Transcriptional Repressor CopR: Use of SELEX To Study the copR Operator Indicates that Evolution Was Directed at Maximal Binding Affinity

Abstract

CopR is one of the two copy number control elements of the streptococcal plasmid pIP501. It represses transcription of the repR mRNA encoding the essential replication initiator protein about 10- to 20-fold by binding to its operator region upstream of the repR promoter pII. CopR binds at two consecutive sites in the major groove of the DNA that share the consensus motif 5′-CGTG. Previously, the minimal operator was narrowed down to 17 bp, and equilibrium dissociation constants for DNA binding and dimerization were determined to be 0.4 nM and 1.4 μM, respectively. In this work, we used a SELEX procedure to study copR operator sequences of different lengths in combination with electrophoretic mobility shift assays of mutated copR operators as well as copy number determinations to assess the sequence requirements for CopR binding. The results suggest that in vivo evolution was directed at maximal binding affinity. Three simultaneous nucleotide exchanges outside the bases directly contacted by CopR only slightly affected CopR binding in vitro or copy numbers in vivo. Furthermore, the optimal spacer sequence was found to comprise 7 bp, to be AT rich, and to need an A/T and a T at the 3′ positions, whereas broad variations in the sequences flanking the minimal 17-bp operator were well tolerated.

Replication of the streptococcal plasmid pIP501 is regulated by two components that act in concert: the transcriptional repressor CopR (10.6 kDa) and the antisense RNA RNAIII (136 nucleotides [nt]) (5). Whereas RNAIII exerts its inhibitory effect by premature termination (attenuation) of the essential repR mRNA (6, 8), CopR has a dual function. On the one hand, it represses transcription from the essential repR promoter pII about 10- to 20-fold (7); on the other hand, it prevents convergent transcription from pII and pIII (antisense promoter), thereby indirectly increasing transcription initiation at pIII (9). Previously, it was found that CopR contacts the DNA asymmetrically at two consecutive major grooves that share the consensus motif 5′-CGTG (28). Thus, the outermost G residues were found to be most important for CopR binding, whereas exchanges of nucleotides adjacent (3′) to the CGTG motif only slightly altered DNA binding. The operator sequence was narrowed down to 17 bp. Furthermore, it was found that CopR binds exclusively as a dimer, and the equilibrium dissociation constants for the CopR dimers and the CopR-DNA complex were calculated to be 0.4 nM and 1.4 μM, respectively (29). A three-dimensional model of the N-terminal 63 amino acids of CopR was built and was used to identify amino acids involved in DNA binding and dimerization (30, 31, 32). By this means, it was found that amino acids R29 and R34, located in the recognition helix (helix III) of the helix-turn-helix motif, make specific contacts to the DNA at G240 (binding site I) and G254 (binding site II) or G242/T243 (site I) and G251 (site II), respectively. Water-mediated contacts were suggested for E35 interacting with the outermost C residues in both binding sites. Unspecific DNA contacts via the sugar-phosphate backbone were proposed for K10 in α-helix I and S28 in the recognition helix (30). Furthermore, it was established that the structured acidic C terminus of CopR that forms a β-strand is necessary for stabilization of the protein (22, 23). A fluorescence energy study revealed that CopR bends the operator DNA slightly (20 to 25°) upon binding, and it was proposed that two pyrimidine-purine dinucleotide steps in the operator sequence that are separated by one helical turn are required for bendability (33). With all these data on hand, we asked whether the copR operator found in nature (on plasmid pIP501) was optimized for strong DNA binding or whether it would be possible to find an operator that is bound more efficiently by CopR and, if so, how such an operator would function in copy number control in vivo.

In 1990, the SELEX procedure was developed independently by two laboratories (13, 36). This procedure uses randomized sequences (DNA or RNA) to select for different criteria, like binding of certain proteins or small metabolites or the ability to perform enzymatic reactions, followed by PCR amplification and further rounds of selection. SELEX not only yielded impressive results that supported the RNA world hypothesis, e.g., selection of RNA molecules that were able to carry out templated RNA polymerization (21) or amino acylation of tRNAs (24) or selection of high-affinity RNA ligands to parasite target molecules (18), but was also employed successfully to study a broad variety of protein-DNA interactions. Examples of the uses of SELEX include the in vitro selection of binding sites for the Escherichia coli trp repressor (12), methionine repressor MetJ (19), integration host factor (IHF) (17), a response regulator of Bradyrhizobium japonicum (14), and bacteriophage Ff gene 5 protein, a single-strand binding protein (37). Consensus sequences for a UP element for bacterial promoters (15) or for embryonic heat shock factor 2 (25) were defined. Furthermore, sequence requirements for efficient termination of conjugation in the oriT gene of E. coli plasmid R1162 were determined (2), and the promoter discrimination between σ^s and σ⁷⁰ RNA polymerases of E. coli was investigated (16). SELEX revealed an unusual DNA binding mode for TRF1, a key player of telomere length regulation (3). In the case of proteins that do not seem to recognize strongly defined consensus sequences, like topoisomerase II, in vitro evolution identified preferred DNA cleavage sites (10). Additionally, SELEX was used for applicative purposes like the selection of DNA aptamers against human immunodeficiency virus type 1 RNase H that display in vitro antiviral activity (1). All these examples demonstrate the power of in vitro selection for the analysis of DNA-protein interactions.

To answer the questions mentioned above, we applied the SELEX procedure with copR operator sequences of different lengths in combination with electrophoretic mobility shift assays (EMSAs) with mutated operator fragments, copy number determinations, and in vitro transcription. Our results demonstrate that in vivo evolution of the copR operator sequence was directed at maximal binding affinity. Furthermore, we defined sequence and length requirements for the spacer and regions adjacent to the two binding sites.

MATERIALS AND METHODS

DNA preparation, manipulation, and copy number determination.

Plasmid DNA was isolated from Bacillus subtilis as reported previously (5). DNA manipulations such as restriction enzyme cleavage and ligation were carried out under the conditions specified by the manufacturer or according to standard protocols (26). A PCR kit from Roche was used for PCR amplifications. DNA sequencing was performed according to the dideoxy chain termination method (27) with a Sequenase kit from Amersham Bioscience. Copy numbers of pIP501 derivatives in B. subtilis were determined as described previously (5), except that gel photographs were scanned and band intensities were quantified with the PCBAS 2.0 program.

Construction of E. coli-B. subtilis shuttle vectors containing mutations in the copR operator.

Plasmid pPRC333 containing the wild-type copR operator region was constructed as follows. First, a PstI site was created at position 582 (4) to facilitate the subsequent construction of mutations in the leader region; a BamHI/PstI fragment spanning nt 160 to 582 was obtained by PCR on pPR1 as a template by using the primer combinations shown below and inserted into the pUC19 BamHI/PstI vector, resulting in plasmid pUC333. The BamHI/PstI fragment of plasmid pUC333 and the PstI/EcoRI fragment of plasmid pUCR3 (20) were jointly cloned into the pPR4 BamHI/EcoRI vector (5), yielding plasmid pPRP333. Subsequently, the copR gene was inserted as a 549-bp EcoRI fragment derived from plasmid pCOP1B2 (7) into the unique EcoRI site of plasmid pPRP333, and the plasmid containing copR in the same direction as the repR gene was designated pPRC333. All PCR-generated fragments were confirmed by sequencing.

