The 12-mer oligonucleotide containing the TGGT sequence (Figure 1) was platinated according to our previously described protocols and purified using HPLC (described in Methods). Collection of NOESY and 2D DQF-COSY spectra (Figures S1 and S2), and chemical shift assignments for OX-DNA and undamaged DNA duplex in the TGGT sequence context (Tables S1 and S2) were performed essentially as described previously for the OX-DNA, CP-DNA adducts and undamaged DNA in the AGGC sequence context [45], [46] (see Methods). The average solution structures of OX-DNA and undamaged DNA in the TGGT sequence context were computed from the NMR data essentially as described previously for OX-DNA, CP-DNA and undamaged DNA in the AGGC sequence context [45], [46] (details given in Methods. Coordinates and NMR restraints of the structures of OX-TGGT adduct and undamaged DNA have been deposited in the Protein Data Bank with accession numbers 2k0t (13 lowest energy structures of OX-DNA), 2k0u (average structure of OX-DNA calculated from the 20 lowest energy structures) and 2k0v (average structure of undamaged DNA calculated from the 20 lowest energy structures)). The stereo view of the average solution structures of the OX-TGGT adduct and undamaged TGGT DNA duplex are shown in Figure 2.

The similarity in the experimental procedures followed in this study and our previous work with CP-GG adducts, OX-GG adducts and undamaged DNA in the AGGC sequence context [45], [46] allow us to perform a systematic comparison of the key structural features of OX-GG adducts and undamaged DNA in two different sequence contexts. Overlays of the OX-TGGT adduct and undamaged TGGT DNA duplex with the corresponding structures in the AGGC sequence context are shown in Figure 3. Based on the RMSD values, sequence context (TGGT versus AGGC) has no major effect on the overall conformation of either undamaged DNA or OX-GG DNA adducts.

In order to probe the possibility that more subtle conformational differences between these DNA structures might exist we employed CURVES version 5.3 [54] to calculate the helical parameters for the central four base-pairs and the central two base-pair steps of our OX-GG and undamaged DNA NMR solution structures (Table S3). We then identified the helical parameters that showed statistically significant differences between distributions of the 14 lowest energy structures for each NMR data set (see Methods) on the basis of their Z-score [55] (described in Text S1). Pair-wise comparisons of OX-DNA and undamaged DNA in the TGGT and AGGC sequence contexts are shown in Figure 4 as a heat map of Z-scores (described in Text S1). The heat maps allowed rapid identification of the helical parameters that differ the most between any two structures, while the DNA helical parameters themselves (Table S3) provide a description of the conformational differences between the structures.

Figure 4A and Table S3 show the comparison of the OX-GG adducts with undamaged DNA in both the TGGT and AGGC sequence contexts. In both sequence contexts, the formation of the OX-GG adduct resulted in a significant increase in roll and dihedral angle at the G6–G7 base pair step compared to undamaged DNA (Figure 4A, Table S3). This is a common feature of all Pt-GG structures reported to date [41], [42], [43], [44], [45], [46], [47], [48], [49], [50] and is thought to be important for the recognition of Pt-GG DNA adducts by HMGB1 and other damage-recognition proteins (see Discussion).

The OX-TGGT adduct is also similar to all other Pt-GG adducts with respect to a significant changes of DNA helical parameters of bases on both the 5′ and 3′ side of the adduct compared to undamaged DNA (Figure 4A, Table S3). However, the exact nature of these distortions appears to be dependent on the sequence context of the adduct. For example, the base pair step on the 5′ side of the OX-TGGT adduct differs from undamaged TGGT DNA primarily in terms of a large negative roll, while the base pair step on the 3′ side the OX-TGGT adduct differs from undamaged TGGT DNA primarily in terms of tilt (Figure 4A, Table S3). In contrast, the base pair step on the 5′ side of the OX-AGGC adduct differed from undamaged DNA primarily in terms of a negative shift and slide, while on the 3′ side, the OX-AGGC adduct differed from undamaged AGGC DNA in primarily in terms of and slide (Figure 4A, Table S3).

