A general strategy to enhance aptamer affinity by suppressing dissociation through symmetric assembly

Abstract

Aptamers are programmable molecular recognition elements with broad utility in diagnostics, therapeutics, and synthetic biology. However, many aptamers suffer from insufficient affinity due to rapid target dissociation, and no general strategy currently exists to overcome this limitation. Here, we report a symmetry-guided assembly approach that enhances aptamer affinity by suppressing the dissociation rate constant (k_off). Three identical aptamer units are spatially organized into a flexible trivalent assembly to enable kinetic cooperativity through rapid rebinding. Applied to aptamers targeting SARS-CoV-2 spike (both trimeric and monomeric S1 subunit), VEGF₁₆₅ (dimeric), and cardiac troponin I (monomeric), the resulting trimers exhibited dissociation constants (K_d) in the low pM range and k_off values in the 10⁻⁶ s⁻¹ range, over 100-fold improvements relative to monomers. In a serum-based VEGF₁₆₅ assay, the trimeric aptamer improved detection sensitivity by 30-fold. This modular, chemistry-based strategy is applicable to existing aptamers and establishes dissociation suppression as a general principle for engineering ultrahigh-affinity aptamers.

Graphical Abstract

Graphical Abstract.

Introduction

Natural evolution has produced a wide array of biomolecules with molecular recognition functions, including antibodies, receptors, enzymes, and riboswitches [1, 2]. These molecules bind their physiological targets with high specificity and affinity, enabling essential biological processes such as metabolism, immune response, and gene regulation [3, 4]. However, the molecular requirements of modern diagnostics, therapeutics, and biological interrogation often exceed the capabilities of naturally evolved systems. This is not a failure of biology, but a reflection of the fact that natural molecules are shaped to meet the needs of living organisms but not the diverse demands of human-designed applications. For example, antibodies are widely used as detection reagents in diagnostic assays, yet limited binding affinity can hinder the accurate identification of low-abundance targets in clinical samples [5–7]. These challenges highlight the need for synthetic molecular recognition elements that are created through artificial selection or in vitro evolution to complement and expand nature’s molecular toolkit.

Among synthetic recognition elements, nucleic acid aptamers have emerged as versatile tools for diagnostics, therapeutics, and molecular sensing [8–14]. These single-stranded DNA or RNA molecules can fold into defined three-dimensional structures to bind targets with high specificity. In addition to their binding capabilities, aptamers offer attractive features such as relatively small size, excellent chemical and thermal stability, low immunogenicity, and straightforward synthesis and modification [15–21]. They are typically identified using SELEX (Systematic Evolution of Ligands by Exponential Enrichment) [22–27], a well-established in vitro selection method that has successfully generated aptamers for a wide range of targets [28–33]. However, SELEX does not always yield aptamers with the affinity required for demanding applications, particularly when strong, sustained binding or ultra-low target detection is needed, highlighting the need for efficient post-SELEX strategies.

A key post-SELEX strategy for improving aptamer performance involves multivalent assembly, where multiple aptamer units are combined to enhance binding. Common approaches include linking aptamers with flexible oligonucleotide spacers [20, 34–36] or displaying them on nanomaterial scaffolds [37–41]. These strategies have demonstrated that multivalency can improve apparent affinity through increased local concentration and cooperative effects. Although substantial affinity enhancements have been reported in some cases [42, 43], many multivalent aptamers do not achieve the best possible affinities, due in part to limited reductions in the dissociation rate constant (k_off), a key determinant of overall binding strength. For example, multimeric versions of an RBD–PB6 aptamer reduced the K_d from 400 pM to 39 pM [34], and a nanoparticle-conjugated anti-IFN-γ aptamer showed a decrease from 59 nM to 5.9 nM [39], yet both exhibited <2.8-fold reductions in k_off. These advances highlight the promise of multivalent design while pointing to the opportunity for new chemical strategies that can more effectively improve aptamer–target interactions.

Suppressing the dissociation rate of aptamer binding is an effective strategy for engineering high-affinity aptamers, yet no general approach currently exists to achieve this. Here, we introduce a symmetric trimerization strategy that employs a chemical scaffold to arrange three identical aptamer units in a flexible trivalent assembly, enabling kinetic cooperativity through rapid rebinding. We hypothesize that this close spatial arrangement facilitates a “standby binding” mechanism, in which non-engaged aptamer arms rapidly rebind the target upon dissociation of the initially bound unit, thereby reducing the overall dissociation rate constant. To test this, we applied the strategy to three protein targets with distinct oligomeric states: the SARS-CoV-2 spike protein (both trimeric and monomeric S1 subunit), the dimeric VEGF₁₆₅, and the monomeric cardiac troponin I. The resulting trimerized aptamers exhibited over a 100-fold reduction in K_d, driven primarily by dramatic decreases in k_off, while the association rate constant (k_on) remained largely unchanged. To our knowledge, this represents the first post-SELEX assembly strategy to achieve substantial affinity enhancement primarily through k_off suppression, establishing symmetric homotrimeric aptamer assembly as a general and effective platform for high-affinity aptamer engineering, regardless of target oligomeric state or structural symmetry.