Mutated operator sequences were constructed by the same procedure using the following primers in combination with primer SB214 (5′-TAG AAG CTA CGA TCA AAG TTG AA): pPRC333-SB333 (5′-AATTGGATCCGATTTCGTGTGAATAATGCA), pPRC334-SB334 (5′-AATTGGATCCGATTTCGTGCGAATAATGCACGAAATCATT), pPRC221-SB221 (5′-AATTGGATCCAAAAGCAATGATTTCGTGTCCCCCCCGCACGAAATCATTGCTTAT), pPRC222-SB222(5′-AATTGGATCCAAAAGCAATGATTTCGTGTGAAAAAAAGCACGAAATCATTGCTTAT), pPRC227-SB227 (5′-AATTGGATCCGATTTCGTGTGAAAAATGCACGAAATCATTGCTTAT), pPRC228-SB228 (5′-AATTGGATCCGATTTCGTGTGAATTAATGCACGAAATCATTGCTT), pPRC292-SB292 (5′-AATTGGATCCGATTTCGTGTGGGGGGGCACGAAATCATTGCTTAT), pPRC294-SB294 (5′-AATTGGATCCGATTTCGTGTGAATAATACACGAAATCATTGCTTA), and pPRC416-SB416 (5′-AATTGGATCCGATTTTGTGCATGTATTGGACGAAATCATTGCTTATTTT).

Construction of lacZ fusion vectors and determination of β-galactosidase activity.

Plasmids pUC333 (wild type), pUC334 (symmetric operator), and pUC221 (spacer with 7C) were used for the isolation of EcoRI/HindIII blunt fragments comprising promoter pII with the upstream copR operator region, promoter pIII, attenuator, and 130 bp downstream (Table 1). These fragments were inserted into the EcoRI/BamHI blunt vector pAC6 (34) to generate transcriptional fusions with the promoterless lacZ gene. The resulting vectors, pAC333, pAC334, and pAC221, were linearized with ScaI, and the corresponding lacZ fusions were integrated into the amyE locus of the B. subtilis chromosome of strain DB104 by double crossover. To provide CopR in trans, the corresponding integrant strains were transformed with plasmid pCOP9 (5), and Cm^r Pm^r transformants were selected. These strains were used for the determination of β-galactosidase activity as described previously (7).

TABLE 1.

Plasmids used in this study

Plasmid	Descriptiona	Reference or Source
pUC19	E. coli cloning vector; Apr; MCS	26
pPR1	Shuttle vector; pIP501 derivative; Apr Pmr	5
pCOP9	pPR1 derivative with copR gene and defect in pIII	5
pCOP1B2	pPR1 with copR gene downstream from oriR; Pmr	7
pUCR3	pUC19 derivative comprising the pUCR1 PstI/Hind fragment and the pPR1 EcoRI/HindIII fragment	20
pTYB11	Vector for expression of N-terminal fusion protein with chitin binding domain and intein tag; Ampr	New England Biolabs
pTYBC11	pTYB11 with copR gene as SapI/PstI fragment	This study
pUC333	pUC19 derivative comprising nt 229-580 of the pIP501 replicon as BamHI/PstI fragment	This study
pUC334	As pUC333, but with T243C	This study
pUC221	As pUC333, but with 7C spacer	This study
pUC222	As pUC333, but with 7A spacer	This study
pUC292	As pUC333, but with 7G spacer	This study
pUC294	As pUC333, but with G512A	This study
pUC227	As pUC333, but with T247A	This study
pUC228	As pUC333, but with 8-bp spacer	This study
pUC416	As pUC333, but with C239T, T243C, C252G	This study
pPRP333	pPR4 derivative carrying the pUC333 BamHI/PstI fragment and the pUCR3 PstI/EcoRI fragment	This study
pPRC333	pPRP333 with a 549-bp copR fragment from pCOP1B2	This study
pPRC334	pPRC333 derivative with pUC334 mutation	This study
pPRC221	pPRC333 derivative with pUC333 spacer mutation	This study
pPRC222	pPRC333 derivative with pUC221 spacer mutation	This study
pPRC227	pPRC333 derivative with pUC227 spacer mutation	This study
pPRC228	pPRC333 derivative with pUC228 spacer mutation	This study
pPRC292	pPRC333 derivative with pUC292 spacer mutation	This study
pPRC294	pPRC333 derivative with pUC294 mutation	This study
pPRC416	pPRC333 derivative with pUC416 mutation	This study
pAC6	Vector for intergration of transcriptional lacZ fusions	34
pAC333	pAC6 with wild-type EcoRI/BamHI fragment (nt 229-362)	This study
pAC334	pAC6 with KS9 EcoRI/BamHI fragment (nt 229-362)	This study
pAC221	pAC6 with 7C EcoRI/BamHI fragment (nt 229-362)	This study

^aNucleotide numbering is according to Brantl et al. (4). MCS, multiple cloning site.

Construction of vector pBTYC11 for overexpression and purification of native CopR.

A promoterless copR gene generated by PCR with primers SB203 (5′-GGT GGT TGC TCT TCC AAC ATG GAA CT GCA TTT AGA GAA) and C-951-30 (5′-GAA TTC CTG CAG TCA CAC GAA ATC ATT GCT) on plasmid pCOP7C as a template and subsequently digested with SapI and PstI was inserted into vector pTYB11 (New England Biolabs) digested with the same pair of enzymes. E. coli strain TG1 was used for transformation of the ligation mixture, and Amp^r transformants were screened for the presence of recombinant pTYB11. The resulting vector was designated pTYBC11, and the inserted copR gene was confirmed by sequencing. In pTYBC11, the N-terminal codons of CopR are fused to the intein tag and hence the C-terminal codons of the chitin binding domain. Expression strain ER2566 (IMPACT-CN protein purification system; New England Biolabs) was transformed with pTYBC11 and used for the overexpression of native CopR.

Preparation of labeled wild-type and mutant CopR targets.

Oligodeoxyribonucleotides listed in Table 6 were 5′ end labeled with [γ-³²P]ATP (26) and purified from 8% denaturing polyacrylamide gels. Double-stranded CopR targets were generated in a Klenow reaction with oligodeoxyribonucleotide SB176 (5′-CCC CTT AAA AAA ATA AGC) and, in the case of SB414, SB417, and SB418, oligodeoxyribonucleotide SB415 (5′-TCGCTGAACATTCGATCTA) as primers.

TABLE 6.