Another approach to analyzing the sequence context effects on conformation is to compare the conformations of the OX-TGGT and OX-AGGC adducts directly without reference to the corresponding undamaged DNA structures (Figure 4B). In this comparison, the most significant conformational differences appear to be the slide of the base pair steps on both the 5′ and 3′ side of the OX-GG adduct. Because DNA sequence can influence the conformation of undamaged DNA also [51], we also compared undamaged DNA in the TGGT and AGGC sequence contexts (data not shown). While some conformational differences were apparent for undamaged DNA in the two sequence contexts (Table S3), they were not large enough to influence the direct comparison of OX-TGGT and OX-AGGC adducts. Thus, these data show that the conformational distortion on the 5′ and 3′ side of OX-GG adducts is significantly affected by the sequence context of the adduct (TGGT versus AGGC), which may be of importance in understanding the sequence specificity of recognition of Pt-DNA adducts by HMGB1 and other damage-recognition proteins (see Discussion).

The temperature dependence of the imino proton resonances of OX-GG adduct and undamaged DNA duplex (Figure 5) was monitored by 1D ¹H-NMR in H₂O buffer solution (100 mM NaCl, 5 mM Na₂HPO₄/NaH₂PO₄, pH 7.0) varying the temperature from 2°C to 45°C. For the OX-GG adduct in the TGGT sequence context, the imino signals of G6 and G7 disappear with increasing temperature while other imino peaks are observed even at 45°C. Moreover, the G6 imino peak for the OX-GG adduct in the TGGT sequence context shows a slightly faster solvent exchange rate relative to the G7 imino peak. These observations are consistent with previous reports for other Pt-GG adducts [41], [44], [45], [46], [47], and indicate that DNA containing both the CP and OX adducts is more solvent accessible on the 5′ side of the adduct than on the 3′ side, which suggests that the DNA may be more distorted and/or flexible on the 5′ side of the adduct.

In addition to the higher solvent exchange rate exhibited by the G6 and G7 imino protons, the T5 imino proton also displayed a fast exchange rate, which is comparable to the exchange rate shown by the G6 and G7 imino protons (Figure 5A). This feature was not observed for undamaged DNA (Figure 5B). A similar rapid exchange rate has previously been reported for the 5′T imino proton of a CP-TGGT DNA duplex [44]. The temperature dependence of the imino proton signals have previously been reported for CP-GG adducts in the CGGC sequence context [47] and for both CP- and OX-GG adducts in the AGGC sequence context [45], [46]. While the 5′ flanking bases do not possess imino proton signals, their complementary bases (G and T) in the opposing strand do possess imino signals; and no loss of signal from their imino protons was observed at increasing temperature. The fast solvent exchange rate seen for T in the 5′ T•A base pair, but not for either T in the 5′A•T base pair or G in the 5′ C•G base pair suggests that the 5′ T•A base pair in the TGGT sequence context is more solvent accessible than either the 5′ A•T or 5′ C•G base pairs in the AGGC and the CGGC sequence contexts. We hypothesize that the distortion and/or flexibility observed on the 5′- side of the Pt-GG adduct is extended to the 5′-flanking residue base-pair in the TGGT sequence context but not in the AGGC or CGGC sequence contexts. This difference in conformational flexibility of the 5′-flanking residue is consistent with the molecular dynamics simulations described in the next section and could influence the sequence context specificity of protein recognition of Pt-GG adducts (see Discussion).

In an effort to better understand the effect of sequence context on the conformational dynamics of Pt-DNA adducts, we performed multiple 10 ns all-atom MD simulations (see Methods for details) of CP-, OX-GG adducts and undamaged DNA in the TGGT sequence context and compared them to our earlier simulations in the AGGC and TGGA sequence contexts [52], [53]. The simulations attained equilibrium within the first few nanoseconds, with the all-atom mass weighted RMSD of undamaged DNA remaining less than 3 Å and the RMSD of CP-DNA and OX-DNA remaining less than 4 Å to the starting structure throughout the simulation (Figure S3). The equilibrium structures were independent of the starting structure and initial velocities indicating that the simulations were well equilibrated. The centroid structures determined from the undamaged DNA and the OX-DNA simulations had RMSDs of 1.8 and 2.8 Å relative to their corresponding NMR structures (Figure 6). Additional evidence that the MD simulations were congruent with the NMR structures is included in Supporting Data Text S1.