Materials and methods

Materials and reagents

All the DNA sequences are listed in Supplementary Table S1. MSA52t, lhTMSA52t, AS, H4r3, Tro4, 2G19, and 2G19-linear trimer were obtained from Integrated DNA Technologies (IDT) and purified by 10% denaturing (8 M urea) polyacrylamide gel electrophoresis (dPAGE) before use. Trebler phosphoramidite (Tris-2,2,2-[3-(4,4′-dimethoxytrityloxy)propyloxymethyl]ethyl-[(2-cyanoethyl)-(N,N-diisopropyl)]-phosphoramidite) was obtained from Glen Research (Virginia, United States). Trimeric aptamers and the Trebler with T15 linker (Trebler–T15) were synthesized using a Mermade 12 oligonucleotide synthesizer (BioAutomation) at McMaster University and purified by 10% dPAGE before use. The dimeric shDMSA52t was prepared by ligating MSA52t with Trebler–T15 using a ligation template (ratio of Trebler–T15:MSA52t:ligation template = 1:2:2) and T4 DNA ligase (5 U, room temperature, 1 h). The ligated shDMSA52t was purified by 10% dPAGE. The trimer and dimer synthesis can yield product mixtures containing 0, 1, 2, or 3 extended arms; dPAGE purification was performed prior to use. The successful preparation and product purity of shTMSA52t, shDMSA52t, and MSA52t used in tests were confirmed using 10% dPAGE (Supplementary Fig. S1). The wild-type SARS-CoV-2 spike protein subunit S1 (mS-protein, Cat. No. 40591-V08B1) and Troponin I (Cat. No. 501-TNNI0010) were purchased from Sino Biological Inc. VEGF₁₆₅ protein (his-tagged, Cat. No. VE5-H5248) was obtained from Acro Biosystems. Sodium borohydride (NaBH₄, 98%), potassium chloride (KCl, ≥99%), sodium chloride (NaCl, ≥99.5%), 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES, ≥99%), magnesium chloride (MgCl₂, ≥99%), tetrachloroauric(III) acid trihydrate (HAuCl₄·3H₂O, ≥99.7%), sodium citrate (≥99.7%), acetic acid (HOAc, ≥99.7%), sodium acetate (NaOAc, ≥99%), tris(2-carboxyethyl)phosphine hydrochloride (≥98%), sulfuric acid (H₂SO₄, 95%–98%), sodium dodecyl sulfate (≥98%), and Tween-20 were obtained from Sigma–Aldrich. Nitrocellulose membranes (Cat. No. 10600125) were obtained from GE Healthcare. Nylon membranes (Cat. No. NEF994001PK) were obtained from PerkinElmer Inc. T4 polynucleotide kinase (PNK) and T4 DNA ligase with 10× buffer were acquired from Thermo Scientific (Ottawa, Canada). [γ-³²P]-ATP was purchased from PerkinElmer. Ninety-six-well black flat-bottom plates (Cat. No. 655076) were purchased from Greiner Bio-One. Ultrapure water (Milli-Q System, Millipore) was used to prepare all aqueous solutions.

Preparation of radioactive DNA

DNA aptamers for dot blot assays were labeled with ³²P at the 5′ end using PNK according to a previously reported protocol with slight modifications [44]. Briefly, the DNA aptamer (2 μl, 1 μM) was mixed with [γ-³²P]-ATP (1 μl), 10× PNK reaction buffer A (1 μl), PNK (1 μl, 10 U/ml), and water (5 μl) in a 200-μl polymerase chain reaction tube, followed by incubation at 37°C for 20 min. The reaction mixture was then purified by 10% dPAGE containing 8 M urea.

Dot-blot binding assays

The binding affinities of ³²P-labeled MSA52t, shDMSA52t, and shTMSA52t for the mS-protein were measured using dot-blot assays. Nitrocellulose and nylon membranes were pre-soaked in binding buffer (50 mM HEPES, 150 mM NaCl, 6 mM KCl, 2.5 mM MgCl₂, 2.5 mM CaCl₂, and 0.01% Tween-20, pH 7.4) for 1 h. Radioactive MSA52t, shDMSA52t, or shTMSA52t (10 μl, 10 pM) in binding buffer was denatured at 90°C for 5 min and annealed at 22°C for 20 min. Different concentrations of the mS-protein (10 μl) were then mixed with the aptamers and incubated at 22°C for 30 min. The reaction mixture was sequentially filtered through a nitrocellulose membrane, a nylon membrane, and a pre-wetted Whatman paper filter using a Whatman Minifold-1 96-well apparatus under vacuum. Bound aptamers on the nitrocellulose membrane and unbound aptamers on the nylon membrane were visualized after 12 h of phosphor screen exposure using a Typhoon 9200 imager (GE Healthcare). Dot intensity was quantified using ImageJ software to determine the bound fraction of the aptamer, which was plotted against mS-protein concentration. K_d values were obtained via non-linear curve fitting using Origin 2020 software by the equation Y = B_maxX/(K_d + X), where Y represents the bound fraction of the aptamer, B_max denotes the maximum bound fraction, and X corresponds to the mS-protein concentration. To evaluate the blocking effect of antisense sequence (AS), mixtures of ³²P-labeled MSA52t, shDMSA52t, or shTMSA52t (10 pM) with unlabeled AS (at varying concentrations) were prepared and analyzed using the same method.

Preparation of multimeric aptamer nMSA52t

The multimeric aptamer nMSA52t was prepared by conjugating MSA52t to gold nanoparticles (AuNPs) using a previously reported salt-aging protocol [39]. AuNPs of 49 nm were synthesized using a seeded growth method [45]. Briefly, AuNP seeds were prepared by heating 150 ml of 2.2 mM sodium citrate solution to reflux in an oil bath for 15 min under vigorous stirring in a three-necked round-bottom flask equipped with a condenser to prevent solvent evaporation. Once boiling, 1 ml of 25 mM HAuCl₄ was injected. After seed formation, the reaction was cooled to 90°C, and 1 ml of 60 mM sodium citrate was added, followed by the addition of 1 ml of 25 mM HAuCl₄. The AuNP concentration was estimated at 1.06 nM [39], based on the initial HAuCl₄ quantity and the assumption of complete precursor conversion. The resulting AuNPs were stored in the reaction solution until conjugation with DNA aptamers.

