Several structural motifs cooperate in determining the highly effective anti-thrombin activity of NU172 aptamer

Sample preparation and crystallization

The human D-Phe-Pro-Arg-chloromethylketone (PPACK)-inhibited thrombin was purchased from Haematologic Technologies (USA). The oligonucleotides (Supplementary Figure S1) were purchased from Sigma-Aldrich (United Kingdom).

Two stock solutions of NU172 were prepared at a concentration of about 0.1 mM, dissolving solid lyophilized oligonucleotides in 10 mM potassium phosphate buffer pH 7.4 and 100 mM KCl (K buffer) or 10 mM sodium phosphate buffer pH 7.4 and 100 mM NaCl (Na buffer). In order to induce folding, all aptamer samples were annealed by heating to 90°C for 5 minutes and then slow cooling down in 50–60 minutes and storing at 20°C overnight.

A standard protocol was followed for the preparation of the complexes between thrombin and NU172 in presence of potassium (thrombin–NU172–K) or sodium (thrombin–NU172–Na) ions (19,28–31). The thrombin-NU172 complexes were concentrated to about 10 mg/ml using 10 kDa-cutoff Centricon mini-concentrator (Vivaspin 500, Sartorius, Goettingen, Germany) and a refrigerated centrifuge (Z216MK, Hermle Labortechnik, Wehingen, Germany).

An initial extensive screening of sitting-drop crystallization experiments at 20°C in 96-well plates (Greiner Bio-One, Monroe, NC, USA) was carried out using an Automated Protein Crystallization Workstation (Hamilton Robotics) and precipitant solutions of a commercially available crystallization screen (Hampton Research Index). Optimization of starting conditions was performed by hanging drop vapour diffusion method mixing 0.5 μl complex solution with 0.5 μl reservoir solution at 20°C. Crystals suitable for X-ray diffraction data collection grew, for both complexes, in 50% (v/v) Tacsimate™ pH 7.0 (Supplementary Figure S2) (32).

Data collection, structure determination and refinement

Thrombin–NU172–K diffraction data were collected at the Institute of Biostructures and Bioimages (IBB), CNR, Naples, Italy, using a Saturn944 CCD detector. The X-ray radiation used was CuKα radiation (λ = 1.5418 Å) from a Rigaku Micromax 007 HF generator. Thrombin–NU172–Na diffraction data were collected at the ID30A-1 beamline of European Synchrotron Radiation Facility (ESRF), Grenoble, France, using λ = 0.9677 Å. In both cases, crystals were cryoprotected by addition of 20% (v/v) glycerol to the crystallization solution, flash-cooled at 100 K in supercooled N₂ gas and maintained at this temperature during the data collection. Thrombin–NU172–K dataset was processed using HKL2000 (33), while that of thrombin–NU172–Na using iMOSFLM (34) and POINTLESS/AIMLESS (35,36).

Crystals of thrombin-NU172-K and thrombin-NU172-Na belong to the orthorhombic space group I222 and diffract X-rays up to 2.50 or 2.80 Å resolution, respectively. Matthews’ coefficient calculations suggested the presence of a 1:1 complex in the asymmetric unit and a solvent content of 74.7% in thrombin–NU172–K complex and 76.1% in thrombin–NU172–Na complex. Detailed statistics on data collection are reported in Supplementary Table S1.

The phase problem was solved by molecular replacement using Phaser (37). The coordinates of the native protein (PDB code: 1PPB) (38) were utilized as search model. The starting model was subjected to few cycles of rigid body refinement followed by several cycles of coordinate minimization and B-factor refinement using REFMAC5 program (39). Analysis of Fourier difference maps, calculated with (F_o – F_c) and (2F_o – F_c) coefficients, and manual model building performed with WinCoot program (40), allowed the rebuilding of the whole aptamer, of a sugar chain at the N-glycosylation site on thrombin, the positioning of ions and of several water molecules.

The final R_factor/R_free values were 16.9/20.5 and 15.9/20.3 for thrombin-NU172-K and thrombin-NU172-Na complexes, respectively. Detailed statistics of the refinement are reported in Supplementary Table S2.

Structural analysis

3DNA-dssr (41) was used to calculate local and overall geometric parameters of the aptamer. The Superpose program (42) from CCP4 package (43) was used to calculate root-mean-square deviations (RMSD). Features of the thrombin-NU172 interface and interactions between the two molecules were calculated by PISA (44), Contact (CCP4 package) (43) and CoCoMaps (45) programs. All the results were verified by visual inspection of the structure with WinCoot (40).

Molecular graphics figures were prepared with PyMOL (DeLano Scientific, Palo Alto, CA, USA). The coordinates of the structures were deposited in the Protein Data Bank (codes 6EVV and 6GN7 for thrombin-NU172-K and thrombin-NU172-Na, respectively).

Circular dichroism measurements

Oligonucleotides were suspended in K buffer at a concentration ranging between 1 and 40 μM and annealed, as reported above. CD spectra were recorded on a Jasco J-715 spectropolarimeter equipped with a Peltier temperature control, using a 1.0 or 0.1 cm path length cell. Spectra were registered at 10°C in the 210–320 nm range with 50 nm/min scanning speed, 2 s response time, 1 nm data pitch and 2.0 nm bandwidth. Each spectrum was obtained averaging three scans.

