Tethered imidazole mediated duplex stabilization and its potential for aptamer stabilization

Abstract

Previous investigations of the impact of an imidazole-tethered thymidine in synthetic DNA duplexes, monitored using UV and NMR spectroscopy, revealed a base context dependent increase in thermal stability of these duplexes and a striking correlation with the imidazolium pK_a. Unrestrained molecular dynamics (MD) simulations demonstrated the existence of a hydrogen bond between the imidazolium and the Hoogsteen side of a nearby guanosine which, together with electrostatic interactions, form the basis of the so-called pK_a-motif responsible for these duplex-stabilizing and pK_a-modulating properties. Here, the robustness and utility of this pK_a-motif was explored by introducing multiple imidazole-tethered thymidines at different positions on the same dsDNA duplex. For all constructs, sequence based expectations as to pK_a-motif formation were supported by MD simulations and experimentally validated using NOESY. Based on the analysis of the pK_a values and melting temperatures, guidelines are formulated to assist in the rational design of oligonucleotides modified with imidazolium-tethered thymidines for increased thermal stability that should be generally applicable, as demonstrated through a triply modified construct. In addition, a proof-of-principle study demonstrating enhanced stability of the l-argininamide binding aptamer modified with an imidazole-tethered thymidine in the presence and absence of ligand, demonstrates its potential for the design of more stable aptamers.

INTRODUCTION

At ambient pressure and temperature, proteins often show unmatched efficiency and selectivity in the high affinity binding of specific targets (cfr. antibodies) and in catalysis (cfr. enzymes). This makes them interesting partners in the development of immunoassays and affinity purification methods on the one hand (1,2) and more energy efficient and sustainable chemical processes on the other hand (3). Besides protein based architectures the unique ability of nucleic acids to form high order single or double stranded structures has led these to enjoy considerable attention as artificial binders and enzymes, so-called aptamers and (deoxy)ribozymes. Aptamers are single stranded nucleic acids selected by an iterative in vitro selection process called ‘SELEX’ (Systematic Evolution of Ligands by EXponential enrichment) and feature a specific tertiary structure featuring double stranded regions as well as hairpin and loop regions, capable of recognising and binding a target molecule with great selectivity and affinity. They are considered as promising alternatives for antibodies and continue to find new applications as therapeutics (4,5), biosensors (6,7), drug delivery systems (8,9) or as bio-imaging (9,10) and diagnostic tools (9,11). Shortly after the development of nucleic acid aptamers, catalytically active deoxyribozymes were generated from random libraries of oligonucleotides by a combinatorial in vitro selection approach (12), similar to SELEX. The scope of deoxyribozymes has expanded from reactions with nucleic acid substrates such as cleavage (13–15) and ligation (16,17) in the early beginning, to diverse chemical transformations including ester hydrolysis (18), Diels–Alder cycloaddition (19) and Friedel–Crafts alkylation (20).

Despite the power of directed evolution approaches in the creation of functional nucleic acids, the success rates of aptamer and (deoxy)ribozyme development is substantially constrained by the limited physicochemical diversity of native nucleic acids, especially when compared to proteins. This can be addressed by introducing protein-like functionalities through modification of the phosphate backbone, sugar moiety or nucleobase. According to literature, this leads to aptamers with increased target affinity (21–29), selectivity (30,31) and pharmacokinetic parameters (32) as well as metal-ion independent deoxyribozymes that outperform their unmodified counterparts (33–36). The established way to introduce extra functionalities is the use of modified (2′-deoxy-)nucleoside triphosphates in the selection protocol (37,38). Alternatively, modifications can be embedded selectively, by using modified monomers during solid phase oligonucleotide synthesis (23,39). Chemical synthesis offers a far larger functional versatility and spatial control when it comes to the introduction of unnatural nucleotides. In addition, it allows for (semi-)rational approaches to develop new functional nucleic acids with improved properties, as an alternative to randomized selection.

The imidazole group is of particular interest in this context and is an often-recurring functionality in functional nucleic acids obtained from SELEX (24,34,40–43) or chemical synthesis (23,44,45) alike. This results from its versatile physicochemicall features, including reversible protonation and the capacity for hydrogen bond formation. In its protonated, imidazolium state, it allows to place a positive charge within the negatively charged duplex environment associated with the phosphate backbone. The resulting favourable electrostatic interaction has been invoked to attribute the stabilising effect of tethered cationic modifications, including imidazole, on DNA/DNA or DNA/RNA duplexes, albeit without further investigation, for instance regarding the influence of the neighboring base pair context (46–50).

We previously described an in-depth study of the effects of incorporating a single histamine-based imidazole-modified thymidine (T^ImH+, Figure 1A) into a 14 base pair duplex model with a sequence optimized to allow a diversity of thymine positions for substitution while ensuring sufficient spectral resolution for NMR investigation (Figure 1B). Histamine was tethered to the C5 position of 2′-deoxyuridine using an amide bond orienting the imidazole ring in the major groove. By combining UV spectroscopy for thermal denaturation analysis and NMR spectroscopy together with in silico molecular modelling for structural characterization, the impact of the local base pair context on the overall duplex structure and stability and the imidazolium pK_a was studied (51). This revealed the possibility for a single T^ImH+ to increase the thermal stability (T_m) of the model duplex by 5–6°C when present in a particular three base pair sequence motif composed of the T^ImH+A base pair followed by a GC/CG sequence such as present in the T₈^ImH+ dsDNA sequence (Figure 1B). Its presence introduces a hydrogen bond between the imidazolium NHε2 and the O6 on the Hoogsteen side of the guanine at position n+2 on the opposite strand (Figure 1C). This places the imidazolium moiety in the major groove and effectively introduces a non-covalent, reversible clip between both strands (Figure 1D). This situation is referred to as an in-motif duplex hereafter. Its formation is accompanied by a significant reduction in the acidity of the imidazolium moiety by 1.6 to 1.9 pK_a units, hence we conveniently refer to the three base pair sequence (Figure 1B), leading to duplex stabilization and reduced imidazolium acidity as the ‘pK_a-motif’ (51). Importantly, when the T^ImH+A base pair in the pK_a-motif is flipped to AT^ImH+, thus creating the T₂₁^ImH+ construct when starting from the T₈^ImH+ one (Figure 1B) a non-motif duplex is obtained wherein the imidazolium is solvent exposed, ΔT_m is limited to ∼2°C while the acidity change ΔpK_a is only about ∼0.5, thereby illustrating the specific impact of the hydrogen bond between T₈^ImH+ and G₁₉ on the physicochemical properties. Throughout the text, we will refer to in-motif and non-motif duplexes as defined above, but also specify the imidazolium or T^ImH+ nucleotide involved as an in-motif and non-motif imidazolium or T^ImH+, respectively.

Figure 1.

(A) Structure of the T^ImH+ nucleoside building block. (B) Overview of the single T^ImH+ model duplex systems used as reference when comparing melting temperatures and pKa values. T* is used to represent the T^ImH+ nucleotide in all sequence representations. For the in-motif T₈^ImH+ sequence the pK_a-motif is shown enclosed within the box. The arrow indicates which guanine is engaged via its Hoogsteen side in the hydrogen bond with the imidazolium moiety, thereby creating an interstrand hydrogen bond. (C) Detail of the structure aspects of the pK_a-motif using in-motifT₈^ImH+. The hydrogen bond between the imidazolium functionality of T₈^ImH+ and the carbonyl oxygen (O6) of G₁₉ on the opposite strand is highlighted. (D) The pK_a-motif places the imidazolium functionality within the major groove as illustrated by the T₈^ImH+ system. Blue and red colours indicate positive and negative (partial) charges with variable intensity scaling with the charge size (visualization with VMD 1.9.1)

The possibility to modulate the pK_a of the imidazolium functionality by introducing either in-motif or non-motif T^ImH+ nucleotides hints towards the possibility to develop new artificial deoxyribozymes by rational decoration of DNA duplexes with catalytic combinations of imidazole functionalities in different charge states and nucleophilicities. Compared to this long term endeavour, a more short term and versatile opportunity for application exploits the stabilizing impact of an in-motif imidazole modification to heighten the thermal stability and/or improve the binding affinity of DNA duplex sequence of specific interest or DNA aptamers. Obviously, progress towards either development requires prior and systematic exploration of the effects on structure, thermal stability and acidity caused by introducing two or more T^ImH+ nucleotides at various positions in a model dsDNA sequence. Indeed, compared to duplexes with a single T^ImH+ studied so far, combinations of multiple in-motif and/or non-motif T^ImH+ nucleotides are expected to produce further thermal stabilization but will also introduce the possibility for electrostatic repulsion between imidazolium units, causing destabilization and mutual pK_a modulation that will depend on the relative position of the T^ImH+ units in the oligonucleotide sequence. To proceed towards a more general deployment of T^ImH+ in DNAzyme and aptamer research, a set of guidelines or design rules to guide the rational introduction of multiple T^ImH+ nucleotides for non-covalent stabilization or to create a particular pK_a combination is required. To this end, we report our investigation of a set of five duplexes based on the same model dsDNA sequence and equipped with two or three T^ImH+ nucleotides. We demonstrate the rather constant and largely additive contribution of each pK_a-motif to the thermal stability of a dsDNA duplex and analyze the extent to which interactions between the multiple T^ImH+ interfere with thermal stabilization and modulate the imidazolium pK_a. Using a set of guidelines derived from our analysis we conclude with a proof-of-concept study wherein rational introduction of a single T^ImH+ nucleotide in the l-argininamide DNA binding aptamer affords an increase in thermal stability exceeding 10°C while maintaining its binding capacity.