MSA52t aptamers were next conjugated to the AuNPs to prepare multimeric nMSA52t using a modified salt-aging protocol [39]. Firstly, 38 μl of 100 μM 5′-thiol-modified MSA52t and 2 μl of 100 μM T20 linker (with a 5′-thiol and 3′-biotin) were incubated with 10 μl of 0.5 mM TCEP for 2 h at room temperature to reduce disulfide bonds. Meanwhile, 1 ml of AuNP solution was centrifuged at 2700 RCF for 10 min, the supernatant was removed, and the AuNP pellet was resuspended in 100 μl of water. The reduced aptamer mixture was then added to the AuNP suspension and incubated overnight. To adjust phosphate concentration, 100 mM sodium phosphate buffer (pH 7.0) was added to achieve a final 10 mM phosphate concentration, followed by sodium dodecyl sulfate to reach 0.1% (wt/vol). The solution was shaken for 30 min before gradually adding a salting buffer (10 mM sodium phosphate, 2 M NaCl, pH 7.0) in six increments over 2 days. After the final salt addition, the conjugates were left to equilibrate overnight. The conjugates were then washed three times by centrifugation at 9300 RCF for 10 min with 1 ml of washing buffer (10 mM sodium phosphate, 150 mM NaCl, pH 7.4) and resuspended in 1 ml of binding buffer. The average number of aptamers per particle prepared using the protocol was determined at ∼250 [39].

Biolayer interferometry binding assay

The binding kinetics of aptamers were analyzed using biolayer interferometry (BLI) with a Sartorius Octet RED96 instrument. First, Octet High Precision Streptavidin (SAX) Biosensors (Cat. No. 18-5117, Sartorius) were hydrated by immersing them in binding buffer (50 mM HEPES, 150 mM NaCl, 6 mM KCl, 2.5 mM MgCl₂, 2.5 mM CaCl₂, and 0.01% Tween-20, pH 7.4) for at least 10 min. Biotin-modified aptamers (200 μl, 40 nM) were annealed by heating at 95°C for 3 min, followed by cooling to room temperature for 20 min. Protein solutions at various concentrations were prepared in the same buffer. All solutions (200 μl) were loaded into a 96-well black flat-bottom plate (Greiner Bio-One, Cat. No. 655076).

To initiate the experiment, biosensors were first immersed in the buffer for 120 s to establish a baseline. They were then immersed in the aptamer solution for 200 s or until a 0.4 nm response was reached. Afterward, the biosensors were returned to the buffer for 240 s to establish a stable baseline and then were immersed in protein solutions of varying concentrations for 1500 s for determining association kinetics. Finally, the sensors were immersed in buffer to measure dissociation kinetics for 1500 s. Control experiments were conducted by immersing aptamer-modified biosensors in buffer without proteins. Control data from reference sensors were used to correct for biosensor drift and other artifacts. Data processing and analysis were performed using the Sartorius Analysis Kit 13. Binding kinetic parameters were obtained by globally fitting the data using a 1:1 binding model available in the analysis software. To prevent multimer formation on the sensor surface, aptamer loading was limited to an unsaturated state, stopping at a 0.4 nm response. According to the Octet BLI manual, each sensor contains ~10⁹ streptavidin-conjugation sites and has a diameter of 600 μm, ensuring that the spacing between coupled aptamers remained greater than 9 nm.

Enzyme-linked aptamer assays for detection of VEGF165 in serum

Microtiter plates (clear, polystyrene, Product No. P7491 from Sigma–Aldrich) were first coated with 50 μl per well of 3 μg/ml monoclonal anti-VEGF₁₆₅ antibodies (clone 26503, Product No. MA5-23719 from Thermo Fisher Scientific). The plates were then blocked with 300 μl of 1% bovine serum albumin (BSA) in PBS for 2 h at room temperature. After blocking, the wells were washed three times with 100 μl of PBS. Next, 100 μl of VEGF₁₆₅ at varying concentrations in binding buffer (50 mM HEPES, 150 mM NaCl, 6 mM KCl, 2.5 mM MgCl₂, 2.5 mM CaCl₂, 0.01% Tween-20, pH 7.4) was added to each well and incubated for 1 h at room temperature, followed by washing with 200 μl of binding buffer. Then, 100 μl of 10 nM biotinylated shTH4r3 or H4r3 aptamer in binding buffer was added and incubated for 30 min to form a sandwich complex with VEGF₁₆₅ and the antibody. After washing with binding buffer, 100 μl of streptavidin–HRP conjugate (Product No. 18-152 from Sigma–Aldrich) was added to each well and incubated for 30 min. Wells were then washed six times with binding buffer, followed by the addition of 100 μl per well of TMB substrate solution and a 15-min incubation. The reaction was stopped by adding 100 μl of 0.1 M HCl to each well, and absorbance at 450 nm (OD450) was measured using a multimode plate reader (TECAN INFINITE 2000). Each assay condition was performed three times (n = 3), and average absorbance values are presented with error bars representing one standard deviation.

Results and discussion

Aptamer assembly and affinity assessment

To begin testing the generalizability of symmetric homotrimeric aptamer design, we first selected the monomeric S1 subunit of the SARS-CoV-2 spike protein (simplified herein as “mS-protein”), a truncated version of the full-length spike protein that does not trimerize, along with MSA52, an aptamer originally selected for binding to the wild-type mS-protein [44]. This system provided an ideal model to evaluate whether symmetric trimeric assembly could enhance binding affinity even in the absence of a multivalent or symmetric target.

