The Zinc-responsive Regulator Zur and Its Control of theznu Gene Cluster Encoding the ZnuABC Zinc Uptake System in Escherichia coli *

The synthesis of the Escherichia colizinc transporter, encoded by the znuACB gene cluster, is regulated in response to the intracellular zinc concentration by thezur gene product. Inactivation of the zur gene demonstrated that Zur acts as a repressor when binding Zn2+. Eight chromosomal mutant zur alleles were sequenced to correlate the loss of Zur function with individual mutations. Wild-type Zur and ZurΔ46–91 formed homo- and heterodimers. Dimerization was independent of metal ions since it also occurred in the presence of metal chelators. Using an in vivo titration assay, the znu operator was narrowed down to a 31-base pair region overlapping the translational start site of znuA. This location was confirmed by footprinting assays. Zur directly binds to a single region comprising a nearly perfect palindrome. Zinc chelators completely inhibited and Zn2+ in low concentrations enhanced DNA binding of Zur. No evidence for autoregulation of Zur was found. Zur binds at least 2 zinc ions/dimer specifically. Although most of the mutant Zur proteins bound to the znu operator in vitro, no protection was observed in in vivo footprinting experiments. Analysis of the mutant Zur proteins suggested an amino-terminal DNA contact domain around residue 65 and a dimerization and Zn2+-binding domain toward the carboxyl-terminal end.

compatible ends were blunt-ended prior to ligation by removal of protruding 3Ј termini or filling in of recessed 3Ј termini. Plasmid pSP93/3 was obtained as one of the gene library plasmids that complemented a zur mutant (3). The 0.6-kb Eco47III fragment from pSP93/3, which encompass the zur operator region, was ligated in either orientation into the EcoRV site of pBCSK ϩ , resulting in pSP97/61 and pSP97/63.
For construction of the wild-type zur plasmids, the 0.86-kb BsrI fragment from pSP93/3 was inserted into the SmaI site of pT7-5 behind the phage T7 ⌽10 promoter (pSP100/3) and in the opposite orientation (pSP100/4). The primers ZUR2 (GTGTTTTTCCTGGAACATGGTA, nt 902 to 881, E. coli genome sequence (4) section 368) and ZUR4 (TA-ATCCCTCCTGCCCGACGTGT, nt 287-308, section 368) were used for cloning the large 0.62-kb region of the zur gene into NdeI/EcoRI-cleaved pT7-7, yielding pSP100/32 (zur behind the phage T7 ⌽10 promoter) and pSP100/34 (zur in the opposite orientation). Sequencing of the pSP100/32 insert revealed that due to failure during oligonucleotide synthesis, three nucleotides in the ZUR2 primer were lacking, resulting in an exchange of the first residue (Val) for Met and loss of the second residue (Phe). The short 0.55-kb form of the zur gene was amplified by PCR with the primers ZUR3 (TGGAAAAGACCACAACGCAGGA, nt 841 to 820, section 368) and ZUR4 and likewise ligated into NdeI/EcoRIdigested pT7-7, resulting in pSP100/22 (zur behind the phage T7 ⌽10 promoter) and pSP100/21 (zur in the opposite orientation). pSP98/35 (zur⌬46 -91) was obtained from an E. coli gene library as a plasmid that derepressed the znuA::MudX fusion of SIP468 in the in vivo titration assay.
To express the mutant zur alleles, genomic DNA extracted from the appropriate strains or pSP98/35 plasmid DNA was amplified by PCR using primers ZUR3 and ZUR4 and cloned into NdeI/EcoRI-cleaved pT7-7, creating the pSP104 plasmids. For negative complementation assays, the PCR fragment recovered from genomic DNA isolated from MC4100 (wild-type zur) and from mutant zur strains or from plasmid pSP98/35 with primers ZUR1 and ZUR4 or primers ZUR1 and TET2 (GTCCTGCTCGCTTCGCTACTTGG, nt 1802-1824 of pACYC184), respectively, was cloned into the EcoRV site of pACYC184, resulting in the plasmids of the pSP102 series.
Strain SIP500, which has a chromosomal znuAЈ-lacZ fusion, was obtained as follows. The 0.69-kb NcoI/SalI DNA fragment of the znuA operator was inserted into the SmaI site of the lacZ operon fusion vector pRS415. The znuAЈ-lacZ fusion of the resultant plasmid (pSP87/1) was transferred to the phage RS45 by homologous recombination, giving rise to SP87/1, which was introduced into MC4100 as described (13). zur-19 was cotransduced with lamB::Tn10 (from MM143) into strain SIP500 (znuAЈ-lacZYA), selecting for Tet r (where Tet is tetracycline and the superscript "r" indicates resistant/resistance) and znuAЈ-lac-ZYA derepressed colonies, to generate strain SIP584.
