Chem. Res. Toxicol. 1995, 8, 821-832
821
Articles Solution Structure of an Oligodeoxynucleotide Containing the Human n-rus Codon 61 Sequence Refined from lH NMR Using Molecular Dynamics Restrained by Nuclear Overhauser Effects Binbin Feng and Michael P. Stone* Department of Chemistry and Center in Molecular Toxicology, Vanderbilt University, Nashville, Tennessee 37235 Received December 12, 1994@
The solution structure of the rus6l oligodeoxynucleotide duplex d(CGGACAAGAAG). d(CTTCTTGTCCG),which consists of codons 60, 61 (underlined), and 62 of the human n-ras protooncogene, was refined from lH NMR data. The sequence contains a run of purines in the coding strand, with one R-Y step, A4*T19-C5G18, and one Y-R step, C5G18+A6*T17 (excluding the 5’4erminal base pair). The NMR data were consistent with a B-like helix as judged by characteristic internucleotide NOES. The NOE intensities between purine H8 and purine anomeric protons were small as compared to the intensities between cytosine H5 and H6 protons, indicative of glycosyl torsion angles in the anti range. Cross-peaks were observed between purine H8 and pyrimidine H5 and CH3 protons on adjacent bases in the direction of purine (5’-3’) pyrimidine, but not in the direction pyrimidine (5’43’) purine. Watson-Crick hydrogen bonding was intact and enabled the assignment of the exchangeable protons. A total of 226 experimental distance restraints were obtained. A restrained molecular dynamics and simulated annealing approach was utilized in the refinement. The data for 5 emergent molecular dynamics (MD) structures calculated from a B-form starting structure and 5 emergent MD structures calculated from a n A-form starting structure refined t o a n average pairwise root-mean-square (rms) difference of 1.2 A, with maximum pairwise rmsd of 1.7 A. The accuracy of the emergent structures was assessed by complete relaxation matrix back-calculation. The was measured between the refined structures and the sixth root residual index of 9.4 x NOE data, suggesting that the former were in reasonable agreement with the data. The refined structures revealed a n increased roll angle of 7” in the codon 61 sequence a t base step C5G18+A6*T17, which relieved the purine-purine clash in the minor groove, and in turn relieved the purine-purine clash in major groove between A4*T19and C5*G18. A 3.7 8, rise between C5*G18and A69T17was calculated, which assisted in relieving the purine-purine clash. The local variations in the B-like conformation did not confer large structural alterations upon the rus6l sequence, but could be important in modulating the reactivity of the first as compared t o the second adenine in codon 61.
Introduction Chemical carcinogenesis is commonly postulated to result from site-specific adduction of DNA by reactive electrophiles. These adducts ultimately induce mutations during DNA replication or repair ( 11. Adducts in crucial DNA sequences seem to be involved in tumorigenesis. The coding sequences for protooncogenes represent one such example. The ras protooncogene product functions on the inner surface of the plasma membrane and regulates signal transduction. Mutations within a limited number of codons, particularly codons 12,13, and 61, are particularly associated with activation of the rus family of oncogenes (reviewed in ref 2). The result of the mutation is a substitution of a key amino acid in the gene product, which results in impairment of GTPase activity. Oncogene activation depends on the identity of the amino acids encoded by a particular DNA sequence and * Author to whom correspondence should be addressed. @
Abstract published in Advance ACS Abstracts, July 15, 1995.