MATERIALS AND METHODS

Building block synthesis

The synthesis of the 2′-deoxy-5-(N-(2-(imidazol-4-yl)ethyl)-carbamoyl)-uridine-derivative was performed according to the procedure of Holmes and Gait (52). In the first step, histamine was coupled to 5-iodo-2′-deoxyuridine using a palladium-catalyzed one-pot carboxamidation reaction. After protecting the imidazole moiety with a tert-butoxycarbonyl group, the 5′-hydroxylgroup was selectively protected with 4,4′-dimethoxytrityl chloride. After conversion to the corresponding phosphoramidite, the imidazole modified thymidine building block was conveniently incorporated into oligonucleotides without further purification (details, see supplementary data).

Oligonucleotide synthesis

Modified oligonucleotides were synthesized on an ABI 394 DNA synthesizer using the standard automated phosphoramidite-based solid phase synthesis protocol (53,54) for a 1 μmol synthesis scale (details, see supplementary data). The oligonucleotides were purified by solid phase extraction (SPE) using Waters C18 Sep-Pak cartridges. This affords an easy way to purify oligonucleotides from impurities such as protecting group by-products and truncated sequences. RP-HPLC analyses were carried out on an Agilent 1100 Series instrument equipped with a Phenomenex Clarity 5μM Oligo-RP column (250 × 4.6 mm, 5 μm, 110 Å) or a Waters XBridge Oligonucleotide BEH C18 column (50 × 4.6 mm, 2.5 μm, 130 Å) and used a linear increasing acetonitrile gradient (5–35% in 15 min) against a triethylammonium acetate buffer (0.1 M TEAA, pH 7) at 50°C. The identity of the synthesized sequences was confirmed using ESI or MALDI mass spectrometry. After purification by solid phase extraction further purification and exchange of remaining triethylammonium counter ions for sodium ions was achieved by precipitation of the oligonucleotides in a 95/5 mixture of isopropanol and 3 M NaOAc(aq) solutions according to Andrus and Kuimelis (55). Unmodified sequences were purchased from Integrated DNA Technologies (IDT, Belgium) and were also desalted using the aforementioned protocol. The concentration of the oligonucleotides was determined spectrophotometrically using a Trinean DropSense^® 96 equipped with a DropPlate^® reader. The double stranded oligonucleotides were hybridized by mixing equimolar amounts (typically ±300 nmol) of complementary single stranded oligonucleotides. Samples were heated to 95°C in 30 minutes and kept at this temperature for 2 min. After removing the heating source, the sample was allowed to slowly reach room temperature. Following this annealing step, samples were dialyzed for at least 24 h against pure water using a dialysis membrane with a molecular weight cut-off of 3.5 kDa to remove remaining single strands or other impurities. The single stranded aptamer sequences did not require the preceding annealing step and were dialyzed immediately after the isopropanol/NaOAc precipitation procedure. Oligonucleotides were at –18°C either as a pellet or starting from an aqueous solution.

UV thermal denaturation experiments

The absorbance was measured at 260 nm in a temperature range from 5°C to 90°C with a heating/cooling rate of 0.3°C/min on a Varian Cary 300 Bio UV/VIS spectrophotometer equipped with a thermostated multicell holder. The annealed oligonucleotide strands (1 μM) were dissolved in a buffer solution containing 100 mM NaCl and 10 mM phosphate (pH 7.0). The melting temperatures were obtained from the melting curves using the first derivative approach in the Cary 300 Bio software. The reported T_m values represent average values of three consecutive melting traces and are reported with one standard deviation. For the aptamer in the presence of ligand, the same conditions were used with a 10 mM l-Arm concentration, expected to generate close to 100% saturation of the aptamer binding site in the unmodified system (56).

NMR studies

Samples used for the assignment of the non-labile protons consisted of the appropriate dsDNA or L-Arm aptamer (approximately 250 nmol) dissolved in D₂O (550 μL) containing 100 mM NaCl, 0.1 mM EDTA, 0.05 mM NaN₃ and 0.05 mM DSS (4,4-dimethyl-4-silapentane-1-sulfonic acid) for internal chemical shift referencing. Spectra were recorded in D₂O at 25°C (298.15 K) with a pH set at 5.0 (corrected for the isotope effect) by adding small aliquots of deuterated sodium hydroxide or deuterium chloride, this to ensure full protonation of the imidazolium group. NMR spectra were measured at a ¹H frequency of 700.13 MHz on a Bruker Avance II spectrometer equipped with a standard 5 mm inverse TXI-Z ATMA probe head operating under Topspin 3.1pl6. Standard pulse sequences from the Bruker library shown between brackets hereafter, were used throughout. All spectra in D₂O had a spectral width of 12.0 ppm and were recorded using excitation sculpting (zgesgp) for the suppression of the residual HDO signal when necessary (57). Typically, 128 scans of 16K data points each were accumulated. Prior to Fourier transformation, the data was apodized with a squared cosine bell function and zero filled to 32K. Zero-order polynomial baseline corrections were applied. 2D measurements consisted of COSY (cosydfesgpphpp), TOCSY (clmlevesgpph) and NOESY (noesyesgpph) spectra, recorded with 512 t₁ increments of 4K data points, 64 scans each. Mixing times for the TOCSY and NOESY were 75 and 200 ms, respectively. Apodization with a squared cosine bell function, followed by zero filling and Fourier transformation to a 2K × 2K data matrix followed by polynomial baseline correction yielded the final spectra. The CCPN data model (58) was used to assist the assignment of all systems, to provide chemical shift data for the non-exchangeable nucleic acid protons and to allow identification of the nOe contacts involving the imidazolium-tethered nucleotides. For all dsDNA constructs, non-exchangeable resonances were assigned with the usual methods for unlabelled oligonucleotiudes (51,59) As usual all were assigned sequence-specifically except for the the H4′ and H5′,H5′ resonances which are isolated from the remainder of the ribose spins system due to a small ³J_H3H5 scalar coupling. The NMR spectra, in terms of sequential nOe contacts and intraresidue COSY intensities, are consistent with a right-handed double helix in the B-family of conformations. For the l-Arm aptamers, partial assignments were limited to those nucleotides present in the double stranded stem. The assignments for all constructs were used for the chemical shift perturbation analysis collected in the supplementary material section.

pKa titration protocol

dsDNA (∼250 nmol) was dissolved in 550 μl H₂O/D₂O (9/1) containing 100 mM NaCl, 0.1 mM EDTA, 0.05 mM NaN₃ and 0.05 mM DSS for internal chemical shift referencing, while the water signal was suppressed using excitation sculpting. Spectra were measured over 25.0 ppm using 32K data points. The pH of the solutions was measured directly in the NMR tube using a 3 mm diameter Ag/AgCl pH electrode. The pH was adjusted by adding small aliquots of concentrated aqueous solutions of HCl or NaOH. The chemical shift of the aromatic (non-exchangeable) proton resonances of Hε1 and Hδ2 resonances were monitored while gradually increasing the pH from ∼5–10 (59,60). The identification of the Hδ2 and Hϵ1 resonances throughout the titration was facilitated by exploiting the long-range ⁴J(Hδ2, Hϵ1) coupling to generate a distinctive cross-peak in a 60 ms 2D TOCSY in a spectral region otherwise devoid of other cross-peaks as described previously (51). pK_a values were calculated by fitting the chemical shift data derived from the Hϵ1 proton to a Henderson-Hasselbalch equation as this proton is generally observed without signal overlap. Error-values were calculated using an in-house written Monte Carlo based algorithm (61). Whenever the calculation of the pK_a values was possible from Hδ2 data, they agreed within error limits with those derived from Hε1.

