Structure-based investigation of a DNA aptamer targeting PTK7 reveals an intricate 3D fold guiding functional optimization

Significance

Numerous efforts have been devoted to designing potent DNA aptamers for various applications. Efficient optimization of sgc8c, one of the foremost DNA aptamers targeting membrane proteins for disease theranostics, has been impeded by a lack of high-resolution structural guidance. We employed a streamlined NMR-based approach to determine the intricate 3D structure and essential binding site of sgc8c without acquiring the structure of sgc8c–protein tyrosine kinase 7 complex. The well-established structure–function relationship guides efficient engineering of sgc8c variants with significantly enhanced thermostability, biostability, and binding affinity to both protein and cell targets. This work provides encouragement to overcome challenges in understanding and optimizing DNA aptamers targeting membrane proteins and offers insights into sophisticated structure–function organization of DNA molecules.

Abstract

DNA aptamers have emerged as novel molecular tools in disease theranostics owing to their high binding affinity and specificity for protein targets, which rely on their ability to fold into distinctive three-dimensional (3D) structures. However, delicate atomic interactions that shape the 3D structures are often ignored when designing and modeling aptamers, leading to inefficient functional optimization. Challenges persist in determining high-resolution aptamer–protein complex structures. Moreover, the experimentally determined 3D structures of DNA molecules with exquisite functions remain scarce. These factors impede our comprehension and optimization of some important DNA aptamers. Here, we performed a streamlined solution NMR-based structural investigation on the 41-nt sgc8c, a prominent DNA aptamer used to target membrane protein tyrosine kinase 7, for cancer theranostics. We show that sgc8c prefolds into an intricate three-way junction (3WJ) structure stabilized by long-range tertiary interactions and extensive base–base stackings. Delineated by NMR chemical shift perturbations, site-directed mutagenesis, and 3D structural information, we identified essential nucleotides constituting the key functional elements of sgc8c that are centralized at the core of 3WJ. Leveraging the well-established structure–function relationship, we efficiently engineered two sgc8c variants by modifying the apical loop and introducing _L-DNA base pairs to simultaneously enhance thermostability, biostability, and binding affinity for both protein and cell targets, a feat not previously attained despite extensive efforts. This work showcases a simplified NMR-based approach to comprehend and optimize sgc8c without acquiring the complex structure, and offers principles for the sophisticated structure–function organization of DNA molecules.

Aptamers are functional nucleic acids capable of binding to various targets, ranging from small molecules to proteins and whole cells, with high affinity and specificity (1–3). In particular, aptamers that target pivotal proteins relevant to diseases have emerged as innovative molecular tools for clinical diagnosis, targeted drug delivery, and protein degradation (4–6). The functional efficacy of an aptamer fundamentally stems from its intricate three-dimensional (3D) structure (7–12). Although deciphering the 3D structure of an aptamer in complex with its protein target is the most direct approach for comprehending and optimizing its functionality, this approach is not consistently surmountable. The inherent conformational flexibility arising from unpaired nucleotides in aptamers often hinders the acquisition of high-quality particles and precise 3D structure modeling for cryoelectron microscopy (cryo-EM) and X-ray crystallography (13, 14). Nuclear magnetic resonance (NMR) spectroscopy is applicable for resolving structures of small nucleic acids with conformational flexibility, but determination of complex structures encounters hurdles such as NMR peak broadening and overlapping in the presence of a large protein partner (15, 16). Recent advancements in 3D structure modeling and prediction for nucleic acid–protein complexes, represented by tools such as CryoREAD (17), OPUS-DSD (18), and RoseTTAFoldNA (19), have bolstered this domain with optimal efficacy on RNA molecules. However, the 3D structures of DNA molecules, which are perceived to lack RNA-like tertiary interactions, remain largely undervalued. Addressing this gap holds immense promise for advancing our comprehension and functional optimization of DNA aptamers in targeted protein interactions.

Examination of DNA aptamer–protein complex structures in the Protein Data Bank (20) reveals that DNA aptamers commonly use nucleotides from unpaired regions to construct the binding site and employ paired regions as a scaffold to maintain the 3D fold (21–24). If binding of the aptamer to its protein target only induces a local conformational change at or near the binding site without destroying the 3D fold, the use of NMR chemical shift perturbations (CSPs) will be an effective approach for studying aptamer functionality. This idea is exemplified analogously by a recent breakthrough work on NMR-guided enzyme evolution (25). In addition, resolving the 3D structure of an aptamer by incorporating Watson–Crick base pairing information, which are accessible through NMR experiments, may facilitate the restriction of other regions exhibiting large conformational plasticity. This approach is recently demonstrated by successful NMR structural determination of the 10-23 DNAzyme (26). We propose that an integration of these NMR-based approaches may enable the structural elucidation and functional optimization of a DNA aptamer without acquiring the structure of aptamer–protein complex.

Sgc8c is a 41-nucleotide (nt) DNA aptamer screened through cell-SELEX (systematic evolution of ligands by exponential enrichment) to target leukemia cells with a high binding affinity, i.e., an equilibrium dissociation constant (K_D) of aptamer–cell interaction at the nanomolar-to-picomolar scale (27, 28). The molecular target of sgc8c is protein tyrosine kinase 7 (PTK7) (29), a transmembrane receptor pseudokinase associated with cell growth (30) that is overexpressed in various types of cancers, including leukemia (31), colon (30), non–small cell lung (32), and gastric cancers (33). PTK7 comprises an extracellular domain featuring seven immunoglobulin-like motifs, a transmembrane domain, and a catalytic domain that lacks kinase activity (34). Specifically, sgc8c binds to the extracellular domain of PTK7 with a nanomolar-scale K_D (35, 36). Owing to its high binding specificity and affinity for both protein and cell targets, sgc8c has become one of the most widely used DNA aptamers in cancer theranostics (35, 37–39). Nevertheless, its inherent conformational flexibility presents hurdles in determining the 3D structure of sgc8c using cryo-EM and X-ray crystallization. Consequently, the structural basis underlying the nuanced functionality of sgc8c remains elusive, underscoring the imperative need for structure-guided functional comprehension and optimization of sgc8c.

