Development of an Efficient G‐Quadruplex‐Stabilised Thrombin‐Binding Aptamer Containing a Three‐Carbon Spacer Molecule

Abstract The thrombin‐binding aptamer (TBA), which shows anticoagulant properties, is one of the most studied G‐quadruplex‐forming aptamers. In this study, we investigated the impact of different chemical modifications such as a three‐carbon spacer (spacer‐C3), unlocked nucleic acid (UNA) and 3′‐amino‐modified UNA (amino‐UNA) on the structural dynamics and stability of TBA. All three modifications were incorporated at three different loop positions (T3, T7, T12) of the TBA G‐quadruplex structure to result in a series of TBA variants and their stability was studied by thermal denaturation; folding was studied by circular dichroism spectroscopy and thrombin clotting time. The results showed that spacer‐C3 introduction at the T7 loop position (TBA‐SP7) significantly improved stability and thrombin clotting time while maintaining a similar binding affinity as TBA to thrombin. Detailed molecular modelling experiments provided novel insights into the experimental observations, further supporting the efficacy of TBA‐SP7. The results of this study could provide valuable information for future designs of TBA analogues with superior thrombin inhibition properties.


Introduction
Oligonucleotide therapies have great potential and have attracted significant interest in recent years. Aptamers are ac lass of short,s ingle-stranded oligonucleotide sequences that are able to bind target molecules with high affinity and specificity because of their ability to adopt three-dimensional structures. Aptamers, often termedc hemical antibodies, are generally developedb ya ni nv itro selection process called systematic evolution of ligands by exponentiale nrichment (SELEX). [1][2][3] The developed full-length aptamers can be chemically fabricated and truncated by using predicted secondary structurest of urther improve their biophysical properties. [4] The thrombin-binding aptamer (TBA) is one of the moststudied G-quadruplex forminga ptamers. [5] It is a1 5-mer G-rich DNA oligonucleotide (5'-GGTTG GTGTG GTTGG-3', Scheme1)t hat inhibits the formation of fibrin clots by bindingt o thrombin, [5c] which makes TBA atherapeutically relevant anticoagulant. However, TBA did not meet clinical expectationsd uring trial stages due to suboptimal dosing profiles;c onsequently,d evelopment beyond phase Ic linicals tudies was not initiated. [6] Since then, the investigation and improvement of TBA have been continued. [5] X-ray crystallography and NMR spectroscopy studies revealed thatT BA forms ac hair-shaped antiparallel G-quadruplex. [7][8][9] The structure consists of two stacked G-tetrads connectedt hrough three edgewise loops:atop central TGT and two bottomT Tl oops (Scheme 1). With regardt ot he structure and interaction with thrombin, an umber of investigations have shown that the loop regions of TBA are crucial for thrombin interaction and G-quadruplex formation. [10,11] Previous structure-activity relationship( SAR) studies revealed that loop The thrombin-binding aptamer (TBA), which shows anticoagulant properties, is one of the most studied G-quadruplex-forming aptamers. In this study,w einvestigated the impact of different chemical modifications such as at hree-carbon spacer (spacer-C 3 ), unlocked nucleic acid (UNA) and3'-amino-modified UNA (amino-UNA) on the structural dynamics and stability of TBA. All three modificationsw ere incorporated at three different loop positions (T3, T7, T12) of the TBA G-quadruplex structure to result in as eries of TBA variants and their stabilityw as studied by thermal denaturation;f olding was studied by circu-lar dichroism spectroscopy and thrombin clottingt ime. The results showedt hat spacer-C 3 introduction at the T7 loop position (TBA-SP7)s ignificantly improved stabilitya nd thrombin clottingt ime while maintaining as imilarb inding affinity as TBA to thrombin. Detailed molecular modelling experiments provided novel insights into the experimental observations, furthers upportingt he efficacy of TBA-SP7.T he results of this study could providev aluable information for future designso f TBA analoguesw ith superior thrombininhibition properties. Scheme 1. G-quadruplex structureo ft he thrombinbinding aptamer. residues T4, T9, T13 and G8 are critical to preserve the G-quadruplex structure, and that T3, T7 and T12 are more flexible moieties that are particularly involved in thrombini nhibition. [12][13][14][15] Krauss et al. suggested that the TT loops form ap incer-like structuret hat binds the protrudingr egion of thrombine xosite I. [11] In addition, it was further suggested