For genomic inactivation of zur, the 2.87-kb fragment from chromosomal DNA of strain MC4100 amplified by PCR using the primers ZUR5 (GTTAGTGGTCGGCAATATCCTC, nt 10382-10403, section 367) and ZUR6 (GTGCGTTCCTTCATTCACCAGA, nt 1992 to 1971, section 368) was inserted into the EcoRV site of pACYC184. In the resulting plasmid (pSP106/28), an internal DraIII/PstI fragment spanning nt 354 -733 (section 368) of the zur gene was replaced with the 2.0-kb BamHI ⍀ fragment (Spc r /Str r (where Spc is spectinomycin and Str is streptomycin) gene box) of pHP45⍀ to generate pSP106/47. The linear 4.5-kb NdeI/NheI fragment of pSP106/47 was introduced into strain SIP602 by transformation. SIP602 was obtained by transduction of recD1901::Tn10 (from CAG12135) into W3110 to allow double selection for Spc r and Str r . The chromosomal zur::Spc r /Str r gene disruption of chloramphenicol-sensitive transformants was confirmed by PCR, and the allele was transduced into strains MC4100 and SIP468, creating strains SIP812 and SIP600, respectively.
Strains with a chromosomal znuЈ-lacZ fusion were obtained by infection with a Mud1(Amp r lac cts) lysate prepared from strain MAL103 as described (14). The Mud1 fusion sites of zinc-induced and EGTA-sensitive Mud1 fusions were amplified and cloned with the primers MUD1 (CACGTACATGCCGCCAAACTCACCA, nt 170 to 146, GenBank TM / EBI Data Bank accession number M33723), which is complementary to the right end of the Mud1 phage, and YEB1, which binds to the znuA gene. Sequencing revealed that the Mud1 phage was inserted in the znu gene cluster behind nt 2122 (section 170) (SIP576) and nt 1409 (section 170) (SIP775). Mud1 insertion inactivates the genes. The Mud1 fusions were isolated by transduction, creating strains SIP578, SIP579, and SIP800.
Overproduction and Purification of Zur-The zur genes were expressed from the pT7-7-derived plasmids using the optimal Shine-Dalgarno sequence. E. coli BL21(DE3) cells freshly transformed with the appropriate plasmid were induced with 1 mM isopropyl-␤-D-thiogalactopyranoside, harvested, resuspended in buffer A (20 mM sodium P i and 300 mM NaCl, pH 7.4), and disrupted by two passages through a French pressure cell (Aminco, Silver Spring, MD) at 138 MPa. For purification of Zur from the soluble fraction, the cell extract was centrifuged (70 min at 13 500 ϫ g), and the supernatant was separated by fast protein liquid chromatography on a Ni 2ϩ -iminodiacetic acid column (HiTrap chelating, Amersham Pharmacia Biotech) equilibrated with buffer A. To isolate Zur from inclusion bodies, the inclusion bodies were recovered from the cell extract by centrifugation and dissolved in buffer containing 50 mM HEPES/NaOH and 1 M urea, pH 8.0. Insoluble cell debris was removed by centrifugation, and the supernatant was slowly diluted 10-fold with buffer A and likewise applied to the column. The resin was washed with buffer A. Zur was eluted with linear imidazole gradients in the 5-500 mM range. The buffer was exchanged, and the protein was concentrated by ultrafiltration (Centricon-10, Amicon, Witten, Germany). Starting with 100 ml of bacterial culture, purification from the soluble fraction yielded Ϸ6 mg of total protein in the supernatant and Ϸ1 mg (17%) of pure protein for wild-type and mutant Zur proteins and Ϸ12 mg of total protein in the supernatant and Ϸ6 mg (50%) of pure protein for Zur⌬46 -91. The yield of total and pure proteins for Zur⌬46 -91 was higher because no Zur⌬46 -91 was lost in inclusion bodies. The purified proteins were stable and remained active for at least 6 months if stored at 4°C.
DNase I Footprinting-For generation of the labeled target DNA, the appropriate operator fragment was amplified by PCR with genomic DNA isolated from MC4100 using primer YEB2 and the fluoresceinlabeled primer YEB3 (CTGTGTTGCACCTCCCCAGAGAG, nt 937-959, section 170) for the ZnuCB-coding strand or primer YEB1 and the fluorescein-labeled primer YEB7 (AGTTCCAGCGACACATCAGAGAG, nt 1160 to 1138, section 170) for the ZnuA-coding strand. Purified Zur protein was allowed to equilibrate in 50 l of binding buffer (10 mM Tris-HCl, 5 mM DTT, and 5% glycerol, pH 8.0) containing metal ions or chelators as indicated for 20 min at room temperature prior to the addition of 30 ng of the fluorescein-labeled DNA probe. After further incubation for 10 min and the addition of 50 l of DNase buffer (10 mM Tris-HCl, 25 mM NaCl, 5 mM MgCl 2 , and 1 mM CaCl 2 , pH 7.8), DNA was digested for 2 min at room temperature with 1 l of DNase I in DNase buffer at a concentration such that Ϸ40% of the DNA molecules were left intact. Reactions were terminated by the addition of 10 l of stop solution (150 mM EDTA and 5% SDS), followed by ethanol precipitation. Samples were electrophoresed on the A.L.F. DNA Sequencer and analyzed with A.L.F. DNA Fragment Manager Version 1.1 software (Amersham Pharmacia Biotech). The products of the standard DNA sequencing reactions using the same labeled primer and pSP92/3 as DNA template were included on the gel. To exclude the possibility that the observed DNA protection was due to impurities in the purified Zur fraction, footprinting assays with the same column fraction recovered accordingly, but without overproduction of Zur, were performed as a negative control.
