A novel zinc snap motif conveys structural stability to 3-methyladenine DNA glycosylase I.

The Escherichia coli 3-methyladenine DNA glycosylase I (TAG) is a DNA repair enzyme that excises 3-methyladenine in DNA and is the smallest member of the helix-hairpin-helix (HhH) superfamily of DNA glycosylases. Despite many studies over the last 25 years, there has been no suggestion that TAG was a metalloprotein. However, here we establish by heteronuclear NMR and other spectroscopic methods that TAG binds 1 eq of Zn2+ extremely tightly. A family of refined NMR structures shows that 4 conserved residues contributed from the amino- and carboxyl-terminal regions of TAG (Cys4, His17, His175, and Cys179) form a Zn2+ binding site. The Zn2+ ion serves to tether the otherwise unstructured amino- and carboxyl-terminal regions of TAG. We propose that this unexpected "zinc snap" motif in the TAG family (CX(12-17)HX(approximately 150)HX(3)C) serves to stabilize the HhH domain thereby mimicking the functional role of protein-protein interactions in larger HhH superfamily members.

Excision repair of damaged bases in DNA is one pathway that cells use to protect the genome from the damaging effects of ionizing radiation, reactive oxygen species, and alkylating agents (1). Lesion repair begins by enzymatic hydrolysis of the glycosidic bond, the critical initiating step of the DNA base excision repair pathway. There are many different types of DNA glycosylases in cells. In general, each DNA glycosylase exhibits specificity for a cognate damaged base while ignoring undamaged bases in DNA. One alkylated base, 3-methyladenine, is highly toxic, because DNA replication is blocked at this lesion site (2). In Escherichia coli, the glycosidic bond of 3-methyladenine is hydrolyzed by two enzymes, 3-methyladenine DNA glycosylase I (TAG) 1 and II (AlkA) (3).
It has been recently discovered that TAG is a member of the helix-hairpin-helix (HhH) superfamily of DNA glycosylases (4).
The HhH motif is a sequence-nonspecific DNA binding module found in DNA polymerases, NAD ϩ -dependent DNA ligases, and some DNA glycosylases (5). This motif consists of two ␣-helices connected by a consensus hairpin loop that interacts nonspecifically with the DNA backbone. Unlike other HhH family members, TAG does not possess the consensus hairpin sequence, (L/F)PG(V/I)G, nor does it contain the conserved aspartate group that was previously believed to be required for catalysis as a water activating group or, alternatively, for stabilization of a transition state with glycosyl cation character.
Thus, TAG appears to be unique with respect to structure and mechanism within this superfamily (4).
Although it has been concluded previously that TAG does not require a metal ion for activity (6), the possibility of a metal binding site was suggested by our recent NMR structure and a coincident bioinformatics study (4,7). According to the solution structure, two conserved sulfur and histidine ligands at opposing ends of the linear sequence are closely positioned in threedimensional space (Cys 4 , His 17 , His 175 , and Cys 179 ) (4, 7), but given the limitations of the NMR method, the presence of a metal ion was not established. As shown in Fig. 1, the sequence alignment of the TAG family shows that these potential metal ligands are also completely conserved across all species. Taken together, these observations clearly suggested the presence of a metal ion binding site. Here we establish this hypothesis and present the structure of a long overlooked Zn 2ϩ binding site in TAG. The metal site stabilizes the structure of TAG by "snapping" together the largely unstructured N-and C-terminal regions. This "zinc snap" motif (CX 12-17 HX ϳ150 HX 3 C) is distinct from that of zinc finger proteins or the zinc binding sites in other DNA repair proteins such as formamidopyrimidine DNA glycosylase, poly(ADP-ribose) polymerase, and UvrA (8 -11). We propose that the minimal HhH domain is intrinsically unstable and that the zinc snap provides the stabilizing forces that are assumed by intramolecular protein-protein contacts in other larger HhH family members.

