Journal list menu

Volume 281, Issue 4 p. 1085-1099
Original Article
Free Access

Specific loop modifications of the thrombin-binding aptamer trigger the formation of parallel structures

Anna Aviñó

Anna Aviñó

Institute for Advanced Chemistry of Catalonia (IQAC), CSIC, Networking Center on Bioengineering, Biomaterials and Nanomedicine (CIBER-BBN), Barcelona, Spain

These authors contributed equally to this workSearch for more papers by this author
Guillem Portella

Guillem Portella

Institute for Research in Biomedicine (IRB Barcelona), Spain

Joint IRB-BSC Program in Computational Biology, Barcelona, Spain

These authors contributed equally to this workSearch for more papers by this author
Ruben Ferreira

Ruben Ferreira

Institute for Advanced Chemistry of Catalonia (IQAC), CSIC, Networking Center on Bioengineering, Biomaterials and Nanomedicine (CIBER-BBN), Barcelona, Spain

Search for more papers by this author
Raimundo Gargallo

Raimundo Gargallo

Department of Analytical Chemistry, University of Barcelona, Spain

Search for more papers by this author
Stefania Mazzini

Stefania Mazzini

Department of Food, Environmental and Nutritional Sciences (DEFENS), Section of Chemical and Biomolecular Sciences, University of Milan, Italy

Search for more papers by this author
Valérie Gabelica

Valérie Gabelica

Physical Chemistry and Mass Spectrometry Laboratory, Department of Chemistry, University of Liège, Belgium

Search for more papers by this author
Modesto Orozco

Corresponding Author

Modesto Orozco

Institute for Research in Biomedicine (IRB Barcelona), Spain

Joint IRB-BSC Program in Computational Biology, Barcelona, Spain

Departament de Bioquímica i Biologia Molecular, Facultat de Biologia, University of Barcelona, Spain

Correspondence

R. Eritja, Institute for Advanced Chemistry of Catalonia (IQAC), CSIC, Networking Center on Bioengineering, Biomaterials and Nanomedicine (CIBER-BBN), Jordi Girona 18-26 08034 Barcelona, Spain.

Fax: +34 93 2045904

Tel: +34 93 4006145

E-mail: [email protected]

M. Orozco, Institute for Research in Biomedicine (IRB Barcelona), Baldiri Reixac 10, 08028 Barcelona, Spain

Fax: +34 93 4037157

Tel: +34 93 4039942

E-mail: [email protected]

Search for more papers by this author
Ramon Eritja

Corresponding Author

Ramon Eritja

Institute for Advanced Chemistry of Catalonia (IQAC), CSIC, Networking Center on Bioengineering, Biomaterials and Nanomedicine (CIBER-BBN), Barcelona, Spain

Correspondence

R. Eritja, Institute for Advanced Chemistry of Catalonia (IQAC), CSIC, Networking Center on Bioengineering, Biomaterials and Nanomedicine (CIBER-BBN), Jordi Girona 18-26 08034 Barcelona, Spain.

Fax: +34 93 2045904

Tel: +34 93 4006145

E-mail: [email protected]

M. Orozco, Institute for Research in Biomedicine (IRB Barcelona), Baldiri Reixac 10, 08028 Barcelona, Spain

Fax: +34 93 4037157

Tel: +34 93 4039942

E-mail: [email protected]

Search for more papers by this author
First published: 04 December 2013
Citations: 25

Abstract

Guanine-rich sequences show large structural variability, with folds ranging from duplex to triplex and quadruplex helices. Quadruplexes are polymorphic, and can show multiple stoichiometries, parallel and antiparallel strand alignments, and different topological arrangements. We analyze here the equilibrium between intramolecular antiparallel and intermolecular parallel G-quadruplexes in the thrombin-binding aptamer (TBA) sequence. Our theoretical and experimental studies demonstrate that an apparently simple modification at the loops of TBA induces a large change in the monomeric antiparallel structure of TBA to yield a parallel G-quadruplex showing a novel T-tetrad. The present results illustrate the extreme polymorphism of G-quadruplexes and the ease with which their conformation in solution can be manipulated by nucleotide modification.

Abbreviations

  • CCS
  • collision cross-section
  • DC
  • self-diffusion coefficient
  • DOSY
  • diffusion-ordered spectroscopy
  • DT
  • drift time
  • IMS
  • ion mobility spectroscopy
  • MD
  • molecular dynamics
  • PDB
  • Protein Data Bank
  • TBA
  • thrombin-binding aptamer
  • T m
  • melting temperature
  • Introduction

    DNA is an extremely polymorphic molecule that is able to adopt a multitude of helical structures [1-3]. Guanine-rich oligonucleotides lead to especially polymorphic nucleic acids, which can adopt both A-type and B-type duplexes, and different types of triplex and quadruplex [1-3]. G-quadruplexes are very stable structures generated by the association of steps of four guanines (G-quartets) held together by eight hydrogen bonds. These structures are stabilized by intramolecular and intermolecular stacking, and strong electrostatic interactions with cations within the structure [4, 5]. One of the most intriguing characteristics of G-quadruplexes is their structural variability [6-9]; depending on the length of the polyG track, the loops that connect the quartets, the environment and the syn/anti preference of the purines in the quartets, the G-quadruplex can appear in both ‘antiparallel’ and ‘parallel’ arrangements. Within the first family of structures, the quadruplex is characterized by opposite neighboring strand orientations and different conformations around the glycosidic bond of the guanosines in the G-quartets. In contrast, in the parallel quadruplex, all of the strands have the same direction, and all of the guanines are in anti conformation. In addition, hybrid structures formed by the combination of both arrangements have been described, as well as multimeric G-quadruplex formed by the association of several strands [4, 5]. It is important to shed light on the factors that govern the specific folding and stability of quadruplexes of a given sequence. Although considerable effort is being made in this regard, no general rules have been established to date.