For ease of synthesis and modular assembly, we used a truncated version of MSA52, designated MSA52t, which retains comparable affinity to the full-length aptamer (Supplementary Table S1). To construct multivalent aptamers, each MSA52t unit was extended with a flexible 15-thymidine (T15) spacer and assembled onto a trident linker. This linker was synthesized using a trebler phosphoramidite [Tris-2,2,2-[3-(4,4′-dimethoxytrityloxy)propyloxymethyl]ethyl-[(2-cyanoethyl)-(N,N-diisopropyl)]-phosphoramidite] [46], allowing site-specific conjugation of two or three aptamer units in flexible orientation. The T15 spacer ensured sufficient flexibility and spatial separation (up to ∼12 nm) to allow each aptamer unit to independently access its target site [47], accommodating a wide range of protein sizes (Fig. 1A) [48]. This strategy enabled the construction of a symmetric homodimer (named shDMSA52t) and a symmetric homotrimer (shTMSA52t), where “sh” denotes “symmetric homo.” The dimeric construct was synthesized by leaving one arm of the trident linker unextended.

Figure 1.

Symmetric aptamer trimerization enhances binding affinity. (A) Schematic illustration of aptamer assembly. The 3′ end of monomeric aptamer MSA52t was linked to the trident linker (trebler phosphoramidite) with a 15-thymidine spacer (T15) and assembled into dimeric (shDMSA52t) or trimeric (shTMSA52t) constructs. This design provides flexible orientation and optimal target accessibility for each aptamer unit. (B) Dot-blot binding assays showing the affinities of MSA52t, shDMSA52t, and shTMSA52t toward the monomeric wild-type S-protein (mS-protein). Apparent dissociation constants (K_d) were determined by nonlinear curve fitting. Trimeric assembly resulted in a ∼344-fold affinity enhancement over the monomer.

We evaluated the binding affinities of MSA52t, shDMSA52t, and shTMSA52t toward the wild-type mS-protein using a dot-blot assay. 5′-³²P-labeled aptamers were incubated with increasing concentrations of the mS-protein, and the resulting complexes were captured on nitrocellulose membranes, while unbound aptamers were retained on nylon membranes. The bound and unbound fractions were quantified by phosphor imaging (Supplementary Fig. S2), and K_d values were calculated by nonlinear fitting (Fig. 1B).

The monomeric aptamer MSA52t exhibited a K_d of 4.3 nM, consistent with the affinity of the full-length sequence [44]. Dimeric shDMSA52t showed a significantly improved K_d of 295.6 pM, corresponding to a 14.5-fold enhancement over the monomer. Most notably, trimeric shTMSA52t achieved a K_d of 12.5 pM, representing a 23.6-fold improvement over the dimer and a 344-fold enhancement over the monomer. These results demonstrate that symmetric homotrimeric assembly can yield dramatic affinity gains even for monomeric targets, showing its promise as a general post-SELEX strategy for improving aptamer performance.

Assessment of binding cooperativity of assembled aptamers

After confirming the affinity enhancement, we next sought to determine whether each aptamer arm within the assembled constructs functionally contributes to target recognition. We hypothesized that the observed affinity enhancement results from cooperative interactions among the individual aptamer units.

To test this, we used a complementary antisense (AS) DNA strand to block individual MSA52t arms and assess the impact on binding. A 49-nucleotide AS sequence (Supplementary Table S1), fully complementary to MSA52t, was incubated with ³²P-labeled aptamers at defined molar ratios prior to the addition of the mS-protein. Binding was then evaluated by dot-blot assay, and K_d values were determined by nonlinear fitting (Fig. 2 and Supplementary Fig. S3).

Figure 2.

Probing the functional contribution of individual aptamer units using antisense blockade. Binding affinities of (A) monomeric MSA52t, (B) dimeric shDMSA52t, and (C) trimeric shTMSA52t to the mS-protein were assessed in the absence and presence of antisense (AS) DNA strands complementary to MSA52t. AS strands were added at 1:1, 2:1, or 3:1 molar ratios relative to the aptamer to block the aptamer arms. Binding was evaluated by dot-blot assays, and apparent K_d values were determined by nonlinear fitting. Inhibition patterns indicate that each aptamer unit contributes cooperatively to the overall binding affinity of the multivalent constructs.

We first validated AS blocking using monomeric MSA52t (Fig. 2A). At a 1:1 AS:MSA52t ratio, binding to the mS-protein was completely abolished, confirming that a single AS strand effectively blocks the aptamer’s binding function. We then evaluated the effect of AS-mediated blocking on shDMSA52t (Fig. 2B). By blocking the arms of shDMSA52t with AS, we can assess whether all the aptamer arms are required for the best possible binding. When a 1:1 AS:aptamer ratio was used, K_d of shDMSA52t increased from 295 pM to 4565 pM. At a 2:1 ratio, no binding was observed. Compared with the monomeric MSA52t, for which binding was completely abolished at 1:1 AS:aptamer ratio, the dimeric shDMSA52t required a 2:1 AS:aptamer ratio to abolish binding. These results indicate that both MSA52t units in shDMSA52t contribute cooperatively to the higher affinity of the dimer.

Similarly, for shTMSA52t (Fig. 2C), progressive inhibition was observed with increasing AS concentrations. Binding affinity decreased from 12.5 pM (no AS) to 298 pM with the 1:1 AS:aptamer ratio and to 4251 pM at the 2:1 ratio. At 3:1, complete inhibition was achieved, indicating that the three aptamer arms of shTMSA52t contribute to target protein binding in a concerted fashion.

Together, these results confirm that each aptamer arm in shDMSA52t and shTMSA52t actively engages the target and contributes to the overall binding affinity. The substantial affinity losses upon partial antisense blocking indicate cooperative interactions among the aptamer units, supporting the idea that multivalent engagement underlies the performance of these assembled constructs.

Analysis of binding kinetics

To understand the mechanistic basis for the enhanced binding affinities observed in the assembled aptamers, we next investigated their binding kinetics. While K_d is widely used to characterize aptamer-target interactions, kinetic parameters, specifically k_on and k_off, can provide deeper insight into affinity enhancement mechanisms [49, 50]. The K_d values can be calculated from these kinetic parameters according to the equation K_d = k_off/k_on.

