°C. Labeled DNA was added to each reaction, and the final reaction volume was 15 l. Binding reactions were incubated at 4°C for 30 min and then loaded onto a 5% acrylamide/bisacrylamide (40:1.2) gel. The gel was run at 4°C in 1 ϫ TE buffer (10 mM Tris-HCl, pH 7.5, 1 mM EDTA) and run at 120 V. The dried gel was exposed to autoradiography film overnight at Ϫ80°C.
Expression of the PrS 2 -OmpR Fusion Protein-The PrS 2 -OmpR pro-tein was expressed using the T7 expression system. BL21(DE3) cells were transformed with pPR010 and grown until mid-log phase at 37°C in Luria Broth containing 100 g/ml ampicillin. Expression of PrS 2 -OmpR was induced by the addition of 1 mM isopropyl-1-thio-␤-D-galactopyranoside, and the cells were grown for another 2.5 h. PrS 2 -OmpR was purified by the same protocol used for wild-type OmpR purification (Forst et al., 1989a). The final PrS 2 -OmpR protein was greater than 90% pure and was stored as a 1 mg/ml stock in 20% glycerol at Ϫ20°C.

Random Selection and Amplification of OmpR Binding
Sites-For the initial randomized template, a single-stranded DNA molecule was synthesized. The first template used contained 20 randomized nucleotides and had the sequence ATAGAATTCGCGGC(N20)ATTATGGA-TCCGCG, where N20 represents a random 20-mer nucleotide sequence composed of equimolar amounts of G, A, T, and C. Its corresponding PCR primers were ATAGAATTCGCGGC (5Ј) and CGCGGATC-CATAAT (3Ј). A second template had the sequence GACGGATC-CTTTTACTTTTGG(N9)CTGTAGGAATTCTCAC, where N9 represents a random 9-mer nucleotide sequence composed of equimolar amounts of G, A, T, and C. The corresponding PCR primers were GACGGATC-CTTTTTACTTTTGG (5Ј) and GTGAGAATTCCTACAG (3Ј). For the initial shift, the single-stranded DNA oligonucleotide containing the random bases was made into a double-stranded DNA template using Klenow in the presence of an end-labeled PCR primer. Approximately 100 pmol of DNA and 350 ng of OmpR were used for the initial binding reaction. Binding reactions were carried out as described above. The wet gel was exposed for 5 h at 4°C. Gel above the free DNA was cut out, crushed, and soaked overnight at 37°C in elution buffer (0.5 M ammonium acetate, 10 mM magnesium acetate, 1 mM EDTA, pH 8.0, and 0.1% SDS). The eluted DNA was ethanol precipitated and resuspended in 25 l of 0.1 ϫ TE buffer. Five l of this DNA was used as template for a PCR reaction. The PCR reaction was 95°C for 1 min, 45°C for 2 min., 72°C for 1 s, cycled 29 times. The PCR product was gel-purified before subsequent gel shifts.

Construction of an OmpR Fusion
Protein-OmpR DNA-binding activity has been extensively characterized through footprinting and mobility shift assays. In vivo and in vitro footprinting studies have shown that OmpR binds to the Ϫ100 to Ϫ40 sequence of ompC (Maeda and Mizuno, 1990) and to the Ϫ380 to Ϫ360 (Rampersaud et al., 1994;Huang et al., 1994) and Ϫ100 to Ϫ40 sequence of ompF . However, it is not known how many OmpR molecules bind to the upstream region of the ompF and ompC promoters to differentially regulate the transcription of these genes in response to changes in environmental osmolarity. Therefore, an OmpR hybrid protein was constructed for use in mobility shift assays to determine the number of OmpR molecules binding to the upstream region of the ompF promoter.
