In the past, we showed that metal species have a high affinity fur the central G in the GGG sequence of the duplex d(A(1)T(2)G(3)G(4)G(5)T(6)A(7)C(8)C(9)C(10)A(11)T(12))(2) (G3-D) and that cisplatin (cis-Pt(NH3)(2)Cl-2) and G3-D formed an N7-Pt-N7 G(4),G(5) intrastrand cross-link preferentially over the G(3),G(4) adduct (similar to 25:1). Thus, a putative G(4) monoadduct was postulated to cross-link in the 3'- rather than the normally more favorable 5'-direction. To evaluate this hypothesis and also to explore why the G3-D G(4),G(5) adduct had atl unusual hairpin structure, we have now introduced the use of N,N'-dimethylthiourea (DMTU) as a monoadduct trap and have extended the study to a G3-D analogue with a hairpin form, d(A(1)T(2)G(3)G(4)G(5)T(6)T(7)C(8)C(9)C(10)A(11)T(12)) (G3-H). Chemical shift and 2D H-1 and C-13 NMR data indicated that the G3-H hairpin has a stem region with B-form structure and a nonhelical loop region. Zn2+ or Mg2+ ions transformed G3-H into a bulged duplex. Downfield shifts of G(4)H(8) and G(4)C(8) NMR signals indicated that Zn2+ binds preferentially to G(4)N7. Reaction of cisplatin or cis-[Pt(NH3)(2)(H2O)(2)](2+) with the bulged duplex and hairpin forms of G3-H gave a similar intrastrand cross-link ratio, G(4),G(5):G(3),G(4) = 7:3. This ratio is insensitive to DNA form or Pt leaving group. For G3-D this ratio is lower in the cis- [Pt(NH3)(2)(H2O)(2)](2+) reaction (similar to 1:1) than in the cisplatin reaction (25:1), indicating that the leaving group influences the cross-linking step for G3-D. The G(4) monoadducts of the cis-Pt(NH3)(2)Cl-2-G3-H and -G3-D and the cis-[Pt(NH3)(2)(H2O)(2)](2+)-G3-D reactions were trapped with DMTU, but no monoadduct was trapped in the cis-[Pt(NH3)(2)(H2O)(2)](2+)-G3-H reaction. The results suggest that the respective monoadducts are more long-lived for G3-D. We postulate that the G(5) in the G3-D Cl-G(4) monoadduct is placed in a favorable position to form the cross-link because of a prior conformational change induced by G(4)-A(7) stacking. This accounts for the very high selectivity for 3'-cross-linking. Nevertheless, in all other cases, regardless of the form or conformation, 3'-direction cross-linking is unusually favored at GGGT sequences, suggesting that the sequence itself contributes greatly to the 3'-cross-linking preference; since telomeres have multiple repeats of this GGGT sequence, this finding may have biological relevance.