Gel electrophoresis (PAGE) experiment

A 20% polyacrylamide gel was prepared using 30% acrylamide/bis-acrylamide solution (24:1). The electrophoretic run was carried out in 90 mM Tris–borate buffer pH 8.0 and 20 mM EDTA (TBE buffer) on a 20 × 20 cm slab, at 4°C, by applying to the heads of the electrophoresis cell a 120 V voltage. For loading, 30 μl of 40 μM oligonucleotide sample, annealed in K buffer, were added to 20 μl of loading buffer (TBE buffer and 20% (v/v) glycerol). After a 4 h run time, bands were visualized by UV shadowing.

Anticoagulant activity experiments

Evaluation of the anticoagulant activity of TBA, NU172 and the mutants Des_NU172 and TBA_GT (Supplementary Figure S1) was performed by means of light scattering (LS) experiments monitoring the increase of the light scattered intensity upon conversion of fibrinogen to fibrin catalysed by thrombin (46). LS measurements were carried out with a home-made instrument. A 1.2 μM solution of fibrinogen in Phosphate Buffered Saline (PBS) was placed in a LS cuvette to which the oligonucleotide was added and left to equilibrate in the instrument for 20 min. Then, thrombin was added up to a final concentration of 5 nM and the light scattered intensity was registered every 20 s for at least 2 h. In the case of NU172 longer observation times were required. Each oligonucleotide, annealed in K buffer, was added to fibrinogen solutions before addition of thrombin at a concentration such that thrombin:oligonucleotide molar ratio was either 1:1, 1:2 or 1:5. All the experiments were performed in triplicate. Comparison of the anticoagulant activities was performed by plotting the normalized scattered intensity nI

as a function of time and by linearly fitting the initial increase of nI upon the lag time. The slopes of the fitted lines give an estimation of the coagulation rate. The thrombin inhibiting properties of the different oligonucleotides were evaluated by calculating the ratio between the coagulation rate in the exclusive presence of thrombin and that in the presence of thrombin and oligonucleotides.

Crystal structure of the thrombin-NU172 complex

Orthorhombic I222 crystals with one 1:1 protein–aptamer complex (Figure 1A and Supplementary Figure S3) in the asymmetric unit were grown either in the presence of K⁺ (thrombin–NU172–K) or in the presence of Na⁺ (thrombin–NU172–Na). They are practically isomorphous and diffract X-rays up to 2.50 and 2.80 Å resolution, respectively. The two independently refined complexes, within the experimental errors, do not show significant differences and, therefore, only thrombin–NU172–K complex is analysed in detail.

As expected, the aptamer binding does not substantially modify the thrombin structure. After superposition of the C_α heavy chain and that of the PPACK inhibited enzyme (PDB code: 1PPB) (38) the value of RMSD is only 0.34 Å. The only structural novelty is the long N-linked sugar chain attached to Asn60G residue (Supplementary Figure S4) that is clearly visible in the electron density map of the complex.

NU172 structural features

Fourier difference maps showed continuous electron density that allowed to build the whole oligonucleotide (Supplementary Figure S5). The refined model reveals the presence of the expected duplex and quadruplex domains (47), enchained to each other with interesting new features. In particular, the quadruplex domain (residues 7–21) is arranged in a chair-like anti-parallel fold, with guanines 8, 11, 17, 20 and 7, 12, 16, 21 forming G-tetrad I and II, respectively, and three edge-wise loops, TT (Thy9-Thy10), GT (Gua18-Thy19) and GTA (Gua13-Thy14-Ade15). A potassium ion is sandwiched between the two tetrads (Figure 1A). The quadruplex motif is very similar to that of TBA (Figure 1B) and its variants (19,29,31), which is formed by two G-tetrads (guanines 2, 5, 11, 14 for tetrad I and 1, 6, 10, 15 for tetrad II), surrounded by a TGT loop on one side and two TT loops on the opposite side. Differently from the analogous TGT loop of TBA-like aptamers that is highly mobile (19,29,31), the GTA loop is particularly well defined. In the latter, Gua13 is hydrogen bonded to Gua18 of the GT loop (Figure 2), a residue that replaces the thymine of the TBA-like aptamers, moreover Thy14 and Ade15 form reverse Hoogsteen base pairs with Ade6 and Thy5, respectively. Remarkably, Ade6 also participates to two non-linear Watson-Crick hydrogen bonds with Thy22, thus creating a TAT triad (Thy22/Ade6/Thy14). This triad and the base-pair Thy5/Ade15 mediate the transition from the quadruplex to the duplex domain, creating a continuous stacking of bases from one domain to the other. The duplex region of NU172 is composed by four Watson–Crick base pairs (Cyt1/Gua26, Gua2/Cyt25, Cyt3/Gua24 and Cyt4/Gua23), which define a regular B-type helix (Figure 2 and Supplementary Figure S6).