OmpR was fused to protein S, a protein present on the spore coat of a Gram-negative soil bacterium, Myxococcus xanthus (Inouye et al., 1979(Inouye et al., , 1981. Protein S has a high ␤-sheet content (90%) and is known to be very stable in solution (Bagby et al., 1994). Two tandem copies of a DNA fragment encoding the amino-terminal 92 amino acid residues of protein S were fused to the 5Ј-end of the ompR gene to encode a hybrid protein with a molecular mass of 47 kDa, which is 20 kDa larger than OmpR. The purpose of fusing protein S to OmpR was to increase the size of OmpR without affecting OmpR DNA-binding activity. Thus, in a mobility shift assay, the addition of PrS 2 -OmpR to DNA-binding reactions should simply result in the production of slower mobility bands relative to the mobility of the DNA fragment bound by the intact OmpR molecule. The T7 expression system was used to express this hybrid protein, referred to as PrS 2 -OmpR (Fig. 1A).
Previous studies have shown that OmpR binds to the Ϫ100 to Ϫ40 region of ompF. The BamHI-HindIII fragment of pCF001, containing the Ϫ100 to Ϫ40 region of ompF and referred to as 100/1, was used for mobility shift assays. The gel retardation analysis of this fragment with increasing amounts of phosphorylated OmpR is shown in Fig. 1B. A total of 3 bands, labeled a, b, and c, represent binding by different numbers of OmpR molecules. Band a represents the DNA fragment bound by the minimum number of OmpR molecules under the binding conditions used in the present study. Higher concentrations of OmpR resulted in the binding of all of the DNA with exclusive formation of band c, suggesting that this band represents the DNA fragment bound by the maximum number of OmpR molecules. The existence of three bands indicates that there are three distinct OmpR-binding units in the 100/1 fragment. A similar gel shift pattern consisting of three bands a, b, and c, is observed when this DNA fragment is incubated with increasing concentrations of PrS 2 -OmpR (Fig. 1C). These three bound complexes correspond to the three complexes formed when using intact OmpR (Fig. 1C, far right lane), but the mobility of the DNA bound by PrS 2 -OmpR is slower due to its higher molecular weight. Formation of the complexes required PrS 2 -OmpR concentrations similar to those of intact OmpR, indicating that fusion of protein S to OmpR did not affect OmpR DNA-binding activity. The PrS 2 -OmpR-DNA complexes appear as doublets, possibly due to conformational effects imparted by the protein S domain. In vitro copper phenanthroline footprinting in the presence of PrS 2 -OmpR resulted in a protection pattern identical to that observed in the presence of the intact OmpR (data not shown). This provides further evidence that both OmpR and PrS 2 -OmpR bind to the 100/1 fragment in an identical manner.
Six OmpR Molecules Bind to the Ϫ100 to Ϫ40 region of ompF-Incubation of the 100/1 fragment in the presence of 100 nM OmpR resulted in the formation of a single band c. Since this band represents binding by the maximum number of OmpR molecules, we next determined the number of OmpR molecules bound to the DNA in this complex. To do this, 100 nM OmpR was incubated with increasing concentrations of PrS 2 -OmpR from 5 to 200 nM prior to the addition of the 100/1 DNA fragment. The DNA-binding reactions were then analyzed by gel electrophoresis. As shown in Fig. 2A, a total of seven bound complexes are observed over the range of PrS 2 -OmpR concentrations. As illustrated to the right of the gel shown in Fig. 2A, this result can be most adequately explained by assuming that there are six OmpR binding sites on the fragment and that both OmpR and PrS 2 -OmpR have equal affinities to each site. The band with the slowest mobility represents complete displacement of OmpR by PrS 2 -OmpR. Equal concentrations of OmpR and PrS 2 -OmpR in the binding reaction result predominantly in the complex formed by binding of three OmpR proteins and three PrS 2 -OmpR proteins ( Fig. 2A, 100 nM OmpR with 100 nM PrS2-OmpR). It is interesting to note that protein S, linked to the N-terminal end of OmpR, appears not to interfere with binding of either OmpR or PrS 2 -OmpR molecules to the DNA fragment.