0893-228x/95/2708-0821$09.00/0
their functional role in the resulting gene product or altered gene product. It is anticipated to be modulated by the propensity toward adduct formation (3-13) and repair (14-18) at specific sites. Evidence suggesting chemical carcinogens induce mutations in, and thereby directly activate, protooncogenes, such as ras, comes from comparison of mutations in carcinogen-induced tumors within a number of tissues to known or suspected mutational spectra of specific carcinogens (14, 19, 20). Alternatively, it has been suggested that NMU-induced mammary tumors in rodents may instead arise from cells with preexisting oncogenic Hrasl mutations (211. The latter hypothesis posits that exposure to the chemical mutagen results in secondary mutations or promotes tumorigenesis. Adult human tumors are suspected to result from a multistep progression of mutations at various loci within the genome (reviewed in ref 22). The low probability that the necessary series of independent mutational events would occur within a single cell seems
0 1995 American Chemical Society
822 Chem. Res. Toxicol., Vol. 8, No. 6, 1995
inconsistent with this model, unless one considers either that the probability of mutations is greater for specific DNA sequences, due to increased probabilities of DNA damage or decreased probabilities of repair, or that there exist mutator phenotypes (23). Studies using native and synthetic DNA fragments revealed a distinct bonding preference of benzo[alpyrene diol epoxide to guanine-rich regions (3-7,24,25). M a toxin B1 also exhibited sequence-dependent effects in chemical reactivity, with runs of guanines being particularly susceptible toward adduction (9-13). A comparison of benzo[alpyrene diol epoxide bonding selectivity revealed a preference for bonding to the clustered guanines in the ras codon 12 sequence (8). The propensity toward DNA adduction by reactive electrophiles could be modulated by sequence-specific changes in the tertiary structure (26,271, electrostatic potential (28, 291, or conformational changes of the DNA duplex induced upon binding of reactive electrophiles (30-33). Conformational analysis probes for sequence-specific structural features in oligodeoxynucleotides. Such features, if present, could influence the reactivity of specific DNA bases to chemical carcinogens. An understanding of structure is also critical to investigations into adductinduced structural changes, which may induce mutations during subsequent DNA replication or repair events (34, 35). NMR techniques, including nuclear Overhauser effect (NOESY)' and correlation (COSY) spectroscopies, have become powerful tools in elucidating solution structures of oligodeoxynucleotide duplexes (36). Interproton distance constraints obtained by relaxation matrix analysis (RMA) allow the determination of the three-dimensional solution structure. The accuracy of emergent solution structures from restrained molecular dynamics calculations can be assessed by the use of the complete relaxation matrix approach to compare calculated NOE intensities with experimentally determined intensities (37-40). In the first phase of an investigation into the specific effects of adduct formation at the exocyclic amino groups of adenine in the human n-ras codon 61 sequence, the ras6l oligodeoxynucleotide d(CGGACAAGAAG).d(CTTCTTGTCCGY was constructed (Chart 1). A combination of NOESY and TOCSY spectra were used to assign the lH resonances. The solution structure was refined using restrained molecular dynamics and simulated annealing protocols, based upon NOE-derived distances. The emerAbbreviations: DSS, sodium 4,4-dimethyl-4-silapentanesulfonate; MD, molecular dynamics; NOE, nuclear Overhauser enhancement; NOESY, two-dimensional NOE spectroscopy; PEM, potential energy minimization; RMA, relaxation matrix; rMD, restrained molecular dynamics; rmsd, root-mean-square deviation; TPPI, time proportional phase increment; TOCSY, total homonuclear correlated spectroscopy. The oligonucleotides discussed in this paper do not have terminal phosphate groups-we abbreviate the nomenclature for oligonucleotides by leaving out the phosphodiester linkage. A, C, G, and T refer to mononucleotide units. A right superscript refers to numerical position in the oligonucleotide sequence starting from the 5'-terminus of chain A and proceeding to the 3'-terminus of chain A and then from the 5'terminus of chain B to the 3'-terminus of chain B. C2, C5, C6, C8, Cl', C2', C2", etc., represent specific carbon nuclei. H2, H5, H6, H8, Hl', H2'.H2". etc., represent the protons attached to these carbons.
Feng and Stone
gent structures refined to an average rms difference of 1.2 A, as determined by painvise analysis, and were compared to the NOE intensity data using complete relaxation matrix back-calculations. The calculated structures agree well with the data and demonstrate the ras61 sequence adopts a conformation in solution that is B-like, but with sequence-dependent variations associated with the R-Y step at A4.T19-C5.G1s and the Y-R step at C5.Gl8+A6*T1'. These local variations in the B-like conformation do not confer large conformational changes in the ras61 sequence. They might, however, provide a basis for altered chemical reactivity of the first as compared to the second adenine in codon 61.
