Solution Structure of a DNA Duplex Containing an α-Anomeric Adenosine: Insights into Substrate Recognition by Endonuclease IV

Solution Structure of a DNA Duplex Containing an α-Anomeric Adenosine: Insights into Substrate Recognition by Endonuclease IV

doi:10.1016/j.jmb.2004.02.035 J. Mol. Biol. (2004) 338, 77–91 Solution Structure of a DNA Duplex Containing an a-Anomeric Adenosine: Insights into S...

1MB Sizes 0 Downloads 15 Views

doi:10.1016/j.jmb.2004.02.035

J. Mol. Biol. (2004) 338, 77–91

Solution Structure of a DNA Duplex Containing an a-Anomeric Adenosine: Insights into Substrate Recognition by Endonuclease IV James M. Aramini1, Stephen H. Cleaver1, Richard T. Pon2 Richard P. Cunningham3 and Markus W. Germann1* 1 Departments of Chemistry and Biology, Georgia State University, Atlanta, GA 30303 USA 2

Department of Medical Biochemistry, University of Calgary, Calgary, Alta. Canada T2N 4N1 3 Department of Biological Sciences, SUNY at Albany Albany, NY 12222, USA

The cytotoxic a anomer of adenosine, generated in situ by radicals, must be recognized and repaired to maintain genomic stability. Endonuclease IV (Endo IV), a member of the base excision repair (BER) enzyme family, in addition to acting on abasic sites, has the auxiliary function of removing this mutagenic nucleotide in Escherichia coli. We have employed enzymatic, thermodynamic, and structural studies on DNA duplexes containing a central a-anomeric adenosine residue to characterize the role of DNA structure on recognition and catalysis by Endo IV. The enzyme recognizes and cleaves our aA-containing DNA duplexes at the site of the modification. The NMR solution structure of the DNA decamer duplex establishes that the single a-anomeric adenosine residue is intrahelical and stacks in a reverse Watson– Crick fashion consistent with the slight decrease in thermostability. However, the presence of this lesion confers significant changes to the global duplex conformation, resulting from a kink of the helical axis into the major groove and an opening of the minor groove emanating from the a-anomeric site. Interestingly, the conformation of the flanking base-paired segments is not greatly altered from a B-type conformation. The global structural changes caused by this lesion place the DNA along the conformational path leading to the DNA structure observed in the complex. Thus, it appears that the a-anomeric lesion facilitates recognition by Endo IV. q 2004 Elsevier Ltd. All rights reserved.

*Corresponding author

Keywords: DNA; a-anomeric adenosine; NMR; endonuclease IV

Present address: J. M. Aramini, Center for Advanced Biotechnology and Medicine, Rutgers University, Piscataway, NJ 08854, USA. Abbreviations used: Endo IV, endonuclease IV; CD, circular dichroism; CORMA, complete relaxation matrix analysis; DQF-COSY, double quantum filtered correlated spectroscopy; MARDIGRAS, matrix analysis of relaxation for discerning the geometry of an aqueous structure; MDtar, molecular dynamics using timeaveraged restraints; NMR, nuclear magnetic resonance; NOE, nuclear Overhauser enhancement; NOESY, NOE spectroscopy; ODN, oligodeoxyribonucleotide; RANDMARDI, random error MARDIGRAS; RDC, residual dipolar coupling; rEM, restrained energy minimization; rMD, restrained molecular dynamics; RMSD, root-mean-square deviation; SANDER, simulated annealing with NMR-derived energy restraints; TOCSY, total correlated spectroscopy; WATERGATE, water suppression by gradient tailored excitation. E-mail address of the corresponding author: [email protected]

Introduction Attack from endogenous and exogenous sources including chemical mutagens, oxygen radicals, ionizing radiation and normal cellular metabolism, creates mutagenic and cytotoxic DNA base lesions. The recognition and extrication of these lesions is essential for genome stability and for the perpetuation of life. Nature has devised intricate defense mechanisms, the most prevalent being the DNA base excision repair (BER) pathway that proceeds in two distinct stages.1 – 3 In the first “damagespecific” stage, distinct DNA glycosylases, each recognizing a specific class of damage, detect and flip the damaged base out of the DNA helix into an active-site pocket. Once sequestered, the N – C10 glycosylic bond is cleaved to remove the damaged base, creating the central BER intermediate, an apurine/apyrimidine (AP) site.4,5 The AP intermediate is inherently toxic and must be rapidly

0022-2836/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.

78 removed via the action of an AP endonuclease.2 These so-called “damage general” enzymes cleave the DNA backbone at the AP site, providing a substrate for the completion of repair using a cascade of enzymes including DNA polymerase and ligase.6 The damage-specific DNA glycosylases are grouped into broad categories depending on the type of DNA damage they detect. This diverse class of enzymes locate, recognize and bind damaged DNA bases within a large pool of undamaged DNA most likely via a sliding mechanism.7,8 Crystal structures of several DNAglycosylases show that at least upon lesion detection the DNA substrate is kinked at the base lesion.8 – 11 Generally the DNA both 50 and 30 of the lesion is B form and the kink is localized at the lesion with the bases flipped out extrahelically into enzyme pockets where the DNA base is removed, resulting in the AP site, which is further processed by the AP endonuclease. The damage general AP endonucleases can be grouped into two conserved families typified by exonuclease III (Exo III) and endonuclease IV (Endo IV). Exonuclease III, in Escherichia coli, and its human homolog APE-1 provide the majority of 50 AP activity in each organism.12,13 The crystal structures of these enzymes are highly homologous.14,15 APE-1 appears to bind a flippedout abasic nucleotide resulting in a widened minor groove, a bend of , 358 and a large displacement of the DNA helical axes at the AP site.16,17 The endonuclease IV family including the yeast homolog APN-1 represent the second group of conserved AP endonucleases. The recent X-ray crystal structure of Endo IV free and in complex with a DNA duplex containing an AP site revealed striking local conformational changes within both the protein and DNA upon protein binding.18 Conserved enzyme residues located on two minor groove binding loops within the carboxyl terminal face of the triose phosphate isomerase (TIM) barrel fold project into the AP site, effectively flipping out the abasic sugar and its orphan nucleotide partner. Interestingly, Endo IV has no preference for the base opposite the AP site, and no specific contacts are made to the orphan base. As a result of the amino acid insertion into the base stack the minor ˚ wider than groove is grossly deformed, ca 5.5 A normal B-DNA and the helical axis is severely bent at the site of the lesion. Despite the gross structural changes, stacking interactions with the base-pairs 50 and 30 of the AP site are preserved. Remarkably, Endo IV also functions as a damagespecific enzyme, since it detects and processes a-anomeric adenosine lesions, which are generated by the abstraction of H10 by hydroxyl radicals under anoxic conditions.19 On the basis of this structure, the a-anomers would appear to be able to fit into a solvent-accessible pocket on the enzyme surface that sterically excludes the binding of the normal b-anomer configuration.18 However, to date, the structural basis for the interaction

Structure of a DNA Duplex Containing a-Adenosine

between Endo IV and a-anomeric adenosine sites is lacking. Thermodynamic and structural studies of a-anomeric moieties embedded within duplex DNA and DNA·RNA hybrids by way of 30 -30 and 50 -50 phosphodiester linkages have shown that polarity-reversed a-nucleotides form stable Watson– Crick base-pairs, and impart localized structural and dynamic changes.20 – 24 However, little is known about the precise structural consequences of placing a-anomeric nucleotides within b-anomeric oligodeoxyribonucleotides (ODNs) using normal phosphodiester linkages. A thermodynamic and molecular modeling study revealed that the stability of a DNA duplex containing a single a-anomeric adenosine residue is comparable to its unmodified control, and predicted the formation of a reversed Watson – Crick aA:bT basepair with limited perturbations to the overall structure of the duplex.25 Elucidating the structural consequences of a single a-anomeric adenosine residue within an otherwise b-anomeric DNA duplex has direct relevance to understanding the action of the Endo IV enzyme in the removal of this lesion from the genome. Specifically, our goal is to determine the local and possible global changes to the DNA structure caused by this lesion in the absence of the enzyme and how they might signal and facilitate recognition by the enzyme. Here, we have investigated the Endo IV substrate, thermodynamic, and structural properties of DNA duplexes containing a single a-anomeric adenosine residue. We show that although the a-anomeric nucleotide stacks within the DNA helix, its presence results in significant changes in topology that may aid in the enzyme recognition process.