Molecular dynamics simulations

Initial models of the T^ImH+-modified duplexes were constructed in Discovery Studio (Accelrys Software Inc., San Diego USA, release 4.0) using standard canonical B-DNA parameters. These structures were neutralized with Na⁺ ions and solvated in a truncated octahedral box of TIP3P water molecules with the edges of the box not closer than 13 Å to any solute atom. Topology files for the T^ImH+ nucleotides were the ones described before (51) and were constructed using xLEAP and AmberTools 12 (62). The solvent molecules and sodium ions were first minimized using the steepest descent minimization (1000 steps) followed by the conjugate gradient minimization method (1000 steps) while applying a large harmonic restraint force constant of 500 kcal.mol⁻¹.Ų on all nucleic acid atoms. In a second minimization step the entire system was minimized using steepest descent minimization (1000 steps) followed by conjugate gradient minimization (1500 steps) without positional restraints. All systems were then heated to 300 K using the Langevin temperature equilibration scheme during a 20 ps NVT run with weak harmonic restraints on the oligonucleotide strands (10 kcal.mol⁻¹.Ų). Finally, the system was equilibrated for 100 ps (NPT ensemble) before performing a 50 ns molecular dynamics simulation. Simulations were performed using the NPT ensemble and applying the SHAKE algorithm (63) to constrain all covalent bonds involving hydrogen enabling a 2.0 fs time step. All simulations were carried out using periodic boundary conditions and the PME method with a 15 Å cut-off (64). Langevin dynamics (65) were applied for temperature control using a collision frequency of 1.0 ps⁻¹. Snapshots were collected every 2 ps, so generating 25 000 frames for each 50 ns run. Isotropic position scaling with a pressure relaxation time of 2 ps was used to maintain an average pressure of 1 atm. All MD simulations were run on a single Nvidia GTX680 GPU using the GPU implementation of PMEMD included in the AMBER v12 simulation package (66,67). After every equilibration and MD run several general system parameters (pressure, temperature, kinetic and potential energy) were extracted from the calculated trajectories and evaluated for errors or inconsistencies. VMD 1.9.2 (68) was used for the visualization of the simulated trajectories, hydrogen bond analysis was performed using the HBonanza python script (69). Hydrogen bonds were determined by the following requirements: (i) the distance between the donor and the acceptor heavy atom (D•••A) must not exceed 3.0 Å and (ii) the angle DHA between the donor heavy atom, the hydrogen atom and the acceptor heavy atom must not be inferior to 150°.

RESULTS AND DISCUSSION

Double introduction of T^ImH+ in duplexes: general analysis and selection of constructs for investigation

With the rational design of multiple imidazole-modified duplex systems in mind, understanding the mutual impact of two (or more) T^ImH+ modifications on duplex stability and imidazolium pK_a as a function of their relative position is essential. Here, we start by a general evaluation of all configurations that may arise when two T^ImH+ nucleotides are introduced in an otherwise unspecified dsDNA sequence as illustrated in Figure 2. At the outset, it should be realized that a second T^ImH+ cannot be introduced within the 3 bp sequence corresponding to the pKa-motif (box in Figure 2B) since this would remove one of the two CG base pairs that defines the pK_a-motif. Also, the specific situation that would place a second T^ImH+ opposite the first one would create a TT mismatch and is therefore excluded from consideration. Building on our previous work (51), we initiate our analysis based upon the expectation that introducing a first T^ImH+ at position n within any duplex will generate either a non-motif or in-motif imidazolium depending on the local base pair context at positions n+1 and n+2 (Figure 2). Indeed, when the base pairs at these positions do not correspond to those associated with the pKa-motif, a non-motif imidazolium will result (Figure 2A). Considering this singly modified non-motif construct, all positions n±i are in principle amenable for introduction of a second T^ImH+. However, when introduced at position n–2 or n–1 on the same strand or n+1 and n+2 on the complementary strand, the second T^ImH+ can itself only give rise to a non-motif imidazolium (Figure 2A, indicated as ‘n’ in the sequence). Indeed, when placed at either of these positions, the local base pair context of the second T^ImH+ will automatically include the first T^ImH+ and therefore prevent formation of the specific three base pair sequence of the pK_a-motif. At all other positions no limitations exist and the second T^imH+ can be introduced both as non-motif or in-motif (Figure 2A, indicated by ‘b’ for both), depending on the precise local base pair context. Next, we consider a singly modified in-motif construct as shown in Figure 2B. Here, the local base pair context at position n+1 and n+2 is fully defined by the sequence requirements of the pKa-motif. Similar to the non-motif case described above, placing a second T^ImH+ at n–1 or n–2 on the same strand again implies creation of a non-motif imidazolium, whereas either an in-motif or non-motif (indicated by ‘b’ Figure 2B) T^ImH+ can be introduced at any other position relative the first in-motif T^ImH+. When introduced on the complementary strand, all positions n–i allow for both type of motifs (indicated by ‘b’), as will positions starting at n+5. At n+4 however, the local base pair context of the first T^ImH+ again prevents occurrence of the pK_a-motif for the second one, leading it to be non-motif. Finally, the position n+3 (Figure 2B, labelled ‘x’) represents a particular case only encountered when starting from an in-motif duplex. Indeed, it is in principle possible for a T^ImH+ introduced at this position to form a pK_a-motif with the G at n+1 on the first strand. However, this is already part of the pK_a-motif of the T^ImH+ at position n. Both pKa-motifs are mutually exclusive for steric reasons as they cannot simultaneously fit within the same major groove area. As we foresee this introduces an ambiguous, uncontrollable state, this position is best avoided. When two T^ImH+ nucleotides capable of adopting the pK_a motif are introduced in the same sequence, a variety of topologies may arise, depending on the relative position of the two (T*GC·GCA) sequences and the fact whether or not the T^ImH+ nucleotides are on the same or opposing strands, as listed in Supplementary Figure S2.

Figure 2.

Overview of possible configurations when combining two T^ImH+ nucleotides within an otherwise unspecified dsDNA sequence. (A) and (B) provide a schematic evaluation of the positions amenable to introduction of a second T^ImH+ in the presence (indicated by T* in the sequence) of a non-motif respectively in-motif T^ImH+ nucleotide. The letter ‘b’ (for both) is used to indicate that a second T^ImH+ introduced at this position may adopt an in-motif or non-motif configuration depending on the local base-pair context. When only a non-motif configuration is possible, this is indicated with ‘n’. The pK_a-motif is shown as explained in Figure 1B. The ‘x’ indicates a position where the outcome of introducing a second T^ImH+ creates an ambiguous situation, because two pK_a-motifs overlap (see text).

