A new way to see RNA

Unlike proteins, the RNA backbone has numerous degrees of freedom (eight, if one counts the sugar pucker), making RNA modeling, structure building and prediction a multidimensional problem of exceptionally high complexity. And yet RNA tertiary structures are not infinite in their structural morphology; rather, they are built from a limited set of discrete units. In order to reduce the dimensionality of the RNA backbone in a physically reasonable way, a shorthand notation was created that reduced the RNA backbone torsion angles to two (η and θ, analogous to φ and ψ in proteins). When these torsion angles are calculated for nucleotides in a crystallographic database and plotted against one another, one obtains a plot analogous to a Ramachandran plot (the η/θ plot), with highly populated and unpopulated regions. Nucleotides that occupy proximal positions on the plot have identical structures and are found in the same units of tertiary structure. In this review, we describe the statistical validation of the η/θ formalism and the exploration of features within the η/θ plot. We also describe the application of the η/θ formalism in RNA motif discovery, structural comparison, RNA structure building and tertiary structure prediction. More than a tool, however, the η/θ formalism has provided new insights into RNA structure itself, revealing its fundamental components and the factors underlying RNA architectural form.     

Crystal structures of two plasmid copy control related RNA duplexes: An 18 base pair duplex at 1.20 A resolution and a 19 base pair duplex at 1.55 A resolution.

The structures of two RNA duplexes, whose sequences correspond to portions of the ColE1 plasmid copy control RNA I and RNA II, have been determined. Crystals containing the 18mers 5'-CA CCGUUGGUAGCGGUGC-3' and 5'-CACCGCUACCAACGGUGC-3' diffract to 1.20 A resolution while those containing the 19mers 5'-GCACCGUUGGUAGCGGUGC-3' and 5'-GCACCGCUACCAACGGUGC-3' diffract to 1.55 A resolution. Both duplexes are standard A form, with Watson-Crick base pairing throughout. Use of anisotropic atomic displacement factors in refinement of the 1.20 A structure dramatically improved refinement statistics, resulting in a final R(free) of 15.0% and a crystallographic R-factor of 11.6%. Perhaps surprisingly, these crystals of the 18 base pair RNA exhibit a 36-fold static disorder, resulting in a structure with a single sugar-phosphate backbone conformation and an averaged base composition at each residue. Since the sugar-phosphate backbone structure is identical in the 36 different nucleotides that are superimposed, there can be no sequence-dependent variation in the structure. The average ribose pucker amplitude is 45.8 degrees for the 18 base pair structure and 46.4 degrees for the 19 base pair structure; these values are respectively 19% and 20% larger than the average pucker amplitude reported from nucleoside crystal structures. A standard RNA water structure, based on analysis of the hydration of these crystal structures and that of the TAR RNA stem [Ippolito, J. A., and Steitz, T. A. (1998) Proc. Natl. Acad. Sci. U.S.A. 95, 9819-9824], has been derived, which has allowed us to predict water positions in lower resolution RNA crystal structures. We report a new RNA packing motif, in which three pro-S(p) phosphate oxygens interact with an ammonium ion.     

Crystal structure of an adenine bulge in the RNA chain of a DNA.RNA hybrid, d(CTCCTCTTC).r(gaagagagag)

Crystal structure of a DNA.RNA hybrid, d(CTCCTCTTC).r(gaagagagag), with an adenine bulge in the polypurine RNA strand was determined at 2.3 A resolution. The structure was solved by the molecular replacement method and refined to a final R-factor of 19.9% (Rfree 22.2%). The hybrid duplex crystallized in the space group I222 with unit cell dimensions, a = 46.66 A, b = 47.61 A and c = 54.05 A, and adopts the A-form conformation. All RNA and DNA sugars are in the C3'-endo conformation, the glycosyl angles in anti conformation and the majority of the C4'-C5' torsion angles in g+ except two trans angles, in conformity with the C3'-endo rigid nucleotide hypothesis. The adenine bulge is looped out and it is also in the anti C3'-endo conformation. The bulge is involved in a base-triple (C.g)*a interaction with the end base-pair (C9.g10) in the minor groove of a symmetry-related molecule. The 2' hydroxyl group of g15 is hydrogen bonded to O2P and O5' of g17, skipping the bulged adenine a16 and stabilizing the sugar-phosphate backbone of the hybrid. The hydrogen bonding and the backbone conformation at the bulged adenine site is very similar to that found in the crystal structure of a protein-RNA complex.     

