DNA Distortion Mechanism for Transcriptional Activation by ZntR, a Zn(II)-responsive MerR Homologue in Escherichia coli *

MerR-like DNA distortion mechanisms have been proposed for a variety of stress-responsive transcription factors. TheEscherichia coli ZntR protein, a homologue of MerR, has recently been shown to mediate Zn(II)-responsive regulation ofzntA, a gene involved in Zn(II) detoxification. To determine whether the MerR DNA distortion mechanism is conserved among MerR family members, we have purified ZntR to homogeneity and shown that it is a zinc receptor that is necessary and sufficient to stimulate Zn-responsive transcription at the zntA promoter. Biochemical, DNA footprinting, and in vitro transcription assays indicate that apo-ZntR binds in the atypical 20-base pair spacer region of the promoter and distorts the DNA in a manner that is similar to apo-MerR. The addition of Zn(II) to ZntR converts it to a transcriptional activator protein that introduces changes in the DNA conformation. These changes apparently make the promoter a better substrate for RNA polymerase. We propose that this zinc-sensing homologue of MerR restructures the target promoter in a manner similar to that of other stress-responsive transcription factors. The ZntR metalloregulatory protein is a direct Zn(II) sensor that catalyzes transcriptional activation of a zinc efflux gene, thus preventing intracellular Zn(II) from exceeding an optimal but as yet unknown concentration.

Zinc is an essential element that must be maintained at certain levels within all cells. However, like many transition elements, zinc is also harmful at elevated concentrations. Zinc starvation and zinc toxicity both lead to transcription of a number of recently characterized Escherichia coli genes that encode Zn(II) uptake or export proteins. The znuABC operon in E. coli encodes three proteins involved in Zn(II) uptake (1). The zur gene, which encodes a zinc-responsive homologue of the Fur (ferric uptake regulation) protein, mediates regulation of this operon. When Zn(II) levels in the media fall to a critical point, Zur is proposed to de-repress transcription of the znuABC operon. In contrast, when Zn(II) levels in the cell are too high, a separate Zn(II) efflux system is activated. In this system, the zntR gene has been shown to be essential for Zn-induced expression of the zinc-exporter, ZntA (2). ZntA is a cation-translocating ATPase that couples Zn(II), Cd(II), or Pb(II) export with ATP hydrolysis (3)(4)(5). Several mechanisms for stress-responsive transcriptional control are known in microbial systems (6), however few have been established at the molecular level for essential metals such as zinc.
Only a handful of Zn(II)-responsive metalloregulatory proteins have been identified in eukaryotic and prokaryotic systems (7,8). One candidate for a Zn-specific metalloregulatory protein is the zinc finger protein MTF-1 that up-regulates expression of mammalian metallothionein genes in response to elevated concentrations of Zn(II) and other heavy metals (9,10). ZAP1 is an unrelated zinc finger protein that exerts positive control on a variety of yeast genes including those involved in zinc export (11,12). Several Zn(II)-responsive repressors have also been characterized, such as the SmtB family (13)(14)(15) and Zur (1,16,17). While the mechanism of Zur repression has not been characterized in vitro, the SmtB protein has been crystallized (18) and zinc-induced repression has been shown (19). None of these proteins, however, show significant homology with E. coli ZntR (2).
The predicted sequence of the ZntR protein exhibits 34% identity with the MerR metalloregulatory protein, an Hg(II)specific receptor that regulates expression of bacterial plasmidencoded mercury detoxification genes (20 -29). In vivo experiments by Brocklehurst et al. (2) have revealed that zntR is a trans activator of zntA transcription, whereas gel shift assays have shown that ZntR binds to the zntA promoter (P zntA ). A construct containing a 2-bp 1 deletion in the spacer region of the zntA promoter displayed constitutive activity, indicating that the wild-type 20-bp spacing between the Ϫ35 and Ϫ10 sites plays a role in regulation. These experiments indicate significant similarities between ZntR and MerR function in vivo and lead us to address the question of how ZntR activates transcription. We investigated this question using tools developed to probe the regulatory mechanism of MerR.