Obviously, many constructs can be considered when introducing two (or more) T^ImH+ nucleotides in a dsDNA sequence depending on the duplex sequence used as template and their relative position. For our current investigation, we chose to start from the fully matched version of the 14mer duplex sequence of the original DNA template used in our previous study of singly modified constructs (Figure 3, model duplex) (51). This model duplex was designed with six AT base pairs, each amenable to T^ImH+ introduction, while affording good spectral resolution for NMR investigations. This choice also affords to use the same unmodified model sequence as a single reference construct and allows for comparison with our previous data. In total, 15 double T^ImH+ constructs are possible using this DNA template, too large a number for each to be subjected to our combined in silico and NMR characterization approach. Here, a total of four doubly imidazole modified constructs depicted in Figure 3 were selected for synthesis and analysis. These will be referred to as (T_xT_y)^ImH+ throughout the text, whereby an italicized T_x indicates an in-motif T^ImH+ nucleotide. In our particular model duplex, non-motif T^ImH+ constructs arise when introducing a T^ImH+ at position 6 (T₆^ImH+) or 7 (T₇^ImH+), while introduction at position 8 (T₈^ImH+) leads to an in-motif duplex. Therefore, comparison of (T₆T₈)^ImH+ with (T₇T₈)^ImH+ allows to judge the impact when bringing two T^ImH+ nucleotides closer together within the same strand while one is in-motif. In (T₆T₇)^ImH+ and (T₇T₈)^ImH+ both T^ImH+ are immediate neighbours, but one T^ImH+ goes from being non-motif to in-motif respectively, allowing to judge the impact of in-motif formation while keeping relative strand position constant. A fourth construct, (T₆T₁₈)^ImH+ places the in-motif T^ImH+ on the opposite strand. This inverts the directional positioning of the imidazolium (see arrow in Figure 3) such that it is expected to interact with G₉(C₂₀) in the major groove, i.e. spatially very near to (C₁₀)G₁₉^, itself involved in-motif formation when a T^ImH+ is introduced at position 8. For this reason the (T₈T₁₈)^ImH+ construct was excluded as it would have overlapping pK_a-motifs (each T^ImH+ occupying position ‘x’ in Figure 2B with respect to the other one) creating an ambiguous state (see above). Considering all other possible doubly imidazole modified constructs based on our 14mer duplex template, the nine involving substitution of either T₂₄ and/or T₂₇ for a T^ImH+ were not considered as this would involve the first three base pairs which we judged to be less representative for the typical major groove environment well inside the sequence. Also, a T^ImH+ at either position would only yield a non-motif situation, leading to non/non or non/in-motif situations already covered by systems described above. This also applies to the (T₇T₁₈)^ImH+ construct which was a priori judged insufficiently distinct from the already selected ones to be explicitly considered. Potential consideration of a (T₈T₁₈)^ImH+ was discarded in view of the mutual exclusion of one pK_a motif by the other as can be derived from Figure 2. No unambiguous double in-motif construct can be designed without introducing changes to the model sequence, a disadvantage of the model duplex which results from the fact that the pK_a motif was only discovered after the model's inception (53). While we could have considered alternative sequences for in/in-motif investigations, the results obtained from analyzing non/non and non/in-motifs described above (T₇T₈T₂₃)^ImH+, led us to immediately move to study a triply modified system that carries two in-motif modifications, each flanking one non-motif imidazole. All these systems were compared to the singly-modified in-motif reference duplex system T₈^ImH+ complemented by the non-motif T₂₁^ImH+ one (Figure 1B). Using all these constructs, the impact of various spatial proximities between two or more imidazolium functionalities on pK_a and duplex stability was investigated.

Figure 3.

Summary of all duplex systems featuring double and triple introduction of a T^ImH+ nucleotide (represented by T*). The top sequence shows the model duplex used in this and previous work (51). The box indicates pKa-motifs and the arrow indicates which guanine is engaged via its Hoogsteen side in the hydrogen bond with the imidazolium moiety. T^ImH+ (T*) are colour labelled as orange or blue depending on whether an in-motif respectively non-motif situation is expected to result at this position taking into account the base pair context in this particular duplex. The bottom sequence represents the triply modified system featuring two adjacent pK_a-motifs separated by a single non-motif T^ImH+. In the labels naming the various constructs, in-motif T^ImH+ are italicized for easy recognition.

In silico evaluation of the doubly modified duplex constructs

Molecular Dynamics (MD) simulations of the four (T_xT_y)^ImH+ constructs were performed to computationally validate the expectations regarding the presence or absence of the pK_a-motif in the presence of multiple T^ImH+, as outlined above. Additionally, this also provides three-dimensional insight in the behavior of the duplex and the imidazolium moieties at the atomic level, thereby assisting the interpretation of experimental data. Given that previous pK_a values are ∼8 or higher in singly modified systems while experimental data is recorded at pH 7, simulations were only performed with both modified nucleotides in their protonated states.

The pKa-motif entails proximity of the imidazolium to the Hoogsteen side of the guanine at position n+2 on the opposite strand and specifically the formation of a T^ImH+(NHε2)---G(O6) hydrogen bond across both strands (Figure 1C). Since calculations are initiated with the imidazolium directed towards the solvent, this distance is typically between 12 and 14 Å to start with, providing unbiased starting positions to evaluate in-motif configurations. Several descriptors were evaluated from the MD simulations and are collected in Table 1. As before, we characterized the ‘occurrence’ of this hydrogen bond in the T^ImH+ modified duplex constructs using the fraction (in %) of the simulation time where both the associated distance d(D---A) between donor and acceptor as well as angle θ(D–H---A) fall within the required range, specified as ≤3.0 Å and ≥150° respectively. Here, we also use the ‘persistence’ in spatial proximity between the interacting partners in the major groove as the fraction (in %) of the simulation time where only the distance criterion was satisfied. Finally, the ‘excursion’ defines the fraction of the simulation time where this distance is larger than 6 Å, indicating an excursion out of the major groove towards the solvent. Using the latter two descriptors, cases where frequent and/or longer time excursions of the imidazolium moiety well away from the guanine O6 acceptor occur during the simulation can be distinguished from those where the hydrogen bond is no longer optimal, yet the imidazolium remains very near to the O6 acceptor (Supplementary Figure S3). For instance, the singly modified in-motif T₈^ImH+ reference duplex, features an occurrence of 27.4% with a persistence of 57.3% and an excursion of only 10.5% (51). These numbers indicate that shortly after the start of the simulation, typically within the first 2–3 ns, the imidazolium engages with the Hoogsteen face of G₁₉ and remains very close, forming geometrically sound hydrogen bonds for considerable time. These values are considered to reflect an in-motif situation.

Table 1.

In silico analysis of hydrogen bond properties for pKa-motifs in singly, doubly and triply modified model duplex systems

		H-bond[a]
System	Motif	Occurrence (%)	Persistence (% ≤ 3.0 Å)	Excursion (% ≥ 6.0 Å)
T21ImH+	non	n.a.	n.a	n.a
T8ImH+	in	27.4	57.3	10.5
T18 ImH+	in	10.3	23.4	27.6
(T6T8)ImH+	non/in	32.8	64.9	7.5
(T7T8)ImH+	non/in	17.5	40.4	3.7
(T6T18)ImH+	non/in	28.5	64.9	3.7
(T6T7)ImH+	non/non	n.a	n.a	n.a
(T7T8T23)ImH+	non/in/in	20.3/40.1	49.2/77.0	6.2/16.2

^[a]Values are determined for the H-bond between the in-motif T_x^ImH+Nϵ2 hydrogen atom and the guanine carbonyl O6 at position n+2 on the opposite strand. In-motif T^ImH+ are in italics. For the definition of occurrence, persistence and excursion, see text. n.a. signifies not-applicable.

Considering the various doubly modified (T_xT_y)^ImH+ constructs, the excursions stay well below 10% in all cases, while the persistencies and occurrences show values that further support the presence of an in-motif situation for T₈^ImH+ or T₁₈^ImH+. Thus, all expected in-motif situations are indeed found to occur in silico.

Experimental validation of the pK_a-motif in doubly modified constructs

Based on our previous work on singly modified T^ImH+ duplex constructs, four experimental criteria were defined that accompany the effective presence of the pK_a-motif. These are: (i) an increase in UV measured T_m value of 6 ± 1°C compared to the unmodified duplex; (ii) an increase in imidazolium pK_a to a value around 8.5–9.0, i.e. an increase with approximately one unit compared to the non-motif situation where values are in the 7.5–8.0 range, itself above the value of the individual nucleotide at 7.2; (iii) a significant downfield shift of the imidazolium Hϵ1 proton in the ¹H NMR spectrum to 9.0 ppm compared to the 8.7 ppm of the Hϵ1 resonance when the group is solvent-exposed and (iv) a clear set of nOe contacts between the Hϵ1 proton of the imidazolium group at position n to the nucleotide at position n+3 on the complementary strand. The latter criterion is very well illustrated by the detail of the NOESY spectrum of the T₈^ImH+ in-motif reference system showing multiple specific short range nOe contacts (Figure 4A). In contrast, the corresponding flipped T₂₁^ImH+ system used as non-motif reference system shows a dearth of such nOe's (not shown, see (51)).

Figure 4.

Overview of the specific nOe contacts of the Hε1 protons to adjacent base pairs in T^ImH+ -modified duplex systems. (A) The trace of specific nOe contacts associated with the pKa-motif in a single T^ImH+ modified duplex as illustrated by the T₈^ImH+ system. nOe's are assigned as labelled and mostly involve the ribose and base hydrogens of T₁₈, i.e. at position n+3 on the opposite strand relative to the T^ImH+ nucleotide considered at position n. (B–F) demonstrate the presence of in-motif and non-motif T^ImH+ nucleotides within doubly (B–E) and triply (F) modified duplex constructs. In most cases, the amide NH (indicated by ◊) of the imidazole linker is visible, and displays clear intra-residue nOe's (dashed boxes) with the ethylene part of the chain or the neighboring thymine methyl group (in B). All 2D NOESY were recorded at 700 MHz and 25°C using a 200 ms nOe mixing time in 90%H₂O/D₂O.