In Vivo Footprinting-Cells were grown in TY medium supplemented with 10 M ZnSO 4 until E 578 Ϸ 0.4 was reached. Bacteria from 6 ml of this culture were harvested by centrifugation, resuspended in 0.4 ml of growth medium, and incubated with 4 l of DMS for 2 min at room temperature. To stop the methylation reaction, cells were washed twice in 40 ml of ice-cold M9 minimal medium. For comparison, plasmid DNA containing the znu operator region was methylated in vitro with DMS as described by Ausubel et al. (17), but 50 mM Tris-HCl, pH 8.0, was used as the reaction buffer. The genomic or plasmid DNA isolated from each sample was used as template for a one-sided PCR with Taq DNA polymerase and the 5Ј-fluorescein-labeled primer YEB3 or YEB7. When a methylated base is encountered (DMS predominantly methylates N-7 of guanine and N-3 of adenine (18)), the chain extension by Taq DNA polymerase terminates due to mismatches and the lack of proofreading activity. This results in 5Ј-fluorescein labeled DNA fragments starting at nt 1160 (section 170) (primer YEB7) or nt 937 (section 170) (primer YEB3) and mostly ending at positions opposite of a methylated residue in the matrix strand. PCR products were precipitated with ethanol and analyzed on the A.L.F. DNA Sequencer.
Determination of the Metal Ion Content of Zur-To saturate Zur with metal ions, the protein was incubated with a 2-3-fold molar excess of Zn 2ϩ and/or Fe 2ϩ in buffer B (20 mM Tris-HCl, 50 mM NaCl, and 2 mM DTT, pH 8.0) for 20 min at room temperature. Unbound metal ions were removed by gel filtration on a Sephadex G-25M column (Amersham Pharmacia Biotech) equilibrated with buffer B. The zinc and iron contents were determined by atomic absorption spectroscopy.
Protein Assay and Amino-terminal Protein Sequencing-Protein concentrations were quantitated in triplicate using the biuret reaction and bicinchoninic acid (reagents from Pierce) as described by Smith et al. (19) or according to Bradford (20) with bovine serum albumin as the protein standard. Zur was amino-terminally sequenced on an Applied Biosystems 477A protein sequencer by Ralph Jack (Universität Tü bingen).
Gel Filtration-Gel filtration was performed on a Superdex 75 HR 10/30 fast protein liquid chromatography column (Amersham Pharmacia Biotech) equilibrated with buffer A or with buffer containing 20 mM Tris-HCl, 300 mM NaCl, pH 8.0, and additives as indicated and calibrated with appropriate standards (bovine chymotrypsinogen A, bovine serum albumin, bovine RNase A, soybean trypsin inhibitor, chicken ovalbumin, and bovine carbonic anhydrase) at a flow rate of 0.5 ml/min.
Computer Analyses-Nucleic acid and protein sequences were analyzed by the PC/GENE Version 6.85 program package (IntelliGenetics, Inc., Mountain View, CA).

Identification of the Translational Start Site of zur-
The translational start site of yjbK (zur) was proposed as Val 1 according to E. coli Genome Sequence Data Base section 368 (4). However, comparison of the Zur and Fur amino acid sequences (3) suggested that the translation of zur starts later at Met 21 (numbering as in Blattner et al. (4)), resulting in a smaller protein. To determine the physiological start site, the DNA fragments for the two alternative forms corresponding to amino acids 1-191 (long form) and 21-191 (short form) of YjbK were cloned in frame directly upstream of the optimal Shine-Dalgarno sequence in pT7-7 (pSP100/32 and pSP100/22, respectively) to initiate translation exactly at the indicated sites. For comparison, pSP100/3 was used for synthesis of the native Zur protein. All three plasmids, including the one coding for the short Zur protein (pSP100/22), expressed physiological active proteins that complemented a zur mutant. Complementation occurred even without induction of T7 RNA polymerase or when the inserts were in the direction opposite to that of the T7 RNA polymerase promoter (pSP100/34, long form on pT7-7; pSP100/21, short form on pT7-7; and pSP100/4, native form on pT7-5), indicating that minimal amounts of Zur protein are sufficient for repression of the znu gene cluster. Labeling with [ 35 S]methionine ( Fig. 1) showed that native Zur (lane 2) was identical to the short form (lane 4) and comprised 171 amino acids (Fig. 6, ZUR_ECOLI). In the sample of the labeled proteins synthesized from the plasmid construct encoding the long form of Zur, small amounts of the native form were observed (lane 6), which made it impossible to determine whether the long form is active in vivo. The amino-terminal sequence of purified wild-type Zur (see below) was determined to be MEKTTTQELLAQAEK.