    G-quadruplexes are structures of great biological relevance, and contribute to the stabilization of telomers [10, 11] and, accordingly, to the preservation of chromosomal integrity. Furthermore, aromatic molecules that are able to bind to G-quadruplexes are excellent inhibitors of telomerase, a major target in cancer [3, 4, 12]. G-quadruplexes acting as aptamers [13] can be powerful inhibitors of different proteins, such as HIV integrase [14] or thrombin [15, 16], with various applications in diagnostics and therapeutics. In particular, much research effort has been focussed in the 15-base thrombin-binding aptamer (TBA) sequence, which was derived in experiments as an inhibitor of thrombin, a key enzyme in the blood clotting cascade [15]. In aqueous solution, TBA forms an intramolecular, antiparallel G-quadruplex with two stacked G-quartets connected by a central TGT, and two lateral TT loops, defining a chair-like conformation [17, 18]. The adopted structure is identical in the presence of potassium and in ammonium ions [19]. Guanines in the two tetrads show an alternating syn/anti conformation (syn-dG at positions dG1, dG5, dG10, and dG14, anti-dG at positions dG2, dG6, dG11, and dG15), and are stabilized by the presence of small monovalent cations in the central channel [19]. The stoichiometry and interaction motif between thrombin and TBA have been controversial for several years [18, 20, 21]. It is currently assumed that the two TT loops interact with the thrombin anion exosite I in a 1 : 1 stoichiometry [22]. The loops connecting G-quartets have a strong influence on the folding and stability of intramolecular quadruplexes. Different loop lengths and sequences can either stabilize or destabilize the G-quadruplex [23-28]. For example, reduction of the two lateral TT loops of TBA to a single thymidine caused complete disruption of the quadruplex structure, whereas extension to TTT did not alter the stability [23]. Significant changes in stability have also been found when thymidines in the lateral loops have been changed to other coding nucleotides [29, 30], or to derivatives with modified backbones [31-35]. For example, the replacement of several thymidines at the loop positions with 2′-F-arabino derivatives always leads to a more stable G-quadruplex [31], whereas substitution by locked nucleic acids, unlocked nucleic acids and acyclic thymidines showed a position-dependent effect on TBA stability [32-35].

    In this study, we systematically analyzed the effect of (coding) nucleotide substitution in the two lateral loops of TBA. The most remarkable finding was made with the double T→G substitution; this leads to a dramatic change in the topology of the quadruplex, which moves from an intramolecular antiparallel arrangement to an intermolecular parallel one. A wide repertoire of experiments and theoretical calculations have allowed us to characterize and rationalize the nature of this conformational change induced by loop changes, which, up to now, has been believed to occur only upon changes in the solvent or in the conformational preferences of nucleotides in the central tetrads [36-40]. The unexpected results reported here illustrate the extreme conformational promiscuity of G-DNA quadruplexes, and the possibility of manipulating their conformational landscape by simple mutations.

    Results and Discussion

    Thermal UV and CD studies of TBA derivatives

    We synthesized up to seven variants of TBA to exhaustively explore the effect of nucleotide modification in the lateral loops (Table 1).

    Table 1. Sequences of the TBA derivatives used in this study.
    Name Sequence
    TBA GGTTGGTGTGGTTGG
    TBA2A[4,13] GGTAGGTGTGGTAGG
    TBA2C[4,13] GGTCGGTGTGGTCGG
    TBA1G[4] GGTGGGTGTGGTTGG
    TBA2G[4,13] GGTGGGTGTGGTGGG
    TBA2G[3,12] GGGTGGTGTGGGTGG
    TBA4G[3,4,12,13] GGGGGGTGTGGGGGG

    We used CD spectroscopy to characterize the nature of the structures of the different TBA derivatives. TBA and most of the derivatives showed clear antiparallel quadruplex spectra (see CD spectra in KCl/cacodylate buffer in Fig. 1A), with positive bands at approximately 250 nm and 295 nm, and a negative band at approximately 270 nm, whereas TBA derivatives containing two or more guanines at the lateral loops (TBA2G[4,13], TBA2G[3,12], and TBA4G[3,4,12,13]) showed very different CD spectra, which match with those expected for a parallel G-quadruplex (note the positive band at 265 nm and the negative one at 245 nm).

    Details are in the caption following the image
    (A, B) CD spectra of modified TBAs in (A) 10 mm cacodylate and 100 mm KCl (pH 7) buffer, and (B) 100 mm ammonium acetate buffer. (C) CD spectra of TBA in 100 mm ammonium acetate buffer with increasing percentages of MeOH. (D) CD spectra of TBA2G[4,13] in 100 mm ammonium acetate buffer with increasing percentages of MeOH.

    As described elsewhere [11, 29-35, 38], TBA forms a very stable structure with a melting temperature (Tm) of ~ 50 °C in potassium buffer (monitored by UV absorbance; Table 2). The melting profiles of TBA were reversible without hysteresis (Fig. S1), suggesting that melting is tracing a single and fast intramolecular process. Replacement of thymine with the other pyrimidine, cytosine (TBA2C[4,13]), at positions 4 and 13 had little impact on the stability of the structure (Table 2; Fig. S1). In contrast, replacement of thymine with adenine (TBA2A[4,13]) induced substantial destabilization of the TBA structure, which again was fully reversible, suggesting that we were still analyzing the same simple intramolecular folding/unfolding process found for TBA.

    Table 2. Melting temperatures and thermodynamic parameters obtained from denaturing curves for TBA derivatives in potassium buffer. ΔTm is the Tm change in the melting process between the modified TBAs and unmodified TBA. Hysteresis refers to the difference between the heating and cooling process with a temperature gradient of 0.5 °C·min−1. Uncertainty of ± 0.5 °C.
    Name TmTm) (°C) ΔH° (kcal·mol−1) ΔS° (cal·kmol−1) ΔG° (25°C) (kcal·mol−1) Hysteresis
    TBA 50.1 −45.1 ± 0.3 −139.6 ± 1.0 −3.5 No
    TBA2A[4,13] 40.1 (−10) −33.2 ± 0.5 −106.1 ± 1.6 −1.6 No
    TBA2C[4,13] 51.5 (+1.4) −42.5 ± 0.7 −131.0 ± 2.4 −3.4 No
    TBA1G[4] 47.9 (−2.2) −39.8 ± 0.9 −124.0 ± 2.9 −2.8 No
    TBA2G[4,13] 59.2 (+9.1) −63.4 ± 0.9 −165.3 ± 2.6a −14.1 Yes
    TBA2G[3,12] 58.5 (+8.4) −58.9 ± 2.5 −152.3 ± 7.6a −13.5 Yes
    TBA4G[3,4,12,13] > 80 Yes
    • a Apparent values obtained with the approximation of a two-state equilibrium.