The dot blot assays were used for initial screening and relative comparison, whereas quantitative affinity analysis was primarily based on BLI measurements. We evaluated the binding kinetics of monomeric MSA52t, dimeric shDMSA52t, and trimeric shTMSA52t using BLI. Aptamers were 3′-biotinylated and immobilized on streptavidin-coated biosensors at low density to avoid multimer formation on the sensor surface (see Supporting information for details). BLI detects changes in the interference pattern of white light caused by the interaction between light reflected from the sensor surface and an internal reference surface [51]. Sensorgrams were recorded upon exposure to the mS-protein at concentrations spanning 0.1–10× their respective K_d values, and kinetic parameters were determined using global fitting (Fig. 3A–C).

Figure 3.

Trimeric assembly enhances affinity primarily by reducing dissociation rate. Binding kinetics of (A) MSA52t, (B) shDMSA52t, and (C) shTMSA52t interacting with the wild-type SARS-CoV-2 S1-protein (mS-protein) were measured by BLI. Concentration-dependent sensorgrams were globally fitted using 1:1 binding model to obtain the kinetic parameters k_on, k_off, and K_d. The experimental curves are shown as solid lines and the fitted curves are shown as black dashed lines. Bar graphs comparing (D) k_on, (E) k_off, and (F) K_d values for each aptamer. While only modest changes in k_on were observed, trimerization led to a dramatic reduction in k_off, resulting in a 348-fold improvement in K_d for shTMSA52t compared to the monomer.

All three aptamers exhibited similar association rates, with k_on values of 1.4 × 10⁵, 1.8 × 10⁵, and 2.0 × 10⁵ M⁻¹ s⁻¹ for MSA52t, shDMSA52t, and shTMSA52t, respectively. In contrast, their dissociation rate constants differed dramatically. The monomeric aptamer dissociated relatively quickly (k_off = 6.1 × 10^-4 s⁻¹), whereas the dimer and trimer dissociated much more slowly, with k_off values of 5.3 × 10⁻⁵ s⁻¹ and 2.5 × 10^-6 s⁻¹, respectively (Fig. 3E).

As summarized in Fig. 3D–F, shDMSA52t showed only a modest 1.2-fold increase in k_on but an 11.5-fold decrease in k_off, resulting in a 14.8-fold improvement in K_d (4357 pM versus 294.4 pM). Notably, shTMSA52t exhibited a 1.4-fold increase in k_on but a striking 244-fold decrease in k_off, leading to a 348-fold affinity enhancement (4357 pM versus 12.5 pM). The kinetically derived K_d values were consistent with those obtained by dot-blot analysis (Fig. 1B), confirming that affinity gains primarily arise from reduced dissociation.

We further evaluated shTMSA52t binding to the full-length trimeric wild-type SARS-CoV-2 spike protein (Supplementary Fig. S4). This pairing resulted in a slightly lower dissociation rate constant (k_off = 2.2 × 10^-6 s⁻¹) but a notably higher association rate (k_on = 4.6 × 10⁵ M⁻¹ s⁻¹) in comparison to the mS-protein, yielding an outstanding dissociation constant K_d of 4.7 pM. This result reflects the high-affinity, symmetry-matched 3-on-3 interaction between the trimeric aptamer and its structurally compatible protein target [47]. In contrast to symmetric matching for the trimeric spike protein, k_off suppression for the monomeric S protein is attributed to binding cooperativity of aptamer units assembled by the symmetric trebler. We also tested shTMSA52t for binding to BSA and VEGF₁₆₅ as control proteins, and used a scrambled trimer control (Trimer Ctrl) and a trebler containing only T15 linkers (Trebler-T15) for binding to mS-protein. No binding was observed in the BLI assays (Supplementary Fig. S5), demonstrating the high specificity of shTMSA52t for binding its intended target.

In the BLI assays, aptamer immobilization was limited to an unsaturated loading level to minimize potential mass-transport effects. To validate the robustness of the measurements, additional BLI experiments were performed using a lower aptamer loading concentration (5 nM instead of 40 nM) (Supplementary Fig. S6). The resulting kinetic parameters were consistent with those obtained under higher loading conditions, confirming the reliability of the BLI measurements. We also tested shTMSA52t with a shortened T5 linker, and a decreased binding affinity was observed (Supplementary Fig. S7). The results indicate that sufficient flexibility of the trebler linker is a prerequisite for achieving high-affinity enhancement. With a shortened linker, the three aptamer units in shTMSA52t may have limited access to the binding domain of the spike protein.

These findings highlight symmetric trimerized aptamer assembly as a powerful post-SELEX strategy for enhancing binding affinity, driven primarily by a dramatic suppression of the dissociation rate. We propose that the close spatial proximity and flexible orientation of identical aptamer arms within the symmetric assembly enable rapid rebinding of dissociated units. This dynamic rebinding mechanism effectively prevents complete dissociation, thereby strongly stabilizing the aptamer–protein complex even in the absence of multiple binding epitopes on the target, as represented by the case of the mS-protein.

Comparison with alternative assembly strategies

To evaluate the performance of our symmetric trimerization approach relative to conventional aptamer assembly methods, we compared it with two commonly used multivalent designs: linear linker-based assembly and nanoparticle conjugation.