Molecular structures of two new anti-HIV nucleoside analogs: 9-(2,3-dideoxy-2-fluoro-beta-D-threo-pentofuranosyl)adenine and 9-(2,3-dideoxy-2-fluoro-beta-D-threo-pentofuranosyl)hypoxanthine.

The x-ray crystal structures of two new anti-HIV compounds, 9-(2,3-dideoxy-2-fluoro-beta-D-threo-pentofuranosyl)adenine (2'-F-dd-araA) and 9-(2,3-dideoxy-2-fluoro-beta-D-threo- pentofuranosyl)hypoxanthine (2'-F-dd-aral), have been determined at two temperatures. Both crystals are in the space group P2(1)2(1)2(1), and their structures were solved by direct methods. Least-squares refinement produced final R-factors of 0.027 for the 2'-F-dd-araA structure and of 0.044 for the 2'-F-dd-aral structure, respectively. The latter structure contains a two-fold disordered conformation of the sugar moiety. All three conformers (one for 2'-F-dd-araA and two for 2'-F-dd-aral) adopt an anti chi CN glycosyl torsion angle. The sugar in the 2'-F-dd-araA structure has a C2'-endo pucker conformation, whereas the sugar in the 2'-F-dd-aral structure has a mixture of C2'-endo and C3'-endo pucker conformations. When the sugar adopts the C2'-endo conformation, the torsion angle about the C4'-C5' bond is in a transgauche+ conformation. In contrast, when the sugar adopts the C3'-endo conformation, the torsion angle about the C4'-C5' bond is in a gauche(+)-gauche- conformation. The C2'-F bond distance is 1.406(3) A, similar to that found in other aliphatic C-F bonds. The results suggest that the 2'-fluoro-2',3'-dideoxyarabinosyl nucleosides do not have a strong preference for either C2'-endo or C3'-endo sugar pucker.     

Solution conformation of a bulged adenosine base in an RNA duplex by relaxation matrix refinement

Bulges are common structural motifs in RNA secondary structure and are thought to play important roles in RNA-protein and RNA-drug interactions. Adenosine bases are the most commonly occurring unpaired base in double helical RNA secondary structures. The solution conformation and dynamics of a 25-nucleotide RNA duplex containing an unpaired adenosine, r(GGCAGAGUGCCGC): r(GCGGCACCUGCC) have been studied by NMR spectroscopy and MORASS iterative relaxation matrix structural refinement. The results show that the bulged adenosine residue stacks into the RNA duplex with little perturbation around the bulged region. Most of the bases in the RNA duplex adopt C(3)'-endo conformation, exhibiting the N-type sugar pucker as found in the A form helices. The sugars of the bulged residue and the 5' flanking residue to it are found to exhibit C(2)'-endo conformation. None of the residues are in syn conformation.     

The occurence of the syn-C3' endo conformation and the distorted backbone conformations for C4'-C5' and P-O5' in oligo and polynucleotides.

The nucleoside constituents of nucleic acids prefer the anti conformation (1). When the sugar pucker is taken into account the nucleosides prefer the C2'endo-anti conformation. Of the nearly 300 nucleosides known, about 250 are in the anti conformation and 50 are in the syn-conformation, i.e., anti to syn conformation is 5:1. The nucleotide building blocks of nucleic acids show the same trend as nucleosides. Both the deoxy-guanosine and riboguanosine residues in nucleosides and nucleotides prefer the syn-C2'endo conformation with an intra-molecular hydrogen bond (for nucleosides) between the O5'-H and the N3 of the base and, a few syn-C3'endo conformations are also observed. Evidence is presented for the occurrence of the C3'endo-syn conformation for guanines in mis-paired double helical right-handed structures with the distorted sugar phosphate C4'-C5' and P-O5' bonds respectively, from g+ (gg) and g- to trans. Evidence is also provided for guanosine nucleotides in left-handed double-helical (Z-DNA) oligo and polynucleotides which has the same syn-C3'endo conformation and the distorted backbone sugar-phosphate bonds (C4'-C5' and P-O5') as in the earlier right-handed case.     