    The results above strongly suggest that, in the lateral loop, there are no clear preferences between pyrimidines, but the presence of a purine is not very well tolerated, probably because of its larger size. However, experiments with TBA2G[4,13], in which two thymines are replaced with two guanines, showed a completely different situation. On the one hand, this substitution increased the thermal stability (ΔTm = + 10 °C; Table 2). On the other hand, melting profiles showed a clear hysteresis (Fig. S1), suggesting that the TBA→TBA2G[4,13] (or the TBA→TBA2G[3,12]) substitutions induce a severe structural change in the quadruplex, resulting in a structure with slower folding kinetics. To further analyze these results, we synthesized and performed melting experiments with a derivative in which only one thymine was replaced with one guanine (TBA1G[4]), and another in which all thymines in lateral loops were replaced with guanines (TBA4G[3,4,12,13]). The first derivative showed the behavior resulting from a pyrimidine → purine transversion, i.e. destabilization of the structure, but maintenance of the melting profile. In contrast, the second derivative showed a dramatic increase in Tm, the resulting structure being so stable that structural transition could not be observed (Table 2; Fig. S1).

    The melting experiments suggest that the TBA structure is quite permissive to pyrimidine transitions in the lateral loops, whereas pyrimidine → purine transversions destabilize the TBA structure, except in the case of two guanine or four guanine substitutions. In that case, a more stable and complex structure, with slower folding kinetics, was formed. Melting experiments at different oligonucleotide concentrations strongly suggested that two guanine substitutions altered the overall molecularity of the quadruplex, as Tm increased with concentration only for the TBA2G derivatives (Fig. S2). In addition, the thermal stability of TBA2G[4,13] also increased with K+ concentration (Table S4). This could be explained by the improvement in stacking interactions between G-tetrads.

    Thermal denaturation and CD were also analyzed in ammonium conditions. The stabilities of the modified TBAs are shown in Table S5. Tm values were lower than in potassium buffer, but the Tm differences relative to unmodified TBA were the same in both buffer conditions. The CD profile of the TBAs in ammonium buffer differed from those in potassium buffer (Fig. 1B). Specifically, TBA2G[4,13] showed CD spectra with two positive bands: a higher band at ~ 260 nm corresponding to a parallel quadruplex structure, and a lower band at ~ 295 nm corresponding to an antiparallel structure. This agrees with partial conversion to the antiparallel state. Note that such a change was not detected in the ultrastable TBA4G[3,4,12,13] parallel quadruplex, where the antiparallel/parallel stability difference seemed to be very large, therefore precluding any ion-dependent polymorphism [17, 19]. In addition, TBA2A[4,13], which was similar to TBA when dissolved in potassium buffer, had CD spectra showing a mixture of parallel and antiparallel structure when in ammonium buffer. The rest of the TBA derivatives, TBA, TBA2G[3,12], TBA1G[4], TBA2C[4,13], and TBA4G[3,4,12,13], had similar CD spectra in ammonium buffer and in potassium buffer.

    TBA and TBA2G[4,13] were also examined with thermal denaturation and CD in the presence of a dehydrating agent (MeOH), which is known to favor compact structures (such as the parallel quadruplex) [41, 42]. The thermal stability of TBA2G[4,13] was higher than that of TBA, and increased with the percentage of MeOH in both compounds (Table S6). The addition of MeOH to the TBA in ammonium buffer increased the amplitude of the CD bands corresponding to antiparallel quadruplex structures (Fig. 1C). This finding has also been made with poly(ethylene glycol) 200 [41]. More importantly, the addition of MeOH to TBA2G[4,13], which, in the absence of MeOH, showed a mixture of parallel and antiparallel conformations, caused the complete formation of a parallel structure (Fig. 1D).

    Melting experiments and CD spectroscopy clearly suggest that addition of two or more guanines to the lateral loops of TBA induces a transition from a monomeric antiparallel to a multimeric parallel G-quadruplex. In addition, stabilities derived from thermodynamic parameters with a simple folded and unfolded state were obtained in order to compare all of the derivatives (Table 2). The results from TBA2G derivatives were approximate, and should be interpreted cautiously, as minor multimeric species have also been observed (see below).

    NMR studies of TBA derivatives

    NMR measurements of TBA, TBA2C[4,13] and TBA2A[4,13] gave rise to similar spectra (Fig. 2), which fitted well with the known characteristics of antiparallel quadruplexes [24, 38]. A temperature increase led to the loss of quadruplex signals at temperatures (TBA and TBA2C[4,13] at ~ 50 °C, and TBA2A[4,13] at ~ 40 °C (Fig. 2A,B) that matched well with the Tm derived from UV melting experiments.

    Details are in the caption following the image
    Imino protons region. NMR melting experiments on (A) TBA2C[4,13], (B) TBA2A[4,13] and (C) TBA2G[4,13] in H2O, 10 mm potassium phosphate buffer, and 5 mm KCl (pH 6.7).

    We measured the NMR spectra of TBA2G[4,13] at different K+ concentrations, ranging from 5 mm to 100 mm KCl, and obtained essentially the same spectra. The spectra of TBA2G[4,13] showed severe and generalized line broadening, suggesting that several species were in equilibrium (Fig. 2C). Visible broad NH signals were found at 10.5–11.5 p.p.m. These signals remained visible even at a high temperature. Quantitative analysis of the TBA2G spectra is difficult, owing to the existence of different conformers.