Results Sequence design The thermodynamic and NMR structural studies were performed on decamer duplexes (control and alphaA) featuring a sequence analogous to that employed in recent studies of the Endo IV complex with AP containing DNA18 (Figure 1). In order to facilitate enzymatic studies with Endo IV, a 19-mer duplex (A19) whose core is based on the decamer sequence was synthesized. This design favors duplex formation at the concentrations and elevated temperatures used in the enzymatic reactions and minimizes unwanted hairpin formation (see below). Endonuclease IV reactivity Varying amounts of Endo IV were incubated with A19 and the reaction mixtures were analyzed on a denaturing polyacrylamide gel (data not shown). The A19 substrate was specifically cleaved by Endo IV at the aA lesion as described for other

79

Structure of a DNA Duplex Containing a-Adenosine

tetrahydrofuran.19 The Km that we determined, 194 nM, is higher than that previously reported and may reflect a difference in sequence context. These observations demonstrate that the core sequence selected for the NMR and thermodynamic studies is capable of supporting enzyme recognition and hydrolysis with normal characteristics and is therefore suitable for structural characterization. Thermodynamic, hyperchromicity, and CD studies

Figure 1. Sequences of the DNA duplexes used in the Endo IV enzymatic experiments (A19) and thermodynamic and structural studies (control and alphaA), respectively. The a-anomeric adenosine residue is shown in outline type. The strand polarity (arrows) and sequence numbering (shown for alphaA) are in the 50 ! 30 direction. The core region of A19 that is identical in sequence with that of the decamers is boxed.

aA-containing DNA substrates.19 The kcat for the cleavage of A19 by Endo IV is 3.1 min21, which is similar to reported values determined for both aA and an AP site-mimetic,

Melting curves and van’t Hoff plots for the control and alphaA duplexes in phosphate buffer containing 500 mM NaCl are shown in Figure 2A and B. It was necessary to resort to a tenfold higher concentration of NaCl compared to that used in our other spectroscopic studies (i.e. [NaCl] ¼ 50 mM), because the van’t Hoff plots for both duplexes are not linear at low salt concentration. This is indicative of melting not proceeding via a purely two-state duplex $ coil mechanism. Moreover, imino 1H NMR spectra of the isolated strands are consistent with their propensity to form hairpins at low salt concentrations (data not shown). In buffer containing 500 mM NaCl, the aA substitution in the decamer sequence results in a significant drop in the Tm of the duplex, although the enthalpy is quite comparable to that of the control

Figure 2. A and B, UV melting curves (CT ¼ 40 mM) and van’t Hoff plots for the control (W) and alphaA (X) decamer duplexes in 10 mM NaPi, 500 mM NaCl, 0.1 mM EDTA (pH 7.0); every tenth point of the melting curves is shown. C and D, UV hyperchromicity and CD curves for the control (W) and aA (X) decamer duplexes in 10 mM NaPi, 50 mM NaCl, 0.1 mM EDTA (pH 7.0); every 30th point is shown.

80

Structure of a DNA Duplex Containing a-Adenosine

Table 1. Thermodynamic data for the control and alphaA DNA duplexes Duplex

DH8 (kJ mol21)

DS8 (kJ mol21 K21)

Tm (K) (CT ¼ 30 mM)

DG8 (kJ mol21) (T ¼ 298 K)

Control AlphaA

304 (27) 272 (19)

0.817 (0.082) 0.737 (0.058)

331.5 (0.1) 325.1 (0.1)

60.4 52.3

Thermodynamic data presented here were obtained under the following buffer conditions: 500 mM NaCl, 10 mM phosphate, 0.1 mM EDTA (pH 7.0). In both cases, final values and uncertainties for DH8; DS8; and Tm at CT ¼ 30 mM were calculated from the van’t Hoff slope and y-intercept obtained by linear regression analysis (OriginGraph 5.0)24 and the following relationship for bimolecular associations of non-self-complementary strands:40 1 R DS8 2 R ln 4 ln CT þ ¼ Tm DH8 DH8

(Table 1). A reduction in thermostability due to this modification is in agreement with the demonstrated disruptive effect of an a-nucleotide within a self-complementary sequence.26 Moreover, on the basis of DG, a DNA nonamer duplex featuring the same aA modification but flanked by purines, was found to be slightly less stable than the control duplex containing a normal bA·T base-pair.25 However, the difference in stability between the control and alphaA duplexes was much smaller in that case, which may be attributed to differing flanking sequences. UV hyperchromicity and circular dichroism (CD) spectra of control and alphaA duplexes in the same buffer (i.e. 50 mM NaCl) used in the NMR data collection are shown in Figure 2C and D. Both techniques suggest that the overall basestacking and local helical conformation of both duplexes are highly comparable, and consistent with a B-type motif.

NMR spectroscopy 1 H and 31P resonance assignments obtained by conventional strategies27 reveal that the presence of an a-anomeric adenosine results in a number of localized spectral perturbations in the alphaA duplex compared to the unmodified control. The most dramatic changes include the apparent absence of an imino 1H signal of the thymidine (T16) opposite aA5 and the significant downfield shift in the 31P signal for the phosphodiester group in the nucleotide (C6) immediately following the a-nucleotide (Figure 3). Note that the imino protons flanking the aA lesion remain sharp and give no evidence of pre-melting at elevated temperatures (data not shown). Numerous protons in this region of the alphaA duplex also experience notable chemical shift perturbations, including the imino protons in the flanking G-C base-pairs (G15 and G17), certain sugar ring protons (i.e. H10 , H20 ,

Figure 3. Imino region of 1 – 1 jump and return 1H NMR spectra (left, 283 K) and 31P spectra (right, 293 K) of the control (1.0 mM) and alphaA (1.0 mM) decamer duplexes in 10 mM NaPi, 50 mM NaCl, 0.1 mM EDTA (pH 6.6). Assignments for the imino protons and the phosphorus atom of C6 are indicated.

81

Structure of a DNA Duplex Containing a-Adenosine

H200 ) of C4, C6, and T16, and changes to the H20 / H200 and H40 resonance positions of aA5 characteristic of a-anomeric deoxyribose moieties.20,21 Similarly, unusually large J2 2 geminal couplings (2 17 Hz) have also been previously observed for a-anomeric nucleotides other DNA duplexes.22 In addition, pseudorotation analysis22,23,28 of vicinal sugar ring spin – spin couplings—(J1 2 , J1 2 , J2 3 , J2 3 ) reveal that, in the limit of a rapid two-state N $ S conformation equilibrium, the mole fraction of S-pucker, fs ; for C6 is somewhat reduced from the high values diagnostic of B-DNA ðfs . 0:7Þ and found for the other non-terminal nucleotides (Table 2). This deviates from the situation for the same residue in the control duplex, whose DQFCOSY cross-peak pattern is qualitatively diagnostic of a normal high S-type sugar (J1 2 . J1 2 , J2 3 extremely small; data not shown). The shifted N $ S equilibrium at C6 compared to the control signifies enhanced local dynamics. However, we do not observe any broad or unusually sharp signals for either 1H or 31P resonances, suggesting that the duplex conformation is relatively static. The analysis of aA5 reveals a high S conformation with a larger pseudorotation angle and lower puckering amplitude than in b anomeric deoxyriboses (Table 2). These observations are similar to our previous results for a anomeric residues in either DNA, or DNA/RNA duplexes with 50 -50 and 30 -30 phosphodiester linkage.21,22,24 The flip in chirality at the C10 position of aA5 results in breaks in the sugar proton to base nuclear Overhauser effect (NOE) pathways and ½di ð6; 8; 20 Þ ! ð½di ð6; 8; 10 Þ ! ds ð10 ; 6; 8Þ 00 ds ð2 ; 6; 8ÞÞ that are normally continuous in regular B-duplexes.27 In addition to the break in the NOE 0 00

0 0

0 00

0 0

0 0

0 00

00 0

0 0

pathway between C4 and aA5 there are numerous unique NOE contacts observed for the aA5 residue. The most salient sequential NOEs are those of H8 to H5 and H6 of C6 as well as a strong intraresidue NOE to H40 . These observations clearly place aA5(H8) in the minor groove and the aA5 base on top of C6 (Figure 4). It is noteworthy that no NOE was observed between aA5(H8) and any C4 base protons. In contrast, we observed an NOE between C4(H6) and aA5(H2); this effectively places H2 in the major groove. Additional NOE contacts were observed for aA5(H2) in H2O solution. The sequential NOEs to the amino groups of C4 and C6 together with the interstrand NOEs to the imino protons of G15 and G17 irrefutably place this base within the helical stack of the duplex. Similarly, NOEs from T17(Me) to G15(H8) and G15(H10 ) as well as NOEs from G17(H8) to T16(H10 ) and T16(H6) show that the partner of aA5 is also stacked. These spectral traits are all reflected in the NMR structure of the alphaA duplex described in the next section. NMR structure of the alphaA DNA duplex The final restrained energy minimized average structure of the alphaA decamer duplex is shown in Figure 5A. The DYANA/AMBER nucleic acid structure determination strategy used here is analogous to methods employed by other groups,29,30 and combines the speed of DYANA for rapid conformational searching and generation of large numbers of structures, with the established force field of AMBER for structure optimization. The relatively tight bundle of structures and the resulting final averaged structure are in very good