Footprinting of the ompF upstream regulatory sequences in the presence of the OmpR2 protein, a DNA-binding mutant, results in protection of the Ϫ100 to Ϫ60 sequence, but no protection of the Ϫ60 to Ϫ40 sequence Tsung et al., 1989). Also, binding studies previously done in this laboratory using a synthetic DNA fragment containing the Ϫ100 to Ϫ64 sequence of ompF, demonstrated that phosphorylated OmpR can bind this DNA (Rampersaud et al., 1989). Mobility shift assays of OmpR and PrS 2 -OmpR with the synthetic DNA fragment containing the Ϫ100 to Ϫ64 fragment are shown in Fig. 2B. A total of five bands are formed, indicating that four OmpR molecules bind to this fragment. In the presence of increasing concentrations of intact OmpR only (Fig. 2B), two binding events are observed as shown by the presence of two complexes.
FIG. 1. OmpR and PrS 2 -OmpR bind to the ompF upstream regulatory sequences to form a total of three complexes. A, schematic of the PrS 2 -OmpR fusion protein in which two N-terminal domains of protein S were fused to the wild-type OmpR to create a protein with a molecular mass of 42 kDa. B, increasing concentrations of OmpR, as indicated above the gel, were phosphorylated in vitro with EnvZ(C) and ATP. Labeled DNA containing the ompF Ϫ100 to Ϫ40 sequence was added to the phosphorylation reactions, and the reactions were incubated on ice for 30 min. Binding reactions were then loaded on a 5% polyacrylamide gel and run at 4°C in 1 ϫ TE buffer. a, b, and c indicate the three OmpR-DNA complexes formed. C, increasing concentrations of PrS2-OmpR were phosphorylated in vitro and incubated with the ompF Ϫ100 to Ϫ40 DNA as described for OmpR in A. A control reaction using wild-type OmpR is shown in the far right lane as a comparison of the mobilities of the three complexes a, b, and c.
Finally, a 20-bp fragment representing the ompF Ϫ100 to Ϫ81 sequence was used in a mobility shift assay with OmpR and PrS 2 -OmpR. With increasing concentrations of phosphorylated OmpR alone, only a single bound complex is formed (Fig. 2C). The addition of increasing concentrations of PrS 2 -OmpR to OmpR leads to the formation of three complexes (Fig.  2C), indicating that the 20-mer is bound by two OmpR molecules.
The results obtained above indicate that two OmpR molecules bind to each of the three 20-bp sequences in the region downstream of Ϫ100 and that a total of six OmpR molecules bind to the region from Ϫ100 to Ϫ40. The binding of the 100/1 fragment using increasing concentrations of OmpR resulted in the step-wise formation of three distinct complexes. These results suggest that OmpR binds to the DNA "two-at-a-time" and that bands a, b, and c in Fig. 1B thus correspond to two, four, and six OmpR molecules binding to the 100/1 fragment. There are not sufficient data at present to indicate that OmpR forms dimers prior to binding DNA. Binding by two OmpR molecules may occur in a highly cooperative manner, such that formation of a complex with an OmpR monomer was not detected by the mobility shift assay. It is evident that there are three OmpR binding units in the Ϫ100 to Ϫ40 sequence of ompF. Each consists of approximately 20 bp, to which two OmpR molecules bind. These units are assigned F1, F2, and F3 and determined to be from Ϫ100 to Ϫ81, Ϫ80 to Ϫ61, and Ϫ58 to Ϫ39, respectively, on the basis of the results above and a consensus sequence as described below (see Fig. 3).
Determination of an OmpR Consensus Sequence-Binding of a 20-mer sequence by two OmpR molecules as shown above poses the question of whether the proteins bind to two direct repeats or two inverted repeats. When the 20-bp OmpR binding sequences F1, F2, and F3 are aligned with another upstream sequence of the ompF promoter (Ϫ380 to Ϫ361 and assigned F4; Rampersaud et al. (1994) Huang et al. (1994)) and an OmpR-binding sequence of the ompC promoter (Ϫ98 to Ϫ89 and assigned C1; data not shown), a consensus sequence can be found as shown in Fig. 4. Interestingly, the 5Ј-half of the consensus sequence is very T-rich; eight out of 10 bases are T. The only non-T bases are A and C at the fourth and fifth positions from the 5Ј-end, respectively, and the same AC sequence is repeated in the 3Ј-half. It should also be noted that the T at position 2, the C residue at position 5, and the AC sequence at positions 14 and 15, are the only residues completely conserved in the five OmpR binding sequences (Fig. 4).
