Autoregulation of Escherichia coli purR Requires Two Control Sites Downstream of the Promotert

The expression of Escherichia coli purR, which encodes the pur regulon repressor protein, is autoregulated. Autoregulation at the level of transcription requires two operator sites, designated purRol and purRo2 (O1 and 02). Operator 01 is in the region of DNA between the transcription start site and the site for translation initiation, and 02 is in the protein-coding region. The repressor. protein binds noncooperatively to 01 with a sixfold-higher affinity than to 02, and saturation of 01 by the repressor precedes saturation of 02. Both Oi and 02 function in the two- to threefold autoregulation in yto; as determined by measurement of 0-galactosidase and mRNA from purR-lacZ translational fusions. Of all the genes thus far known to be regulated by the Pur repressor, only purR employs a two-operator mechanism. de IMP polycistronic

In Escherichia coli, the genes encoding the enzymes for the de novo synthesis of IMP are arranged as individual loci and small polycistronic operons. The gene organization and map locations (in minutes) are as follows (2,33): purB, 25.2; purHD, 90.3; purEK, 12.2; cvp purF dedF (11,37), 50.0; purMN, 53.5; purC, 53.3; purL, 55.2. In addition, guaBA at min 53.9 is required for the two-step conversion of IMP, to GMP, and purA at min 95.0 and purB are required for conversion of IMP.to AMP. The addition of exogenous purines to the growth medium causes repression of all genes in the pathway (15,36). However, the AMP and GMP branches.appear to be under separate regulation from the pathway leading to IMP (15,54). Genes for the pathway to IMP, except.for purB (23), are coregulated by the purRenc.oded repressor (23,34) and a corepressor that is a small molecule (25). These genes constitute the E. coli pur regulon. Gene purR encoding the pur regulon repressor has been cloned (42) -and sequenced, and operator binding sites have been identified (29,42). Each of the coregulated pur genes has a 16-base-pair (bp) conserved operator sequence that is located in the promoter region (1, lla, 32, 43, 45, 47, 47a, 48 submittedfor publication). Mutational analysis (41) and DNase I footprinting (23,42) have established that the Pur repressor binds to these operator sequences for the negative regulation of pur regulon gene expression (23,34).
Gene purR contains two operatorlike sequences located downstream of the promoter (42). In this report, we show that purR is autoregulated and that the operatorlike sequences purRo1 and purRo2 (01 and 02) are authentic control sites. The mechanism of autoregulation involves the independent binding of Pur repressor toO1 and 02-Operator site 02, located within the purR coding sequence, binds the repressor in vitro with a sixfold lower affinity than 01 and yet makes an important contribution to in vivo autoregulation under the conditions studies. During the preparation of this manuscript, Meng et al. (34) reported the autoregulation of * Corresponding author.
t Journal paper no. 12559 from the Purdue University Agricultural Experiment Station. E. coli purR. Our work supports and extends their analyses and gives a more complete picture of the autoregulation of purR.
MATERIALS AND METHODS Bacterial strains and plasmids. Strains and piasmids are described in Table 1.
-Plasmid conmstructions. A series of plasmids (Table 1) was produced in the construction of pPR1006, the source of fragments for binding assays. A 3.2-kilobase BalI-to-PstI fragment ( Fig. 1A) from purR+ plasmid pPR1003 was subcloned into the HincIl and PstI. sites of pUC119, yielding plasmid pPR1004. The. 3' noncoding sequences were removed by deleting sequences between the PstI site and the HpaI site located at position 1266 relative to the transcription start site of purR. The resulting plasmid, pPR1004-2, contained the entire coding sequence ofpurR and 136 of 155 nucleotides of the 5' transcribed but untranslated sequence cloned in the opposite direction to the lac promoter. An NdeI site was constructed at the initiating ATG codon in plasmid pPR1004-2 by oligonucleotide-directed mutagenesis (31) to produce pPR1005. Plasmid pPR1006 is a derivative of pPR1005 in which most of the coding sequence was removed by deleting from an internal HincII site at nucleotide 321 to the downstream polylinker SphI site.