Thrombin-NU172 interface

Resembling TBA and its variants (19,28,29,31,48,49), the two-residue loops of the G-quadruplex domain act as a pincer-like system that captures the protruding region of exosite I (Figure 1 and Supplementary Figure S7). However, recognition of exosite I by NU172 presents interesting new features. Indeed, in addition to the two-residue TT and GT loops, the three-residue GTA loop of the G-quadruplex is also involved in the binding. On the other hand, the duplex domain makes no contacts with the thrombin surface.

Focusing on exosite I, it can be described as formed by two sub-regions, referred to as A and B, respectively (48). In the former, a hydrophobic crevice on the protein surface forms a well-shaped joint that interlocks the Gua18–Thy19 loop. As mentioned above, Gua18 shares two hydrogen bonds with Gua13 of GTA loop. The strong interaction fixes Gua13 in a position most favourable to interact with Arg75, whose guanidinium group is fully surrounded by NU172 side chains and, in addition to Gua13, is hydrogen bonded to Thy19 and Thy10 (Figure 3A and B). This network of side chains contacts recruits the three-residue loop of the G-quadruplex for the binding to thrombin and represents a unique feature of NU172 that significantly enlarges the contact area between NU172 and the A-region, compared to the complexes between thrombin and other aptamers (Table 1 and Supplementary Table S3). As far as the TT loop is concerned, it interacts with B-region, which is less extended than the A-region (48), through a π-π stacking between Thy9 and Tyr76 (Figure 3C, Table 1 and Supplementary Table S3).

It should be recalled that TBA can bind thrombin in two different ways that are practically isoenergetic. In one case, a TT loop interacts with the A-region and the second TT loop interacts with the B-region, in the second case these interactions are reversed through a rotation of 180° about the pseudo twofold axis of the aptamer (48,50). In the case of NU172, the replacement of a T for a G in one TT loop discriminates between the two binding modes, strongly favouring the one in which the GT loop interacts with the A-region of the protein.

Finally, it is worth to stress that protein-aptamer contacts do not change in the thrombin–NU172–Na structure (Supplementary Table S3), in contrast to what observed for thrombin-TBA complex (19), where the replacement of K⁺ with Na⁺ causes a small reorganization of the protein–aptamer interface that may explain the observed modification of the aptamer functionality.

Characterization of NU172 mutants

In order to gain a deeper knowledge of the structural features that are at the basis of the improved affinity of NU172 with respect to TBA-like aptamers, two mutants were constructed and characterized in solution. In particular, attention was focused on the role of the GT loop and that of the duplex module (Supplementary Figure S1), by studying the mutant TBA_GT of TBA with a T→G substitution in the appropriate loop (Thy12–Thy13), and the deletion mutant Des_NU172, in which only the G-quadruplex domain of NU172 is conserved. In particular, we tested their aggregation status, folding in solution and in vitro anticoagulant activity in comparison to TBA and NU172.

The polyacrylamide native gel electrophoretogram (Figure 4) exhibits a single band for TBA and NU172 (lanes 1 and 2, respectively), indicating the presence of a single structural species, with a relative mobility in fully agreement with their size. Vice versa, several bands are observed in lanes 3 and 4, suggesting that Des_NU172 and TBA_GT form non-homogeneous solution containing species with different molecularity.

Similar conclusions can be drawn by the CD spectra recorded under the same conditions (Figure 5A). While the spectra of TBA and NU172 show features that indicate the presence of a canonical G-quadruplex structure, those of Des_NU172 and TBA_GT suggest conformational heterogeneity. Interestingly, at lower concentration (1 μM) the spectrum of TBA_GT becomes superimposable to that of TBA, whereas the spectrum of Des_NU172 still indicates the presence of more than one conformer (Figure 5B).

These CD data, together with the electrophoretic results, indicate that the substitution of Thy12 with a guanine residue in the TT loop of TBA and, in particular, the deletion of the duplex region in NU172 induce a marked destabilization of the antiparallel unimolecular G-quadruplex conformation and an increase in its conformational flexibility.

Anticoagulant activity of NU172 and its mutants

The anticoagulant activities were evaluated following the thrombin catalysed conversion of fibrinogen to fibrin by means of LS experiments (46). In Figure 6 the intensity of the scattered light, which measures the progression of fibrin formation, is reported as a function of time in the absence and in the presence of the oligonucleotides (1:2 thrombin:oligonucleotide molar ratio). The various curves clearly indicate the inhibition of thrombin coagulation activity exerted by all tested oligonucleotides. Remarkably, NU172 acts as a very potent thrombin inhibitor. Indeed, for a rather long time no change of the intensity is observed. In the case of the other oligonucleotides, the increase of the intensity can be noted much sooner and in few hours the samples gelifies, similarly to the samples in which only thrombin is present.

An estimate of the anticoagulant activity of each oligonucleotide was obtained by calculating the ratio between the initial slope of the normalized scattered intensities in the absence and in the presence of the oligonucleotide (Supplementary Table S4). The results show that NU172 is able to decrease the coagulation rate of about two orders of magnitude already at 1:1 thrombin:oligonucleotide molar ratio, while the two mutants are thrombin inhibitors as effective as TBA.