Structures of DNA duplexes containing O6-carboxymethylguanine, a lesion associated with gastrointestinal cancer, reveal a mechanism for inducing pyrimidine transition mutations.

N-nitrosation of glycine and its derivatives generates potent alkylating agents that can lead to the formation of O(6)-carboxymethylguanine (O(6)-CMG) in DNA. O(6)-CMG has been identified in DNA derived from human colon tissue, and its occurrence has been linked to diets high in red and processed meats. By analogy to O(6)-methylguanine, O(6)-CMG is expected to be highly mutagenic, inducing G to A mutations during DNA replication that can increase the risk of gastrointestinal and other cancers. Two crystal structures of DNA dodecamers d(CGCG[O(6)-CMG]ATTCGCG) and d(CGC[O(6)-CMG]AATTCGCG) in complex with Hoechst33258 reveal that each can form a self-complementary duplex to retain the B-form conformation. Electron density maps clearly show that O(6)-CMG forms a Watson-Crick-type pair with thymine similar to the canonical A:T pair, and it forms a reversed wobble pair with cytosine. In situ structural modeling suggests that a DNA polymerase can accept the Watson-Crick-type pair of O(6)-CMG with thymine, but might also accept the reversed wobble pair of O(6)-CMG with cytosine. Thus, O(6)-CMG would permit the mis-incorporation of dTTP during DNA replication. Alternatively, the triphosphate that would be formed by carboxymethylation of the nucleotide triphosphate pool d[O(6)-CMG]TP might compete with dATP incorporation opposite thymine in a DNA template.

INTRODUCTION

Diets high in red and especially processed meats are known to be risk factors of colorectal cancer, which is one of the most common cancers worldwide (Globocan 2008 Cancer fact sheet, http://globocan.iarc.fr/factsheets/cancers/colorectal.asp). Promutagenic lesions O6-methylguanine (O6-MeG) and O6-carboxymethylguanine (O6-CMG) (Figure 1a) are commonly found in colorectal DNA, and their frequency might be indicative of a risk factor of colorectal cancer (1–3). One route that can lead to the formation of such lesions involves the initial nitrosation of amino acids, such as glycine and derivatives thereof, e.g. N-glycyl-peptides and the bile acid conjugate glycocholic acid. Nitrosation derives from reaction at neutral or alkaline pH with dinitrogen trioxide (N2O3), which in turn is generated by the oxidation of NO (4), from dietary nitrite or after exposure to ionizing radiation (5). N-Nitrosoglycine is converted into diazoacetate or α-lactone (6,7), potent mutagens that can alkylate guanine in DNA to form O6-CMG and O6-MeG (8). In humans, O6-methylguanine-DNA methyltransferase (MGMT) repairs DNA containing a wide variety of different O6-alkylguanine lesions by transferring the alkyl group to the thiolate side chain of the active site Cys (9). Recently, we have shown for the first time that DNA containing O6-CMG is also a substrate for MGMT (10). In vivo and in vitro evidence suggests that O6-CMG predominantly induces G:C→A:T transition mutations (3,11), implying that O6-CMG within a DNA template not only directs the incorporation of complementary dCTP but also allows the mis-incorporation of non-complementary dTTP into the newly synthesized DNA. In general, DNA polymerases accept only the Watson–Crick–type pairs in the B-form conformation (12–14). To understand the mechanism of such mutations, it is necessary to reveal the pairing or interaction geometry of O6-CMG with T and C bases.

Figure 1.

Chemical structures of O6-CMG with the atomic numbering and CMG (a), and the sequences and numbering schemes of O6-CMG-containing DNA duplexes (b). X indicates O6-CMG.

Here, we describe the crystal structures of DNA duplexes containing O6-CMG at residue positions that place the modified base opposite T or C in the palindromic B-form Dickerson–Drew sequence d(CGCGAATTCGCG) (15).