    Spectra of TBA, TBA2A and TBA2C can be analyzed to obtain general structural information (see proton assignments in Tables S7 and S8). These derivatives showed NH signals between 11.5 p.p.m. and 12.5 p.p.m. that corresponded to an antiparallel conformation. In addition, the C4 and C13 NH imino signals in TBA2C present at 5 °C disappeared when we increased the temperature to 15 °C (data not shown). The same effect was seen with the imino protons of the thymines in the loops, indicating that they were more exposed to the solvent. We did not find chemical shift signals for the imino protons of A4 and A13. The 1D NMR spectrum of TBA2A showed a main antiparallel monomeric species in solution, together with another minor monomeric species, as proved by the low-intensity signals present both in the NH imino and in the aromatic protons region (Fig. 2B). The presence of A4 and A13 affected the NH imino protons of thymines in the loops. No NH signals were observed for T12, T7, and T3, which implies exposure to a major solvent. Similarly, the path of NOE connectivity for TBA2C between aromatic protons and H1′ of the ribose of G5, G10 and G14 was characterized by strong intensities, as expected for a syn conformation, whereas weak intraresidue intensities found for G2 and G6 were associated with an anti conformation. No sequential connections were found between G2 and T3, T3 and C4, C4 and G5, T7 and G8, T9 and G10, or C13 and G14. We observed characteristic long-range interactions, i.e. H8 of G11 with the NH imino proton of C13, and H8 of G2 with H2′, H2″ and NH of C4. These contacts were confirmed by the NOEs between H6 of C4 and H1′, H2′ and H2″ of G2. We also detected interactions between H8 of G8 and H1′, H2′′ and NH of G6, and between H1′ of T9 and H8 of G15. Moreover, the methyl group of T9 gave an NOE signal with NH imino protons of G6, G10, G1, and G15 (Fig. S4). All of these findings pointed to a similarity in the backbone conformation with the unmodified TBA. We also detected other interactions between aromatic protons of C4 and G2, and C13 and G11.

    The spectral assignment for TBA2A[4,13] was difficult, owing to the large overlap of NH and aromatic proton signals. Nevertheless, some NOE contacts were found between H8 of G2 and ribose protons of G6, H8 of G15 and H1′ of T9, and MeT9 and NH imino protons of G15, G10, G1, and G6. Finally, NMR diffusion-ordered spectroscopy (DOSY) experiments allowed us to determine self-diffusion coefficients (DCs): for TBA, TBA2C, and TBA2A, we obtained a DC value of 1.86 × 10−6 cm2·s−1 (expected value for the monomeric form: 1.77 × 10-6 cm2·s−1), whereas for TBA2G we obtained a smaller value, of 1.26 × 10−6 cm2·s−1 (expected value for the dimeric form: 1.26 × 10−6 cm2·s−1), suggesting the existence of a dimer structure for TBA2G.

    ESI-MS and ion mobility spectroscopy (IMS) of TBA derivatives

    Figure 3 shows the ESI-IMS-MS results for unmodified TBA and TBA2G[4,13], in aqueous solution (top) and 60% methanol solution (bottom). The data are represented in two dimensions, with m/z separation horizontally, and ion mobility drift time (DT) separation vertically.

    Details are in the caption following the image
    ESI-IMS-MS 2D spectra for (A) natural TBA in 100% aqueous 100 mm NH4OAc, (B) TBA2G[4,13] in 100% aqueous 100 mm NH4OAc, (C) natural TBA in 40% aqueous 100 mm NH4OAc and 60% MeOH, and (D) TBA2G[4,13] in 40% aqueous 100 mm NH4OAc and 60% MeOH. Peak annotations showing theoretical peak locations: M, monomer; D, dimer; T, trimer; Q, tetramer. The insets in (B) and (D) are the mass spectra extracted for peak D5−, and show the number distribution of ammonium ions preserved in the dimers in both conditions.

    ESI-IMS-MS of modified and unmodified TBA spectra from purely aqueous ammonium acetate in the negative mode (Fig. 3A) gave ion signals corresponding to the intramolecular G-quadruplex at charge states 3−, 4− (major), and 5−. For TBA2G[4,13] (Fig. 3B), an additional charge state distribution was detected, corresponding to dimeric G-quadruplexes (D), at charge states 5−, 6−, and 7−. ESI-IMS-MS spectra recorded for TBA in dehydrating conditions (Fig. 3C) were similar to those obtained in aqueous conditions. There was a very slight relative increase in a dimer signal, indicated by shadows on the 2D plot, but the relative signal was too weak for extraction of the corresponding mass data. In contrast, for TBA2G[4,13] (Fig. 3D), the addition of methanol led to the disappearance of the monomer, and enlargement of the peaks corresponding to the dimer, which was the major species. Trimer and tetramer ion series were also detectable.

    In the M3− monomeric forms of TBA and TBA2G, we found 0, 1 or 2 ammonium ions bound to TBAs, in agreement with previous results [42]. For the charge state 3−, the collision cross-section (CCS) did not change with the ammonium ion count [43]. As the lowest charge state is also the least prone to gas-phase conformational changes resulting from Coulomb repulsion, we used M3− to compare gas-phase experimental and theoretical CCSs. With the same reasoning, we used the CCS of the dimer D5− peak to determine the plausible gas-phase structure of the dimer. At least four ammonium ions were preserved in the dimer in both conditions, with or without methanol (see blue insets in Fig. 3). Interestingly, more ammonium ions were preserved in the presence of MeOH (see insert in Fig. 3D; the distribution maximum is with five ammonium ions). This may be attributable to increasing cation activity in 60% : 40% methanol/water as compared with 100% water.

    We obtained the CCSs for different gas-phase ions from ion mobility experiments. According to our previous results [44-48], such observations provide valuable information on the shapes of monomers and dimers, both in the gas phase and in solution. The results shown in Table 3 match well with those derived from extended MD simulations in the gas phase (see next section), confirming that the detected signal corresponded to the expected monomeric antiparallel (low CCSs) and mostly dimeric parallel G-quadruplexes (large CCS found for TBA2G[4,13]).

