Learn more: PMC Disclaimer | PMC Copyright Notice
The solution structure of the N-terminal domain of E3L shows a tyrosine conformation that may explain its reduced affinity to Z-DNA in vitro
Associated Data
Abstract
The N-terminal domain of the vaccinia virus protein E3L (ZαE3L) is essential for full viral pathogenicity in mice. It has sequence similarity to the high-affinity human Z-DNA-binding domains ZαADAR1 and ZαDLM1. Here, we report the solution structure of ZαE3L and the chemical shift map of its interaction surface with Z-DNA. The global structure and the Z-DNA interaction surface of ZαE3L are very similar to the high-affinity Z-DNA-binding domains ZαADAR1 and ZαDLM1. However, the key Z-DNA contacting residue Y48 of ZαE3L adopts a different side chain conformation in unbound ZαE3L, which requires rearrangement for binding to Z-DNA. This difference suggests a molecular basis for the significantly lower in vitro affinity of ZαE3L to Z-DNA compared with its homologues.
Vaccinia virus is a member of the large double-stranded DNA family of poxviruses. It has been used globally as a vaccine to eradicate smallpox, a devastating disease caused by variola virus that is presently an important threat of bioterrorism (1). The vaccinia virus protein E3L, which is conserved in variola and related viruses, plays a key role in circumventing the IFN-mediated defense of host cells (2). E3L contains two domains, of which the C-terminal double-stranded RNA-binding domain is essential and sufficient for evading IFN host defense in cultured cells. In animal models, however, full pathogenesis requires the N-terminal domain of E3L (2), which has sequence homology to the family of Z-DNA-binding protein domains (Zα) but shows only comparatively low affinity to Z-DNA in vitro (3). When the ZαE3L domain is removed from vaccinia virus and is replaced by either ZαADAR1 or ZαDLM1, the virus retains full pathogenicity in the mouse model. Mutational studies show that these domains bind to Z-DNA (4).
The structurally defined Zα domains are 62-residue (α plus β) helix–turn–helix proteins with an additional β-sheet that bind to left-handed Z-DNA with medium nanomolar affinity in vitro (5, 6). The 3D structures of the human Zα domains of the RNA-editing enzyme ADAR1 (ZαADAR1) and of the tumor-related protein DLM1 (ZαDLM1) were solved complexed with Z-DNA (7, 8). Further, the solution structure of ZαADAR1 was determined in the unbound state (9), and residues essential for binding to Z-DNA were identified by alanine-scanning mutagenesis (5). Single-point mutations in two such residues, which strongly reduce the affinity of ZαADAR1 to Z-DNA in vitro, were recently shown to abrogate vaccinia virus pathogenicity in a mouse model when introduced in homologous positions in the N-terminal domain of E3L (ZαE3L) (4). Considering this close correlation in the function of such residues between ZαADAR1 and ZαE3L the lack of correlation in their in vitro affinity to Z-DNA is intriguing.
Here, we report the solution structure of ZαE3L and a chemical shift map of its interaction surface with Z-DNA. The 3D structure and interaction surface of ZαE3L is grossly very similar to ZαADAR1 and ZαDLM1 but differs in the side chain conformation of a pivotal Z-DNA-contacting residue Y48. In contrast to ZαADAR1 and ZαDLM1, Y48 of free ZαE3L is exposed to solvent and shows selective vanishing of its NMR signals consistent with a conformational rearrangement when Z-DNA is bound. Therefore, the additional cost in energy for rearranging Y48 may account for the substantially lower in vitro affinity of ZαE3L to Z-DNA as compared with its homologues ZαADAR1 and ZαDLM1. In vivo, this energy may be provided by other factors, rendering wild-type ZαE3L–E3L as pathogenic as the ZαADAR1–E3L chimera (4).
Materials and Methods
Protein Preparation. Residues 1–78 of the vaccinia virus gene E3L (GenBank no. AAA02759), comprising the ZαE3L domain, with four additional vector-encoded residues at the N terminus, were expressed as a fusion protein with an N-terminal (His)6 tag from a pET-28 vector (Novagen) in Escherichia coli strain BL21(DE3). Alternatively, residues 5–70 of E3L were expressed similarly for DNA interaction assays.