that the stability and rigidity of TBA is important for the interaction between the aptamer and exosite I. [16] Al arge number of structural modifications have been tested to improve the activity and stabilityo fT BA, often resulting in decreased stability of TBA compared to unmodified TBA. [12-14, 17, 18] However,i nt he last year,s everal studies succeeded in improving upon TBA.P asternak et al. reported that the systematic introductiono fasingle UNA-U nucleotide at position T3, T7 or T12 increased the thermodynamic stability of TBA, and the presence of unlocked nucleic acid (UNA)-U at position 7s howed high potency with regard to inhibition of fibrin clot formation. [14] Borbone and colleagues reported similarr esults by modifying TBA at position T7 with an acyclicp yrimidine analogue. [13] In two recent studies, Virgilio et al. was ablet oi mprove the functionality of TBA. In the first study,t hey introduced 5'-fluoro-2'-deoxyuridine residues to positions T4 and T13, achieving remarkable improvements in melting temperatures and anticoagulant activity. [19] In the second study,t hey added an extra residue at the 3'-end or at both ends of the originalT BA sequence, linked through3 '-3' or 5'-5' phosphodiester bonds, resulting in strong improvements in the thermal stabilityo fTBA. [20] Herein, we systematically investigated the effect of three differentc hemical modifications-3'-amino-modified UNA (amino-UNA), unlocked nucleic acid (UNA) and at hree-carbon spacer (spacer-C 3 ;S cheme 2)-at loop positions T3, T7 and T12, with the goal of improving the binding interactions with thrombin.I nt he amino-UNA monomer included in this study, the 3'-OHg roup was replaced with aN H 2 group;t hus, the linkage to the next nucleotide becamea5 '-2' internucleotide phosphate linkage. As mall library of modified TBAs equences was synthesised (Table 1) by varyingt hree substitutions of residues T3, T7 and T12 with amino-UNA,U NA and spacer-C 3 modifications (Table 1). The sequences were tested for quadruplex folding by circular dichroism( CD) spectroscopy,t hermal stabilityo ft he formed structure by melting temperature (T m ) analysis, at hrombin clotting time assay and thrombinb inding affinity analysis of the most potent aptamerc andidates by biolayer interferometry.

Results
Physical properties and structural characterisation of modified TBA sequences Modified TBA sequences were characterisedb yC Ds pectroscopy to understand the effects of spacer-C 3 ,a mino-UNA and UNA modifications on the overall G-quadruplex structure of TBA ( Figure 1). All spectra showed the characteristic TBA absorbance [7] -two maximaa t% 240 and % 295 nm andaminimum at % 265 nm-thus suggesting the formation of antiparallel G-quadruplex structuresb yt he modification of TBA loops with spacer-C 3 ,amino-UNA and UNA.
Next, the thermals tability of the quadruplexes formed by the TBA derivatives were analysed by UV melting studies (Figure S1 in the Supporting Information). The results showed a noteworthy increase( + % 6 8C) in thermals tabilityf or substitutions with as pacer-C 3 (TBA-SP3,T BA-SP7 and TBA-SP12), but the TBA variants with UNA monomers (TBA-U3, TBA-U7 and TBA-U12) showed an increase of + +2t o+ +5 8C ( Ta ble 2a nd Figure S1). However,t he introduction of amino-UNA to the loops (TBA-A3,T BA-A7 and TBA-A12) decreased the thermal stability in comparison with the unmodified TBA (Table 2).

Biological effect of the modified TBA sequences
To investigate the ability of the modified TBA sequences to inhibit the enzymatic activity of thrombin and to evaluatet he impact of quadruplex thermal stability on biological potency, Scheme2.Structure of the three monomer modificationsu sed in this study: 1) spacer-C 3 (X);2 )amino-UNA (aU);and 3) UNA (uU).   the sequences were tested in at hrombin clotting time assay where the change in absorption, due to thrombin-induced clottingo ff ibrinogen in phosphate-buffered saline (PBS), was spectrophotometrically measured at 380 nm. The time needed to clot 50 %o ft he availablef ibrinogen ("clotting time") was then determined. The clottingt imew ithout aptamer interference was (29 AE 6) s. The majority of the modified TBA sequences showed decreased inhibitory efficiency compared to un-modifiedT BA, but TBA-SP7 showed remarkablye nhanced clotting time ( Table 2). To validate this remarkable differenceo f TBA-SP7 in the clotting time assay,f urtheri nvestigation of the thrombin aptamer bindinga ffinity was investigated by biolayer interferometry using the BLITz instrument platform (ForteBio). However,T BA and TBA-SP7 exhibiteds imilar affinities (1665 nm for TBA-SP7 and 1733 nm for TBA;F igure S2).