Suboptimal spacing (19 bp) between the Ϫ35 and Ϫ10 promoter elements is a key feature of the merT promoter (P T ) regulated by MerR (30) and is at the heart of the DNA distortion mechanism. Promoters regulated by members of the MerR family typically have spacer elements longer than the consensus length of 17 bp which makes them poor substrates for RNA polymerase (RNAP) (31). MerR controls transcription of the mer genes while bound to the spacer region at a palindrome located between the Ϫ35 and Ϫ10 sites of P T (21,22). Several members of the MerR family are also known to bind in the spacer regions of their target promoters, including SoxR (32), TipAL (33), BmrR (34), and Mta (35).
In the MerR DNA distortion mechanism, the transition from repression to activation involves several alterations in the local DNA structure. A series of specific distortions are proposed to make the promoter a more optimal substrate for RNAP ( Fig. 1) (24,36,37). Evidence for these conformational changes comes from both physical studies and comparison of nuclease cleavage patterns for repressed and activated states of the MerR⅐DNA complex. The DNase I cleavage pattern observed for the MerR⅐DNA complex is very similar to that observed for the CAP⅐DNA complex (37). Inspection of the CAP⅐DNA complex crystal structure reveals that sites of cleavage are immediately 3Ј to the DNA distortion sites. These distortion sites are two 40°kinks in the DNA which widen the exposed minor groove and bend the DNA toward the CAP protein (37,38). These results led to a model in which apo-MerR bends the DNA toward itself, producing two kinks in the DNA structure that appear as hypersensitive bands within the protected region in DNase I footprinting experiments (see Fig. 1A). Addition of Hg(II) to MerR relieves these bends at the two kink sites, resulting in a loss of the DNase I hypersensitive bands. This model was corroborated by phase bend studies that confirmed bending of the DNA toward the protein for apo-MerR and a relaxation of these bends by Hg-MerR (37).
Studies of E. coli RNAP with MerR and the mer operator further corroborate this mechanism. RNAP is able to bind to the Ϫ35 site of P T with apo-MerR bound, however the Ϫ10 site remains inaccessible (22). RNAP binds adjacent to apo-MerR in this closed complex. In the DNA distortion mechanism, Hg(II) binding to MerR leads to underwinding of the DNA (36), resulting in a Cu-OP hypersensitive site at the center of the operator (24). In vivo methylation and permanganate footprinting are also consistent with a change in the conformation of the RNAP⅐MerR complex (39). One effect of this underwinding, which is different from unbending, is to bring the Ϫ35 and Ϫ10 sites into proper alignment for RNAP binding to both simultaneously so that transcription can proceed (Fig. 1B). The underwinding may also facilitate the energetics of strand separation, a requisite step in open complex formation (37). Thus, underwinding and the relaxation of bending act together to remodel the promoter, making it a better substrate for RNAP and dramatically increasing the transcription initiation rate (28).
In order to delineate which mechanistic aspects are conserved among MerR family members (40 -43), we have examined the mechanism that ZntR employs in transcriptional regulation of zntA. Protein/DNA footprinting and in vivo transcriptional assays reveal that, like MerR, ZntR functions primarily as a transcriptional activator, but it only weakly, if at all, represses expression of the zntA gene. Our results are consistent with a ZntR mechanism in which a series of MerRlike DNA unbending and underwinding steps restructure the promoter, allowing efficient transcription of zntA.

EXPERIMENTAL PROCEDURES
Plasmid Construction and Bacterial Strains-All DNA manipulations were carried out using standard procedures (44). Plasmid construction was confirmed by DNA sequencing, and all cloning was carried out in the E. coli strain DH5␣.