To produce 15N- and 15N/13C-labeled ZαE3L, bacteria were grown in M9 medium containing 1 g/liter 15NH4Cl and 1.5 g/liter 13C-glucose. Cultures were induced with 1 mM isopropyl β-d-thiogalactoside (IPTG) when they reached OD600 of 0.8. After 4 h induction, cells were harvested, resuspended in 50 mM NaH2PO4 (pH 8.0), 300 mM NaCl, 10 mM imidazole, supplemented with Roche Complete Protease Inhibitor Mix, and lysed by French pressing. The lysate was centrifuged at 48,000 × g for 30 min. The supernatant was applied on a Ni-NTA column (Qiagen, Hilden, Germany). After washing with 50 mM NaH2PO4 (pH 8.0), 300 mM NaCl, 20 mM imidazole, the (His)6 tag protein attached to the Ni2+ matrix was eluted by thrombin digestion overnight at room temperature in PBS. The cleaved protein was loaded on a Resource Q column (Amersham Biosciences), and the bound protein was eluted with a gradient of 0–1 M NaCl (20 mM NaH2PO4, pH 6.5). Alternatively, bacteria were lysed with Bugbuster (Novagen) following the manufacturer's protocol, and protein was purified as described (4). Briefly, the (His)6-fusion protein was purified by using HisBind resin (Novagen). Bound protein was stepped off with 300 mM imidazole in 500 mM NaCl, 20 mM Tris (pH 8.0). The (His)6 tag was removed by thrombin digestion for 3 h at room temperature in 20 mM Tris·HCl (pH 8.0), 150 mM NaCl, 2.5 mM CaCl2, 1 mM EDTA. ZαE3L protein was further purified by using a HiTrap Q column (Amersham Biosciences) and a 25–500 mM NaCl gradient. Matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry of the 15N- and 15N/13C-labeled ZαE3L yielded single peaks at 9,291 Da and 9,672 Da, respectively, which agree very well with the calculated molecular masses of 9,292 Da and 9,685 Da, respectively.
NMR Spectroscopy. NMR experiments were carried out at 25°Con 2.2-mM u-15N-labeled and 2-mM u-13C, 15N-labeled ZαE3L samples in 20 mM sodium-phosphate (pH 6.5), 20 mM NaCl, 0.1 mM NaN3 with 5% and 100% D2O, respectively, on 600-MHz NMR spectrometers. 1H, 15N, and 13C resonance assignments were obtained from the following 3D heteronuclear correlation experiments (10): CBCA(CO)NH, CBCANH, HBHA(CO)NH, H(CCO)NH, C(CCO)NH, HCCH-COSY, and HCCH-TOCSY. Interproton distance restraints were derived from 3D 15N-heteronuclear single quantum coherence (HSQC)-NOESY (150-ms mixing time), 3D 13C-HSQC-NOESY (40 and 70 ms mixing times). Spectra were processed with xwinnmr (Bruker) and analyzed with sparky 3.105 (11). Spectra were referenced by external calibration on 2,2-dimethyl-silapentane-5-sulfonic acid (DSS), sodium salt (12).
Interaction Mapping. For interaction mapping, a shortened ZαE3L construct (comprising residues 5–70 of GenBank no. AAA02759) was used that lacks the first four N-terminal and the last eight C-terminal residues. These residues are nonstructured in the 3D structure of ZαE3L. The 1H and 15N backbone chemical shifts are virtually identical between this construct and the 1–78 residues construct, indicating that both constructs share the same 3D fold. 1D 1H and 2D 15N-HSQC NMR spectra were recorded on the following four samples in 20 mM [bis(2-hydroxyethyl)amino]tris(hydroxymethyl)methane (Bistris; pH 6.7), 50 mM NaCl, 5% D2O at 25°C: (i) 40 μM ZαE3L, (ii) 40 μM ZαE3L with 300 μM [Co(NH3)6]3+, (iii) 40 μM ZαE3L with 10 μM d(CG)6T4(CG)6, and (iv) 40 μM ZαE3L with 10 μM d(CG)6T4(CG)6 and 300 μM [Co(NH3)6]3+. After NMR data acquisition, CD spectra were recorded on each sample at room temperature by using a 2-mm cuvette on a Jasco J-720 CD spectrometer (Jasco, Easton, MD). CD control spectra of 10 μM d(CG)6T4(CG)6 in the B-DNA conformation and in the Z-DNA conformation were recorded in a 1-mm cuvette at room temperature in 20 mM Bistris (pH 6.7), 50 mM NaCl buffer alone, and with 5 M NaCl, respectively. The assignments and concentration dependence of the chemical shift changes of ZαE3L were confirmed by a titration experiment on a 20-μM ZαE3L sample with increasing concentrations of d(CG)6T4(CG)6 (1, 2, 4, 7, and 12 μM) and a constant [Co(NH3)6]3+ over DNA excess of 30 in 20 mM Bistris (pH 6.7), 50 mM NaCl, and 5% D2O at 25°C. Spectra were processed and analyzed as described above. Chemical shift changes were averaged according to the formula [(Δ 1H)2 + (Δ 15N/5)2]0.5 (13).
