The six zinc fingers of metal-responsive element binding transcription factor-1 form stable and quasi-ordered structures with relatively small differences in zinc affinities.

Six Cys(2)His(2) zinc fingers (F1-6) comprise the DNA binding domain of metal-responsive element binding transcription factor-1 (MTF-1). F1-6 is necessary for basal and zinc-induced expression of metallothionein genes. Analysis of NMR structural and dynamic data for an F1-6 protein construct demonstrates that each zinc finger adopts a stable betabetaalpha fold in the presence of stoichiometric Zn(II), provided that all cysteine ligands are in a reduced state. Parallel studies of protein constructs spanning the four N-terminal core DNA binding fingers (F1-4) and two C-terminal low DNA affinity fingers (F5-6) reveal similar stable zinc finger structures. In both the F1-6 and F5-6 proteins, the finger 5 cysteines were found to readily oxidize at neutral pH. Detailed spectral density and hydrodynamic analysis of (15)N relaxation data revealed quasi-ordered anisotropic rotational diffusion properties of the six F1-6 zinc fingers that could influence MTF-1 DNA binding function. A more general effect on the rotational diffusion properties of Cys(2)His(2) zinc fingers was also uncovered that is dependent upon the position of each finger within multifinger domains. Analysis of NMR (1)H-(15)N-heteronuclear single quantum coherence spectral peak intensities measured as a function of added Zn(II) in conjunction with Zn(II) binding modeling studies indicated that the Zn(II) affinities of all MTF-1 zinc fingers are within approximately 10-50-fold. These analyses further suggested that metal sensing by MTF-1 in eukaryotic cells involves multiple zinc fingers and occurs over a 100-fold or less range of accessible Zn(II) concentration.

Six Cys 2 His 2 zinc fingers (F1-6) comprise the DNA binding domain of metal-responsive element binding transcription factor-1 (MTF-1). F1-6 is necessary for basal and zinc-induced expression of metallothionein genes. Analysis of NMR structural and dynamic data for an F1-6 protein construct demonstrates that each zinc finger adopts a stable ␤␤␣ fold in the presence of stoichiometric Zn(II), provided that all cysteine ligands are in a reduced state. Parallel studies of protein constructs spanning the four N-terminal core DNA binding fingers (F1-4) and two C-terminal low DNA affinity fingers (F5-6) reveal similar stable zinc finger structures. In both the F1-6 and F5-6 proteins, the finger 5 cysteines were found to readily oxidize at neutral pH. Detailed spectral density and hydrodynamic analysis of 15 N relaxation data revealed quasi-ordered anisotropic rotational diffusion properties of the six F1-6 zinc fingers that could influence MTF-1 DNA binding function. A more general effect on the rotational diffusion properties of Cys 2 His 2 zinc fingers was also uncovered that is dependent upon the position of each finger within multifinger domains. Analysis of NMR 1 H-15 N-heteronuclear single quantum coherence spectral peak intensities measured as a function of added Zn(II) in conjunction with Zn(II) binding modeling studies indicated that the Zn(II) affinities of all MTF-1 zinc fingers are within ϳ10 -50-fold. These analyses further suggested that metal sensing by MTF-1 in eukaryotic cells involves multiple zinc fingers and occurs over a 100-fold or less range of accessible Zn(II) concentration.
Organisms ranging from bacteria to mammals maintain intracellular Zn(II) levels within a functional range through homeostatic mechanisms that include metal-dependent regulation of gene expression. It has recently been proposed that the high Zn(II) binding affinities reported for the bacterial Zur and ZntR proteins suggest that Zn(II) homeostasis in Escherichia coli is tightly regulated, such that essentially no labile or "ac-cessible" cytoplasmic Zn(II) is available (1). In this model, Zn(II) substitution reactions between Zur and ZntR and as yet unidentified metal chaperone transporters could mediate the metal response. In higher eukaryotic organisms, multiple homeostatic mechanisms regulate intracellular Zn(II) levels (2,3). However, the accessible concentration range over which Zn(II) is regulated in eukaryotic cells is unknown. In higher eukaryotic organisms, Zn(II)-inducible expression of genes encoding for the zinc transporter-1 and cysteine-rich metallothionein class of proteins is mediated through the metal-responsive element binding transcription factor-1 (MTF-1) 1 shown schematically in Fig. 1. Metallothioneins are the most well studied target of MTF-1 and are important for Zn(II) homeostasis, protection against oxidative stress, and heavy metal detoxification.
The MTF-1 transcription factor was first cloned from mouse (4) and has since been identified in humans (5,6), Drosophila melanogaster (7), Takifugu rubripes (8), zebrafish (9,10), and chicken. 2 Mouse MTF-1 (mMTF-1) is a 75-kDa, 675-amino acid protein that contains a six-Cys 2 His 2 zinc finger DNA binding domain (F1-6) (11) and three distinct transcriptional activation domains (Fig. 1). With the exception of the Drosophila protein, 92% sequence identity exists between all other known orthologs of MTF-1 in the zinc finger domain (8). The human and mouse MTF-1 zinc finger domains differ by a single amino acid. By contrast, considerable sequence divergence occurs in the remaining domains of MTF-1. The high sequence conservation of the MTF-1 DNA binding domain is consistent with a proposed Zn(II) regulatory role. Transcriptional specificity of MTF-1 is mediated through F1-6, which recognizes a metalresponsive DNA promoter element (MRE). The minimal MRE consensus sequence is TGCRCnC on the 5Ј-end of a larger 12-base pair site that contains a loosely conserved GC-rich region on the 3Ј-end (12)(13)(14)(15). Although variable in orientation, MREs are often found in multiple copies within the proximal promotor sites of target genes.
Multiple lines of evidence point to zinc finger metal coordination as an essential mechanistic contributor to Zn(II)-dependent activation of MTF-1. Electrophoretic mobility shift assays on nuclear extracts from various higher eukaryotic cell lines demonstrated that MTF-1 binding to the MRE is reversibly activated by the presence of 30 M Zn(II) at 30 -37°C (6, 16 -18). These and other experimental observations led to an estimation of in vitro Zn(II)-responsive MRE binding by MTF-1 as being in the low micromolar range (6, 16 -18). Because these in vitro systems contained numerous other Zn(II)-binding proteins in relatively high concentration, a quantitative estimation of the Zn(II) affinities of the MTF-1 zinc fingers from these studies was not possible. However, recent Co(II) affinity measurements of each MTF-1 zinc finger in the context of the fulllength F1-6 DNA binding domain reported all K d s to be in the 20 -100 nM range (19). MTF-1 can act as a heterologous Zn(II)responsive transcriptional activator in yeast, which has no chromosomal MTF-1 gene (20). Similarly, Zn(II)-responsive activation of gene expression mediated by indigenous MRE sites was recently demonstrated using a transfected D. melanogaster MTF-1 gene (7). Exposure of mammalian cells to Zn(II) also results in rapid nuclear translocation of MTF-1. Very recent in vivo studies in mammalian cells suggest that additional Zn(II) binding sites within a cysteine-rich C-terminal region of MTF-1 may also contribute to MTF-1 activity (21).