In this work, we showcase a streamlined NMR-based structural investigation to guide efficient structure–function analysis of sgc8c without the need for obtaining sgc8c–PTK7 complex structure. NMR CSPs and site-directed mutagenesis experiments revealed that nucleotides from the shortest paired region of sgc8c constitute the key binding element for PTK7. The solution NMR structure of sgc8c revealed an intricate three-way junction (3WJ) fold stabilized by long-range hydrogen bonding and extensive base–base stacking interactions, with pivotal residues being recruited to the center of 3WJ. Leveraging insights from the binding site and 3D structure of sgc8c, we efficiently designed two sgc8c variants that exhibit enhanced thermostability, biostability, and notably improved binding affinity to both protein and cell targets. Our results highlight the importance of long-range tertiary interactions, which are common in RNA but rare in DNA molecules, to the delicate structure and sophisticated function of sgc8c, offering principles for designing advanced biomaterials and techniques centered around functional DNA molecules.

Results

Secondary Structure and Key Binding Site of sgc8c.

Sgc8c was screened through cell-SELEX by incubating the DNA pool with CCRF-CEM cells at ~4 °C in saline buffer supplemented with Mg²⁺ (27, 28). To make the condition of structural determination more relevant to that of cell-SELEX, we prepared sgc8c in sodium phosphate buffer (NaPi, pH 7) with 5 mM MgCl₂ and first acquired the one-dimensional (1D) ¹H NMR spectrum at 4 °C. Ten imino proton (guanine G H1 and thymine T H3) signals were observed at 12.5 to 14.0 ppm, suggesting that sgc8c formed at least 10 Watson–Crick base pairs (Fig. 1A). The 1D ¹H NMR spectrum acquired at 25 °C shows peak broadening of several imino signals (SI Appendix, Fig. S1), which could be attributed to the occurrence of unstable base pairings or the presence of conformational dynamics at a higher temperature. This offers a possible explanation for the challenges encountered in assembling sgc8c at room temperature (40) and resolving its crystal structure.

Fig. 1.

The secondary structure and binding site of sgc8c. (A) 2D NOESY NMR spectrum of sgc8c shows NOEs between two imino protons from neighboring Watson–Crick base pairs. Imino protons include guanine G H1 and thymine T H3. [DNA] = 0.5 mM, [NaPi, pH 7] = 10 mM, [MgCl₂] = 5 mM, 90% H₂O/10% D₂O, T = 4 °C. (B) Schematic for the simplified secondary structure of sgc8c. (C) The 1D ¹H NMR spectra show G H1 and T H3 signals of sgc8c in the presence of 0, 0.5, and 1 equivalent of PTK7. [DNA] = 0.025 mM, [NaPi, pH 7] = 10 mM, [MgCl₂] = 5 mM, 90% H₂O/10% D₂O, T = 4 °C. (D) The NMR CSPs profile constructed by measuring the chemical shift changes (absolute value) of G H1, T H3, and thymine methyl protons (T H7) of sgc8c before and after adding 1 equivalent of PTK7 at 4 °C. (E) The K_D of aptamer–PTK7 interaction for G9·C13, C10·G12, A5·T38, C18·G36, and T11 mutants determined by SPR at 25 °C. N/A indicates no measurable binding. The original SPR spectra are shown in SI Appendix, Figs. S8 and S9. (F–J) Flow cytometry results for G9·C13 mutants (F), C10·G12 mutants (G), A5·T38 mutants (H), C18·G36 mutants (I), and T11 mutants (J) to CCRF-CEM cells. T = 4 °C for incubation of aptamers and cells.

We then conducted more detailed NMR investigations at 4 °C to characterize the secondary structure of sgc8c. To aid resonance assignments of thymine H7 and guanine H1 of sgc8c, we acquired 1D ¹H NMR spectra for 20 sgc8c variants, including 9 with a single-nucleotide substitution of thymine to 2′-deoxyuridine (dU) and 11 with a single-nucleotide substitution of guanine to 8-oxo-7,8-dihydroguanine (o⁸G) (SI Appendix, Figs. S2 and S3 and Table S1). The H1 signals of G9, G12, G20, G35, G36, and G40 were identified at 12.5 to 13.5 ppm, suggesting that they formed G·C Watson–Crick base pairs. Using 10 sgc8c variants with a single-nucleotide substitution of cytosine to 5-methylcytosine (m⁵C), we probed C3·G40, G9·C13, C10·G12, C18·G36, C19·G35 and G20·C34 Watson–Crick base pairs (SI Appendix, Figs. S4 and S5), which were further confirmed by two-dimensional (2D) nuclear Overhauser effect spectroscopy (NOESY) of sgc8c, including the corresponding G H1-C H41/H42 NOEs (SI Appendix, Fig. S6A) and imino–imino proton NOEs between neighboring base pairs (Fig. 1A). The T2·A41, T4·A39, A5·T38 and A6·T37 Watson–Crick base pairs were identified through the corresponding T H3-A H2 NOEs (SI Appendix, Fig. S6B) and imino–imino NOEs between neighboring base pairs (Fig. 1A). Probing the 10 Watson–Crick base pairs shaped the secondary structure of sgc8c containing three paired regions (Fig. 1B), which is different from the plain stem–loop structures predicted by Mfold and NUPACK (SI Appendix, Fig. S7).