Random selection and PCR amplification (Oliphant and Struhl, 1988) was used to determine a consensus sequence for OmpR binding. Initially, an oligonucleotide was designed for OmpR binding in which 20 bp were randomized. After repeated attempts failed, a new PCR template was designed that contained the 5Ј 11 bp of the F1 sequence. The final PCR template thus used was 46 bp in length, having the following sequence: GACGGATCCTTTTACTTTTGG(N9)CTGTAGGAATTCTCAC. The ompF Ϫ100 to Ϫ90 sequence (underlined) is followed by nine randomized bases (N9) on the 3Ј-side. After four cycles of binding and PCR amplification, 5% of the DNA was retarded in the mobility shift gel electrophoresis. Essentially 100% of the DNA was retarded after six cycles of binding and PCR amplification. This shifted DNA was eluted, PCR-amplified in the presence of labeled fragment, and used in a mobility shift assay in the presence of 200 nM OmpR and increasing concentrations of PrS 2 -OmpR. As shown in Fig. 5, the results demonstrate that the selected sequence is bound by two OmpR molecules as judged from the formation of three bands.
The PCR products were subcloned and 20 clones were sequenced. The results are shown in Fig. 6. Most striking are the 100% conserved ACA sequence at positions 3, 4, and 5 of the N9 sequence, and the highly conserved TTACA sequence at positions 1-5. It should be noted that the C at position 4 occurs 10 bp downstream of the C in the template, which represents the C at position Ϫ96 within the F1 box. Both C residues correspond to the C residues that are fully conserved in all five OmpR binding sequences shown in Fig. 3. Therefore, it is reasonable to assume that the C residue in the N9 sequence was selected by the distance from the C residue in the portion of the sequence held constant. This accurate determination of the distance can be explained by sequential binding of two OmpR molecules to the 20-mer sequence containing N9; an OmpR molecule first recognizes the motif in the 5Ј-half, TTAC, FIG. 2. Six OmpR proteins bind to the ؊100 to ؊40 sequence of ompF. Varying concentrations of OmpR and PrS 2 -OmpR were mixed, phosphorylated in vitro with EnvZ(C) and ATP, and incubated with the 100/1 fragment, containing the Ϫ100 to Ϫ40 sequence of ompF (A); the 100/64 fragment, containing the Ϫ100 to Ϫ64 sequence of ompF (B); and the 100/81 fragment, containing the Ϫ100 to Ϫ81 sequence of ompF (C). The concentrations of OmpR and PrS 2 -OmpR are indicated for each lane. The binding reactions were incubated on ice for 30 min and loaded onto a 5% acrylamide gel and run at 4°C in 1 ϫ TE buffer. To the right of each gel is a diagrammatic representation of the number of OmpR and PrS 2 -OmpR proteins present in each bound complex. ϫ symbolizes PrS 2 -OmpR and E symbolizes OmpR.
which governs the position of the second OmpR binding. This results in selection of the TTAC sequence within the N9 region as the most preferred sequence. It should also be noted that the consensus sequence in the N9 region from 20 independent clones was TTACATXTX (Fig. 6), which is identical to the 3Ј-half of the F1 sequence (TTACATATT; see Fig. 3) except for 2 Xs. Interestingly, the first X in F1 is A, and the PCR reaction selected A or T at this position in 17 out of 20 cases. Similarly, the second X in F1 is T, and the PCR reaction again picked A or T in 15 out of 20 cases. This result indicates that A or T is the highly preferred base at these positions.