To explore the differences in conformational dynamics between the MD simulations for each structure, we calculated the helical parameters of the central four base-pairs and the central three base-pair steps for the ensembles of all the snapshots from each of the MD simulations using CURVES version 5.3 [56], [57]. We then compared the conformational dynamics of these simulations by constructing histograms of the helical parameters that showed statistically significant differences between distributions of any two ensembles (Figure S4) as determined using Kolmogorov-Smirnov (KS) ratio [55] (Text S1). From these comparisons two distinct effects of sequence context were observed. In the first case, formation of the OX-AGGC adduct, but not the OX-TGGT adduct, induced a significant change in conformational dynamics compared to undamaged DNA (Figure S4, Panels A and B), which tended to occur primarily on the 5′ side of the adduct. In the second case, formation of the OX-TGGT adduct, but not the OX-AGGC adduct induced a significant change in the conformational dynamics compared to undamaged DNA (Figure S4, Panels C and D). These differences in conformational dynamics were more subtle and tended to occur on the 3′ side of the adduct. The inclusion of undamaged DNA in the analysis was valuable because it allowed us to exclude differences in conformational dynamics between the OX-TGGT and OX-AGGC adducts that were primarily due to the effect of sequence context on undamaged DNA (Figure S4, Panels E and F).

We have previously reported the formation of hydrogen bonds between Pt-amines and adjacent bases in our simulations of CP-DNA and OX-DNA adducts in both the AGGC and TGGA sequence contexts [52], [53]. We see a similar occurrence on the 3′ side of CP-DNA and OX-DNA adducts the TGGT sequence context, with hydrogen bonds observed between the 3′ Pt-amine and either the O6 atom of the 3′ Guanine (the G7-O6 hydrogen bond) or the O4 atom of 3′ Thymine (the T8-O4 hydrogen bond) (Figure 7). No significant hydrogen bond formation was observed between the Pt-amines and bases on the 5′ side of the CP- or OX-TGGT adducts. Even though similar hydrogen bonds are observed with both CP- TGGT and OX-TGGT DNA, the frequency of hydrogen bond formation is different for the CP- and OX-DNA adducts. The G7-O6 hydrogen bond is formed more frequently by OX-DNA (72% of the time compared to 32% for CP-DNA, Table 1), while the T8-O4 hydrogen bond is formed more frequently by CP-DNA (49% compared to 19% for OX-DNA).

The characterization of the hydrogen bonds between platinum amines and the adjacent bases is of importance because each of those hydrogen bonds is associated with minor DNA conformations that may influence protein recognition of Pt-DNA adducts [52]. For example, we have shown that minor DNA conformations associated with hydrogen bond formation between the platinum amines and adjacent bases were consistent with the slight preferential recognition of CP-GG adducts by HMGB1a in the AGGC sequence context [52] and the very strong preferential recognition of CP-GG adducts by HMGB1a in the TGGA sequence context [53].

Therefore, in the TGGT sequence context, we clustered the structures according to those forming the G7-O6 hydrogen bond, the T8-O4 hydrogen bond and those forming no hydrogen bonds and determined the helical parameters for each of those structures using CURVES version 5.3 (see Methods). As described previously, we then identified the helical parameters that showed statistically significant differences between distributions of any two ensembles on the basis of heat maps (Figure S5) of their Kolmogorov-Smirnov (KS) ratios [55] (described in Text S1). We observed no major differences in conformation between structures forming the G7-O6 hydrogen bond and those forming no hydrogen bonds for both CP- and OX-DNA adducts (data not shown). We had observed a similar result in the TGGA sequence context [53]. These data suggest that the formation of a hydrogen bond between the 3′ Pt-amine and G7-O6 does not require significant distortion of the Pt-GG adduct. However, in the CP-TGGT adducts, we found significant differences in the helical parameters between structures forming the T8-O4 hydrogen bonds and those forming either the G7-O6 hydrogen bond or no hydrogen bonds (Figure S5, panel A and Figure 8). For example, we observe more positive values for shift of the G7-T8 base-pair step and T8-A17 opening and more negative values for G7-C18 propeller twist and G7-C18 shear, which promote formation of the T8-O4 hydrogen bond. In OX-TGGT adducts, we observed shifts in helical parameters similar to CP-TGGT adducts for the structures forming the T8-O4 hydrogen bond (Figure S5 panel B and Figure 9). However, these effects on T8-O4 hydrogen bond formation are overshadowed by the heterogeneous distribution of structures forming the G7-O6 hydrogen bond. In the case of the Pt-TGGT adducts, the differences in conformation associated with the formation of the T8-O4 hydrogen bond do not appear to favor formation of the CP-GG-HMGB1a complex based on the comparison of the conformational distributions associated with formation of G7-O6 and T8-O4 hydrogen bond with the conformation observed in the crystal structure of the CP-GG-HMGB1a complex (the vertical line in Figures 8 and 9). These data are consistent with the very limited ability of HMGB1a to discriminate between CP-GG and OX-GG DNA adducts in the TGGT sequence context (see Discussion).