Oligodeoxyribonucleotides used in the gel shift analyses

Oligodeoxyribonucleotide	Sequencea
KS1	5′-GGGGAAAAGCAATGATTTCGTGTGAATAATGCACGAAATCATTGCTTATTTTTTTAAGGGG-3′
KS9	5′-GGGGAAAAGCAATGATTTCGTGCGAATAATGCACGAAATCATTGCTTATTTTTTTAAGGGG-3′
KS3	5′-GGGGAAAAGCAATGATTTCGTGTGAATAATACACGAAATCATTGCTTATTTTTTTAAGGGG-3′
SB174	5′-GGGGAAAAGCAATGATTTCGTGTTAATAATGCACGAAATCATTGCTTATTTTTTTAAGGGG-3′
SB175	5′-GGGGAAAAGCAATGATTTCGTGTGAAAAATGCACGAAATCATTGCTTATTTTTTTAAGGGG-3′
SB180	5′-GGGGAAAAGCAATGATTTTGTGTGAATAATGCACGAAATCATTGCTTATTTTTTTAAGGGG-3′
SB183	5′-GGGGAAAAGCAATGATTTCGTGTAAAAAAAGCACGAAATCATTGCTTATTTTTTTAAGGGG-3′
SB184	5′-GGGGAAAAGCAATGATTTCGTGTTTTTTTTGCACGAAATCATTGCTTATTTTTTTAAGGGG-3′
SB187	5′-GGGGAAAAGCAATGATTTCGTGTGGGGGGGGCACGAAATCATTGCTTATTTTTTTAAGGGG-3′
SB188	5′-GGGGAAAAGCAATGATTTCGTGTGAAAATGCACGAAATCATTGCTTATTTTTTTAAGGGG-3′
SB189	5′-GGGGAAAAGCAATGATTTCGTGTGAATTAATGCACGAAATCATTGCTTATTTTTTTAAGGGG-3′
SB190	5′-GGGGAAAAGCAATGATTTCGTGTCCCCCCCGCACGAAATCATTGCTTATTTTTTTAAGGGG-3′
SB198	5′-GGGGAAAAGCAATGATTTCGCGTGAATAATGCACGAAATCATTGCTTATTTTTTTAAGGGG-3′
SB199	5′-GGGGAAAAGCAATGATTTCGTATGAATAATGCACGAAATCATTGCTTATTTTTTTAAGGGG-3′
SB414	5′-TAGCTTAGGCAGTCACATGAAACTGTGCATGTATTGGACGAATTAGATCGAATGTTAGCGA-3′
SB417	5′-TAGCTTAGGCAGTCACATGAAACTGTGCATGTATTGCACGAATTAGATCGAATGTTAGCGA-3′
SB418	5′-TAGCTTAGGCAGTCACATGAAACCGTGCATGTATTGGACGAATTAGATCGAATGTTAGCGA-3′

^aBoldface type indicates binding sites I and II.

Overexpression and purification of His₆ CopR and native CopR.

Plasmid pQEC1 was used for overexpression and purification of His₆ CopR as described previously (28, 29). For the preparation of native CopR, strain ER2566 (pTYBC11) was used. After induction with IPTG (isopropyl-β-d-thiogalactopyranoside) at an optical density at 560 nm of 1.0, the strain was grown at 12°C overnight until an optical density at 560 nm of 1.8 to 1.9 was reached, to prevent the accumulation of insoluble protein. Afterwards, cells were pelleted and sonicated. The supernatant was centrifuged at 4°C for 10 min at 13,000 rpm in a Beckman J2-21M centrifuge, bound with chitin for 40 min at 4°C with constant stirring, and subsequently filled into a column. After two washing steps, self-cleavage of the fusion proteins by the intein domain was induced by the addition of 50 mM dithiothreitol in a buffer containing 500 mM NaCl and 20 mM Tris-HCl (pH 9.0) and incubation at room temperature for 16 h. Afterwards, the native CopR protein was eluted with 500 mM NaCl and 20 mM Tris-HCl (pH 8.0), and the first two 250-μl fractions containing native CopR were stored with glycerol (final concentration, 50%) at −20°C.

CopR-DNA binding reaction and band shift assay.

Binding reactions were performed in a final volume of 20 μl containing 0.5× Tris-borate-EDTA (TBE) buffer (pH 8.0), 0.9 nM of end-labeled DNA fragment, and 3 to 150 nM His₆ CopR. Herring sperm DNA (0.1 μg/μl) was added as a nonspecific competitor. After incubation at 30°C for 30 min, the reaction mixtures were separated on 8% native polyacrylamide gel runs at room temperature for 1.5 h (16 V/cm) in 0.5× TBE buffer. Visualization and quantification of the bands were performed on a Fuji PhosphorImager. In some cases, 75 mM NaCl was included in the reaction mixture, the gel, and the electrophoresis buffer.

In vitro selection (SELEX) procedure.

To generate double-stranded templates for SELEX I, SELEX II, and SELEX IV, between 8 and 333 pmol of the following 61-bp oligodeoxyribonucleotides containing random sequences flanked by fixed regions were used as templates in a primer extension reaction employing an 18-bp primer, SB179 (5′ GAT GCA TGG ATC CAT GAT), complementary to the 3′ end of randomized DNA pools: SB206 (I) [5′-ACAGGAAACAGCTATGACCATGATTACGCCGATGGAATTCAAGCTTAATGATTTCGTGT(N₇)GCACGAAATCATGGATCCATGCATCACTGGCCGTCGTTTTACAACGTC-GTGACTG], SB265 (II) [5′-ACAGGAAACAGCTATGACCATGATTACGCCGATGGAATTCAAGCTTAATGATTT(N₁₇)AAATCATGGATCCATGCATCACTGGCCGTCGTTTTACAACGTCGTGACTG], SB369 (III) [5′-TACGGTAACTGGACTGCATAACGATGCATTTGACTCATTCAAGCTTCATCCATA(N₃₀)TAGTCGTGGATCCTTGACATGACAGGTATGTAGTCATAAGCACTTAGCAA], and SB330 (IV) [5′-ACAGGAAACAGCTATGACCATGATTACGCCGATGGAATTCAAGCTTAAT(N₅)CGTGTGAATAATGCACG(N₅)ATGGATCCATGCATCACTGGCCGTCGTTTTACAACGTC-GTGACTG].

For construction of the double-stranded randomized DNA pool for SELEX III, instead of SB179, primer SB370 (5′-ATG TCA AGG ATC CAC GAC) was used. Subsequently, the DNA pools were amplified by PCR with primers SB223 (5′-CAG TCA CGA CGT TGT AAA ACG ACG GCC AGT GAT GCA TGG ATC CAT GAT) and SB224 (5′-ACA GGA AAC AGC TAT GAC CAT GAT TAC GCC GAT GGA ATT CAA GCT TAA TG) in the cases of SELEX I, II, and IV and primers SB371 (5′-TAC GGT AAC TGG ACT GCA TAA CGA TGC ATT TGA CTC ATT CAA GCT TCAT C) and SB372 (5′-TTG CTA AGT GCT TAT GAC TAC ATA CCT GTC ATG TCA AGG ATC CAC GAC) in the case of SELEX III. Primers carry an overhang to obtain longer PCR products (121 bp for SELEX I, II, and IV and 134 bp for SELEX III). After phenol-chloroform extractions and ethanol precipitation, the PCR products were 5′ end labeled with 10 μCi of [γ-³²P]ATP. The radioactively labeled DNA fragments were purified from 8% native polyacrylamide gels, visualized by phosphorimaging, excised, eluted two times in elution buffer containing 1 mM EDTA (pH 8.0), 500 mM NaAc, 10 mM MgAc, and 0.1% sodium dodecyl sulfate for 1 h at 50°C, and precipitated with ethanol afterwards.