Unfortunately, the first three criteria can no longer be used as proxy for the presence of the pKa-motif when considering constructs with two or more T^ImH+ residues, because they are not exclusively sensitive to the individual formation of each pKa-motif. Indeed, the introduction of a second positively charged T^ImH+ will also contribute to the T_m value, change the environment and therefore the Hϵ1 chemical shift, while electrostatic interactions between both imidazolium moieties are bound to shift the acid-base equilibrium and therefore the pK_a of either imidazolium, all in ways that cannot be predicted. Since nOe contacts require a close proximity of the interacting ¹H’s, these do provide an independent measure of motif formation for each individual T^ImH+, independent from the presence of another T^ImH+ nucleotide. Therefore, ¹H resonance assignment was performed for all constructs, followed by assigning the nOe contacts involving the imidazolium moieties, as shown in Figure 4. In most cases, the NH resonance of the amide in the linker of both T^ImH+ constructs is also visible. For constructs where in-motif T^ImH+ are predicted and in silico validated (vide supra), i.e. (T₆T₈)^ImH+, (T₇T₈)^ImH+ and (T₆T₁₈)^ImH+, clear nOe's connecting Hϵ1 with sugar and base ¹H resonances of the nucleotide at position n+3 on the opposing strand, a thymine in all cases, are visible (Figure 4B–D). In addition, intra-nucleotide nOe's involving the amide NH of the linker are also clearly present and mostly involve the ethylene protons of the linker located between 2 and 3 ppm, or a neighbouring thymine methyl group (only Figure 4B). For imidazolium groups proposed to be non-motif these intranucleotide nOe's are the only ones present, supporting their orientation towards the solvent. In addition, analysis of the chemical shift perturbations caused by the introduction of T^ImH+ compared to the unmodified sequences (Figures S4-S10) indicates a characteristic and remarkably consistent shift by 0.45 ± 0.03 ppm to lower value for the 6-CH₃ group of the thymine located at the position complementary to n+3 (i.e. the position marked by x in Figure 2), showing nOe's with the T^ImH+ residue. This characteristic feature was already observed in singly modified systems and can be attributed to an aromaticity induced shift resulting from the immediate spatial vicinity of the interacting imidazolium moieties to the thymine methyl group. It constitutes therefore an independent marker for the occurrence of an in-motif system. We conclude therefore that all predicted in-motif and non-motif T^ImH+ nucleotides in the singly and doubly modified constructs are experimentally observed, again demonstrating the general validity of the pK_a-motif. Using these results, we can now interpret the impact on duplex stability (T_m) and imidazolium pK_a while being certain about the motif-state of the particular imidazolium of interest, thus providing guidance for further analysis.

Initial analysis of duplex stability and imidazolium pK_a in doubly modified systems

The T_m values and pK_a for the imidazolium functions, the latter obtained from monitoring their aromatic resonances as a function of pH, are collected in Table 2. As reported previously (51), the singly modified non-motif T₂₁^ImH+ reference system shows a ΔT_m of 2.1°C and a pK_a of 7.62, while the in-motif T₈^ImH+ gives characteristically higher values of both 5.2°C and 8.72 respectively. For completeness, and to allow comparison with the (T₆T₁₈)^ImH+ construct, we also determined these properties for the in-motif T₁₈^ImH+ construct as this had not been investigated before. Gratifyingly, it features, within error, an identical ΔT_m and similarly increased pK_a value as for the in-motif T₈^ImH+ construct (Table 2).

Table 2.

Trends in thermal stability using experimental and predicted melting temperatures and overview of experimental pK_a values in singly, doubly and triply modified model duplex systems

Duplex	Motif	T m a	ΔTmb	pKa	pKa	0 ≤ fi ≤ 1c	ΔmaxTmp d	ΔTmp e	(ΔTmp – ΔTm)
		(°C)	(°C)	Ti	Tj>i		(°C)	(°C)	(°C)
Model	n.a.	58.9 ± 0.2	ref
T21-model	n.a.	57.2 ± 0.2	ref*
T8 ImH+	in	64.1 ± 0.6	5.2 ± 0.6	8.72 ± 0.02	n.a.	∼1	5.2	ref	ref
T18 ImH+	in	64.2 ± 0.6	5.3 ± 0.6	8.57 ± 0.03	n.a.	∼1	5.2	5.2	-0.1
T21ImH+	non	59.3 ± 0.5	2.1 ± 0.5*	7.62 ± 0.03	n.a.	0.81	2.6	ref	n.a.
(T6T8)ImH+	non/in	65.0 ± 0.2	6.1 ± 0.3	7.82 ± 0.05	8.90 ± 0.05	0.87	7.8	7.6	1.5
(T7T8)ImH+	non/in	63.9 ± 0.4	5.0 ± 0.4	6.96 ± 0.08	8.70 ± 0.07	0.48	7.8	6.5	1.5
(T6T18)ImH+	non/in	64.7 ± 0.3	5.8 ± 0.4	8.01 ± 0.04	8.53 ± 0.03	0.91	7.8	7.7	1.9
(T6T7)ImH+	non/non	59.6 ± 0.1	0.7 ± 0.2	7.50 ± 0.05	7.64 ± 0.04	0.76/0.82	5.2	4.1	3.4
T7T8T23-model	n.a.	62.5 ± 1.5	3.6 ± 1.7
(T7T8T23)ImH+	non/in/in	71.8 ± 0.9	9.3 ± 1.8∧	7.40 ± 0.08	8.59 ± 0.08	0.72	13.0	12.3	3.0
					9.03 ± 0.02

^a1 μM strand concentration, 100 mM NaCl, 10 mM sodium phosphate, pH 7.0 The reported errors are one standard deviation as described in the materials section.

^bThe reference system used for calculations is always the model except when T₂₁-model is used as indicated by an * or when the triply modified system is used indicated by ∧.

^cProtonation degree calculated using the experimental pK_a; underlined pK_a values are those for in-motif imidazolium moieties.

^dMaximum predicted increase in T_m and [e] predicted increase taking protonation in c into account.

Before evaluating the ΔT_m values observed in the doubly modified systems (Figure 3), the pK_a values need to be addressed as they allow to assess the protonation state of the imidazolium moieties in the duplex constructs at conditions used for T_m determination, i.e. at pH 7. For all in-motif T^ImH+, the pK_a values in doubly modified constructs remain within the expected 8.5–9.0 interval, very close or identical to the values for the respective singly modified reference systems (Table 2). Thus, it appears that irrespective of the presence of a second, non-motif T^ImH+, the in-motif pK_a values are well maintained and in all cases, the in-motif imidazolium approaches 100% protonation. We conclude therefore that the acid-base equilibrium for an in-motif T^ImH+ is mostly insensitive to the presence of a non-motif T^ImH+ even when they are immediately adjacent on the same strand. In contrast, non-motif T^ImH+ pK_a values show a distinct behavior depending on the nature and position of the second T^ImH+ construct. When a second non-motif is introduced immediately adjacent to one on the same strand, like in (T₆T₇)^ImH+, the pK_a values appear little affected compared to that of the non-motif T₂₁^ImH+ reference system (Table 2). However, when an in-motif is introduced immediately adjacent a non-motif T^ImH+ on the same strand, as in (T₇T₈)^ImH+ the pK_a is significantly lowered. At 6.96, the non-motif pK_a value of T₇^ImH+ is even lower than that of the isolated T^ImH+ nucleoside, and the lowest value reported so far for any T^ImH+ modified duplex. On the other hand, separating both T^ImH+ by one nucleotide while maintaining the non/in-motif combination, as in (T₆T₈)^ImH+, causes a moderate raise in pKa of the non-motif T₆^ImH+ compared to this reference value. The fact that moving the non-motif T^ImH+ from position 6 to 7 with respect to in-motif T₈^ImH+ lowers the pKa by almost one order of magnitude while only having a marginal effect (within error) on the T₈^ImH+ pK_a is quite striking. While singly modified T₆^ImH+ and T₇^ImH+ analogues were not investigated, there is no reason to believe this variation is linked to a pK_a variation caused by the local base pair context when shifting the non-motif T^ImH+ along the strand. Indeed, the previously reported difference in pKa values between singly modified but TT mismatched sequences mT₆^ImH+ and mT₇^ImH+ is less than 0.20 pK_a units (51). Finally, comparing (T₆T₈)^ImH+ with (T₆T₁₈)^ImH+, where both in-motifs involve a guanine in the same region of the major groove (G₁₉ opposite C₁₀, versus G₉) but on opposing strands, the pK_a for the non-motif T₆^ImH+ is close to identical within error, indicating that the orientation of the pK_a-motif does not play a role in determining the pK_a of the non-motif T^ImH+. We conclude therefore that a pK_a difference of two units can be achieved within a duplex environment, providing opportunities for catalysis design.