The 12-mer oligonucleotide containing the TGGT sequence (Figure 1) was platinated according to our previously described protocols and purified using HPLC (described in Text S1). Following hybridization with complementary strand, the OX-GG 12-mer duplex was further characterized following the same LC/MS procedure reported previously by Wu et al. [45], [46] Using this procedure, the Pt-GG-intrastrand cross-link is digested to Pt[d(GpG)] and Pt-G monoadducts to Pt(dG). In addition, both Pt-GNG-intrastrand cross-links and Pt-GG-interstrand cross-links are digested to dG-Pt-dG [60]. The digest of undamaged DNA showed peaks corresponding to the four normal deoxynucleosides (data not shown). The digest of the same 12-mer duplex containing the OX-DNA adduct showed the same four deoxynucleosides and one additional peak eluting just after dT (Figure 10A). This additional digestion product had the same retention time and UV spectrum as a synthetic OX[d(GpG)] standard (data not shown). This additional peak was further identified as OX[d(GpG)] by the presence of the expected molecular ions in both the positive (m/z 904.97) (Figure 10B) and negative (m/z 902.99) mass spectra. Lastly, the MS-spectra also showed the expected Pt-isotope pattern, confirming the presence of a Pt compound. No digestion products were detected with the masses and isotopic pattern expected for the dG-OX-dG or OX(dG) adducts. These data demonstrate that the 12-mer duplex employed in this NMR study consisted exclusively of the OX-GG-intrastrand cross-link, confirming the purity of the substrate used for structure determination.

For both undamaged DNA and the OX-DNA adduct in the TGGT sequence context, two NMR samples were prepared; one, H₂O sample, in a 5% D₂O/95% H₂O buffer (100 mM NaCl, 5 mM phosphate buffer, pH 7.0) that was used for detection of exchangeable protons with varying temperature (2 to 40 °C) and the other, D₂O sample, in 100% D₂O buffer (same buffer composition as that of the H₂O buffer) for detection of non-exchangeable protons. For both the samples, the duplex DNA concentration was 1.2 mM. NMR spectra were acquired on Varian Inova 500, 700, or 800 MHz spectrometers. The carrier frequency for protons was set on the H₂O signal. 1D proton spectra were recorded using a Varian Inova 800 MHz NMR spectrometer at temperatures ranging from 2 to 40 °C for detection of exchangeable protons. NOESY spectra were recorded in D₂O buffer at 25 °C at 700 MHz using a mixing time of 200 ms, 32 transients and 400 complex FIDs corresponding to spectral width of 25 and 12 ppm in both dimensions for both samples in H₂O and D₂O. To determine the optimal mixing time for quantification of NOE data, a series of 2D NOE data were collected in which the mixing times were varied (150 ms, 200 ms, 250 ms, and 300 ms). Close inspection of the data revealed that NOE data collected at a mixing time of 200 ms optimized signal/noise while minimizing spin diffusion effects. The WATERGATE pulse sequence was employed for water suppression in both H₂O and D₂O samples [61]. Distance constraints were obtained from the 200 ms NOESY spectra in both H₂O and D₂O [43]. The assignments were obtained initially from NOE connectivities and were confirmed by analysis of 2D DQF-COSY [44] (2048 × 720 complex points, 12 ppm spectral width in both dimensions, 32 transients). 2D DQF-COSY data were also used for determination of J coupling constraints for determination of sugar pucker as described previously [45], [46]. The J-coupling constants were estimated by simulating DQF-COSY cross-peaks using the program Chords 2.0 (Spectrum Research, LLC). All J_{H2′–H2″} values were set as −14.0 Hz. All other coupling constants were determined by adjusting their values in steps of 0.1 Hz. This step was repeated until the simulated multiplet looked very similar to its experimental counterpart. NMR data were processed with NMRPipe and analyzed with Felix (version 2000, Molecular Simulations, Inc., San Diego, CA). The structure calculations are essentially as reported previously [45], [46] and are described in detail in Text S1.

