The unlabeled and the site-specific low-enrichment (2% ¹⁵N-labeled) oligonucleotides were synthesized and purified as described previously (52,53). Unless otherwise stated, the strand concentration of the NMR samples was typically 0.5–5 mM; the solutions contained 70 mM of KCl and 20 mM of potassium phosphate (pH 7). Sequences used in this work are shown below:

Name	Sequence
Natural-Form 1	d[TAGGGTTAGGGTTAGGGTTAGGG]
BrG16-Form 1	d[TAGGGTTAGGGTTAG(BrG)GTTAGGG]
Natural-Form 2	d[TAGGGTTAGGGTTAGGGTTAGGGTT]
BrG15-Form 2	d[TAGGGTTAGGGTTA(BrG)GGTTAGGGTT]
Two forms	d[TAGGGTTAGGGTTAGGGTTAGGGT]

Experiments were performed on 600 MHz spectrometers at 25°C, unless otherwise specified. Resonances for G residues were assigned unambiguously by using site-specific low-enrichment labeling and through-bond correlations at natural abundance (52–55). Resonances for T residues were assigned following systematic T-to-U replacements. Assignments of resonances for A residues were obtained from NOE connectivities with neighboring T and G residues in the fold. NMR spectral assignments were completed by through-bond (COSY, TOCSY) and through-space (NOESY) correlation experiments as described previously (54). Interproton distances were measured by using NOESY experiments at different mixing times.

The structures of Form 1 and Form 2 human telomere quadruplexes were calculated using the X-PLOR program (56). NMR-restrained molecular dynamics computations were performed as described previously (43). The structures were first calculated for ^BrG-substituted sequences. Inclusion of K⁺ ions within the top and bottom caps of Form 2 (57) resulted in well-converged structures, which are consistent with experimental data. The ensembles of structures for natural sequences were computed from those for ^BrG-substituted sequences by refining them against the experimental restraints obtained for natural sequences.

The coordinates for four quadruplex structures formed by the 23-nt and 25-nt natural and ^BrG-substituted human telomere sequences have been deposited in the Protein Data Bank (accession codes natural-Form 1: 2JSM; ^BrG16-Form 1: 2JSK; natural-Form 2: 2JSL; ^BrG15-Form 2: 2JSQ).

Previously, we systematically examined human telomere sequences containing four G-tracts and showed that small changes to flanking sequences can perturb the equilibrium between different coexisting G-quadruplex forms (44). In K⁺ solution, d[TAGGG(TTAGGG)₃] (a natural human telomere sequence) forms up to 70% of Form 1 (43), while d[TAGGG(TTAGGG)₃TT] (also a natural human telomere sequence) with two Ts at 3′-end forms up to 70% of Form 2 (44). Here we will call these sequences natural-Form 1 and natural-Form 2, respectively. It should be noted that two G-quadruplex conformers coexist in slow exchange at comparable proportions in d[TAGGG(TTA GGG)₃T] (another natural human telomere sequence) having one T at 3′-end (Figure S1, Supplementary Data).

Substitution of proton by bromine at position C8 of a guanine has been shown to favor syn glycosidic conformation of the nucleotide (47). When all five guanines that adopted syn conformations in Form 1 were substituted by 8-bromoguanines, Form 1 predominated (46,51). However, each G-to-^BrG substitution removes a proton (useful in NMR studies). For the natural-Form 1 and natural-Form 2 sequences, we could identify single G-to-^BrG substitutions (at position G16 and G15, respectively) that further favored the corresponding major form, thereby significantly improving NMR spectra (Figure 2). They will be called ^BrG16-Form 1 and ^BrG15-Form 2, respectively. Figure 2.

Imino proton spectra of (A) d[TAGGG(TTAGGG)₃] (natural-Form 1), (B) ^BrG16-Form 1, (C) d[TAGGG(TTAGGG)₃TT] (natural-Form 2) and (D) ^BrG15-Form 2, in K⁺ solution with assignments listed over the spectra.

We have previously unambiguously assigned imino and H8 protons of guanines in natural sequences (43,44). Corresponding assignments for single ^BrG-substituted sequences could be obtained by comparing spectral patterns of the modified and natural sequences (Figures 2 and 3). These assignments were also independently confirmed by some low-enrichment site-specific labeling and natural abundance through-bond correlation experiments (Figure 4) (52–55). Assignments of resonances for T residues were obtained from T-to-U substitution samples (54). NMR spectral assignments were completed by through-bond (COSY, TOCSY) and through-space (NOESY) correlation experiments as described previously (54). Figure 3.

The H8/6-H1′ proton region of NOESY spectra (mixing time, 300 ms) of (A) d[TAGGG(TTAGGG)₃] (natural-Form 1), (B) ^BrG16-Form 1, (C) d[TAGGG(TTAGGG)₃TT] (natural-Form 2) and (D) ^BrG15-Form 2 in K⁺ solution. The assignments and H8/6-H1′ NOE sequential connectivities are shown.

The structures of Form 1 (Figures 5, 6 and S2, Table 1) and Form 2 (Figures 7, 8 and S3, Table 2) human telomere quadruplexes adopted by ^BrG-substituted and natural sequences were calculated on the basis of NMR restraints using the X-PLOR program (56). In all cases, the structure of the G-tetrad cores is better defined than that of the loops (Figures 5 and 7). The structures formed by the ^BrG-substituted and the corresponding natural sequences are quite similar. However, the structures calculated for the ^BrG-substituted sequences (Figures 5A and 7A) are slightly better defined than those calculated for the natural sequences (Figures 5C and 7C) thanks to larger numbers of defined NOE peaks associated with cleaner NMR spectra. Figure 5.