Computational analysis
The effect of the introductiono fd ifferent monomersw as further investigated by using the stochastic dynamics (SD) simulations. The model for TBA ( Figure 2A)s hows guanine and thymine base stacking at positions 7a nd 8i nt he TGT loop, whereas the nucleobase of the thymidine residue at position 9 is not involved in the stacking and is directed upwardsa nd away from the G-quadruplex core. T3 and T12 adopt ap incerlike structure. It hasb een reported that the TT loops (T3 and T12) constitute the thrombin-binding motifs of TBA [11] and that these pincer-like loops bind the protruding region of thrombin exosite I. Therefore, we focused on the structural changes in the pincer-like orientation of T3 and T12, as well as on T9 of the TGT loop. On the other hand, TBA-SP7 ( Figure 2B)s howed strong bindingi nteractions at thrombin exosite Ia sc ompared to the TBA;t his is mainly due to the fact that there are more interactions with TBA-SP7,e specially at G14, in addition to G2, T3 and T4, whicha re common to both TBA and TBA-SP7 analogues. Noticeably,t here is au nique p-p interaction between the residue Tyr118 and T3 of TBA-SP7. These interactions strongly stabilise the TBA-SP7 conformation compared to the other TBAa naloguesd esigned in this study.F or ab etter overview,amolecular model was used to examinec onformational changes of TBA upon incorporation of spacer-C 3 ,a mino-UNA and UNA at positions T3, T7 andT 12. Conformational changes of various tested aptamersb inding to TBA are provided in the Supporting Information ( Figure S2).
Ta ble 3d isplays the conformational changes and the results for thrombin clotting time activity (CT 50 ;t he time needed to clot 50 %o ft he availablef ibrinogen). The introduction of amino-UNA in TBA alwaysc hanged the orientation of the T9 nucleobase from an upward direction to an orientation more towardst he side and away from the guanine tetrad. Additionally,t he pincer-like structure of the TT loops was disrupted for the amino-UNA introductiona tT 3a nd T12. Spacer-C 3 incorporation at T3 and T12 (TBA-SP3 and TBA-SP12) resulted in the loss of the pincer-like structure and the wild-type orientation of T9. However,i nt he case of TBA-SP7,o nly the organisation   Ar eduction of up to 50 %o fa nti-thrombina ctivity( CT 50 )i si ndicated by (-), and reduction up to 75 %i sm arked (-);asignificant increasei sd enoted by (+ ++ ++ +). of the TGT loop was affected. Similart ot he data obtainedf or the spacer-C 3 variants, UNA variants TBA-U3 and TBA-U12a lso showedthe loss of the T3,T9and T12 orientations.Interestingly,t he structure of TBA-U7 was very similar to the unmodified TBA, although not identical. All TBA derivatives showni nT able1were used for the prime molecular mechanics-generalised Born surface area (MM-GBSA)-based binding affinity calculations. CT 50 values were comparedw ith binding energieso btained from the MM-GBSA calculations. The summary of the free energies obtained for each complex is provided in Ta ble S1, which shows all the energy terms of Equation (1) together with the experimentally measureda ctivities, as well as the predicted binding free energies (DG bind ). TBA-SP7,w hich had high potency (CT 50 = 168 s) among the compounds tested, wasc orrectly predicted as a highly stable compound by the MM-GBSAm ethod (À224.96 kcal mol À1 ). The coulomb energy for all of the compounds,o btained by the MM-GBSA method,i sh ighly favourable comparedt ot he overall binding energy; however,s ignificantly high solvation energy( E solvation )c ompensates for the electrostatic contribution to the overall bindingp rocess. As ar esult, the van der Waals (vdW) energy dominates the overall binding affinity for all the compounds. It is important to emphasise that the covalent binding energy (E covalent )c ontribution to the overall binding affinity of each complex is either negligible or unfavourable, except for the TBA-SP7 complex. In the case of the TBA-SP7-thrombin complex, the covalent binding energy contribution is remarkably low (À165.4 kcal mol À1 ), comparedt or est of the compounds in this series, and points to this contribution as key to making TBA-SP7 the most thermostable compound among those tested with respectt o thrombin binding ( Figure S3). The covalentb inding energy was calculated from the energy differenceb etween the thrombin-bound and free conformationso fa ptamer and essentially refers to the contributions due to changes in intramolecular structure of aptamer upon bindingt ot hrombin.I nt he case of TBA-SP7,t he covalent binding energy was as much as À165.4 kcal mol À1 and stabilises the thrombin:aptamer association to larger extent.T he energyo ft he aptamer was relatively low when it was bound to thrombin compared to the conformation of aptamer existingi ns olvent( comparison of E covalent and DG covalent is provided in Figure S4 Ai nt he Supporting Information). We further analysed the data in order to understand the covalent energy contribution andf ound no correlation (R 2 = 0.03) when DG covalent was excluded from DG bind in the dataset. This effect was mainly due to the TBA-SP7 construct, which behaveda sa no utlier; however,w hen TBA-SP7 and TBA were excluded from the dataset, the correlation improved to R 2 = 0.82 (Figure S4 B). Thisa nalysis clearly indicated that the high stability of TBA-SP7 was mainly due to the DG covalent energy,w hich also influenced the overall binding process to thrombin.