On the structural features of hairpin triloops in rRNA: from nucleotide to global conformational change upon ligand binding

RNA structure can be viewed as both a construct composed of various structural motifs and a flexible polymer that is substantially influenced by its environment. In this light, the present paper represents an attempt to reconcile the two standpoints. By using the 3D structures both of four (16S and 23S) portions of unbound 50S, H50S, and T30S ribosomal subunits and of 38 large ribonucleoligand complexes as the starting point, the behavior, which is induced by ligand binding, of 73 hairpin triloops with closing g-c and c-g base pairs was investigated using root-mean-square deviation (RMSD) approach and pseudotorsional (eta,theta) convention at the nucleotide-by-nucleotide level. Triloops were annotated in accordance with a recent proposal of geometric nomenclature. A simple measure for the determination of the strain of a triloop is introduced. It is believed that a possible classification of the interior triloops, based on the 2D eta-theta unique path, will aid to conceive their local behavior upon ligand binding. All rRNA residues in contact with ligands as well as regions of considerable conformational changes upon complex formation were identified. The analysis offers the answer to: how proximal to and how far from the actual ligand-binding sites the structural changes occur?     

Mechanisms of chain folding in nucleic acids. The (omega, omega) plot and its correlation to the nucleotide geometry in yeast tRNAPhe1.

The (omega', omega) polot depicting the internucleotide P-O bond rotation angles in yeast phenylalanyl transfer RNA has established the interdependence of the phosphodiesters and the nucleotide geometries in the folding of the polynucleotide backbone. The plot distinguishes the regions characteristic of secondary helical structures and tertiary structural loops and bends. The folding of the polynucleotide chain is accomplished either solely by rotations around the P-O bonds or in concert with rotations around the nucleotide C4'-C5' bond with or without changes in the sugar ring pucker. In spite of differences in nucleotide sequence and intraloop tertiary interactions in the anticodon and pseudouridine loops, a characteristic repeating structural unit is found for the sugar-phosphate backbone of the tetranucleotide segment around the sharp turns.     

NMR structure of an antisense DNA.RNA hybrid duplex containing a 3'-CH(2)N(CH(3))-O-5' or an MMI backbone linker.

The solution structure of an antisense DNA.RNA hybrid duplex, d(CGCGTT-MMI-TTGCGC).r(GCGCAAAACGCG) (designated R4), containing an MMI backbone linker [3'-CH(2)N(CH(3))-O5'], is elucidated. The structural details of the MMI linker, its structural effects on the neighboring residues, and the molecular basis of the MMI effects are examined. The lipophilic N-methyl group of MMI is peripheral to the helix, assuming a conformation that is most stable with regard to the N-O torsion angle. The MMI linker promotes a 3'-endo conformation for the sugar moieties at both 3'- and 5'-adjacent positions and a backbone kink involving distant residues along the 3'-direction. Comparison of R4 with other analogous hybrid duplexes previously studied in this laboratory reveals a new family of low-energy helical conformations that can be accommodated in stable duplexes and a common feature of C3'-modified sugars for adopting a C3'-endo pucker. The results of these studies emphasize the interplay of several factors that govern the formation of stable hybrid duplexes and provide a basis for the understanding of the biological role of the MMI modifications, which are important building blocks for a family of promising chimeric antisense oligonucleotides.     