Chromosomal Gene Inactivation of zur-zur mutants that cause a derepression of the znuACB gene cluster and an increase in zinc uptake have been previously isolated (3). To decide whether Zur operates as an activator or as a repressor, a zur disruption mutant (zur::Spc r /Str r ) was constructed in which zur was deleted from its chromosomal site. The derivative pSP106/47, which carries the zur::Spc r /Str r disruption, did not complement the zur-19 mutant strain SIP584, whereas the zur-containing plasmid pSP106/28 did. Strain SIP812 (MC4100 zur::Spc r /Str r ) was viable; therefore, zur is not an essential gene under the conditions tested. In strain SIP600 (zur::Spc r /Str r znuA::MudX), the znuAЈ-lacZ fusion was constitutively derepressed, indicating that Zur acts as a repressor.
zur Mutations-To correlate the loss of Zur function with individual mutations in the amino acid sequence of the Zur protein, the zur alleles of MC4100 (wild-type zur); of the constitutive 1-methyl-3-nitro-1-nitrosoguanidine mutants SIP557, SIP559, SIP561, SIP562, SIP564, SIP565, SIP566, and SIP567; and of the plasmid pSP98/35 were sequenced after the cloning of two independently generated PCR products. The sequence of wild-type zur corresponded to that reported by Blattner et al. (4). All mutant zur genes contained a single point mutation (exclusively G:C to A:T transitions, as common for 1-methyl-3nitro-1-nitrosoguanidine mutagenesis) associated with one-amino acid substitution, and only zur-11 and zur-36 were identical (Table II). The mutant zur genes were tested for negative complementation to see whether interaction of wild-type Zur with mutant Zur eliminates the function of wild-type Zur (Table II).
Overproduction, Purification, and Characterization of Wildtype and Mutant Zur Proteins-The wild-type Zur protein synthesized represented a major portion of the total cellular protein ( Fig. 2A, lane 2), and most of the Zur protein accumulated in inclusion bodies, with only a minor portion in the soluble fraction. High levels of thioredoxin (encoded on plasmid pT-Trx), lowering of the incubation temperature (21), or variation of growth medium (22) does not significantly increase the solubility of Zur.
Because of its metal ion-binding properties and the high His and Cys content (11%), Zur was purified by immobilized metal affinity chromatography. Ni 2ϩ and Zn 2ϩ turned out to be the most suitable metal ions. Zur from the soluble fraction and from inclusion bodies was purified to electrophoretic homogeneity by nickel affinity chromatography. SDS-PAGE analysis of purified wild-type Zur revealed a single protein with a molecular mass of Ϸ19 kDa (Fig. 2A, lane 4). This is consistent with the molecular mass of Zur predicted from the amino acid sequence (19.3 kDa).
The gene products of the mutant zur genes zur-11, zur-19, zur-23, zur-25, zur-27, zur-31, zur-35, and zur⌬46 -91 were purified accordingly. Only Zur⌬46 -91 did not form inclusion bodies. Purified Zur⌬46 -91 showed an apparent molecular mass of Ϸ14 kDa (Fig. 2B, lane 4) as predicted (14.0 kDa). Despite nearly identical molecular masses, some of the mutant Zur proteins (Zur19, Zur31, and Zur35) migrated faster on SDS-PAGE than wild-type Zur. Zur migrated more slowly under reducing conditions than under nonreducing conditions, which indicates that the isolated Zur protein was in the oxidized form with at least one intramolecular disulfide bridge even when the protein isolation and purification procedures were carried out in the presence of 10 mM 2-mercaptoethanol. Stronger reducing agents such as DTT or higher concentrations of 2-mercaptoethanol were not used for Ni 2ϩ -immobilized metal affinity chromatography to avoid reduction of Ni 2ϩ . Under nonreducing conditions, protein bands with lower intensity were observed at Ϸ38 and Ϸ57 kDa and at Ϸ29 and Ϸ43 kDa for wild-type Zur and Zur⌬46 -91, respectively; these two bands probably represent the dimer and trimer formed by intermolecular disulfide bridges.
Zur Forms a Dimer-To confirm that Zur forms dimers, to determine the dimer/monomer ratio under various conditions, and to see whether the mutant Zur proteins differ, the purified proteins were analyzed by gel filtration. Wild-type Zur eluted as a major peak with a molecular mass (ϮS.E.) of 46.0 Ϯ 2.5 kDa (apparently corresponding to the dimer form), and only trace amounts eluted with a molecular mass of 29.0 Ϯ 2.9 kDa (representing the monomer). The higher than expected molecular masses could be caused by a more rod-like rather than globular shape of the protein. The relative difference between the expected and observed molecular masses was smaller for the dimer than for the monomer. This is in agreement with the explanation provided that the monomers associate at the extended interfaces to form a dimer that is more globular than the monomer. Due to the high ionic strength (300 mM NaCl) of  a Each zur allele in both orientations on vector pACYC184 (pSP102 plasmids) was introduced into the wild-type zur strain SIP468 (znuA::MudX) by transformation. When negative complementation occurred, the repression of the znuAЈ-lacZ fusion by wild-type Zur was abolished, resulting in red instead of white colonies on MacConkey lactose agar plates. the buffer, it was ruled out that Zur was strongly retained because of its negative charge at this pH. Thus, wild-type Zur exists as a dimer under these conditions. Addition of the metal ion chelator TPEN (30 M) or MnCl 2 (250 M), each in the presence of DTT (2 mM), yielded the same results, indicating that dimerization was independent of metal ions. For Zur⌬46 -91, nearly all of the protein eluted at the dimer size (42.6 kDa) as well. This shows that Zur⌬46 -91 forms dimers despite the deletion of 46 amino acids in the amino-terminal part, indicating that this region is not essential for dimerization. For some of the mutant Zur proteins, the dimer/monomer ratio was shifted, e.g. for Zur25, the portion of the dimer form was lower than that of wild-type Zur, and for Zur27, the monomeric state prevailed.