Binding reactions were performed in a final volume of 20 μl containing the labeled DNA fragment and 84 nM His₆ CopR in 0.5× TBE buffer. In all cases, incubation without His₆ CopR was used for comparison. As a reference for the excision of the shifted CopR target, which was not visible in the first round of selection with SELEX II and III, primers with wild-type operator sequence but that were the same length as the SELEX primer were used in each SELEX experiment and treated in the same way (Klenow reaction, PCR amplification, labeling, and EMSA). After round 3 of each SELEX procedure, a 0.6 μM wild-type DNA fragment (KS1) (28), which has no binding sites for amplification primers, was added as a competitor in each binding reaction to promote selection towards a high-affinity and high-specificity pool. After 30 min at 30°C, bound and unbound DNA species were separated on 8% native polyacrylamide gels at 230 V. Band shifts were detected by phosphorimaging. The bound species were excised and eluted in elution buffer (see above) followed by phenol-chloroform extractions to remove the CopR protein and ethanol precipitation.

The recovered bound ligand sequences were dissolved in water and subsequently PCR amplified by using primers SB225 (5′-CAG TCA CGA CGT TGT AAA) and SB226 (5′-ACA GGA AAC AGC TAT GAC) for SELEX I, SELEX II, and SELEX IV products and primers SB372 (5′-TAC GGT AAC TGG ACT GCA) and SB374 (5′-TTG CTA AGT GCT TAT GAC) for SELEX III products. Twenty cycles of PCR amplification were performed for 30 s each at 95, 52, and 72°C. PCR products were phenol-chloroform extracted and precipitated with ethanol followed by 5′ labeling as described above. These steps were repeated 10 times. After the 10th round of SELEX, amplification products were digested with BamHI and HindIII and inserted into the pUC19 BamHI/HindIII vector. After transformation of E. coli strain TG1, the DNA of individual white transformants was sequenced.

In vitro transcription with B. subtilis RNA polymerase.

Linear templates for in vitro transcription were generated by PCR from the corresponding pUC derivatives pUC333, pUC334, pUC294, pUC292, pUC221, and pUC228 with primer SB214 and the universal sequencing primer and gel purified. In vitro transcription assays were performed in a final volume of 20 μl containing 60 mM Tris HCl (pH 7.8), 12 mM MgCl₂, 1 mM dithiothreitol, and 1 ng of different linear DNA templates as well as 200 μM (each) ATP, GTP, and CTP; 20 μM UTP; 5 μCi of [α-³²P]UTP; and 240 ng of native CopR. After incubation at 30°C for 15 min, 0.5 μl of B. subtilis RNA polymerase (0.24 μg) was added, and the incubation continued at 30°C for 30 min. Transcription was stopped by phenol-chloroform extraction and ethanol precipitation, and the products were dissolved in water and 50% formamide loading dye, heat denatured for 5 min at 95°C, and separated on a 6% denaturing polyacrylamide gel. Dried gels were analyzed and quantitated in a Fuji PhosphorImager.

RESULTS AND DISCUSSION

Use of SELEX to study the spacer region reveals two consensus positions at the 3′ end.

To answer the question of whether there are any sequence preferences in the spacer region between the two CopR binding sites, in a first SELEX approach (SELEX I), an operator sequence that was randomized at the 7-bp spacer region but that contained wild-type binding sites I and II was used. The SELEX experiment was started with 8 pmol of the randomized sequence, which corresponds to ≈2.9 × 10⁹ copies of every possible sequence, and this pool was amplified by PCR prior to labeling as described in Materials and Methods. As expected, a shifted band was already visible in round 1 (Fig. 1). After round 10, the selected DNA fragments were amplified by PCR, digested with BamHI and HindIII, and inserted into the pUC19 vector. Twenty-eight clones were sequenced, and the results are shown in Table 2. In all but one case, spacers were found with an A or T at position 6, and in all but two cases, a T in position 7 flanked by otherwise random sequences in positions 1 to 5 was found. However, all spacers were AT rich, with only three of them having more than two G or C residues.

FIG. 1.

In vitro selection (SELEX) with four different randomized sequences. Autoradiograms of the EMSAs in the 10th round of in vitro selection are shown for all four randomized targets used in the SELEX I to IV experiments. Bound DNA fragments were excised and, after removal of CopR, were PCR amplified and subcloned into pUC19 to obtain individual operator fragments for sequencing as described in Materials and Methods.

TABLE 2.

Sequences selected with SELEX I (N₇)

Clonea	Sequenceb
RS	5′-AAGCTTAATGATTT CGTGT(N7)GCACG AAATCATGGATCC-3′
1	5′-AAGCTTAATGATTT CGTGTTTCATATGCACG AAATCATGGATCC-3′
2	5′-AAGCTTAATGATTT CGTGTGAAATTTGCACG AAATCATGGATCC-3′
3	5′-AAGCTTAATGATTT CGTGTGAAATTTGCACG AAATCATGGATCC-3′
4	5′-AAGCTTAATGATTT CGTGTGTACTATGCACG AAATCATGGATCC-3′
5	5′-AAGCTTAATGATTT CGTGTTTAACTTGCACG AAATCATGGATCC-3′
6	5′-AAGCTTAATGATTT CGTGTCTACATTGCACG AAATCATGGATCC-3′
7	5′-AAGCTTAATGATTT CGTGTTCACCATGCACG AAATCATGGATCC-3′
8	5′-AAGCTTAATGATTT CGTGTACACCTTGCACG AAATCATGGATCC-3′
9	5′-AAGCTTAATGATTT CGTGTTTAACTTGCACG AAATCATGGATCC-3′
10	5′-AAGCTTAATGATTT CGTGTAAAGTATGCACG AAATCATGGATCC-3′
11	5′-AAGCTTAATGATTT CGTGTTTAACTTGCACG AAATCATGGATCC-3′
12	5′-AAGCTTAATGATTT CGTGTTTAATATGCACG AAATCATGGATCC-3′
13	5′-AAGCTTAATGATTT CGTGTGTCTTATGCACG AAATCATGGATCC-3′
14	5′-AAGCTTAATGATTT CGTGTGTCTTATGCACG AAATCATGGATCC-3′
15	5′-AAGCTTAATGATTT CGTGTTTAGTTTGCACG AAATCATGGATCC-3′
16	5′-AAGCTTAATGATTT CGTGTGTCTTATGCACG AAATCATGGATCC-3′
17	5′-AAGCTTAATGATTT CGTGTATGTCAGGCACG AAATCATGGATCC-3′
18	5′-AAGCTTAATGATTT CGTGTGCCTGTCGCACG AAATCATGGATCC-3′
19	5′-AAGCTTAATGATTT CGTGTTAATTGTGCACG AAATCATGGATCC-3′
20	5′-AAGCTTAATGATTT CGTGTTTTAGTTGCACG AAATCATGGATCC-3′
21	5′-AAGCTTAATGATTT CGTGTTTCTGTTGCACG AAATCATGGATCC-3′
22	5′-AAGCTTAATGATTT CGTGTTTATCTTGCACG AAATCATGGATCC-3′
23	5′-AAGCTTAATGATTT CGTGTTTATTATGCACG AAATCATGGATCC-3′
24	5′-AAGCTTAATGATTT CGTGTTTATTATGCACG AAATCATGGATCC-3′
25	5′-AAGCTTAATGATTT CGTGTTTATACTGCACG AAATCATGGATCC-3′
26	5′-AAGCTTAATGATTT CGTGTTTATTATGCACG AAATCATGGATCC-3′
27	5′-AAGCTTAATGATTT CGTGTGTAACTTGCACG AAATCATGGATCC-3′
28	5′-AAGCTTAATGATTT CGTGTAATCTTTGCACG AAATCATGGATCC-3′
WT	5′-AAGCTTAATGATTT CGTGT GAATAAT GCACG AAATCATGGATCC-3′

^aRS, randomized sequence; WT, wild-type sequence.