ZntR Cloning and Purification-The zntR gene was cloned from the E. coli chromosome using the primers ZntR N-terminal (5Ј-CTA GTG GAG TAC ATA TGT ATC GCA TTG G-3Ј) and ZntR C-terminal (5Ј-GTA ATC CTG CGG ATC CAA AAA ATC AAC AAC C-3Ј). The 461-bp PCR fragment was digested with NdeI and BamHI and inserted into pET11c (Novagen) digested with the same enzymes. The resulting overexpression plasmid (pET11cZntR) was transformed into BL21(DE3) cells (Novagen). The cells were grown in 9 liters of Luria Bertani broth with shaking at 37°C and induced with 400 M IPTG at A 600 ϭ 0.6. The cells were harvested by centrifugation 2.5 h after IPTG induction and then stored at Ϫ80°C. The pelleted cells were frozen and thawed three times (45) and then resuspended in 250 ml of Tris buffer A (50 mM Tris-Cl, pH 8.0, 2 mM EDTA, and 5 mM DTT). The cell debris was removed by centrifugation, and the protein in the supernatant was precipitated with 45% (NH 4 ) 2 SO 4 . The precipitated protein was then resuspended in 10 ml of Tris buffer A and desalted on a Sephadex G-25 column. The crude protein extract was subsequently loaded onto a Heparin column equilibrated with Tris buffer B (20 mM Tris, pH 8.0, 5 mM DTT). Following elution with a NaCl gradient, the protein fractions were collected and (NH 4 ) 2 SO 4 was added to a final concentration of 1.2 M. The protein was then loaded onto a Phenyl-Sepharose High Performance hydrophobic column (Amersham Pharmacia Biotech) in Tris buffer C (50 mM Tris, pH 8.0, 1.2 M (NH 4 ) 2 SO 4 , 5 mM DTT). The protein was eluted with a decreasing salt gradient, coming off at 0.5-0.3 M (NH 4 ) 2 SO 4 , and the protein-containing fractions were concentrated to 2 ml. As a final purification step, the protein was loaded onto a High Load Superdex 75 gel filtration column (Amersham Pharmacia Biotech) equilibrated with Tris buffer D (50 mM Tris, pH 8.0, 250 mM NaCl, 5 mM DTT). The ZntR fractions were concentrated to ϳ1-2 ml and showed a single band on overloaded SDS-PAGE gels. Protein was stored at Ϫ80°C in Tris buffer D with 5% glycerol. The molecular mass of ZntR (calculated: 16179.3 Da with first Met) was found by matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) (PerSpective Biosystems Voyager-DE) to be 16178.1 Da using a sinapinic acid matrix with myoglobin as the calibration standard.
Protein-Metal Ratio Determination-A calculated extinction coefficient for the apo-protein (6700 1/M) at 280 nm (46) in 20 mM MES, pH 6.0, 6 M guanidine-HCl was used to standardize the Bradford assay (Bio-Rad). The Bradford assay was found to overestimate the ZntR concentration by 2.65 when using IgG as the standard. With this correction factor, protein concentrations were routinely determined by the Bradford assay. Metal concentrations were determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES).
Zn(II) Binding Assays-To prepare Zn-ZntR, 2-4 equivalents of ZnSO 4 was added to 10 -20 M ZntR in 50 mM Tris, pH 8.0, 250 mM NaCl, 1-2 mM DTT in an Amicon ultrafiltration cell. The solution was stirred at 4°C for 30 min and then washed three times with buffer to remove excess metal. Zn-ZntR was also prepared under anaerobic conditions in a glove box. The ultrafiltration method was the same as the aerobic procedures, however the buffer used was 50 mM phosphate, pH 7.8, with no thiol reductant.

FIG. 1. MerR DNA distortion model.
A, apo-MerR bends the DNA, producing two kinks (open arrows) that are symmetrically spaced around the center of the operator (vertical line). Upon binding Hg(II), MerR relaxes these bends and unwinds (circular arrow) the center of the operator. B, with apo-MerR bound to the operator, RNAP is only able to access the Ϫ35 site and transcription is repressed. The unkinking and untwisting introduced by Hg-MerR allows RNAP to access both the Ϫ35 and Ϫ10 sites, and transcription is activated.
G-3Ј). The 644-bp fragment was digested with EcoRI and BamHI and inserted into pUC19 (New England Biolabs) digested with the same enzymes creating pUC19Znt. Two more primers (Zntfoot1-5Ј-GCT GCG CAA CTG TTG GGA AGG GC-3Ј and Zntfoot2-5Ј-GAG AGA GTT GGC GCC CGG GAA CAT GCG-3Ј) were then used for PCR with pUC19Znt to clone out a smaller portion of the promoter. The 486-bp PCR fragment was digested with EcoRI and AvaI and inserted into pUC19 digested with the same enzymes. The resulting plasmid, pUC19Zntfoot, was used for labeling in the footprinting assays.
