High-Affinity Host–Guest Recognition for Efficient Assembly and Enzymatic Responsiveness of DNA Nanostructures

Abstract

The combination of multiple orthogonal interactions enables hierarchical complexity in self-assembled nanoscale materials. Here, efficient supramolecular polymerization of DNA origami nanostructures is demonstrated using a multivalent display of small molecule host–guest interactions. Modification of DNA strands with cucurbit[7]uril (CB[7]) and its adamantane guest, yielding a supramolecular complex with an affinity of order 10¹⁰ M⁻¹, directs hierarchical assembly of origami monomers into 1D nanofibers. This affinity regime enables efficient polymerization; a lower-affinity β-cyclodextrin–adamantane complex does not promote extended structures at a similar valency. Finally, the utility of the high-affinity CB[7]–adamantane interactions is exploited to enable responsive enzymatic actuation of origami nanofibers assembled using peptide linkers. This work demonstrates the power of high-affinity CB[7]–guest recognition as an orthogonal axis to drive self-assembly in DNA nanotechnology.

1. Introduction

Self-assembly in nature arises from the combination of multiple different noncovalent intermolecular forces. Inspired by the functional diversity of these systems, nanoscale materials have sought to integrate different orthogonal interactions. Over the past 30 years, the field of DNA nanotechnology has yielded a suite of programmable structures driven by Watson–Crick pairing.^[1] The synthesis of 1D nanofibers has been a particularly active area of research, due to their diverse applications as structural units for hierarchical materials.^[2] Recently, orthogonal noncovalent interactions have been introduced beyond the canonical forces underlying DNA structural organization, including self-assembling peptides,^[3] proteins,^[4] hydrophobic packing,^[5] and “peg in hole” base stacking.^[6] These hybrid structures introduce the added control of different assembly “modes” with orthogonal molecular triggers or leverage interactions beyond DNA hybridization for the incorporation of other species.

The supramolecular recognition of a macrocycle host binding a small-molecule guest has not been extensively explored in DNA nanotechnology. In a recent report, the Walther group demonstrated the self-assembly of DNA origami cuboids into 1D fibers through the interaction of β-cyclodextrin (βCD) macrocycles with adamantane (Ad) guests; efficient assembly required between 18 and 36 host–guest complexes per origami.^[7] Although this elegant work showed the potential of host–guest motifs in DNA nanotechnology, the high valency of necessary interactions may limit certain applications of this system. Among host–guest motifs, the family of cucurbit[n]uril (CB[n]) macrocycles have particular utility for efficient recognition compared to cyclodextrin.^[8] Within this family, cucurbit[7]uril (CB[7]) is especially promising for its solubility in water and exceptionally high binding affinity, with K_eq of the order of 10¹⁷ M⁻¹ for certain guests.^[9] For comparison, among the best interactions for βCD is its binding to an Ad guest, with a K_eq of order 10⁴ M⁻¹.^[10] The dimensions and cavity volumes of βCD and CB[7] macrocycles are nearly identical.^[11] Accordingly, the significantly enhanced affinity afforded by CB[7] expands the supramolecular design tool set, especially with the demonstration of routes to modify CB[7] with functional handles for inclusion on materials.^[12] In recent years, supramolecular chemistry has found extensive use in materials design.^[13] The Seitz group demonstrated the distance-dependent effect of multivalency in CB[7]–Ad binding using DNA “rulers.”^[14] In addition, Jayawickramarajah and co-workers have demonstrated the use of host–guest chemistry, including both cyclodextrin and cucurbituril, for the synthesis of complex DNA-based materials,^[15] whereas Tan and co-workers have used these interactions for theranostic applications.^[16] Building off our work with heterodimeric coiled-coil peptides,^[] here we used a high-affinity CB[7]–Ad complex to drive 1D fiber assembly requiring only eight interactions per origami. At comparable valency, limited short oligomers were observed with βCD–Ad. The CB[7]–Ad motif thus provides an efficient orthogonal interaction for integration with DNA nanotechnology to enable hierarchical assembly through lightly modified programmable interfaces. This recognition motif may likewise be useful to display prosthetic moieties such as proteins, peptides, or nanoparticles on DNA nanostructures. Toward this goal, we also demonstrated the use of CB[7]–Ad recognition to incorporate enzymatically responsive peptide “linkers” to form a DNA origami nanofiber that can be depolymerized by exposure to matrix metalloproteinase-8 (MMP-8).