We next examined whether sgc8c could maintain its 3D fold after binding to PTK7, which is a prerequisite for our NMR-based approach of aptamer structure–function analysis in the absence of a complex structure. We acquired 1D ¹H NMR spectra for sgc8c upon the addition of PTK7. The imino proton signals of sgc8c generally remained intact after the addition of 0.5 or 1 equivalent of PTK7 (Fig. 1C), suggesting that sgc8c could retain its original 3D fold after binding to PTK7. The overall peak broadening of sgc8c at 1 equivalent of PTK7 was due to a lower tumbling rate upon binding to a large protein. We constructed the NMR CSPs profile by measuring changes in the chemical shifts of imino protons (G H1 and T H3) and thymine methyl protons (T H7) of sgc8c before and after adding 1 equivalent of PTK7. These types of NMR signals were recorded because they were well separated and could be clearly monitored. NMR chemical shifts are sensitive to structural changes, and residues that show large CSPs usually represent binding sites (25). G9 H1, G12 H1, T38 H3, G36 H1, and T11 H7 showed relatively large CSPs (Fig. 1D), suggesting that these residues might be in the binding site.

The important functional sites of biomolecules are typically sensitive to mutations. To further verify the residues involved in the binding site of sgc8c, we performed site-directed mutagenesis by substituting i) G9·C13 to C·G, A·T and T·A, ii) C10·G12 to G·C, A·T and T·A, iii) A5·T38 to T·A, G·C and C·G, and iv) C18·G36 to G·C, A·T and T·A. We assessed their binding affinity for the PTK7 protein and CCRF-CEM cells which highly express PTK7. The K_D of the aptamer–PTK7 interaction was determined using surface plasmon resonance (SPR) at 25 °C, a regular temperature condition for measuring nucleic acid–protein binding affinity in vitro. Following the routine of cell-SELEX (27), we assessed the binding to CCRF-CEM cells using flow cytometry by incubating the aptamer and cells at 4 °C. Compared to the K_D of aptamer–PTK7 interaction for wild-type sgc8c (0.41 ± 0.04 nM), the G9·C13 and C10·G12 mutants showed no measurable binding to PTK7, whereas A5·T38 and C18·G36 were tolerant to mutagenesis with K_D values below 1.3 nM (Fig. 1E and SI Appendix, Fig. S8). In line with the SPR results, binding to CCRF-CEM cells was abolished or significantly weakened for the G9·C13 and C10·G12 mutants (Fig. 1 F and G) but nearly unaffected for the A5·T38 and C18·G36 mutants (Fig. 1 H–I). These results suggested that G9·C13 and C10·G12 were involved in the binding site. In addition, T11 displayed the largest thymine H7 CSP (Fig. 1D) and it was also proximal to G9·C13 and C10·G12. We further investigated the role of T11 in binding. Substitution of T11 to A or G resulted in weakened binding to the PTK7 protein (K_D of 3.7 ± 0.8 and 62 ± 21 nM, respectively, Fig. 1E and SI Appendix, Fig. S9) and to CCRF-CEM cells (Fig. 1J), suggesting that base–base stacking involving T11 did not serve as a key binding interaction; otherwise, A or G mutation would provide favorable base–base stacking through their larger aromatic rings. Substitution of T11 to C or dU maintained favorable binding affinity for PTK7 (K_D of 0.4 ± 0.2 and 0.8 ± 0.3 nM, respectively), suggesting that the methyl group of T11 did not participate in key binding interaction. It appears that the #11 nucleotide prefers a pyrimidine nucleobase to exert the aptamer functionality. The results of NMR CSPs and site-directed mutagenesis collectively revealed that nucleotides from the shortest paired region, including G9·C13, C10·G12 and T11, primarily constitute the key binding site and functional element of sgc8c.

3D Structure of the sgc8c DNA Aptamer.

Given that binding to PTK7 did not perturb the original 3D fold of sgc8c, we proceeded to determine the solution NMR structures of free sgc8c using restrained molecular dynamics (rMD) simulations with NMR experimental restraints. Ten structures were selected as the final representative ensemble (SI Appendix, Fig. S10 and Table S2). Sgc8c adopts an intricate 3WJ fold stabilized by long-range nucleobase-backbone/nucleobase interactions and extensive base–base stackings. This 3WJ consists of three paired regions, including the terminal P1 (nucleotides #1-6 and 37-41), the shortest P2 (nucleotides #9-13), and the longest P3 (nucleotides #18-36). There are two junction sites, namely, J1 (nucleotides #7-8) and J2 (nucleotides #14-17) (Fig. 2). One notable structural feature of sgc8c is the coaxial stack between P1 and P3 through A6·T37/C18·G36 base pairs, aligning almost colinearly and spanning approximately 53 Å. Such an extensive helical stacking configuration has been observed in the 3WJs and four-way junctions (4WJs) of RNA molecules such as the Pepper (10) and Chili aptamers (41) and, most recently, in the 4WJ of Lettuce DNA aptamer (42). The 3D fold of sgc8c was corroborated by imino–imino proton NOEs between neighboring base pairs, particularly the G36 H1-T37 H3 NOE that supports the coaxial stacking of P1 and P3 (Fig. 1A).

Fig. 2.

The 3D structure of sgc8c DNA aptamer. (A) Cartoon representation using one of the 10 representative solution NMR structures of sgc8c (PDB ID: 8Y0F). Arrows denote 5'-to-3' chain direction. Residues from P1, P2, P3, and junctions (J1 and J2) are shown in yellow, pink, blue, and red colors, respectively. The superimposed 10 solution NMR structures and their NMR refinement statistics are shown in SI Appendix, Fig. S10 and Table S2, respectively. (B) Schematic for the secondary structure of sgc8c. Lines with dashes and Leontis–Westhof symbols denote Watson–Crick base pairs and noncanonical hydrogen bond(s), respectively.