Table 2. Coupling constant and pseudorotation data for the alphaA DNA duplex J1 2

J1 2

J2 3

J2 3

J2 2 (^0.5 Hz)

Fm (^18)

P

fS

G1 T2 C3 C4 aA5 C6 G7 A8 C9 G10

6.1 ^ 0.7 9.4 ^ 0.5 8.5 ^ 0.5 8.5 ^ 0.5 7.7 ^ 0.5 6.7 ^ 0.5 10.4 ^ 0.5 10.3 ^ 0.7 8.9 ^ 0.5 8.2 ^ 0.5

6.7 ^ 0.5 5.8 ^ 0.4 6.0 ^ 0.7 6.0 ^ 0.7 ,2 6.5 ^ 0.4 5.5 ^ 0.4 5.6 ^ 0.5 5.8 ^ 0.5 6.0 ^ 0.4

6.2 ^ 0.5 5.8 ^ 0.4 6.0 ^ 0.5 6.0 ^ 0.5 5.7 ^ 0.7 6.5 ^ 0.4 5.6 ^ 0.4 5.7 ^ 0.5 5.8 ^ 0.4 6.2 ^ 0.4

5.0 ^ 1.0 ,2.5 3.0 ^ 1.5 3.0 ^ 1.5 ,2 4.8 ^ 0.7 ,2.5 ,2.5 ,3.0 3.5 ^ 1.0

214.0 214.5 214.5 214.5 217.0 214.5 214.5 214.5 214.0 214.5

36 36 35 35 30 ^ 2 37 37 36 36 36

135– 190 140– 170 135– 180 135– 180 155– 205 120– 150 140– 165 140– 170 140– 175 130– 165

0.44– 0.64 0.82– 0.94 0.72– 0.87 0.72– 0.87 0.84– 1.00 0.51– 0.63 0.90– 1.00 0.88– 1.00 0.76– 0.90 0.68– 0.79

C11 G12 T13 C14 G15 T16 G17 G18 A19 C20

8.3 ^ 0.5 10.0 ^ 0.7 9.4 ^ 0.5 8.5 ^ 0.5 nd 8.9 ^ 0.7 9.1 ^ 0.5 9.1 ^ 0.5 8.5 ^ 0.5 nd

5.9 ^ 0.4 5.6 ^ 0.5 5.8 ^ 0.7 5.9 ^ 0.5 nd 5.8 ^ 0.5 5.7 ^ 0.4 5.7 ^ 0.4 5.9 ^ 0.5 nd

6.0 ^ 0.4 5.6 ^ 0.5 5.8 ^ 0.5 6.0 ^ 0.4 nd 5.8 ^ 0.7 5.8 ^ 0.4 5.7 ^ 0.4 6.0 ^ 0.5 nd

3.5 ^ 1.0 ,2.5 ,2.5 3.0 ^ 1.5 nd ,3.0 ,3.0 ,3.0 3.0 ^ 1.0 nd

214.5 214.0 214.5 214.0 nd 214.5 214.5 214.0 214.5 nd

37 36 36 36 nd 37 37 37 36 nd

135– 165 140– 175 140– 175 135– 165 nd 135– 180 135– 165 140– 170 135– 170 nd

0.68– 0.80 0.85– 1.00 0.82– 0.95 0.72– 0.83 0.76 ^ 0.15 0.74– 0.92 0.77– 0.89 0.77– 0.91 0.72– 0.85 0.69 ^ 0.15

Residue

0 0

0 00

0 0

00 0

0 00

Pseudorotation analyses were performed graphically using contour plots of the individual coupling constants as a function of fS and PS at various F values assuming the following: b-anomeric sugars: PN ¼ 188; FN ¼ FSP ; aA5, PN ¼ 08; FN ¼ FS : Due to 1H 0 chemical shift overlap, fS values for G15 and C20 were estimated P 0 from the sum of J1 2 and J1 2 , ð 1 Þ assuming a two-state equilibrium in which PN ¼ 188; PS ¼ 1538; and FN ¼ FS ¼ 378 (i.e. fS ¼ ½ 1 2 9:3=6:8Þ: nd, not determined. 0 0

0 00

82

Figure 4. View into the minor groove of the core of alphaA showing the interproton distance restraints used in the structure calculations. The a-anomeric adenosine residue is shown in red.

agreement with the NMR data, as judged by the low constraint violations and R x factors (Table 3). All of the spectroscopic properties of the alphaA duplex discussed above are manifest in the final average structure. The global conformation of the molecule is typical of a B-type duplex; it does however, feature a kink in the helical axis exactly at the position of the a-anomeric nucleotide lesion (Figure 5B). Residual dipolar coupling restraints and solvent effects on the structure To further support the presence of the kink at the site of the lesion we employed residual dipolar coupling data for 18 deoxyribose and seven base (three for the aA residue) restraints as well as restrained molecular dynamics (rMD) calculations in explicit solvent (Table 4). Residual dipolar coupling (RDC) data are uniquely suited for the determination of global structural parameters in macromolecules.31,32 The in vacuo NOE structure

Structure of a DNA Duplex Containing a-Adenosine

was refined using the 25 RDC restraints. The RDC refined structure is very similar to the initial structure (all-atom root-mean-square deviation (RMSD) of 0.65) with a total alignment constraint energy violation of only 3.09 kcal/mol. This demonstrates that the measured RDC restraints are fully consistent with the in vacuo NOE structure. We have also recorded RDC data for the control duplex. If the common RDCs measured for the control are forced on the final alphaA duplex, high dipolar alignment violation energy ðEialign ¼ 56:9 kcal=molÞ are observed compared to the corresponding alphaA RDC with the final structure ðEialign ¼ 0:74 kcal=molÞ: In addition, simulations were carried out in an explicit solvent box and for a longer time (Table 5). All structures (Figure 6), whether refined using conventional NMR restraints, or including RDC restraints, either solvated or in vacuum, are structurally very close (Figure 6) and all feature a kink and an enlarged minor groove. Analysis of the aA DNA duplex In all structures, the base of aA5 is indeed stacked within the duplex, but this has some interesting structural consequences. The flip in chirality at C10 results in a significant shift of this base into the minor groove, and alteration of the stacking interactions with flanking bases in the same strand. Specifically, aA5 slides away from C4 and sits directly above C6 (Figure 5C). This may account for the decrease in thermostability observed for the alphaA duplex and the chemical shift perturbations in the vicinity of the modification. At first glance, our structure suggests that aA5 is in the correct orientation to form a reverse Watson– Crick base-pair with T16. However, the aA5(N1) –T16(N3) distance in the in vacuo structure as well as in rMD calculations in water

Figure 5. Final restrained energy minimized average NMR structure of the alphaA decamer duplex. The central aA nucleotide is shown in red, the rest of the a-containing strand is colored green, and the complementary strand is in blue. A, View into the minor groove. B, 908 rotation from A, showing the kink of the helical axis into the major groove. C, View looking down the trinucleotide base-stack encompassing aA5. The phosphorus atom of C6, whose 31P resonance is unusually shifted, is indicated by an arrow.

83

Structure of a DNA Duplex Containing a-Adenosine

Table 3. Statistics of the final AMBER restraints used in the NMR structure determination of the aA DNA decamer and structural parameters for the final ensemble and restrained energy minimized mean structure Restraintsa

Number

Quantitative distance restraints (RANDMARDI): Non-exchangeables (total) 230 Intraresidue 129 Interresidue (sequential) 101 ˚) kwelll (A 0.64 ^ 0.64 Exchangeables (total) 31 ˚ kwelll (A) 1.44 ^ 0.12 Qualitative non-exchangeable 23 distance restraints (S/M/W) Endocyclic torsion angle restraints (n0 to n4) Watson–Crick distance restraints Watson–Crick flat angle restraints Backbone torsion angle restraints (a, b, g, 1, z) Total Average restraints per residue Average RANDMARDI restraints per residue

k 25 25 25 20 25

80

50

25

25

25

10

88

10

502 25.1 13.1

Structural parametersb kEAMBERl kEconstraintl

Ensemble 21031.86 ^ 15.0 64.43 ^ 0.49

Final 21035.5 64.0

Ddav Daav Dnav ˚ Distance violations .0.3 A

0.0450 ^ 0.0003 0.1972 ^ 0.0318 0.1347 ^ 0.0097 6.2 ^ 0.42

0.046 0.207 0.136 6

0.0483 0.0492 0.0485

0.0473 0.0487 0.0477

0.0454 0.0440 0.0450

0.0443 0.0472 0.0451

0.0418 0.0368 0.0402

0.0409 0.0403 0.0407

0.01030 ^ 0.00005 2.389 ^ 0.011

0.0103 2.392

Rx tm ¼ 75 ms Intra (126) Inter (61) Total (187) tm ¼ 150 ms Intra (166) Inter (75) Total (241) tm ¼ 250 ms Intra (196) Inter (100) Total (296) RMSD from ideal geometry Bonds Angles RMSD All atom Heavy atom