    Table 3. CCSs for the TBA derivatives ions at 3− charge state in ammonium acetate buffer determined experimentally, and computed from MD simulations. Experimental: average values and standard deviations obtained from measurements on the SYNAPT G1 HDMS at a bias voltage of 15 V. The prediction error (95% confidence) is 1.6% in all cases. MD simulation: MD values obtained from ensemble of > 4 ms obtained for replicates with different charge distributions and initial structures (see 4) defined for the ionic state determined from ESI-MS measurements.
    Ω (Ǻ2) experimental Ω (Ǻ2) MD simulation
    TBA 564 ± 1 569 ± 4
    TBA2A[4,13] 564 ± 1
    TBA2C[4,13] 548 ± 1
    TBA2G[4,13] 563 ± 1
    TBA2G[4,13]a 899 ± 2 928 ± 4
    • a CCSs calculated for the dimer ion at 5− charge state.

    All-atoms resolution theoretical studies

    The spectroscopic evidence just described suggests that the solution structure of TBA derivatives is the result of a subtle balance between at least three species: a single-stranded oligonucleotide, a monomeric antiparallel quadruplex, and a dimeric parallel quadruplex (Fig. 4). Clearly, as the concentration of oligomers increases, dimeric structures become more numerous, but the entropic advantages of an intramolecular process favors the antiparallel quadruplex even in the high NMR concentration regime. However, the complete equilibrium can be modulated by the introduction of mutations at the loops. The physical reasons are, however, unclear, which forced us to perform theoretical calculations.

    Details are in the caption following the image
    The effects of nucleotide substitutions on the structures of TBA derivatives. The transition between the antiparallel monomer and the parallel dimer structures occurs via an unfolded intermediate, as direct conversion is kinetically very unfavorable. The left and right halves of the sketch show that the main structure corresponds to the parallel dimers if positions 3 and 12, and 4 and 13, are occupied by two or four guanines (in the case of a complete substitution with guanines, the equilibrium is prominently displaced towards the parallel dimer). In the upper and lower halves, we illustrate the equilibrium that corresponds to TBA, TBA2C[4,13], TBA2A[4,13] and TBA1G[4], where the parallel dimer is not stable, and the main structure is the monomeric antiparallel fold. The lower thermal stability of TBA2A[4,13] and TBA1G[4] than of TBA and TBA2C[4,13] is indicated by a slight shift towards the unfolded state.

    Extended MD simulations in aqueous solution of TBA, TBA2A[4,13], TBA2C[4,13] and TBA2G[4,13] resulted in stable trajectories that did not significantly deviate from the starting structures generated from the known NMR conformation of TBA in water (Fig. S6). Our simulations support the experimental observation that nucleotides changes (T→A, T→C, or T→G) at the lateral loop might destabilize the structure, but does not lead to disruption of the antiparallel arrangement in physiological-like conditions. We also performed MD simulations for models of the dimeric parallel quadruplexes for TBA, TBA2A[4,13], TBA2C[4,13], TBA2G[4,13] and TBA4G[3,4,12,13] sequences. The intermolecular TBA2G[4,13] and TBA4G[3,4,12,13] quadruplexes remain fully stable for 0.4 μs, in a conformation that resembled very closely a canonical parallel quadruplex (Fig. 5), with the central channel occupied by three or four cations. In contrast, the parallel quadruplexes generated from TBA, TBA2A[4,13] and TBA2C[4,13] sequences led to unstable trajectories, characterized by a large increase in rmsd (above 0.5 nm from the original model structure), loss of ions in the central channel, and complete corruption of the structure at the end of the trajectory, where dissociation of the two strands was more evident for TBA (see embedded representations in Fig. 5). Furthermore, the CCS of the main charge states for TBA (M3−) and TBA2G dimers (D5−) computed from MD simulations in the gas phase agreed well with the experimental measurements (Table 3).

    Details are in the caption following the image
    (A) Time evolution of the rmsd with respect to the core tetrads of each initial structure for the parallel dimers of TBA4G[3,4,12,13], TBA2G[4,13], TBA2A[4,13], TBA2C[4,13], and TBA (i.e. the rmsd excluding the atoms that belong to the loops). Whereas TBA4G[3,4,12,13] and TBA2G[4,13] adopt a stable conformation in a folded state after partially losing one layer of G-tetrad, TB2A[4,13], TBA2C[4,13] and TBA rapidly and completely unfold, losing almost all of the G-tetrads (TBA2A[4,13] retains two of them, owing to the time scale of the simulations). The embedded molecular models show sketches of the end state of each MD simulation. The potassium ions are shown as purple spheres. (B) Side (above) and top (below) close-up views of the central tetrads for TBA4G[3,4,12,13], TBA2G[4,13] and TBA after relaxation of the loops, prior to the unrestrained MD simulations. The rightmost figures illustrate a top view of the T-tetrad (above) and a G-tetrad (below). For clarity, thymine carbons are colored in green, and guanine carbon atoms are colored in gray.

    MD simulations strongly support the idea that a dimer of TBA4G[3,4,12,13] can form a very stable parallel quadruplex, thanks to an uninterrupted tract of six G-tetrads that maintains a strong hydrogen bond network and coordinates around five ions in the central cavity. According to our MD simulations, the presence of one T-tetrad in the central part of the intermolecular parallel G-quadruplex is well tolerated (the case of the TBA2G[4,13] sequence), despite the reduction in the number of of hydrogen bonds as compared with a G-tetrad. Ion coordination between the T-tetrad and G-tetrad is tolerated, and is consistent with ESI-MS data. In contrast, our simulations strongly suggest that TBA, TBA2C[4,13] and TBA2A[4,13] do not tolerate well the inclusion of two consecutive non-G-tetrads in the central steps, and thus they do not form intermolecular parallel quadruplexes, and the intramolecular antiparallel quadruplex dominates the conformational landscape of these oligonucleotides.