It has been well established that divalent metal ions such as Mg²⁺ have stabilizing effects on DNA/RNA 3WJ and 4WJ structures (10, 42–44) by forming electrostatic interactions and neutralizing negatively charged phosphodiester backbones. We analyzed the distribution of Mg²⁺ during rMD simulations of sgc8c and found that Mg²⁺ ions were predominantly localized at junction sites and vicinal DNA strands, where electrostatic potentials were more negative (SI Appendix, Fig. S11). We then experimentally examined the effects of Mg²⁺ on the structure and function of sgc8c. In the absence of Mg²⁺, the 1D ¹H NMR spectrum of sgc8c did not show imino proton signals at 12.5 to 14.0 ppm, suggesting that sgc8c was unable to fold into a stable structure under this condition (SI Appendix, Fig. S12A). Consistent with the NMR data, the K_D of sgc8c–PTK7 interaction in the absence of Mg²⁺ was ~922-fold greater than that in the presence of 5 mM Mg²⁺ (SI Appendix, Fig. S12B). These results highlight a critical role of Mg²⁺ in maintaining the structure and function of sgc8c.

Long-Range Tertiary Interactions Are Crucial to the Structure and Function of sgc8c.

In sgc8c, P1 and P3 form a scaffold through a long helical stack, leaving the shortest P2 as an accessible binding site. More intriguingly, we also identified several long-range nucleobase-backbone/nucleobase tertiary interactions, which are rarely observed in DNA molecules. Specifically, C7 from J1 folded into the minor groove of P1, and used its amino proton H41/H42 to form hydrogen bond(s) with A39 N3, T38 O2, and/or T38 O3'. G17 from J2 folded into the major groove of P1, and used its imino proton H1 and amino proton H21 to form hydrogen bonds with C3 OP2 and C3 OP1, respectively (Fig. 3A and SI Appendix, Fig. S13). These long-range interactions appear to stabilize the helical stack of P1 and P3. To investigate the importance of these long-range interactions involving the C7 and G17 nucleobases, we substituted C7 or G17 to an abasic site (dS). Compared with sgc8c, the 1D ¹H NMR spectra of C7dS and G17dS showed altered imino signal patterns (Fig. 3B), suggesting that C7dS and G17dS were incapable of maintaining the original 3D fold, thus confirming the importance of the C7 and G17 nucleobases to the structure of sgc8c. Consistently, C7dS and G17dS had no measurable binding to PTK7 (Fig. 3C and SI Appendix, Fig. S14) and CCRF-CEM cells (SI Appendix, Fig. S15). In addition, substitutions of C7 to natural T/A/G, and G17 to natural T/A/C also abolished or significantly weakened the binding affinity for both the PTK7 protein and CCRF-CEM cell targets (Fig. 3C and SI Appendix, Figs. S14 and S15). To specifically examine the importance of long-range hydrogen bond interactions involving C7 and G17, we designed i) one mutant by substituting C7 to dU that lacks the N4 amino group, and ii) two mutants by substituting G17 to 2-aminopurine (2AP) that lacks the N1 imino group, or xanthosine (X) that lacks the N2 amino group. The C7dU, G17-2AP, and G17X mutants had K_D values of 575 ± 83 nM, 232 ± 76 nM and no measurable binding to PTK7, respectively, and their binding to CCRF-CEM cells were also significantly weakened (Fig. 3C and SI Appendix, Fig. S16). Overall, these long-range tertiary interactions, which rarely exist in DNA molecules, are crucial to the structure and function of sgc8c.

Fig. 3.

Long-range tertiary interactions are crucial to the structure and function of sgc8c. (A) Cartoon representation of sgc8c shows C7 H41 formed a hydrogen bond with A39 N3, and G17 H1 and H21 formed hydrogen bonds with C3 OP2 and C3 OP1, respectively. In the other representative NMR structures, C7 H41/H42 also formed hydrogen bonds with T38 O2 or T38 O3' as shown in SI Appendix, Fig. S13. (B) 1D ¹H NMR spectra show imino proton signals (G H1 and T H3) of C7dS and G17dS in comparison with sgc8c. [DNA] = 0.2 mM, [NaPi, pH 7] = 10 mM, [MgCl₂] = 5 mM, 90% H₂O/10% D₂O, T = 4 °C. (C) The K_D of aptamer–PTK7 interaction for C7, G17, C19·G35, and G20·C34 mutants determined by SPR at 25 °C. The original SPR spectra are shown in SI Appendix, Figs. S14, S16, and S17. (D) Cartoon representation shows base–base stackings involving C19·G35 and G20·C34 Watson–Crick base pairs from P3.

The longest paired region P3 of sgc8c features extensive base–base stackings, including three consecutive C·G Watson–Crick base pairs located at the center of 3WJ, along with stacking of unpaired G21 and A33 (Fig. 3D). As previously mentioned, C18·G36 was tolerant to mutation (Fig. 1 E and I), possibly because it faced the junction, allowing greater flexibility in adjusting its coaxial stack with A6·T37. In contrast, the mutation of i) C19·G35 to G·C, T·A or A·T, and ii) G20·C34 to C·G, T·A or A·T was detrimental to the binding to both the PTK7 protein and CCRF-CEM cell targets (Fig. 3C and SI Appendix, Figs. S17 and S18).

Structure-Guided Functional Optimization of sgc8c.