^bBoldface type indicates binding sites I and II.

The results of SELEX I show that an optimal spacer region should comprise 7 bp, be AT rich, and contain preferably an A or a T in position 6 and a T in position 7. This is in agreement with a previous fluorescence resonance energy transfer study which revealed a slight (20 to 25°) bend of the copR operator region upon binding of the CopR protein (33). We argued that two pyrimidine-purine steps one helical turn apart might be required for flexibility and hence bendability. SELEX I seems to confirm this argument; one of these pyrimidine-purine steps is at the boundary between the spacer and binding site II and requires the T in position 7 of the spacer which indeed was also found with SELEX II (see below) and found to be a T or C in 9 of 14 sequenced clones by SELEX III (see below). The other pyrimidine-purine step is provided by T241 and G242 in binding site I, which was not found to be altered in any SELEX-derived sequence.

Use of SELEX to study a randomized 17-mer sequence selects both the asymmetric wild-type operator sequence and an operator sequence with perfect symmetry.

To find out whether binding sites that are bound more efficiently by CopR than by the wild-type operator exist, a randomized sequence of 17 bp, the minimal wild-type operator length, was used in the SELEX II experiment. This SELEX II experiment was started with 20 pmol of the randomized sequence, corresponding to ≈700 copies of every possible sequence, which was PCR amplified and labeled prior to selection (see above). Selection was performed for the first three rounds without a competitor, and after a shifted band emerged in round 3 (Fig. 1), the nonlabeled wild-type operator was added as a competitor and seven additional rounds of selection were performed. After 10 rounds, selected fragments were cloned into the pUC19 vector as described above, and 24 independent clones were sequenced. The results (Table 3) show that all selected sequences contained wild-type binding sites I and II. Furthermore, the 3′ nucleotide of binding site I, which is a T in the wild-type site, was in 17 of 24 sequences replaced by a C, making both binding sites perfectly symmetric. Such a perfectly symmetric operator sequence with T243C (Fig. 2), termed KS9, had been analyzed previously (28) and was found to be bound at least as efficiently as the wild-type sequence. Our three-dimensional model of the N-terminal 63 amino acids of CopR (30) predicts that this nucleotide position in binding site I is contacted by R34 of the recognition helix (Fig. 2) and that the contact would be stronger with a C instead of a T.

TABLE 3.

Sequences selected with SELEX II (N₁₇)

Clone	Sequencea	No. of clones containing indicated sequence
RS	5′-AAGCTTAATGATTTN17AAATCATGGATCC-3′
1	5′-AAGCTTAATGATTT CGTGC TTCCTAT GCACG AAATCATGGATCC-3′	17
2	5′-AAGCTTAATGATTT CGTGT TTCTGCT GCACG AAATCATGGACAG-3′	3
3	5′-AAGCTTAATGATTT CGTGT TTATCAT GCACG AAATCATGGATCC-3′	2
4	5′-AAGCTTAATGATTT CGTGT TTCCTCT GCACG AAATCACTGATTC-3′	1
5	5′-AAGCTTAATGATTT CGTGT AAATTTT GCACG AAATCATGGATCC-3′	1
WT	5′-AAGCTTAATGATTT CGTGT GAATAAT GCACG AAATCATGGATCC-3′

^aBoldface type indicates the randomized region.

FIG. 2.

Wild-type copR operator sequence. Binding sites I and II are boxed, and the consensus binding motif 5′-CGTG is highlighted in boldface type. The −35 box and the −10 box of the repR promoter pII are shown in italics. Arrows indicate nucleotides contacted specifically by the corresponding amino acids (shown in the one-letter code) of the recognition helix of CopR. Dashed lines indicate water-mediated contacts. The minimal operator sequence comprises 17 bp, i.e., binding sites I and II with their spacer regions.

Interestingly, the spacer regions of all 24 clones were AT rich (in 22 of 24 cases, ≤2 G's or C's were found) and contained a T in position 7 and, in 18 cases, an A in position 6. The latter data are in agreement with the results described above.

Use of SELEX to study a 30-mer sequence selects the symmetric and the asymmetric wild-type operator sequence and a novel sequence with three nucleotide exchanges.

Since in vitro selection of a 17-bp randomized sequence neither allows extended spacer lengths to be found nor is able to obtain any information on the variability of the flanking sequences, a randomized 30-mer sequence was used in a third SELEX (SELEX III) experiment. This experiment was started with 333 pmol (8 μg) of randomized oligodeoxyribonucleotide SB369, which was PCR amplified and labeled prior to selection. Here, in round 4, a bound fraction appeared, and the next rounds were again performed in the presence of a competitor sequence. After 10 rounds, 14 independent clones were sequenced as described above. Table 4 presents the results. Interestingly, whereas the asymmetric wild-type binding sites in two cases (numbers 1 and 4) and the symmetric binding sites in one case (number 13) were found to be similar to those in SELEX II, the other 11 clones contained binding sites with one (number 5), two (numbers 2 and 3), three (numbers 6, 7, 8, 10, 11, 12, and 14), or even four (number 9) nucleotide exchanges. One of these exchanges was the C in position 243 found in the symmetric wild-type operator by SELEX II. However, the other exchanges were within the consensus binding motif 5′-CGTG (Fig. 2). In binding site I, the C in wild-type position 239 was replaced by a T, and in binding site II, the C in wild-type position 252 was altered to a G (Fig. 3). Additionally, in one case (number 9), the G in position 255 was replaced by an A. These results were surprising, since previous mutations in binding sites I and II seemed to suggest that the four positions of the consensus motif (5′-CGTG) cannot be altered without significant loss of binding affinity (28; unpublished data).

TABLE 4.