Materials and Methods Sample Preparation. The oligodeoxynucleotides d(CGGACAAGAAG) and d(CTTCTTGTCCG1 were bought from the Midland Certified Reagent Co. (Midland, TX). The concentrations of the single-stranded oligodeoxynucleotides were determined from the extinction coefficients, 1.17 x lo5 M-l cm-l and 9.08 x lo4 M-l cm-l, respectively (41). Equal molar amounts were mixed in 10 mM NaHzP04,O.l M NaC1, and 50 mM NazEDTA at p H 6.9. The annealed duplex was eluted from a hydroxylapatite column using sodium phosphate and desalted by gel filtration using Bio-Gel P-2. All samples were prepared in 0.1 M NaC1, 0.01 M NaH2P04, and 0.05 mM NaZEDTA (pH 7.4). For observation of nonexchangeable resonances, samples were repeatedly lyophilized and finally dissolved in 0.5 mL of 99.996% D20, giving a 2 mM solution. For assignments of water-exchangeable protons, the samples were dissolved in a 9 : l HzO/DzO buffer ofthe same composition as described above. NMR Spectroscopy. Spectra were recorded a t a lH frequency of 500.13 MHz. The data were processed using FELIX 2.3 (Biosym Technologies, S a n Diego, CAI, running on Iris workstations (Silicon Graphics, Inc., Mountain View, CA). A phase-sensitive NOESY spectrum in D20, used for the assignments of the nonexchangeable protons, was recorded using the standard pulse sequence and the TPPI method for phase cycling; the mixing time was 250 ms. The residual water resonance was saturated during the relaxation delay and the mixing period. TOCSY experiments used t h e standard pulse sequence with a 100 m s MLEV-17 spinlock field (2 G). Chemical shifts were referenced internally relative to DSS. One-dimensional experiments in 90% HzO were carried out using a jump return 1-1 echo pulse sequence for water suppression (42). A twodimensional phase-sensitive NOESY experiment in 90% H20 was carried out using a jump return 1-1 sequence as the read pulse for water suppression (43, 441, a t 20 "C. Convolution difference was used during processing to minimize the residual water signal ( 4 5 ) . The mixing time was 250 ms. Relaxation Matrix Analysis. Three NOESY experiments were r u n at mixing times of 150, 200, and 250 ms. Footprints were selected manually with FELIX 2.3 to fit NOE cross-peaks at the contour level that showed the weak NOES but not t h e random noise. NOE cross-peaks were integrated for the three sets of spectra using the same contour levels. For overlapped cross-peaks, footprints were estimated and larger upper and lower error bounds were assigned to the resulting distances. For each of the three mixing times, a hybrid intensity matrix was constructed using MARDIGRAS (38). Complete iterative relaxation matrix (RMA) calculations were performed on each matrix, yielding 3 sets of internuclear distances. The distances were averaged and upper and lower error bounds were calculated t o provide the distance restraints t h a t were used in subsequent MD calculations. Restrained Molecular Dynamics Calculations. INSIGHT11 (Biosym Technologies, S a n Diego, CA) was used t o build the starting structures and for molecular visualization. Energy minimization and molecular dynamics calculations were carried out using X-PLOR (46). The total energy was the sum of the empirical energy of the molecule and the effective energy,
Chem. Res. Toxicol., Vol. 8, No. 6, 1995 823