In addition to prime MM-GBSA calculations, an additional extensivem olecular dynamics simulationw ith explicit solvation was performed for the highly active construct,T BA-SP7,i n order to comparet he binding mode with TBA (wild-type) with respectt ob inding free energies. Based on the 15 ns simula-tions, the relative binding energies of TBA and TBA-SP7 in thrombin were estimated. From Figure 3, it is cleart hat TBA-SP7 binds (DG bind = À57.19 AE 0.2) relatively more strongly than TBA (DG bind = À38.26 AE 0.3) in the thrombinb inding site. An individual energyc ontributions plot also reveals that nonpolar contribution dominates the binding, as TBA-SP7 showed significantly higher E vdW and E surf values comparedt oTBA.

Discussion
Data obtained from CD measurements clearly showedt hat the introduction of amino-UNA,U NA or spacer-C 3 modifications in place of the thymidine nucleotide at positions T3, T7 and T12 was compatible with quadruplex formation, resulting in as imilar structuralt opology to that of unmodified TBA. The appearance of an absorption minimuma t2 65 nm and am aximum at 295 nm has been reported as characteristic for an antiparallel single-stranded G-quadruplext opology and corresponds to Next, UV meltings tudies were initiated to analyse the thermal stability of the TBAd erivatives. Notably,U NA or spacer-C 3 substitution at T3,T 7o rT 12 increasedt he thermald enaturation temperature, whereas the temperature decreased in the case of amino-UNA modifications. This could possibly be explained by the influence on the G-quadruplexs tructure of the 2'-5' internucleotide linkage between the sugar and the phosphate group or the 3'-aminog roup. As amino-UNAi sanovel construct, there are no studies that can be used to explain this phenomenon directly. However, ar ecent study by Aher et al. investigated the impact of 2'-5'-phosphodiester linkages and loop lengths on the folding topology of TBA. [21] They showed that all variantsw ith a2 '-5'-linkage exhibited decreased melting temperatures. This effect was attributedt oa ni ncreased number of bonds between the 5'-O and 3'-O-phosphorus or/ and an extended backbone geometry enabled by the 2'-5' linkage.E ven though amino-UNAd iffers from the 3'-deoxy-2'-5'-linked non-genetici soDNA studied by Aher et al.,i ts eems plausible that the decreases in melting temperatures that we observed might be the result of asimilareffect.
UNA and spacer-C 3 are less rigid than normaln ucleotides and add more flexibility in the region in which they are introduced. The loop regions are known to make contributions to the stability of the G-quadruplex due to p-stacking between loop nucleobases and G-quartets.O ur data suggest that the increased flexibility of the loop regionsm ight have led to the improved ability of loop nucleobases to interact with the Gquartets,therefore resulting in enhancedt hermal stability. [14] For ab etter overview,t he impact of modifying the TT loops and the TGT loop on thrombin clottingt imes are discussed separately:

TT loops modification
Modification of the thymine residues at positions 3a nd 12 lead to ar eduction in the inhibitory effect, whichw as observed in the thrombinc lotting experiments for all variants (amino-UNA, spacer-C 3 and UNA). Krauss et al. showedt hat the two loopsa ct as ap incer-like system that binds the protruding region of thrombin exosite I. [10,11] Thus,t he lack of thrombini nhibition of TBA-SP3 and TBA-SP12i st ob ee xpected, as these oligonucleotides no longer possess the crucial thyminest hat facilitatet he pincerf ormation. This is supported by our modelling results, which suggested that the substitution of either T3 or T12 with spacer-C 3 led to profound structuralr earrangements,a ffecting the formation of ap incer-like structure (Figure S2 Da nd F). Although they interactw itht he guanine tetrad, these modifications are no longera vailable for interactions with thrombin. This could explain not only the lack of inhibitory efficacy for both variants TBA-SP3 and TBA-SP12 but also the increasei nt hermal stabilitya sp-stacking interactions of either T3 or T12 with the guanine tetrads that could increase the stability of the structural core. For TBA-A3 and TBA-A12, as well as for TBA-U3 and TBA-U12, our modelling data suggested as imilar change in structure ( Figure S2 A, C, Ga nd I). Even though the UNA and amino-UNA variantss till have the thymine residues at the modification site, it could be theorised that the flexibilityo ft hese nucleotides induces tabilisation of the G-quadruplex through ah igher stacking energy with the quartet. Consequently,t his could result in ah igher energy barrier to form the pincer-like structure, which could negatively affect TBA and collectivelya ccount for its reduced activity towards thrombin.