Sequences selected with SELEX III (N₃₀)^a

Clone	Sequence
RS	5′-AAGCTTCATCCATAN30TAGTCGTGGATCC-3′
1	5′-AAGCTTCATCCATA CGTGCGTAGGAAACACGGGCCAGACCACAT TAGTCGTGGATCC-3′
2	5′-AAGCTTCATCCATA GATCGTGCATATTTAGCACACCTTTGCATT TAGTCGTGGATCC-3′
3	5′-AAGCTTCATCCATA TGTCGTGCATATTTAGCACACCTTGACCGG TAGTCGTGGATCC-3′
4	5′-AAGCTTCATCCATA CAATCGTGCGTAGGAAACACGGGCAGACCC TAGTCGTGGATCC-3′
5	5′-AAGCTTCATCCATA GATCGTGCATATTTAGCACACCTTGGTCGG TAGTCGTGGATCC-3′
6	5′-AAGCTTCATCCATA TGCATGAAACTGTGCATGTATTGGACGAAT TAGTCGTGGATCC-3′
7	5′-AAGCTTCATCCATA AAGCTTCTGTGCATGTATCGGACGAATTGT TAGTCGTGGATCC-3′
8	5′-AAGCTTCATCCATA TGCATGAAACTGTGCATGTATTGGACGAAT TAGTCGTGGATCC-3′
9	5′-AAGCTTCATCCATA ATGAAACTGTGCATGTATCGGACAATTCAG TAGTCGTGGATCC-3′
10	5′-AAGCTTCATCCATA TGCATGAAACTGTGCATGTATTGGACGAGT TAGTCGTGGATCC-3′
11	5′-AAGCTTCATCCATA TGCATGAAACTGTGCATGTATCGGACGAAT TAGTCGTGGATCC-3′
12	5′-AAGCTTCATCCATA TGCATGAAACTGTGCATGTATCGGACGAGT TAGTCGTGGATCC-3′
13	5′-AAGCTTCATCCATA TTATCGGATGGCGTGCCTATCTTGCACGGT TAGTCGTGGATCC-3′
14	5′-AAGCTTCATCCATA TTCATGAAACTGTGCATGTATAGGACGAAT TAGTCGTGGATCC-3′
WT	5′TGATTTCGTGTGAATAATGCACGAAATCAT-3′

^aFor a comparison of binding sites of sequences determined by SELEX III to that of the wild type, see Fig. 3. Boldface type indicates binding sites I and II.

FIG. 3.

Comparison of binding sites of all SELEX III sequences to those of the wild type. Nucleotide positions different from those of the wild type (shown in boldface type) are underlined. See Table 4 for sequences determined by SELEX III. sy, symmetric operator sequence; WT, wild-type sequence on complementary strand.

Although several other operators in addition to the symmetric and the asymmetric wild-type operator with two, three, or even four nucleotide exchanges were found, these exchanges did not affect nucleotides for which direct covalent contacts with amino acids in the recognition helix of the CopR proteins were proposed (Fig. 2).

With regard to the selected spacer sequences, the results of SELEX I and II for an AT-rich spacer and the necessity of T in position 6 were confirmed, whereas in position 7, a C was found in five cases. Interestingly, no alterations of the spacer length were selected; i.e., neither an 8-bp nor a 6-bp spacer emerged. The flanking sequences could not be analyzed in sufficient detail, since the binding sites were found in the 5′ portion in some cases and in the 3′ portion of the randomized 30-mer sequence in other cases.

The 5′ and 3′ flanking sequences of binding sites I and II can vary widely.

In order to obtain information on the sequences flanking binding sites I and II, a fourth SELEX (SELEX IV) experiment was performed using a 17-bp wild-type operator with a randomized 5-bp spacer region on either side. This SELEX was started with 200 pmol of randomized sequence, corresponding to ≈115 × 10⁶ copies of every sequence variant which were PCR amplified and labeled prior to selection. As expected, a shifted band was visible from SELEX round 1, and after 10 rounds of selection (rounds 4 to 10 with a wild-type competitor), fragments were cloned and 14 clones were analyzed by sequencing. Table 5 shows clearly that no specific sequences are preferred either 5′ or 3′ of binding sites I and II, respectively. Even the AT content does not seem to play a role in these regions.

TABLE 5.

Sequences selected with SELEX IV (N_5-5)^a

Clone	Sequence
RS	5′-AAGCTTAATGATTTN5CGTGT GAATAAT GCACGN5AAATCATGGATCC-3′
1	5′-AAGCTTAATGATTT TTTGGCGTGT GAATAAT GCACGTAAAT AAATCATGGATCC-3′
2	5′-AAGCTTAATGATTT CTTTCCGTGT GAATAAT GCACGTAAGA AAATCATGGATCC-3′
3	5′-AAGCTTAATGATTT ACTACCGTGT GAATAAT GCACGCACTA AAATCATGGATCC-3′
4	5′-AAGCTTAATGATTT TTCTCCGTGT GAATAAT GCACGTTGAA AAATCATGGATCC-3′
5	5′-AAGCTTAATGATTT GCTACCGTGT GAATAAT GCACGTTAAA AAATCATGGATCC-3′
6	5′-AAGCTTAATGATTT AATCTCGTGT GAATAAT GCACGATTTA AAATCATGGATCC-3′
7	5′-AAGCTTAATGATTT TGATGCGTGT GAATAAT GCACGGGTCC AAATCATGGATCC-3′
8	5′-AAGCTTAATGATTT GTGCCCGTGT GAATAAT GCACGCTGAA AAATCATGGATCC-3′
9	5′-AAGCTTAATGATTT AGCTTCGTGT GAATAAT GCACGCGCTG AAATCATGGATCC-3′
10	5′-AAGCTTAATGATTT CTAGCCGTGT GAATAAT GCACGTACAT AAATCATGGATCC-3′
11	5′-AAGCTTAATGATTT TGAAGCGTGT GAATAAT GCACGCATTA AAATCATGGATCC-3′
12	5′-AAGCTTAATGATTT TAGTCCGTGT GAATAAT GCACGCAGGT AAATCATGGATCC-3′
13	5′-AAGCTTAATGATTT AACAACGTGT GAATAAT GCACGTGGCC AAATCATGGATCC-3′
14	5′-AAGCTTAATGATTT CCGGGCGTGT GAATAAT GCACGACCTC AAATCATGGATCC-3′
WT	5′-AAGCTTAATGATTT CGTGT GAATAAT GCACG AAATC AT-3′

^aBoldface type indicates binding sites I and II with the spacer region; italicized type indicates the randomized region.

Binding curves of mutated copR operators confirmed the SELEX results.

Four independent SELEX experiments yielded a set of data concerning binding sites, the spacer region, and surrounding regions of the copR operator. However, no information was obtained about the ability of an 8-bp spacer to bind CopR or about the importance of the pyrimidine-purine dinucleotide steps within the operator regions for efficient binding. In this context, binding efficiencies for spacers comprising exclusively A or T residues or, on the other hand, only G or C residues were of interest, even when such spacers were not selected in vitro. Furthermore, the exchange of single nucleotides such as those found in the mutated operator selected with the SELEX III experiment within the consensus regions and their effects on binding of CopR should be investigated. A comparison of binding curves of operators found with SELEX II and SELEX III should enable us to estimate whether CopR complexes with the selected operator sequences indeed had the same or even lower equilibrium dissociation rate constants as those with the wild-type operator, i.e., if an operator sequence was selected that is bound more tightly by the repressor.