Refined Structure of the n-ras61 Oligonucleotide composed of the restraint energy terms. The empirical energy function was derived from the CHARMM force field (47) developed for nucleic acids and treated all hydrogens explicitly (48). The effective energy term comprised restraining energy terms that used experimental information, and t h e empirical energy consisted of the energy terms for bonds, bond angles, dihedral angles, chirality or planarity, hydrogen bonding, and nonbonded interactions. The nonbonded interactions included van der Waals and electrostatic terms, which used the pure Lennard-Jones and Coulomb functions, respectively. The electrostatic term was based on a reduced charge set of partial charges (-0.32/residue) and a dielectric constant o,f 4.0. The cutoff radius for nonbonded interactions was 11 A, and the nonbonded list was updated if any atom moved more t h a n 0.5 A. The SHAKE algorithm (49)was used to fix all bond lengths involving hydrogens. The effective energy function was composed of two terms describing the distance and dihedral restraints, both in the form of a standard square well potential (50). Empirical base pairing distance and planarity restraints were used as follows. For G C base pairs, r[cytosine N4guanosine 061 1 2 . 9 1 0.05 A, r[cytosine N3-guanosine N11 = 2.95 0.05 A, rlcytosine 02-guanosine N21 = 2.86 & 0.05 A, r[cytosine N3-guanosine N2] = 3.65 i 0.05 8,and r[cytosine 02-guanosine 061 = 5.42 5 0.05 A. For A.T base pairs, r[adenosine N6-thymidine 041 = 2.95 f 0.05 A, r[adenosine N1-thymidine N3] = 2.82 0.05 A, r[adenosine N1-thymidine 041 = 3.63 i 0.05 A, and r[adenosine N6-thymidine 021 = 5.40 0.05 A. The value of the torsion angle for Watson-Crick base pair was a s follows: [purine C2-purine N1-pyrimidine N3pyrimidine C21 = 0 10”. Empirical base-step distances were r[H8-H8] = 5.00 f 0.20 A, r[H6-H6] = 5.00 f 0.20 A, and 0.20 A. Two em irical internucleotide r[H8-H6] = 4.80 distances r[G3H2’-A4 H81= 3.80 0.2 and r[G3H3’-A4 H81 = 4.95 k 0.2 A were used between nucleotides G3and A4. The integration time step used in t h e molecular dynamics calculations was 1 fs. Structure coordinates were archived every 0.1 ps. Back-calculation of NMR data was performed using CORMA (37). The refined structures were analyzed using DIALS AND WINDOWS 1.0 (51).
A
I
*
*
*
0
I
b
cd
A4
6.0
*
5.5
5.0
D1 (PPm
+
*
* B
Results Spectral Assignment. (a) Nonexchangeable Protons. The one-dimensional lH NMR spectrum of the at 20 “C duplex d(CGGACAAGAAG).d(CTTC’M’GTCCG) exhibited well-resolved proton resonances for the base (7.0-8.2 ppm), sugar H1’ (5.0-6.3 ppm), sugar H3’ (4.44.9 ppm), sugar H2’,2” (1.6-2.7 ppm), and CH3 (1.2-1.7 ppm) protons. At lower temperatures the resonances were less resolved. A contour plot of the phase-sensitive NOESY spectrum collected at a mixing time of 250 ms for the unmodified duplex is shown in Figure 1. Sequential proton assignments were carried out according to standard methods (36, 52). The salient features of the sequential assignment were as follows. The aromatic protons of C5,located between two A*Tbase pairs, were shifted upfield as compared to the remainder of the cytosine H5 and H6 protons, presumably due to increased stacking interactions. The H5 and H6 protons of C12, from the terminal C*Gbase pair, were located downfield of the remainder of the cytosine H5 and H6 protons, which were clustered at 5.4-5.7 and 7.3-7.5 ppm, respectively. This clustering of the cytosines resulted in the overlap of a number of cross-peaks. Specifically, H6 of C1 T14 C l j T16, CZo,and CZ1and H8 of G8 and Gll were superimposed. Likewise, the H1’ protons of A4,TI3, TI4, and GZ2were superimposed, as were those from AIO, Gll, C1j, T16, G18, T19, and CZo. The remainder of the deoxyribose protons were assigned from TOCSY and NOESY spectra, based upon the sequential assignments of the H1’ protons. The assignments of the H2’ and H2” ,
I
,
I
6.0
I
5.5
I
5.0
D1 (PPm Figure 1. Expanded plot of a phase-sensitive NOESY spectrum in DzO buffer a t a mixing time of 250 ms showing the sequential NOE connectivities from t h e aromatic to H1’ protons. (A) (B) Sequential NOE connectivities for nucleotides C1-G”. Sequential NOE connectivities for nucleotides Clz-Gz2. The base positions are indicated at t h e intranucleotide aromatic to sugar H1’NOE cross-peak.