In the realm of aptamer research, a continuous demand persists for enhancing the thermostability and biostability of aptamers to prevent undesired folding, misdirected targeting, and degradation in biological environments that are far more complex and formidable than in vitro conditions. However, enhancing the functionality of conformationally flexible DNA aptamers is challenging, especially when high-resolution structural guidance is lacking. Fig. 4A displays the 3D structure of sgc8c using a surface mode to highlight our experimentally verified nucleotides that are essential to the structure and function, including G9·C13, C10·G12 and T11 from P2 serving as the binding site, C7 and G17 from junction sites participating in long-range tertiary interactions, and C19·G35 and G20·C34 from P3 supporting the scaffold. Notably, sgc8c can recruit more than 10 nucleotides from distinct regions and assemble them into its key structural and functional framework. A recent study showed that modification of T4, T11, T32, T37, or T38 in sgc8c with an alkyne group significantly weakened the binding to PTK7 (35), but this could not be fully rationalized using the predicted plain stem–loop structures. These thymine residues are within or proximal to the surface of our defined essential site (SI Appendix, Fig. S19), and thus introducing an alkyne group might cause steric hindrance and structure–function perturbations. The insights derived from our established structure–function relationship of sgc8c suggest that optimizing the functionality could be achieved by modifying the apex of P3 and the bottom of P1.

Fig. 4.

Structure-guided functional optimization of sgc8c. (A) 3D structure of sgc8c shows the essential site composed of G9·C13, C10·G12, T11, C7, G17, C19·G35 and G20·C34 in a surface mode. The essential site is crucial to the structure and function of sgc8c. (B) 1D ¹H NMR spectra show imino proton signals (G H1 and T H3) of GAA_25-28 and GAA_25-29 in comparison with sgc8c. [DNA] = 0.2 mM, [NaPi, pH 7] = 10 mM, [MgCl₂] = 5 mM, 90% H₂O/10% D₂O, T = 4 °C. (C–E) The T_m determined by optical melting experiments, K_D of aptamer–PTK7 interaction determined by SPR at 25 °C, and K_D of aptamer–cell interaction determined by flow cytometry (37 °C incubation) for GAA_25-28 and GAA_25-29. (F) Denaturing polyacrylamide gel electrophoresis (PAGE) shows DNA bands of sgc8c, L1-2, L1-3, L1-4, L+2bp, and L+3bp without and with treatment of combined Exonuclease I (0.15 U/μL) and Exonuclease III (0.5 U/μL) at 37 °C for 15 min. (G–H) The K_D of aptamer–PTK7 interaction determined by SPR at 25 °C, and K_D of aptamer–cell interaction determined by flow cytometry (37 °C incubation) for L1-3, L1-4, L+2bp, and L+3bp. The secondary structures of GAA_25-28, GAA_25-29, L1-3, L1-4, L+2bp, and L+3bp, the melting, SPR, and fluorescence curves used to determine T_m and K_D are shown in SI Appendix, Figs. S20 and S21.

The P3 region in sgc8c is capped by an AATA tetraloop (nucleotides #25-28) with C29 flipping out (Fig. 4A). The AATA tetraloop does not belong to families of ultrastable DNA loops (45). To enhance the thermostability, we engineered two variants by replacing the AATA with GAA, either retaining or deleting C29. These two variants were named GAA_25-28 and GAA_25-29, respectively (SI Appendix, Fig. S20A). The GAA loop is known to be thermodynamically ultrastable due to the formation of a sheared G·A loop-closing base pair that exhibits extensive base–base stacking with neighboring residues (46). We next examined the structure, thermostability, and binding function of GAA_25-28 and GAA_25-29. The 1D ¹H NMR spectra of GAA_25-28 and GAA_25-29 showed similar imino signal patterns compared with those of sgc8c (Fig. 4B), confirming that these two variants maintained the original 3D fold and reinforcing the reliability of our determined solution NMR structures. The melting temperatures (T_m) of GAA_25-28 and GAA_25-29 increased by ~7 and 11 °C, respectively (Fig. 4C and SI Appendix, Table S3). The binding affinity of GAA_25-28 and GAA_25-29 to PTK7 was enhanced and unaffected, with K_D values of 0.22 ± 0.03 and 0.4 ± 0.2 nM, respectively (Fig. 4D). We further examined whether the enhanced thermostability could improve the binding affinity to cell targets at a physiological temperature, which will benefit in vivo applications of aptamers such as targeted drug delivery and tumor imaging. The aptamers were incubated with CCRF-CEM cells at 37 °C for 30 min and then subjected to flow cytometry. The K_D of aptamer–cell interaction for GAA_25-28 and GAA_25-29 reached a picomolar scale, i.e., 0.7 ± 0.1 and 0.9 ± 0.2 nM, respectively (Fig. 4E), which were ~3.7- and 2.9-fold smaller than that for sgc8c (2.6 ± 0.2 nM). Compared to GAA_25-29, GAA_25-28 exhibited a slightly lower thermostability but a higher binding affinity, revealing that the retention of C29 benefits the binding function of sgc8c. The melting, SPR, and fluorescence curves used to determine T_m and K_D are shown in SI Appendix, Fig. S20 B–D.