    To gain a more quantitative explanation of the role of different nucleotides in the stabilization of the parallel quadruplex, we computed the changes that nucleotide mutation in the loop produces in the relative stability of unfolded single-strand oligonucleotide, the antiparallel G-quadruplex, and a model parallel quadruplex. We used a standard thermodynamic cycle to extract the differential binding free energy between the sequences from the free energy associated with changes to nucleotides in different environments (unfolded, antiparallel, and parallel; see 4, Appendix S1, and Fig. S3). The results (summarized in Table 4) clearly suggest that the mutations of T to purines (G or A) do not dramatically affect the intrinsic stability of the antiparallel quadruplex (i.e. do not change the relative stabilities of the antiparallel and unfolded conformations). As expected from unbiased MD, the same T→G mutation that was innocuous regarding the stability of the antiparallel conformation dramatically affected the stability of the dimeric parallel quadruplex. In fact, the mutation TBA→TBA2G[4,13] stabilized the intermolecular parallel quadruplex by > 47 kJ·mol−1 (> 60 kJ·mol−1 if we take the completely unfolded state as the reference). Assuming additivity, the multiple mutation TBA→TBA4G[3,4,12,13] would imply extreme stabilization of > 90 kJ·mol−1 for formation of the hypothetical parallel quadruplex (120 kJ·mol−1 if the unfolded oligomer is taken as the reference). Note that these estimates were performed in the absence of external loops (already considered in unbiased MD simulations), suggesting that the extreme stabilization of parallel quadruplexes stems from the formation of an alternative tetrad (T4 in the case of TBA2G[4,13], and G4 in the case of TBA4G[3,4,12,13]; Fig. 5), and not from spurious loop effects. Theoretical simulations thus provide a quantitative and clear explanation for the surprising effects of loop mutations on the stability and overall topology of G-quadruplexes.

    Table 4. Changes in free energy of different process associated with mutations at loops. For example: ΔΔGantiparallel → parallel dimer (T→G) means the change in relative stability of antiparallel and parallel quadruplexes occurring as a consequence of the mutation of a T in the lateral loop of TBA into a G. The negative value (− 48 kJ·mol−1) means that the parallel dimer is highly stabilized by the T→G change.
    Mutation Free energy difference (kJ·mol−1)
    ΔΔGantiparallel → parallel dimer (T→G) − 48 ± 2
    ΔΔGunfolded → parallel dimer (T→G) − 62 ± 1
    ΔΔGunfolded → antiparallel (T→G) − 7 ± 1
    ΔΔGunfolded → antiparallel (A→G) 3 ± 2

    Proposed structures for TBA derivatives

    Replacement of T4 and T13 with A, C and G generated distinct G-quadruplex structures, depending on the nucleobase. Experimental results showed that substitution with A and C generates antiparallel quadruplex structures. TBA2C[4,13] resulted in a stable quadruplex derivative similar to TBA, an observation that implies that the two cytosines are well accommodated in the overall antiparallel structure. By contrast, TBA2A[4,13] showed a destabilized structure, as reflected by our spectroscopic data. Figure 4 summarizes the hypothetical structures for TBA derivatives observed in this study.

    Double guanine substitution derivatives (TBA2G[4,13] and TBA2G[3,12]) resulted in a more stable parallel quadruplex structures. DOSY experiments and the agreement between simulated and experimental CCSs for TBA2G[4,13] suggested that the parallel G-quadruplex originates from the dimerization of two TBA2G[4,13] oligonucleotides. Note that the CCS of TBA2G[4,13] is less than the sum of the constituent TBA monomers, both theoretically and experimentally. In addition, PAGE analysis of TBA derivatives showed a main band with lower migration for TBA2G derivatives, indicating a larger structure than that of TBA (Fig. S5).

    Consequently, we propose a parallel structure with five G-tetrads and a T-tetrad, and two TGT propeller loops (Fig. 4). Spectroscopic data showed that the preferred structure is a dimer, but other parallel multimers are also present. NMR imino proton signals are broad, and several ammonium ions are observed in the ESI-MS spectra. The multimer quadruplex may correspond to the stacking of several parallel dimers. Experimentally, five ammonium ions were observed in the ESI-MS spectra, suggesting that additional ammonium ions could be accommodated between T-tetrads and G-tetrads. To our knowledge, this coordination has not been described to date.

    The complete rearrangement of two unfolded structures is required for the formation of the parallel dimer. This dimer was previously observed in TBA derivatives carrying riboguanosines or acyclic thymidine derivatives [35, 49]. In addition, some authors have described the formation of a bimolecular antiparallel quadruplex for unmodified TBA [50], in contrast to earlier findings and NMR results [18, 19, 29].

    In the proposed dimer, the four thymidines are intercalated by stacking between two G-tetrads. This T-tetrad also contributes to the overall stability of the structure. Nevertheless, the broad signal of the dimer at 10–11 p.p.m. does not allow assignment of the NH imino protons of the T-tetrad. A similar tetrad has been found by the use of NMR in a parallel-stranded DNA quadruplex formed by Saccharomyces cerevisae telomeric repeats d(TGGTGGC). The T-tetrad is formed by four thymidines held together by O4·H3 hydrogen bonds in a plane, and is well accommodated in the center of the G-quadruplex and flanked by two G-tetrads on each side. The T-tetrad is less stable than the G-tetrad (only four hydrogen bonds and a smaller surface for stacking interactions) [51]. This thymidine arrangement was also investigated in theoretical studies [52]. TBA4G[3,4,12,13] forms a very stable parallel quadruplex structure, for which we propose a preferentially very compact dimer structure with six G-tetrads.

    In addition to the proposed dimer, multimers are also observed. The presence of multimeric species may be attributable to stacking of dimers on the G-tetrads at the end of the quadruplex. These multimeric forms are not observed in the chair-like TBA structure, owing to the presence of the TT and TGT loops near the G-tetrads, which prevents stacking. In this way, the presence of multimeric species is in agreement with the parallel dimer structure suggested for the TBA2G and TBA4G derivatives.

    In the case of TBA1G[4], where T4 was replaced with G4, it is not possible to form an extra G-tetrad, and the parallel dimer is expected not to form. CD confirmed the antiparallel structure and Tm, and the thermodynamic parameters showed that this derivative was destabilized. However, Tm was increased at high concentrations. This observation and the PAGE experiment (Fig. S5) implied that equilibrium of complex multimeric species was present in solution.