0.45 ^ 0.21 0.40 ^ 0.22

a A general description of the structure determination protocol is given in the main text. In the first stage of the protocol, exchangeable distance restraints were applied in a quantitative ˚ ), and the endocyclic deoxyribose torsion fashion (2.0–5.0 A angles for aA5 and C6 as well as the backbone torsion angles from 1 of C4 through to g of C6 were left unconstrained. In subsequent cycles the restraints were refined and augmented to give the final set of restraints described below. AMBER calculations: in SANDER, restraints are applied in the form of flatwell parabolic pseudo-energy terms, where the force constants, k; for the distance and (torsion) angle restraints are in units of ˚ 2) and kcal/(mol rad2), respectively. Average wellkcal/(mol A P width, kwelll ¼ ð lr3 2 r2 lÞ=N; and standard deviations. All vicinal intraresidue sugar interproton distances were omitted from the structure calculations. For the exchangeable distance

˚ ) than in normal is consistently longer (0.2 – 0.3 A Watson– Crick base-pairs. This provides a rationale for our difficulty in observing an imino 1H signal for T16, namely this proton exchanges too rapidly to be detected on the NMR time-scale because it is presumably not base-paired. Further investigations revealed a very broad signal at 11 ppm for low pH and temperature conditions (pH 5.0 – 6.0, 5 –10 8C) that is indicative of an unpaired T imino proton. Analysis of the backbone and local helicoid parameters reveals other perturbations due to the presence of an aA within an otherwise b-anomeric DNA duplex; selected parameters are shown in Figure 7. The majority of the backbone torsion angles experience only small changes at the site of the lesion. For instance the backbone torsion angles

restraints, only RANDMARDI-derived upper bounds for interresidue data were used; lower bounds were defined set to ˚ . Qualitative non-exchangeable distance restraints: S/M, 2.0 A ˚ ; W, 4.0–7.0 A ˚ . Parabolic windows of 2 A ˚ were used 2.0–4.5 A for all distance restraints. Endocyclic deoxyribose ring torsion angle restraints were calculated on the basis of pseudorotation theory using the following empirical ranges of P; F and fS : (1) non-terminal b-anomeric residues except C6 (P ¼ 130 – 1808; F ¼ 368; fS ¼ 1:0): n0 ¼ 224:6ð^11:5Þ8; n1 ¼ 35:3ð^3:0Þ8; n2 ¼ 232:6ð^6:6Þ8; n3 ¼ 17:5ð^13:7Þ8; n4 ¼ 4:4ð^15:6Þ8; (2); aA5 (P ¼ 155 – 2058; F ¼ 308; fS ¼ 1:0): n0 ¼ 29:3ð^12:4Þ8; n1 ¼ n3 ¼ 24:3ð^7:7Þ8; n4 ¼ 24:3ð^7:7Þ8; n2 ¼ 230:0ð^3:0Þ8; 29:3ð^12:4Þ8; (3); C6 (P ¼ 120 – 1558; F ¼ 378; fS ¼ 0:51 – 0:63): n0 ¼ 218:8ð^6:4Þ8; n1 ¼ 11:5ð^5:0Þ8; n2 ¼ 0:2ð^10:5Þ8; n3 ¼ 211:8ð^12:0Þ8; n4 ¼ 18:9ð^8:9Þ8; 108 parabolic windows. Heavyatom Watson–Crick distance restraints and weak linear (170– 1908) angle restraints were applied to all complementary pairs ˚ ; Gexcept aA5:T16: G:C base-pairs: G-O6:C-N4, 2.81–3.01 A ˚ ; G-N2:C-O2, 2.76–2.96 A ˚ ; A:T base-pairs: N1:C-N3, 2.85–3.05 A ˚ ; A-N6:T-O4, 2.85–3.05 A ˚ ; parabolic A-N1:T-N3, 2.72–2.92 A ˚ . Broad right-handed backbone restraints47 windows ¼ 0.5 A were applied throughout the duplex except for z of C4, a,z of aA5, a of C6: a ¼ 290 to 2308, b ¼ 135–2158, g ¼ 30 – 908; 1 ¼ 140 – 3008; z ¼ 150 – 3158; 108 parabolic windows. DYANA calculations: quantitative and qualitative lower and upper distance bounds were identical with those used in AMBER and were augmented with planar Watson–Crick base-pair restraints generated by DYANA and sugar ring closure restraints; total lower and upper distance bounds were 312 and 363, respectively. A total of 153 backbone and sugar ring torsion angle restraints were used: (i) narrow backbone torsion angle restraints for B-DNA:48 a ¼ 246ð^15Þ8; b ¼ 2147ð^15Þ8; g ¼ 36ð^15Þ8; 1 ¼ 155ð^15Þ8; z ¼ 296ð^15Þ8; aA5: b ¼ 135 – 2158; g ¼ 20 – 608; 1 ¼ 140 – 2108; z of C4, a,z of aA5, a of C6 were left unconstrained; (ii) B-type glycosidic torsion angle restraint on all b-anomeric residues (x ¼ 298(^15)8); (iii) n1, n2 and d (n3 þ 1258) bounds were identical to those used in the AMBER calculations. b Definitions of structural parameters: kEAMBERl, average AMBER (non-constraint) energy (kcal/mol); kEconstraintl, average constraint energy (kcal/mol); Ddav ; Daav ; Dnav ; average distance ˚ ), bond angle (deg.), and torsion angle (deg.) deviations from (A the lower or upper bounds; R x, sixth-root CORMA R-factor (intraresidue, interresidue, and total); RMSD, average rootmean-square deviation. Standard deviations for kEAMBERl, kEconstraintl, and RMSD are shown in parentheses. R x-factors were calculated for each NOESY spectrum in 2H2O assuming a the following expression: Rx ¼ tc ¼ 3:2 ns, and using P P lao ðiÞ1=6 2 ac ðiÞ1=6 l= lao ðiÞ1=6 l; where ao and ac represent the experimental and calculated NOE cross-peak volumes, respectively.55 The numbers of NOE volumes used to obtain the R x-factors are shown in parentheses. Pairwise RMSDs were computed within MOLMOL.

84

Structure of a DNA Duplex Containing a-Adenosine

Table 4. Residual dipolar coupling data for rMD/rEM in vacuo calculations on alphaA 1

Residue G1 T2 T2 C4 aA5 aA5 aA5 C6 A8 C9 C9 C9 C10 C10 C11 C11 T13 T13 C14 C14 G15 T16 G17 C20 C20

DCH

DCH calc. (Hz)

C30 –H30 C40 –H40 C6–H6 C10 –H10 C10 –H10 C40 –H40 C2–H2 C40 –H40 C8–H8 C10 –H10 C30 –H30 C6–H6 C10 –H10 C40 –H40 C10 –H10 C40 –H40 C10 –H10 C6–H6 C40 –H40 C6–H6 C8–H8 C10 –H10 C10 –H10 C30 –H30 C40 –H40

21.300 22.745 2.375 20.180 4.927 1.858 6.098 21.390 5.260 2.960 1.814 0.869 2.895 4.913 20.633 20.436 21.214 4.389 21.634 5.018 2.650 4.301 1.522 1.807 0.484

1

1

DCH obs. (Hz)

D1DCH (Hz)

Edip (kcal/mol)

21.340 22.790 2.510 20.120 3.990 1.320 6.500 20.570 5.000 2.780 2.160 0.510 3.010 5.110 20.420 20.620 21.140 4.500 21.390 4.990 2.500 4.170 1.390 2.270 0.000

0.040 0.045 20.135 20.060 0.937 0.538 20.40 20.820 0.260 0.180 20.346 0.359 20.115 20.197 20.213 0.184 20.074 20.111 20.244 0.028 0.150 0.131 0.132 20.463 0.484

0.002 0.002 0.018 0.004 0.877 0.289 0.162 0.672 0.067 0.032 0.119 0.129 0.013 0.039 0.045 0.034 0.005 0.012 0.059 0.001 0.023 0.017 0.018 0.214 0.234

Total 13

1

3.09 13

The 25 residual dipolar C– H couplings (18 sugars and seven bases), determined from F2 coupled high resolution C– 1H HSQC experiments. Coupling constants for aligned and non-aligned samples (0.6 mM DNA duplexes) were obtained from 2D spectra using Sparky and peak fitting. Typically, the S/N ratio in a corresponding F2 slice is 30:1, depending on the coupling partners. We estimate an error of ^0.3 Hz for the RDC data.