Structural and functional studies of MTF-1 zinc fingers have reported conflicting observations relating to which fingers are responsible for metal sensing and whether all fingers adopt stable canonical ␤␤␣ structures (20,(22)(23)(24)(25)(26). CD spectra at 222 nm were reported to be essentially identical for MTF-1 F1-6 proteins with 3.5 and 5.5 equivalents of Zn(II) (23). This result suggested to the authors that two or three of the fingers do not adopt the predicted folded structures. An NMR study of the three C-terminal fingers of MTF-1 (F4 -6) reported that fingers 4 and 6 adopt stable canonical folds in contrast to finger 5, which was reported to have only a fractional population of the ␤␤␣ folded form in the presence of stoichiometric quantities of Zn(II) (24). However, this putative finger 5 instability appears to be inconsistent with the submicromolar Co(II) binding K d measured for finger 5 (19). DNA binding studies have provided evidence that the four N-terminal MTF-1 fingers (F1-4) are the core MRE binding group, whereas the C-terminal fingers (F5-6) have been proposed to be regulatory Zn(II)-sensing fingers that stabilize DNA complex formation (22).
A different picture of Zn(II) sensing by MTF-1 has emerged from functional studies that suggest zinc finger 1 may be a regulatory finger, whereas fingers 5 and 6 may be necessary for transcriptional function in the context of the chromatin complex (25). In vivo studies using MRE-driven reporters in dko7 (MTF-1 Ϫ/Ϫ ), Drosophila SL2, and yeast cell lines demonstrated that finger 1, but not fingers 5 and 6, of MTF-1 was needed for Zn(II)-inducible activation of DNA binding and subsequent gene expression (20). A role for the N-terminal fingers of MTF-1 in Zn(II)-responsive transcriptional activity has also been reported by Koizumi et al. (26). In vitro DNA binding of MTF-1 was also shown to be activated in a in vitro transcription translation-coupled reticulocyte lysate system (TnT) lysate in the presence of 5-15 M Zn(II) at 37°C from undetectable levels without added Zn(II) (20). Deletion mutants of MTF-1 missing finger 1 exhibited constitutive, albeit poor, DNA binding in the absence of exogenous Zn(II) and were not further activated upon the addition of more Zn(II). By contrast, no significant difference in Zn(II) induction or relative MRE binding activities was observed in identical assays of mutant proteins in which finger 5 or fingers 5 and 6 were deleted from the MTF-1 zinc finger domain.
To reconcile conflicting structural and metal binding observations and to gain insights into the higher eukaryotic cellular Zn(II)-responsive range, we have characterized the structure, dynamics, and Zn(II) binding properties of all six MTF-1 zinc fingers (F1-6). Chemical shift, 3 J HN-H␣ scalar coupling, and 15 N-{ 1 H}-NOE dynamics NMR measurements clearly demonstrate that each Cys 2 His 2 domain of MTF-1 adopts a stable ␤␤␣ structure at submillimolar protein concentrations in the presence of 5 mM ␤-mercaptoethanol (BME) and stoichiometric quantities of Zn(II). Parallel studies show that the four Nterminal (F1-4) and two C-terminal (F5-6) zinc fingers of MTF-1 also form stable structures that are similar to those of the same fingers in the context of the full-length F1-6 DNA binding domain. An unusual tendency of zinc finger 5 toward cysteine oxidation is also reported. 15 N relaxation measurements and subsequent spectral density and hydrodynamic analyses reveal that the rotational diffusion properties of F1-6 are relatively well correlated (quasi-ordered) compared with several other multizinc finger proteins, which suggests interfinger interactions could play a functional role in MTF-1 DNA binding. Position-specific effects on the rotational diffusion motions of zinc fingers that are likely general phenomena of the Cys 2 His 2 class of zinc fingers are also reported. Titration experiments monitored by 1 H-15 N-HSQC spectra peak intensities combined with theoretical binding studies also suggest that the Zn(II) affinities of all six MTF-1 fingers are within ϳ10 -50fold. These results are also consistent with a eukaryotic MTF-1 metalloregulatory mechanism involving multiple zinc fingers that occurs over a 100-fold or less concentration range of accessible Zn(II).

MATERIALS AND METHODS
Plasmid Construction and Expression-cDNA encoding residues 136 -312 that encompass F1-6 was amplified by polymerase chain reaction (PCR) from a murine expressed sequence tag clone 1528077 (American Type Culture Collection). The DNA primers used for PCR amplification also incorporated the XbaI-NdeI region of pET21a(ϩ) (Novagen) and an EcoRI restriction site onto the 5Ј-and 3Ј-ends of the gene, respectively. Gel-purified PCR fragments were digested and sub-FIG. 1. Schematic representation of the mMTF-1 gene. A, proceeding from N to C terminus, mMTF-1 contains six Cys 2 His 2 zinc fingers that comprise the DNA binding domain, followed by acidic, proline-rich, and serine/threonine-rich regulatory domains. B, amino acid sequence of the mMTF-1 zinc finger region (residues 139 -312) aligned by individual fingers (F) with predicted canonical secondary structures and linker regions indicated above the sequence. Commonly occurring DNA base-contacting residues are also indicated above the sequence as numerical positions relative to the start of the predicted ␣-helices.
sequently ligated into the XbaI and EcoRI sites of pET-21a(ϩ) to form the expression vector pMZF1-6. DNA constructs coding for mMTF-1 F1-4 (pMZF1-4, residues 136 -252) and F5-6 (pMZF5-6, residues 257-312) were constructed using the same cloning strategy. Data provided by the University of Kansas Medical Center Biotechnology Support Facility or the University of Missouri-Kansas City School of Biological Sciences Genomics Facility were used to verify all plasmid DNA sequences in both strand directions.