EXPERIMENTAL PROCEDURES
Enzyme Purification-Wild-type TAG was overexpressed from pET21a(TAG) in BL21(DE3) and was purified by two sequential chromatography steps as described previously by Drohat et al. (4). A His 6 -tagged version was overexpressed from pET28a(TAG) in BL21(DE3)pLysS and purified using a nickel-nitrilotriacetic acid-agarose column. The amino-terminal tag was cleaved and removed using biotinylated thrombin, streptavidin agarose resin, and a second nickelnitrilotriacetic acid-agarose column. Final purification was achieved using a Poros 20-S HPLC column (Applied Biosystems, Foster City, CA) using a linear gradient of 0.1-1 M NaCl. The purity of TAG was Ͼ95% by both procedures.
Atomic Absorption Spectroscopy (AAS)-The presence of zinc in the holoenzyme was determined using a PerkinElmer model 370 atomic absorption spectrophotometer. The purified enzyme was dialyzed extensively against buffer A (50 mM Tris-HCl, 200 mM NaCl, pH 7.5) containing 4 ml of hydrated chelex-100 resin (Sigma)/500 ml of buffer. * This work was supported by National Institutes of Health Grant GM46835. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
A 2-ml volume of TAG (2 or 5 M) was injected into the atomic absorption spectrophotometer with monitoring at the absorption maximum for Zn 2ϩ (213.5 nm). AAS on apoTAG was performed by the following procedures. Partial removal of Zn 2ϩ from holoTAG was achieved using nondenaturing conditions by dialysis of the enzyme against low pH buffer B (20 mM MES, 200 mM NaCl, pH 5.5) containing 50 mM EDTA and chelating resin. Complete removal of Zn 2ϩ was achieved by denaturing the enzyme in 0.1 M HCl and then separating the denatured apoenzyme from the released Zn 2ϩ by gel filtration chromatography (PD-10, Amersham Biosciences). The TAG concentration was determined by UV absorbance using a calculated extinction coefficient of 30,800 M Ϫ1 cm Ϫ1 in a buffer containing 20 mM sodium phosphate (pH 6.0) and 6 M guanidinium chloride. 2 Standard solutions of ZnCl 2 were used to calibrate the atomic absorption spectrophotometer.
UV-visible Absorption Spectroscopy of Co 2ϩ Complex-A solution of TAG with 78% of the Zn 2ϩ removed was prepared by dialysis against buffer B as described above. To remove EDTA and raise the pH, the enzyme (100 l, 25 M total protein) was exchanged into buffer C (20 mM Tris-HCl, 100 mM NaCl, pH 7.5) using two sequential Bio-Gel P-6 gel filtration columns (Bio-Rad). An absorption spectrum of the 78% apoTAG sample (as determined from AAS) was collected in the wavelength range of 320 to 800 nm using a Beckman DU 640 spectrophotometer. Then, 1 molar eq of CoCl 2 was added quickly, and the spectrum was reacquired. A polynomial base-line subtraction was used to correct the spectrum for light scattering by apoTAG.
NMR Spectroscopy-The NMR sample, prepared as described in a previous report (4), contained ϳ0.5 mM TAG, 10 mM phosphate buffer (pH 6.6), 100 mM NaCl, 3 mM dithiothreitol, 0.34 mM NaN 3 , and 10% D 2 O in a total volume of 300 l. The two-dimensional 1 H-15 N LR HSQC experiment was conducted at 20°C on a Varian Unity Plus 600-MHz spectrometer equipped with four frequency channels and pulse-field gradients as described (12). This experiment was simply a conventional HSQC used for backbone amide correlations collected with an optimized two-bond 2 J NH value of 22 Hz in order to observe signals from the weak two-bond couplings in the histidine rings and suppress the signals from the one-bond J NH amide couplings. The 15 N dimension was collected with 256 complex points, a 120-ppm sweep width, 128 scans, and the 15 N carrier set at 205 ppm. The 1 H dimension was collected with 1024 complex points and a 13.3-ppm sweep width centered at 4.82 ppm. Decoupling of 15 N was accomplished with the WALTZ16 sequence (13) using a 5.1-kHz field.