    Conclusions

    G-quadruplexes are very polymorphic structures whose major conformation in solution can change for multiple reasons. By combining experimental and theoretical tools, we demonstrate here that the conformation of the thrombin aptamer sequence can be changed completely from a chair-like intramolecular antiparallel quadruplex to a parallel quadruplex by mutations in the lateral loop region. The present results might force us to revise our assumptions regarding the nature of chromosomal quadruplexes, as apparently small sequence alterations can make intermolecular parallel quadruplexes competitive with the intramolecular antiparallel ones, opening new potential roles of G-quadruplexes in genomic recombination and telomere fusion processes.

    Experimental procedures

    The oligonucleotides prepared for this study (Table 1) correspond to the TBA modified at positions 3, 4, 12 and 13 that were originally occupied by thymidines. The modified TBAs were synthesized on an automated RNA/DNA synthesizer with β-cyanoethylphosphoramidite chemistry, and following standard protocols. The oligonucleotides were purified by HPLC and characterized by MS.

    UV absorbance and CD measurements

    Oligonucleotides at 3 μm were resuspended in either potassium buffer (10 mm sodium cacodylate, 100 mm KCl, pH 7.0) or ammonium buffer (100 mm ammonium acetate, pH 7.0), and were then annealed at 85 °C and slowly cooled to room temperature. TBA and TBA2G[4,13] were also studied under dehydrating conditions. For this purpose, the oligonucleotides were dissolved in ammonium buffer with increasing percentages of MeOH prior to analysis. The thermal curves were obtained by following the absorption change at 295 nm from 15 °C to 80 °C with a linear temperature ramp of 0.5 °C·min−1 on a JASCO V-650 spectrophotometer equipped with a Peltier temperature control. Thermal denaturation experiments were performed at a range of concentrations (1.5, 3, 15 and 30 μm). The thermal denaturation–renaturation profile (hysteresis) was studied at 3 μm at 0.5 °C·min−1.

    Thermodynamic parameters were calculated from UV melting experiments as described elsewhere [53], with matlab (R2009b version; Math-Works, Natick, MA, USA). This analysis assumes a simple two-state equilibrium between the folded and unfolded states. A self-complementary bimolecular process was assumed for thermodynamic analysis of TBA2G[4,13] and TBA2G[3,12], and an intramolecular process for the rest of the derivatives.

    The CD spectra were recorded on a JASCO spectropolarimeter J-810. Spectra were registered at 25 °C over a range of 220–320 nm, with a scanning speed of 50 nm·min−1, a response time of 4 s, data pitch of 0.5 nm, and a bandwidth of 1 nm. The samples (4 μm) were dissolved in potassium or ammonium buffer, and annealed before recording of spectra. To analyze the effect of crowding agents, TBA and TBA2G[4,13] were dissolved in ammonium buffer with increasing percentages of MeOH (20%, 40%, and 60%) prior to analysis.

    NMR spectroscopy

    NMR spectra were recorded on a Bruker AV600 spectrometer operating at a frequency of 600.10 MHz for 1H. 1H-NMR spectra were recorded at variable temperatures ranging from 5 °C to 65 °C. Chemical shifts (δ) were measured in p.p.m. and referenced to external 2,2-dimethyl-2-silapentane-5-sulfonate sodium salt set at 0.00 p.p.m. The estimated accuracy is within 0.02 p.p.m.

    For NMR experiments, TBAs were dissolved in H2O/D2O (90 : 10 v/v) at a concentration range of 0.7–1.0 mm in the presence of 10 mm potassium phosphate buffer and 5 mm KCl (pH 6.7). The sequential 1H assignments for all oligonucleotides were performed by applying well-established procedures for the analysis of TBA [29]. All of the protons were attributed for TBA2C[4,13], and protons were partially attributed for TBA2A[4,13]. The melting experiments were performed with 1H-NMR spectroscopy at different temperatures. Phase-sensitive NOESY spectra (tmix = 300 ms) were acquired at 5 °C in TPPI mode, with 4000 × 512 complex free induction decays, a spectral width of 15 000 Hz for H2O solutions, a recycling delay of 1.5 s, and 120 scans. TOCSY spectra were acquired with the use of an MLEV-17 [54] spin-lock pulse (field strength, 10 000 Hz; total duration, 60 ms). All spectra were transformed and weighted with a 90° shifted sine-bell squared function to 4000 × 1000 real data points. For the H2O suppression, the excitation sculpting sequences from standard Bruker pulse program libraries were used. Pseudo-2D DOSY [55] experiments were performed with the pulseprogram stebpgp1s, a diffusion delay of 0.12–0.45 s, a gradient pulse of 1.5 ms, 64 increments, and a temperature of 25 °C. Raw data were processed with the standard DOSY software present in the Bruker library (topspin v. 1.3). A calibration curve was obtained with samples with Mr ranging from 180 to 23.500 as standards, as previously reported [56]. The value of the DC obtained for TBA, TBA2C and TBA2A was 1.86 × 10−6 cm2·s−1 (10−9.73 m2·s−1). The expected value for the monomeric form was 1.77 × 10−6 cm2·s−1 (10−9.75 m2·s−1). The value obtained for TBA2G was 1.26 × 10−6 cm2·s−1 (10−9.90 m2·s−1. The expected value for the dimeric form was 1.26 × 10−6 cm2·s−1 (10−9.90 m2·s−1).