sugar protons were made from their NOES with the H1’ sugar protons of the same residue and were based on the assumption that, for a B-like DNA, the Hl’-H2’’ distances were shorter than the Hl’-H2’ distances. Therefore, the stronger cross-peaks were assigned to the H2” protons. The adenine H2 protons were assigned from weak cross-peaks to H1’ of the adjacent 3’-nucleotide,and from the strong cross-peaks to the imino protons of the
824 Chem. Res. Toxicol., Vol. 8, No. 6, 1995
Feng and Stone Table 2. Assignments (ppm) of the Exchangeable Imino and Amino Protons of rase1 imino base pair
G Hl/T H3
G2CZ1 G3CZ0 A4Tl9 C5GI8 AV7 A7TI6 GQ5 A9TI4 A10T13
13.01 12.66 13.52 12.31 13.57 13.83 12.33 13.71 14.10
amino
C NHz(~I"
c NHzb,b
6.87 6.61
8.54 8.24
6.26
8.00
6.82
8.14
a Non-hydrogen-bonded cytosine amino proton. Hydrogenbonded cytosine amino proton.
Ti3 T 'h3H I
I
14.0
13.5
I
I
13.0
I
12.5
12.0
D1 (PPW Figure 2. An expanded plot of a phase-sensitive NOESY spectrum at a mixing time of 250 m s showing t h e sequential NOE connectivities for t h e imino protons of base pairs G2.Cz1AlO.Tl3.
Table 1. Assignments (ppm) of the Base and Deoxyribose Protons of the rase1 Sequence H2 H8 H6 H5 H1' H2' H2" H3' H4' CH3 C' 7.41 5.71 5.56 1.65 2.17 4.53 3.90 G2 7.75 5.28 2.52 2.56 4.83 4.13 G3 7.66 5.49 2.50 2.63 4.90 4.25 A4 7.63 8.00 6.04 2.47 2.74 4.90 4.31 c5
A6 A7 GB A9 AIo G" C '2 T13 TI4
C'5 T'6 T17 G'S TI9 C*O C2' G22
6.97 7.98 7.27 7.86 7.45 7.15 7.88 7.49 7.84 7.40
7.03 5.09 5.22 5.60 5.71 5.18 5.75 5.86 5.82
1.67 2.49 2.40 2.30 2.43 2.39 2.18
2.06 2.65 2.60 2.44 2.70 2.68 2.10
4.64 4.86 4.87 4.81
3.93 4.18 4.23 4.16
f
4.86 4.26 4.45 4.04
7.73 5.77 5.70 2.15 2.44 4.51 7.53 6.06 2.12 2.50 4.79 7.35 6.01 2.15 2.46 4.80 7.49 5.50 5.87 2.01 2.42 7.30 5.84 2.00 2.40 7.23 5.65 2.04 2.36 4.74 7.72 5.78 2.46 2.62 4.80 7.13 5.87 2.00 2.36 7.40 5.49 5.78 1.97 2.25 7.34 5.55 5.47 1.83 2.17 7.81 6.02 2.48 2.22 4.55
3.95 4.14 4.12 4.06 4.10 4.00 4.23 4.09 3.99 3.95 4.04
1.55 1.53 1.48 1.54 1.20
complementary thymines observed in the HzO NOESY spectrum. (b)Exchangeable Protons. A contour plot of the far downfield region of the IH NMR spectra acquired in HzO buffer is shown in Figure 2. This region of the spectrum revealed the hydrogen-bonded imino resonances and dipolar connectivities between adjacent imino protons. With the exception of the terminal base pairs C1GzZand G11.C12,the imino resonances from each base pair were resolved and identified. The imino resonance from base pair C5-Gl8overlapped with that from base pair G8-Cl5, while the imino resonance of A6.T1' overlapped with that of A4.T19. Overall, the spectrum and observed chemical shifts were consistent with a B-form helix. Table 1lists the collected IH chemical shift data for the nonexchangeable protons. The assignments of the exchangeable protons are collected in Table 2.