For this purpose, gel shift assays were performed with the mutated operator sequences listed in Table 6. The corresponding binding curves are shown in Fig. 4, and the calculated K_d values for the CopR operator complexes are shown in Table 7.

FIG. 4.

Binding curves with mutated operator sequences. (A) Autoradiograms of the EMSAs with the wild type and the indicated mutated targets are shown. Below each autoradiogram, the lowest and highest concentrations of purified His₆-tagged CopR used in the experiments are indicated. In the case of SB189, SB198, and SB199, the protein was added to the DNA, complex formation was allowed for 30 min, and the complexes were subsequently loaded onto the gel at different time points (0, 20, and 30 min). Consequently, the autoradiograms show the complexes after 40, 20, and 10 min (left, center, and right, respectively) of migration through the native polyacrylamide gel. A significant portion of the protein-DNA complexes dissociate during gel electrophoresis, indicating unstable complex formation. (B) Binding curves of the wild type and mutated targets. These binding curves were used for the calculation of the K_d values of the CopR-DNA complexes (as described in reference 29) shown in Table 7.

TABLE 7.

Equilibrium dissociation rate constants of mutated copR operators

Buffer and Oligodeoxyribonucleotide	Kd (nM)	Nucleotide alteration(s)
Buffer with NaCla
KS1	0.40 ± 0.05	Wild type
SB175	0.82 ± 0.05	T247A (pyrimidine-purine step in spacer)
SB174	0.42 ± 0.05	G244T (pyrimidine-purine step in spacer)
SB183	0.36 ± 0.04	Spacer containing 7A
SB184	0.30 ± 0.07	Spacer containing 7T
SB187	6.21 ± 1.90	Spacer containing 7G
SB190	4.11 ± 0.90	Spacer containing 7C
Buffer without NaClb
KS1	0.40 ± 0.06	Wild type
KS9	0.18 ± 0.05	T243C
KS3	1.54 ± 0.11	G251A
SB414	1.28 ± 0.16	C239T, T243C, C252G
SB417	0.68 ± 0.07	C239T, T243C
SB418	0.30 ± 0.05	C252G

^aBinding buffer, electrophoresis buffer, and gel contained 75 mM NaCl.

^bEMSAs were performed in the absence of NaCl.

With regard to the spacer region, three important results were obtained. First, an 8-bp spacer sequence did not prevent CopR binding; however, only unstable complexes were formed (Fig. 4A). On the other hand, a 6-bp spacer did not allow CopR binding (data not shown). Second, the alteration of a pyrimidine-purine step within the spacer did not affect binding. Third, spacers composed of only A's or T's were bound equally well and comparable to those of the wild-type. In contrast, spacers comprising only C's or G's were bound with significantly lower efficiency (10- to 15-fold lower). These results are in correlation with the SELEX I experiment, where AT-rich spacers were selected, and in correlation with SELEX III, where only 7-bp spacers were found.

A comparison of binding constants of asymmetric and symmetric operators as selected with SELEX II demonstrated that the K_d value of the latter one (former KS9) (28) was twofold lower than that of the asymmetric wild-type operator; i.e., this operator was bound slightly more efficiently than the wild-type operator. This is most probably the reason why the symmetric operator was selected with SELEX II in the majority of the cases. The mutated operator selected with SELEX III (SB414) yielded a K_d value that was threefold higher than that of the asymmetric wild-type operator. A mutated operator containing only one of the two mutations within the 5′-CGTG motif (SB417) bound CopR almost as efficiently as the wild-type with a K_d of 0.6 nM.

Exchanges of other single nucleotides in binding site I or II impaired or prevented CopR binding or led to unstable complexes that dissociated during electrophoretic separation, as was found for SB198 and SB199 containing nucleotide exchanges in positions 241 and 242 of binding site I, respectively (Fig. 4A). Such exchanges were not found with SELEX II or III. The importance of the G's in positions 240 and 254 had been demonstrated before (28). G-to-A exchanges at these two positions abolished binding or decreased the binding affinity drastically.

Copy numbers of pIP501 derivatives demonstrated that selection in vivo (evolution) was directed at maximal binding efficiency.

A set of pIP501 derivatives was constructed to evaluate the effects of mutated operator sequences on copy number control in vivo. The corresponding plasmids were introduced into B. subtilis strain DB104 by transformation, and copy numbers were determined (Fig. 5 and Table 8).

FIG. 5.

Copy number determination of pIP501 derivatives with mutated operator sequences. BamHI-linearized aliquots of undiluted plasmid DNAs separated on 1% agarose gels were prepared from 1-ml culture volumes of B. subtilis strains containing wild-type or mutant pIP501 derivatives grown to the same optical density in late logarithmic phase. In all cases, four to six transformants grown in parallel were used for plasmid preparations. Gel photos were scanned and quantified by using PCBAS 2.0 software. The results are shown in Table 8.

TABLE 8.

Comparison of copy numbers of wild-type and mutant pIP501 derivatives in B. subtilis

Plasmid	Relative copy no.	Nucleotide alteration(s)
pPRC333	1.0	Wild type
pPRC334	0.9	T243C (KS9)
pPRC227	0.9	T247A (spacer without pyrimidine-purine)
pPRC294	2.8	G251A (KS3)
pPRC221	0.8	Spacer with 7C
pPRC222	1.3	Spacer with 7A
pPRC292	0.6	Spacer with 7G
pPRC228	1.5	8-bp spacer
pPRC416	2.9	C239T, T243C, C252G

A comparison of the copy numbers of plasmids pPRC333 (wild type) and pPRC334 (T243C symmetric operator) shows that in vivo, a symmetric and an asymmetric operator are equally efficient in regulation and that the slight differences (twofold) in the K_d values do not result in significant effects. In contrast, the copy number of pPRC416 carrying three mutations, the T243C and 1-bp exchanges in binding sites I and II, was about threefold higher, which was in correlation with the K_d value that was threefold higher than that of the asymmetric wild-type and a K_d value even sixfold higher than that of the symmetric wild-type operator, and hence, CopR binding was slightly impaired. From these data, it can be concluded that evolution in vivo was directed at maximal binding. A direct correlation between binding and regulation was also apparent from the analysis of the other symmetric operator variant and supported the importance of G251 in binding site II; the 3- to 4-fold-higher K_d value of KS3 was accompanied by a 2.8-fold-higher copy number of pPRC294, i.e., impaired regulation in vivo.

Some of the pIP501 derivatives that carried mutations in the spacer region showed somewhat unexpected results. Whereas pPRC228 that has an 8-bp spacer, for which unstable binding was observed, replicated at a 1.5-fold-higher copy number than pPRC333, and pPRC derivatives with only A or T residues in the spacer behaved like those of the wild type, pPRC292 carrying only G residues and pPRC221 carrying only C residues replicated at even lower copy numbers than did the wild type, which was in strong contrast to the calculated 16- or 10-fold-higher K_d values that indicated significantly impaired binding. Stacking effects within the spacer region or supercoiling effects (see below) might be responsible for this unexpected behavior.

In vitro transcription with B. subtilis RNA polymerase confirmed the in vivo data.