Primer Extension-Total RNA was isolated from exponentially dividing cells of strain DH5␣ using the Qiagen RNeasy RNA isolation kit. Cells induced with zinc were exposed to 1 mM ZnSO 4 for 1 h prior to RNA isolation. Primer ZntA:PE1 (5Ј-GCT TTC TTG CCG TGA TTG TCA GG-3Ј), labeled with [␥-32 P]ATP by T4 polynucleotide kinase (New England Biolabs), was used with 10 g of total RNA for primer extension analysis. Primer and RNA were heated to 65°C for 5 min, quenched on ice, added to a reaction mixture of M-MuLV reverse transcriptase (New England Biolabs), and placed at 42°C for 1 h. Sequencing of pUC19Znt was carried out using labeled primer ZntA: PE1 as directed in the CircumVent Thermal Cycle Dideoxy DNA Sequencing Kit (New England Biolabs). Primer extension and sequencing samples were run together on an 8% polyacrylamide, 8 M urea, 1.2ϫ TBE sequencing gel. P zntA Labeling for Footprinting-To label the non-template strand of the promoter, pUC19Zntfoot was digested with BamHI, labeled with [␣-32 P]dATP using the Klenow fragment of DNA polymerase (New England Biolabs), and then digested with EcoRI. To label the template strand, the fragment was digested first with EcoRI, labeled, and then digested second with BamHI. In each case, the labeled fragments were purified by gel electrophoresis.
DNase I Footprinting-DNase I footprinting reactions were carried out in 50 l of footprinting buffer (10 mM Tris, pH 8.0, 2 mM MgCl 2 , 1 mM CaCl 2 , 100 mM K-glutamate, 1 mM DTT, 100 g/ml bovine serum albumin, 2.5 g/ml sonicated salmon sperm DNA, 1 M TPEN, 5% glycerol). Labeled DNA, ZntR, ZnSO 4 , and RNAP were added to the reaction mixture in sequential order with a 5-min incubation at 37°C after each addition. The mixture was then incubated for 30 min at 37°C before DNase I digestion. The dinucleotide initiator UpA (100 M) and ATP, GTP, and UTP (10 M) were added 15 min before DNase I digestion. All RNAP complexes were also challenged with heparin (50 g/ml) prior to DNase I digestion. 2.5 l of 20 nM DNase I was added to the reaction mixture and allowed to incubate for 1 min at 37°C. The reaction was terminated with the addition of 175 l of DNase I stop solution (560 mM NH 4 OAc, 30 g/ml yeast tRNA, 86% EtOH) and then was precipitated on dry ice for Ͼ 30 min. The pellets were washed with 70% EtOH, dried under vacuum, and then resuspended in loading buffer (0.5ϫ TBE, 80% formamide, 0.05% xylene cyanol FF, 0.05% bromphenol blue). The samples were loaded onto a 7% polyacrylamide sequencing gel containing 7 M urea and 1.2ϫ TBE. The same DNA fragment was also cleaved in a Maxam-Gilbert guanine-specific reaction (44) for use as a DNA sequence ladder. After drying, the gel was applied to a Molecular Dynamics PhosphorImager. The signals were digitized and analyzed using the ImageQuant Version 1.2 program (Molecular Dynamics).
KMnO 4 Footprinting-KMnO 4 footprinting reactions were carried out in 20 l of footprinting buffer minus salmon sperm DNA and glycerol and with 0.5 mM DTT instead of 1 mM. Incubation times were the same as described above for DNase I footprinting. RNAP complexes were challenged with 50 g/ml heparin before the addition of KMnO 4 . The KMnO 4 procedures are the same as described previously (22).
5-Phenyl-1,10-Phenanthroline-Cu Footprinting-Phenylphenanthroline footprinting procedures were the same as described previously with a few modifications (24). The reactions were carried out in 20 l of footprinting buffer minus salmon sperm DNA and with 1 mM mercaptopropionic acid instead of DTT. Incubation times were the same as described above for DNase I footprinting. Ten microliters of 10 mM CuSO 4 and 10 l of 10 mM 5-phenyl-1,10-phenanthroline were mixed together and immediately diluted to 400 l with H 2 O. One microliter of the diluted Cu-phenanthroline mixture and 1 l of 100 mM mercaptopropionic acid were added to the protein/DNA solution and incubated for 4 -5 min at room temperature. One microliter of 56 mM 2,9-dimethyl-1,10-phenanthroline was added, and the mixture was incubated for 4 -5 min longer. Three microliters of 3 M NaOAc and 60 l of EtOH were added to the reaction, and the DNA was precipitated on dry ice Ͼ 30 min. Preparation and electrophoresis of the precipitated DNA are the same as described above for DNase I footprinting.