Expression and Purification of F1-6, F1-4, and F5-6 -DNA plasmids coding MTF-1 zinc finger proteins F1-6 or F1-4 were transformed into E. coli strain BL21(DE3), from which a single colony was used to inoculate 20 ml of Luria-Bertani broth supplemented with 200 g/ml ampicillin. After overnight growth, gently centrifuged cells were resuspended into 1 liter of minimal medium (1.28% Na 2 HPO 4 , 0.3% KH 2 PO 4 , 0.05% NaCl, 1 mM MgCl 2 and 1 mM CaCl 2 ) containing 50 g/ml ampicillin, trace metals, and 10 mg/liter thiamine. Uniform (U) isotopic enrichment of the medium was obtained by using 1 g/liter U-15 NH 4 Cl and either 2 g/liter unlabeled glucose or 2 g/liter U-[ 13 C] glucose when producing U-[ 15 N]-or U-[ 15 N, 13 C]-labeled protein, respectively. Cultures were grown at 37°C until the absorbance at 600 nm reached 0.6, induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside, supplemented with an additional 50 M ZnSO 4 , and grown overnight (ϳ16 h). Cells were harvested by centrifugation and resuspended in lysis buffer (25 mM MOPS, pH 7, 50 M ZnSO 4 , 300 mM NaCl, 50 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 0.5% Triton X-100). Resuspended cells were then lysed by sonication whereby the soluble and insoluble fractions were separated by centrifugation at 18,000 rpm for 30 min (Sorvall SS-34 rotor). The soluble fraction was discarded, and the insoluble fraction containing the zinc finger inclusion body proteins was solubilized with 7 M guanidine hydrochloride, 25 mM MOPS, pH 7, and 200 mM dithiothreitol. The solubilized protein solution was centrifuged at 18,000 rpm for 60 min (Sorvall SS-34 rotor) and diluted 100-fold directly into 0.5% trifluoroacetic acid and 10% CH 3 CN. This protein solution was centrifuged at 10,000 rpm (Sorvall SS-34 rotor) for 30 min at 4°C to remove precipitated cellular materials. This solution was then passed through a 0.2-m filter and loaded onto a preparative 40 ϫ 100-mm Waters Delta-Pak 300-Å pore size C18 high performance liquid chromatography (HPLC) column pre-equilibrated with 0.1% trifluoroacetic acid and 20% CH 3 CN. A linear 20 -50% CH 3 CN gradient was applied to the column at 1.5%/min, whereby the MTF-1 zinc finger proteins eluted at ϳ30 -37% CH 3 CN. Pure fractions of the eluted protein were screened by analytical high performance liquid chromatography, pooled, quantitated by ultraviolet absorbance at 280 nm (Varian Cary100), and lyophilized. Expression and purification of F5-6 protein was identical to F1-6 and F1-4 except for the following protocol changes: cells were grown at 20°C and harvested 3-5 h postinduction. F5-6 protein was localized in the soluble portion of the cell lysate, which was isolated and adjusted to contain 0.5% trifluoroacetic acid and 5% CH 3 CN for high performance liquid chromatography purification as described. A linear 5-45% CH 3 CN gradient was applied to the column at 1.5%/min. Protein extinction coefficients and concentration determinations for zinc titration experiments were determined by amino acid analysis at the W. M. Keck Facility at Yale University. A purity of Ͼ95% was estimated for each protein sample using sodium dodecyl sulfate-polyacrylamide gel electrophoresis, mass spectrometry, and analytical high performance liquid chromatography. Mass spectrometry confirmed essentially complete 15 N and 13 C isotopic enrichment of the MTF-1 zinc finger proteins (University of Missouri-Kansas City School of Biological Sciences Proteomics Facility). Undetectable residual Zn(II) levels from 100 M MTF-1 protein samples as determined by atomic absorption spectroscopy (Varian AA400, ϳ1 ppm Zn(II) threshold) corresponded to Ͻ0.1 molar equivalent Zn(II)/mole zinc finger protein (F1-6, F1-4, and F5-6) in all apo preparations.
NMR Spectroscopy-NMR samples were prepared from lyophilized protein that was dissolved in argon-saturated d 13 -MES, pH 6.9, with 5% D 2 O (or 99% D 2 O for 13 C-NOESY-HSQC experiments), 5 mM BME, 0.2 mM 3-(trimethylsiyl)-1-propane-sulfonic acid, and 1 mM NaN 3 . MTF-1 zinc finger NMR samples were treated with 1.05 molar equivalents of Zn(II)/zinc finger except for those apo protein samples used in titrations. All MTF-1 zinc finger protein concentrations were 0.7-1.0 mM except for F5-6, which was soluble to ϳ0.3 mM. NMR sample tubes were equilibrated with argon prior to sealing with an airtight cap. All NMR data collected for these studies originated from a 14.1 T Varian Inova spectrometer (599.7 MHz for 1 H, 150.8 MHz for 13 C, and 60.8 MHz for 15 N) equipped with either a conventional or cryogenically cooled ( 1 H, 13 C, 15 N) triple resonance probe.
NMR spectra described below were recorded at 30°C for F1-6 and 20°C for F1-4 and F5-6. Standard double and triple resonance spectra were used to assign the backbone 1 H N , 15 N H , 13 C, and side chain 13 C␤ resonances for F1-6, F1-4, and F5-6 (27). Specifically, 1 H-15 N-HSQC, HNCACB, CBCA(CO)NH, HNCA, HNCO, and (HCA)CO(CA)NH spectra were recorded using water flip-back methods (28,29) and/or 15 N gradient coherence selection with sensitivity enhancement (30) for solvent suppression. Aliphatic side chain assignments of F1-6 were obtained from three-dimensional CT-HCCH-COSY, HCCH-TOCSY, using mixing times of 100 ms. Additional aliphatic and aromatic F1-6 side chain information was obtained from three-dimensional 15 N-and 13 C-NOESY-HSQC experiments (150-ms mixing times), the latter of which was recorded twice with the carbon frequency centered on the aliphatic and aromatic regions. Scalar 3 J HN-H␣ coupling constants for F1-6 were calculated from a three-dimensional HNHA spectrum using a homonuclear H N -H␣ 3 J-coupling dephasing delay () of 12.5 ms as described by Vuister and Bax (31).
For all of the NMR spectra, the proton carrier was set to the water frequency to reduce the amplitude of spurious water echoes. In general, 512 complex points were acquired in the direct 1 H dimension, while 128, 45, and 45 complex increments were recorded where appropriate for the 1 H, 13 C, and 15 N indirect dimensions, respectively. All NMR data processing and analysis was carried out with the NMRPipe and NMRView software packages, respectively (32,33). Forward linear prediction was used in all spectra recorded with constant time evolution to double the time domain. All indirect dimensions of these NMR spectra were zero filled to 256 complex points and apodized with a 90°-shifted sinebellsquared window function prior to Fourier transformation.