For the linear strategy, we constructed a trimeric aptamer, lhTMSA52t (linear homo TMSA52t), by linking three MSA52t units with two poly(T15) spacers (Fig. 4A, sequence in Supplementary Table S1). The construct was biotinylated at the 5′ end and immobilized on BLI sensors for kinetic analysis. lhTMSA52t exhibited a k_on of 3.6 × 10⁵ M⁻¹ s⁻¹ and a k_off of 6.9 × 10⁻⁵ s⁻¹, yielding a K_d of 191.7 pM. Compared to monomeric MSA52t, this represents a 2.5-fold increase in association rate and an 8.8-fold reduction in dissociation rate, resulting in a 22.7-fold affinity enhancement (Fig. 4C–E, MSA52t versus lhTMSA52t).

Figure 4.

Binding kinetics of aptamer assemblies constructed via non-symmetric multivalent strategies. BLI sensorgrams showing the binding kinetics of the mS-protein to (A) lhTMSA52t, a linear trimeric aptamer assembled by connecting three MSA52t units with poly(T15) spacers, and (B) nMSA52t, a multivalent construct formed by conjugating 5′-thiol-modified MSA52t onto gold nanoparticles. Concentration-dependent sensorgrams were globally fitted using 1:1 binding model to obtain the kinetic parameters k_on, k_off, and K_d. The experimental curves are shown as solid lines and the fitted curves are shown as black dashed lines. Bar graphs comparing (C) k_on, (D) k_off, and (E) K_d values for lhTMSA52t and nMSA52t with the monomeric MSA52t. (F) k_off comparison across different aptamer assembly strategies. The linear trimer and nanoparticle-conjugated multimer strategies showed less significant affinity improvements, primarily due to ineffective suppression of dissociation rates.

When directly compared to the symmetric trimerized aptamer shTMSA52t, lhTMSA52t had a similar k_on but a dissociation rate 27.6 times faster. As a result, shTMSA52t achieved a K_d 15.3-fold lower than lhTMSA52t (12.5 pM versus 191.7 pM). The results demonstrate the superior ability of the symmetric trimer to reduce k_off and achieve tighter binding. The reduced performance of lhTMSA52t may be due to restricted target accessibility, especially for inner aptamer units tethered in a linear fashion. In contrast, the trident linker used in shTMSA52t ensures equal spacing and flexible orientation, maximizing the availability of each aptamer arm for target engagement. To confirm the effectiveness of the symmetric trimerization strategy, we evaluated another aptamer named 2G19, which binds VEGF₁₆₅ [43]. Using this aptamer, we constructed a conventional linear-linker trimer (2G19–linear trimer) and a symmetric trimer (2G19–symmetric trimer), and measured their binding to VEGF₁₆₅ by BLI (Supplementary Fig. S8). The 2G19–linear trimer showed an increased k_on (3.5 × 10⁵ M⁻¹ s⁻¹) but a similar k_off to the monomer, resulting in a 2.5-fold affinity improvement (K_d = 7.7 nM for the trimer versus 19.3 nM for the monomer). In contrast, the 2G19–symmetric trimer exhibited a markedly reduced k_off (3.0 × 10^-6 s⁻¹) and a similar k_on, leading to a 572-fold affinity enhancement over the monomeric aptamer (K_d = 33.7 pM for the trimer versus 19.3 nM for the monomer). These results confirm the effectiveness of the symmetric trimerization strategy in enhancing aptamer affinity.

We next evaluated a nanoparticle-based multivalent system by conjugating 5′-thiol-modified MSA52t to gold nanoparticles, forming nMSA52t (Fig. 4B). This construct was immobilized on BLI sensors using a 20-mer poly(T) spacer with a 5′ thiol (for gold nanoparticle binding) and 3′ biotin group (for sensor immobilization). BLI measurements showed a k_on of 1.3 × 10⁶ M⁻¹ s⁻¹, a 9.2-fold increase over the monomer (Fig. 4C). However, its k_off remained high at 5.2 × 10^-4 s⁻¹, similar to MSA52t (Fig. 4D), resulting in only a 10.9-fold improvement in affinity (Fig. 4E). These results are consistent with previous reports, such as gold nanoparticle-conjugated aptamers targeting interferon-gamma (IFN-γ) [39], where K_d improvements were also modest despite efforts to optimize aptamer density.

Together, these findings suggest that while linear and nanoparticle-based assemblies can enhance the association rate by increasing local aptamer concentration, they do not substantially reduce the dissociation rate and thus fall short in achieving strong affinity enhancement. By contrast, our symmetric trimerization approach uniquely enhances binding affinity by substantially suppressing dissociation (244-fold, Fig. 3E), an effect not observed with conventional multivalent strategies. Among the tested assemblies, the symmetric trimerization strategy achieved the strongest suppression of k_off (Fig. 4F). The linear trimer and nanoparticle-conjugated multimer strategies showed less significant affinity improvements primarily due to ineffective suppression of the dissociation rate. These findings indicate the critical role of linker architecture and spatial arrangement in enabling cooperative, kinetically stabilized target binding. The linear trimerization and nanoparticle conjugation strategies suffer from limitations of steric hindrance or uncontrolled orientation of binding units. In contrast, multivalent assemblies based on controlled geometry and well-defined architectures, e.g. DNA junctions or origamis [52, 53], can enhance affinity by enforcing specific spatial organization of binding arms.

Expanding the assembly strategy to other aptamers and targets

To further evaluate the generalizability of symmetric aptamer trimerization, we extended our strategy to two additional aptamer–target pairs: H4r3, which binds human vascular endothelial growth factor 165 (VEGF₁₆₅) [54], and Tro4, which targets cardiac troponin I [55]. Although both VEGF₁₆₅ and troponin I are important clinical biomarkers, we selected them for this study primarily because VEGF₁₆₅ is a homodimeric protein and troponin I is a monomeric protein. These targets complement the previously examined monomeric S1 and trimeric spike proteins, allowing us to test the effectiveness of trimeric aptamer assembly across a diverse range of target oligomeric states.