Conformational properties of branched RNA fragments in aqueous solution

The conformational properties of branched trinucleoside diphosphates ACC, ACG, AGC, AGG, AUU, AGU, AUG, ATT, GUU, and aAUU [XYZ = X(2'p5'Y)3'p5'Z] have been studied in aqueous solution by nuclear magnetic resonance (1H, 13C), ultraviolet absorption, and circular dichroism. It is concluded from these studies that the purine ring of the central residue (X; e.g., adenosine) forms a base-base stack exclusively with the purine or pyrimidine ring of the 2'-nucleotidyl unit (Y; 2'-residue). The residue attached to the central nucleoside via the 3'-5'-linkage (Z; 3'-residue) is "free" from the influence of the other two heterocyclic rings. The ribose rings of the central nucleoside and the 2'- and 3'-residues exist as equilibrium mixtures of C2'-endo (2E)-C3'-endo (3E) conformers. The furanose ring of the central nucleoside (e.g., A) when linked to a pyrimidine nucleoside via the 2'-5'-linkage shows a higher preference for the 2E pucker conformation (e.g., AUG, AUU, ACG, ca. 80%) than those linked to a guanosine nucleoside through the same type of bond (AGU, AGG, AGC, ca. 70%). This indicates some correlation between nucleotide sequence and ribose conformational equilibrium. The 2E-3E equilibrium of 2'-pyrimidines (Y) shows significant, sometimes exclusive, preference (70-100%) for the 3E conformation; 3'-pyrimidines and 2'-guanosines have nearly equal 2E and 3E rotamer populations; and the ribose conformational equilibrium of 3'-guanosines shows a preference (60-65%) for the 2E pucker. Conformational properties were quantitatively evaluated for most of the bonds (C4'-C5', C5'-O5', C2'-O2', and C3'-O3') in the branched "trinucleotides" AUU and AGG by analysis of 1H-1H, 1H-31P, and 13C-31P coupling constants. The C4'-C5' bond of the adenosine units shows a significant preference for the gamma + conformation. The dominant conformation about C4'-C5' and C5'-O5' for the 2'-and 3'-nucleotidyl units is gamma + and beta t, respectively, with larger gamma + and beta t rotamer populations for the 2'-unit. The increased conformational purity in the 2'-residue, compared to the 3'-residue, is ascribed to the presence of an ordered (adenine----2'-residue) stacked state. The favored rotamers about C3'-O3' and C2'-O2' are epsilon- and epsilon'-, respectively. The conformational features of AUU and AGG were compared to those of their constitutive dimers A3'p5'G, A2'p5'G, A3'p5'U, and A2'p5'U and monomers 5'pG and 5'pU.     

Analysis of the possible helical structures of nucleic acids and polynucleotides. Application of (n-h) plots.

The two helical parameters n and h where n is the number of nucleotide residues per turn and h is the height per nucleotide residue have been evaluated for single stranded helical polynucleotide chains comprising C(3') -endo and C(2') endo class of nucleotides. The helical parameters are found to be especially sensitive to the C(4')-C(3') (sugar pucker) and the C(4')-C(5') torsions. The (n-h) plots display only one important helix forming domain for each class of nucleotides characterized by the sugar pucker and the C(4')-C(5') torsion. A correlation between the (n-h) plots and the known RNA (A,A') and DNA (A,B,C) helical forms has been established. It is found that all forms of helices except the C-DNA possess a favorable combination of P-O torsions. The analysis of the (n-h) plots suggests that C-DNA can have a conformation very similar to B-DNA. Although the (n-h) plots predict the stereochemical possibility of both right-handed and left-handed helices, nucleic acids apparently prefer right-handed conformation because of the energetics associated with the sugar-phosphate backbone and the base.     