The standard approach used to assign all MTF-1 zinc finger proteins involved correlating intra-(i) and inter-residue (i-1) 13 C␣, 13 C␤, and 13 CO resonances to the corresponding backbone 1 H N -15 N H resonances through scalar couplings for each non-proline amino acid sequentially in the protein (34). The independent assignment of 13 C␣/␤ and 13 CO resonances was useful for clarifying ambiguous residue connections that are common in identical residue positions of repeating zinc finger motifs. Backbone 13 C␣ and 13 CO random coil values have been reported previously (35,36). Referencing for 1 H was internal relative to 3-(trimethylsiyl)-1-propane-sulfonic acid, whereas 15 N and 13 C referencing was external according to standard procedures (37). All MTF-1 zinc finger backbone and side chain NMR resonance assignments have been deposited in the Biomagnetic Resonance Data Bank. F1-6, F1-4, and F5-6 15 N-{ 1 H}-NOE measurements, 15 N longitudinal (R 1 ) and transverse (R 2 ) relaxation time constants were measured by the steady-state, inversion-recovery and Carr-Purcell-Meiboom-Gill methods, respectively (38 -40), using standard water flip-back methods (28). All 15 N relaxation spectra were recorded as 128 ϫ 512 complex matrices, with 8 (R 1 and R 2 ) or 16 ( 15 N-{ 1 H}-NOE) scans/t 1 point and spectral widths of 8000 and 2200 Hz in the 1 H and 15 N dimensions, respectively. Indirect dimensions were zero filled to 256 complex points and apodized with a 90°-shifted sinebell-squared window function prior to Fourier transformation. The R 1 and R 2 spectra were collected with relaxation delays as follows: These spectra typically included one or more multiple R 1 and R 2 data sets that were used to estimate the precision of the peak intensities. Similarly, at least three interleaved steady-state 15 N-{ 1 H}-NOE experiments were collected with and without 1 H saturation to allow for estimation of experimental uncertainty. A recovery delay of 5 s was used for all 15 N relaxation experiments. The R 1 and R 2 time constants were calculated as two parameter exponential decay curves described by I(t) ϭ I 0 e ϪtR , where I(t) is the measured peak intensity as a function of the known relaxation delay time t, I 0 is the fitted initial peak intensity at t ϭ 0, and R is the calculated 15 N relaxation time constant (R 1 or R 2 ). All exponential data curves were fitted with the program CurveFit1. 30 using the program mapsdf generously provided by P. E. Wright (Scripps Research Institute, La Jolla, CA). Criteria used to identify rigid nitrogens with no significant contributions from subnanosecond-to-nanosecond internal (t e ) or microsecond-to-millisecond segmental motions due to chemical exchange (R ex ) were those of Barbato et al. (44). Rotational Diffusion Calculations-Both isotropic and anisotropic models were used to characterize the overall rotation diffusion properties of the F1-6, F1-4, and F5-6 MTF-1 zinc finger constructs using the TENSOR2 program (45). Because three-dimensional structural models were needed for each MTF-1 zinc finger as input for TENSOR2, a homology modeling approach using the MODELLER program version 6v2 was used to generate these structures (46). Suitable zinc finger structural templates for MODELLER were chosen from a recently compiled list of zinc finger proteins (47) for which structural models were available in the RCSB Protein Data Bank (www.rcsb.org/pdb/). Single or multiple templates used for final modeling calculations had at least 35% sequence identity to the respective MTF-1 zinc finger target and a z-score of at least 3.5 and produced homology models with a backbone root mean square deviation between template(s) and target of Յ1 Å. (Table I).
In addition to the selection criteria for rigid amides already described, residues with calculated differences in backbone and/or dihedral angles of Ͼ35°between the zinc finger models and the corresponding dihedral angles measured for F1-6 using the TALOS method (48) were also excluded. Confidence in the isotropic and anisotropic diffusion models was assessed in TENSOR2 using Monte Carlo sampling methods to compare experimental 2 values ( 2 exp ) with those Monte Carlo simulated values ( 2 MC ) determined at the 90% confidence limit from the optimal fitted diffusion parameters (45). Because of the inherent deviations in individual residue chemical shift anisotropy, R 2 and R 1 uncertainties were set at 5% of the time constant value for residues with lower fitted uncertainties (generally 1-3%) (49).
Zinc Titrations-Zn(II)-coupled binding/folding of each apo F1-6 finger was monitored by 1 H-15 N-HSQC fractional peak intensities, which were calculated from the ratio of each peak intensity corresponding to a given residue 1 H-15 N correlation to the maximum peak intensity value measured from all of the 1 H-15 N-HSQC experiments for that residue.
Theoretical Zn(II) Occupancy-Calculations of Zn(II) occupancies for each MTF-1 zinc finger within F1-6 were carried out using the Berkeley Madonna software package (version 8.0.1, ©1997-2000, Robert I. Macey and George F. Oster, Kagi Shareware, Berkeley CA). Solutions to differential equations using the Runge-Kutta 4 integration method provided steady-state metal-bound concentrations for each finger. Six non-cooperative simultaneous coupled metal binding-folding equilibria described in Equation 2 were assumed (50) where F i and Zn(II)F i are the apo and metal bound forms for each zinc finger of MTF-1, respectively. Equilibrium Zn(II) occupancies were calculated for each finger using the full range of K a values reported previously for zinc finger binding to Co(II) or Zn(II) (10 11 -10 6 M Ϫ1 ) (19,51). An estimation of the minimum high/low Zn(II) occupancy ratio from all six F1-6 fingers for a given range of affinities was sought. Therefore, within each K a range modeled, one finger had the highest Zn(II) affinity and the remaining five fingers had the lowest affinity because this scenario always produced this minimum high/low Zn(II) occupancy ratio. Experimental concentrations of 180 M for each F1-6 zinc finger (F i ) and 270 M Zn(II) were used for modeling calculations. These concentrations of 1.5 molar equivalents Zn(II)/mole F1-6 protein were chosen because they represent the minimum Zn(II)/protein ratio in which resonances from all six MTF-1 fingers in the 180-M experimental protein sample could be observed.