Formation of wild-type Zur/Zur⌬46 -91 heterodimers was demonstrated by nondenaturing polyacrylamide gel electrophoresis. Although wild-type Zur and Zur⌬46 -91 clearly differ in mass, they are nearly identical in their charge/mass ratio, the main criterion for separation upon nondenaturing gel electrophoresis. Therefore, it was difficult to find conditions to separate these two proteins. Besides the protein bands for the wild-type Zur homodimer (Fig. 2C, lane 2) 1 and 4). This band was less than twice as intense as the homodimer bands expected statistically; therefore, homodimers seem to be favored above heterodimers.
Subcloning of the znu Operator Region Affording Zinc Regulation in Vivo-To localize the DNA region within the znu operator that is responsible for Zur regulation, the in vivo titration assay described previously (3) was employed. Introduction of the znu operator on a high copy number plasmid (pSP93/2) titrated Zur and hence derepressed znuAЈ-lacZ expression in strain SIP468, resulting in red colonies on Mac-Conkey lactose agar plates. In contrast, SIP468 harboring the pBCSK ϩ vector formed white colonies under these conditions. The region responsible for the Zur regulation was narrowed down by subcloning as depicted in Fig. 3. The smallest active fragment covered nt 1025-1055 of section 170. In this region, a nearly perfect palindrome was found (Fig. 4C). The segment TGTGATATTATAACA within this palindrome cloned in either direction into pBCSK ϩ (pSP105/26 and pSP105/27) was not active, indicating that this sequence was not sufficient for effective Zur titration. However, plasmid pSP105/25, which contains three copies of this sequence, derepressed znuAЈ-lacZ fusion expression in the operator titration assay, either because of the higher number of copies or because the adjacent nucleotides restored an active Zur-binding site. Plasmids pSP97/61 and pSP97/63 (pBCSK ϩ with the zur operator region in the two orientations) were inactive in the in vivo titration assay, which indicates that Zur does not interact with its own operator to autoregulate the expression of its gene.
Zur Binds to the znu Operator-The data obtained by in vivo titration assays did not rule out the possibility that regulation via Zur is indirect, mediated by other factors. DNase I footprinting assays with purified wild-type Zur protein were used to determine whether Zur directly binds to this DNA region and to verify the DNA-binding site on the znu operator. A single region of the znu operator was protected from DNase I cleavage, extending from nt 1033 to 1061 of section 170 on the ZnuCB-coding strand (Fig. 4A) and from nt 1057 to 1029 of section 170 on the ZnuA-coding strand (Fig. 4B). The protected region comprises the imperfect palindrome identified by the in vivo titration assays described above (Fig. 4C). Thus, Zur appears to bind to the DNA as a dimer. A molar excess of Ͻ15 Zur dimers/DNA molecule was sufficient for protection under optimized conditions. Higher concentrations of the repressor (a protein/DNA ratio of up to 1700:1) did not result in an extended protection zone. Zur bound to DNA only in the presence of a reducing agent such as DTT. No difference was observed between Zur purified from solubilized inclusion bodies and Zur purified from the soluble fraction.
Zur31 and Zur⌬46 -91 were the only mutant Zur proteins that did not protect the znu operator in DNase I footprinting assays even when applied in concentrations 20-fold higher than needed for DNA protection by wild-type Zur (Table II). A surplus of Zur⌬46 -91 (10-fold) or Zur25, but not Zur27, over wild-type Zur was able to abolish binding of wild-type Zur to the DNA. Competition for Zn 2ϩ did not account for this effect because the addition of more Zn 2ϩ did not change the result. An excess of Zur⌬46 -91 probably titrates wild-type Zur away from the znu operator by forming heterodimers (see above). Since the individual components Zur25 and wild-type Zur bind to DNA, it is surprising that the heterodimer seems to bind less efficiently or not at all.
Metal Ion Dependence of the DNA-binding Activity of Zur-DNase I footprinting assays were also employed to study the metal ion dependence of Zur-DNA binding. Purified Zur as isolated bound to the znu operator even without the addition of metal ions, which indicates that the buffers and the protein preparation contained sufficient zinc. Zur acquired zinc either in the cell or during purification; the zinc/protein ratio in the sample of purified Zur was 1:5 as determined by atom absorp- tion spectroscopy (see below). Addition of 5 M Zn 2ϩ or Mn 2ϩ enhanced DNA protection by this partially Zn 2ϩ -saturated Zur preparation. The zinc chelators TPEN (Ͻ2 M) and EDDS (Ͻ100 M) abolished DNA binding of Zur. The high concentration of EDDS necessary for Zur inactivation demonstrates the strong affinity of reduced Zur for Zn 2ϩ . The greater efficiency of TPEN compared with that of EDDS reflects its by far higher affinity for Zn 2ϩ , as has also been observed in plate assays (3). Other metal chelators such as EGTA and EDTA mainly complex the Mg 2ϩ essential for DNase I activity and therefore could not be used in adequate concentrations to chelate Zn 2ϩ . Cu 2ϩ , Fe 2ϩ , and Mn 2ϩ at Ͼ250 M impaired DNA protection. Interestingly, Mn 2ϩ (Ͼ250 M, but not at 25 M) altered the DNA fragmentation pattern generated by DNase I digestion when P i was present. Ni 2ϩ was not stable under the strong reducing conditions required.