In regard to the modification with amino-UNA, additional explanations couldb ef ound. Recently reported crystal structures showedt hat residues in the TT loop interact with two hydrophobic clefts of the anion-binding exosite Io ft hrombin (ABE1). [10,11] The substitutions of these loop residues (T3 and T12) with an amino-UNA led to the introduction of ap olar amino group. Thus, the hydrophobic interactions betweent he loop regions and thrombin could be reduced, resultinginadecreased inhibitory effect. Furthermore, reduction in the inhibitory ability of TBA-A3,T BA-A7 and TBA-A12 could be the result of the observed lack of stability. Finally,r esults from the incorporationo fU NA into TBA at positions T3 and T12 were in agreement with the work of Pasternak etal.,w ho have previously reported that UNA incorporation at T7 (TBA-U7) increased thrombin inhibition in anti-thrombina ssays by using blood plasma samples, whereas UNA incorporation at positions T3 and T12 (TBA-U3 and TBA-U12)r educed the effect. [14] TGT loop modification The thrombin clottingt ime assay showed significantly prolongedc lottingt imes for the variant TBA-SP7, whereas TBA-A7 and TBA-U7 failed to show high thrombininhibition. Only marginal interactions of thrombin with the TGT loop of TBA were reported by Krauss et al. [10] Additionally,H ee tal. reported the relevance of the phosphate groups for the TBA-thrombin interaction. [22] In regardt ot hese reports, an explanation for the reduced inhibition could be found in the amino group of amino-UNA, as this group could have as teric influence on these interactions in TBA-A7. However,i tw as also suggested that the stability and rigidity of TBA is important for the interaction betweena ptamer and thrombine xosite I. [16] As the variants with amino-UNA exhibited the lowest melting temperatures of all tested variants, including TBA, this could also be an explanation for the reduced inhibitory effect of the amino-UNA variants.I nc ontrast, the spacer-C 3 variant is expected to exhibit the pincer-like structure while also exhibiting increased stability;c ollectively this could account fort he improved efficacy for TBA-SP7 in the thrombinc lotting time assay.I ti sw orth mentioning that, for ad irect comparison, the buffer condition must be the same in all above-mentioned assay;h owever,w e adopted the suggested buffers ystem reported by ap revious group, [14] which only allows for comparison between the unmodified TBA and the modified TBAs discussed here.
For TBA-U7,o ur resultsd eviate from the published data by Pasternake tal. [14] They reported that UNA incorporation at T7 (TBA-U7) increased the thrombin inhibition in anti-thrombin assays using bloodp lasma samples.T he differenta ssays for measuring the effect of TBA derivatives on thrombina ctivity can be taken as ab asis for this discrepancy.T he thrombinc lotting time assay measured the effecto fT BA on thrombini n PBS, whereas the anti-thrombin assay measured the activity in human plasma in whicho ther plasma componentsc ould interact with thrombinand decrease the concentration availablef or thrombini nteraction. Likewise, interactions with these other plasma components could destabilise the structure of TBA. Thus more stable forms, like TBA-U7,c ould exhibit superior inhibitory activity in plasma, while being inferior in ab uffer system like PBS. Another explanation for this discrepancy can be found in the different TBA concentrations used (50 nm of TBA-U7 used in this study compared to 330 nm used by Pasternak et al.). Earlier studies have already shown that the concentration strongly affects the inhibitory capability of the TBA aptamer. [23,24] These discrepancies highlight the importance of testing modified aptamersi nv ariouss ettings in order to grasp possible drawbacks of prior in vivo andp reclinical investigations.
Althoughe lectrostatic interactions contributed favourably to the overall binding of TBA, very highly unfavourable solvation energy compensated for the electrostatic energy,a nd as ar esult, nonpolar interactionss uch as vdW and lipophilic energies dominate the overall binding energyo ft hese analogues (Figure S5). In general, amino-UNA analogues bound with moderate affinity to thrombin( CT 50 ranges from 45 to 55). From the binding affinity calculation, TBA-SP3 (DG = À46.5 kcal mol À1 ) and TBA-SP12 (DG = À57.7 kcal mol À1 )s hould bind poorly to thrombin as compared to TBA-SP7 (DG = À225.0 kcal mol À1 ), and this could be mainly due the loss of crucial interactions at the interface between the TBA-SP3 analoguea nd residues at the thrombin active site, for example, Tyr117, Asn78 and Ile79 with T3 and Thr74,A rg75 and Tyr76 with T12. Moreover,n onpolar interactions (E vdW = À58.1 kcal mol À1 )w ere significantly favourable for binding with thrombin (Table S1) as compared to rest of the analogues. It was clearly seen that TBA-U7 binds significantly better than TBA-U3 andT BA-U12. Again, this might be due to the effect of interaction with thrombinr esidues. When positions T3 and T12 werem odified with uUNA nucleotide, interactions between residues such as Thr74, Arg75,T yr76, Asn78,I le79 and Tyr117 with T3/T12w ere lost. This is also significantly reflected in the overall binding affinity of U3 (DG = À33.5 kcal mol À1 )a nd U12 (DG = À30 kcal mol À1 ) compared to U7 (DG = À65.7 kcal mol À1 ). We note from individual energy components contributions to the overall binding affinity that the value of E vdW wasq uite low for U3 (À6.4 kcal mol À1 )a nd U12 (À51.7 kcal mol À1 )c ompared to that of U7 (À75.0 kcal mol À1 ).