A series of plasmids was produced in the construction of purR-lacZ fusions and the mutagenesis of 01 and 02. Plasmid pPR2000 contains the promoter region and was constructed by cloning the XhoI (position -245)-to-BglI (position 258) fragment (Fig. 1A) from pPR1002 into the Hincll site of pUC119. A purR-IacZ fusion was constructed in two steps by subcloning purR' DNA from pPR2000 into the low-copy-number vector pGB2 by using EcoRI and HindIII polylinker sites. A Pstl fragment containing lac'Z from pMC1871 was ligated downstream of the purR sequences in the PstI site. The fusion gene created at this step (Fig. IB) contained 35 codons of purR and 13 codons from a polylinkerjoined to codon 8 of lacZ. This purR-lacZ translational fusion was under the same controls as the purR gene. The copy number of this plasmid is expected to be maintained between four and six copies per cell (21).
Substitutions were made in the operators 01 and 02 by oligonucleotide-directed mutagenesis with pPR2000 as the source of single-stranded DNA. The purR-lacZ fusions were created by using the same cloning scheme described for the wild-type fusion. The plasmid designations for the purR-lacZ containing 01-02, 0102-7 and 01-02are pPR2004, pPR2005, and pPR2006, respectively.
Assay of I-galactosidase. Cells were grown to the midlog phase in minimal medium with and without adenine. The P-galactosidase activity of purR-lacZ fusions was determined by the Miller assay (35). Since purR and purR+ strains were each grown with and without added purine, it was possible to calculate repression by two methods ( Table 2). In every case, the values calculated by the two methods agreed closely and supported the reliability of these measurements of repression.
Repressor-operator binding. Protein extracts were prepared as described previously (42) (42), with the following changes: 10 fmol of each DNA fragment was used for binding; extract was added containing 0, 0.5, 1.0, 2.5, 5.0, 7.5, 10.0, and 20.0 p,g of protein; N-2-hydroxyethylpiperazine-N'-2-ethanesulfonic acid (pH 8.0) and glycerol were added to 10 mM and 10%, respectively. Samples were resolved by electrophoresis at 100 V for 1.5 h. For DNase I footprinting, plasmid pPR1006 was digested with either EcoRI or HindIII, end labeled with T4 polynucleotide kinase and [y-32P]ATP, and then treated with the other restriction enzyme to cut on the opposite side of the polylinker. Fragments were electroeluted from 5% polyacrylamide gel slices. For DNase I footprinting, repressoroperator complexes were prepared as described above by using 0, 1, 2.5, 5, 10, and 15 p.g of protein. After complex formation, 2.5 mM MgCl2 and 0.2 mg of DNase I per ml were added. Digestion was stopped after 1.5 min by the addition of 5 mM EDTA and placing samples on ice. Samples were extracted with phenol-chloroform (1:1), 20 ,ug of glycogen was added, and DNA was ethanol precipitated. Pellets were suspended in 2.5 ,u1 of water and 2.5 of formamide dye (80% formamide, 0.1% EDTA, 0.1% each bromophenol blue and xylene cyanol). Half of this sample was loaded onto an 8% polyacrylamide-urea sequencing gel for electrophoretic resolution.
RNA analyses. Cells were grown to the midlog phase in minimal medium with or without adenine, and 10 mCi of [5,6-3H]uridine (specific activity, 38 Ci/mmol) was added for 1.5 min. Cells were harvested by filtration through Whatman type GF/C glass fiber filters for 20 to 30 s, and RNA was isolated as described previously (10). Concentrations of RNA samples were determined by A260.