MTF-1 F5 Cysteines Are Hypersensitive to Thiol
Oxidation-MTF-1 zinc finger constructs encompassing F1-6, F1-4, and F5-6 were isolated as lyophilized apo proteins in the free thiol (reduced) state. Despite exhaustive efforts to maintain anaerobic samples, analysis of preliminary NMR triple resonance spectra recorded for the F1-6 and F5-6 proteins in the pres-ence of stoichiometric Zn(II) produced evidence of progressive cysteine thiol oxidation. Although well dispersed peaks corresponding to residues within folded zinc finger domains were evident in these early spectra, a second set of poorly dispersed resonances with increasing intensity over the course of data collection was also observed. Peaks with the greatest loss of intensity by far were determined to correspond to finger 5 in the folded state. Nearly complete assignments of the poorly dispersed resonances from the F5-6 spectra revealed a sequential run of finger 5 residues with essentially random coil chemical shifts and downfield-shifted (40 -42 ppm) cystine 13 C␤ values (data not shown). Because Zn(II) coordination to the two cysteine ligands is a necessary stabilizing force for canonical ␤␤␣ structure formation within each finger (52), these observed random coil shifts are consistent with a disulfide-bonded zinc finger 5 polypeptide. To prevent thiol oxidation, 5 mM BME was added to zinc finger protein samples used to record all subsequent NMR spectra in these studies. BME was chosen for its relatively low Zn(II) binding affinity (53). 1 H-15 N-HSQC spectra recorded over a 2-3-week period for F1-6, F1-4, and F5-6 protein samples treated with BME had stable intensities and monodispersed resonances including those corresponding to the finger 5 polypeptide. Addition of 5 mM BME to a partially oxidized F1-6 sample was sufficient to regenerate a stable NMR sample with the same spectral properties. An 1 H-15 N-HSQC spectrum for F1-6 annotated with residue assignments for most resonances is shown in Fig. 2. Corresponding 1 H- 15 2. 1 H-15 N-HSQC spectrum of MTF-1 F1-6. A, full spectrum with selected residue annotations. B, expanded view of boxed region in panel A with additional residue annotations. Spectrum was recorded using an 800-M protein sample with 6.3 molar equivalents Zn(II) and 5 mM BME at pH 6.9 and 30°C. a U-15 N, 13 C-labeled F1-6 protein sample in the presence of 5 mM BME and 1.05 molar equivalents of added Zn(II)/finger at 30°C and pH 6.9 as reported recently (54). With the exception of Cys-290 13 CO, all backbone 13 C and side chain 13 C␤ NMR resonance assignments were also obtained. Approximately 90% of all side chain 1 H and 13 C resonances were also recently determined for F1-6 (54). Here, backbone 1 H-15 N resonance assignments were obtained for F1-4 (107 of 114 residues) and F5-6 (52 of 55 residues) in addition to Ͼ95% of all backbone 13 C and side chain 13 C␤ resonances for these proteins at 20°C and pH 6.9.
Comparison of the chemical shifts obtained for the MTF-1 F1-6 protein to those of random coil shown in Fig. 3 is a good indicator of secondary structure (35,55). The most prominent repeating feature within each finger is the C-terminal ␣-helix that is strongly predicted by the presence of positive ⌬␦ 13 C␣ and 13 CO values in these regions. The only deviation from predicted structure of any F1-6 zinc finger indicated from the data in Fig. 3 is the sequence position of the first finger 5 ␣-helical residue at His-274. For a canonical zinc finger, the first ␣-helical residue occurs 2 amino acids C-terminal to the second ␤-strand (Ser-273 for finger 5). Another prominent feature of the chemical shift data descriptive of canonical zinc finger structure is the significantly upfield-shifted 13 C␣ and downfield-shifted 13 C␤ resonances for the residues in the Ϫ1 position relative to the start of each predicted ␣-helix attrib-uted to an N-cap (56). By first approximation, negative ⌬␦ 13 C␣ and 13 CO values are expected for the 6 residues within each finger predicted to comprise the small anti-parallel ␤-sheet. Although the 13 C␣ and 13 CO trends in these regions are only partially obeyed, chemical shift comparisons to random coil throughout each finger have a very similar pattern to those reported for the three N-terminal fingers of TFIIIA (TF1-3) and the four zinc fingers from the human Wilms tumor suppressor protein (WT1-4) (57,58). Moreover, the backbone 13 C␣ shift differences from random coil for each F1-6 cysteine at residue position 3 of ␤-strand 1 are positive (except finger 6), likely because of the effects of thiolate coordination to Zn(II) (58). The topology of F1-6 was further analyzed using the TALOS method (48), which produced dihedral angle restraints consistent with ␤␤␣ secondary structures for all fingers (Fig. 3). Backbone 3 J HN-H␣ scalar coupling constants measured for residues in canonical ␣-helical and ␤-sheet regions were generally Ͻ5 and Ͼ7 Hz, respectively, which are qualitatively consistent with the respective secondary structure predictions (Fig. 3).
F1-4 and F5-6 Adopt Stable Structures Similar to the Corresponding Fingers of F1-6 -Potential context-dependent effects on the structural stability of the MTF-1 zinc fingers were investigated (24). Comparisons of weighted HN ([(␦H) 2 ϩ(␦N[sw N /sw H ]) 2 ] 1/2 ) and backbone 13 C chemical shift differences between residues from constructs containing the proposed core DNA binding fingers (F1-4) and lower DNA affinity fingers (F5-6) (22) are presented in Fig. 4. Most of these chemical shifts in F1-4 and F5-6 are the same as those corresponding shifts from F1-6 within the limits of the respective spectral resolutions (0.02 ppm for 1 H N , 0.15 ppm for 15 N H , 0.3 ppm for 13 C␣, and 0.1 ppm for 13 CO). The structural differences in the C-terminal residues of finger 4 from F1-4 and N-terminal residues of finger 5 from F5-6 suggested from HN, 13 C␣, and 13 CO chemical shift comparisons reflect expected differences in chemical environments between the different protein constructs in these regions. However, the N-terminal region of subtle context-dependent structural change in F5-6 suggested from the chemical shift data in Fig.  4 does appear to extend through the ␤-sheet of finger 5. It is  15 N, 13 C, and side chain 13 C␤ chemical shift differences from random coil values for each F1-6 residue (139 -312) (35). Predicted secondary structure schematic drawings for each finger are indicated above the plot. Lower panel, scalar 3 J HN-H␣ coupling constants for F1-6 residues obtained from three-dimensional HNHA (31) with secondary structure features predicted from backbone , angles using the TALOS method (48) indicated schematically above the plot. Bars and secondary structure schematics corresponding to residues in canonical ␤-sheet and ␣-helical structures in upper and lower panels are indicated with gray and black bars, respectively. 13 CO, and 13 C␣ chemical shift differences between residues from F1-6 with those of F1-4 (A) and F5-6 (B). Predicted secondary structure schematic drawings for each finger are indicated above the plot. also worth noting that the linker region between fingers 5 and 6 and ␤-strand 2 of finger 6 in the F5-6 protein may have some subtle structural or dynamic difference manifested by changes in HN and 13 C␣ or 13 CO shifts (Fig. 4).