NMR Structural Calculations-Starting from the published NMR structure of TAG (Protein Data Bank code: 1LMZ), which was originally refined without any constraints to Zn 2ϩ , a set of 100 structures that included restraints from the enzyme ligands to Zn 2ϩ were calculated using the program XPLOR-NIH, version 2.0.2 3 (14). These calculations employed the same NOE constraints, dihedral angle, and H-bond restraints as the original structures. Additionally, because of new assignments obtained from the 1 H-15 N LR-HSQC experiment, a total of 11 new NOEs and three new hydrogen bonds were used as restraints in the calculation. From this set, 25 low energy structures were selected with distance restraint violations less than 0.5 Å and dihedral angle restraint violations less than 5 o . Constraints between the protein ligands and the zinc ion were added using the procedure of Neuhaus et al. (15).
The constraint values were based on tetrahedral coordination, using literature values for the HisN ⑀,␦ -Zn and CysS ␥ -Zn bond lengths of 2.0 a Two batches of TAG, which were purified by different methods as described under "Experimental Procedures," contained identical equivalents of bound zinc.
b The samples were prepared by dialysis against different buffers: buffer A (50 mM Tris-HCl, 200 mM NaCl, pH 7.5) ϩ Chelex-100, buffer A ϩ 40 M CoCl 2 , or buffer B (20 mM MES, 200 mM NaCl, pH 5.5) ϩ 50 mM EDTA and Chelex-100. c TAG was denatured using 0.1 M HCl and separated from the released zinc ions using gel filtration column.
FIG. 2. Absorption spectra of apo-and Co 2؉ -TAG. A sample of TAG was prepared in which 80% of the bound Zn 2ϩ was removed, 1 eq of Co 2ϩ was then added, and the UV-visible spectrum was collected (solid line). The spectrum of apoTAG is shown as a broken line. The sequences of the TAG family were obtained by running a BLAST search using the E. coli TAG sequences. The alignment of the partial sequences was generated using T-COFFEE, version 1.41 (25). In the consensus sequences, an asterisk indicates identical amino acid residues and a colon indicates highly conserved residues. The metal-binding Cys 2 His 2 cluster, CX 12-17 HX n HX 3 C, is highlighted in black. The NCBI gene identification (GI) numbers of the various TAG sequences are shown next to the names of the species. and 2.3 Å, respectively, which were obtained from EXAFS (extended x-ray absorption fine structure) measurements (16). The following additional distance constraints were used (15)

TAG Has One Tight Zn 2ϩ
Binding Site-To determine whether TAG possesses a tight metal binding site, AAS measurements were performed (Table I). Two independently purified samples of TAG that had been dialyzed extensively against a neutral pH buffer containing excess chelating resin showed the presence of 1 molar eq of Zn 2ϩ by AAS analysis (Table I). It was found that exhaustive dialysis of these same samples against a buffer that contained 40 M CoCl 2 (pH 7.5) still retained 1 eq of Zn 2ϩ , indicating that the site was inert to exchange under these conditions. Surprisingly, even extensive dialysis of the enzyme against a low pH buffer (pH 5.5) that contained 50 mM EDTA removed only about 0.8 eq of the bound metal. In fact, complete removal of Zn 2ϩ required denaturation of the enzyme using 0.1 M HCl (Table I). These results indicate that previous reports of TAG not requiring a metal ion for activity were due to the very high binding affinity of Zn 2ϩ for this site, which cannot be removed by standard chelating agents that are included in the purification buffers (6,14). The previously reported dramatic loss in TAG activity below pH 6 may be attributed to metal ion loss and partial protein unfolding (17). We found that the apoTAG was unstable and aggregated extensively at pH 7.5 (data not shown).