for C4 – P – aA5 are essentially unperturbed (a þ 158, b 2 318, g 2 318) compared to the average values in this DNA duplex. The major deviation occurs at aA5 –P – C6, where a modest alteration in 1 (þ 218) and a large change in z (2 658) is obtained while a and b are in the normal range with 2 758 (a2) and 1828 (bt), respectively. Thus, the backbone conformation is (a2, bt, 1t, zt). The observation of normal 3JHP coupling patterns in the 31 P – 1H correlation experiment is indicative of standard b and 1 conformations. Interestingly, only the aA5 – P –C6 31P signal exhibits the large downfield shift. It is well established that torsion angles a and z primarily determine 31P chemical shifts.33 We note that for a zt conformation a significant downfield shift is expected. In addition, Hyperchem TNDO/2 calculations suggest that

ring current effects of aA5 may also contribute to the unusual chemical shift. Stacking the aA within the duplex can be achieved at the expense of only one backbone torsion angle, namely z of aA5, which undergoes a z2 to zt change in conformation. The accommodation of the a-nucleotide also results in a concomitant increase in minor groove width downstream (30 ) of aA5 from that typical of B-DNA. Next, the presence of the a-nucleotide causes a significant increase in roll and decrease in twist for the base-pair steps preceding and following the lesion, respectively. Finally, the deoxyribose ring of C6 is somewhat compressed (reduced amplitude, F) and the glycosidic torsion angle, x, is intermediate between A and B-type duplexes, consistent with the enhanced conformational

Table 5. RMS deviation of alphaA duplex structures

NOE NOE þ water NOE þ RDC

NOE þ water

NOE þ RDC

NOE þ RDC þ water

0.72 – –

0.66 0.70 –

0.69 0.64 0.62

Pairwise, all-atom RMSD. NOE, final in vacuo structure (Figure 5); PDB 1S0T. NOE þ water, equilibration of the in vacuo structure in ˚ between the solute and the edge, including sodium counterions) followed by 1 ns of rMD runs using all a solvent box (minimum 10 A 502 restraints followed by rEM. Using only empirical restraints (i.e. minus backbone and Watson–Crick restraints) the all-atom RMSD ˚ , compared to the in vacuo structure. NOE-RDC, addition of 25 RDC restraints to the in vacuo structure followed by 30 ps of is 1.2 A rMD and rEM; PDB 1S74. NOE þ RDC þ water, equilibration of the NOE-RDC structure in water and sodium ions followed by 1 ns of rMD and rEM; PDB 1S75.

85

Structure of a DNA Duplex Containing a-Adenosine

dynamics of this moiety predicted by the J-coupling data (data not shown). The potential influence of the deoxyribose ring conformation of C6 on the helical kink was further assessed in a series of 100 ps solvated rMD calculations with RDC restraints. In these calculations four different types of restraints were applied. The conformation of the C6 deoxyribose was confined to S, N, or the average conformation predicted by the coupling constants for that residue (S – N) used in previous calculations. In addition, one simulation was run without any torsional restraints. The resulting structures all show similar kink angles that center around 188: S, 18.08; N, 15.28; S –N, 17.98; no torsional restraints, 22.48. In comparison, the average kink angle observed in the 1 ns simulation was 18(^ 3)8. These calculations suggest that the C6 sugar conformation does not affect the kink angle. We note that sugar repuckering is expected to be much faster than helical axis changes, which requires the concerted movement of a large number of atoms.

Discussion Figure 6. Structure comparison: grey, conventional in vacuo NOE structure. Blue, in vacuo NOE structure with RDC restraints. Red, solvated structure with RDC restraints.

With the recent structural insights into DNA repair enzymes, an emerging view of the activity of these enzymes is that their specificity for damaged over undamaged DNA hinges upon recognizing pre-existing deformations and/or deformability in the structure of the substrate.34,35 Both pre-existing deformations and enhanced

Figure 7. Plots of z, minor groove width, roll, and twist for the final rEM average structure of the alphaA decamer duplex. In each case, the average value for B-DNA is shown as a broken line. The minor groove width corresponds ˚ .59 to the minimum interstrand P– P distance minus 5.8 A

86

dynamics or deformability would facilitate damage detection and formation of the recognition complex. As we discuss below, our solution structure of the alphaA DNA decamer containing an a-anomeric adenosine, a lesion that is removed by E. coli AP endonuclease IV is consistent with this hypothesis. On the basis of enzymatic, thermal denaturation, and optical spectroscopic data we conclude that (1) Endo IV recognizes the aA residue in our substrates leading to strand cleavage, (2) the alphaA decamer duplex is slightly less stable than the unmodified control, and (3) it exhibits B-type conformational features with little overall change in base stacking. These conclusions are further supported and extended by the solution structure of alphaA, which reveals that the a-anomeric nucleotide is fully stacked within the duplex, although the base stacking interactions of aA5 with its flanking nucleotides are altered due to the change in chirality at C10 . This alteration in the stacking most likely gives rise to the lower melting temperature observed for this duplex compared to the control duplex. Although these results are generally in line with earlier predictions,25,36 the alphaA structure provides a detailed picture of the local and resulting global structural effects of this modified residue not observed in the modeling studies. Specifically, the stacked a-anomeric adenosine residue results in an 188 kink of the helical axis into the major groove, which is associated with a large increase in roll and decrease in twist in the C4/aA5 and aA5/C6 base-pair steps, respectively. These changes are accompanied by an enlarged minor groove 30 to the modification. Examination of an

Structure of a DNA Duplex Containing a-Adenosine

earlier NMR structure of an AP-DNA duplex (1A9I) reveals that an abasic site may also result in a kink into the major groove.37 While bends and kinks are notoriously difficult to accurately determine in nucleic acids using conventional NMR restraints, the kink in the alphaA structure is driven by numerous local restraints, in particular NOE contacts, at the site of the lesion. Further evidence that independently supports a helical axis kink and the global conformation of this duplex is obtained from the in vacuo rMD calculations with RDC restraints and solvated rMD calculations with and without RDC restraints. All structures generated under these various conditions show minimal drift from the final in vacuo rMD structure. It is interesting to note that even though the lesion and the perturbations caused by it are local, they change the global appearance of the DNA. This suggests that a local change can give rise to a structural marker for repair enzyme recognition. In this case, the kinked helical axis and rather enlarged minor groove around the lesion in the alphaA structure have significant ramifications for its recognition by Endo IV. Both of these features, which we observe for the free duplex, are also present in the crystal structure of an apurinic DNA substrate with Endo IV (1QUM).18 The kink of the DNA into the major groove at the point of the lesion is 188 for alphaA compared to 678 for 1QUM (Figure 8, top). The widening of the minor groove due to the lesion follows a similar trend: 1QUM . alphaA . Arnott B-DNA (Figure 8, bottom). Helical parameter analysis of the 1QUM and alphaA reveal that the base-paired segments above and below the lesions retain B-like traits, particularly reflected by x-displacement, rise, and

Figure 8. Ribbon (top) and Connelly surface (bottom) representations of three DNA duplexes: (left) AP-containing DNA duplex in complex with Endo IV (PDB ID 1QUM;18 (center) the alphaA decamer duplex; (right) Arnott B-DNA model. In the ribbons, the sugar-phosphate backbones are shown is grey, purines are red, pyrimidines are blue and the helical axes are yellow; the cleaved abasic residue in 1QUM is highlighted in wireframe. In the Connelly surface images, the view into the minor groove (m) of each duplex is shown. Notice the hole in the 1QUM structure due to the extruded abasic residue and its orphaned complementary nucleotide. The helical axis kinks for 1QUM and alphaA were obtained by CURVES fits of the base-paired segments above and below the lesions.

87

Structure of a DNA Duplex Containing a-Adenosine

sugar puckering. This is again similar to the flanking segments of the AP-DNA:Endo IV complex where the flanking segments also retain a B-type conformation. In addition, the unique local structural and dynamic properties in the alphaA decamer structure, namely the conformational flexibility of the C6 deoxyribose and the altered phosphodiester backbone, are also expected to aid damage recognition. From this evidence, we postulate that the alphaA structure enhances enzyme recognition by the distortions that are already on the path towards their values in the subsequent complex. Thus, the kink and enlarged minor groove facilitate enzymatic access by both reducing the energetic cost of forming the initial distortion and further reducing the energetic cost of driving the nucleic acid into its final conformation. This, combined with the fact that an a-anomeric lesion can be sterically accommodated when modeled into the 1QUM structure (data not shown), accounts for the additional specificity of this enzyme for aA lesions. We further speculate that the flanking sequences can affect the extent of these distortions and thereby influence recognition by Endo IV. This hypothesis is currently under investigation in our laboratories.