Two OmpR Proteins Tandemly Bind to the F1 Sequence-The F1 sequence (Fig. 3) contains a direct repeat of TTACX-TXT, which was also selected by PCR amplification. This suggests that two OmpR molecules bind tandemly or in the same orientation on the F1 sequence. However, in the F1 sequence, the second of two G residues gives a palindromic nature to F1. The second G residue (G 11 ) is highly conserved (in three of the five sites; see Fig. 4) and the GT sequence (residues 11 and 12 of F1) is palindromic to AC (residues 14 and 15 of F1). When G 11 is changed to T (G 11 3 T; see Fig. 7A), OmpR was still able to bind the sequences (Fig. 7B), albeit weaker than the wildtype F1 sequence (Fig. 7B). This result indicates that G 11 is dispensable, and thus that OmpR does not bind inverted repeat sequences.
Next we analyzed the role of the two fully conserved C residues by individually substituting them with T. Interestingly, the change of C to T at positions 5 and 15 of the 20-mer caused a decrease in binding to approximately 10% of the wild-type 20-mer sequence (Fig. 7B, lanes C5T and C15T). This result indicates that although two OmpR molecules bind to the F1 sequence at the same time, a single OmpR molecule is unable to individually bind to either the 5Ј-or 3Ј-half.
The results described above support the notion that two OmpR molecules bind tandemly to two direct repeats, each of which exists in a 10-bp subunit sequence; one is the 5Ј-half (which we will call subunit a), and the other is the 3Ј-half (which we will call subunit b). Examination of all five binding sequences shown in Fig. 4 reveals an interesting feature; 33 Ts out of a total of 50 bases (66%) are found within subunit a, while only 15 Ts out of a total of 50 bases (30%) are found within subunit b. To examine the significance of this finding, the order of subunits was reversed by placing subunit b of F1 as the 5Ј-half and subunit a of F2 as the 3Ј-half. The 20-mer fragment thus encompasses the ompF upstream sequence from Ϫ90 to Ϫ71 (90/71 fragment, Fig. 7A). This fragment was not bound by OmpR (data not shown). Since the substitution of G 11 FIG. 3. The ؊100 to ؊39 sequence of ompF is composed of three OmpR binding sites, each 20 base pairs long, and designated F1, F2, and F3.

FIG. 4. Computer alignment of OmpR binding sites.
Five OmpR 20 base pair sites, which are each bound by two OmpR proteins, were aligned using the GCG version 7.2 Pileup program. The sites are as follows: F1 (ompF Ϫ100 to -81); F2 (ompF -80 to -61), F3, ompF Ϫ58 to Ϫ39; F4, ompF Ϫ380 to Ϫ361; and C1, ompC Ϫ58 to Ϫ89). Boldface letters indicate residues that are fully conserved, upper case letters indicate highly conserved residues, and lower case letters indicate poorly conserved residues. The numbers above the aligned sequence indicate the positions of the residues within the 20-bp motif as discussed in the text.
FIG. 5. The sequence obtained through random selection and PCR amplification is bound by two OmpR molecules. The selected pool of sequences was PCR-amplified using a PCR primer endlabeled with [␥-32 P]ATP and incubated with in vitro phosphorylated OmpR and PrS 2 -OmpR. The concentrations of each protein are indicated above the gel.
FIG. 6. A highly conserved TTACATNTN sequence was selected within the random pool of binding templates. The pool of amplified sequences after 6 rounds of selection was subcloned, and 20 clones were sequenced. The residues labeled constant represent the 11 bp in the 5Ј-half of the F1 site, which were present in the initial template for OmpR binding selection. N9 indicates the randomized portion of the initial binding template. The highly conserved residues that were selected are in boldface, and a consensus sequence is indicated at the bottom.
with T in F1 resulted in a substantial reduction of OmpR binding (see Fig. 7B), we examined whether the substitution of the T residue with G at position 11 of the 90/71 fragment recovers OmpR binding ability. The resultant fragment (Fig.  7A, 90/71(T11G)) was poorly bound by OmpR (Fig. 7B). These results indicate that the T-richness in the 5Ј-half of the OmpRbinding 20-mer fragments plays an important role in the coordinate binding of two OmpR molecules to the fragment.