Structure Calculation. Nuclear Overhauser enhancement (NOE) distance restraints derived from 15N- and 13C-edited NOESY experiments were manually assigned and further analyzed (calibration and removal of redundant distance restraints) by using the program dyana 3.1 (14). Seventy-two backbone dihedral angle constraints were derived from Cα chemical shifts according to the rules (15): -120° < Ψ < -20° and -100° < φ < 0° for Δ(Cα) > 1.5 ppm, and -200° < Ψ < -80° and 40° < φ < 220° for Δ(Cα) < -1.5 ppm. The dihedral angle constraints are in agreement with the preliminary structure calculated solely from the NOE restraints. Further 20 hydrogen bonds within α-helices and four hydrogen bonds within β-strands were derived from the NOE-based preliminary structure and confirmed by the analysis of the Cα and Cβ chemical shift values by using the program talos (16). Structures were calculated by 4,000 steps of simulated annealing with torsion angle dynamics and subsequently 1,000 steps of minimization in dyana 3.1. For better convergence during structural refinement, the Ψ and φ dihedral angles of the residues 59, 60, and 61 preceding the cis peptide bond between I62 and P63 were preset to a range that is wider by 5° or more than the range of the structural ensemble calculated without such preset angles.
Results and Discussion
Structure Determination. The 3D structure of the N-terminal Z-DNA-binding domain of the vaccinia virus gene product E3L (residues 1–78) was determined by multidimensional NMR spectroscopy in solution. Complete chemical shift assignments were obtained from 3D triple-resonance and double-resonance NMR spectra except for the first two vector-encoded residues. Chemical shifts of residues 1–10 and 69–78 are not well dispersed. This finding is confirmed by the 3D structure demonstrating that the folded core domain comprises residues 11–68 (subsequently referred to as ZαE3L). Residues 9–10 and 69–70 show a preferred orientation, as evidenced by a few medium-range NOE, whereas all other N- and C-terminal residues are unstructured. Structural statistics are listed in Table 1. The coordinates of the ensemble of the 20 lowest energy structures of ZαE3L (residues 9–70) have been deposited in the Protein Data Bank with the PDB ID code 1OYI.
Table 1.
| NOE upper distance limits* | 487 |
| Dihedral angle constraints* | 78 |
| Residual target function, Å2*† | 0.9 ± 0.1 |
| Residual distance constraint violations*†‡ | |
| Number ≥ 0.1 Å | 5.5 ± 0.3 |
| Maximum, Å | 0.24 ± 0.05 |
| Residual dihedral angle constraint violations*† | |
| Number ≥ 2 deg | 0 ± 0 |
| Maximum, deg | 0.1 ± 0.1 |
| rms deviations from ideal geometry§ | |
| Bond lengths, Å | 0.01 |
| Bond angles, deg | 1.7 |
| rms deviation from the mean coordinates†¶ | |
| Backbone (N, Ca, C), Å | 0.8 ± 0.2 |
| All heavy atoms, Å | 1.6 ± 0.2 |
| Ramachandran analysis†¶ | |
| Residues in most favored regions | 72.5% |
| Residues in additional allowed regions | 22.6% |
| Residues in generously allowed regions | 4.8% |
| Residues in disallowed regions | 0.1% |
Structure Description. The ZαE3L domain is composed of three α-helices (designated α1, α2, and α3) and three β-strands (designated β1, β2, and β3) in an α1β1α2α3β2β3 linear order (Fig. 1). Helices α1 and α3 pack against a short anti-parallel, triple-stranded β-sheet, in which β3 is sandwiched between β1 and β2. Strands β1 and β3 are connected by only two backbone hydrogen bonds between residues A27 and W66. Strands β2 and β3 are bridged by three hydrogen bonds comprising residues Y57 and S59 of β2 and R65 and F67 of β3. The ensemble of the 20 lowest energy structures shows that only loop 2 between α2 and α3 is less rigid whereas all other loops are tightly structured rendering ZαE3L a rigid body. Loop 4 between β2 and β3 is made rigid by the two sequential prolines 63 and 64, of which the former adopts a rare cis peptide bond. The side chains of the other residues in this loop (D60, D61, I62, and R65) are flexible in the structural ensemble, giving this loop the shape of a solvent accessible finger with a rigid backbone. Mutational (5) and structural studies (7, 9) have demonstrated that this distinctly conserved feature is essential for selective interaction of the homologous ZαADAR1 domain with Z-DNA. In the co-crystal structure of ZαADAR1 and Z-DNA, the protein makes several important van der Waals contacts to Z-DNA.