We assembled H4r3 and Tro4 into symmetric homotrimers, designated shTH4r3 and shTTro4, and analyzed their binding kinetics using BLI (Fig. 5). For VEGF₁₆₅, monomeric H4r3 exhibited a k_on of 1.8 × 10⁴ M⁻¹ s⁻¹ and a k_off of 1.5 × 10^-4 s⁻¹, yielding a K_d of 8.3 nM (Fig. 5A). Upon trimerization, shTH4r3 showed a moderate increase in k_on (k_on of 3.4 × 10⁴ M⁻¹ s⁻¹, Fig. 5B; 1.8-fold increase over the monomer, Supplementary Fig. S9) and a dramatically reduced dissociation rate constant (k_off = 2.7 × 10⁻⁶ s⁻¹, Fig. 5B), resulting in a K_d of 79 pM (Fig. 5B). This represents a 105-fold improvement in affinity (Fig. 5F), driven primarily by a 55-fold decrease in k_off (Fig. 5E).

Figure 5.

Symmetric trimeric assembly enhances aptamer affinity for VEGF₁₆₅ and Troponin I. BLI sensorgrams showing the binding kinetics of (A) monomeric H4r3 and (B) symmetric homotrimeric shTH4r3 for VEGF₁₆₅, and (C) monomeric Tro4 and (D) symmetric homotrimeric shTTro4 for troponin I. Concentration-dependent sensorgrams were globally fitted using 1:1 binding model to obtain the kinetic parameters k_on, k_off, and K_d. The experimental curves are shown as solid lines and the fitted curves are shown as black dashed lines. Bar graph comparison of (E) k_off and (F) K_d for each aptamer pair. Homotrimeric aptamer assembly dramatically suppressed k_off, resulting in >100-fold affinity enhancements for both targets.

A similar enhancement was observed for troponin I. Monomeric Tro4 exhibited a k_on of 4.0 × 10⁴ M⁻¹ s⁻¹ and a k_off of 7.2 × 10⁻⁵ s⁻¹, yielding a K_d of 1.8 nM (Fig. 5C). Upon trimerization, shTTro4 displayed a modest increase in k_on (6.9 × 10⁴ M⁻¹ s⁻¹, Fig. 5D; 1.7-fold increase over the monomer, Supplementary Fig. S9) and a much-reduced dissociation rate constant (k_off = 1.2 × 10⁻⁶ s⁻¹), resulting in a K_d of 17.3 pM (Fig. 5D). This corresponds to a 104-fold improvement in affinity (Fig. 5F), again primarily due to a 60-fold suppression of k_off (Fig. 5E).

Together, these results demonstrate that symmetric homotrimeric assembly significantly improves aptamer affinity across multiple aptamer–target systems. While only modest increases in k_on were observed, the enhancement in binding strength was consistently driven by substantial reductions in k_off, in line with our findings for the MSA52t system. These findings support the utility of this strategy as a broadly applicable post-SELEX platform for engineering high-affinity aptamers.

Comparison with reported aptamers

To contextualize the performance of our symmetric homotrimeric aptamers, we compared them with a broad panel of protein-binding aptamers reported in the literature. We compiled 40 aptamers for which the complete data on K_d, k_on, and k_off were available. These included monomeric, dimeric, and multimeric constructs targeting a variety of proteins. Kinetic and K_d data for these aptamers, along with those from our symmetric assemblies and two alternative trimeric designs (lhTMSA52t and nMSA52t), are summarized in Supplementary Table S2. Comparative visualizations are provided in Fig. 6 and Supplementary Fig. S10. The reported aptamers exhibited wide-ranging kinetic parameters: K_d values from 30 pM to 25 μM, k_on values from 2.5 × 10² to 3.8 × 10⁸ M⁻¹ s⁻¹, and k_off values from 3.45 s⁻¹ to 2.0 × 10⁻⁵ s⁻¹.

Figure 6.

Comparative analysis of K_d, k_on, and 1/k_off values for the top 12 aptamers compiled in Supplementary Table S2. Symmetric homotrimeric aptamers developed in this study (shTMSA52t, shTTro4, shTH4r3) are included. (A) Affinity ranking based on K_d values, with lower values indicating stronger binding. (B) Comparison of association rate constants (k_on) shows that several top-performing aptamers achieve high affinity despite moderate k_on values. (C) Inverse dissociation rate constants (1/k_off) highlight the dramatically slower target dissociation exhibited by the symmetric trimerized aptamers, distinguishing them from all other designs. The shTMSA52t aptamer targeting the trimeric wild-type spike protein (ranked #1 in affinity) is included for reference. Complete comparisons of all 45 aptamers are provided in Supplementary Table S2 and Supplementary Fig. S10.

We first examined the top 12 highest-affinity aptamers in the dataset. Among them, shTMSA52t exhibited the strongest binding, with a K_d of 4.7 pM for the trimeric wild-type S protein and 12.5 pM for the monomeric wild-type S1 protein, ranking first in the compiled dataset (Fig. 6A). shTTro4, targeting monomeric troponin I, ranked second with a K_d of 17.3 pM. shTH4r3, which binds dimeric VEGF₁₆₅, showed a K_d of 79 pM, ranking eighth overall, just behind AptD4 (30 pM), RBD-PB6 trimer (39 pM), AptD3 (57 pM), and RBD-PB6 dimer (72 pM). Notably, several of these top-performing literature aptamers (e.g. AptD1–AptD4) incorporate chemically modified bases [56], while our trimeric aptamers consist entirely of natural DNA sequences.

To understand the mechanistic origin of the enhanced affinity, we compared k_on and k_off values across the top-performing constructs. As shown in Fig. 6B, the symmetric homotrimeric aptamers did not display unusually high k_on values, suggesting that rapid association was not the dominant contributor to their high affinity. However, their k_off values were uniquely low. When plotted as 1/k_off (Fig. 6C), the symmetric trimers stood out dramatically, indicating that slow dissociation is the principal driver of their ultrahigh affinity.