Calculation of Internuclear Distances. Classical B-DNA and A-DNA oligodeoxynucleotides(53)were used as the reference structures. These were energy minimized for 100 iterations by the conjugate gradient method to give the starting structures IniA and IniB. A total of 226 experimental intensities, supplemented with calculated intensities from IniB, were used in the MARDIGRAS calculations. An isotropic correlation time z, of 5 ns was used for both sugar and base protons. The determination of tc was from a fluorescence anisotropy study of an 11-mer oligodeoxynucleotide adducted to BPDE.3 Excluding the terminal base pairs, an average of 20 intranucleotide, internucleotide, and empirical distances were obtained for each base pair in ras61. Including NOE and empirical distance restraints, a total of 279 distance restraints were used for the MD calculations, of which 50 were base pairing restraints, and 10 were base-step restraints, between nucleotides C1 and G2, G2 and G3, G3 and A4, A9 and AlO,AIo and GI1, C12 and TI3, TI3 and TI4, C19 and CZo,CZoand CZ1,and CZ1and GZ2. Two were internucleotide distance restraints between nucleotides G3 and A4. Individual distance restraints calculated by MARDIGRAS were removed from the distance set if the values were calculated to be greater than 5 A. Additional restraints were removed if they were inconsistent with a reasonable structure. These inconsistent distance restraints were usually due to errors in intensity measurements arising from spectral overlap, or from cross-peaks close to the water resonance whose intensities were altered by the water presaturation pulse used during acquisition of NOESY spectra. After culling poor distance restraints, the final experimental distance set consisted of 217 (collected in Table 3) of the original 226 distances calculated by MARDIGRAS. Restrained Molecular Dynamics. The NOE distance restraints were separated into four sets (best, medium, bad, and worst) according to the quality of distances derived from MARDIGRAS [force constants 45-50,30-45,20-30, and e20 k c a l / ( m ~ l . ~ ~ and ) ] ,those four sets distances were assigned force constants loo%, 80%, 60%, and 40% of the maximum during the MD calculations (the distance restraints are tabulated in Table S1 of the supporting information). Class 1 consisted of the 153 best distance restraints, class 2 contained 29 restraints, class 3 contained 13 restraints, and class 4 consisted of the 34 worst restraints. The MD calculations were initialized by assigning a random set of velocities to all the atoms that fit a MaxwellBoltzmann distribution at 1000 K. Sets of 5 MD calculaDr. Irene S. Zegar, personal communication.
Chem. Res. Toxicol., Vol. 8, No. 6, 1995 825
Refined Structure of the n-ras61 Oligonucleotide Table 3. The Distribution of Experimental Restraints among Nucleotide Units of ras61
nucleotide
total experimental restraints
intranucleotide restraints
internucleotide restraintsa
9 7 8 7 12 5 5 8 3
0 3 3 2 5 3 2 3 2
9 10 11 9 17 8 7 11
5 7 5
1
6 10
~~
C' G2 G3 A4 c5
A6 A7 G8
As A10
G" C'2
TI3 T14
9 4
C'5
4
T'6 T17
6 10 6 6 8 9 9
G'8
TI9 C20
C2' G22
3 0 3 3 2 4
5 3 6 5 4 3
5
5 12
7 6
10 15 9 12 13 13 12
The internucleotide NOES are listed in the direction n n - 1.