    ESI-MS and IMS

    The ESI-MS experiments were carried with a SYNAPT HDMS (Waters, Manchester, UK) in ion mobility mode. The capillary voltage was set to − 2.2 kV, the cone voltage was 30 V, the extraction cone was 4 V, the source pressure was 3.15 mbar, the source and desolvation temperatures were 40 °C and 60 °C, respectively, and the trap and transfer voltages were 6 V and 4 V, respectively. The ion mobility cell was supplied with N2 to reach a pressure of 0.532 mbar in the IMS cell (instrument pirani reading). The wave height was 8 V, and the wave speed was 300 m·s−1. The bias voltage for ion introduction into the IMS cell was 10 V. The SYNAPT HDMS was calibrated in CCS with oligonucleotides, as described previously [57]. TBA, TBA2G[4,13], TBA2C[4,13] and TBA2A[4,13] were folded at 200 μm in 100 mm ammonium acetate buffer, and injected at a final strand concentration of 5 μm and at a rate of 140 μL·h−1·L−1 at room temperature. Higher mobility (higher charge or lower CCS) translates into shorter DT. Here, ion mobility is used to separate the charge state series corresponding to different stoichiometries. For example, if a given m/z can correspond to a monomer Mz and a dimer D2z (double the mass and double the charge), given that CCSs of dimers are almost always less than twice than those of monomeric forms, and that the charge is double for the dimer, the mobility is higher for the dimer than for the monomer. The higher the molecularity, the lower the DT. As a consequence, multimers separate on different diagonals on the 2D plots.

    Molecular dynamics (MD) simulations

    Model preparation

    We obtained the initial structure of TBA in its native antiparallel fold from the Protein Data Bank (PDB) structure 148D. The same PDB structure was used to model the TBA2G, TBA2A and TBA2C mutations in monomeric antiparallel form by exchanging bases at positions 4 and 13 from thymine to guanine, adenine, or cytosine, respectively. We used the PDB structure 2JT7 – sequence (TG4T)4 – as a template to model a parallel G-quadruplex fold built of four eight-mer chains of the sequence TG6T. We changed the third and/or fourth G-tetrads of the original oligonucleotide to T-tetrads to obtain (TGGTTGGT)4 and (TGGTGGGT)4, which were used as model systems for the dimeric forms of TBA and TBA2G. We used these structures for free energy calculations (see below), as the lack of loops allows the effects of the mutations to be isolated. We also built parallel G-quadruplexes with the sequences (GGTTGG)4, (GGTAGG)4, (GGTCGG)4, (GGTGGG)4, and (GGGGGG)4, based on PDB 2JT7. In these parallel quadruplexes, we connected the 3′-termini of strand 1 with the 5′-termini of strand 2 with a manually placed TGT fragment, thus completing the TBA, TBA2A[4,13], TBA2C[4,13], TBA2G[4,13] and TBA4G sequences. Using the same operation between strands 3 and 4, we finally obtained the dimeric G-quadruplex models. More details can be found in the supporting information (see Appendix S1).

    We immersed each previously described DNA structure in an equilibrated water box with a sufficicient amount of potassium ions to both fill the cavities between each G-tetrad and also ensure overall charge neutrality. The structures were optimized, thermalized and pre-equilibrated with a slightly modified version of our standard protocol [58] (Appendix S1) prior to an extended equilibration of 10 ns. Production runs were in the range of 0.2–0.7 μs (Table S1) for detailed enumeration of all MD simulations. All simulations were performed with gromacs-4.5 [59], the force fields describing the interactions for DNA were generated on the basis of the parm99-parmBSC0 [60, 61] parameters, and SPC/E [62] model was used to describe the water molecules.

    Gas-phase simulations

    We used the last structures of our aqueous MD simulations of TBA and the parallel dimer of TBA2G as initial models for gas-phase calculations. For TBA, we used one starting conformation, whereas for the dimer of TBA2G we used two extracted from our two different MD simulations. As in previous studies [44-47], we selected the most probable ionic states on the basis of our MS results (see below), which suggested a charge of 5− on TBA and a charge of 10− for the TBA2G parallel dimer. As there are two cations inside TBA, and five cations in the TBA2G parallel dimer, the total charges of the systems are, respectively, 3− and 5−. We selected two (TBA) and three (TBA2G dimer) charge distributions that optimized the placement of the protonated phosphates by reducing Coulomb repulsion between anionic phosphates (Table S2). For TBA, we performed five different MD simulations (using different starting velocities) of 400 ns for each one of the two different charge distributions. For the TBA2G dimer, we performed three replicates for each one of the three charge distributions, using two different starting configurations, all of them 400 ns in length. From each trajectory, we extracted 400 equally spaced structures, and computed the predicted CCSs with the exact hard-sphere scattering approximation [63]; these were then averaged to obtain the ensemble CCS.

    Differential binding free energies

    We determined the effects of the T→G substitution in the loops on the differential stabilities of monomer antiparallel and dimeric parallel quadruplexes. For this purpose, we used standard thermodynamic cycles (Figs S3, S6, and S7; full details in Appendix S1) to compute the free energy associated with the T→G conversion in three different environments: (a) a single-stranded d(GGTGG) oligonucleotide; (b) an antiparallel TBA quadruplex (T4 and T13); and (c) a parallel quadruplex (TGGTTGGT)4 (T4). From such free energy, the effects of the T→G mutation on the relative stabilities of unfolded, antiparallel and parallel quadruplexes was determined by taking advantage of the state function properties of the free energy (Fig. S3). We also determined the difference in free energy of formation between TBA2A and TBA by transforming A to G in a single-stranded d(GGAGG) oligonucleotide and in the antiparallel TBA2A. The free energy associated with the T→G or A→G change in the different structures were computed with a thermodynamic integration method in its discrete formalism (DTI) [64].

    Acknowledgements

    This work was supported by the Spanish Ministry of Science and Innovation, MICINN (CTQ2010-20541, CTQ2012-38616, BIO2009-10964, and Consolider E-Science), the Generalitat de Catalunya (2009/SGR/208), the University of Milano (PUR 2009 Funds), the funds de la Recherche Scientifique-FNRS (VG research associate position and FRFC grant 2.4528.11), and PRIN09 (2009Prot-2009J54YAP_005). R. Ferreira is a recipient of an FPI predoctoral contract (MICINN) and an STSM from COST (G4net, MP0802). G. Portella is a recipient of a Sara Borrell postdoctoral fellowship. Collaborative research was funded by a Cost action (G4net, MP0802) and an Italian-Spanish collaborative action (IT2009-0067). CIBER-BBN is an initiative funded by the VI National R&D&i Plan 2008-2011, Iniciativa Ingenio 2010, Consolider Program, CIBER Actions and financed by the Instituto de Salud Carlos III with assistance from the European Regional Development Fund.