Zur Binds Zn 2ϩ Specifically-The zinc content of wild-type Zur was determined. When purified on a Ni 2ϩ column under nonreducing or mildly reducing conditions (10 mM 2-mercaptoethanol) and after removal of imidazole, the Zur preparation barely contained any zinc (ϳ1 zinc ion/5 protein molecules). Under these conditions, Zur was in the oxidized form as judged by SDS-PAGE. Only when Zur was reduced by DTT and loaded with ZnSO 4 and then unbound Zn 2ϩ was removed by gel filtration was zinc detected at 1.2-1.7 zinc ions/Zur molecule, i.e. one Zur dimer bound at least 2 zinc ions. Zur⌬46 -91 contained as much zinc as wild-type Zur, indicating that zinc binding is not abolished by the deletion. In the presence of equimolar concentrations of Zn 2ϩ and Fe 2ϩ , the content of bound Zn 2ϩ decreased to 80% of the value measured without Fe 2ϩ ; the remaining 20% of the bound ions were Fe 2ϩ . This result suggests that wildtype Zur has much higher affinity for Zn 2ϩ than for Fe 2ϩ .
In Vivo Footprinting Assays-In DNase I footprinting assays, wild-type Zur and all the mutant Zur proteins except Zur31 and Zur⌬46 -91 bound to the znu operator, although the mutant Zur proteins were not able to repress the znuAЈ-lacZ fusion in vivo. To determine whether the DNA binding of the mutant Zur proteins was due to the nonphysiological conditions or whether the mutant proteins occupy the Zur-binding site on the znu operator even in vivo, an in vivo footprinting assay was designed. Whole cells of the zur-19 mutant strain SIP575 and its parental strain (MC4100) were treated with DMS, and the methylated genomic DNA was isolated, followed by one-sided PCR using Taq DNA polymerase. With genomic DNA, peaks were of low intensity and difficult to evaluate (data not shown). Therefore, footprinting was also carried out with cells transformed with the znu operator region on the pA-CYC184 vector (pSP97/47), and to compensate for the multicopy effect, the cells were cotransformed with the zur allele on pT7-5 (pSP100/3 for wild-type zur and pSP99/224 for zur-19) (Fig. 5). The results were essentially the same as with genomic DNA, but the peaks were more distinct.
With the ZnuC-coding strand (Fig. 5A), signals for Zur19 (trace 2) were identical to those for in vitro methylated pSP97/47 in the absence of Zur (negative control, trace 1). With wild-type Zur (traces 3a and 3b), signal intensities were the same as with the negative control only for nucleotides before nt 1034 and after nt 1057 of section 170. In the region in between, most signals diminished (downward-pointing arrows), but one increased (upward-pointing arrows), indicating that most residues were protected by Zur binding in vivo, whereas one residue was more susceptible to methylation by DMS (hypersensitive). This effect was greater with increased expression of wild-type Zur (trace 3b). The enhanced reactivity of the hypersensitive residue can be explained either by protein binding-induced structural alterations of the DNA that lead to a more exposed reaction site or by the formation of a so-called hydrophobic pocket by the protein on the DNA that attracts DMS and hence increases its local concentration. The latter interpretation seems less likely since such sites of increased reactivity are often observed at the boundary of the protein on the DNA. Hypersensitivity was not observed in DNase I footprinting assays, probably because of steric hindrance of the much larger DNase I protein in comparison with the small DMS molecule.
With the ZnuA-coding strand (Fig. 5B), most of the nucleotides located in the region around nt 1039 -1054 of section 170 (Zur-binding site of the znu operator) were protected from methylation by wild-type Zur (trace 3, downward-pointing arrows) compared with the in vitro methylated pSP97/47 DNA in the absence of Zur (negative control, trace 1). These same nucleotides were not protected in cells with Zur19 (trace 2), which indicates that Zur19 does not bind to the znu operator in vivo. Thus, the DNA binding of Zur19 in DNase I footprinting assays was caused by the nonphysiological conditions. The protection zone in the in vivo footprinting assay was generally smaller than in the DNase I footprinting assay presumably because the DMS molecule is much smaller than the bulky DNase I protein.

DISCUSSION
Native Zur is a dimer, even in the absence of zinc or other metal ions, as indicated by gel filtration experiments in the presence of chelators. A regulatory sequence of dyad symmetry was defined within the znu operator; this sequence is sufficient for conferring zinc-dependent and Zur-dependent repression. Zur protein was shown to bind directly and specifically to DNA with this sequence using three independent assays of DNA binding: in vivo titration, DNase I footprinting, and in vivo footprinting.