RNA from the purR-lacZ fusion was quantitated by RNA-DNA hybridization (14). Hybridizations were in vials containing one 13-mm nitrocellulose filter with 2 p.g of immobi- 66°C. Nonspecific RNA was removed by washing at room temperature as follows: two washes in 2x SSC-10 p.g of RNase A per ml for 30 min; two washes in 2x SSC for 30 min. Filters were dried for.10 min on each side under a heat lamp, placed in scintillation vials, and counted after complete dissolution in toluene-based scintillation fluid. All hybridizations were in duplicate to linearized pMC1871, which contains lac'Z in pBR322, and to linearized pBR322 as a control.

RESULTS
Autoregulation of purR. Two pur regulon operatorlike sequences flank the initiating methionine codon in the purR gene (42). Operator 01 is located between positions 96 and 111 relative to the transcription initiation site, and operator 02 iS in the coding sequence at positions 184 to 199 (Fig. 1A).
To determine whether the Pur repressor regulates expression of the purR gene, a purR-lacZ translational fusion was constructed in a low-copy-number vector and named pPR2002. Plasmid pPR2002 was transformed into purR+ and purR strains, and ,-galactosidase activities were measured from cells grown in the presence and absence of adenine ( Table 2). The data demonstrate a 2.5to 2.7-fold repression of purR-lacZ in the purR+ strain. Essentially no regulation was obtained in the purR strain. Thus, purR expression is autoregulated.
Repressor-operator interaction. Gel retardation assays were performed to determine whether the Pur repressor can interact with either one or both of the operatorlike sequences. A BalI-to-HincII fragment containing purR' DNA from nucleotides 20 to 321 was cloned into the SmaI site of the pUC119 polylinker and used for binding experiments. This DNA fragment was excised as a 324-bp EcoRI-to-HindIIT segment, end labeled, and incubated with extract prepared from a purR+ strain. The results of repressor binding to the labeled DNA are shown in Fig. 2A. After electrophoresis, DNA was found in three positions: the band with the fastest mobility, which was free DNA, plus two positions with protein-DNA complexes Bi and B2. With increasing protein there was initial formation of band Bi and subsequent formation of band B2 with a slower mobility than Bi. The binding curves in Fig. 3 show the concentration dependence for conversion of free DNA to complexes Bi and B2. The half-maximal formation of Bl required 1.1 ,ug of protein, whereas the half-maximal formation of B2 required approximately 6.9 jig of protein.
No binding was obtained with extract from a purR mutant (data not shown). The results in Fig. 2A suggest that binding of the repressor to one operator site results in a complex that migrates as band Bi, whereas binding of repressors to 01 and 02 forms a protein-DNA complex that migrates as band B2.
To determine whether there are different affinities of the repressor for 01 and 02 and to determine whether the repressor initially interacts with 01 or with 02, binding assays were conducted with "half' fragments containing only one operator. Fragment 1 contained 01 (Fig. 2B), and fragment 2 contained 02 (Fig. 2C). Saturation of 01 and 02 with repressor is shown in Fig. 2B and C and is quantitated in Fig. 3. The saturation curves in Fig. 3 show that the concentration dependence for binding of the repressor to operator 01 was similar to the concentration dependence for band Bi and identifies band Bi as the repressor-Ol complex in the DNA fragment containing 01 and 02. The concentration dependence for saturation Of 02 in fragment 2 was similar to the formation of B2 in DNA containing 01 and 02. Thus, the Pur repressor initially saturates 01, followed by saturation of 02. In this analysis there is no evidence for cooperativity for binding of the repressor to sites 01 and 02- The gel retardation assays presented in Fig. 2B and C show decreased electrophoretic mobility for fragments 1 and 2, respectively, incubated with 10 or 20 ,ug of protein. It is likely that this decreased mobility is due to protein-protein or nonspecific protein-DNA interactions, since no additional protected regions appeared in a DNase I footprint (see below).