MATERIALS AND METHODS

Oligodeoxyribonucleotide synthesis and purification

Oligodeoxyribonucleotides (ODNs) with the sequences d(CGCG[O6-CMG]ATTCGCG) and d(CGC[O6-CMG]AATTCGCG) were synthesized and purified by reversed-phase HPLC as described previously (16), and they were characterized by ESI–mass spectrometry. For crystallization, the samples in pure water were purified on an ÄKTAprime plus (GE Healthcare) using a Superdex 30 pg 16/60 column at flow rate of 0.5 ml/min with a gradient of 0–100% of pH 7.2 buffer solution (50 mM NaH2PO4 and 150 mM NaCl); the ODN-containing fractions monitored by a UV monitor were confirmed by PAGE analysis with TBE. Finally, the ODNs were desalted by a series of C18 (Waters Corp.), AG50W-X8 (BioRad Co.) and Chelex 100 (BioRad Co.) columns, in turn. The eluted solutions were dried in vacuo at room temperature to store the samples.

Crystallization and data collection

Initial screenings of crystallization conditions were performed at 277 K by the hanging-drop vapor diffusion method using a kit for nucleic acids reported by Berger et al. (17). Two-microliter droplets were equilibrated against 700 μl of the reservoir solution. The optimized conditions for obtaining O6-CMG5T and O6-CMG4C crystals were as follows. For O6-CMG5:T, a droplet of 40 mM sodium cacodylate buffer solution at pH 7.0 containing 1 mM ODN, 10% (v/v) 2-methyl-2,4-pentanediol (MPD), 12 mM spermine tetrahydrochloride, 80 mM sodium chloride, 12 mM potassium chloride, 20 mM magnesium chloride and 1 mM Hoechst33258 (2′-(4-hydroxyphenyl)-6-(4-methyl-1-piperazinyl)-2,6′-bi-1H-benzimidazole) was equilibrated against 35% (v/v) MPD. For O6-CMG4C, a droplet of 40 mM sodium cacodylate buffer solution at pH 7.0 containing 1 mM ODN, 10% (v/v) MPD, 12 mM spermine tetrahydrochloride, 40 mM lithium chloride, 80 mM strontium chloride and 1 mM Hoechst 33258 was equilibrated against 35% (v/v) MPD.

O6-CMG5T and O6-CMG4C crystals suitable for X-ray data collections were picked up from their droplets with a nylon loop (Hampton Research) and transferred into liquid nitrogen. X-Ray diffraction experiments of these crystals were performed at 100 K with synchrotron radiation (λ = 1.00 Å at BL-5 A and 0.98 Å at BL-17A) of the Photon Factory in Tsukuba (Japan). Diffracted intensities were recorded on a CCD detector Quantum 315r positioned 200.0 and 155.4 mm from O6-CMG5T and O6-CMG4C crystals, respectively. A total of 180 frames of the patterns for one crystal were taken at 1° oscillation steps with 1 s exposure per frame. Raw diffraction images were indexed, and intensities around Bragg spots were integrated using the computer programs, HKL2000 (18) for O6-CMG5T data and Mosflm(19)-Scala (20) of the CCP4 suite (21) for O6-CMG4C data. To compensate for the overloaded reflections, the intensity data were merged with those collected at different exposure doses. The crystal data and statistics of data collection are summarized in Table 1.

Table 1.

Crystal data and statistics of data collection and structure refinement

Crystal code	O6-CMG5T	O6-CMG4C
Crystal data
Space group	P212121	P212121
Unit cell (Å)
a	25.1	24.6
b	40.4	41.2
c	65.7	63.1
Data collection
Resolution range (Å)	50.0–1.94	25.1–1.54
Outer shell (Å)	1.97–1.94	1.62–1.54
Observed reflections	27 766	62 239
Unique reflections	4877	9559
Completeness (%)a	91.0 (82.9)	95.8 (100)
Rmerge (%)b	5.7 (52.2)	8.9 (28.5)
I/σ	8.1 (0.7)	14 (6.4)
Redundancyc	5.7 (1.4)	6.5 (6.7)
Structure refinement
Resolution range (Å)	34.4–1.94	25.1–1.54
Used reflections	3842	9017
Rfactor (%)d	23.9	19.4
Rfree (%)e	26.7	23.6
Rms deviation
Bond distances (Å)	0.05	0.03
Bond angles (degree)	4.3	3.8
No. of DNA duplex	1	1
No. of Hoechst33258	1	1
No. of ions	1Mg2+	3Sr2+
No. of water molecules	43	135

aValues in parentheses indicate those in the outer shell.