Zur occupied its binding site only in the presence of zinc or other divalent metal cations at low concentrations, as shown by the DNase I footprinting data. Zur protected a 29-nt approximate palindrome on each strand of the znu operator with a 3Ј-stagger of 4 nt. This footprint resembles that of typical DNA-binding dimers such as classical helix-turn-helix proteins, e.g. the CI repressor from bacteriophage (23). The observed 3Ј-stagger is indicative of coverage of the minor groove at the ends, but provides no information about the protein-DNA recognition contacts. Even at high concentrations of Zur, no polymerization and extension of the protected region occurred, as has been observed with Fur at several Fur-dependent operators in footprinting experiments (e.g. Refs. 24 and 25) and by electron and atomic force microscopy (26).
Zur is active only in the reduced form. As a cytoplasmic protein, it has predominantly reduced thiols rather than oxidized disulfides due to the reducing conditions in the cytoplasm (27). In vitro, the cysteine residues of Zur are easily oxidized to disulfides, as judged by the slower migration on SDS-polyacrylamide gels under reducing conditions compared with nonreducing conditions. Oxidized Zur does not bind DNA or considerable amounts of Zn 2ϩ .
Zur seems to be widespread among bacteria, even in Grampositive bacteria and cyanobacteria, as indicated by sequence similarity searches. In B. subtilis, apart from Fur and PerR (peroxide stress response regulator) (11), the third homologue , strain MC4100 (wild-type zur) bearing pSP97/47 (trace 3a), and strain MC4100 bearing pSP97/47 and pSP100/3 (wild-type zur) (traces 3b) were exposed to DMS in vivo, and plasmid pSP97/47 was methylated with DMS in vitro (traces 1). Plasmid DNA from each sample was isolated and subjected to one-sided PCR with Taq DNA polymerase and the 5Ј-fluorescein-labeled primer YEB7 (A) or YEB3 (B) as described under "Experimental Procedures." The signals in the scan represent nucleotides that reacted with DMS, thereby generating truncated labeled extension products. Nucleotides protected by a DNA-interacting protein are not exposed to DMS and do not give a signal in this assay. Residues protected from DMS modification upon Zur binding are indicated by downward-pointing arrows; signals that increased are indicated by upward-pointing arrows. The nucleotide sequence resulting from a standard dideoxynucleotide sequencing reaction performed with the same end-labeled primer is shown at the bottom. Due to the different migration of samples and the standard on the sequencing gel, the signals may be shifted by 1-2 nt relative to their true position. Nucleotides are numbered according to E. coli genome section 170. Note that due to the different principle of the two methods, opposite DNA strands are analyzed in the in vivo footprinting assay and in the DNase I footprinting assay with the same fluorescein-labeled primer.
YqfV can act as Zur (5). The sequence similarity to E. coli Zur is not strong owing to the distant relationship between Gramnegative and Gram-positive bacteria. A zur gene has also been found in L. monocytogenes (6). Based on sequence similarity, we expect that the proposed Fur protein in Staphylococcus epidermidis (28) is Zur rather than Fur (Fig. 6, FUR_STAEP). It is possible that other proteins designated as Fur homologues will turn out to be Zur proteins. Synechocystis sp. (29) may also have a Zur protein. Of the three homologues, Slr1738 and Sll1937 are more similar to Zur than to Fur from E. coli. It is noteworthy that sll1937 is located directly upstream and is divergently transcribed of the genes encoding the ZnuACB equivalents in Synechocystis sp.
In many of the partially sequenced genomes of various bacterial species, a Zur equivalent is found, e.g. in Salmonella typhimurium LT2 (Washington University Genome Sequencing Center (WUGSC)), Salmonella typhi (Sanger Center), Salmonella paratyphi (WUGSC), Klebsiella pneumoniae (WUGSC), Yersinia pestis (Sanger Center), Vibrio cholerae (Institute for Genomic Research), Bordetella pertussis (Sanger Center), Caulobacter crescentus (Institute for Genomic Research), Pseudomonas aeruginosa (Pseudomonas Genome Project), Neisseria gonorrhoeae (University of Oklahoma's Advanced Center for Genome Technology), and Neisseria meningitidis (Sanger Center). Since these organisms also pos-sess a system homologous to Znu, we propose that these proteins likewise are regulators of zinc uptake. Some of these Zur-equivalent proteins are aligned in Fig. 6.
Since all mutant Zur proteins could be purified in large amounts, it can be ruled out that the Zur phenotype is the result of low protein expression or of an unstable Zur protein that is quickly degraded by cellular proteases. This suggests that the amino acids altered in the mutant Zur proteins are important for the structure and/or function of native Zur. The alignment of the amino acid sequences of the mutant zur products with putative Zur proteins from other organisms reveals that all of the point mutations occur in well conserved areas (Fig. 6).
Zur⌬46 -91, which has an internal deletion (amino acids 46 -91) in the amino-terminal part of Zur, bound as much Zn 2ϩ as wild-type Zur and formed dimers and heterodimers with wild-type Zur, but did not protect DNA from DNase I cleavage. Apparently, this region is responsible for binding of DNA, but is not essential for dimerization or Zn 2ϩ binding.