Relative contributions of in-motif versus non-motif T^ImH+ on duplex stability

Our experimental data shows that like in all previously reported singly modified constructs, modification does not destabilize the duplex compared to the unmodified construct. In all cases a higher T_m value with ΔT_m ranging from 0.7°C up to 6.1°C is observed. To explore the contribution of stabilizing and destabilizing interactions in doubly modified constructs, we start with the assumption that the change in T_m values of +5.2 and +2.1°C observed in the singly modified reference systems may be considered as characteristic ‘reference’ values for an in-motif and non-motif T^ImH+ at pH 7. If we first assume the lack of any kind of mutual interference between T^ImH+ nucleotides these contributions may be used in an additive fashion to predict Δ_maxT_m^p, the maximum predicted value for doubly (and higher) modified constructs:

(1)

where ΔT_mⁱⁿ and ΔT_m^non represent the increase observed in the singly modified reference systems. However, care must be taken to account for the impact of the degree of protonation f_i of the imidazolium groups in the various constructs. Here, these are accessible from the experimental pK_a values (Table 2). For in-motif imidazolium groups in singly (51) and doubly modified constructs, pK_a values are above 8.5 and the degree of protonation is of no concern as it approaches f_i = 1. For a single non-motif group with pK_a around 7.5 however, the protonation degree drops to approximately 0.75. Thus, in/non motif combinations will lead to two distinct duplex populations at pH 7 depending on whether the non-motif T^ImH+ is protonated or not. This situation is further complicated for double non-motif combinations like (T₆T₇)^ImH+, where four species corresponding to doubly non-protonated, singly protonated or doubly protonated forms co-exist in solution thereby further convoluting the measurement and interpretation of the data. This notwithstanding, a simplified approach can be devised to correct ΔT_m^p values for the actual protonation degree. We propose that in most cases, ΔT_m^p, i.e. the predicted increase in T_m of a doubly imidazole-modified duplex compared to the corresponding unmodified sequence taking into account the protonation degree can be approximated as follows:

(2)

Herein, the assumption is made that any in-motif T^ImH+ is fully protonated (f_i = 1) while partial protonation of the non-motif T^ImH+ is handled by scaling ΔT_m^non with the protonation degree of the associated imidazolium. Based on the ΔT_m values observed for the T₈^ImH+ and T₂₁^ImH+ reference systems, we propose to set ΔT_mⁱⁿ to 5.2°C, while for ΔT_m^non a value of 2.6°C is proposed for a fully protonated non-motif system based on the 2.1°C increase occurring at 80% protonation of the singly modified T₂₁^ImH+ construct. Note that this approach implicitly assumes that non-protonated states do not contribute to change the T_m value. The values of Δ_maxT_m^p and ΔT_m^p for each specific construct is listed in Table 2. Because ΔT_m^non is only 2.6°C, inclusion of the protonation degree when calculating ΔT_m^p causes a small i.e. −0.1 to −0.2°C to moderate reduction i.e. −0.9 to −1.3°C, the latter due to the low pK_a (f_i = 0.48) of T₇^ImH+ in (T₇T₈)^ImH+. These protonation degree corrected ΔT_m^p values can now be compared with the experimental ΔT_m values, by calculating (ΔT_m^p – ΔT_m) as collected in Table 3.

Table 3.

Experimental trends in thermal stability of the aptamer constructs in presence and absence of the L-Arm ligand

System	Motif	T m (°C)a	ΔTm (°C)	Change (reference)c
apt	n.a.	55.2 ± 1.6*	ref*	–
apt(ImH+)	Non	56.4 ± 1.3	1.2 ± 2.0	+ T3ImH+ (*)
f2-apt	n.a.	53.5 ± 0.7∧	f2-ref∧	–
			−1.7 ± 2.0	+ double flip (*)
+ L-Armb		62.2 ± 1.7	8.7 ± 1.9	+ L-Arm (∧)
f2-apt (ImH+)	in	64.9 ± 1.1	11.4 ± 1.3#	+ T3ImH+ (∧)
+ L-Arm[b]		69.4 ± 0.8	4.5 ± 1.4	+ L-Arm (#)
			15.9 ± 1.3	+(T3ImH+ & L-Arm) (*)

^a1 μM strand concentration, 100 mM NaCl, 10 mM sodium phosphate, pH 7.0 The reported errors are one standard deviation as described in the materials section.

^bThe L-Arm concentration used was 10 mM representing a 10 000:1 L-Arm/aptamer ratio.

^c *, ∧ and # refer to the same labels introduced in column T_m in order to identify the temperature used as a reference to gauge the impact of the change mentioned.

Relatively good agreement is found when an in/non-motif construct is considered, as the difference between experiment and prediction is quite close, on average −1.6 ± 0.20°C. The similar difference suggests that some mutual interference occurs between the in- and non-motif imidazolium in these systems causing a destabilising ‘cost’ in melting temperature. The electrostatic repulsion between the positively charged imidazolium groups is the most obvious choice for this destabilising effect. Therefore, we evaluated whether the distance between both imidazolium moieties in the simulation trajectory can be correlated with the observed trends in T_m values. The average distance between the centroid of both imidazolium rings is 7.3 Å for (T₆T₈)^ImH+ and 7.4 Å for (T₇T₈)^ImH+ whereas for (T₆T₁₈)^ImH+ this is about 10.7 Å. For the double non-motif case, ΔT_m^p is simplistically calculated from the protonation weighted contribution of each individual T^ImH+ to be 4.1°C. Here, the discrepancy with the experimental ΔT_m value is −3.4°C, more than double the value seen for the other systems. This not only reflects the simplicity of the approximation in this case, but likely in also indicates that a considerable electrostatic repulsion occurs between both solvent-exposed charged rings, which are one average 7.3 Å from each other in the simulation. However, assuming the MD trajectories are representative for the actual solution behavior we can only conclude that a simple distance dependent electrostatic term only involving two fully charged imidazolium moieties does not suffice. An in-depth in silico investigation involving a full scale electrostatic computation which includes solvent screening effects is thus warranted but falls beyond the scope of current investigation.

First conclusions and guidelines

From all the above, a number of conclusions can be made concerning the mutual impact of two T^ImH+ modifications on duplex stability and imidazolium pKa as a function of their relative position in doubly modified (T_xT_y)^ImH+ systems. With respect to the impact on thermal stability, a first general guideline is that simple additivity of the individual stabilizing contributions ΔT_mⁱⁿ and ΔT_m^non observed in the non-motif and in-motif singly modified reference systems represents an excellent first estimate for the maximum gain in stability that can be expected experimentally. Indeed, even when both T^ImH+ nucleotides are in close sequence proximity, i.e. like in (T₆T₈)^ImH+ or in (T₇T₈)^ImH+, predictions overestimate the experimental ΔT_m values by 1.5 to 2°C only depending on whether protonation degrees are taken into account or not. While not explicitly investigated, we suggest that as the distance is increased between the T^ImH+ nucleotides in a particular duplex, the experimental ΔT_m may even further tend towards the predicted ones. The only important exception that must be considered occurs when two non-motif T^ImH+ nucleotides are immediately adjacent. Here more severe deviations can be expected as strikingly illustrated by (T₆T₇)^ImH+.

With respect to mutual effects on the acidity of the imidazolium moiety, the observation that an in-motif imidazolium is only weakly sensitive (on the order of |ΔpK_a| < 0.3) to the presence of other nearby T^ImH+ irrespective whether that occurs in a pK_a-motif or not is the basis for the second guideline: the pK_a for in-motif T^ImH+ is expected to be >8.5 and little affected by nearby T^ImH+ residues. The case for non-motif imidazolium moieties on the other hand, is quite different. Indeed, here one should recognize that the only guideline that can be formulated is that it is not possible to predict what to expect.