Previous studies improved the biostability of sgc8c by conjugation with nanoparticles (47, 48), design of circular bivalent aptamers (49), use of crosslinked nucleotides (36), and incorporation of a small-molecule aptamer (50). However, steric hindrance-based modifications inevitably sacrificed the binding affinity. Compared to natural _D-conformation DNA (_D-DNA), chirally inverted mirror-image DNA (_L-DNA) is nuclease-resistant and biostable (51). Here, we sought to enhance the biostability of sgc8c variants with improved binding affinity. We first designed three mutants by substituting nucleotides in the terminal 1-bp (A1 and T2·A41), 2-bp (A1, T2·A41 and C3·G40), and 3-bp (A1, T2·A41, C3·G40 and T4·A39) to the corresponding _L-DNA, which were abbreviated as L1-2, L1-3, and L1-4, respectively (SI Appendix, Fig. S21). Upon incubation with combined Exonuclease I and Exonuclease III at 37 °C for 15 min, sgc8c and L1-2 were completely cleaved, whereas L1-3 and L1-4 remained intact (Fig. 4F). However, the binding affinity of L1-3 and L1-4 for both protein and cell targets were lower than that of sgc8c (Fig. 4 G and H). These results could be attributed to the mirrored residues that perturbed the important long-range hydrogen bond interactions, particularly those involving C3 and A39 (Fig. 3A). To preserve these crucial tertiary interactions, we extended two and three _L-DNA base pairs and named them L+2bp and L+3bp, respectively (SI Appendix, Fig. S21). L+2bp and L+3bp exhibited excellent resistance to exonuclease cleavage, and retained high binding affinity to both protein and cell targets comparable to that of sgc8c (Fig. 4 F–H).

Through an exhaustive endeavor to simultaneously enhance the binding affinity, thermostability, and biostability, which has not been achieved previously, we amalgamated modifications from GAA_25-28 and L+2bp or L+3bp to engineer two variants, which were named sgc8c-x1 and sgc8c-x2, respectively (Fig. 5A). Integration of these two modifications led to improvements in multiple facets: sgc8c-x1 and sgc8c-x2 exhibited greater thermostability (T_m of 53 ± 1 and 52 ± 2 °C, respectively, versus 45 ± 1 °C for sgc8c) (Fig. 5B and SI Appendix, Fig. S22A), enhanced binding affinity to the PTK7 protein (K_D of 0.18 ± 0.06 and 0.13 ± 0.01 nM, respectively, versus 0.41 ± 0.04 nM for sgc8c at 25 °C) (Fig. 5C and SI Appendix, Fig. S22B), enhanced binding affinity to CCRF-CEM cells (K_D of 1.3 ± 0.2 and 0.7 ± 0.2 nM, respectively, versus 2.6 ± 0.2 nM for sgc8c at 37 °C) (Fig. 5D and SI Appendix, Fig. S22C), and superior resistance to exonucleases (SI Appendix, Fig. S23) and fetal bovine serum (FBS) (Fig. 5E). After incubation with 10% FBS for 48 h, sgc8c lost its ability to bind CCRF-CEM cells due to degradation, whereas sgc8c-x1 and sgc8c-x2 retained high binding capacity (Fig. 5F). By leveraging sgc8c-x1 as a model, we acquired its 2D NOESY NMR spectrum and confirmed that sgc8c-x1 adopted a 3D fold similar to that of sgc8c (SI Appendix, Fig. S24). These results illustrate the efficacy of guided functional optimization for a membrane protein-binding DNA aptamer, by the streamlined NMR-based approach in the absence of complex structural information. Moreover, the optimized sgc8c-x1 and sgc8c-x2 hold significant promise for diverse aptamer-based biomedical applications.

Fig. 5.

Structure and function of the optimized sgc8c-x1 and sgc8c-x2. (A) Schematic for the secondary structures of sgc8c-x1 and sgc8c-x2. Modifications relative to the wild-type sgc8c are shown in dotted box. (B–D) The T_m, K_D of aptamer–PTK7 interaction at 25 °C, and K_D of aptamer–cell interaction (37 °C incubation) for sgc8c-x1 and sgc8c-x2 in comparison with sgc8c. The melting, SPR, and fluorescence curves used to determine T_m and K_D are shown in SI Appendix, Fig. S22. (E) PAGE shows DNA bands of sgc8c, sgc8c-x1, and sgc8c-x2 after treatment with 10% FBS for different time intervals. (F) Flow cytometry results for sgc8c, sgc8c-x1and sgc8c-x2 to CCRF-CEM cells after treatment with 10% FBS for different time intervals.

Discussion

Numerous endeavors have been devoted to optimizing the functionality of sgc8c, one of the foremost DNA aptamers targeting membrane proteins for various biomedical applications (36, 47–50). However, the efficient optimization of sgc8c has been impeded in the past two decades by a lack of high-resolution structural guidance. This challenge is also prevalent among other membrane protein-targeting DNA aptamers, such as the programmed death-ligand 1 aptamer (52). Drawing inspiration from recent breakthrough work of NMR-guided enzyme evolution (25), here we showcase a streamlined NMR-based approach to establish the structure–function relationship and functional optimization of sgc8c without the need to acquire complex structural information. Our findings have two implications for the structural and functional study of DNA aptamers with intricate 3D folds.

First, under the confirmed prerequisite that sgc8c maintains its original 3D fold after binding to PTK7 (Fig. 1C), we explored the binding site on sgc8c using NMR CSPs and site-directed mutagenesis. The rationale is that nucleotides involved in the binding site have large CSPs, and their mutations usually affect essential functions. G9 and G12 showed large CSPs, and the G9·C13 and C10·G12 base pairs were intolerant to mutations (Fig. 1 E–G). For T11 that also exhibited a large CSP, it was sensitive to purine (A, G) mutation, while it could be a pyrimidine (C, dU) to maintain favorable binding affinity (Fig. 1 E and J). These results implicate that G9, C10, T11, G12, and G13 were involved in the binding site. We further determined the solution NMR structure of sgc8c and identified several key residues, including C7 and G17 engaged in long-range tertiary interactions, and C19·G35 and G20·C34 involved in important base–base stackings (Fig. 3). The gathered information on the binding site and 3D structure serves as a comprehensive guide for us to generate two key strategies for optimizing the functionality of sgc8c. First, sequence modifications aimed at improving the thermostability and binding affinity should focus on the apex of P3, which is remote from the essential site (Fig. 4A). Second, to enhance biostability, elongating P1 rather than directly modifying nucleotides in P1 is recommended, as the backbone of P1 is involved in several important long-range tertiary interactions (Figs. 3A and 4 G and H). As a consequence, a combination approach integrating the thermostable GAA triloop and biostable _L-DNA can surpass the binding affinity of sgc8c and simultaneously increase the thermostability and biostability (Fig. 5). We assert that for DNA aptamers that can retain their original 3D folds after binding to the protein target, this streamlined NMR-based structural investigation represents a valuable approach for comprehending and optimizing DNA aptamer functionality without the need for complex structural information, which is often challenging and time consuming to acquire.