Binding of the repressor to the purF operator was also determined. Saturation curves for binding of the repressor to a DNA fragment containing the purF operator and to the fragment containing 0102 were superimposable, as measured by the disappearance of free DNA (data not shown). Saturation of fragment 1 (95% binding) containing 01 required 10 ,ug of repressor compared with 5 ,ug of repressor required for binding to 0102 or PurFo. Thus two purR operators were required to obtain repressor binding that was identical to that of the purF operator. As shown below, this requirement for both purR operators is not due to cooperative binding but reflects the availability of two binding sites for formation of a mixed population of complexes: repressor-01, repressor-02, (repressor)2-0102. Protection of operator DNA by Pur repressor. DNase I footprinting assays were performed to verify that binding of the repressor is specific for operator sites 01 and 02 and to search for evidence for interaction between operator sites. Sites 01 and 02 were saturated with increasing concentrations of repressor before treatment with DNase I. Figure 4A shows the protection of the noncoding (upper) strand, and Fig. 4B shows the protection of the coding (lower) strand. These results demonstrate clearly that the repressor initially binds to 01 and at higher concentrations saturates 02. The boundaries for DNase I protection extend 3 to 5 nucleotides beyond the conserved operator sites and are indicated by brackets under the sequence in Fig. 4C. One boundary was difficult to define due to poor digestion by DNase I on both strands. This boundary is indicated by a dashed bracket in Fig. 4C. These results demonstrate that the repressor binds to 01 and 02-We could detect no other sites of specific interaction of the Pur repressor with the DNA fragment.
There is no evidence in the DNA footprint for DNA bending that might result from interaction between repressor molecules bound at sites 01 and 02-Interaction between sites due to loop formation has been recognized by patterns of alternating DNase I-hypersensitive and -insensitive regions with a periodicity of about 10 bases reflecting the helical rotation (24). Loops are generally formed when the binding sites are separated by an integral number of helical turns (9,30). The separation of the two purR operators is 88 nucleotides from dyad center to dyad center. This converts B2. From a second experiment ( Fig. 2C and D), the percentages of the two half-fragments found in the DNA-protein complex are plotted: A, EcoRI-NdeI fragment; *, NdeI-HindIII fragment.
to 8.5 helical turns with a periodicity of 10.4 bp per turn. According to this analysis, repressors bound at operators 01 and 02 act independently of each other. Operator mutations. The protein-DNA binding experiments indicate that the Pur repressor binds with a higher affinity to 01 than to 02, yet the position of 02 within the coding sequence suggests that it has an in vivo role. To dissect the effects of each operator on in vivo regulation, mutations were made within each operator sequence. A comparison of pur regulon operator sequences from the genes purF, purMN, purL, purC, purEK, purHD, and purR (Fig. 5) revealed a number of invariant positions and highly conserved positions. Mutations were made in 01 and 02 to abolish repressor binding. The wild-type and mutant operators are shown in Fig. 1. Plasmid pPR2004 contains a three-base substitution in positions 9 through 11 of 01: G104A, T1O5C, and T106A. This mutation changes one invariant position, G-9, and the two conserved T residues at positions 10 and 11 (operator numbering is from 1 to 16 from left to right). The T105C and T106A replacements mimic a purF operator-constitutive mutation (41). Plasmid pPR2005 contains a four-base mutation in positions 11 through 14 of 02: T194C, T195A, C196G, and C197T. This change was designed to maintain the purR coding sequence at these positions yet alter two invariant positions: T-12 and C-14. Finally, a double mutant was constructed that contains replacements in both operators. Repressor binding assays with DNA fragments that contained single operator mutations demonstrated gel retardation to only a single position (data not shown). DNA fragments carrying mutations in either of the two operator sequences bound the repressor with identical affinity to the half-fragments reported above (data not shown). These results verify that each of the mutations eliminated the ability of the repressor to bind to mutant operator.