MTF-1 Zinc Fingers Have Similar Fast Internal Dynamics with Quasi-correlated Position-specific Anisotropic Rotational
Diffusion Properties-A summary of 15 N relaxation data recorded at 60 MHz 15 N frequency for F1-6, F1-4, and F5-6 is provided in Fig. 5. Only those residues with well resolved 1 H-15 N resonances corresponding to a single amino acid were used in the subsequent analyses. The two-dimensional 15 N relaxation spectra of the repeating zinc finger domains contained a rather large number of overlapped resonances because numerous residue positions in different fingers contain identical or chemically similar residues that are likewise in similar chemical environments within each domain (see Fig. 1). As the largest multi-Cys 2 His 2 zinc finger protein used for NMR studies to date with six repeating domains, it is not surprising that resonance overlap was most severe for the F1-6 protein. In addition, a small number of resonances with insufficient intensity for accurate measurement in one or more of the three 15 N-{ 1 H}-NOE, R 1 , or R 2 data sets were also excluded. Residues for which 15 N relaxation parameters were determined based on the above criteria are indicated in Fig. 5. 15 N-{ 1 H}-NOE measurements provide information about protein fast internal dynamics (backbone flexibility) that occur on the nanosecond-picosecond timescale such as librational or inter-residue motions. In general, higher frequency lower amplitude internal motions (larger 15  Corresponding values for linker residues of F1-6 (0.45 Ϯ 0.12), F1-4 (0.45 Ϯ 0.08), and F5-6 (0.60 Ϯ 0.01) suggest that residues within the typically disordered linkers connecting adjacent fingers are more flexible than those residues from the finger regions, although the single F5-6 linker appears to be decidedly less so. R 2 , R 1 time constants and R 2 /R 1 ratios calculated for residues from F1-6, F1-4, and F5-6 are presented in Fig. 5. The R 2 /R 1 ratio is directly affected by rotational diffusion motions that occur on the one to tens of nanoseconds timescale. Both the F1-6 and F1-4 zinc finger proteins have uncharacteristically large and heterogeneous R 2 /R 1 ratios when compared with other zinc finger proteins studied, including ADR1 (transcription factor containing two zinc fingers that regulates alcohol dehydrogenase and other peroxisomal enzymes) (61) and WT1-4 3 (Fig. 5). By contrast, R 2 /R 1 ratios measured for F5-6 are much smaller and more homogeneous. The generally larger R 2 /R 1 ratios observed in helical residues within each internal MTF-1 finger of F1-6 and F1-4 ( Fig. 5) are consistent with a predicted greater alignment of these N-H vectors with the long diffusion axis of anisotropic multifinger domains (62) (fingerfinger alignment and rotational diffusion is analyzed in detail later in this section). A systematic trend of increasing average R 2 /R 1 ratios starting from the N-and C-terminal fingers of F1-6 and F1-4 that progresses to the internal fingers is also evident in Fig. 5. Localized effects of R ex chemical exchange processes on the microsecond-millisecond timescale also affects the T 2 or transverse relaxation times (1/R 2 ) for some of the residues in the MTF-1 zinc fingers. Most notably, Thr-142 and Gly-145 in zinc finger 1 from both F1-4 and F1-6 have R 2 values (22-24 s Ϫ1 ) essentially twice the per finger averages calculated for the respective proteins. Interestingly, these two residues within the putative metalloregulatory finger 1 are adjacent to the two metal binding cysteines (20). A possible correlation between the microsecond-millisecond dynamics of Thr-142 and Gly-145 and the putative metal-sensing mechanism of finger 1 is currently under investigation. 15 N relaxation data recorded for F1-6 at an ϳ50% lower protein concentration (ϳ400 M) produced very similar R 2 and R 1 time constants (data not shown), diminishing the possibility that aggregation could account for the heterogeneous and generally larger R 2 /R 1 ratios observed for the MTF-1 F1-4 and F1-6 proteins.
Reduced spectral density maps of J(0), J(N), and J(0.87H) presented in Fig. 6 were calculated to qualitatively characterize the complex dynamic properties of the tethered multifinger MTF-1 proteins. The progressively larger average J(0) and lower average J(N) and J(0.87H) spectral densities for the internal fingers in F1-6 and F1-4 suggest correspondingly more restricted rotational diffusion for these interior domains, which are tethered to one or more adjacent zinc fingers on both ends. Overall, the Fig. 6 spectral density maps and Fig. 7 F1-4 (B), and F5-6 (C). Predicted secondary structure schematic drawings for each finger are indicated above plot. Error bars depict the quality of fit for experimental measurements as described under "Materials and Methods." of J(N) and J(0.87H) versus J(0) demonstrate that the timescales of rotational diffusion motions within F1-6 and F1-4 are heterogeneous, but more similar between adjacent domains. It is these heterogeneous rotational diffusion properties of the F1-6 and F1-4 fingers that are nonetheless more correlated between adjacent domains that we refer to as quasiordered. The distinct grouping of residues from internal and external fingers within F1-6 and F1-4 in Fig. 7 underscores this quasi-ordering, although effects of anisotropy, R ex , and possible fast internal motions (t e ) muddle the interpretation of these results somewhat. Because only J(0) spectral densities are dependent on T 2 , they alone are affected by R ex and are also very sensitive to anisotropy. J(N) spectral densities are dependent on T 1 (1/R 1 ) and therefore are also affected by anisotropic rotational diffusion, albeit to a lesser extent. By contrast the timescales of the two F5-6 zinc finger rotational diffusion motions appear to be very similar and faster than those of even the most rapidly tumbling domains from the larger MTF-1 zinc finger proteins.  (44) (no significant R ex or e ) from each zinc finger of the different MTF-1 constructs. A total of 12 F1-6, 11 F1-4, and 2 F5-6 non-rigid amides were excluded from these m calculations. Given the heterogeneous R 2 /R 1 ratios measured for the remaining residue amides that are consistent with significant anisotropic rotational diffusion motions (Fig. 5), m only represents an approximate weighted average of the inverse diffusion constants from each of the three principal axes of diffusion ( (62). Despite the isotropic approximation inherent to m , the rotational diffusion properties of zinc fingers from MTF-1 F1-6 and F1-4 appear to be more restricted and thus more correlated than those of the WT1-4 protein, which has m values of 1) 5.2 Ϯ 0.8, 2) 6.6 Ϯ 0.8, 3) 6.3 Ϯ 1.0, and 4) 5.4 Ϯ 0.9 ns (60). It should be noted that WT1-4 also appears to have similar but more subdued position-specific rotational diffusion properties (internal fingers are more restricted).
A summary of these calculations is given in Table II. With the exception of finger 6 from F5-6, all of the MTF-1 zinc fingers from F1-6, F1-4, and F5-6 were fit successfully to the anisotropic diffusion model at the 90% confidence interval. By contrast, only data from fingers 1 and 4 from F1-4 and finger 6 from F1-6 and F5-6 produced accepted isotropic models, albeit with mostly lower quality fits than those obtained with an anisotropic model (compare 2 exp with 2 MC for both models in Table II). A plot of experimental R 2 /R 1 ratios bracketed with 5% uncertainties (49) is superimposed with back-calculated R 2 /R 1 ratios in Fig. 8. The rotational anisotropy in F1-6, which is more pronounced in the internal fingers, clearly accounts for much of the experimental R 2 and R 1 heterogeneity. One possible domain orientation relative to the axes of diffusion for each F1-6 finger is also presented in Fig. 8. It should be noted that a 180 rotation about any of the diffusion axes shown is equally consistent with the relaxation data (62). However, given the motional restrictions of individual fingers tethered by small 4 -5-amino acid linkers, it is clear from Fig. 8 that for F1-6 the principal axes of diffusion (D zz ) are reasonably well aligned with the long axis of the helices in each of these internal fingers. Overall, the motion along the long axes of the internal F1-6 and F1-4 zinc fingers is quasi-ordered on the 5-8-ns timescale, whereas motions along the minor axes of diffusion  (D xx and D yy ) are significantly more restricted. Two experimental observations are also consistent with significant F5-6 finger-finger correlated rotational dynamics. First, the rotational diffusion rates determined for both F5-6 fingers are over two times greater than the 1.8 ns reported earlier for a single zinc finger (63). Second, the 15 N-{ 1 H}-NOE values measured for residues in the F5-6 linker are consistent with a relatively ordered peptide (Fig. 5).