Our design modifies two single stranded (ss) DNA handles (termed DNA1 and DNA2) with CB[7] or Ad at their 5’ end. At a 1:1 ratio, Ad–DNA1 and CB[7]–DNA2 will form a heterodimer with the supramolecular host–guest motif linking the two oligonucleotides (Figure 1A). This heterodimer can then assemble DNA origami nanostructures bearing complementary handles to yield extended 1D nanofiber arrays (Figure 1B). In this design, the origami faces serve as programmable molecular pegboards, displaying a multivalent pattern of matched host–guest complexes, similar to cooperative interactions between amino acids on protein–protein interfaces. Compared with sticky ends, interactions based on CB[7]–guest recognition have the added advantages of tunable interaction affinities spanning ≈15 orders of magnitude.^[8b,9] As a model DNA origami nanostructure, we used a cuboid with dimensions 32 × 19.5 × 16 nm,^[2a,3e] allowing for precise control of both the number and spatial distribution of complementary handles. For design of the origami cuboids, and location of the handles, see Sections S6 and Figure S5 (Supporting Information).

Figure 1.

Self-assembly of DNA nanostructures using host–guest interactions. A) DNA strands modified with small molecule adamantane or cucurbit[7]uril moieties self-assemble by host–guest interactions. B) DNA origami cuboids bearing complementary handles assemble into long, 1D nanofibers by multiple interfacial host–guest complexes.

2. Results and Discussion

DNA1 (10 nt) and DNA2 (21 nt) bearing a 5’-thiol were linked to adamantane via a maleimide–Ad conjugate. Separately, CB[7]–N₃ was synthesized by reported methods^[12a] and linked to DNA1 or DNA2 via strain-promoted azide–alkyne cycloaddition (SPAAC) with a 5’-dibenzocyclooctyne (DBCO)-functionalized oligonucleotide. For comparison, we conjugated βCD–N₃ to DNA1 via SPAAC. The structures of the DNA conjugates used are shown in Figure 2A; for synthesis, purification, and characterization, see Section S4 (Supporting Information). Next, we probed the heterocomplex formation of the conjugates using native polyacrylamide gel electrophoresis (Figure 2B). Compared with the individual DNA conjugates (lanes 2 and 3), a 1:1 mixture of Ad–DNA1 and βCD–DNA1 (1 μm each) did not show a band shift indicative of complex formation (lane 4); bands corresponding to the individual conjugates remained, suggesting this concentration is below the effective K_d for the βCD–Ad complex. Comparatively, Ad–DNA1 and CB[7]–DNA1 (lanes 6 and 7) showed almost complete shift to a higher-molecular-weight species (lane 8, yellow arrow) indicative of stable complex formation resulting from the high-affinity CB[7]–Ad interaction. To quantify the relative binding affinity between the Ad–DNA1 and CB[7]–DNA1 complex (cmplx), competition studies using 1-hydroxyadamantane (Ad-OH, a K_eq of 2.3 × 10¹⁰ m⁻¹ with CB[7])^[17] were performed. Exposing cmplx to increasing molar equivalents of Ad-OH decreased the intensity of the high molecular weight band, with a concomitant increase in bands for Ad–DNA1 and CB[7]–DNA1 (Figure 2C). The reduction in cmplx as a function of Ad-OH “inhibitor” concentration was fit to a standard 3-parameter least squares regression, and the concentration ratio at the IC₅₀ was multiplied by the known affinity of Ad-OH to approximate the relative association constant (K_eq,rel) of 1.3 × 10¹⁰ m⁻¹ for binding of Ad–DNA1 to CB[7]–DNA1 (Figure 2D). This value is of the same order as values realized using competition NMR for a similar amide-linked Ad in binding to CB[7],^[18] indicating no dramatic reduction in CB[7]–Ad recognition due to DNA conjugation of both host and guest,^[19] and thereby validating the use of this high-affinity interaction with DNA nanotechnology.

Figure 2.