Geometry and Ligands of the Zn 2ϩ Site-One valuable method for determining the ligands and coordination geometry for a Zn 2ϩ binding site in enzymes is to perform metal replacement with Co 2ϩ , a transition metal that gives diagnostic features in its UV-visible spectra (18 -20). For example, thiolate-Co 2ϩ bonds show a diagnostic charge transfer band at about 350 nm with a typical extinction coefficient of 900 -1300 M Ϫ1 cm Ϫ1 /thiolate ligand (21,22). Additional bands in the 576 -670 nm range are indicative of the coordination geometry; tetrahedral coordination gives a strong signal (E Ͼ300 M Ϫ1 cm Ϫ1 ), whereas pentacoordinate and octahedral sites have much lower molar absorptivities (50 Ͻ E Ͻ 225 M Ϫ1 cm Ϫ1 and E Ͻ 30 M Ϫ1 cm Ϫ1 , respectively) (18,(23)(24)(25).
The UV-visible spectrum of Co 2ϩ -TAG (Fig. 2) is consistent with two sulfur ligands and a tetrahedral coordination geometry. Because of the extensive aggregation of apoTAG during the Co 2ϩ spectroscopy, the concentration of the soluble protein in the cuvette could not be known precisely. However, comparing the ratio (R) of extinction coefficients of the 350 nm and 617 nm bands (R ϭ E 350 /E 617 ) after base-line correction still provides quantitative information on the number of sulfur ligands and the coordination geometry. The expected ratio for tetrahedral geometry is 1 Ͻ R Ͻ 4 for one thiolate ligand, and 2 Ͻ R Ͻ 9 for two thiolate ligands. The latter value encompasses the E 350 / E 617 ϭ 4 measured for TAG. This value differs considerably from the expected ratios for pentacoordinate and octahedral geometry for two thiolate ligands, which would be 8 Յ R Յ 52 and R Ն 60, respectively. Although these estimates have some uncertainty, they are consistent with the strict conservation of 2 Cys and 2 His residues in the TAG family (Fig. 1), and the structural results described below.
Electronic Properties of the Zn 2ϩ Ligands Determined by Heteronuclear NMR Spectroscopy-Although the Co 2ϩ replacement studies described above indicate the ligation of two sulfur atoms to the Zn 2ϩ , the identities of the remaining two ligands are not revealed by this method. Because our previous NMR structure and the homology comparison suggested that His 17 and His 175 were the remaining ligands (Fig. 1), we performed a 1 H-15 N LR-HSQC NMR experiment to investigate the electronic properties of these histidines (Fig. 3A; the histidine spin systems in Fig. 3A were obtained from three-dimensional 13 Cedited aromatic and aliphatic NOESY experiments). This simple LR-HSQC experiment correlates the carbon-bound protons of the histidine rings with the imidazole nitrogen atoms and can unambiguously establish the tautomeric and protonation states of histidines in proteins (12,26). The characteristic upside-down and sideways L-shaped patterns for the peaks in the two-dimensional spectrum for His 175 and His 17 and their well separated 15 N chemical shifts indicate that these histidines are FIG. 3. NMR studies of the zinc ligands. A, two-dimensional 1 H-15 N LR-HSQC spectrum of Zn 2ϩ -TAG at pH 6.6 and 20°C. The individual histidine spin systems are connected by solid black lines. The cross-peaks of the residue His 122 , which are too weak to be observed using the current contour level, are marked with an asterisk. B, the 15 N chemical shifts for the N ␦ 1 and N ⑀ 2 side-chain atoms and the tautomeric states of the 5 histidine residues in TAG. C, the 13 C chemical shifts for the ␤-carbons of the 8 cysteine residues. The metal ligand is marked in bold characters in B and C.
in the N E 2-H and N ␦ 1-H tautomeric forms, respectively, and that both are neutral (Fig. 3, A and B). Thus the nitrogens that are available to coordinate the zinc are the N ␦ 1 of His 175 and the N ⑀ 2 of His 17 . The 15 N ␦ 1 shift for His 175 is by far the most deshielded nitrogen of all histidines in TAG, consistent with strong chelation to an electropositive Zn 2ϩ . The 15 N ⑀ 2 chemical shift of the ligating nitrogen of His 17 is also the most deshielded of the five histidines, also indicating strong chelation to the metal from this position. Similarly, the 13 C ␤ chemical shifts of Cys 4 and Cys 179 are the most deshielded of the 8 cysteine residues of TAG, consistent with thiolate metal coordination from these side-chains (Fig. 3C).