Materials and Methods Oligonucleotide synthesis and sample preparation Aa-nomeric adenosine was purchased from RI Chemicals. Solid-phase syntheses of the decamer and 19-mer strands used here were performed via standard phosphoramidite chemistry on an Applied Biosystems 380B DNA synthesizer. The deprotected ODNs were purified by FLPC on a 15Q PE anion-exchange column and subsequently desalted by gel-filtration (G-10). Strands used in the enzymatic assays were further purified by preparative 15% PAGE followed by gel-filtration. The purity of each ODN was confirmed by 1H NMR spectroscopy prior to use. Control and alphaA DNA duplexes were obtained by combining equimolar amounts of the appropriate strands, whose stock solution concentrations were determined spectrophotometrically (A260 nm at 85 8C) using molar extinction coefficients (1260, M21 cm21) calculated from literature values for the constituent mononucleosides:38 d(GTCCAC GACG), 103,600; d(CGTCGTGGAC), 101,000. Optical (UV and CD) and NMR samples were prepared in 10 mM phosphate buffer, pH 6.5– 7.0, containing 50 mM NaCl, 0.1 mM EDTA; the NaCl concentration was raised to 500 mM in the thermal denaturation experiments. NMR samples were exchanged into either 90% H2O/ 10% 2H2O or 99.996% 2H2O (Cambridge Isotopes). Phage pf1 was purchased from Asla and prepared following the manufacturer’s instructions. Briefly, the phage was exchanged into the 2H2O NMR buffer through a series of ultracentrifugation washing steps. Samples for residual dipolar coupling measurements were prepared by adding pf1 to the existing NMR samples. The phage concentration was adjusted 10.3 mg/ml, resulting in a deuterium splitting of 12.9 Hz at 293 K.

Enzymatic studies Procedures for the purification of E. coli Endo IV and assays using this enzyme are described elsewhere.39 Briefly, the oligonucleotide containing the aA residue was radioactively labeled with [32P]ATP from New England Nuclear using T4 polynucleotide ligase from New England Biolabs according to the supplier’s directions and annealed with a slight excess of the complementary strand. The DNA was incubated with E. coli Endo IV in 50 mM Mops– NaOH (pH 8.0), 200 mM NaCl and 1 mM b-mercaptoethanol (BME) for 30 minutes at 28 8C. Reactions were terminated by the addition of 2 £ formamide buffer and the reaction products were electrophoresed on a denaturing 15% (w/v) polyacrylamide gel. The radioactivity of the cleaved and uncleaved oligonucleotides was visualized and quantified on a Molecular Dynamics Storm phosphorimager. Kinetic parameters were determined for reactions carried out under steady-state conditions and analyzed using the Michaelis – Menten equation with the program GraFit 4.0 (Erithacus Software). UV thermal denaturation and CD studies UV thermal denaturation experiments on the control and alphaA DNA duplexes were performed at l ¼ 260 nm on a thermostatted multi-cell Varian CARY 3E spectrophotometer. Melting profiles for total strand concentrations ðCT Þ ranging from 10 mM to 80 mM were acquired at 0.4 deg. C/minute, with a two seconds signal averaging time, and 0.2 deg. C data collection interval. Assuming a two-state helix $ coil transition, melting temperatures ðTm Þ were calculated from three melting curves at each concentration using a six parameter fitting program.20,21 The enthalpy ðDH8Þ and entropy ðDS8Þ of the transitions were subsequently calculated from the linear van’t Hoff relationship between reciprocal Tm and ln CT :40 Hyperchomicity profiles were computed from wavelength scans (l ¼ 330 ! 220 nm; 100 nm/minute) acquired on both the coil (85 8C) and duplex (5 8C) forms. Circular dichroism spectra were recorded on a Jasco J-810 spectropolarimeter at room temperature. For each spectrum, four scans (l ¼ 330 ! 200 nm; 50 nm/ minute) were accumulated. NMR spectroscopy The majority of NMR experiments were acquired on a Bruker AMX600 NMR spectrometer using 5 mm IDTG600 triple resonance (Nalorac Corp.) and broadband inverse (Bruker) probeheads. F2 coupled 13C– 1H HSQC experiments were recorded on a Bruker Avance 500 MHz instrument equipped with a triple resonance cryoprobe. The array of 2D experiments used to obtain 1 H and 31P resonance assignments of both duplexes and in the structure determination of alphaA: NOESY (tm ¼ 75; 150, and 250 ms; ten seconds relaxation delay), DQF-COSY, TOCSY, 31P– 1H correlation60 in 2H2O at 293 K and WATERGATE41 NOESY (tm ¼ 150 ms; four seconds relaxation delay) in H2O at 283 K, were acquired and processed under identical conditions using standard parameters that we have described in earlier studies.20 – 22,24 The program SPARKY 3.74 (UCSF) was employed for all 2D spectral manipulation and peak assigning, as well as NOESY cross-peak integration. A rotational correlation time ðtc Þ of 3.2 ns for the alphaA duplex, required for MARDIGRAS and CORMA

88

Structure of a DNA Duplex Containing a-Adenosine

calculations, was determined by the truncated driven NOE method.42 1H and 31P spectra were referenced to internal 2,2-dimethylsilapentane-5-sulfonate (DSS) and external 85% H3PO4 (capillary in 2H2O). Structural restraints The final restraints employed in the DYANA and AMBER structure calculations (vide infra) are detailed in Table 3. Quantitative interproton distance restraints for non-exchangeable (2H2O data) and exchangeable (H2O data) protons were calculated from the four NOESY data sets using the complete relaxation matrix approach RANDMARDI (MARDIGRAS 3.2),43 – 45 as described in our recent structure determinations.22,24 Briefly, at each stage of the structure refinement, 30 RANDMARDI cycles were performed on each NOESY data set as a function of rotational correlation time (tc ¼ 2:0; 3.2, and 4.5 ns) and input structure(s). Upper distance bounds involving exchangeable protons were derived from RANDMARDI calculations on the H2O data assuming no exchange and with the non-exchangeable distances fixed to the average values computed from the 2H2O data. For each quantitative distance restraint, the final lower (non-exchangeable only) and upper bounds represent either the average of the lower and upper bounds from all RANDMARDI runs on appropriate NOESY data set(s) or the average ^ their standard deviations, depending on the success rate of the program at finding a solution. All vicinal (intraresidue) sugar ring interproton distance restraints were discarded but their corresponding NOE cross-peaks were included in CORMA calculations. Qualitative non-exchangeable interproton distance restraints were included for interresidue cross-peaks whose volumes could not be accurately determined or were problematic in the RANDMARDI calculations. Endocyclic deoxyribose ring torsion angle restraints (n0 to n4) for non-terminal residues were determined by pseudorotation analysis.28 of vicinal 1 H– 1H J-coupling constants, J1 2 , J1 2 , J2 3 , J2 3 , obtained from SPHINX/LINSHA simulations46 of DQF-COSY cross-peaks as described in our earlier work.22 – 24 For all high S-type sugars (all non-terminals except C6), bounds for n0 to n4 were calculated assuming a mole fraction of S-pucker ðfS Þ equal to 1 and a broader pseudorotation phase angle ðPÞ range than that predicted by pseudorotation analysis. In the case of C6, a near 50:50 sugar, fS was incorporated into the calculations.22,47 Watson – Crick hydrogen bond restraints were applied to all base-pairs in the duplex except aA5:T16, on the basis of imino 1H NMR data. Based on the qualitative analysis of salient cross-peaks in the 31P – 1H, DQF-COSY, and NOESY,48,49 broad right-handed backbone restraints47 were applied to all linkages, with the exception of the a,z torsion angles flanking aA5. For the RDC restraints, one bond 13C – 1H couplings were determined from the difference in 1JCH values from F2 coupled HSQC experiments on the aligned and unaligned alphaA and control duplexes, each at a concentration of 0.6 mM. 0 0

0 00

0 0

00 0

Structure determination protocol In vacuo structure calculations Structure calculations were performed using the torsion angle dynamics program DYANA50,51 and the SANDER program within AMBER 6.0.52 DYANA library and AMBER coordinate and topology files for the aA

residue were constructed using Chem3D (Cambridge Software) and LEaP, respectively. All DYANA calculations were performed using a standard simulated annealing protocol consisting of 4000 TAD steps, of which the first 800 were at high temperature followed by slow cooling over the remainder of the run, and 1000 steps of conjugate gradient energy minimization. The SANDER rMD and rEM calculations were conducted in vacuo using the 1994 all atom parametrization with the charge on each phosphate group attenuated to 20.2,47 ˚ cutdistance-dependent dielectric constant ð1 ¼ rÞ; 8 A off for non-bonded interactions, and SHAKE was applied only to bonds involving hydrogen atoms in the rMD runs.22,24 Arnott A and B-form alphaA duplex starting models, used to produce initial non-exchangeable interproton distances by RANDMARDI, were generated using NUCGEN and LEaP and energy minimized (2000 steps steepest descent) with SANDER. The structure of the alphaA duplex was elucidated using a protocol consisting of three cycles of RANDMARDI distance refinement and DYANA structure calculations, followed by rMD and rEM in AMBER. In each cycle of the refinement, the final ensemble of structures and the average structure were obtained by rEM (200 steps of steepest descent followed by 1800 steps conjugate gradient) of the best ten out of 100 DYANA structures, selected on the basis of target function, followed by coordinate averaging in MOLMOL, and rEM on the resulting mean structure. After each cycle, the final average structure was used as input for RANDMARDI for the refinement of quantitative interproton distance restraints. When no further improvement in CORMA R x factors (described below) was observed, the best ten DYANA structures from the final cycle were further refined by 100 ps rMD at 300 K followed by coordinate averaging of snapshots collected over the last 5 ps, and rEM to yield the final ensemble of alphaA structures. From this, the final average in vacuo structure of the alphaA duplex was obtained by coordinate averaging of the final ten structures followed by rEM.