3D solution structure of ZαE3L.(A) Stereoview of the ensemble of the 20 lowest energy structures of E3L. The first and last residue of the 3 α-helices (red) (α1, α2, α3) and 3 β-strands (cyan) (β1, β2, β3) are numbered. (B) Stereoview of the backbone ribbon of the mean structure illustrating the (α plus β) helix–turn–helix fold of ZαE3L. The N- and C-termini are labeled with N and C, respectively. (C) The secondary structure of ZαE3L is shown with the amino acid sequence directly under it.
Chemical Shift Mapping of the E3L/Z-DNA Interaction Surface. Alternating d(CG)n oligomers have been successfully used to study the interaction of the Z-DNA-binding domains ZαADAR1 (7, 9) and ZαDLM1 (8) with Z-DNA. In contrast to these high-affinity Z-DNA-binding homologues, ZαE3L does not flip the B-DNA conformation of these substrates into the Z-conformation when incubated with each other under physiological buffer conditions at micromolar concentrations (3). The CD spectrum of 40 μM ZαE3L in the presence of 10 μM d(CG)6T4(CG)6 clearly indicates a B-DNA conformation for this DNA substrate (Fig. 2A, blue curve). Further, the 2D 15N-HSQC NMR spectrum of this sample shows no chemical shift changes when compared with the spectrum of the protein alone (see Fig. 4, which is published as supporting information on the PNAS web site). This result indicates that ZαE3L does not interact with d(CG)6T4(CG)6 in the B-DNA conformation under these conditions because chemical shifts are sensitive even to weak interactions.
(A) The conformation of substrate DNA in the presence of ZαE3L. The CD spectrum of d(CG)6T4(CG)6 substrate DNA in the presence of ZαE3L (blue curve) shows a conventional B-DNA conformation. In contrast, the CD spectrum of d(CG)6T4(CG)6 in the presence of ZαE3L and [Co(NH3)6]3+, which is known to promote the conversion from B- to Z-DNA (17), shows an inversion of ellipticity at 295 and 255 nm characteristic for duplex DNA in the Z-conformation (red curve). For control, ZαE3L alone (green curve) and ZαE3L in the presence of [Co(NH3)6]3+ (gray curve) show zero ellipticity between 320 and 250 nm. (B) Chemical shift map of the ZαE3L/substrate DNA interaction. The 15N-HSQC NMR spectrum of ZαE3L in the presence of d(CG)6T4(CG)6 and [Co(NH3)6]3+ (red peaks) shows several large chemical shift changes compared with ZαE3L alone (green peaks). In addition, there are three vanishing resonances (residues underlined). This finding indicates a selective interaction between ZαE3L and d(CG)6T4(CG)6 in the vicinity of the affected residues. The 15N-HSQC NMR spectrum of ZαE3L in the presence of d(CG)6T4(CG)6 substrate DNA shows no chemical shift alterations as compared with ZαE3L alone (Fig. 4), indicating no discernible interaction under these conditions. Further, the spectrum of ZαE3L in the presence of [Co(NH3)6]3+ is identical to ZαE3L alone (not shown). (C) The Z-DNA-binding site of ZαE3L. In the presence of d(CG)6T4(CG)6 and [Co(NH3)6]3+, the H atoms of ZαE3L that show large averaged chemical shift changes (≥0.082 ppm; bold labels in B) are shown as cyan spheres. H atoms with medium shifts (≥0.068 ppm; italic labels in B) are shown as dark red spheres. Atoms of ZαE3L that show vanishing resonances (underlined labels in B) are shown as dark blue spheres in the 3D structure of ZαE3L. Chemical shifts are listed in Table 2, which is published as supporting information on the PNAS web site. These data indicate a contiguous Z-DNA-binding site made up of helices α3 and α2 and the area around W66.