Negative complementation by an excess of the mutant Zur protein can be explained by competition for the corepressor Zn 2ϩ or by formation of less active heterodimers with wild-type Zur. Zur25 (Leu 48  . The partially sequenced Np20 from P. aeruginosa (38) and IviXI from V. cholerae (39), which have been identified as being implicated in virulence, are part of ZUR_PSEAE and ZUR_VIBCH, respectively. The amino acid changes in each of the eight mutant Zur proteins and the mutant designations are given above the corresponding residues in E. coli wild-type Zur.
sors, but were able to complement negatively, i.e. to bind Zn 2ϩ or to form dimers. In all these mutant proteins, residues in the amino-terminal or central region are changed or deleted. Thus, this region or at least the individual residues are not crucial for binding of Zn 2ϩ or dimerization. In contrast to Zur27, Zur25 inhibited DNA binding of wild-type Zur in DNase I footprinting experiments and showed negative complementation, both of which can be accounted for by the prevalence of Zur25 dimer formation. All the mutant Zur proteins except Zur31 and Zur⌬46 -91 were in principle able to bind DNA since they protected DNA from DNase I cleavage in vitro, albeit not in vivo. This indicates that crucial DNA contacts made by the Zur protein are located in the region around residue 65. Moreover, this region is highly conserved in all Zur sequences and corresponds to the second helix in the proposed helix-turn-helix motif in Fur that has previously been suggested to mediate interactions with DNA (30). However, the mutation in Zur27 seemed to influence the dimer formation, although the mutation is close to the assumed DNA-binding site.
The zinc content estimated for Zur was at least 2 Zn 2ϩ ions/Zur dimer, which suggests that at least one metal-binding site of the Zur monomer is occupied by Zn 2ϩ . For establishing the metal/Zur stoichiometry, determination of the absolute protein concentration was critical. As shown in the alignment in Fig. 6, Zur lacks the cluster of histidines around amino acid 90 that is conserved in Fur and that is assumed to be involved in Fe 2ϩ binding of Fur (31,32). Of the nine cysteines present in E. coli Zur, Cys 17 , Cys 113 , Cys 152 , Cys 158 , and possibly Cys 88 are probably not involved in Zn 2ϩ binding because they are not conserved in at least one or even all Zur proteins from other bacterial genera (Fig. 6). Interestingly, E. coli Fur possesses one structural tight-binding zinc site/monomer in addition to the regulatory iron-binding site that senses the intracellular iron concentration (33). The zinc is coordinated tetrahedrally by at least one histidine (33) and Cys 93 and Cys 96 (34). These cysteines are perfectly conserved in all Zur proteins proposed (Fig. 6) and may therefore serve as Zn 2ϩ ligands also in Zur.
As observed in the DNase I footprinting studies, Zn 2ϩ at very low concentrations acts as a corepressor in Zur and can be replaced in vitro by other divalent metal ions such as Mn 2ϩ . A similar unspecificity for divalent cations in vitro is exhibited by Fur with Co 2ϩ , Cd 2ϩ , Cu 2ϩ , Fe 2ϩ , Mn 2ϩ , or Zn 2ϩ and by DtxR and IdeR with Cd 2ϩ , Co 2ϩ , Fe 2ϩ , Mn 2ϩ , Ni 2ϩ , or Zn 2ϩ acting as corepressors in DNase I footprinting experiments (8,24,35,36), although the response of Fur-or IdeR-regulated genes is clearly iron-specific in vivo. It is unknown which factors determine binding of the individual metal ion to the regulator in the cell; therefore, the results from the DNase I footprinting experiments can explain the in vivo situation only to a limited extent. This limitation is also documented by the binding of most of the mutant Zur proteins to the znu operator in vitro in DNase I footprinting assays, but not in vivo, as was indicated by depression of the znuЈ-lacZ fusion in the zur mutants and the lack of protection of the znu operator by at least Zur19 in in vivo footprinting experiments.
The znu gene cluster was derepressed by divalent metal ions such as Fe 2ϩ , Mn 2ϩ , Cu 2ϩ , and Cd 2ϩ . This can be explained either by direct binding of these divalent cations to Zur, thereby changing the conformation into an inactive form that no longer represses znuACB, or by these metal ions exerting their derepressing effect indirectly by lowering the intracellular zinc level. The latter possibility may be achieved by the metal ions competing with Zn 2ϩ for an unspecific transporter and thereby inhibiting uptake of Zn 2ϩ . Yet in regulation studies with znu mutants, the znu system does not play a part because it is inactive. At high concentrations, the derepressing divalent cations did indeed impair Zur binding in DNase I footprinting experiments, but with the restriction as pointed out above regarding the discrepancy between the results for DNase I footprinting and the in vivo situation. Purified Zur bound Fe 2ϩ far less efficiently than Zn 2ϩ , and Fe 2ϩ did not considerably hinder binding of the corepressor Zn 2ϩ to Zur, at least under the in vitro conditions and concentrations tested, thereby providing no hint for a direct interaction of iron with Zur.