In Vitro Run-off Transcription Assays-Transcription template was prepared using the (Ϫ48) and (Ϫ47) pUC sequencing primers (New England Biolabs) and the plasmid pUC19Zntfoot. To label the template, [␣-32 P]dATP was incorporated into the DNA fragment during the PCR reaction. The PCR fragment was then digested with BbvI to make a 171-bp fragment, gel purified, and quantitated by comparison with DNA standards stained with ethidium bromide. Transcription run-off experiments were carried out in 20 l of chelexed footprinting buffer minus salmon sperm DNA. The TPEN concentration was 10 M rather than 1 M as used in the footprinting assays. Procedures were the same as described previously (22) with a few modifications. Two nanomolar labeled DNA was used in the reaction rather than 4 -6 nM, 5 Ci of [␣-32 P]UTP was used rather than 10 Ci, and the reaction was terminated with 200 l of stop buffer (10 mM EDTA, pH 8.0, 100 g/ml RNA) prior to phenol-chloroform extraction. The samples were loaded onto the same type of sequencing gel used in the footprinting assays, and the signals were digitized in the same manner as the footprinting gels. RNA levels in each lane were determined relative to the lanes with RNAP alone, whereas the labeled DNA was used to normalize the DNA loading for each lane. The Zn titration transcription results were fitted to a sigmoidal function using the Igor Pro Version 2.04 software program (Wavemetrics, Inc., Lake Oswego, OR). Fig. 2 shows an SDS-PAGE gel of different steps in the ZntR purification process. A 9 liter prep typically yielded ϳ180 mg of pure protein. The purified protein was found by ICP-AES to contain less that 0.05 Zn/monomer and less than 0.05 Cu/monomer. Zn-ZntR was easily prepared by adding excess Zn(II) to ZntR in aerobic buffer containing DTT. The protein was found to bind 0.9 Ϯ 0.3 Zn/monomer. In aerobic buffer without DTT present, the complex was unstable and precipitated. In anaerobic conditions with 2 equivalents of Zn added but no DTT present, the protein bound 2.1 Ϯ 0.4 Zn/monomer.

ZntR Purification and Metal Content-
Transcription Start Site-To map the start of transcription and determine the approximate location of essential promoter elements, primer extension analysis of the zntA transcript was carried out (Fig. 3). While Brocklehurst et al. (2) have recently reported a primer extension analysis of zntA, in our hands, the start of transcription appears to occur one base 5Ј to the previously reported result. We find that the zntA start of transcription mapped to a T rather than an A (see Fig. 3  tained using a temperature of 37°C for the reverse transcription reaction (44) while our reverse transcription was carried out at 42°C to reduce possible secondary structure that can abort termination of the cDNA before reaching the 5Ј end of the transcript. Possibly, this temperature difference results in a slightly truncated primer extension product at 37°C as com- pared with 42°C. Alternatively, there may be two different transcripts formed upon induction of zntA. We did observe a slightly shorter primer extension product in a faint band below the primary product ( Fig. 3 and other data not shown). Resolution of these competing hypotheses awaits further experimentation. A transcription start site at the T shown in Fig. 3 is consistently observed in our experiments.