In addition, we also performed combined molecular dynamics (to account for sampling effects at particulart emperatures) and MM-GBSA calculations to compute the binding free energy for TBA and TBA-SP7 to thrombin. Both van de Waals and surface area energies in the binding energy clearly showedt hat nonpolarc ontribution is the main driving force to discriminate the TBA-SP7 from TBAb inding to thrombin (Figure 3). Energy decomposition analysisf rom free energy calculation examines the role of individual residue contributions to the overall binding affinity of the aptamers by decomposition of the binding free energy into aptamer-residue pairs. As seen from the decomposition analysisp lot ( Figure 3E), TBA-SP7 showed relativelyd ifferent interaction patterns than TBA with thrombinr esidues during the simulation. For instance, residues Asp63, Arg75, Arg77_A, Asn78,I le79 and Tyr117 showedafavourable energy contribution ( % 2.00 Kcal mol À1 difference) to the binding affinity of TBA-SP7 compared to TBA, especially Tyr117, which showeda na dditional energy contribution for TBA-SP7 (À2.25 kcal mol À1 )a nd whereas for TBA, this residue contributed very low( 0.004 kcal mol À1 ). Moreover,s ome residues from the aptamer also equallyc ontributed to the overall binding affinity of the complex as thrombinr esidues, for instance, T3,T 4, G5, T12 and T13 resides. These aptamer-thrombin structurali nteraction observations are also in good agreement with Prime MM-GBSA calculations. Analysis from the backboneR MSDa nd radius of gyration also suggested that the TBA-SP7-thrombin complex was relatively more stable than the TBA-thrombin complex ( Figure S6);i np articular,w hen it was bound to thrombin, the overall RMSD remained less than 2.0 ,w hereas the thrombin-TBA complex was quite flexible throughout the simulation. In addition, the radius of gyration (of the Ca atoms) of aptamers( TBA and TBA-SP7)f rom simulations was also computed and compared to the initial pose in ordert op rovide am easureo fo verall compactness of molecular shape of aptamers. [25][26][27] As seen from Figure S6 B, the overall molecular shape of TBA-SP7 looks very similart oT BA when the aptamer binds to the thrombin. Another recent report [28] showed the TBA structurals tability and flexibility by using the static modes (SM) calculation including polythymines, PEG spacers and alkyl chain modifications in the TBA. We concluded that the introductiono f spacers in the aptamer greatlya ffects not only the structural stabilityo fT BA but also its ability to positioni tself within thrombine xosite I. Our observations, based on the MM-GBSA calculations, also suggest that overall structurals tabilityo f aptamer-thrombin is not only governed by electrostatic energy (provides internal stability of aptamers)b ut also nonpolar interactions, especially van der Waals ands urface area energy complementarity,w hich is essential for aptamersb inding to the thrombine xosite I. This observation led us to believe that this is another demonstration of the malleability of G-quadruplex-forming aptamers, as previously observed for other aptamers. [28,29]

Conclusion
In summary,w eh ave systematically investigated the importance of the flexibility of TBA loops by incorporating three different chemical modifications:s pacer-C 3 ,a mino-UNA and UNA. All three modifications rendered high flexibility and less rigidity to TBA compared to natural nucleotide monomers, and this increasedf lexibility favoured stronger interactions of the nucleobase with the guaninet etrads, resultingi nq uadruplexf ormation with high thermalstability.Ina ddition, the flexibility possibly favoured structural rearrangements of TBA-types equences, especially towards the pincer-like structure of the TT loops and the orientation of T9. Unlike amino-UNA-and UNA-modified TBA sequences, one introduction of spacer-C 3 at positionT 7o f the TGT loop showed significant improvement in thermals tability and thrombinc lottingt ime comparedt ou nmodified TBA while maintaining the G-quadruplex structure. We believe that these findings will be very useful in designingn ext-generation thrombin-binding aptamers.

Experimental Section
Oligonucleotide synthesis: All oligonucleotides were synthesised by using standard phosphoramidite chemistry on an automated oligonucleotide synthesiser.A ll synthesised oligonucleotides were purified by ion exchange HPLC, and their composition was confirmed by MALDI-TOF mass spectrometry.S ynthesis of oligonucleotide sequences containing the 3'-amino-UNA-U nucleotide monomer has not been reported elsewhere. To wards the construction of these novel oligonucleotide sequences with an oligonucleotide synthesiser,the synthesis methodology for 3'-amino-UNA-U nucleoside phosphoramidite derivative (4,S cheme 3) is briefly described below.