bRmerge = 100 × Σ|I – | / Σ, where I is the jth measurement of the intensity of reflection h and is its mean value.

cDiffraction patterns of 1° oscillation ranges were collected in 180 frames of X5T and X4C.

dR-factor = 100 × Σ||Fo| – |Fc|| / Σ|Fo|, where |Fo| and |Fc| are the observed and calculated structure factor amplitudes, respectively.

eCalculated using a random set containing 5% observations that were not included during refinement (26).

Crystal data and statistics of data collection and structure refinement

aValues in parentheses indicate those in the outer shell.

bRmerge = 100 × Σ|I – | / Σ, where I is the jth measurement of the intensity of reflection h and is its mean value.

cDiffraction patterns of 1° oscillation ranges were collected in 180 frames of X5T and X4C.

dR-factor = 100 × Σ||Fo| – |Fc|| / Σ|Fo|, where |Fo| and |Fc| are the observed and calculated structure factor amplitudes, respectively.

eCalculated using a random set containing 5% observations that were not included during refinement (26).

Structure determination and refinement

Using the program autoMR in the CCP4 suite (21), the phases of the two data sets were separately estimated by the molecular replacement method with the unmodified ODN structure d(CGCGAATTCGCG) (22) as a probe. The atomic parameters were refined using the maximum-likelihood least-squares technique in REFMAC5 (23) of CCP4 and CNS (24). The crystal structures were constructed and modified by adding other molecules and ions using the program Coot (25) in CCP4. The resultant structures were validated by interpretation of OMIT maps at every nucleotide residue. Electron densities assignable to a magnesium ion and three strontium ions were found in O6-CMG5T and O6-CMG4C, respectively, and these cations were included in the subsequent refinements. The structural restraints applied initially on DNA and Hoechst33258 were released. The Rfactor and Rfree values converged with further rounds of the structure refinements.

The statistics of structure refinements are summarized in Table 1. The atomic parameters of O6-CMG5T and O6-CMG4C crystal structures were deposited in the protein data bank (PDB-IDs = 4ITD and 4IJ0). Figure 2 shows the 2Fo-Fc electron density maps of the modified nucleotide and its partner in the pair formation, together with the corresponding Fo-Fc omit maps calculated without the pairs. These maps were depicted by the program Dino (27). All the residues were traced on the electron density maps. The O6-CMG residues assigned on Fo-Fc omit maps of O6-CMG5T and O6-CMG4C fit well into the 2Fo-Fc maps. All the global and local helical parameters, as well as the torsion angles and pseudo-rotation phase angles of sugar rings, were calculated using the program 3DNA (28). Some of them are given in Table 2.

Figure 2.

Fo-Fc omit maps (in the left column) of O6-CMG5:T20 pairs and T8:O6-CMG17 pairs (a), and 2Fo-Fc maps (in the right column) of O6-CMG5:T20 pairs and T8:O6-CMG17 pairs (b). The corresponding maps of O6-CMG4:C21 pairs and C9:O6-CMG16 pairs (c) and those of O6-CMG4:C21 pairs and C9:O6-CMG16 pairs (d). Broken lines show possible hydrogen bonds. Electron densities are contoured at the 2.6, 3.0, 2.5 and 2.0 σ levels for the top-to-bottom pairs in the left column, and the corresponding 2Fo-Fc maps are contoured at the 0.8, 0.8, 1.0 and 1.0 σ levels in the right column.

Table 2.