Characterization of host–guest complexation. A) Structures of Ad–DNA, CB[7]–DNA, and βCD–DNA. B) Native polyacrymalmide gel electrophoresis (PAGE) analysis of host–guest complexation. DNA strands are depicted as single-stranded, but had their complement added prior to PAGE to form double stranded (ds) DNA and enhance staining. Lane M: dsDNA ladder (bp); 1,5: DNA1; 2,6: Ad–DNA1; 3: βCD–DNA1; 4: Ad–DNA1 + βCD–DNA1; 7: CB[7]–DNA1; 8: Ad–DNA1 + CB[7]–DNA1. C) Native-PAGE competition experiment between inhibitor (Ad–OH) and the Ad–DNA1 + CB[7]–DNA1 complex (cmplx). Lane M: dsDNA ladder (bp). Ad: Ad–DNA1; CB: CB[7]–DNA1; cmplx: Ad–DNA1 + CB[7]–DNA1; subsequent lanes: cmplx + indicated equivalents of Ad-OH. Both the PAGE gels were run at a constant temperature of 10 °C and a constant voltage of 200 V. D) Plot of cmplx remaining as a function of Ad-OH inhibitor added.

We next turned to assembling the DNA origami cuboids and attached Ad to DNA2 to avoid competing with the two host-modified DNA1 conjugates. We first investigated a “one pot” annealing protocol, whereby all the components (i.e., the M13 scaffold, staple strands, staples bearing handles, and the host or guest DNA conjugates) were mixed in a single tube and annealed from 65 to 4 °C over 40 h (Figure 3A). Given the high-affinity interaction between CB[7] and Ad, we presume this recognition is minimally impacted by elevated temperatures. Accordingly, the CB[7]–Ad complex forms first under these conditions, followed by assembly of the core origami structure (T_m ≈ 55 °C), and finally hybridization to the cuboids of the DNA handles (T_m ≈ 40–45 °C).

Figure 3.

One-pot assembly of DNA origami nanofibers. A) Protocol for one-pot annealing; colored squares on origami indicate the location of Ad (pink), CB[7] (blue), or βCD (green). B) Agarose gel electrophoresis (AGE) of cuboid assembly. The red arrow indicates free staples. Lane 1: unmodified cuboids; 2: cuboids + Ad–DNA2; 3: cuboids + CB[7]–DNA 1; 4,6: cuboids + Ad–DNA2 + CB[7]–DNA 1; 5: cuboids + Ad–DNA2 + βCD–DNA1; 7: cuboids + Ad–DNA1 + CB[7]–DNA2; 8,9: cuboids with poly(T) handles with Ad–DNA2 + CB[7]–DNA1 (lane 8) or Ad–DNA1 + CB[7]–DNA2 (lane 9). C,D) Negative stain TEM images of samples in lanes 4 and 5, respectively. E,F) Histograms of array length by mass fraction (bars) and cumulative fraction (lines) of samples in lanes 4 and 5, respectively.

We monitored cuboid assembly by agarose gel electrophoresis (AGE), as well as by negative stain transmission electron microscopy (TEM). By AGE, unmodified cuboids showed a distinct band for the nanostructure, along with a large higher-mobility band for excess staple strands (Figure 3B, lane 1). Cuboids bearing only CB[7]–DNA1 or Ad–DNA2 (eight handles) also showed only monomer bands (lanes 2 and 3). However, cuboids with eight handles on both ends showed an aggregated band in the loading well, indicating the formation of large structures (lane 4). By contrast, cuboids with Ad–DNA2 and βCD–DNA1 (eight handles) lacked this aggregate band and were primarily monomers or very short oligomers (lane 5). Control experiments where the handles on Ad and CB[7] were swapped still yielded aggregates (Figure 3B, lane 7), whereas co-assembly with origami bearing mismatched poly( T) handles gave only short oligomers due to blunt-end stacking (Figure 3B, lanes 8 and 9). Studies with 1–7 handles per side yielded shorter fibers (Figure S6, Supporting Information), so all further experiments employed eight handles. We analyzed the structures prepared with Ad–DNA2 and CB[7]–DNA1 (lane 4) or βCD–DNA1 (lane 5) by TEM, and found that origami assembled with the CB[7]–Ad interaction formed long, 1D assemblies linked at the interfaces bearing host–guest motifs (Figure 3C). Fitting the length distribution of these fibers (Figure 3E) resulted in an average extent of polymerization (${\overline{X}}_n$) of 7.9 ± 3.3 monomers. Some fibers surpassed 20–30 monomers, with the longest observed measuring 45 monomers and ≈2 μm in length. By contrast, origami assembled with the βCD–Ad interaction yielded primarily monomers and the rare short oligomer ($\overline{X_{\text{n}}}=1.2\pm 1.9$), consistent with the Walther group’s work, which showed minimal 1D assembly when nine handles were used.^[7] Taken together, our results demonstrate the advantage of high-affinity CB[7]–Ad motifs as efficient interactions to direct assembly of DNA cuboids, compared with the similarly sized βCD–Ad recognition motif with affinity ≈6 orders of magnitude lower.