In order to find out whether the unexpected results of the copy number determinations with only G or C spacers were due to interaction effects with RNA polymerase, in vitro transcription experiments were performed with B. subtilis RNA polymerase in the presence or absence of purified native CopR. Figure 6 demonstrates that repression was performed equally well for the wild type and the T243C mutation (KS9), which coincides with the results of the EMSAs and in vivo copy number determinations (Table 8). For G251A (KS3), a four- to five-fold-lower repression effect was found, which again was in correlation with the results of the EMSA (K_d fourfold higher than that for the wild-type) and copy number determination (≈2.8-fold-higher copy numbers) (Table 8). On the other hand, an operator with only C in the spacer region behaved like the wild type with only 7% transcriptional read-through upon CopR binding. This coincides with the copy number regulation, which was also like that of the wild type. Here, factors other than simple binding apparently play a role, since the 10-fold-higher K_d value indicates that binding of CopR is significantly impaired. The same holds true for the operator with the spacer with G only, which is only 2-fold worse in repression in the in vitro transcription assay and even shows an approximately 2-fold-lower copy number than that of the wild type but has a 16-fold-higher K_d value significantly impaired in CopR binding. As the reasons for these discrepancies, supercoiling effects cannot be excluded, which are not considered in EMSA or SELEX, where linear templates are used. The operator with an 8-bp spacer showed about a fourfold decrease in repression, in line with the formation of an unstable complex in the EMSA (Fig. 4A).

FIG. 6.

In vitro transcription with DNA fragments containing mutated copR operators. In vitro transcription in the presence or absence of native CopR was performed with B. subtilis RNA polymerase as described in Materials and Methods. Above the lanes, the mutations in the operator sequences are indicated. RNAII_F, full-length RNAII (to the end of the used DNA template fragment), RNAII_T, terminated RNAII terminated at the attenuator. Bands were quantitated after phosphorimaging with the PCBAS 2.0 program, and the percentage of read-through in the presence of CopR was calculated (shown below the gel). WT, wild type; M, marker.

LacZ fusions indicate that discrepancies between K_d values and copy numbers are due to supercoiling effects.

To analyze the role of supercoiling for repression in vivo versus that of repression in vitro (EMSA, SELEX, and in vitro transcription), transcriptional lacZ fusions of the repR promoter pII containing either wild-type or mutated operators were constructed and integrated into the amyE locus of the B. subtilis chromosome. Resulting B. subtilis integrant strains DB104::pAC333 (wild type), DB104::pAC334 (T243C, symmetric operator), and DB104::pAC221 (7C spacer) were transformed with plasmid pCOP9 to provide CopR in trans, and β-galactosidase activities were determined as described previously (7). The results are shown in Table 9. A comparison of β-galactosidase activities determined in the absence and presence of CopR revealed different degrees of repression. In the wild-type case (pAC333), about 17-fold repression was observed, whereas the repression effect was slightly higher (23-fold) in the case of pAC334 containing the symmetric operator which was selected with SELEX II. This result is in agreement with the calculated copy numbers, which were identical in both cases, and indicates that the twofold difference in the K_d values determined with linear templates is overcome by the supercoiling effect on a circular template in vivo. For pAC221 comprising the 7C spacer region, lacZ values in the absence of CopR were about 2.5-fold lower than those of pAC333 and pAC334. This finding suggests that the C-rich spacer region upstream of the −35 box of pII might lower the transcription efficiency by the RNA polymerase. On the other hand, the four- to sixfold-lower repression effect compared to those of the wild-type and symmetric operator was in accordance with the calculated 10-fold-higher K_d value for this operator determined with EMSA. Apparently, supercoiling effects are responsible for the decrease of the expected (from the K_d values) difference in repression. With these data, the surprising discrepancies between K_d values and copy numbers for the 7C spacer variant could be explained. The results of the lacZ measurements indicate that efficient binding in vivo is affected by supercoiling and that the copy numbers are influenced by both the K_d value and supercoiling effects.

TABLE 9.

β-Galactosidase activities of chromosomal lacZ fusions^a

Strain	CopR in transb	β-Galactosidase activity (Miller units)	Repression effect (fold)
DB104::pAC333	−	467 ± 49
DB104::pAC333	+	27.7 ± 6.7	17
DB104::pAC334	−	438 ± 62
DB104::pAC334	+	19.3 ± 1	23
DB104::pAC221	−	180 ± 29
DB104::pAC221	+	42.6 ± 11	4

^aThe values are average values derived from three independent determinations with independently grown cultures.

^bβ-Galactosidase activity was determined in the presence (+) and absence (−) of CopR.

Evolutionary considerations.

One instrument to adjust the copy number of pIP501 is the K_d value of the CopR operator complex. Evolution of the copR operator in vivo apparently resulted in a low copy number of the corresponding plasmid pIP501 (approximately 5 copies). This copy number, however, is not the lowest that can be obtained, as shown with pPRC221 and pPRC292. However, in these two cases, copy numbers are lower than those of the wild-type, since the mutated pII promoters with C- or G-rich upstream regions are less efficient than the wild-type promoter per se (see above), and the reduced repression effect on these weak promoters due to high K_d values of the mutated Cop operator complexes is, in a supercoiled context, still sufficient to decrease replication efficiency slightly below the wild-type level.

As demonstrated previously with a series of pIP501 derivatives, there seems to be both an upper and a lower limit for copy numbers found in vivo since no derivative could be constructed that replicated at more than 50 to 100 copies/cell with B. subtilis as one of its gram-positive hosts (5). When pIP501 evolved in its original host, Streptococcus agalactiae, which was living under certain physiological conditions, selection was apparently for low, but not the lowest possible, copy number, which proved to be optimal under the environmental conditions encountered by this host. This finding is supported by the independent in vivo selection of three different (but identical in the binding sites) operators of the three representatives of the inc18 family of streptococcal plasmids that replicate via the theta mechanism in a broad range of gram-positive bacteria (4): pIP501, pSM19035, and pAMβ1. These plasmids are 97% identical on the nucleotide level, and their replication regions reveal the same modular structure: 5′ cop gene, rep gene, and origin. The cop genes (copR [pIP501], copF [pAMβ1], and copS [pSM19035]) encode almost identical proteins that differ only in a few amino acids in the C terminus. Furthermore, the cop operators contain identical binding sites I and II (11, 35). The only differences between the copR and copS operators are found in the spacer region (G244A and T247A) and between the copR and copF operators in the regions flanking the binding sites (T236G and A260G). These data suggest that during evolution, identical Cop binding sites emerged in three independent plasmids. The results of SELEX I, EMSAs with spacer mutations, and copy number determinations with pPRC227 (T247A) confirmed that the two positions in the spacer region that are different in the copS operator are not required for efficient binding or regulation in vivo. Additionally, the results of SELEX IV demonstrated that the sequences of the flanking regions of the cop operator can vary widely so that the differences between copR and copF operator are negligible, too.

In summary, in vitro selection of the copR operator proved to result in the same sequences as those found with in vivo selection and demonstrated that evolution was directed at maximal binding affinity. From our experience, SELEX may be, at least in the case of simple transcriptional repressors, a powerful method to answer evolutionary questions.