Average local helical parameters and base pair parameters

	X-Displacement (Å)	Inclination (°)	Helical twist (°)	Helical rise (Å)
O6-CMG5T	0.57	−0.26	36	3.3
O6-CMG4C	0.67	0.73	36	3.3
B-DNAb	0.05	2.1	36.5	3.3
A-DNAb	−4.2	15	32.5	2.8

aCalculated with the program 3DNA (29)

bHigh-resolution A- and B-form DNA structures by Olson et al. (30)

Fo-Fc omit maps (in the left column) of O6-CMG5:T20 pairs and T8:O6-CMG17 pairs (a), and 2Fo-Fc maps (in the right column) of O6-CMG5:T20 pairs and T8:O6-CMG17 pairs (b). The corresponding maps of O6-CMG4:C21 pairs and C9:O6-CMG16 pairs (c) and those of O6-CMG4:C21 pairs and C9:O6-CMG16 pairs (d). Broken lines show possible hydrogen bonds. Electron densities are contoured at the 2.6, 3.0, 2.5 and 2.0 σ levels for the top-to-bottom pairs in the left column, and the corresponding 2Fo-Fc maps are contoured at the 0.8, 0.8, 1.0 and 1.0 σ levels in the right column.

Average local helical parameters and base pair parameters

aCalculated with the program 3DNA (29)

bHigh-resolution A- and B-form DNA structures by Olson et al. (30)

RESULTS AND DISCUSSION

Overall structure of DNA duplexes

At the initial stage of crystallization condition survey, it was difficult to crystallize the two ODNs. However, by adding Hoechst33258 as a dye to stabilize (30) the duplex formation in ODN solutions, single crystals suitable for X-ray analyses were obtained. As shown in Figure 3, the two homododecamers in both O6-CMG5T and O6-CMG4C crystals are associated with each other to form a right-handed double helix. Their average local helical parameters (Table 2) are close to those of the high-resolution B-form DNA duplexes (22,30,31). However, superimposition onto the unmodified structures (Figure 3c) reveals the local variations in the backbone conformations of the duplexes, the root-mean-square (rms) deviations being 1.2 Å for O6-CMG5T and 1.4 Å for O6-CMG4C. Although the largest deviations occurs at the O6-CMG residues in both O6-CMG5T and O6-CMG4C duplexes, their sugar puckers fluctuate around the C2′-endo conformation, which is the conformation typically found in B-type DNA. These data indicate that carboxymethylation of guanine residues does not significantly affect the overall DNA conformation. The slightly large rms deviation of O6-CMG4C may be related to the interaction geometry between O6-CMG and C, which will be described later in detail. Figure 3 shows Hoechst33258 molecules bound in the minor grooves of O6-CMG5T and O6-CMG4C. They seem to stabilize the duplex structures with no drastic changes in the base pair geometry, as the modified sites are in the major groove. Similar examples are already found in other structures of DNA duplexes crystallized in the presence of this duplex-stabilizing dye (32,33). Only a slight change is found in the groove width; the central shortest P … P distance between the two ODN backbones is 8.9 Å in the absence of Hoechst33258 (23), whereas those of O6-CMG5T and O6-CMG4C are 9.4 and 9.9 Å, respectively. Such expansions ∼1.0 Å are commonly found in the referred examples (32,33).

Figure 3.

Overviews of O6-CMG5T duplex (a) and O6-CMG4C duplex (b) and their superimposition onto the unmodified duplex (black) (c). In (a) and (b), the O6-CMG residues and Hoechst33258 molecules are colored in red and green, respectively. In (c), O6-CMG5T, O6-CMG4C and the unmodified duplexes are colored in red, green and black, respectively, and the modified residues are in blue.

Overviews of O6-CMG5T duplex (a) and O6-CMG4C duplex (b) and their superimposition onto the unmodified duplex (black) (c). In (a) and (b), the O6-CMG residues and Hoechst33258 molecules are colored in red and green, respectively. In (c), O6-CMG5T, O6-CMG4C and the unmodified duplexes are colored in red, green and black, respectively, and the modified residues are in blue.