Effect of operator mutations on autoregulation ofpurR. The purR-lacZ fusion in pPR2002 was recreated in plasmids pPR2004, pPR2005, and pPR2006 with mutations in 01, 02, and 0102, respectively. These plasmids were transformed into purR+ and purR strains, and P-galactosidase levels were measured from cells grown in the presence and absence of a repressing level of adenine (Table 2). Cells carrying pPR2005, 0102-7 showed somewhat less repression than that of the wild-type strain (1.9to 2.0-fold versus 2.5to 2.7-fold). Repression was further decreased in the strain carrying pPR2004, 01-02. However, the 1.5to 1.6-fold repression of purR-lacZ in this strain demonstrated the in vivo function of operator site 02. The double operator mutant pPR2006 showed little or no repression. In each case, repression of purR-lacZ was essentially abolished in a purR strain, indicating that it is the Pur repressor that is mediating the response to adenine. mRNA levels. To investigate whether repression of purR expression was by a transcriptional or posttranscriptional mechanism, we determined purR-lacZ mRNA levels in purR+ and purR strains. RNA was pulse-labeled with [3H]uridine, and purR-lacZ mRNA was determined by DNA-RNA hybridization. Values for mRNA synthesis, uncorrected for decay, are summarized in Table 2. The range of values for repression of mRNA synthesis was calculated by the method used for enzymes. These results demonstrate that repression of mRNA synthesis by the Pur repressor generally paralleled the repression of enzyme formation. Maximal repression of 1.6to 2.2-fold was obtained in the wild type (0102), with marginally less repression from 0102 and Ol02purR-lacZ genes. These data demonstrate transcriptional regulation of purR by binding of the Pur repressor at operators 01 and 02. Although these results do not exclude low-level translational control, transcriptional regulation is sufficient to account for the modulation of enzyme production.

DISCUSSION
The results presented here demonstrate that the Pur repressor autogenously regulates expression of purR. Meng et al. (34) recently reported similar findings. This work supports their conclusions and extends our understanding of this autoregulation from studies of the isolated operators. Two species of DNA-protein complexes were detected in each study. We have identified the components of each of the complexes. At low repressor levels, the primary interaction with operator DNA occurs at 01 and causes a mobility shift of free DNA to Bi. Thus, Bi is not a mixture of repressor-Ol and repressor-02 complexes; Bi is largely repressor-O1 complex. Binding of additional repressor to Bi saturates 02 and results in a shift to B2. Our work demon- strates that interaction of the repressor with 01 and 02 iS noncooperative. Binding of the repressor to either 01 or 02 in vitro is independent of the state of the second operator and is only dependent on protein concentration. In some cases, such as with lac 01/03, loop formation only occurs in supercoiled DNA (see reference 16 for a review). Although we have not excluded the possibility that purR 0102 loops may form in vivo, the oligomeric state of the Pur repressor does not favor DNA looping. The Pur repressor has a molecular weight intermediate between the monomer and the dimer, suggestive of a monomer-dimer equilibrium (42a), in contrast to tetrameric lac repressor. Meng et al. (34) detected a third repressor-operator species by using a gel retardation assay. We did not identify a third species of reduced mobility, but we obtained smearing at high protein concentrations that was ascribed to protein-protein interaction or nonspecific protein-DNA binding. The footprint analysis ( Fig. 4A and B) clearly shows that the Pur repressor preferentially saturates 01 before 02. These results extend our understanding of repressor-operator complex formation by showing different operator affinities for the repressor and demonstrate the noncooperative nature of in vitro binding.
The mutational analysis of operator function indicates an in vivo role for each operator. The full level of repression attained in the wild-type purR-lacZ fusion relies on both operators, not just 01. This is in spite of the closer position of 01 to the promoter and its greater affinity for the Pur repressor. The unique position of operator sequences relative to the promoter suggested the possibility of translational regulation. However, the close correspondence between the response of mRNA and enzyme to repression by purR+ and excess adenine provides evidence for transcriptional regulation. Although values for the lower repression of mRNA relative to enzyme ( Table 2) leave open the possibility for some translational regulation, for the following reasons we favor the conclusion that most or all regulation is transcriptional. (i) Translational regulation should require some type of interaction between repressors bound to 01 and 02 or should require at least operator 02, which is located in the coding sequence. However, we can find no evidence for an 01-02 interaction in vitro or in vivo, nor can we find an absolute requirement for 02 Substantial repression ofpurR-lacZ occurred in each of the mutants possessing only one operator. (ii) Small differences in RNA levels are difficult to quantitate. For this reason the somewhat lower values for repression of mRNA relative to enzyme are unlikely to be significant. The work of Meng et al. (34) was confined to autoregulation of wild-type purR in a lacZ transcriptional fusion and did not include measurements of mRNA.