Structure of Zn 2ϩ Binding Site-Because a zinc binding site was entirely unanticipated in TAG, our previous NMR structure did not include any restraints to a Zn 2ϩ atom (Protein Data Bank code: 1LMZ). Therefore, with unambiguous evidence for a Zn 2ϩ site in hand, we further refined the solution structure using the same constraints employed previously but including distance and angle constraints between the Zn 2ϩ and Cys 4 , His 17 , His 175 , and Cys 179 (see "Experimental Procedures"). Inclusion of these constraints was justified on the basis of the UV-visible spectroscopy measurements, which indicated two sulfur ligands with tetrahedral geometry, and the LR-HSQC experiments, which indicated coordination of His 17 N ⑀ and His 175 N ␦ to the zinc. The addition of the zinc ion and these new constraints did not increase the Lennard-Jones potential energy for the ensemble of 25 lowest energy structures, nor did it introduce new NOE violations. For the new ensemble, about 85% of the dihedral angles are in the most favored region, with 11% in the additionally allowed region. These statistics are within error of those reported previously; further structural statistics are reported in Table II.
The new NMR structure containing Zn 2ϩ is depicted as a ribbon diagram in Fig. 4A and confirms that the Zn 2ϩ binding site tethers the amino and carboxyl termini of TAG. This structure is nearly identical to the deposited TAG structure with r.m.s.d. values of 0.66 Å for backbone atoms and 0.86 Å for all heavy atoms for residues 11-174. The 3-methyladenine binding site, marked with an asterisk in Fig. 4A, is removed from the metal site, precluding any direct involvement of the metal in catalysis. However one metal ligand, His 175 , is involved in a strong H-bond to the carbonyl oxygen of the active site group, Trp 21 (His 175 ␦(NH ⑀ 2) ϭ 14.75 ppm), suggesting that the zinc could indirectly organize the structure of the active site. The detailed structure of the Zn 2ϩ binding site is shown in Fig.  4B (27).
Role of the Zinc Snap Motif in the Helix-Hairpin-Helix Superfamily-The Zn 2ϩ binding site in TAG is distinct from zinc finger motifs found in bacteria to eukaryotes (28), as well as the zinc binding motifs found in the bacterial 8-oxoguanine glycosylase, MutM (8,9). The four Cys zinc finger motif in MutM is in the arrangement CX 2 CX 16 CX 2 C and serves to stabilize two ␤-strands that deliver three polar side chains to their interac- FIG. 4. NMR solution structure. A, three-dimensional structure of TAG including restraints to a Zn 2ϩ ion. The signature HhH structural motif is highlighted in green, and the remaining elements of the conserved helical domain of the HhH glycosylase fold are colored blue. The structure elements that are unique to TAG are shown in orange-red, and the 3-methyladenine binding pocket is indicated with an asterisk. B, the tetrahedral Zn 2ϩ binding site of TAG. tions with the phosphodiester backbone (9). In contrast, the zinc motif in TAG consists of two sets of CX n H submotifs, which are separated by 143-151 amino acid residues. This arrangement differs significantly from zinc finger motifs, which cluster in the region of a turn or a loop. The split zinc binding motif, with ligands donated from both termini of TAG, acts as a "snap" for closing the ends of the protein. The zinc binding site is distant from the previously modeled DNA binding region of TAG (4) and serves to stabilize tertiary structure, but it is not involved directly in DNA binding. Given the small size of the domain, the zinc ion provides a critical stabilizing element. This is supported by our observation that the apoprotein appears to be poorly folded, leading to aggregation and eventual precipitation. Difficulties in preparation of the TAG mutant, H17A, because of low expression levels and aggregation (data not shown), are consistent with the conclusion that the Zn ϩ2 binding motif is important for stabilization of the protein.
Structural comparisons between TAG and other HhH glycosylases show that all other family members have appendages to their amino and carboxyl termini that interact extensively and are likely to contribute to stabilizing the HhH fold. Therefore, the newly revealed zinc snap motif allows the TAG family to efficiently stabilize the HhH structure without additional protein scaffolding.