Synthesis of amino-UNA-U phosphoramidite:C ompound 1 (Scheme 3) was treated with methanesulfonyl chloride in pyridine to activate the alcohol. The crude mesylate was converted into an azide by dissolving it in acetonitrile and reacting it with sodium azide and 15-crown-5 ether in am icrowave reactor.T he benzoyl protecting group was removed by hydrolysis with sodium hydroxide in methanol. The final reduction of the deprotected azide was achieved under Staudinger conditions by employing trimethylphosphine as the reducing agent in am ixture of tetrahydrofuran and water,o btaining the desired 3'-amino UNA-U phosphoramidite in 62 %y ield over four steps. The free amine was protected by using ethyl trifluoroacetate in MeOH with dimethylaminopyridine (DMAP) as an ucleophilic catalyst. The final amidite was furnished from 2-cyanoethyl N,N,N',N'-tetraisopropyldiamidophosphite and diisopropylammonium tetrazolide in dichloromethane (detailed procedures are provided in the Supporting Information).
Melting temperature studies: Melting temperatures were recorded on aB eckman DU 800 spectrophotometer equipped with as ixposition microcell holder and at hermostat. Oligonucleotides (12.5 mm final concentration) were dissolved in buffer containing potassium chloride (100 mm)a nd sodium cacodylate (10 mm), pH 7. The samples were renatured for 10 min at 95 8Ca nd then slowly cooled to room temperature. Absorbance versus temperature curves in 10 mm quartz microcuvettes were recorded at 295 nm. At emperature range of 15-85 8Cw as used at 0.5 8Cmin À1 . Three spectra were recorded and averaged for each sample. The spectrum for buffer only (no sample added) was subtracted from each sample spectrum. Melting curves were analysed by using nonlinear curve fitting with the program MeltWin 3.5. [30] Circular dichroism (CD) spectra: CD spectra were recorded on aJasco J-600A spectropolarimeter by using a0.7 mL quartz cuvette with a2mm path length. Oligonucleotides (2.5 mm final concentration) were dissolved in phosphate buffer (10 mm,p H6.9) containing KCl (5 mm). All samples were renatured for 10 min at 80 8Ca nd then slowly cooled to room temperature prior to measurement at 15 8Cw ith a2 20-320 nm wavelength range. Five spectra were recorded and averaged for each sample. The buffer spectrum was subtracted from each sample spectrum.
Thrombin clotting time assay: Thrombin clotting times were measured spectrophotometrically on aP erkinElmer Lambda 35 UV/Vis spectrometer.O ligonucleotides were incubated for 1min at 37 8C in PBS (1 mL) containing fibrinogen from human plasma (2 mg mL À1 ,F3889, Sigma-Aldrich). Then, human thrombin (100 mL, 10 NIH per mL;T 8885, Sigma-Aldrich) was added to the solution. The time required for fibrin polymerisation was determined from the UV scattering curve, which was registered as af unction of time (wavelength fixed at 380 nm;t otal time:6 00 s; time interval:0.1 s) for each sequence. Each measurement was performed in triplicate at 50 nm concentrations of each oligonucleotide. The clotting time value, reported as average AE standard deviation (AV. AE S.D.) was derived from the midpoint of each scattering curve. This corresponded to the time at which 50 %o ft he final absorbance was observed.

Computational modelling
Preparation of aptamers and protein:T he molecular structure of the human thrombin-aptamer (TBA) complex was obtained from the Protein Data Bank (PDB ID:4 DII, 2.06 resolution). [10] The TBA was imported into the Maestro module available in the Schrçdinger Suite (Schrçdinger,L LC), and subsequently,t he atomic coordinates of the aptamer were separated from the thrombin coordinates, and both were optimised separately.T he thrombin structure was optimised by using the Protein Preparation Wizard. [31] This protein structure optimisation includes adding hydrogen atoms, assigning bond orders and building disulfide bonds. The protonation states of the ionisable residues (pH 7) were predicted by the PROPKA tool provided in the Protein Preparation Wizard. An optimised structure model was finally found by energy minimisation (i.e.,p osition of the hydrogen atoms) with the OPLS2005 force field.
For the aptamer,t he protonation states were predicted by using the PROPKA tool in the presence of aK + ion. Finally,t he structure was energy-minimised (only hydrogen atoms) by using the OPLS2005 force field. The structures of TBA derivatives (amino-UNA, spacer-C 3 and UNA) used in these experiments were built manually by using the Maestro module available in the Schrçdinger Suite from an optimised aptamer obtained from the X-ray crystal structure. [10] TBA derivatives were energy-minimised to avoid any unfavourable contacts or steric clashes between the atoms. Subsequently,o ptimised structures were used for conformational analysis and MM-GBSA-based binding affinityp rediction calculations.