Stereo views of Form 1 structure. (A) Ten superpositioned refined structures of ^BrG16-Form 1. (B) Ribbon view of a representative structure. (C) Ten superpositioned refined structures of natural-Form 1. Anti and syn guanines are colored cyan and magenta, respectively; in (A) and (C), adenines are colored green; thymines, orange; backbone, gray; O4′ atoms, red; phosphorus atoms, yellow. In (B), O4′ atoms are colored yellow.

In both Form 1 and Form 2, there are one double-chain-reversal and two edgewise TTA loops. The double-chain-reversal loop is situated in a medium groove. The edgewise loops always connect an anti guanine to a syn guanine, across narrow or wide grooves; they cap the top and the bottom of the G-tetrad core, respectively (Figure 1). The detailed structures of these elements in Form 1 and Form 2 are shown in Figures 6 and 8, respectively. Figures S2 and S3 show the distribution of sequential and long-range NOEs used to derive the loop structures in these forms.

In Form 1, the T18–T19–A20 loop (Figure 6A) closes a narrow groove and caps the top of the G-tetrad core. Residue A20 from this loop is aligned with residues T1 and A2 from the 5′-end to form the (T1–A2) · A20 triad platform (Figure 6A and D). Residue T19 is stacked on top of this platform, while T18 is projected aside. An adenine triple was observed by Dai et al. (50) in the top cap of Form 1 for an end-modified sequence, in which the natural residue T1 was replaced by an A. This base triple involved adenines which are equivalent to A2, A8 and A20 in our sequence. Such an adenine triple does not form in our structure of the natural human telomere sequence. The T12–T13–A14 loop (Figure 6C) closes a wide groove and caps the bottom of the G-tetrad core. Hoogsteen base pair A14–T12 was observed among computed structures and stacks over the terminal G-tetrad (Figures 5 and 6C). This configuration of the bottom loop is clearly different from that observed for Form 1 with modified bases at the 3′-end (43,50). The structure of the T6–T7–A8 double-chain-reversal loop shows the stacking between T7 and A8 (Figure 6B). We observe differences in the conformations of both edgewise TTA loops between our Form 1 solution structure (Figures 6A and C) and the corresponding solution structure reported by Matsugami et al. (51). In our case, there is much greater stacking of the loop residues over the terminal G-tetrads.

In Form 2, the T12–T13–A14 loop (Figure 8A) closes a narrow groove and caps the top of the G-tetrad core. The residues in this loop interact intimately with the 5′-end residues T1–A2 (Figure 8A and D). NOEs detected within this region are consistent with a structure where T1, T12 and A14 residues, together with the carbonyl groups of the top G-tetrad, can potentially coordinate a K⁺ ion within the loop (Figure S4). This structure would explain the upfield chemical shifts of many protons of T1 (Figure 3 and Table S2). However, other structures might also be possible for this region, as manifested by the broadening of some resonances at 25°C (44). For example, in some stages of computation we observed the interaction of A20 from the double-chain-reversal loop with residues in the top cap. The structure of the T18–T19–A20 double-chain-reversal loop shows some stacking between T18 and T19 bases (Figure 8B). The T6–T7–A8 loop (Figure 8C) closes a wide groove and caps the bottom of the G-tetrad core. Residues T7 and A8 interact with T24 from the 3′-end to form a (T7–A8) · T24 triad platform. Residue T25 is stacked below this platform. The configuration of the bottom cap in Form 2 involving the interaction between the loop and the 3′-end residues is quite different from that observed in Form 1 [see above and Refs. (43,50,51)]. A K⁺ ion can also be potentially coordinated between the bottom G-tetrad and the cap (Figure S4).

In general, K⁺ can be coordinated within all edgewise loops presented here, and the equilibrium between base pairings and K⁺ coordination is probably the best description of their structure in solution. These observations reinforce a previous proposal by our group of K⁺ cation coordination within edgewise loops in G-quadruplexes (57).

Four-repeat human telomere sequences can form at least two different intramolecular (3 + 1) G-quadruplexes in K⁺ solution. These structures can coexist and be in dynamic equilibrium with other G-quadruplex forms. Such an equilibrium between conformers is reminiscent of what was reported previously by our group for two-repeat human telomere (22) and two-repeat Tetrahymena telomere (53) G-quadruplexes in solution. As all possible human telomere G-quadruplexes might contain TTA loops, it is important to gather structural patterns of different TTA loop conformations. We have obtained from this work several different configurations of TTA loops in the context of G-quadruplexes formed by natural sequences. Generally, the bases of edgewise loops tend to maximize pairing by forming non-canonical pairs, triples and triads, which in turn stack over the terminal G-tetrads. In favorable cases, edgewise loop residues could also align to potentially coordinate a K⁺ cation, embedded within the loop turn. A range of topologies have been observed for TTA double-chain-reversal loops, with a common theme that two out of the three bases generally tend to stack on each other. Note that even though the present structures were solved for natural human telomere sequences, sequence extension towards either 5′- or 3′-ends can affect the structure of the loops and caps (44). It should also be noted that the presence of a ligand may push the equilibrium towards one particular structure, which may or may not be represented by the free native sequence (58).