Formation of O6-CMG:T and O6-CMG:C base pairs

All unmodified base pairs in the dodecamer sequences except those containing O6-CMG are of the standard Watson–Crick type. Fo-Fc omit and 2Fo-Fc electron density maps of the O6-CMG5T duplex indicate that the O6-CMG residues form base pairs with the two opposite T residues in the palindromic sequence (Figure 2a and b). The shapes of the densities of these base pairs except the carboxymethyl groups are similar to the canonical Watson–Crick A:T pair found in the original Dickerson–Drew dodecamer (16). On the other hand, although Fo-Fc omit maps and 2Fo-Fc electron density maps of the O6-CMG4C duplex (Figure 2c and d) show that the O6-CMG residues form base pairs with the opposite C residues, the two geometries of the O6-CMG:C pairs are different from the standard G:C pair geometry found in the unmodified dodecamer.

Geometries of the O6-CMG:T and O6-CMG:C base pairs

The interaction geometries of base paired O6-CMGs are shown in Figure 4 and summarized in Table 2. In the O6-CMG5:T20 pair, the interatomic distances between the N1 atom of O6-CMG and the N3 atom of T and between the N2 atom of O6-CMG and the O2 atom of T are 3.3 and 3.1 Å, respectively. In the O6-CMG17:T8 pair, the corresponding distances are 3.4 and 3.0 Å, respectively. These values suggest base pair formation between O6-CMG and T. Thus, although the natural base G forms a wobble base pair with T (34), the O6-CMG:T pair is similar to the canonical Watson–Crick A:T pair. In this O6-CMG:T pair, although the electronegative O6 atom of O6-CMG is exposed to the O4 atom of T, the ensuing repulsion between these two sites is reduced by propeller-twisting between the paired bases. The twisting angles are −18° on average at the two sites. The N1(O6-CMG) … N3(T) distances look slightly longer, perhaps because of the O6(O6-CMG) … O4(T) repulsion, which separates the O6 and O4 atoms at an average distance of 3.6 Å. The interaction geometries of O-CMG:T pairs are similar to those of O-MeG:T pairs found in the O-MeG-containing B-DNA (35). As the carboxymethyl groups of the O6-CMG residues protrude into the major groove, they do not drastically alter the overall DNA conformation. However, as shown in Figure 2, the electron densities of the terminal carboxyl groups are not clear, suggesting that they are disordered in the solvent region.

Figure 4.

Watson–Crick–type pairs between O6-CMG and T (left column) and reversed wobble pairs between O6-CMG and C (right column). Broken lines suggest possible hydrogen bonds, and values indicate atomic distances in Å. The corresponding chemical structures of interacting O6-CMGs are shown in the bottom. The carboxyl groups of O6-CMG:T pairs are invisible because of disorder. Character W indicates a water oxygen atom.

Watson–Crick–type pairs between O6-CMG and T (left column) and reversed wobble pairs between O6-CMG and C (right column). Broken lines suggest possible hydrogen bonds, and values indicate atomic distances in Å. The corresponding chemical structures of interacting O6-CMGs are shown in the bottom. The carboxyl groups of O6-CMG:T pairs are invisible because of disorder. Character W indicates a water oxygen atom.

In the O6-CMG4:C21 pair, the interatomic distances between the N1 atom of O6-CMG and the N4 atom of C and between the N2 atom of O6-CMG and the N3 atom of C are 3.0 and 2.9 Å, respectively. In the O6-CMG16:C9 pair, the corresponding distances are the same 3.1 and 2.9 Å, respectively. These values indicate that O6-CMG can form a base pair with C through hydrogen bonds at these two sites. In both base pairs, the purine moiety of O6-CMG moves towards the major groove side, whereas C remains by its original position (Figure 5). This deformation occurs at both modified sites.

Figure 5.

A comparison of paired-base positions when O6-CMG4C and the unmodified duplexes are superimposed between the phosphate backbones. An arrow indicates that only the modified base wobbles towards the major groove side.