DNA/Protein Interactions-The footprinting results for both strands of P zntA are shown in Fig. 4. On the template strand (Fig. 4A), ZntR alone protects bases Ϫ37 to Ϫ12 from DNase I cleavage, whereas bases Ϫ11 and Ϫ30 show hypersensitivity to cleavage (lane 2). On the non-template strand (Fig. 4B), bases Ϫ35 to Ϫ10 are protected from cleavage, whereas bases Ϫ17 and Ϫ42 are hypersensitive to DNase I cleavage (lane 2). Upon addition of Zn(II), the length of the protected region on both strands does not change (Fig. 4, lanes 3). However, the hypersensitive sites within the footprinted region decrease (Ϫ30 on template and Ϫ17 on non-template), whereas the protection of Ϫ27 on the template strand and Ϫ20 on the non-template strand is diminished. The other hypersensitive bands outside the footprint do not change (Ϫ11 on template and Ϫ42 on non-template). When RNAP is added to ZntR without Zn(II) present (lanes 4), there is little change in the footprint from ZntR alone (lanes 2). A similar result is observed with the addition of nucleotide triphosphates ATP, UTP, and GTP and the dinucleotide initiator UpA (lanes 5) to ZntR and RNAP. As shown in lanes 6, the addition of ZntR, Zn(II), and RNAP to the reaction mixture results in an extended footprint covering Ϫ55 to Ϫ52 and Ϫ46 to ϩ21 on the template strand and Ϫ44 to ϩ25 on the non-template strand. Weaker protection is observed at ϩ22 to ϩ25 on the template and at Ϫ55 to Ϫ50 and ϩ26 to ϩ28 on the non-template strand. Hypersensitive bands are apparent at Ϫ60, Ϫ39, and Ϫ38 on the template strand and at Ϫ58, Ϫ47, and Ϫ46 on the non-template strand. The addition of ATP, UTP, GTP and UpA allows the production of a 13-base RNA transcript of zntA before the incorporation of the first cytosine residue. The resulting halted transcription complex cannot proceed any further without the addition of CTP. Fig. 4, lanes 7 shows the footprint of this complex that is formed with the addition of ZntR, Zn(II), RNAP, UpA, and the three NTPs to the reaction mixture. On the template strand, Ϫ36, Ϫ27, and Ϫ15 are less protected, whereas ϩ22 to ϩ32 are more protected. Similarly on the non-template strand, the regions between Ϫ45 to Ϫ32 and Ϫ17 to Ϫ12 are less protected, whereas ϩ26 to ϩ35 are more protected. The hypersensitive bands seen in lanes 6 (ZntR, Zn(II), and RNAP) are also diminished, while new hypersensitivities appear at Ϫ34 to Ϫ36 on the non-template strand. These results are consistent with a 3Ј movement of RNAP after formation of the initiated, open complex.
Reagents that delineate the melted-out region also indicate that the addition of Zn(II) to ZntR causes the transcription bubble to move in the 3Ј direction. KMnO 4 footprinting indicates the position of unpaired thymidines (and sometimes other bases) in regions of non-base-paired DNA. Fig. 4, lanes 10 -15 indicate that KMnO 4 reactivity is greatly enhanced with ZntR, Zn(II), and RNAP in the reaction mixture. Without the addition of UpA and NTPs, unpaired bases are found at Ϫ10, Ϫ11, Ϫ12 on the template strand and Ϫ9, Ϫ5, Ϫ1, ϩ1 and ϩ4 on the non-template strand (lanes 15). As shown in lanes 16, adding UpA, ATP, UTP, and GTP shifts the reactive bases downstream (Ϫ10 to Ϫ12, ϩ1 to ϩ3, and ϩ11 to ϩ13 on the template and Ϫ1, ϩ1, ϩ4 on the non-template). Since it is unlikely that the transcription bubble extends from Ϫ12 to ϩ13, the footprinting data may indicate a mixture of the RNAP open complex and halted complex.
More subtle distortions of DNA structure have been detected with copper-phenanthroline complexes (24). 5-Phenyl-1,10phenanthroline-Cu cleaves DNA by binding to the minor groove and forming a copper-oxygen species, therefore it is sensitive to changes in minor groove geometry (47). Hypersensitivity to 5-phenyl-OP-Cu cleavage was visible only with ZntR, Zn(II), and RNAP together (Fig. 4, lanes 22). The hypersensitive areas were found in the center of the palindrome protected by ZntR (Ϫ22 to Ϫ24 on the template strand and Ϫ23 to Ϫ25 on the non-template) and near the transcription start site (Ϫ2 to Ϫ5 on the template strand and Ϫ1 to Ϫ4 and ϩ5 to ϩ7 on the non-template). These hypersensitivities were all diminished upon 5Ј movement of RNAP with the addition of UpA and the three NTPs (data not shown). The footprinting data for ZntR are summarized in Fig. 5.
