Identification of I:A Mismatch Base-pairing Structure in DNA*

containing deoxyinosine residues at positions corresponding to ambiguous nucleo- tides derived from an amino acid sequence been successfully used as hybridization probes. It is assumed that the hypoxanthine residue can make base pairs with multiple bases. In order to obtain direct evidence for 1:A base-pairing, a self-complementary deoxyoli- gonucleotide, d(G-G-I-A-C-C), was synthesized and its properties were examined by NMR spectroscopy. Three hydrogen-bonded imino proton resonances are observed at low temperatures in HzO suggesting the formation of a self-duplex with complete base pairing.

Deoxyoligonucleotides containing deoxyinosine residues at positions corresponding to ambiguous nucleotides derived from an amino acid sequence have been successfully used as hybridization probes. It is assumed that the hypoxanthine residue can make base pairs with multiple bases. In order to obtain direct evidence for 1:A base-pairing, a self-complementary deoxyoligonucleotide, d(G-G-I-A-C-C), was synthesized and its properties were examined by NMR spectroscopy. Three hydrogen-bonded imino proton resonances are observed at low temperatures in HzO suggesting the formation of a self-duplex with complete base pairing. Nuclear Overhauser effect (NOE) experiments showed that a signal at 15.1 ppm originated from the imino proton (Hl) of the dI residue (13) which is hydrogenbonded to the dA residue (A4). Both the I3 and A4 residues were assumed to have taken an anti glycosidic conformation since irradiating the H1 of I3 gave NOEs both to its own H2 and to that of A4, an NOE also being observed between the H2 protons of I3 and A4. Comparison of the slP NMR spectra of d(G-G-I-A-C-C) and d(G-G-I-C-C-C) showed the backbone structure of d(G-G-I-A-C-C) to have been disturbed by the presence of purine:purine base pairs in the middle of the hexamer duplex.
Deoxyribooligonucleotides containing deoxyinosine residues at ambiguous nucleotide positions have been successfully used as hybridization probes for the cDNA of a protein whose partial amino acid sequence is known (1, 2). The basis for this method is the deoxyinosine residue being expected to make base pairs with multiple bases. It is assumed that an inosine residue in the first position of the tRNA anticodon can form base pairs with C, A, and U of mRNA codons (3).
The thermal stabilities of deoxyribooligonucleotide duplexes, containing deoxyinosine residue matched with each of the four normal DNA bases, have been examined by optical melting techniques (4,5). The results of these studies suggest the order of stability to be: I:C > I:A > I:G = I:T. However, no direct evidence for these mismatched base pairs has been presented so far.
In the present paper, we report some NMR studies on the * This study was supported by a Grant-in-Aid for Specially Promoted Research from the Ministry of Education, Science, and Culture of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
3 To whom correspondence may be addressed. self-complementary deoxyribooligonucleotides, d(G-G-1-A-C-C) and d(G-G-I-C-C-C). The results reveal that the I(anti):A(anti) base pair, where both nucleoside residues take an anti glycosidic conformation, is indeed formed in the d(G-G-I-A-C-C) duplex and that the I(anti):A(anti) base-pairing conformation is similar to the G(anti):A(anti) conformation observed in solution (6,7); this is in contrast to the G(anti):A(syn) conformation found in crystals (8).
NMR experiments were performed on a JEOL GX500 spectrometer (500 MHz for 'H and 202 MHz for 31P). The 'H chemical shifts were determined relative to internal 2-methyl-2-propano1, for which in turn sodium 3-(trimethylsilyl)-propane-l-sulfonate had been used as its reference compound; the 31P chemical shifts were determined relative to external trimethyl phosphate (5% in ethanol). Proton spectra in HZO were obtained with a time-shared 1-1 pulse sequence (10) to suppress the strong H 2 0 signal, the nuclear Overhauser effect (NOE') difference spectrum representing the spectrum with an onresonance preirradiation pulse subtracted by the spectrum with an off-resonance preirradiation pulse. A single-frequency preirradiation pulse was applied for 0.3-0.5 s, giving an irradiated signal saturation of approximately 60%. The oligomer concentrations for the NMR samples were 205 Azso units/0.4 ml for d(G-G-I-C-C-C) (10 mM) and 168 AZso units/0.4 ml for d(G-G-I-A-C-C) (7 mM).