Efforts to fit the MTF-1 F1-6 and F5-6 15 N relaxation data using the model free approach (64,65) with isotropic and anisotropic models in TENSOR2 were unsuccessful. Similar problems have been encountered with other proteins, such as an ϳ200-amino acid fragment of the prion protein that does not exhibit a single global correlation time (66). As with the prion protein fragment, estimations of the order parameter S 2 calculated for each residue directly from J(0) and J(N) often exceeded unity, and as such were not meaningful. A model free analysis was successful for the F5-6 protein in that most residues (except Lys-267) were fit to model 1 or 2 (67) with no R ex and t e less than 100 -200 ps. However, the interpretation of these data alone did not seem to add any new insights into the structural and functional properties of the MTF-1 metalloregulatory DNA binding domain.

MTF-1 Zinc Fingers Have Similar Zn(II) Binding
Affinities-The relative Zn(II) binding affinity of each MTF-1 finger within F1-6 was investigated using 1 H-15 N-HSQC-based Zn(II) titrations starting from the apo form of the protein.
Representative spectra and results of the titration are presented in Fig. 9. Analyzed 1 H-15 N-HSQC peaks corresponded to residues in the folded protein that were in slow exchange on the NMR timescale (the vast majority of all resonances). A fractional peak intensity corresponding to a given residue and Zn(II) concentration, which is defined as measured intensity divided by maximum peak intensity corresponding to the same residue measured from spectra at all Zn(II) concentrations, was averaged over all observable residues from each finger (Fig. 9, C and D). Moreover, because zinc finger folded structure is fully coupled to Zn(II) binding (52), this average fractional intensity was used to estimate the relative extent of folding for each finger as a function of added Zn(II). Data from a minimum of 9 residues/finger show that each finger has a measurable Zn(II) occupancy by 0.75 equivalents except for finger 6, which is observable at 1.5 equivalents (Fig. 9B). Based on relative Zn(II) occupancies of each finger at substoichiometric Zn(II) levels (Fig. 9D), the relative affinities for the MTF-1 fingers appear to be in qualitative agreement with those reported recently for Co(II): (finger, F, highest to lowest) F4 Ͼ F2 Х F5 Ͼ F1 Х F3 Х F6 (19). However, given the overlapping uncertainties of some Zn(II) titration points between 0.75-2.2 equivalents, the ranking of closely spaced fingers is approximate. Finally, a surprising heterogeneity is evident in the finger 2 binding data in which residues from the loop connecting adjacent ␤-strands and the C-terminal region of the ␣-helix (light blue data points) appear to have less ensemble structural order at substoichiometric quantities of Zn(II). The cause of this structural heterogeneity is unknown and is currently under investigation. DISCUSSION The results of the NMR studies presented here clearly show that all six MTF-1 zinc fingers are capable of forming stable canonical ␤␤␣ secondary structures at submillimolar protein concentrations in the presence of stoichiometric quantities of Zn(II) and millimolar BME (Fig. 3). Chemical shift comparisons of smaller N-and C-terminal zinc finger proteins shown in Fig. 4 also suggest that context-dependent effects on the structure and stability of each MTF-1 finger are not likely to be significant. However, a hypersensitivity of zinc finger 5 to cysteine thiol oxidation indicates that the metal binding properties of this finger may be somewhat unusual. A 50-fold greater thiolate reactivity for a subset of approximately three MTF-1 fingers has been reported previously (23). One possible explanation for hypersensitivity to thiol oxidation is labile Zn(II) binding, which is consistent with previous studies in which unusually high susceptibility to thiol alkylation by finger 5 was also observed (68). However, the observations reported here of a relatively stable ␤␤␣ fold for finger 5 appear on the surface to contradict a Zn(II) lability theory. Although the cause of the increased susceptibility of cysteine oxidation in finger 5 is still under investigation, elevation of both "on" and "off" rates for Zn(II) binding to the metal site could be consistent with all experimental observations to date (i.e. stability and lability). Indeed, a similar explanation for Zn(II) binding lability measured from rapid loss of metal binding upon dialysis has been offered for Zn(II) binding lability in finger 2 of a two-zinc finger zinc-responsive transcriptional activation domain of the Zap1 protein from Saccharomyces cerevisiae (69). A potential contribution to the metalloregulatory function of MTF-1 (via protein inactivation) from a more labile zinc finger could also be envisaged whereby a drop in cellular Zn(II) levels would produce a rapid and potentially irreversible (thiol oxidation) loss of ␤␤␣ structure. It is tempting to speculate that the reported instability of finger 5 in the absence of any reducing agents from an earlier NMR study could be because of hyper-reactive cysteine thiol oxidation (25). However, a spectrophotometric assay from that study indicated that all cysteines were in the free thiol state, suggesting that another explanation may be needed to reconcile the conflicting structural observations for finger 5.
Rotational diffusion motions of proteins are influenced by molecular weight, hydration, and shape. It seems highly unlikely that large hydration differences exist between different DNA binding zinc finger proteins that have similar surface charge potentials. The effective shape of a multizinc finger protein is defined by the ensemble average of finger-finger orientations that reside somewhere between two extremes: (i) individual tumbling domains restricted only by tethering to adjacent zinc fingers via flexible peptide linkers, and (ii) a single more anisotropic rigid body of connected modules. The significantly greater isotropically approximated rotational diffusion values calculated for MTF-1 F1-4 (13.7 kDa) compared with WT1-4 (14.1 kDa) indicate that the two zinc finger proteins, each with four canonical domain structures, have different ensemble average solution conformations. F1-6, F1-4, and F5-6 are characterized by significant anisotropic rotational diffusion properties (Table II) with partial alignment of the principal axes of diffusion for each internal finger (Fig. 8, data not shown for F1-4 and F5-6). Rotational diffusion properties based on an anisotropic model were not reported for WT1-4. However, the more homogeneous R 2 /R 1 ratios calculated for WT1-4 3 suggest less anisotropy compared with the MTF-1 F1-4 and F1-6 fingers. Overall, the most plausible model suggested from the 15 N relaxation data from all three F1-6, F1-4, and F5-6 proteins is one in which each F1-6 zinc finger is characterized by significant interdomain interactions that produce an elongated ensemble-averaged multifinger conformation. Indeed, an ensemble extended three-zinc finger domain from TF1-3 has been reported previously (62).