To investigate the interaction between ZαE3L and d(CG)6T4(CG)6 in the Z-DNA conformation, [Co(NH3)6]3+, which is known to promote the conversion from B- to Z-DNA (17), was added to a final concentration of 300 μM under otherwise identical conditions. The CD spectrum of this sample shows large alterations of the molar ellipticity at 295 and 255 nm, which are characteristic for d(CG)n oligomers in a left-handed Z-DNA conformation (Fig. 2 A, red curve). The corresponding 15N-HSQC NMR spectrum shows a large number of chemical shift changes and three vanishing signals (Fig. 2B) with respect to ZαE3L alone, indicating that ZαE3L interacts with d(CG)6T4(CG)6 in the Z-DNA conformation. The concentration dependence of these chemical shift alterations has been confirmed by an independent chemical shift-mapping experiment at 20 μM ZαE3L with increasing d(CG)6T4(CG)6 concentrations at a constant [Co(NH3)6]3+ to d(CG)6T4(CG)6 ratio of 30 (Fig. 5, which is published as supporting information on the PNAS web site).
The 1H and 15N chemical shift perturbations were quantified and averaged (see Table 2). The atoms showing averaged perturbations ≥0.068 ppm are displayed in the 3D structure of unbound ZαE3L (Fig. 2C). The chemical shift changes map to a contiguous surface encompassing helix α3, the N terminus of helix α2, and the indole proton Hε of W66. The three vanishing signals, which are indicative of intermediate exchange between bound and unbound ZαE3L on the NMR time scale, comprise the backbone amides of Y48 and K45 and the Hδ21 proton of N44 (Figs. (Figs.2B2B and 5). The other side-chain amide proton of N44, Hδ22, does not vanish, suggesting a selective interaction with the Hδ21 proton. The only other shift changes in side-chain amides were observed for the two Hε21 and Hε22 protons of Q35, which are located within a suitable distance for a water-mediated hydrogen bond to the carbonyl of Q31. The backbone amide of Q31 at the N terminus of helix α2 also shows a strong chemical shift alteration. Therefore, the shift changes in the side chain of Q35 probably reflect subtle rearrangements at the N terminus of α2 when Z-DNA is bound rather than direct contact with the Z-DNA. Contacts with the sequential prolines 63 and 64 are not discernible by this mapping method because prolines lack protons bound to nitrogen. Taken together, chemical shift mapping by 15N-HSQC NMR indicates that ZαE3L selectively interacts with DNA in the left-handed Z-conformation through residues in helices α3 and α2 and strand β3, with prominent side-chain perturbations in atoms Hδ21 of N44 and Hε of W66. This result suggests that ZαE3L utilizes a Z-DNA interaction surface that is globally very similar to those of its homologues ZαADAR1 and ZαDLM1 (7, 8).