OmpR Binds Cooperatively to ompF Regulatory Sequences- Rampersaud et al. (1994) proposed that OmpR binds to the upstream regulatory sequences in a hierarchical manner, showing that the distal Ϫ380 to Ϫ360 site (F4; see Fig. 3) is bound only after prior binding to the Ϫ100 to Ϫ40 sequence. This work suggested that OmpR-OmpR interactions enhance binding of OmpR to this far upstream F4 sequence. Therefore, we further examined the role of OmpR-OmpR interactions in enhancing binding to the F1, F2, and F3 sequences. In mobility shift assays, the 100/1 fragment is completely bound in the presence of 50 nM OmpR (Fig. 8). However, when the F1 box is removed, the resultant fragment (referred to in Fig. 8 as 80/1) is bound poorly by OmpR. Concentrations up to 200 nM OmpR failed to shift all of the labeled DNA. When the 100/1 fragment was incubated in the presence of increasing concentrations of OmpR, three complexes are formed, representing binding by two, four, and six OmpR molecules (Fig. 8, ompF 100/1; also  Fig. 1B). However, when the 80/1 fragment containing only the F2 and F3 sequences is incubated with increasing concentrations of OmpR, only a single complex is formed (Fig. 8, ompF 80/1). It was found that this single complex contains four OmpR molecules as determined by a gel shift with PrS 2 -OmpR (data not shown). This indicates that OmpR will bind both F2 and F3 sequences at the same time, but cannot form a stable complex by binding to only one of the sites. Thus, OmpR-OmpR interactions are essential for OmpR binding to lower affinity F2 and F3 sequences, and OmpR molecules bound to F2 must interact with OmpR molecules bound to F3 in order to form a stable OmpR⅐DNA complex.
This cooperative binding can be further supported by mutational analysis of the 100/1 fragment containing either the C 96 3 T or the C 86 3 T mutation, which corresponds to C 5 3 T and C 15 3 T (C5T and C15T in Fig. 7A), respectively. The C 86 3 T mutation completely eliminated OmpR binding to the F1 20mer (Fig. 7B, lane marked C15T). The mutated sequence was bound by six OmpR molecules as determined by gel shift analysis with PrS 2 -OmpR (data not shown). However, unlike the wild-type ompF sequence in which incubation with increasing OmpR concentrations leads to the formation of three bands, incubation of increasing OmpR concentrations with the mutated sequences produced only two complexes, b and c (Fig. 9). Gel shift assays with PrS 2 -OmpR showed that this mutated fragment is bound by six proteins (data not shown). This suggests that although OmpR can bind to the mutated sequence at Ϫ96 or Ϫ86, this binding requires OmpR binding to the downstream F2-F3 region.

DISCUSSION
The present results unambiguously demonstrated that six OmpR molecules bind to the Ϫ100 to Ϫ39 region of ompF. We determined that this region contains three individual binding sites, 20 base pairs in length, which are each bound by two OmpR proteins. The binding sites were designated F1, F2, and F3, representing the Ϫ100 to Ϫ81, Ϫ80 to Ϫ61, and Ϫ58 to Ϫ39 sequences of ompF, respectively (Fig. 3). Recent data have shown a far upstream OmpR binding site located at the Ϫ380 to Ϫ361 sequence of ompF (Huang et al., 1994;Rampersaud et al., 1994), which also shares sequence identity with F1, F2, and F3. Based on the results reported in this paper, we defined this distal site as F4. Although the number of OmpR molecules bound to this site was not determined due to poor binding to this sequence by OmpR, it is reasonable to assume that two OmpR proteins bind to this 20-base pair motif. Similarly, the ompC Ϫ98 to Ϫ79 sequence is another independent binding site bound by two OmpR molecules (data not shown) and assigned C1.