A purine nucleoside unequivocally constrained in the syn form. Crystal structure and conformation of 8-(alpha-hydroxyisopropyl)-adenosine

Crystals of 8-(alpha-hydroxyisopropyl)-adenosine dihydrate, C13H19N5O5.2H2O, belong to the monoclinic space group P21. Cell dimensions are a = 8.259 (1), b = 11.117 (2), c = 9.663 (1) A, beta = 109.65 (2) degrees. Intensity data were collected on a four-circle diffractometer and the structure was solved by direct methods. Block diagonal least-squares refinement led to R = 0.031 for 1467 reflections. The glycosyl torsion angle chiCN is 241.4 degrees, corresponding to a syn conformation. The conformation of the exocyclic C(4')-C(5') bond is gauche-gauche and the sugar pucker is C(2') endo. It is considered that the bulky, tetrahedral, neutral 8-substituent, with an effective van der Waals radius of 3.5--4.0 A, provides an adenosine analogue which should exhibit the syn conformation about the glycosidic bond in solution as well as in solid state, irrespective of the nature of the sugar pucker. It should therefore be suitable for studies of interactions with enzyme systems requiring the anti conformation of the nucleoside or nucleotide.     

The crystal and molecular structure of 2'-deoxy-2'-fluoroinosine monohydrate.

The structure of the hydrate of 2'-deoxy-2'-fluoroinosine has been determined by single-crystal x-ray diffraction. The nucleoside crystallizes in space group P2(1)2(1)2(1) with unit cell dimensions, a = 33.291, b = 10. 871, c = 6.897A. There are two nucleosides and two water molecules in the asymmetric unit. The structure was solved by direct methods and refined to a residual R = 0.095. The two independent nucleosides in the asymmetric unit show different conformations about the glycosidic bond, while other structural details are similar. The base orientation to the sugar is syn in molecule A, whereas anti in molecule B. The exocyclic C(4')-C(5') bond conformation defined with respect to C(3')-C(4')-C(5')-O(5') is gauche+ in both molecules A and B. The sugar ring pucker defined by the pseudorotation phase angle P is a twisted conformation in both, C(3')-endo-C(4')-exo with P = 29 degrees in molecule A and C(4')-exo-C(3')-endo with P = 41 degrees in molecule B. It is shown by comparison with x-ray results of other 2'-fluoronucleosides and unmodified nucleosides including inosines that, in addition to a strong preference of the C(3')-endo type pucker, twisted conformations involving C(4')-exo puckering may be one of characteristic features of 2'-fluoronucleosides.     

A molecular dynamics simulation of the flavin mononucleotide–RNA aptamer complex

We report on an unrestrained molecular dynamics simulation of the flavin mononucleotide (FMN)–RNA aptamer. The simulated average structure maintains both cross‐strand and intermolecular FMN–RNA nuclear Overhauser effects from the nmr experiments and has all qualitative features of the nmr structure including the G10–U12–A25 base triple and the A13–G24, A8–G28, and G9–G27 mismatches. However, the relative orientation of the hairpin loop to the remaining part of the molecule differs from the nmr structure. The simulation predicts that the flexible phosphoglycerol part of FMN moves toward G27 and forms hydrogen bonds. There are structurally long‐lived water molecules in the FMN binding pocket forming hydrogen bonds within FMN and between FMN and RNA. In addition, long‐lived water is found bridging primarily RNA backbone atoms. A general feature of the environment of long‐lived “structural” water is at least two and in most cases three or four potential acceptor atoms. The 2′‐OH group of RNA usually acts as an acceptor in interactions with the solvent. There are almost no intrastrand O2′H(n)⋮O4′(n + 1) hydrogen bonds within the RNA backbone. In the standard case the preferred orientation of the 2′‐OH hydrogen atoms is approximately toward O3′ of the same nucleotide. However, a relatively large number of conformations with the backbone torsional angle γ in the trans orientation is found. A survey of all experimental RNA x‐ray structures shows that this backbone conformation occurs but is less frequent than found in the simulation. Experimental nmr RNA aptamer structures have a higher fraction of this conformation as compared to the x‐ray structures. The backbone conformation of nucleotide n + 1 with the torsional angle γ in the trans orientation leads to a relatively short distance between 2′‐OH(n) and O5′(n + 1), enabling hydrogen‐bond formation. In this case the preferred orientation of the 2′‐OH hydrogen atom is approximately toward O5′(n + 1). We find two relatively short and dynamically stable types of backbone–backbone next‐neighbor contacts, namely C2′(H)(n)⋮O4′(n + 1) and C5′(H)(n + 1)⋮O2′(n). These interactions may affect both backbone rigidity and thermodynamic stability of RNA helical structures. © 1999 John Wiley & Sons, Inc. Biopoly 50: 287–302, 1999     