The Backbone Structure of ZαE3L Is Similar to ZαADAR1 and ZαDLM1. As expected from the primary sequence homology between ZαE3L, ZαADAR1, and ZαDLM1, the three Z-DNA-binding domains share the same α1β1α2α3β2β3 topology. The structure of ZαE3L shows a backbone rms deviation of 1.24 Å to ZαADAR1 and of 1.21 Å to ZαDLM1 (superposition of helices and strands only), indicating that the sequence homology is paralleled by a high overall structural homology. In particular, the three α-helices and three β-strands overlay very well between ZαE3L, ZαADAR1, and ZαDLM1 (Fig. 3A). Structural differences are observed for the loops connecting α1 and β1 (loop 1), α2 and α3 (loop 2), and β2 and β3 (loop 4), of which the latter shows the most marked deviation in its backbone conformation. This finding is not unexpected because loop 4 contains profound differences on the primary sequence level. In ZαDLM1 loop 4 (all subsequent amino acid numbers refer to the homologous residues of ZαE3L) is shorter by two residues than in ZαE3L and ZαADAR1. Moreover, the six residues preceding P63 of loop 4 are poorly conserved between ZαE3L and ZαADAR1. The cis proline 63 of this loop is the sole conserved residue in ZαE3L, ZαADAR1, and ZαDLM1. Proline 63 is of particular importance for the Z-DNA-binding activity because it confers direct Z-DNA contacts in the co-crystal structures of ZαADAR1 (7) and ZαDLM1 (8). Furthermore, the strongest loss of virulence is found when this residue is mutated to alanine in wild-type ZαE3L (4). In the 3D structures of ZαE3L and ZαADAR1, P63 adopts identical positions at the tip of loop 4 although in ZαE3L the entire loop 4 is rotated away from helix α3, resulting in a distance of ≈5.9 Å between the N atoms of P63 of ZαE3L and ZαADAR1. This offset in the interaction surface may be compensated by a subtle adjustment in the binding geometry between ZαE3L and Z-DNA. In conclusion, the overall backbone structure of ZαE3L is very similar to its homologues ZαADAR1 and ZαDLM1, with the exception of loop 4, which shows a displacement that is not expected to markedly affect binding to Z-DNA.
Superposition of the 3D structures of ZαE3L (blue), ZαADAR1 (light green), and ZαDLM1 (gray) in stereo. (A) The secondary structure elements of ZαE3L, ZαADAR1, and ZαDLM1 superimpose well except for the loop containing P63 (side chain labeled). Also, the side chains of K40, R41, N44, K45, and W66, which contact the bound Z-DNA in the co-crystal structures of ZαADAR1 and ZαDLM1, show very similar positions. Only the side chain of Y48 adopts a distinct position in ZαE3L (red) as compared with ZαADAR1 and ZαDLM1.(B) Distinct Y48-W66 distance between ZαE3L and ZαADAR1. Superposition of Y48 and W66 between the 20 lowest energy structures of E3L (blue), unbound ZαADAR1 (red), and bound ZαADAR1 (light green). Whereas Y48 residues are within van-der-Waals distance to W66 in both ZαADAR1 structures, it adopts two solvent exposed rotamer positions in ZαE3L, which are 7.2 and 10.8 Å apart from W66.
Y48 Adopts a Distinct Conformation in ZαE3L. To provide a molecular understanding for the significantly lower affinity of ZαE3L to Z-DNA as compared with ZαADAR1 and ZαDLM1, the structural comparison of the Z-DNA contacting residues between these homologues is of paramount interest. Of the total of eight common Z-DNA-contacting residues in the co-crystal structures of ZαADAR1 and ZαDLM1, the three residues P63, P64, and W66 possess rigid side chains whose conformations are identical between ZαADAR1 and ZαE3L within the limitations of the backbone comparison described above. Of the remaining five Z-DNA contacting residues in helix α3, the side chains of K40, R41, N44, and K45 adopt similar positions in ZαE3L, ZαADAR1, and ZαDLM1 (Fig. 3A). Only the side chain of Y48 is markedly different, showing a distinct solvent exposed conformation in ZαE3L. A closer view of Y48 and W66 in the ensemble of the 20 lowest energy structures of ZαE3L superimposed on ZαADAR1 and ZαDLM1 demonstrates (Fig. 3B) that the phenolic ring of Y48 probably adopts two major rotamer positions, which are ≈7.2 Å and 10.8 Å apart from W66 (distance Y48(Cz) - W66(Cz2)). In contrast, the Y48 - W66 distance in bound (9) and unbound ZαADAR1 (7) measures only 3.9 and 4.5 Å, respectively. The experimental foundation of this marked difference is the observation of several long-range NOEs between the aromatic rings of Y48 and W66 in the NMR structure of unbound ZαADAR1 but none of such NOEs in the 13C-edited NOESY spectra of ZαE3L. Neither does Y48 of ZαE3L show long-range NOEs to other residues. The general sensitivity of the NOESY experiments on ZαE3L is confirmed by the observation of 21 long-range NOEs for the aromatic protons of W66. Therefore, Y48 of ZαE3L adopts two flexible solvent-exposed rotamer positions whereas Y48 of ZαADAR1 is tightly packed against W66 in both the unbound and bound state.