Refinement with residual dipolar coupling restraints The final in vacuo structure was further refined with the addition of RDC restraints using AMBER 6.1. The initial alignment tensor was obtained using a 4000 step restrained energy minimization against the RDC restraints keeping the DNA structure fixed. This initial alignment tensor was then used in conjunction with the RDC and all previous structural restraints for further refinement using rMD. The RDC restraints were applied to the starting structure via a six-step protocol. In the initial step, the RDC restraints were given a very low weight (1.0 £ 1025) and the temperature was ramped to 400 K over the first 3 ps of the 4 ps run. Then in a series of four 2 ps equilibration runs, the weight of the RDC restraints was increased to 0.6 (0.01, 0.1, 0.3 and 0.6). Finally, in the last rMD step the weight of the RDC restraints was set to 1.0, and the system was allowed to evolve for 18 ps, during which the temperature was ramped back down to 0 over the last 10 ps. At each step in this process the last alignment tensor for a particular run was used as the starting alignment tensor for the next run. The final in vacuo RDC structure was generated by restrained energy minimization of the final rMD structure.

Structure of a DNA Duplex Containing a-Adenosine

Solvated calculations The final in vacuo structures, with and without RDC restraints, were equilibrated in a solvent box that included sodium counterions.61 One nanosecond solvated rMD runs with particle mesh Ewald treatment of the electrostatic interactions were set up using (a) all 502 distance and angle restraints and (b) all 502 distance and angle restraints with 25 additional RDC restraints and (c) empirical restraints only (i.e. minus the backbone and Watson– Crick restraints). The final solvated structure was generated by averaging structures from the last 5 ps, followed by restrained energy minimization. Structure analysis Sixth-root R x factors for the final average alphaA structure and ensemble were calculated using the CORMA 5.2 program;53 – 55 R x factors serve as markers of structure quality and represent the goodness-of-fit between the experimental and calculated (model-based) NOE cross-peak volumes. Backbone torsion angles, deoxyribose ring and local helical parameters were calculated using CURVES 5.156 and fitparam.57 Molecular graphics Figures were prepared using WebLab Viewer (Molecular Simulations Inc.) and MOLMOL.58 Atomic cordinates Structural coordinates have been deposited as follows: PDB ID 1S0T, RCSB ID RCSB021223; in vacuo NMR structure. PDB ID 1S74, RCSB ID RCSB021450; in vacuo with RDC restraints. PDB ID 1S75, RCSB ID RCSB021451 solvated with RDC restraints.

Acknowledgements We thank Dr Nick Ulyanov (University of California, San Francisco) for providing the analysis program fitparam and helpful discussions. We thank Dr Anthony Mazurek for suggestions and Drs Deltlef Moskau and Christian Richter of Bruker-Biospin for help with the initial RDC measurements. This work was supported by grants from NIH and the Georgia Cancer Coalition (MWG).

References 1. Lindahl, T. (1993). Recovery of antediluvian DNA. Nature, 365, 700. 2. Mol, C. D., Parikh, S. S., Putnam, C. D., Lo, T. P. & Tainer, J. A. (1999). DNA repair mechanisms for the recognition and removal of damaged DNA bases. Annu. Rev. Biophys. Biomol. Struct. 28, 101– 128. 3. McCullough, A. K., Dodson, M. L. & Lloyd, R. S. (1999). Initiation of base excision repair: glycosylase mechanisms and structures. Annu. Rev. Biochem. 68, 255–285. 4. Cunningham, R. P. (1997). DNA glycosylases. Mutat. Res. 383, 189– 196. 5. Krokan, H. E., Standal, R. & Slupphaug, G. (1997). DNA glycosylases in the base excision repair of DNA. Biochem. J. 325, 1 – 16.

89

6. Hosfield, D. J., Daniels, D. S., Mol, C. D., Putman, C. D., Parikh, S. S. & Tainer, J. A. (2001). DNA damage recognition and repair pathway coordination revealed by the structural biochemistry of DNA repair enzymes. Prog. Nucleic Acid Res. Mol. Biol. 68, 315–347. 7. Verdine, G. L. & Bruner, S. D. (1997). How do DNA repair proteins locate damaged bases in the genome? Chem. Biol. 4, 329– 334. 8. Chen, L., Haushalter, K. A., Lieber, C. M. & Verdine, G. L. (2002). Direct visualization of a DNA glycosylase searching for damage. Chem. Biol. 9, 345– 350. 9. Parikh, S. S., Mol, C. D., Slupphaug, G., Bharati, S., Krokan, H. E. & Tainer, J. A. (1998). Base excision repair initiation revealed by crystal structures and binding kinetics of human uracil-DNA glycosylase with DNA. EMBO J. 17, 5214– 5226. 10. Barrett, T. E., Scharer, O. D., Savva, R., Brown, T., Jiricny, J., Verdine, G. L. & Pearl, L. H. (1999). Crystal structure of a thwarted mismatch glycosylase DNA repair complex. EMBO J. 18, 6599– 6609. 11. Bruner, S. D., Norman, D. P. & Verdine, G. L. (2000). Structural basis for recognition and repair of the endogenous mutagen 8-oxoguanine in DNA. Nature, 403, 859– 866. 12. Saporito, S. M., Smith-White, B. J. & Cunningham, R. P. (1988). Nucleotide sequence of the xth gene of Escherichia coli K-12. J. Bacteriol. 170, 4542– 4547. 13. Demple, B., Herman, T. & Chen, D. S. (1991). Cloning and expression of APE, the cDNA encoding the major human apurinic endonuclease: definition of a family of DNA repair enzymes. Proc. Natl Acad. Sci. USA, 88, 11450– 11454. 14. Gorman, M. A., Morera, S., Rothwell, D. G., deLa Fortelle, E., Mol, C. D., Tainer, J. A. et al. (1997). The crystal structure of the human DNA repair endonuclease HAP1 suggests the recognition of extrahelical deoxyribose at DNA abasic sites. EMBO J. 16, 6548– 6558. 15. Mol, C. D., Kuo, C.-F., Thayer, M. M., Cunningham, R. P. & Tainer, J. A. (1995). Structure and function of the multifunctional DNA-repair enzyme exonuclease III. Nature, 374, 381– 386. 16. Mol, C. D., Hosfield, D. J. & Tainer, J. A. (2000). Abasic site recognition by two apurinic/apyrimidinic endonuclease families in DNA base excision repair: the 30 -ends justify the means. Mutat. Res. 460, 211 – 229. 17. Mol, C. D., Izumi, T., Mitra, S. & Tainer, J. A. (2000). DNA-bound structures and mutants reveal abasic DNA binding by APE1 and DNA repair coordination. Nature, 403, 451–456. 18. Hosfield, D. J., Guan, Y., Haas, B. J., Cunningham, R. P. & Tainer, J. A. (1999). Structure of the DNA repair enzyme endonuclease IV and its DNA complex: double-nucleotide flipping at abasic sites and three-metal-ion catalysis. Cell, 98, 397– 408. 19. Ide, H., Tedzuka, K., Shimizu, H., Kimura, Y., Purmal, A. A., Wallace, S. S. & Kow, Y. W. (1994). Alpha-deoxyadenosine, a major anoxic radiolysis product of adenine in DNA, is a substrate for Escherichia coli endonuclease IV. Biochemistry, 33, 7842 –7847. 20. Aramini, J. M., Kalisch, B. W., Pon, R. T., van de Sande, J. H. & Germann, M. W. (1996). Structure of a DNA duplex that contains alpha-anomeric nucleotides and 30 -30 and 50 -50 phosphodiester linkages: coexistence of parallel and antiparallel DNA. Biochemistry, 35, 9355–9365.