Although two OmpR molecules bind to a 20-base pair se- FIG. 7. Mutating the conserved residues of the 20-base pair OmpR binding motif decreases binding by OmpR. Twenty-base pair oligonucleotides were synthesized and annealed to form doublestranded DNA molecules for OmpR binding assays. The sequences are shown in A. wt is the wild-type F1 sequence. Numbers above this sequence indicate the positions of the residues within the 20-bp sites as discussed in the text. Boldface residues indicate the point mutations within the mutated 20-bp sequences. 90/70(T11G) is the ompF Ϫ90 to Ϫ71 sequence in which the T at position Ϫ80 is changed to a G. The mobility shift assays are shown in B. The 20-mers were labeled with [␥-32 P]ATP and T4 kinase and incubated with or without in vitro phosphorylated OmpR (100 nM).
FIG. 8. Cooperative interaction between OmpR molecules bound to F2 and F3 are required for binding to these low affinity sites. Increasing concentrations of in vitro phosphorylated OmpR were incubated with the 100/1 fragment (containing the F1, F2, and F3 sites) and with the 80/1 fragment (containing the F2 and F3 sites). The three complexes a, b, and c were formed with the 100/1 fragment, but only one complex was formed with the 80/1 fragment. Binding assays with PrS 2 -OmpR showed that the single complex with the 80/1 fragment was bound by four OmpR proteins (data not shown).
quence, each 20-mer is considered a single site since binding of a single OmpR molecule was never detected. Three complexes observed in mobility shift assays with OmpR and the DNA fragment containing the F1, F2, and F3 sequences (Fig. 1B) were thus formed by binding of two (to F1), four (to F1 and F2), or six OmpR (to F1, F2, and F3) proteins. Currently, there are not sufficient data to suggest that OmpR dimerizes in solution prior to binding DNA. Methods such as gel filtration and light scattering detect only OmpR monomers. 2 Co-precipitation of OmpR (not fused to protein S) with PrS 2 -OmpR bound to M. xanthus spores occurred only in the presence of DNA. 3 It is likely that significant OmpR-OmpR interactions occur only upon binding to DNA.
The finding that two proteins bind to a 20-base pair site led to the question of whether the proteins bind to two direct repeats or to two inverted repeats. Random selection of an optimal OmpR binding sequence by PCR amplification was done using a synthetic fragment, which contained 11 base pairs from the 5Ј-half of the F1 site followed by nine random base pairs. The sequence TTACATNTN was selected within the randomized portion. This suggests that the most preferred OmpR binding sequence consists of two direct repeats consisting of TTACATNTN, since sequences similar to TTACATNTN are seen in the 5Ј-half of F1 and the 5Ј-and 3Ј-halves of F2, F3, F4, and C1. Note that T 2 , C 5 , A 14 , and C 15 are invariable in all of the 20-base pair binding motifs. Substitution of the C 5 with T in a 20-bp template resulted in decreased OmpR binding, while substitution of the C 15 to T in a 20-base pair binding template resulted in a complete loss of OmpR binding activity. These data are consistent with the data of Rampersaud et al. (1989) who showed that incubation of phosphorylated OmpR with a fragment containing the Ϫ100 to Ϫ60 sequence of ompF resulted in the formation of two bands. Introduction of the C 96 3 T or C 86 3 T mutation within this fragment resulted in a decrease in affinity of OmpR for this DNA as well as the loss of the faster complex (that bound by only two proteins). Again, the C 86 3 T mutation was more detrimental to binding than the C 96 3 T mutation. However, a T 96 3 A mutation within this fragment increased binding by OmpR, confirming that the ACA sequence obtained through the random selection is a preferred binding sequence in both the 5Ј-and 3Ј-half of the 20-base pair motif. The present results clearly indicate that two OmpR proteins bind to direct repeat sequences rather than inverted repeat sequences. Note that in F1 there are not inverted repeat sequences consisting of more than three bases. Furthermore, the existence of two absolutely conserved C residues in the two halves of the 20-base pair motif makes it unlikely that OmpR proteins bind to an inverted repeat structure.