A comparison of paired-base positions when O6-CMG4C and the unmodified duplexes are superimposed between the phosphate backbones. An arrow indicates that only the modified base wobbles towards the major groove side.

Although this pairing mode has been referred to as a wobble base pair (36,37), we refer to the O-CMG:C pair as a reversed wobble pair, as in the typical wobble pair, G moves to the minor groove, and U or T shifts to the major groove. Such a reversed wobble pair has been found between G and 5-formyluracil, an analog that derives after oxidation of T with oxygen radicals (38). Another example is found in an ODN-containing O6-ethylguanine (O-EtG) at the fourth position of the Dickerson sequence, where one of the two base pairs is of the reversed wobble type (39,40). In many cases, this wobbling makes the C1′ … C1′ distance longer by 1.0 Å and the λI and λII angles asymmetric (Table 2), as compared with those of the unmodified pairs.

The carboxyl groups of O6-CMG (in its pairing with C) are clearly visible on the high resolution maps, as shown in Figure 2. They adopt a syn conformation against the N1 atom of O-CMG. In contrast, the alkyl groups of O-EtG:C pairs (39,40) adopts an anti conformation and protrudes into the major groove. The longer carboxymethyl groups of O6-CMG protrude into the major groove to make contacts with water molecules. The reversed wobble O6-CMG:C pairs seem to be further stabilized by two additional water-mediated hydrogen bonds between an oxygen atom of the carboxyl group of O6-CMG and the N4 atom of C and between the N2 atom of O6-CMG and the O2 atom of C. Three Sr atoms are all hydrated with eight water molecules and bound to the minor groove, major groove and phosphate backbone, respectively, so that they do not influence the base pairs formed by O6-CMG. The same preference is also seen for the modified base in the O6-CMG:C base pair. The change in preference for the orientation of the alkyl group in O6-CMG compared with O6-MeG may result from additional interactions of the carboxyl group of the alkyl side chain with the O4 and N4 atoms of T and C, respectively, that are implied from the crystal structures.

Biological implications

From the present study, it has been found that O6-CMG can form base pairs with both thymine and cytosine, and the pairing modes are Watson–Crick type and reversed wobble type, respectively. To examine a possibility if the two pair formations of O6-CMG:T and O6-CMG:C are acceptable in DNA polymerase, these pairs were incorporated into the active sites of human DNA polymerase η in complex with DNA (41), using the X-ray structure of PDB-ID = 4ED8. The plausible models were energetically refined by the computer program REFMAC5 using the X-ray intensity data of 4ED8 that were downloaded from PDB and truncated at low resolution (5 Å). As shown in Figure 6 and Supplementary Data (see the next page), the in silico structural models suggest that the Watson–Crick–type pair of O6-CMG:T can be accommodated in the template site, consistent with this damaged base being able to induce pyrimidine transition mutations. The reversed wobble pair of O6-CMG:C can also potentially be incorporated into the active site but would require a slight rotation of the base pair.

Figure 6.

In silico structural models of human DNA polymerase η (41) in complex with B-DNA containing O6-CMG paired with dTTP (a) and dCTP (b). In the minor groove, the hydrophilic Gln39 forms a hydrogen bond with the template bases. In addition, hydrophobic amino-acid residues (Phe18, Ile48, Leu89, Tyr92 and Ile114) are packed closely to form a pocket for the paired bases so that there is no space to accept any modification of the paired bases. In the major groove side, however, there is a widely opened space for modified bases. Broken lines indicate possible hydrogen bonds. The viewing directions are slightly different between (a) and (b).

In silico structural models of human DNA polymerase η (41) in complex with B-DNA containing O6-CMG paired with dTTP (a) and dCTP (b). In the minor groove, the hydrophilic Gln39 forms a hydrogen bond with the template bases. In addition, hydrophobic amino-acid residues (Phe18, Ile48, Leu89, Tyr92 and Ile114) are packed closely to form a pocket for the paired bases so that there is no space to accept any modification of the paired bases. In the major groove side, however, there is a widely opened space for modified bases. Broken lines indicate possible hydrogen bonds. The viewing directions are slightly different between (a) and (b).