In our prior work using coiled-coil peptides,^[3e] the modularity of disparate supramolecular modes enabled hierarchical assembly, whereby origami formed by a primary annealing step could be subsequently assembled into fibers through a second, lower-temperature incubation. Thus, we next probed two alternate assembly pathways to optimize formation of 1D arrays directed by CB[7]–Ad motifs: 1) separately forming origami cuboids with Ad or CB[7] on both sides, and then co-assembling them into an “alternating copolymer” (Figure 4A); and 2) purifying origami cuboids bearing complementary DNA handles, and then assembling them with preformed DNA1–Ad/CB[7]–DNA2 complex (Figure 4E). For both routes, the second annealing was conducted at 45–40 °C over 12 h, followed by rapid cooling to 4 °C. Analysis of the “copolymer” route (pathway 1) by both AGE (Figure 4B) and TEM (Figure 4C,D) revealed the formation of 1D nanofibers. The fibers are morphologically similar to those of the one-pot system: long and straight, yet somewhat shorter (${\overline{X}}_n=5.7\pm 2.4$ monomers). By contrast, purified cuboids combined with preformed DNA1–Ad/CB[7]–DNA2 complex (pathway 2) showed dramatically longer fibers by TEM (Figure 4G,H), with ${\overline{X}}_n=21.0\pm 2.5$ monomers, and the longest observed fiber reaching 72 cuboids (≈3.3 μm) in length. Interestingly, these results parallel those obtained from our work with peptide heterodimers,^[3e] with the sequential assembly of purified cuboids giving the longest fibers. The similarity in respective length distributions suggests a universality in DNA cuboid self-assembly arising from disparate motifs—coiled-coil assembly versus host–guest binding—with each capable of high-affinity (i.e., sub-nanomolar) interactions.

Figure 4.

Hierarchical assembly of nanofibers. A) Protocol for copolymer formation (pathway 1); colored squares on origami indicate the location of Ad (pink) or CB[7] (blue). B) AGE of cuboid assembly. Lane M: dsDNA ladder (kbp); 1,2: cuboids with Ad–DNA1 or CB[7]–DNA1 on both sides, respectively; 3,4: cuboids with Ad–DNA2 or CB[7]–DNA2 on both sides, respectively; 5: lane 1 + lane 4, after second anneal; 6: lane 2 + lane 3, after second anneal. C) Negative stain TEM images of lane 6 fibers. D) Histograms of array length by mass fraction (bars) and cumulative fraction (lines) for lane 6 fibers. E) Protocol for assembly of purified origami by preformed host–guest complex with DNA handles (pathway 2). F) AGE of cuboid assembly. Lane M: dsDNA ladder (kbp); 1: unpurified cuboids; 2: purified cuboids; 3: purified cuboids + DNA2–Ad/CB[7]–DNA1 complex, after second anneal. G) Negative stain TEM images of fibers. H) Histograms of array length by mass fraction (bars) and cumulative fraction (lines) for fibers.