Detection of VEGF165 in serum using symmetric aptamer with suppressed dissociation rate

To evaluate the functional advantage of symmetric aptamers with suppressed dissociation kinetics, we performed an enzyme-linked aptamer assays (ELAA) for VEGF₁₆₅ detection in serum using the trimeric aptamer shTH4r3 and its monomeric counterpart H4r3 (Fig. 7A). We hypothesized that the slow dissociation rate of shTH4r3 would improve target retention and thus enhance the sensitivity of detection compared to the faster-dissociating monomeric aptamer.

Figure 7.

ELAA for detection of VEGF₁₆₅ in serum using symmetric (shTH4r3) and monomeric (H4r3) aptamers. (A) Schematic illustration of the ELAA. VEGF₁₆₅ is first captured by an immobilized anti-VEGF₁₆₅ antibody. A biotinylated aptamer (shTH4r3 or H4r3) was used as the detection probe to form a sandwich structure with the captured VEGF₁₆₅ protein, followed by streptavidin–HRP labeling. Upon addition of the TMB substrate and H₂O₂, HRP catalyzes a colorimetric reaction. (B) Photographs of the assay showing increasing color intensity with rising VEGF₁₆₅ concentrations (0 to 1000 pM), with shTH4r3 consistently producing stronger signals than H4r3. (C) Dose–response curves based on absorbance at 450 nm (A450) for each aptamer. The limit of detection (LOD) was defined as the lowest VEGF₁₆₅ concentration yielding a signal greater than three times the standard deviation of the blank. The LOD for shTH4r3 was 1.1 ± 0.3 pM, ~30-fold lower than that of H4r3 (32.4 ± 4.2 pM). Error bars represent standard deviation (n = 3).

In this assay, an anti-VEGF₁₆₅ antibody was immobilized on the plate to capture VEGF₁₆₅ from solution. After incubation with varying concentrations of VEGF₁₆₅, biotinylated shTH4r3 or H4r3 aptamers were introduced as detection probes. Bound aptamers were subsequently labeled with streptavidin–horseradish peroxidase (HRP), followed by the addition of TMB substrate. HRP catalyzes the oxidation of TMB in the presence of hydrogen peroxide, producing a yellow product that was quantified by absorbance at 450 nm (A450), which correlates with VEGF₁₆₅ concentration. Because the antibody and aptamer bind different areas on VEGF₁₆₅, their combination avoids competitive binding [57].

As shown in Fig. 7B, the color intensity was notably stronger with shTH4r3 than with H4r3 across all tested VEGF₁₆₅ concentrations. The corresponding dose–response curves (Fig. 7C) confirmed enhanced sensitivity with shTH4r3. The limit of detection (LOD) for shTH4r3 was 1.1 ± 0.3 pM, ~30-fold lower than that of H4r3 (32.4 ± 4.2 pM). These results support that suppressing dissociation through symmetric aptamer trimerization improves sensitivity in complex biological matrices. This strategy provides a generalizable approach for enhancing aptamer-based assays, particularly for detecting low-abundance targets in clinically relevant samples.

Conclusion

In this study, we developed a symmetric aptamer trimerization strategy that significantly enhances binding affinity by suppressing dissociation rates. By assembling three identical aptamer units in flexible orientation through a symmetric linker, we created homotrimeric aptamers that outperformed their monomeric counterparts in binding strength and stability. This approach was effective across protein targets with distinct oligomeric states—including monomeric, dimeric, and trimeric proteins—and consistently yielded over 100-fold improvements in affinity.

The symmetric trimerization strategy provides a generalizable and robust post-SELEX approach for enhancing aptamer affinity. In contrast to traditional multimeric designs that rely on avidity or increased local concentration, the symmetric assemblies enable cooperative kinetic stabilization through spatial alignment of aptamer arms. The resulting constructs not only exhibit ultrahigh affinity, but also show exceptional binding stability, making them highly promising for applications in biosensing, diagnostics, and targeted therapeutics. While this strategy has been demonstrated to be successful for protein-binding aptamers, its applicability to other types of targets, such as aptamers for small molecules, remains to be explored.

Compared with trimeric aptamer assemblies based on rigid geometry, such as the recent MEDUSA (multivalent evolved DNA-based supramolecular assembly) [58] that precisely control the relative orientation and spacing of binding arms to optimize multivalent engagement, our flexible trimerization strategy allows adaptive exploration of the target surface. This flexibility enables dynamic local rebinding and provides greater tolerance to variations in epitope spacing and target conformational dynamics. Flexible trebler linkers can create both advantages and disadvantages. First, multiple binding arms may experience spatial arm–arm interference, and simultaneous engagement of all arms is not necessarily achieved without constraint. Second, there is a trade-off associated with flexibility: while flexible linkers allow binding arms to explore conformational space and adapt to different spatial arrangements of target sites, excessive flexibility can impose an entropic penalty. Upon binding, flexible arms transition from a freer, disordered state to a more constrained state, resulting in a loss of conformational entropy that is thermodynamically unfavorable. Third, for multivalent constructs with sufficient flexibility, a local rebinding effect—where a dissociated arm rapidly rebinds to the target—may occur even in the absence of a perfectly symmetric geometric arrangement, which could be examined in future studies.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was supported by funding from the National Key R&D Program of China (2024YFA0918900 to ZZ), Canadian Institutes of Health Research (GA5-177777 to YL), Natural Sciences and Engineering Research Council of Canada (ALLRP 570428-2021 to YL). Funding to pay the Open Access publication charges for this article was provided by Canadian Institutes of Health Research (GA5-177777).

Data availability

The data underlying this article are available in the article and in its online supplementary material.