Investigation of (T₇T₈T₂₃)^ImH+—the first triply modified duplex

To build upon these guidelines and demonstrate their applicability we chose to investigate a triply modified system using the same approach outlined for the doubly modified ones. Obviously, many other systems could be considered, but we chose to start from the current doubly modified ones, to allow for optimal comparison. We settled for the (T₇T₈T₂₃)^ImH+ construct since it allows to investigate the outcome of adding a second pK_a-motif by introducing a T^ImH+ close to the non-motif/in-motif combination already investigated in the (T₇T₈)^ImH+ construct (Figure 3). In addition, it creates the possibility to address the response of the non-motif T₇^ImH+ acidity to another nearby in-motif imidazolium possibly decreasing it even further below the pK_a value of 6.96 in (T₇T₈)^ImH+. To this end the T₆A₂₃ and C₄G₂₅ base pairs were flipped while the intermediate A₅T₂₄ base pair was replaced by a CG one compared to the model duplex (Figure 3). A priori in silico analysis is consistent with the formation of two pK_a-motifs (Figure 5), involving hydrogen bonds between T₈^ImH+ and G₁₉, respectively T₂₃^ImH+ and G₄ (Figure 3) as evident from Table 1. In the triply modified construct, values for T₂₃^ImH+ reach the highest values for occurrence and persistence observed so far, the latter reaching 77%. At 16%, the excursion degree is in fact completely located within the first 8 ns of the simulation, representing the time it took for the imidazolium to lodge itself into the major groove (Supplementary Figure S3). The presence of both pKa-motifs is evident from the intense and specific nOe contacts connecting the imidazole Hϵ1 of T₈^ImH+ and T₂₃^ImH+ with sugar and base ¹H resonances of T₁₈ respectively C₃ (Figure 4F), the bases located on the opposite strand and complementary to position n+3 from the respective T^ImH+ (Figure 3). As before, T₇^ImH+ does not show such contacts. Gratifyingly, the 6-CH₃ of T₁₈ and the H5 of C₃ both show the expected perturbation in chemical shift by 0.43 and 0.40 ppm respectively (Supplementary Figure S10), indicating this feature is quite general and not dependent on the nature of the pyrimidine base as the n+3 position.

Figure 5.

Specific hydrogen bond pattern of the T₈^ImH+ and T₂₃^ImH+ imidazolium functionalities with G₁₉ respectively G₄, in triply modified (T₇T₈T₂₃)^ImH+. The hydrogen bonds are highlighted in blue dotted lines. Visualized using VMD 1.9.1.

Using our guidelines we predict that the in-motif imidazolium moieties should have a pK_a above 8.5, which is indeed the case, T₂₃^ImH+ reaching 9.03 (Table 2). The impact on the pK_a of T₈^ImH+ is marginal as expected, being in fact identical within error to the value in the (T₇T₈)^ImH+ construct. Remarkably, the pK_a of T₇^ImH+ is now 7.40 ± 0.08, i.e. 0.44 pK_a unit up from the value in this construct. Clearly, the expectation that the acidity of the non-motif T₇^ImH+ may further increase by introducing a second positively charged pK_a motif in the vicinity is not borne out. It further underlines the need for a more quantitative, structure driven rather than sequence based approach to predict the pK_a of the imidazolium groups in a DNA duplex. Recently, constant pH molecular dynamics have gained popularity to compute pK_a values of titratable residues and study pH-dependent conformational dynamics of biomolecules (70,71), thus opening possibilities for future in silico investigation of the effects impacting the pK_a of T^ImH+ modified oligonucleotides.

When it comes to predicting the change in melting temperature, the introduction of a third T^ImH+ raises the complication as to which and how and which additional mutual interactions should be taken into account. In the absence of mutual interference, a Δ_maxT_m^p value of (2*5.2 + 2.6) = 13.0°C can be predicted assuming full protonation of T₇^ImH+. Using the experimental pK_a of T₇^ImH+, this is corrected to 12.3°C. At +9.3°C compared to the T₇T₈T₂₃ reference duplex the experimental value is distinctly higher than those of singly and doubly modified T^ImH+ constructs, and approximately 3.0°C lower than the predicted maximum value. At first sight, one could rationalize this by noting that the imidazolium moieties of T₈ and T₂₃, being separated in space by 5 base pairs as a result of their hydrogen bonding to G₄ respectively G₁₉ i.e. in opposite directions (Figure 3) are too distant to one another to be significantly contributing to destabilizing interactions. Indeed, their mutual distance is 17.4 Å on average. In this case, the remaining interactions between T₇^ImH+ and T₈^ImH+, respectively T₇^ImH+ and T₂₃^ImH+, which appear equally distant along the duplex sequence, could be assumed to each contribute a destabilising term of −1.5°C or −3°C in total. While this appears to agree with the experimental finding, this is most likely coincidental. Indeed, the relative disposition of each pair is clearly different as evident from Figure 5 and features average inter-imidazolium distances of 8.9 Å respectively 16.9 Å, thus quite distinct values. Nevertheless we conclude that when introducing multiple motifs in each other's vicinity as done here, the gain in stabilization is very much evident and outweighs the loss due to possible destabilizing interactions involving the imidazolium functionalities. Thus, complementing the results from the previous section we conclude that adding multiple pK_a-motifs in short to medium sized oligonucleotides duplexes should lead to sizable and mostly additive stabilization against thermal denaturation.

Broadening the scope: application of the pK_a-motif for non-covalent stabilization of aptamers

The possibility to stabilize short duplex segments using the pK_a-motif provides so far unchartered possibilities for DNA-based aptamer development. In literature reporting aptamer research, a ligand binding event involving an aptamer is generally presented as occurring between an unfolded state and a fully folded state. However, a three step model is generally more appropriate (56,72,73), wherein unfolded (or partially folded) states exist in an equilibrium with one or a discrete collection of pre-folded binding competent states, typically featuring one or more double stranded stems. Binding, either through induced-fit or conformational selection processes then lead to the final bound aptamer state. As demonstrated by various authors (56,72,73), this coupling of a ‘switch’ or selection (K_S) and ‘sense’ (K_D) equilibrium allows for tuning of dynamic range and sensitivity of the aptamer based detection. Based upon this three-state model, the introduction of a covalent or non-covalent clip or staple, a term reminiscent of the concept of stapled peptides that is often used for conformational stabilization of short peptide sequences, at the appropriate position in the aptamer that stabilizes the binding-competent state can be considered a valuable asset. Indeed, any modification that renders population of the binding competent state more robust against changes in temperature, pH, salt conditions, etc. will increase the application potential. However, the introduction of a T^ImH+ based non-covalent staple should not overly interfere with ligand binding, for instance through modification of the binding pocket, steric effects or repulsive interactions involving charged moieties in the ligand and the tether.

As a proof-of-principle investigation that a T^ImH+ nucleotide could act as such a stabilising non-covalent staple in aptamers, the well-characterized L-argininamide (L-Arm) aptamer was chosen. It consists of a 24 deoxynucleotide single strand containing a 7 bp helical stem and a 10-nucleotide loop region (Figure 6), binding L-Arm with a binding constant K_A = 5998 (± 920) M⁻¹ (74,75). Mutagenesis experiments identified nucleotides essential for binding within the loop or base pairs just below the loop and their role could be rationalized from the solution structure of the complex as determined by Patel et al. (PDB-ID 1OLD) (76). It was found that the L-argininamide is enveloped within a cavity formed by the 10-nucleotide loop and interacts through intermolecular hydrogen bonds and stacking interactions. Given the double positive charge of L-Arm, and to change the sequence of the aptamer only to a minimal extent, only T₃ and T₁₉ in the double stranded stem were considered of interest for introduction of a T^ImH+ residue. Using the solution structure (76) T₁₉ was judged too close to the stem-loop interface and so the binding site, therefore we decided to introduce a T^ImH+ at position 3. To this effect, the C₄G₂₁ and G₅C₂₀ base pairs were flipped so as to introduce the pK_a-motif in the aptamer sequence (Figure 6). Thus, a total of four aptamer constructs were assessed experimentally: the unmodified aptamer referred to as apt and the double flipped or f₂-apt aptamer together with the corresponding T₃^ImH+ modified versions apt^ImH+ and f₂-apt^ImH+, expected to represent a non-motif and in-motif aptamer respectively.