Previously, we had assigned each of the direct repeats in F1 as two binding sites, Fa and Fb, which share the 10-base pair sequence NTTACNTNTN (Tsung et al., 1989). It should be noted that Fa is highly T-rich in comparison with Fb. The T-richness in the 5Ј-halves of the 20-base pair motifs is common in all sequences as shown in Fig. 4 and appears to contribute asymmetry to the 20-base pair motif, since exchanging the position between the two halves resulted in the loss of OmpR binding. Maeda et al. (1991) also proposed an asymmetric consensus motif with the sequence TTTTACTTTTTGTAACATAT. Interestingly, Makino et al. (1989) proposed that the OmpR homologue PhoB binds to an asymmetric 18-base pair sequence, the pho box, which is tandemly repeated upstream of promoters controlled by PhoB (Kimura et al., 1989;Kasahara et al., 1991).
Binding of protein dimers to direct repeats has been reported for AraC (Carra and Schleif, 1993), GABP (Thompson et al., 1991), and Mat␣2 (Smith and Johnson, 1992). In each case, the authors propose the presence of a flexible linker domain between the dimerization domain and the DNA-binding domain. The flexible linker allows AraC (Carra and Schleif, 1993) and Mat␣2 (Smith and Johnson, 1992) to bind to the nucleotide sequence repeats in either a direct or inverted orientation, as well as to repeats with variable spacing. In contrast, there appears to be little or no flexibility in the positioning of the two OmpR binding sequences within a 20-base pair motif. One possibility is that the nucleotide sequence of the binding motif confers a strict conformation on the DNA to position the two direct repeats such that two bound OmpR proteins can cooperatively interact for stabilization of the complex. T-rich sequences are known to cause bending of DNA, and bending of the ompF upstream regulatory sequence has been experimentally shown. (Mizuno, 1987;Slauch and Silhavy, 1991).
Note that identification of the fully conserved C residues of an OmpR binding site aided in the more precise identification of the F3 site, which is actually two base pairs removed from the F2 site. This positioning of the F3 site may function to place the bound OmpR proteins on the same face of the DNA helix. Since a full helical turn is 10.5 base pairs in length, the 40 base pairs of the F1 and F2 sites may fall short of four full turns. The location of the F3 site may compensate for this.
Computer alignment of the five OmpR binding sites (Fig. 4) agrees well with the established affinity hierarchy, F1 Ͼ C1 Ͼ F2 Ͼ F3 Ͼ F4 (Rampersaud et al., 1994). 4 Comparison of these sequences suggests that F3 and F4 have the most mismatches and thus are likely to have the lowest affinity for OmpR. Both F3 and F4 lack the highly conserved G residue at position 11 of the 20-base pair site. F4, which seems to have the lowest affinity for OmpR, also lacks the T-richness in its 5Ј-half.
Cooperative interactions also affect the ability of OmpR to bind to the sites upstream of the ompF promoter. Rampersaud et al. (1994) showed that binding of the F4 site depended upon prior binding of OmpR to the downstream region. The present data show that cooperative interactions are also important for binding to the lower affinity F2 and F3 sites. Deletion of the F1 site resulted in a decreased affinity of OmpR for the F2 and F3 sites. It was clear that OmpR was not able to bind to the F2 or F3 sites alone. Clearly, binding of these low affinity sites was enhanced by the presence of OmpR-OmpR interactions. According to the model proposed above, six OmpR proteins bind tandemly to six direct repeats upstream (Ϫ100 to Ϫ39) of the ompF promoter. In addition to these six OmpR molecules, two other molecules are bound to F4, making a total of eight molecules involved in the regulation of ompF expression. Multiple transcription factor binding sites upstream of a promoter is also known in other systems (for review, see Collado-Vides et al. (1991)). In the case of ompF, this allows more flexibility and fine-tuning in the regulation of transcription, as more OmpR binding shifts the function of OmpR from activation to repression. Binding is clearly regulated by the concentration of phosphorylated OmpR in the cell. An increase in medium osmolarity is accompanied by an increase in the number of phosphorylated OmpR molecules (Forst et al., 1990), leading to binding of additional, lower affinity sites. In addition, the need for cooperative interactions ensures that binding of proteins to high affinity sites precedes binding to adjacent low affinity sites.