One advantage of molecular recognition using CB[7]–Ad is the easier incorporation of Ad into some synthetic platforms compared with a DNA handle. In our experience, it can be challenging to attach two oligonucleotide handles to a short peptide—due both to the electrostatic repulsion between the two DNA strands, and the need for subsequent purification—although some examples with DNA^[19] do exist. As an alternative approach, we reasoned that by synthesizing a peptide with only one DNA handle, but with Ad at the other end, it should be possible to use that peptide as a “linker” to DNA cuboids bearing a CB[7]–DNA conjugate. Adamantane can be synthetically incorporated directly during solid phase peptide synthesis through amide bond formation, using selective deprotection of a 4-methyltrityl (Mtt)-lysine residue on-resin, followed by amide coupling of the revealed ϵ-amine to adamantane carboxylic acid. While we note that two peptide nucleic acid (PNA) handles can be attached to peptides on resin, as outlined by a number of reports,^[20] the use of adamantane provides a simpler and more economical alternative for subsequent interaction with the second DNA handle using host–guest supramolecular association.

Toward this aim, we synthesized a peptide with the sequence [K(Ad)]–GSGPQGIWGQGSG–[azK] (which we term “Ad–(peptide)–azK”), where [K(Ad)] denotes the lysine–adamantane modification and azK denotes the noncanonical amino acid azidolysine. This azidolysine allows an oligonucleotide handle to be conjugated to the C-terminus of this peptide via SPAAC, as previously described,^[2a] to yield the conjugate Ad–(peptide)–DNA1. For the structure, synthesis, and characterization of Ad–(peptide)–azK and Ad–(peptide)–DNA1, see Section S4 and Figure S13 (Supporting Information). This peptide contains a sequence (GPQG↓IWGQ) that is cleavable by MMP-8, an enzyme overexpressed in the tumor microenvironment.^[21] Thus, a DNA origami cuboid bearing handles complementary to DNA2 on one side and handles complementary to DNA1 on the other can be assembled into nanofibers by the DNA1–CB[7]/Ad–(peptide)–DNA2 host–guest complex; addition of MMP-8 should degrade these fibers back to the constituent monomers (Figure 5A). As a negative control, we synthesized a peptide with the same amino acid composition, but where the order of the residues was scrambled in order to abrogate MMP-8 cleavage (Ad–(scram)–azK);^[22] the oligonucleotide conjugate of this peptide is Ad–(scramble)–DNA2. Fibers linked by this control molecule should not degrade upon MMP-8 exposure. The Ad–(peptide)–DNA2 is depicted schematically (to scale) in Figure 5B.

Figure 5.

Enzymatic responsive DNA nanofibers. A) Protocol for peptide-incorporated nanofiber formation; colored squares on origami indicate the location of Ad (pink) or CB[7] (blue). B) Schematic (to scale) of Ad–(peptide)–DNA2 conjugate; the boxed region denotes the Ad moiety attached to a terminal lysine residue of the peptide. C) AGE of cuboid assembly. Lane M: dsDNA ladder (kbp); 1, cuboids with handles DNA1* and DNA2* on either side; 2: lane 1+ 2 equiv. preannealed Ad–(peptide)–DNA2 + DNA1–CB[7]; 3: lane 2 + activated MMP-8 protein; 4: lane1 + 2 equiv. preannealed Ad–(scramble)–DNA2 + DNA1–CB[7]; 5: lane 4 + activated MMP-8 protein. D) Negative stain TEM images of lane 2 fibers. E) Negative stain TEM images of lane 3 cleaved fibers. F) Negative stain TEM images of lane 4 fibers. G) Negative stain TEM images of lane 5 noncleaved fibers. H) Histograms of array length by mass fraction (bars) and cumulative fraction (lines) for lane 2 fibers. I) Histograms of array length by mass fraction (bars) and cumulative fraction (lines) for lane 3 sample. J) Histograms of array length by mass fraction (bars) and cumulative fraction (lines) for lane 4 scramble fibers. K) Histograms of array length by mass fraction (bars) and cumulative fraction (lines) for lane 5 noncleavable scramble DNA fibers.