Dynamics in the U6 RNA intramolecular stem-loop: a base flipping conformational change

The U6 RNA intramolecular stem-loop (ISL) structure is an essential component of the spliceosome and binds a metal ion required for pre-messenger RNA splicing. The metal binding internal loop region of the stem contains a partially protonated C67-(+)A79 base pair (pK(a) = 6.5) and an unpaired U80 nucleotide that is stacked within the helix at pH 7.0. Here, we determine that protonation occurs with an exchange lifetime of approximately 20 micros and report the solution structures of the U6 ISL at pH 5.7. The differences between pH 5.7 and 7.0 structures reveal that the pH change significantly alters the RNA conformation. At lower pH, U80 is flipped out into the major groove. Base flipping involves a purine stacking interaction of flanking nucleotides, inversion of the sugar pucker 5' to the flipped base, and phosphodiester backbone rearrangement. Analysis of residual dipolar couplings as a function of pH indicates that base flipping is not restricted to a local conformational change. Rather, base flipping alters the alignment of the upper and lower helices. The alternative conformations of the U6 ISL reveal striking structural similarities with both the NMR and crystal structures of domain 5 of self-splicing group II introns. These structures suggest that base flipping at an essential metal binding site is a conserved feature of the splicing machinery for both the spliceosome and group II self-splicing introns.     

The RNA i-motif

Oligodeoxynucleotides with stretches of cytidine residues associate into a four-stranded structure, the i-motif, in which two head-to-tail, intercalated, parallel-stranded duplexes are held together by hemiprotonated C.C+ pairs. We have investigated the possibility of forming an i-motif structure with C-rich ribonucleic acids. The four C-rich RNAs studied, r(UC5), r(C5), r(C5U) and r(UC3), associate into multiple intercalated structures at acidic pH. r(UC5) forms two i-motif structures that differ by their intercalation topologies. We report on a structural study of the main form and we analyze the small conformational differences found by comparison with the DNA i-motif. The stacking topology of the main structure avoids one of the six 2'-OH/2'-OH repulsive contacts expected in a fully intercalated structure. The C3'-endo pucker of the RNA sugars and the orientation of the intercalated C.C+ pairs result in a modest widening of the narrow grooves at the steps where the hydroxyl groups are in close contact. The free energy of the RNA i-motif, on average -4 kJ mol(-1) per C.C+ pair, is half of the value found in DNA i-motif structures.     

Synthesis and biophysical properties of arabinonucleic acids (ANA): circular dichroic spectra, melting temperatures, and ribonuclease H susceptibility of ANA.RNA hybrid duplexes

Arabinonucleic acid (ANA), the 2'-epimer of RNA, was synthesized from arabinonucleoside building blocks by conventional solid-phase phosphoramidite synthesis. In addition, the biochemical and physicochemical properties of ANA strands of mixed base composition were evaluated for the first time. ANA exhibit certain characteristics desirable for use as antisense agents. They form duplexes with complementary RNA, direct RNase H degradation of target RNA molecules, and display resistance to 3'-exonucleases. Since RNA does not elicit RNase H activity, our findings establish that the stereochemistry at C2' (ANA versus RNA) is a key determinant in the activation of the enzyme RNase H. Inversion of stereochemistry at C2' is most likely accompanied by a conformational change in the furanose sugar pucker from C3'-endo (RNA) to C2'-endo ("DNA-like") pucker (ANA) [Noronha and Damha (1998) Nucleic Acids Res. 26, 2665-2671; Venkateswarlu and Ferguson (1999) J. Am. Chem. Soc. 121, 5609-5610]. This produces ANA/RNA hybrids whose CD spectra (i.e., helical conformation) are more similar to the native DNA/RNA substrates than to those of the pure RNA/RNA duplex. These features, combined with the fact that ara-2'OH groups project into the major groove of the helix (where they should not interfere with RNase H binding), help to explain the RNase H activity of ANA/RNA hybrids.     