crMDA>
-
tions were initiated using random seeds, starting from both IniA and IniB. Simulated annealing was carried out for 7 ps, with the force constants of 10 kcaYmo1-A2 for empirical hydrogen bonding and base pair planarity restraints, and 50 kcal/mol.A2 as the maximal constant for NOE restraints. To control the temperature, the molecules were weakly coupled to a temperature bath with a target temperature of 1000 K and a coupling constant of 0.05 ps (54). The force constants were scaled up to 100 and 150 kcal/mol.A2for the empirical restraints and NOE restraints, respectively, during the next 7 ps with the temperature maintained at 1000 K. The temperature was maintained at 1000 K for 6 ps, then decreased to 300 K for a period of 3 ps, and then kept at 300 K for 2 ps using the high force constants. The force constants were scaled down to 20 and 50 kcal/mol.A2 for the empirical restraints and NOE restraints, respectively, for a period of 4 ps and kept at those values for 14 ps. The structures from the last 5 ps were averaged, and energy minimized for 300 iterations by the conjugate gradient method to give the final structures. The superposition of the 5 emergent MD structures from IniA and IniB is shown in Figure 3. The large difference between IniA and IniB was confirmed by the measured rmsd of 4-10 A between these two structures (Figure 4A). The sets of 5 structures calculated from both IniA and IniB converged to a B-like structure, as confirmed by comparison of the rms differences between IniA, IniB, and the emergent rMD structure. Large rms differences were observed between IniA and rMD (Figure 4B), while IniB showed close agreement with rMD, with rms differences of the order of l A (Figure 4C). The final structures from each set were tested for convergence by all-atom pairwise rmsd comparison, excluding the end base pairs. Satisfactory convergence was obtained, as judged by a maximum pairwise rmsd of less than 1.7 A, and average pairwise rmsd of 1.2 A, excluding the terminal base pairs. The resulting 10 MD structures were averaged and energy minimized to obtain the final rMD structure.
crMDB>
rMDA vs. rMDB
Figure 3. Stereo drawings of (A) 5 MD structures which emerged from calculations using IniA as the starting structure, (B) 5 MD structures which emerged from calculations using IniB as the starting structure, and ( C )the superposition of the rMDA and rMDB structures. The structure drawn in bold is t h a t of rMDB.
826 Chem. Res. Toxicol., Vol. 8, No. 6, 1995 A.
IniA vs.
15.0
- E
Feng and Stone B.
lniB
IniA vs. rMD
15.0 1
I
I
- 1
I
12.0
12.0
0.0 0 2 4 6 8 1012141618202224 residue
0.0 0 2
C.
D.
iniB
VI.
rMD
4
6
8 1012141618202224 residue
dMD>
VI.
rMD
15.0 12.0
12.0 9.0
6.0 3.0
0.0 0
2
4
6
8 1012141618202224
residue
0 2
4
6
8 1012141618202224
resldue
Figure 4. Per residue rmsd comparisons between the initial structures and the final structures. (A) Comparison of IniA vs IniB. (B) Comparison of IniA vs rMD. (C) Comparison of IniB vs rMD. (D) (rMD) vs rMD is t h e comparison of each of t h e final structures with the energy minimized average of t h e 10. The error bars represent the standard deviation observed a t each nucleotide. Table 4. Comparison of Sixth-Root Residual Indices R1” for Starting Models and Resulting MD Structures as a Function of NOE Mixing RiX( X structure tm = 150 ms t , = 200 ms t , = 250 ms IniA IniB rMDA rMDB rMDfinai
20 11 9.7 9.7 10
19 11 9.4 9.3 9.7
18 11 9.2 9.1 9.4
a Only the inner 9 base pairs were used in the calculations to - ( U ~ ) ~ ~ ’ ~ ~ ~ I ( ~ O ) where ~ ~ ’ ~ I , (a01 exclude end effects. R1“ = Z.l(a~),”~ and (a,) are the intensities of observed (nonzero) and calculated NOE cross-peaks. c IniA, starting energy-minimized A-DNA IniB, starting energy-minimized B-DNA, rMDA, average of 5 rMD structures starting from IniA; rMDB, average of 5 MD structures starting from IniB; rMDfinal,average of 10 MD structures starting from IniA and IniB.