DNA replication relies on cognate Watson–Crick–type base pair formation in the active site of a DNA polymerase (13–15). Typically, there is not enough space for a wobble type or other non-complementary base pairs. In addition, as the polymerase is bound in the minor groove of DNA, extrusion of the carboxymethyl group into the major groove is unlikely to interfere with binding to the DNA polymerase or with nucleotide incorporation opposite the damaged base. Taking the Watson–Crick–type O6-CMG:T and the reversed wobbling-type O6-CMG:C pairings into consideration, it is deduced that when O6-CMG is in the template, it can accept a thymine and, to a much lesser extent, a cytosine residue in the newly synthesized DNA.

Based on these two cases of such mis-incorporations, three possible routes of pyrimidine transition at the modified G site could be proposed as shown in Figure 7. In the case that the template strand is damaged, the original G:C pair can be replaced with an A:T pair. In the first replication, a thymine residue is introduced in the daughter strand by accepting both dTTP and dCTP, and then the synthesized strand is used as a second template. In the second replication, dATP is bound against the mutated thymine residue. After two steps of replication, a pyrimidine transition mutation can be achieved.

Figure 7.

Three possible schemes of pyrimidine transition mutations. A template guanine base is O-carboxymethylated (X = O-CMG) in the left box. In the right box, dGTP is Ocarboxymethylated (dXTP) to be incorporated opposite T and C residues, respectively.

Three possible schemes of pyrimidine transition mutations. A template guanine base is O-carboxymethylated (X = O-CMG) in the left box. In the right box, dGTP is Ocarboxymethylated (dXTP) to be incorporated opposite T and C residues, respectively.

Alternatively, the triphosphate that would be formed by carboxymethylation of the nucleotide triphosphate pool d[O6-CMG]TP might compete with dATP incorporation opposite thymine in a DNA template. Once d[O6-CMG]TP could pair with a T residue in a template strand, leading to the insertion of A in the opposite strand, not only C but also T will be introduced opposite the template O6-CMG residue in the second replication. At the third replication, the incorporated C residue directs the insertion of G in the opposite strand. After three cycles of DNA replication at least, the pyrimidine transition mutation will be completed. Another case is when the d[O6-CMG]TP residue is initially paired opposite C in a template strand and then the introduced O6-CMG residue accepts a T residue. In the third cycle, an A is inserted opposite the T residue. Through the three cycles, the original G:C pair is converted to an A:T pair.

CONCLUSION

In this study, we have determined the crystal structures of two O6-CMG-containing DNA duplexes. The carboxymethylated guanine base can form a Watson–Crick–type pair with T (in the O6-CMG5T crystal) and a reversed wobble pair with C (in the O6-CMG4C crystal). In silico structural modeling suggests that both the Watson–Crick–type O6-CMG:T and the reversed wobble-type O6-CMG:C pairing modes, found in the present study, could be accepted by the DNA polymerase. In other words, O6-CMG residues in a damaged DNA template would direct the incorporation not only of the complementary dCTP but also of the non-complementary dTTP into the newly synthesized DNA strand. Finally, we conclude that the G:C→A:T transition mutations, demonstrated by in vivo and in vitro experiments (3,12) as a factor in the etiology of gastrointestinal cancer, likely occur as a consequence of the Watson–Crick–type pairing of O6-CMG with T.

ACCESSION NUMBERS

4ITD and 4IJ0.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online: Supplementary Figures 1–2.

FUNDING

Biotechnology and Biological Sciences Research Council, UK (to O.J.W., D.M.W.) (in part); Cancer Research-UK (to G.P.M.) (in part); Engineering and Physical Sciences Research Council, UK (to C.L.M., D.M.W.) (in part).

Conflict of interest statement. None declared.