Chemical shift assignments for OX-DNA and undamaged DNA duplex in the TGGT sequence context were obtained essentially as described previously for the OX- and CP-DNA adduct in the AGGC sequence context [45], [46]. Assignment of the non-exchangeable base and sugar protons were obtained by analysis of NOESY and 2D DQF-COSY spectra as described in Methods. For example, the NOESY region for OX-DNA in Figures S1 and S2 shows NOE correlations between the base (purine H8/pyrimidine H6) and the H1′, H2′, and H2′′ sugar protons. Sequential connectivities can be observed without interruption from C1 to C12 in the GG strand and from G13 to G24 in the CC strand (Figures S1 and S2, panels A and B). Similar connectivities in the NOESY spectrum were also observed for the undamaged DNA in the same TGGT sequence context (Figures S1 and S2, panels C and D). Upon completion of the sequential assignments for the H8/H6, H1′, H2′, and H2′′ proton signals, assignments of other (H3′, H4′, H5′, and H5′′) sugar protons were obtained by following standard procedures with the NOESY and DQF-COSY spectra [62]. Assignment of the exchangeable protons was obtained by analyzing distance connectivities between the imino and base/amino proton regions of NOESY spectra in H₂O buffer at 2°C. The chemical shift assignments for the OX-DNA duplex and undamaged DNA are shown in Tables S1 and S2, respectively.

We obtained 481 and 665 experimental distance constraints for OX-DNA duplex and undamaged DNA respectively. These experimental distance constraints were used as input for CNS to calculate the respective solution structures of OX-DNA duplex and undamaged DNA as described in Methods. The structural statistics for OX-DNA duplex and undamaged DNA are listed in Table 2. Of 20 calculated structures for OX- and undamaged DNA, 14 with lowest energies were accepted as a family. The root mean square deviation (RMSD) for the superimposition of the heavy atoms for all 14 final structures was 0.83 Å for OX-DNA duplex and 1.21 Å for undamaged DNA.

We performed simulations on a 12-mer DNA sequence (similar to the one used for NMR experiments), which was either undamaged or covalently bound to CP or OX at the N7 of G6 and G7. There were two sets of simulations with different starting structures for each of CP-, OX- and undamaged DNA. The starting structures for CP- and OX-DNA were NMR and crystal structures of the adducts in the TGGT sequence context (1A84 (crystal) and 1AIO (NMR) for CP-DNA and 1IHH (crystal) and the average structure from this study (NMR) for OX-DNA). For undamaged DNA, the average NMR structure obtained in this study in the TGGT sequence context and the B-DNA structure in the TGGT sequence context (generated using Insight II) were used as the two starting structures.

We performed 5 sets of 10 ns simulations for each starting structure of CP-, OX-, and undamaged DNA. The 5 sets had the same starting structures but the initial velocities were randomized. We employed simulation protocols identical to our published work on the AGGC [52] and TGGA sequence contexts [53].

All the trajectories were analyzed for the presence of hydrogen bonds between all possible donors and acceptors based on a distance cut-off of 3.5 Å between the donor and acceptor and an angular cut-off of 135° between donor-H-acceptor. In addition to the Watson-Crick interactions the only hydrogen bonds that were observed for more than 10% of simulation time were between Pt-amines and the surrounding base pairs. All the frames of the trajectory were classified based on the type of hydrogen bonds formed between Pt-amines and the adjacent bases.

Helical parameters were calculated for each snapshot of the trajectory using the CURVES program, version 5.3 [56], [57]. The analysis was performed using protocols identical to our earlier studies [52], [53]. The detailed methodology used for analysis is described in the Text S1.