Complete Relaxation Matrix Calculations. The agreement between the experimental NOE cross-peak intensities and the values calculated from the restrained MD-generated structures was measured by the sixth-root residual index Rlr, calculated using C O R M (37). The sixth-root R factor represents a residual index that measures the difference between NOE intensities calculated from the model structure and the intensities measured from the NOESY spectra. Table 4 shows R factors calculated for the starting structures and the rMD structures, calculated a t three values of the NOE mixing time. Consistent results were obtained a t each of the three mixing times, with the lowest values being observed for the 250 ms data. This perhaps reflected the improved sensitivity of the longer mixing time data. At 250 ms, the R factor for IniB was 0.11 as compared to a value of 0.18 for IniA. The R factor decreased when the starting structures IniA and IniB were compared to the structures rMDA and rMDB that emerged from the calculations. As compared to the two starting structures, which exhibited different R factors, the values calculated for the emergent rMDA and rMDB structures were 0.092 and 0.091, respectively, at 250 ms.
Refined Structure. (a)Backbone Torsion Angles. The calculated values for the backbone torsion angles a-5 are shown graphically in Figure 5. The plots show the data obtained from the IniA- and IniB-derived structures separately, as well as the overall average. The dashed line represents the expected value for canonical B-DNA, whereas the dotted line represents the canonical A-DNA value. In determining the average values for specific torsion angles, points that deviated greatly were discarded. Inspection of the data suggested that these were artifacts arising from the high temperature MD simulations, which occurred as a result of insufficient restraints at the backbone torsion angles. (The distribution of data points is documented in DIALS AND WINDOWS format in the supporting information). m e r removing data points that were judged to be unreasonable, the backbone torsion angles converged reasonably well starting from either IniA or IniB. The torsion angles for ,l3 and y converged outside the canonical range for Aor B-form DNA, with the ,8 angles being consistently lower whereas the y angles were consistently greater than expected for the canonical right-handed helices. The remaining torsion angles 6, E , 5, and a converged to values consistent with right-handed helices. A discussion of the precision vs the accuracy of these angles is to be found below. The pseudorotation angle 4 remained >loo”, consistent with predominantly S conformation of the deoxyribose moieties. The values of the glycosidic torsion angle x were consistently greater than 240°, indicative of an anti conformation. (b)Intra-Base-PairHelical Parameters. The plots of the calculated intra-base-pair helical parameters are shown in Figure 6A. Most parameters converged well from both IniA and IniB starting structures and appeared to be in reasonable agreement with the canonical values for B-form DNA. This was most evident in the data for base-pair inclination, stretch, and x-displacement, since for most of the intra-base-pair parameters, A-form and B-form DNA are similar. A notable exception was the values calculated for propeller twist, which did not agree
Chem. Res. Toxicol., Vol. 8, No. 6, 1995 827
Refined Structure of the n-ras61 Oligonucleotide
700
160.0
0
2
4
6
8 10121418182022 nuclrolldr
.......
160.0
0 2
4
6
8 10121418162022 nuolrolldr
..... 45 0
1
1.....................................
I
270 0
210
.....................................
190 0
0
2
4
6
8
1012 1416182022 nuel*otlds
0 2
4
6
8 10121416182022
0 2
4
8
8 10121416182022 nuel*ollde
nucleolldr
Figure 5. Helicoidal analysis. Backbone torsion angles (a-c), glycosidic torsion angles (x),and sugar pseudorotation angle (41, for the structures rMDA (O), rMDB (O),and rMD (0).The dashed line represents the expected value for canonical B-DNA, whereas the dotted line represents the expected value for canonical A-DNA.
well with either canonical A-form or B-form DNA. The propeller twist values were consistently calculated to be lower than the canonical values. Deviations in the values of tip angle, buckle, and stagger were observed for the terminal base pairs C1.GzZand G 1 W 2 . The precision of the calculated values for propeller twist, tip angle, stagger, and buckle was lower than for the other intrabase-pair parameters. The difference of base-pair opening between the rMD structure and classical B-DNA was minimal (