Second, tertiary interactions are commonly observed in RNA molecules, and they serve as crucial chemical forces that govern the folding of higher-order structures. In contrast, DNA molecules are perceived to lack the capacity to form RNA-like tertiary interactions for intricate folding. Recent groundbreaking work revealed an unusually delicate 4WJ fold of a 53-nt Lettuce DNA aptamer through extensive tertiary contacts (42), offering unprecedented insights into the sophisticated structure and function of DNA molecules. Compared to the plain stem–loop (22, 23, 53, 54) and small G-quadruplex structures (24, 55, 56) of DNA aptamers that target proteins (SI Appendix, Fig. S25), the remarkable feature of sgc8c lies in its delicate 3WJ fold stabilized by a hydrogen-bond network via long-range tertiary interactions, notably between G17 and C3, and between C7 and T38/A39 (Fig. 3A and SI Appendix, Fig. S13). Aligned with these structural attributes, mutations of nucleotides engaged in these tertiary interactions perturbed the structure and binding function. In addition, it is imperative to note that contemporary DNA/RNA secondary structure prediction tools primarily rely on evaluating thermodynamic contributions from base–base pairing and stacking interactions (57, 58). The lack of tertiary interactions in these predictions may lead to inaccuracies, particularly for intricate structures such as sgc8c (SI Appendix, Fig. S7).

In summary, this work delineates an intricate 3WJ structure adopted by the 41-nt sgc8c DNA aptamer, in which P1 and P3 form a long helical stack to constitute the scaffold, and P2 protrudes to serve as the primary binding site. Notably, sgc8c adeptly recruits more than 10 nucleotides from different regions to craft its key structural and binding element at the center of the 3WJ. We established the structure–function relationships of sgc8c collectively using 3D structure, NMR CSPs, and site-directed mutagenesis data. The integration of a thermostable GAA triloop at the apex of P3, along with the elongation of P1 by two to three _L-DNA base pairs, can efficiently increase thermostability, improve biostability, and enhance binding affinity to both protein and cell targets. This streamlined NMR-based structural investigation approach will benefit functional optimization of protein-binding DNA aptamers without the need for determining the complex structure. Most importantly, our results highlight a pivotal role of tertiary interactions in shaping the intricate structure and sophisticated function of DNA molecules, providing principles for the structural and functional organization of nucleic acids.

Materials and Methods

Sample Preparation.

DNA oligonucleotides containing xanthosine (X) were purchased from Accurate Biology (Hunan, China), and all the other DNA oligonucleotides were purchased from Sangon Biotech (Shanghai, China) and were of HPLC purification grade. For NMR experiments, the samples were further purified in our laboratory using diethylaminoethyl sephacel anion exchange column and centrifugal desalting. DNA samples were quantified using a NanoDrop microvolume spectrophotometer. The DNA sequences used in this study are shown in SI Appendix, Table S1. The PTK7 protein (His tag) was purchased from Acrobiosystems (Cat #PT7-H52H3, USA).

NMR Spectroscopy.

NMR samples were prepared to contain ~0.2 mM DNA (for 1D experiments) to ~0.5 mM DNA (for 2D experiments) in 10 mM sodium phosphate (NaPi, pH 7) buffer with 5 mM MgCl₂ unless otherwise specified. 1D and 2D imino NMR spectra were collected for sgc8c in a 90% H₂O/10% D₂O solvent using excitation sculpting for water suppression (59). 2D ¹H-¹H NOESY (mixing time of 200 to 300 ms), total correlation spectroscopy (TOCSY), and double-quantum filtered correlation spectroscopy (DQF-COSY) for nonlabile protons were acquired in a 99.96% D₂O solvent. NMR spectra were acquired at 4 °C unless otherwise specified. All NMR data were collected on a Bruker AVANCE 600 MHz spectrometer and analyzed using TopSpin 4.1.3 software.

Structural Calculations.

The DNA structure was calculated by rMD simulations on GROMACS 2021.7 using NOE-derived distance restraints, as well as hydrogen bond distance, hydrogen bond angle, and planarity restraints for Watson–Crick base pairs in 2-41, 3-40, 4-39, 5-38, 6-37, 9-13, 10-12, 18-36, 19-35, and 20-34. For the rMD simulations, the DNA was solvated in a cubic periodic box of TIP3P water with a distance no less than 30 Å between the edge of water box and solute. The DNA was simulated using the OL15 force field (60) in 10 mM Na⁺ and 5 mM Mg²⁺. The NOE-derived distance restraints and hydrogen bond restraints were employed during energy minimization, heating, equilibration, annealing, and production. In brief, after energy minimization, the system was linearly heated from 0 to 278.15 K in 1 ns under 1 atm pressure and then continually equilibrated for 1 ns, followed by a 10-ns periodic annealing in which the system was linearly heated to 353.15 K in 500 ps, kept for 500 ns, cooled to 278.15 K in 500 ps and held for 500 ps, with the thermal cycle repeated five times. The production was performed at 278.15 K and 1 atm for 1 µs. Then, DNA structures in the production trajectory were clustered using the GROMOS method (61), and the centroid structures from the 10 maximum clusters were selected as the initial structures. For each initial structure, rMD simulations were performed 10 times using the aforementioned parameters with a 10-ns production process, and the last structure was extracted from each trajectory. A total of 100 structures were clustered into 10 groups using the GROMOS method. Then, the 10 centroid structures were selected and subjected to restrained energy minimization using AMBER20 to serve as the final representative structures (SI Appendix, Fig. S10). The NMR restraints and refinement statistics are shown in SI Appendix, Table S2. 3D structural figures were prepared using PyMOL (62) unless otherwise stated.