In the co-crystal structures of ZαADAR1 and ZαDLM1, Y48 is the only residue that mediates direct contacts to a base of the bound Z-DNA. In this interaction, the base adopts the syn conformation, which is characteristic for the left-handed Z-conformation of double-stranded DNA. This close interaction geometry suggests that in ZαE3L the solvent-exposed side chain of Y48 rearranges when ZαE3L binds to Z-DNA. By analyzing resolved 1H chemical shifts in 1D1H-NMR spectra of 20 μM ZαE3L with increasing concentrations of d(CG)6T4(CG)/[Co(NH3)6]3+, we found that the aromatic Hδ and Hε atoms of Y48 vanish when ZαE3L binds to Z-DNA. Moreover, the chemical shifts of both methyl groups of L47 (-0.21 and -0.314 ppm in unbound ZαE3L), which are packed in van-der-Waals distance underneath the indole ring of W66, alter in this experiment. These data indicate that the chemical environment around the side chains of L47, Y48, and W66 changes when Z-DNA is bound. Further, the observation of selectively vanishing signals for the HN, Hδ, and Hε of Y48 and the HN of K45 (Fig. 5), which is connected to HN of Y48 through an (i, i + 3) α-helical hydrogen bridge, is in agreement with conformational rearrangements of the Y48 when ZαE3L binds to Z-DNA. The cost in energy for such a rearrangement may account for the substantially lower affinity of ZαE3L to Z-DNA, as compared with ZαADAR1 and ZαDLM1, where Y48 is prepositioned to bind Z-DNA (9).
Mutational experiments suggest that Y48 plays a key role for both binding to Z-DNA in vitro as well as viral pathogenicity in mice. In ZαADAR1, the mutation of Y48 to alanine leads to both a profound loss in Z-DNA affinity and a significant reduction in binding specificity to the Z-conformation of DNA, as evidenced by Biacore and CD spectroscopy (4, 5, 7). In ZαE3L, the mutation of Y48 to alanine abrogates viral pathogenicity in mice by three log10 units (4). It is therefore intriguing to consider Y48 a conformational switch that has to be turned inward toward the protein to enable binding to Z-DNA. In vivo, this switch may be turned on by activating proteins and induced by preformed segments of Z-DNA.
The importance of the tyrosine–Z-DNA interaction is further illustrated by a second domain in ADAR-1 called ZβADAR1. Although it has many sequence similarities to ZαADAR1, it lacks the tyrosine on helix 3, has an isoleucine instead, and shows no in vitro Z-DNA binding (3, 4). When put into vaccinia virus instead of ZαE3L, the chimeric virus shows no pathogenicity (4). However, if a mutant ZβADAR is made, changing isoleucine to tyrosine, it then binds Z-DNA in vitro, and the chimeric virus becomes pathogenic.
A yeast one-hybrid system has been developed in which reporter gene (β-galactosidase) expression depends on binding of a protein to Z-DNA near the promoter (3). The protein is fused to a transcriptional activator domain, which turns on the gene. When ZαE3L is used, the response of the reporter gene is the same as when ZαADAR1 or ZαDLM1 are used (3). As with vaccinia virus infection, this yeast in vivo system shows that ZαE3L is active in binding Z-DNA. The experiments reported here suggest that residue Y48 undergoes a conformational change on binding Z-DNA in vitro. Thus, the change in the tyrosine side chain conformation may act as a switch to turn on in vivo activity.
Acknowledgments
We thank Heidemarie Lerch and Eberhard Krause for matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry and Heike Nikolenko and Michael Bienert for providing a CD spectrometer. We further thank T. D. Goddard, G. Cornilescu and F. Delaglio, and P. Güntert for providing the software sparky 3.105, talos, and dyana 3.1/molmol 2.1–2.6, respectively. This work was supported by grants from the National Institutes of Health (to A.R.).
Notes
Abbreviations: Zα, Z-DNA-binding protein domain; HSQC, heteronuclear single quantum coherence; NOE, nuclear Overhauser enhancement.
Data deposition: The atomic coordinates of the 20 lowest energy structures of a total of 450 have been deposited in the Protein Data Bank, www.pdb.org (PDB ID code 1OYI).