RESULTS
I:C Base Pair in d(G-G-I-C-C-C)-It is known that the hydrogen-bonded imino proton resonances of guanine, thymine, and uracil residues in nucleic acid duplexes are observed in the low field region (11-15 ppm downfield from trimethylsilylpropane sulfonate) of the proton NMR spectrum measured in H,O (11). As a reference compound, d(G-G-I-C-C-C) was first examined by NMR spectroscopy, its proton spectrum in H 2 0 showing a signal at 15.4 ppm for one proton and a signal with a shoulder at 13.0 ppm for two protons (Fig. lA).
Since d(G-G-I-C-C-C) contains one deoxyinosine (dI) and two deoxyguanosine (dG) residues, the appearance of signals for three protons in this region suggests that all the base residues are involved in the hydrogen bonding to form a completely base-paired duplex.
These imino proton resonances were assigned by NOE experiments as shown in Fig. 1, B and C. NOE arises from cross-relaxation between close protons and can be observed as a change in intensity of one signal upon irradiating another.
Irradiation of the signal at 15.4 ppm produces a large NOE to the signal at 7.8 ppm in addition to a pair af small NOEs at 6.6 and 7.6 ppm (Fig. 1B). This result indicates that the signals at 15.4 and 7.8 ppm belong to the N1-H (Hl) and C2-H (Hz) of the dI residue since these two protons are very close to each other (see Fig. 6). A similar chemical shift (15.  Fig. 2. Three clearly resolved resonances were observed, suggesting that this hexamer also forms a duplex with complete base pairing. The broad peak which appeared in this region's highest field can be assigned to the H1 of the terminal dG residue (Gl) when taking the fraying effect into consideration (14,15).
These imino proton resonances were assigned unambiguously by NOE experiments at 1 "C ( Fig. 3). Irradiation of the signal at 15.1 ppm produced large NOE peaks at 7.8 and 8.1 ppm (Fig. 3B) while irradiation of the other two signals resulted in pairs of small NOE peaks (Fig. 3, C and D). These results clearly show that the 15.1 ppm resonance belongs to I3-H1, and that the 7.8 ppm signal with a larger NOE is assigned to I3-H2. The signal at 8.1 ppm can be assigned to either A4-H2 or A4-H8. The imino proton resonance of dG in the lower field was assigned to G2-H1 since a very small NOE was observed on irradiating 13-H1. The irradiation of G2-H1 gave rise to weak inter-base pair NOE peaks at I3-H2 and A4-H2 (Fig. 3C).
I:A Base-paired Duplex Conformation-The assignment for A4-HZ and 13-HZ resonances were made by NOE experiments between nonexchangeable protons measured in D20 (Fig. 4). It is known that the H8 of the purine residue and the H6 of the pyrimidine residue in Aand B-form double helices have NOEs with both their own H1' and the H1' of the 5'-adjacent residue (16,17) and consequently that all the H8/H6 and H1' resonances can be connected along the oligonucleotide chain (Fig. 4A). The NOE difference spectra for H8 and H2 of I3 and A4 are shown in Fig. 4
suggesting a B-form type structure (Fig. 4, D and E ) . However, the irradiation of A4-HZ and I3-H2 signals produced almost no NOE in the sugar proton region, although resulting in a strong NOE on each other's H2 (Fig. 4, B and C). A very smdi NOE was observed between A4-H2 and C5-Hl' (Fig.   4B). These results not only support the H2 assignments but also suggest an anti glycosidic conformation for the 13 and A4 residues. Therefore, it was concluded that the I(anti):A(anti) base pair is indeed formed in the d(G-G-I-A-C-C) duplex.
In order to see the phosphodiester backbone conformation, the hexamers' 31P NMR spectra were measured (Fig. 5). The d(G-G-I-A-C-C) duplex spectrum showed a more dispersed pattern of signals and a profoundly downfield-shifted signal when compared with that of the d(G-G-I-C-C-C) duplex. This result indicates that considerable change in the phosphodiester backbone takes place to accomodate the purine:purine base-pairing structure in the DNA duplex.

DISCUSSION
A study of molecular model building shows the C1'-C1' distance for the paired residues in the I(anti):A(anti) structufe to be 12.7 A (Fig.6), with that of the I:C pair being 10.6 A.
The larger distance between the paired residues will disturb the ordinary DNA double helix backbone which consists of only purine:pyrimidine base pairs, the expected distortion being reflected in a dispersed signal pattern and the appearance of a profoundly downfield-shifted resonance in the 31P NMR spectrum (Fig. 5). Similar phenomena are to be ob-  (6); it would seem that the G:A base pair is rather unstable. For d(C-G-A-G-A-A-T-T-C-G-C-G), the G:A imino proton resonance disappears earlier than in the terminal G:C pair, upon raising the temperature. It would appear that the I A base pair in the d(G-G-I-A-C-C) duplex is relatively stable since the Gl-Hl resonance disappears earlier than the 13-H1 resonance (Fig. 2).
In contrast to results in solution, a G(anti):A(syn) base pair of the Hoogsteen type is found in d(C-G-C-G-A-A-T-T-A-G-C-G) duplex crystals (8). The Cl'-Cl' separation in this conformation is 10.7 A, close to that for standard Watson-Crick base pairs. A uniform cylinder-like structure for the double helix may be favorable for crystal packing, and moreover, the relatively unstable nature of the G(anti):A(anti) base pair may allow the duplex to adopt the alternative structure.
The present results may contribute to a better understanding of the mutations induced by nitrous acid and its equivalent in the environment. with a computer-drawn figure of the I:A base pair, and Dr. Yoshimasa Kyogoku for valuable discussions. We thank Dr. D. S. Jones for reading the manuscript.