90

21. Aramini, J. M., van de Sande, J. H. & Germann, M. W. (1997). Spectroscopic and thermodynamic studies of DNA duplexes containing alpha-anomeric C, A, and G nucleotides and polarity reversals: coexistence of localized parallel and antiparallel DNA. Biochemistry, 36, 9715– 9725. 22. Aramini, J. M., Mujeeb, A. & Germann, M. W. (1998). NMR solution structures of [d(GCGAAT-30 -30 alphaT-50 -50 -CGC)2] and its unmodified control. Nucl. Acids Res. 26, 5644–5654. 23. Aramini, J. M., Mujeeb, A., Ulyanov, N. B. & Germann, M. W. (2000). Conformational dynamics in mixed alpha/beta-oligonucleotides containing polarity reversals: a molecular dynamics study using time-averaged restraints. J. Biomol. NMR, 18, 287– 302. 24. Aramini, J. M. & Germann, M. W. (1999). Solution structure of a DNA·RNA hybrid containing an alpha-anomeric thymidine and polarity reversals: d(ATGG-30 -30 -alphaT-50 -50 -GCTC)·r(gagcaccau). Biochemistry, 38, 15448– 15458. 25. Ide, H., Shimizu, H., Kimura, Y., Sakamoto, S., Makino, K., Glackin, M. et al. (1995). Influence of alpha-deoxyadenosine on the stability and structure of DNA. Thermodynamic and molecular mechanics studies. Biochemistry, 34, 6947–6955. 26. Aramini, J. M., van de Sande, J. H. & Germann, M. W. (1998). Structure and stability of DNA containing inverted anomeric centers and polarity reversals. In ACS Symposium Series 682 (Leontis, N. B. & Santa Lucia, J. Jr, eds), pp. 92 – 105, American Chemical Society, Washington DC. 27. Wu¨thrich, K. (1986). NMR of Proteins and Nucleic Acids, Wiley, New York. 28. van Wijk, J., Huckriede, B. D., Ippel, J. H. & Altona, C. (1992). Furanose sugar conformations in DNA from NMR coupling constants. Methods Enzymol. 211, 286– 306. 29. Bertini, I., Clemente, A., Rombeck, I., Rosato, A., Turano, P., Lippert, B. & Quadrifoglio, F. (1999). Three-dimensional solution structures of two DNA dodecamers through full relaxation matrix analysis. Magn. Reson. Chem. 37, 564– 572. 30. Mujeeb, A., Parslow, T. G., Zarrinpar, A., Das, C. & James, T. L. (1999). NMR structure of the mature dimer initiation complex of HIV-1 genomic RNA. FEBS Letters, 458, 387– 392. 31. Tjandra, N., Tate, S., Ono, A., Kainosho, M. & Bax, A. (2000). The NMR structure of a DNA dodecamer in an aqueous dilute liquid crystalline phase. J. Am. Chem. Soc. 122, 6190– 6200. 32. MacDonald, D., Herbert, K., Zhang, X., Pologruto, T. & Lu, P. (2001). Solution structure of an A-tract DNA bend. J. Mol. Biol. 306, 1081– 1098. 33. Gorenstein, D. G. (1994). Conformation and Dynamics of DNA and Protein-DNA complexes by 31 P NMR. Chem. Rev. 94, 1315– 1338. 34. Rachofsky, E. L., Seibert, E., Stivers, J. T., Osman, R. & Ross, J. B. (2001). Conformation and dynamics of abasic sites in DNA investigated by time-resolved fluorescence of 2-aminopurine. Biochemistry, 40, 957– 967. 35. Isaacs, R. J., Rayens, W. S. & Spielmann, H. P. (2002). Structural differences in the NOE-derived structure of G-T mismatch DNA relative to normal DNA are correlated with differences in the 13C relaxationbased internal dynamics. J. Mol. Biol. 319, 191– 207. 36. Bielecki, L. & Adamiak, R. W. (2001). Structure and dynamics of a DNA duplex containing single alpha-

Structure of a DNA Duplex Containing a-Adenosine

37. 38. 39.

40.

41.

42. 43. 44.

45.

46.

47.

48.

49.

50.

51. 52.

53.

anomeric deoxyadenosine residue. Acta Biochim. Pol. 48, 103– 111. Beger, R. D. & Bolton, P. H. (1998). Structures of apurinic and apyrimidinic sites in duplex DNAs. J. Biol. Chem. 273, 15565 –15573. Fasman, G. D. (1975). Editor of Handbook of Biochemistry and Molecular Biology. Nucleic Acids, 3rd edit., vol. 1, CRC Press, Boca Raton, FL. Haas, B. J., Sandigursky, M., Tainer, J. A., Franklin, W. A. & Cunningham, R. P. (1999). Purification and characterization of Thermotoga maritima endonuclease IV, a thermostable apurinic/apyrimidinic endonuclease and 30 -repair diesterase. J. Bacteriol. 181, 2834– 2839. Breslauer, K. J. (1995). Extracting thermodynamic data from equilibrium melting curves for oligonucleotide order– disorder transitions. Methods Enzymol. 259, 221– 242. Piotto, M., Saudek, V. & Sklena´r, V. (1992). Gradienttailored excitation for single-quantum NMR spectroscopy of aqueous solutions. J. Biomol. NMR, 2, 661– 665. Lane, A. N. (1995). Determination of fast dynamics of nucleic acids by NMR. Methods Enzymol. 261, 413– 435. Borgias, B. A. & James, T. L. (1989). Two-dimensional nuclear Overhauser effect: complete relaxation matrix analysis. Methods Enzymol. 176, 169–183. Borgias, B. A. & James, T. L. (1990). MARDIGRAS: a procedure for matrix analysis of relaxation for discerning geometry of an aqueous structure. J. Magn. Reson. 87, 475– 487. Liu, H., Spielmann, H. P., Ulyanov, N. B., Wemmer, D. E. & James, T. L. (1995). Interproton distance bounds from 2D NOE intensities: effect of experimental noise and peak integration errors. J. Biomol. NMR, 6, 390–402. Widmer, H. & Wu¨thrich, K. (1987). Simulated twodimensional NMR cross peak fine structures for 1H spin systems in polypeptides and polydeoxynucleotides. J. Magn. Reson. 74, 316– 336. Mujeeb, A., Kerwin, S. M., Kenyon, G. L. & James, T. L. (1993). Solution structure of a conserved DNA sequence from the HIV-1 genome: restrained molecular dynamics simulation with distance and torsion angle restraints derived from two-dimensional NMR spectra. Biochemistry, 32, 13419– 13431. Wijmenga, S. S., Mooren, M. M. W. & Hilbers, C. W. (1993). NMR of nucleic acids; from spectrum to structure. In NMR of Macromolecules; A Practical Approach (Roberts, G. C. K., ed.), pp. 217– 288, Oxford University Press, New York. Pikkemaat, J. A. & Altona, C. (1996). Fine structure of the P –H50 cross-peak in 31P– 1H correlated 2D NMR spectroscopy. An efficient probe for the backbone torsion angles b and g in nucleic acids. Magn. Reson. Chem. 34, S33– S39. Gu¨ntert, P., Mumenthaler, C. & Wu¨thrich, K. (1997). Torsion angle dynamics for NMR structure calculation with the new program DYANA. J. Mol. Biol. 273, 283– 298. Gu¨ntert, P. (1998). Structure calculation of biological macromolecules from NMR data. Quart Rev. Biophys. 31, 145– 237. Case, D. A., Pearlman, D. A., Caldwell, J. W., Cheatham, T. E. III, Ross, W. S., Simmerling, C. L., et al. (1999). AMBER 6, University of California, San Francisco. Keepers, J. W. & James, T. L. (1984). Two-dimensional

91

Structure of a DNA Duplex Containing a-Adenosine

54.

55.

56. 57.

nuclear overhauser effect spectra. J. Magn. Reson. 57, 404–426. Borgias, B. A. & James, T. L. (1988). COMATOSE, a method for constrained refinement of macromolecular structure based on two-dimensional nuclear Overhauser effect spectra. J. Magn. Reson. 79, 493– 512. Thomas, P. D., Basus, V. J. & James, T. L. (1991). Protein solution structure determination using distances from two-dimensional nuclear Overhauser effect experiments: effect of approximations on the accuracy of derived structures. Proc. Natl Acad. Sci. USA, 88, 1237– 1241. Lavery, R. & Sklenar, H. (1996). CURVES 5.1. Helical Analysis of Irregular Nucleic Acids, Laboratoire de Biochimie Theorique CNRS, Paris, France. Ulyanov, N. B. & James, T. L. (1995). Statistical

58. 59.

60.

61.

analysis of DNA duplex structural features. Methods Enzymol. 261, 90 – 120. Koradi, R., Billeter, M. & Wuthrich, K. (1996). MOLMOL: a program for display and analysis of macromolecular structures. J. Mol. Graph. 14, 51 – 55. Bhattacharyya, D. & Bansal, M. (1992). Groove width and depth of B-DNA structures depend on local variation in slide. J. Biomol. Struct. Dynam. 10, 213 –226. Sklena´r, V., Miyashiro, H., Zon, G., Miles, T. & Bax, A. (1986). Assignment of the 31P and 1H resonances in oligonucleotides by two-dimensional NMR spectroscopy. FEBS Letters, 208, 94 – 98. Cheatham, T. E., III & Kollman, P. A. (1996). Observation of the A-DNA to B-DNA transition during unrestrained molecular dynamics in aqueous solution. J. Mol. Biol. 259, 434– 444.

Edited by M. F. Summers (Received 28 October 2003; received in revised form 22 January 2004; accepted 13 February 2004)