The expression and regulation of the purR gene appear unique in two respects. First, the 5' end of the mRNA has been mapped and corresponds to a thymidine residue 155 nucleotides upstream of the ATG initiation site (42). Most E. coli transcripts begin with a purine nucleotide, usually an adenine (22). Sequences upstream of the mapped mRNA 5' end show poor homology with c70 promoter consensus elements (22). Although there is a very good promoter consensus sequence located between nucleotides -50 and -85, corresponding mRNA 5' ends were not detected. Second, the location of the two purR operators is unusual. Both purR operators lie downstream of the promoter. Repressors that bind within the promoter are thought to function by preventing formation of closed or open transcription complexes (39). We propose that the large difference in repression ofpurF and purR by the Pur repressor reflects the location of operator sites. The Pur repressor regulates the expression of a purF-lacZ fusion by approximately 28-fold (41), compared with 2to 3-fold for repression of purR-lacZ, notwithstanding that in vitro binding of the repressor to the two control regions is indistinguishable. We conclude that the mechanism for repression of purR expression is much less efficient than that for purF. Repression of purR may occur by inhibition of transcription elongation rather than at the step of transcription initiation as in purF. The advantages to the cell of this mechanism of repression of purR are not presently understood.
There are several examples for multiple operator sites in gene regulation (for reviews, see references 16 and 38). There are two operators in both the lac and gal operons. In these cases one operator site is upstream of (8) or overlapping (13) the site for transcription initiation, and the second is in the protein-coding region (12,27,40); repressor binding is cooperative, and DNA loops are formed (18,26,30). Gralla (reviewed in reference 16) has proposed two roles for the lac 02 operator element within the lacZ gene. It contributes to repression of transcription initiation by enhancing the interaction of the repressor bound to 01 via loop formation and also is thought to block elongation of lac transcription that has escaped repression at 01. Alternative loop formation also plays a role in the regulation of araBAD and araC expression by the AraC regulatory protein (9,20). AraC protein bound to araO2 operator located within araC can interact with AraC bound either at araO1 or araI (20). Both of these latter operators are located in or near the promoters PC and PBAD. Single and double loop formation has also been suggested to explain the regulation of the deo operon by DeoR (7,19). In these examples, repressors bind to an operator site in or near the promoter, and binding to a second operator is stabilized through loop formation (16). Of the four bacterial repressors that utilize two operator sites, only araC is autoregulated; gaiR, lacI, and deoR are not. Gene trpR is autoregulated by the Trp repressor, which binds to an operator site overlapping the -10 and + 1 regions of the promoter (3,17,28). The regulation ofpurR by the Pur repressor is distinct from these examples in both operator location and in lack of loop formation. The regulation of purR expression appears to be the first example of autoregulation that occurs by binding to two operator sequences, both downstream of the promoter. The Pur repressor is now known to regulate a number of genes or operons involved in de novo synthesis of purine and pyrimidine nucleotides. These include purF, purMN, purL, purC, purHD, purEK, guaBA, pyrC, and pyrD (5,23,34,52,53). In each case, except for guaBA and pyrD, the Pur repressor has been shown to interact with a single operator site located in the promoter-transcription initiation region. Binding of the Pur repressor to guaBA and pyrD control sites has not been reported. Of all the genes thus far known to be regulated by the Pur repressor, only purR employs a twooperator mechanism.