Conformational analysis: In order to reveal the structural stability of various TBA derivatives, we initiated short stochastic dynamic simulations on these modified TBAs by using the Macro Model (version 9.1), as it is implemented in the Schrçdinger Suite. All calculations were carried out by using the AMBER* force field [32] with aG B/SA solvation model. [33] In the process of thermalisation, initial velocities were generated from aM axwell-Boltzmann distribution at 300 K. The SHAKE algorithm was utilised to constrain the lengths of all bonds involving hydrogen atoms. Coordinates were saved every 1.5 fs from the 0.5 ns simulation for the analysis. For minimisation, the PRCG (Polak-Ribiere Conjugate Gradient) protocol was applied with ac onvergence threshold of 0.05 kJ mol À1 .R epresentative low-energy structures for each TBA derivative were selected for further inspection.
MM-GBSA calculations: In order to understand the binding affinity differences of various TBA complexes, MM-GBSA calculations were initiated by using the Schrçdinger Suite (Prime MM-GBSA). [34,35] MM-GBSA method implementation in the Schrçdinger Suite is slightly different from the corresponding MM-PBSA method described in our previous work. [29] Whereas the MM-PBSA method is based on molecular dynamics simulations, the Prime MM-GBSA method calculates the binding affinityo ft he aptamer (DG bind )b y energy minimisation procedures. The binding energy of the aptamer was extracted from an energy-optimised thrombin-aptamer complex and Prime MM-GBSA by using the VSGB 2.0 solvation (implicit) model. [33] The Prime MM-GBSA energy of the aptamer binding is estimated as [Eq. (1)]: This procedure is very efficient and is frequently used to rank protein-ligand complexes in the virtual screening process of lead identification/optimisation. During binding affinity prediction, the program offers the option to treat the aptamer and protein as flex-ible;i nt his study,a5.0 region of the protein around the ligand was treated as flexible.
Extensive molecular dynamic simulations (eMD): Molecular dynamics (MD) simulations were carried out for the two chosen thrombin-aptamer complexes, namely thrombin:TBA (wild-type) and thrombin:TBA-SP7. The starting structure for the thrombin:TBA used in the MD simulation is based on the structure reported in the Protein Data Bank (PDB ID:4 DII). [10] The starting structure for thrombin:TBA-SP7 was obtained by modifying residue 7( DT7) of the aptamer with as pacer group (-CH 2 -CH 2 -CH 2 -), and this structure was relaxed by using the AMBER force field. [36] Both complexes were solvated in water and neutralised with as ufficientn umber of counter ions (seven Na + ions). The K + ion stabilising the aptamer geometry was also retained in both of the thrombin-aptamer complexes. The simulation box was chosen as orthorhombic and contained more than 20 000 water molecules. The force field for describing the aptamer and thrombin was FF99SB, [37] as it contains parameters for both amino acids and nucleic acids, which are the fundamental units of proteins and DNA/aptamers, respectively.T he water solvent was described by using the TIP3P force field. The space group parameters were based on general AMBER force field (GAFF), and the charges for this nonstandard residue were prepared based on fitting to the molecular electrostatic potential obtained by employing the Merz-Singh-Kollman Scheme. [38] Structures were optimised with the HF/6-31G* level of theory,a si mplemented in Gaussian09. [39] This protocol is often employed to describe any nonstandard residue or organic molecule in force field MD simulations. The solvated complexes were first energy-minimised, then the simulations were carried out under constant temperature and pressure (room temperature, 1atm pressure). The temperature was controlled by using aL angevin thermostat; [40] the pressure was controlled by connecting the systems to aB erendsen'sb arostat. [41] The time scale for integration of equation of motion was 2f s. The initial timescale for the equilibration of the systems was 2ns, and the total production run was 15 ns. The calculations were carried out by using Amber14 software. [42] The time evolution of total energies and system densities served as indicators for the equilibration runs. Further,t he stability of both aptamer-thrombin complexes was verified by the potential energy landscape, root mean square displacement (RSMD) and radius of gyration. Free-energy calculations (from 400 configurations), the RMSD and the radius of gyration analysis were performed for the last 5nst rajectory of the production run. The MMPBSA.py tool [43] of the Amber software was employed for this purpose. Various contributions to the total binding free energy,n amely vdW,e lectrostatic, polar solvation and nonpolar solvation-free energies were obtained, and the results are presented in Table ST2. The entropy calculations were also performed by using normal mode analysis (for 50 configurations collected at equal intervals from the last 2ns run). The binding free energy analysis showed that the thrombin-TBA-SP7 made as tronger complex than the thrombin-TBA (wildtype) complex itself. The reason for this was analysed by using decomposition analysis, in which the individual residue contributions to total binding free energy can be computed.