The clear increasing trend of R 2 /R 1 and J(0) shown in Figs. 5 and 6, respectively, that starts from N-and C-terminal fingers and progresses inward to internal fingers demonstrates that finger tethering plays a predictable and more general role in Cys 2 His 2 zinc finger rotational diffusion properties. Specifically, individual domain motions become increasingly more restricted for internal fingers that are tethered to adjacent single or multiple domains on both ends. The rotational diffusion motions of the MTF-1 zinc fingers are complex, because greater finger-finger interactions also increase the effective molecular weight experienced by the residues within each affected finger. This complexity more than likely contributed to the lack of success fitting model free parameters (64,65) to the MTF-1 F1-6 15 N relaxation data.
An elongated ensemble conformation resulting from more restricted domain motion compared with other previously stud- ). All resonances that correspond to residues from Zn(II)-bound protein are annotated with lettering color coded by finger as defined in panel B. B, plots of individual residue fractional 1 H-15 N-HSQC peak intensities as a function of added Zn(II) that are divided into residues from each MTF-1 zinc finger. Two distinct populations of finger 2 fractional peak intensities were observed, which are color coded light and dark blue. Fractional intensity is defined as the measured resonance intensity corresponding to a given residue and Zn(II) concentration divided by the maximum intensity of the same residue 1 H-15 N-HSQC peak from spectra recorded at all Zn(II) concentrations. C, average per finger fractional 1 H-15 N-HSQC peak intensities from panel B. Error arcs around each data point represent the S.E. from the averages. D, expanded regions of early titration points from panel C clearly show the differences in Zn(II) occupancy (relative affinity) for different MTF-1 zinc fingers. ied zinc finger proteins would have functional implications for MTF-1 DNA binding properties (60,61), although a direct link to MTF-1 metal sensing is not obvious. Because the linkers connecting adjacent zinc fingers are flexible in the DNA-free state (lower 15 N-{ 1 H}-NOE values), the structural determinants of the quasi-ordered MTF-1 zinc fingers suggest as yet unidentified interdomain contacts may be present within each finger. By contrast, the relative domain orientations of DNAbound zinc finger structures along the major groove of DNA are typically at least partially determined by highly specific residue-DNA base contacts and DNA-induced canonical ordered and less flexible linker conformations (60,70,71). Therefore, potential enthalpic DNA binding penalties could result from the disruption of interdomain contacts within the free F1-6 protein concomitant with DNA-induced changes in finger orientations. Even if the free F1-6 interdomain interactions orient the fingers in such a way that is optimal for DNA binding, a loss of at least some of the DNA-induced finger-finger packing stabilization is likely if the free fingers are already ordered (62). However, a stabilizing entropic DNA binding effect for F1-6 stemming from reduced finger-finger mobility in the free state could compensate the destabilizing enthalpic contributions, although the extent of this compensation is unknown.
The narrow 25-fold range in Co(II) affinities reported for the MTF-1 zinc fingers by Berg and coworkers (19) could suggest that multiple fingers are involved in metalloregulation if Zn(II) binding properties follow a similar trend. However, the Zn(II) affinities of individual MTF-1 zinc fingers are unknown, and parallel studies of Zn(II) and Co(II) binding by other Cys 2 His 2 zinc fingers have reported variable Zn(II)/Co(II) affinity ratios in the range of 10 2 -10 4 (52,72). Therefore, we sought to estimate an upper limit for the range of MTF-1 F1-6 Zn(II) affinities that would be consistent with our titration results. Theoretical Zn(II) occupancies of the MTF-1 zinc fingers were calculated from a general binding model of six metal sites, each at an experimental protein concentration of 180 M and a total Zn(II) concentration of 1.5 equivalents (270 M) using K a s spanning the maximum reported for any zinc finger (10 11 M Ϫ1 ) to the lowest measured MTF-1 Co(II) affinity (10 6 M Ϫ1 ) (19). The Zn(II) concentration used for modeling is the minimum concentration at which resonances from all six MTF-1 fingers could be observed (Fig. 9B). For all solutions to this model, any of the six zinc fingers one order of magnitude higher in K a compared with the lowest affinity finger would have at least 4.0-fold higher Zn(II) occupancy at 1.5 equivalents. Zinc fingers two orders of magnitude higher in K a would be at least 8.1-fold higher. Actual high/low Zn(II) occupancy ratios would probably be substantially greater because this is the predicted outcome of having several higher affinity fingers, and fingers 1-5 appear to be higher affinity than finger 6 ( Fig. 9, C and D). Close inspection of the NMR data recorded at 180 M protein indicates that the average peak intensity at 1.5 equivalents of Zn(II) for all observable resonances from the highest affinity MTF-1 zinc finger (finger 4) was 0.063 Ϯ 0.014, whereas the absolute threshold for peak detection was Ն0.010. By comparison, average peak intensities for the lowest affinity finger at 1.5 equivalents of Zn(II) was 0.012 Ϯ 0.002. Given the strong correlation of peak intensity with Zn(II) occupancy for each finger shown clearly in Fig. 9B, this ϳ5-fold finger 4/finger 6 peak intensity ratio provides strong evidence that the relative affinities for Zn(II) of all MTF-1 zinc fingers are remarkably similar (within ϳ10 -50-fold).
Overall, the Zn(II) binding studies presented here strongly support the role of multiple similar affinity zinc fingers in governing metal sensing by MTF-1 in eukaryotic cells. Moreover, because two of the medium Zn(II) affinity fingers (finger 1 and finger 3) are also needed for high affinity DNA binding and concomitant MTF-1 function (20,22), an upper limit for the concentration range over which accessible Zn(II) is sensed and responded to by MTF-1 can be estimated using the binding model described by Equation 2. Assuming similar affinities for all F1-6 fingers indicated from Fig. 9D (within 50-fold), F1-6 occupancies ranging from Ͻ10 to Ͼ90% of the maximum attainable will always occur within a Zn(II) concentration range of ϳ100-fold or less for all finger K a values within the range of 10 6 -10 11 M Ϫ1 regardless of MTF-1 or accessible Zn(II) concentrations. In this model, no explicit consideration is given for potential finger-finger or DNA binding cooperative effects on Zn(II) binding affinities, although these effects would be expected to decrease the Zn(II)-responsive range for metal sensing. Indeed, evidence from Co(II) binding and thiol alkylation studies suggest modest cooperativity from both finger-finger and DNA binding (19,68). The absolute affinity and in vivo copy number range for the regulatory fingers needed to establish the intracellular accessible Zn(II) "set point" concentration and actual concentration range that is sensed by MTF-1 is currently under investigation.