Transcription Assays-In vitro run-off transcription assays using the P zntA template are shown in Fig. 6. RNAP alone (Fig.  6A, lanes 1 and 9) produces very low levels of the zntA tran-script. When ZntR is titrated with excess TPEN present (10 M) but no added Zn(II), zntA transcript levels increase very slightly up to 20 nM ZntR, and by 500 nM ZntR, they drop off to a value that is ϳ20% lower than RNAP alone (lanes 2-8 in Similar transcription results were obtained using ApA rather than UpA as the dinucleotide initiator (data not shown). Without TPEN in the transcription buffer, ZntR did not require added Zn(II) for activation (data not shown), suggesting that the buffer still contains sufficient levels of Zn(II) to activate ZntR despite chelexing overnight. Any zinc present in the buffer was too low to detect by ICP-AES analysis (detection limit for Zn ϭ 200 nM). DISCUSSION The in vitro results outlined here have shown that the purified ZntR protein is a Zn(II) receptor that is both necessary and sufficient to directly mediate zinc-responsive activation at P zntA . At the heart of the transcriptional activation mechanism employed by ZntR are DNA distortions that are proposed to make the promoter a more optimal substrate for RNAP. This DNA distortion mechanism is apparently a widespread attribute of MerR family members which generally mediate responses to a large variety of physical and chemical stresses.
A transcriptionally active, heparin-resistant open RNAP complex at P zntA is only formed when ZntR, Zn(II), and RNAP are present in the reaction mixture. The in vitro activation and the sequence of metal-dependent changes in the ZntR footprinting data show many similarities with those obtained for MerR (22,24,25,37) and SoxR, another MerR homologue in E. coli that responds to oxidative stress (40). SoxR contains a redox-active Fe-S cluster that activates transcription in the oxidized form but not in the reduced form (48). As shown in Fig.  7, MerR, SoxR, and ZntR all bind to a palindromic sequence located between the Ϫ35 and Ϫ10 regions of their target promoters. The spacing between these promoter elements is suboptimal in each case (19 bp for merT and soxS, and 20 bp for zntA). Without the activating metal, all three proteins produce DNase I hypersensitive areas within the footprinted region. With metal bound (or in the oxidized state of the Fe-S cluster in the case of SoxR), some of these hypersensitivities are significantly decreased (37,49). Ansari et al. (37) proposed that two of the DNase I hypersensitive sites in MerR footprints (marked with a asterisk in Fig. 7) arise from protein-induced bending that creates kinks in the DNA structure much like those observed in the CAP⅐DNA complex. Phasing and topoisomerase assays indicate that, upon Hg(II) binding, the protein-induced bends are relaxed and the operator is untwisted (36,37). These changes realign the promoter elements in a manner that leads to more effective RNAP binding. The footprinting data for ZntR indicates similar, but not identical, DNA distortions occurring upon Zn(II) binding. Furthermore, like Hg-MerR, Zn-ZntR greatly enhances RNAP binding and open complex formation at the target promoter. While promoter remodeling by sensor proteins may be one of several general mechanisms for transcriptional activation, a comparison of the detailed conformational changes reveals a variety of subtle differences in DNA structure from system to system.
Analysis of the crystal structure of DNase I⅐DNA complexes provides an understanding of how protein-induced bending affects DNase I cleavage. DNase I widens the minor groove while bending the DNA away from itself toward the major groove (50). Therefore, distortions in DNA structure that create a widened minor groove result in excellent substrates for DNase I and exhibit enhanced cleavage by this endonuclease (51)(52)(53). DNase I hypersensitivity within a footprint suggests that the DNA in the nucleoprotein complex is bent toward the major groove or that the DNA is highly flexible (52). Like MerR, SoxR, and ZntR, the CAP protein also exhibits these DNase I hypersensitivities within the footprinted region (54). The crystal structure of the CAP⅐DNA complex indicates the exact position of the protein-induced kinks (38), and the DNase I hypersensitivities are found immediately 3Ј to these sites (37). Therefore, the footprinting similarities seen between the MerR family and CAP coupled to the crystallographic data available for the CAP⅐DNA complex strongly support the bending aspect of the DNA distortion mechanism proposed for MerR and ZntR. Support for the localized DNA unwinding in the MerR-like DNA distortion mechanism was obtained from chemical nuclease probes. The footprints of ZntR, MerR, and SoxR show hypersensitivity to 5-phenyl-OP-Cu cleavage at the center of the palindromic binding site with the activating metal bound (24,55). For Zn-ZntR and the oxidized Fe 2 S 2 form of SoxR, this nuclease hypersensitivity is apparent only in the ternary open RNAP complexes, whereas Hg-MerR displays this feature both with and without RNAP. This hypersensitivity is proposed to be caused by a protein-induced underwinding of the DNA helix. Additional 5-phenyl-OP-Cu hypersensitive sites are also found around the transcription start site for each metal-activated protein when RNAP is added to the reaction. This cleavage by Cu-OP at positions Ϫ3 to Ϫ7 on the template strand and ϩ4 to ϩ5 on both strands correlates well with the presence of an open, transcriptionally active complex at other promoters (56,57).