Cuboids bearing eight ssDNA handles with sequence DNA1* on one side, and eight ssDNA handles with sequence DNA2* (i.e., complementary to the oligonucleotide on Ad–(peptide)–DNA2) were annealed and purified away from excess staples and peptides. To this sample, we added two equivalents of preannealed Ad–(peptide)–DNA2 + DNA1–CB[7] and annealed the sample from 45 to 40 °C over 12 h to promote nanofiber formation. By AGE, cuboids incubated with DNA1–CB[7] and either Ad–(peptide)–DNA2 or Ad–(scramble)–DNA2 both showed high molecular weight species in the loading well (Figure 5C, lanes 2 and 4). TEM analysis showed long nanofibers for both of these samples (Figure 5D,G), with ${\overline{X}}_n=39.9\pm 1.9$ and 38.7 ± 1.7 monomers, respectively (Figure 5H,J). Intriguingly, the use of the peptide as the linker gave fibers almost twice as long, on average, as the best system using CB[7]–Ad interactions without the peptide (Figure 4H). It is possible that the added flexibility of the peptide linker facilitated interaction with the origami more readily, or perhaps the added length mitigated their electrostatic repulsion. We next incubated these nanofibers with MMP-8 (12 h at 37 °C in assay buffer) and analyzed them again by both AGE and TEM. The fibers assembled using the scrambled peptide did not show any degradation by either method, with high-molecular-weight bands in the loading well (Figure 5C, lane 5), and long fibers seen by TEM (Figure 5F). The fibers assembled using the cleavable peptide, however, showed an almost complete transition to higher-mobility bands (at roughly the same retention as the cuboid monomers; Figure 5C, lane 3), and the TEM showed only scattered monomers and short oligomers (Figure 5E), with a length distribution of ${\overline{X}}_n=1.9\pm 2.1$ (Figure 5I). These results confirmed that not only could the Ad–(peptide)–DNA2 conjugate assemble cuboids (through a combination of DNA hybridization and host–guest interactions), but that the peptide remained accessible to enzymes for cleavage. The scrambled peptide control confirmed that it was the enzymatic activity (and not, for example, the buffer conditions) that mediated this effect, as long nanofibers were seen both before and after exposure to the enzyme.

Taken together, our results suggest that a peptide bearing a DNA handle and an Ad moiety can form a complex with CB[7]–DNA, and effectively bridge structures bearing complementary handles to the host–guest complex. In effect, we use the CB[7]–Ad host–guest interaction as the second “bioconjugation reaction” to attach a second DNA handle to the peptide, in a way that is challenging to do with covalent chemistry. Furthermore, the host–guest interaction is reversible, so this second interaction could, in principle, be broken by addition of free Ad guest. We thus see the approach outlined here as an alternative strategy for implementing peptide linkers and latches for DNA nanostructures; coupled with the enzymatic responsiveness of many peptides, this should afford novel stimulus-responsive peptide–DNA nano- and biomaterials.

3. Conclusion

In conclusion, we have demonstrated a high-affinity CB[7]–Ad recognition motif as an efficient and effective orthogonal interaction with promise for the construction of DNA nanostructures. First, we demonstrated the use of these interactions for the efficient hierarchical 1D assembly of DNA nanostructure to form micrometer-length supramolecular polymers. Compared to prior reports using host–guest recognition of Ad by similarly sized βCD macrocycles, with distributions of 1D arrays of ≈12 cuboids in length when using 36 handles,^[7] the optimized assembly pathway revealed for CB[7]–Ad generates significantly longer 1D arrays while requiring only eight handles on each interface. We then demonstrated that the use of these interactions could be incorporated into a peptide–DNA “linker” to create enzymatically degradable DNA origami nanofibers. Although we used nanofiber degradation as a proof of principle, this approach for peptide linking of DNA structures could be translated to enzymatically responsive “latches” or “staples” for DNA nanocages, e.g., to open them and release a cargo in the tumor microenvironment.

A key advantage of the CB[7]–Ad motif is the ease of incorporation into synthetic molecules like peptides. Although we demonstrated a peptide linker consisting of both an adamantane moiety and a DNA handle, peptides bearing a single adamantane alone should readily decorate DNA nanostructures bearing CB[7] “docking” sites, and provide an attractive alternative to peptide–DNA chemical conjugation altogether. Encapsulating hydrophobic drug molecules, or guest-modified prodrugs, into these CB[7] moieties and anchoring them into nanovehicles could afford an approach for precise loading of a desired therapeutic onto DNA nanostructures. Thus, CB[7]–Ad interactions hold promise as orthogonal interactions in parallel with Watson–Crick base pairing for incorporating molecules into DNA nanostructures without having to conjugate them to DNA handles first.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.