The spatial distribution of Mg²⁺ ions to sgc8c was calculated by rMD simulations in a manner similar to that used for the structural calculations mentioned above. NOE-derived distance restraints, hydrogen bond distance, hydrogen bond angle, and planarity restraints for Watson–Crick base pairs were employed. The concentration of Mg²⁺ was set to be 30 mM for efficient spatial distribution sampling. After energy minimization, a 200-ps heat process from 0 K to 278.15 K, a 300-ps equilibrium, and a 100-ns production simulation were conducted. The trajectories of sgc8c were aligned using least squares fitting, and the Mg²⁺ ions were shown as points every 4 ps. The distribution of Mg²⁺ was analyzed and plotted using the Visual Molecular Dynamics (VMD) program (SI Appendix, Fig. S11).

Optical Melting Experiment.

Optical melting experiments were conducted on a Chirascan V100 circular dichroism spectropolarimeter. The samples were prepared to contain 5 μM DNA in 10 mM NaPi (pH 7) with 5 mM MgCl₂, and the spectra were acquired using a 0.5 mm path length quartz cuvette with a 1 nm bandwidth at 25 °C. The blank correction was made by subtracting the buffer spectrum. The UV absorbance at 260 nm (A₂₆₀) was recorded from 20 to 80 °C at a heating rate of 1 °C/min. The thermodynamic parameters, including T_m, changes in entropy, enthalpy, and Gibbs free energy, were determined by fitting the A₂₆₀-temperature melting curve with a two-state transition model (63). The thermodynamic parameters are summarized in SI Appendix, Table S3.

SPR Experiment.

The K_D of DNA aptamer–PTK7 interaction was determined using a Biacore-8 K SPR instrument (Cytiva) at 25 °C. 3 μg of PTK7 (His tag) was immobilized onto a CM5 sensor chip cell by amine coupling procedures, and the other cell was coupled with His tag as a control. Aptamers at various concentrations in running buffer (DPBS supplemented with 5 mM MgCl₂ and 0.01% Tween 20) were injected into the analyte channel at a flow rate of 30 μL/min, a contact time of 120 s and a dissociation time of 300 to 600 s. After each analysis cycle, the analyte channels were regenerated via a 30-s injection of 1.5 M NaCl. The K_D of the aptamer–PTK7 interaction was obtained by fitting the sensorgram with a 1:1 binding model (64) using Biacore evaluation software (GE Healthcare). The K_D values are summarized in SI Appendix, Table S4.

Biostability Assay.

To evaluate the nuclease resistance of the DNA aptamers, 500 nM FAM-labeled DNA aptamers were incubated with a mixture of Exonuclease I (0.15 U/μL) and Exonuclease III (0.5 U/μL) in DPBS buffer supplemented with 5 mM MgCl₂ at 37 °C for 15 min, heated at 95 °C for 10 min to inhibit exonuclease activity, and then stored at 4 °C. To evaluate the biostability of the DNA aptamers in serum, 500 nM FAM-labeled DNA aptamers were incubated in RPMI-1640 medium supplemented with 10% FBS and 5 mM MgCl₂ at 37 °C for different time intervals (0, 1, 2, 6, 12, 24, 36, and 48 h) and then stored at −20 °C. The integrity of the DNA aptamers was resolved by 10% denaturing PAGE and visualized using Amersham ImageQuant 800.

Cell Culture and Flow Cytometry.

CCRF-CEM cells were cultured in RPMI-1640 medium supplemented with penicillin (60 mg/L), streptomycin (100 mg/L), and 10% FBS at 37 °C in a 5% CO₂ incubator. To evaluate the binding of the sgc8c mutants to CCRF-CEM cells, the cells were washed with DPBS and then incubated with 200 nM aptamer in binding buffer (DPBS supplemented with 4.5 mg/mL glucose, 1 mg/mL BSA, 0.1 mg/mL HS-DNA, and 5 mM MgCl₂) at 4 °C for 30 min. The flow cytometry assays were performed and the results were analyzed using a Beckman CytoFLEX LX system.

To determine the K_D of the aptamer–cell interaction for sgc8c, GAA_25-28, GAA_25-29, L1-3, L1-4, L+2bp, L+3bp, sgc8c-x1, and sgc8c-x2, CCRF-CEM cells were washed with DPBS and then incubated with various concentrations of the aptamer in the binding buffer mentioned above at 37 °C for 30 min. The K_D was determined by fitting the curve of the fluorescence-aptamer concentration using the equation Y = B_max·X/(K_D+X) (28). The K_D values of the aptamer–cell interactions are summarized in SI Appendix, Table S5.

Statistical Analysis.

Statistical analyses were performed using GraphPad Prism 9.0. The statistical data are presented as means ± SD by three replicate experiments.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The atomic coordinates for the solution NMR structures of sgc8c have been deposited to the Protein Data Bank (PDB ID: 8Y0F) (65). All study data are included in the article and/or SI Appendix.