Recent results indicate that RNAP open complex formation in general involves wrapping of the DNA strand around RNAP to form a left-handed superhelix (58 -61). The DNase I footprinting data for the RNAP/P zntA /Zn-ZntR complex are consistent with a similar model in which a series of upstream bends lead to spooling of DNA onto the polymerase. A model for the interaction of Zn-ZntR with the wrapped RNAP⅐DNA open complex is given in Fig. 8. A MerR-like DNA distortion mechanism conforms nicely with this model for RNAP/DNA interaction. In the repressor conformation, apo-MerR may bend DNA away from RNAP, preventing the wrapping of DNA around the RNAP protein core. Relaxation of these bends by Hg-MerR coupled with unwinding and changes in writhe may stimulate transcription by facilitating the wrapping characteristic of a productive open complex. Apo-and Zn-ZntR would follow this same mechanism, although the extent of bending and underwinding remains to be seen in this system.
Another unusual aspect of the MerR regulation mechanism is the ability of RNAP to form closed complexes at promoters bound to apo-MerR (Fig. 1B) (24). DNase I footprinting analyses indicated that with apo-MerR bound to the DNA, RNAP binds upstream of the MerR binding site, only accessing the Ϫ35 site (24,37). In this state, apo-MerR has also been shown to weakly repress transcription at merT (24). ZntR⅐P zntA did not readily form a closed RNAP complex in this manner at 37°C (data not shown). While the transcription assays indicate strong activation of zntA transcription with Zn-ZntR, apo-ZntR only weakly represses at concentrations exceeding 100 nM. The slight increase in transcription at lower apo-ZntR concentrations may be attributed to ZntR scavenging any exogenous zinc from the buffer. As more apo-ZntR is added, Zn-ZntR is outcompeted by the apo protein for the DNA binding site.
Without a strong zinc chelator like TPEN (log K ϭ 18.0) (62), which can sequester contaminating Zn(II), ZntR shows activation with no additional metal added (data not shown). The fact that TPEN is required to see the lowest levels of transcript in the absence of Zn(II) suggests that ZntR is very sensitive to free Zn(II) concentrations in the buffer. As Zn(II) is titrated into the assay buffer, the onset of transcriptional activation is observed just before Zn(II) levels exceed the TPEN concentration, indicating that ZntR can compete to a limited extent with TPEN for Zn(II). This provides further evidence that ZntR has a strong affinity for Zn(II). While MerR binds 1 mol of Hg per MerR dimer (22,63), ZntR can bind 1-2 mol of Zn(II) per monomer depending on the buffer conditions. However, the minimal metal occupancy required for function remains to be established.
In vivo transcriptional data for ZntR (2) reveals another similarity to MerR-the cooperative or ultrasensitive transcriptional response to metal ions (25). Brocklehurst et al. (2) have shown that in vivo expression of zntA exhibits a sigmoidal response to changes in Zn(II) concentrations in the media. This phenomenon has been observed for both in vivo (7,64) and in vitro (25) activation of P T by the MerR protein. The Zn(II) induction curve from our transcription results, however, should not be interpreted as ultrasensitivity because of the presence of TPEN in the assay. TPEN is a strong Zn(II) chelator and thus buffers the Zn(II) concentration. Once the buffer capacity is exceeded, Zn(II) binding to ZntR will be stoichiometric and thus will not reflect metal effects in a simple manner. The interesting ultrasensitivity feature of both the MerR and ZntR regulation systems is another important similarity of these stress response systems, however the molecular basis of this cooperative response phenomenon remains elusive.
The transcription and footprinting data presented here provide strong evidence that the stress-responsive transcriptional activation of P zntA by ZntR involves a MerR-like DNA distortion mechanism. ZntR acts as a genetic switch that becomes a strong activator upon Zn(II) binding. The metal selectivity and Zn(II) affinity of ZntR are being evaluated and may shed light on the molecular basis of metal ion recognition by this MerR family member.