Figure 6.

(A) The sequence and secondary structure of the L-argininamide (L-arm) binding DNA aptamer (apt) and of the L-Arm ligand (at pH 7). (B) Sequences and putative secondary structures of the two imidazolium-modified apt variants apt^ImH+ and f₂-apt^ImH+, featuring a non-motif and an in-motif T^ImH+ respectively, and the f₂-apt system serving as a non-modified reference for the f₂-apt^ImH+ construct. T* is used to indicate the T^ImH+ nucleotide. The pK_a-motif is shown as explained in Figure 1B.

The 2D NOESY spectra show the expected pattern for the modified aptamers: no relevant NOE’s for non-motif apt^ImH+ while a string of nOe's connecting the Hϵ1 from T₃^ImH+ with T₁₉ is clearly visible in f₂-apt^ImH+. This includes a nOe to T₁₉-Me6 (Figure 7)), itself displaying a notable chemical shift perturbation of 0.29 ppm (Supplementary Figures S11 and S12), both in excellent agreement with the observations for pKa-motif formation in doubly and triply modified cases described above. The more reduced value of 0.29 compared to 0.45 may reflect the impact of the shorter 7mer stem compared to the 14mer duplex studied before. With respect to thermal stability, the impact from the introduction of the double flip in f₂-apt is marginal, the T_m being identical within error (Table 3). This is also the case for apt^ImH+ where the introduction of T₃^ImH+ leads to a non-motif construct. In f₂-apt^ImH+ however, the change in melting temperature caused by introducing an in-motif T₃^ImH+ amounts to 11.4 ± 1.3°C when compared to the unmodified reference f₂-apt reference strand, an unprecedented high value for singly modified oligonucleotides. We also note from the ¹H NMR spectra that in f₂-apt^ImH+, additional base pair resonances become visible in the stem that can be assigned to the C₅A₆A₇/G₂₀T₁₉T₁₈ base pairs, whereas these remain absent in the apt^ImH+ construct (see Supplementary Figures S11 versus S12). Since these are positioned at the end of the dsDNA stem forming the hairpin, the strikingly higher T_m most likely reflects a zippering up of that end of the stem, thus contrasting with the T^ImH+ dsDNA constructs described before, where modifications were well away from the duplex ends. While all this clearly indicates that the introduction of a single pKa-motif induces major stabilization of the aptamer through tethering of its duplex stem, this does not guarantee that ligand binding is maintained. This was checked by repeating the T_m measurements on f₂-apt and f₂-apt^ImH+ in the presence of a 100-fold excess of L-Arm under conditions where complete saturation of the unmodified aptamer is normally obtained (Table 3). Upon complexation with L-Arm, the thermal stability of f₂-apt is increased by 8.7°C, a value that compares well with those for apt in the literature (75). Gratifyingly, the thermal stability of f₂-apt^ImH+ increases further, achieving 15.9°C more than the uncomplexed and unmodified f₂-apt. While the additional increase upon ligand addition to f₂-apt^ImH+ is only half that seen for f₂-apt, i.e. 4.5°C as opposed to 8.8°C, its size supports the attribution of this effect to an additional stabilization associated with ligand binding. It is interesting to note that in their original NMR investigation, Lin and Patel remarked that the DNA aptamer undergoes an adaptive conformational transition upon complex formation with L-Arm (76). They based this observation on the appearance of a structured loop but also of the A₇T₁₈ and A₆T₁₉ base pair resonances in the stem preceding the hairpin. We propose that the apparition of the same signals (vide supra) prior to L-Arm binding in the f₂-apt^ImH+ and the unusually high ΔT_m value of 11.4°C accompanying the introduction of the in-motif T₃^ImH+ in the f₂-aptamer provide initial support for an interpretation whereby the stabilization of the duplex stem by the pK_a-motif assists in pre-organizing the aptamer into a more stable, binding competent state. Both this pre-organization and the mutual electrostatic interaction between T^ImH+ and L-Arm may modulate the binding affinity, possibly in opposing ways. To respond to these questions, further in depth investigations should be performed.

Figure 7.

Overview of the specific nOe contacts involving the Hε1 protons of the (A) f₂-apt^ImH+ and (B) apt^ImH+ constructs. A specific trace of nOe's characteristic for an in-motif configuration is evident (compare with Figure 4A) for f₂-apt^ImH+, while these are totally absent for apt^ImH+ indicating a non-motif configuration. Partial resonance assignment only allows for certain nOe contacts to be assigned. 2D NOESY were recorded at 700 MHz and 25°C using a 200 ms nOe mixing time in 90%H₂O/D₂O.

CONCLUSIONS

In previous work, the thermal stabilizing effects occurring upon introduction of a single imidazole-tethered thymidine T^ImH+ were reported (51). The discovery of the pK_a-motif that generates a non-covalent staple mediated by hydrogen bond formation between the in-motif imidazolium moiety and the Hoogsteen side of a guanine on the opposite strand was described and shown to provide remarkable stabilizing properties, while increasing its pK_a value; effects not observed for non-motif imidazolium constructs. In this study, the possibility of and effects associated with the introduction of multiple imidazole-tethered thymidine T^ImH+ moieties into dsDNA sequences and a model aptamer have been investigated. The possible configurations for combining in-motif and/or non-motif T^ImH+ have been outlined and subsequently applied to design four doubly modified sequences and one triply modified one. We demonstrated that all in- and non-motif combinations expected from the sequence design are reproduced in silico using MD simulations, and can be experimentally validated using specific nOe contacts in NOESY spectra. Irrespective of the presence of a second or third T^ImH+, the individual in-motif pKa values are well maintained in a pK_a range of 8.5–9, approaching 100% protonation at physiological pH. In contrast, non-motif T^ImH+ pK_a values show a distinct and currently unpredictable behavior depending on the nature and position of the other T^ImH+ nucleotides in the construct. Nevertheless, the fact that up to two units difference in pK_a can be achieved within a duplex environment, provides interesting further opportunities for catalysis design and should motivate further computational research into the factors affecting the pK_a values.

Of more immediate application potential is the thermal stabilization effect associated with the pK_a-motif. In all dsDNA constructs, we observed that introduction of multiple T^ImH+ never decreases the thermal stability and that with a single exception, the increase in T_m can to a good approximation be predicted from suitable addition of the ΔT_m values associated with singly modified in-motif (+5.2°C) and non-motif (+2.6°C at 100% protonation) constructs. It is therefore clear that introduction of multiple pK_a-motifs in constructs containing short to medium sized oligonucleotide duplex sequences can be exploited to generate a predictable degree of thermal stabilization, each pK_a-motif raising the T_m value stepwise as is quite strikingly illustrated in Figure 8. Indeed, modified homologous duplexes containing the same number of in-motif T^ImH+ nucleotides cluster together around the same T_m value, with clear stepwise changes occurring as the number of in-motif T^ImH+ is increased.

Figure 8.

Graph of the melting temperatures of the singly, doubly and triply modified dsDNA constructs shown along the abscis. The data is shown clustered as a function the number of in-motif T^ImH+ present increasing from 0 to 2.

Furthermore, we demonstrated the usefulness of the T^ImH+ based pKa motif for the design of a modified aptamer with enhanced stability prior to ligand binding. Next to expanding the chemical diversity, modification of aptamers with side chains reminiscent of those of amino acids has been previously shown to enhance stability against nucleases and improve target affinity (77). Here, we show that incorporation of a imidazole-tethered thymidine nucleotide at a well-considered position in the L-Arm aptamer greatly increased the thermal stability by stabilizing the dsDNA stem at the basis of the hairpin binding site. It stabilizes a conformation resembling the one competent for ligand binding, while maintaining L-Arm binding capacity. More detailed investigations into the nature of all factors contributing to this effect and the potential for stabilization of alternative aptamer–ligand complexes are ongoing and will be reported upon in due course. Given a recent report on the sequence independent increased affinity of imidazole-tethered aptamers for the negatively charged ligand ATP, we foresee interesting applications of our T^ImH+ motif in such systems, where additional attractive electrostatic interactions with the ligand can further enhance the aptamer performance (23).

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

University of Ghent [1.5.186.03 to V.G., B.V.G., J.V.D.B.]; Research Foundation – Flanders (FWO) [G.0422.13 to J.C.M. and A.M. for a position filled by D.B.]. Funding for open access charge: FWO [G.0422.13].

Conflict of interest statement. None declared.