Recognition of nucleic acid double helices by homopyrimidine 2', 5'-linked RNA.

We have studied the effect of a 2',5'-RNA third strand backbone on the stability of triple helices with a 'pyrimidine motif' targeting the polypurine strand of duplex DNA, duplex RNA and DNA/RNA hybrids. Comparative experiments were run in parallel with DNA and the regioisomeric RNA as third strands adopting the experimental design of Roberts and Crothers. The results reveal that 2',5'-RNA is indeed able to recognize double helical DNA (DD) and DNA (purine):RNA (pyrimidine) hybrids (DR). However, when the duplex purine strand is RNA and the duplex pyrimidine strand is DNA or RNA (i.e. RD or RR), triplex formation is not observed. These results exactly parallel what is observed for DNA third strands. Based on T m data, the affinities of 2',5'-RNA and DNA third strands towards DD and DR duplexes were similar. The RNA third strand formed triplexes with all four hairpins, as previously demonstrated. In analogy to the arabinose and 2'-deoxyribose third strands, the possible C2'- endo pucker of 2',5'-linked riboses together with the lack of an alpha-2'-OH group are believed to be responsible for the selective binding of 2',5'-RNA to DD and DR duplexes, over RR and RD duplexes. These studies indicate that the use of other oligonucleotide analogues will prove extremely useful in dissecting the contributions of backbone and/or sugar puckering to the recognition of nucleic acid duplexes.     

Determination of nucleic acid backbone conformation by 1H NMR.

In DNA or RNA duplexes, the six-bond C3'-O3'-P-O5'-C5'-C4'-C3' backbone linkage connecting adjacent residues contains six torsion angles (epsilon, zeta, alpha, beta, gamma, delta) but only four protons. This seriously limits the ability to define the backbone conformation by NMR using purely 1H-1H distance geometry (DG) methods. The problem is further compounded by the inability to assign two of the four backbone protons, namely the poorly resolved H5' and H5' protons, and invariably leads to DG structures with poorly defined backbone conformations. We have developed and tested a reliable method to constrain the beta, gamma, and epsilon (and indirectly alpha and zeta) backbone torsion angles by lower-bound NOE distances to unassigned H5'/H5' resonances combined with either 1H line widths or the conservative use of sigma J measurements; the method relies only on 1H 2-D NMR data, does not involve any structural assumptions, and leads to much improved backbone convergence among DG structures. The C4'-C5' torsion angle gamma is constrained by lower-bound NOE distances from H2' and from H6/H8 to any H5'/H5', as well as by sigma JH4, coupling measurements in the 3.9-4.4 ppm region; delta is constrained by H1'-H4' NOE distances and by H3'-H4' and H3'-H2' J couplings in COSY data; epsilon is partially constrained by H3' line width and/or further constrained by subtracting the minimum possible sigma JH3'-H from the observed sigma JH3' (COSY) to arrive at the maximum possible JH3'-P, which is then converted to H3'-P distance bounds. The angle beta is partially constrained via H5'-P and H5'-P distance bounds consistent with the maximum H5'-P and H5'-P J couplings derived from the observed H5' and H5' line widths, while alpha and zeta are indirectly constrained by lower distance bounds on the observed (n)H1' to (n + 1)H5'/H5' NOEs combined with the prior partial constraints on beta, gamma, delta, and epsilon. The combined effects of these additional constraints in determining distance geometry structures have been demonstrated using a 12-base duplex, [d(GCCGTTAACGGC)]2. Coordinate RMSDs per atom between structures refined with these constraints from random-embedded DG structures, from ideal A-DNA, and from B-DNA starting structures were less than 0.4 A for the central 8 base pairs indicating good convergence. All backbone angles for the central 8 base pairs are very well constrained with less than 10 degrees variation in any of the 48 torsion angles.