Orderly Entangled Nanostructures of Metal–Peptide Strands

Abstract

Construction of entangled nanostructures from molecular rings or strands has long attracted chemists, yet synthetic approaches for highly entangled nanostructures remain unexplored to date. Here, we introduce our recent achievements in construction of such nanostructures by utilization of metal–peptide strands. Our folding-and-assembly strategy, that is based on a cooperative processes of peptide self-folding and metal-induced self-assembly, has afforded unprecedented topological nanostructures through threading of multiple metal–peptide rings. Starting from the initial design of the system, we discuss remarkable examples such as polyhedral links, torus knots, and a poly[n]catenane, and state the perspectives in this account review.

Topology, Self-assembly, Folding

1. Introduction

Orderly entangled structures are ubiquitous in macro-scale fabrications, yet uncommon in nanoscale. Current typical examples would be limited to DNA nanoarchitectures,¹ which are based on originally entwined duplexes. To construct a wide variety of entangled nanostructures efficiently, development of a synthetic method is desired. When looking at natural protein structures for a hint, an ultimate example exists. A virus capsid structure is constructed by topological linkages of 72 protein rings in a chain mail manner.² This spectacular example inspired us to use peptidic building blocks, yet in fact multiple entanglements are rarely seen in natural protein structures. A recent study reported the rate of knotted protein structures was around 6% and their crossing numbers, degree of the topological complexity, were no more than six.³ In other words, only simple entanglements exist for protein nanostructures. Since spontaneous formation of knotted topologies from a long polypeptide strand generally requires large entropy costs, highly entangled topological structures are unlikely to form. Thus, even for polypeptide strands, an excellent biological building block, we considered an artificial trick was necessary for the purpose of constructing advanced entangled topologies.

In the last decade, we have incorporated peptide folding phenomena into studies on metal-directed self-assembly. We have explored nanostructures through complexation of short peptide ligands and metal ions (Figure 1a). The fundamental purpose is to create a new concept that is a fusion of self-folding⁴ and self-assembly fields.^5–7 Respective fields have well developed, yet their interface has been little studied. During the course of such studies, called a folding-and-assembly (F&A) strategy,⁸ we have found the emergence of unique topological nanostructures, that are not accessible by previous synthetic methods⁹ (Figure 1b). Before introducing the examples in detail, formation of metal–peptide entanglements can be concisely described as follows. Coordination bonds direct a metal–peptide strand into a macrocyclic structure while each peptide region is folded into a specific conformation programmed by the sequence. In this process, reversibility of the coordination bond enables threading of strands/macrocycles (described as “magic rings”¹⁰) and transiently folded conformation of strands fit each other through cooperative non-covalent interactions, reminiscent of jigsaw puzzle pieces. As a result, a multiply entangled structure forms as a thermodynamically stable nanostructure even overcoming the considerable entropy costs. Thus, simple mixing of metal ions and short peptide ligands can afford a variety of topological nanostructures according to the peptide sequence design. In this account review, we start from the initial ideas, introduce recent remarkable examples, and summarize with the future perspectives on highly entangled metal–peptide nanostructures.

Figure 1.

(a) General scheme for a metal–peptide nanostructure through a folding-and-assembly process and (b) highly entangled topologies described in this review.

2. Initial Idea and Design

The initial attempt was to test a short peptide as an organic ligand of metal-directed self-assembly. Short peptides are unable to show stable self-folding, and at the same time, such flexible building blocks were rarely seen in metal-directed self-assembly, where rigid organic ligands are generally used for defining length and directions.^6,7 In fact, in porous crystals fields, peptide building blocks were restricted almost exclusively to tiny dipeptide fragments in earlier work.¹¹ Conversely, we expected a well-defined nanostructure would be obtained if concomitant processes of folding and self-assembly were to work well. Another hint came from tandem repeats often found in protein structures.¹² Nature efficiently constructs three-dimensional nanostructures by repeating the peptide secondary structures. Likewise, we expected the facile formation of a pseudo polypeptide strand if a peptide fragment were to be accurately polymerized through metal crosslinking (Figure 2a). Based on these ideas, the combination of the GPP sequence from the collagen sequence unit and linear-coordinating Ag⁺ ion was designed (Figure 2b). Through complexation of the ditopic tripeptide ligand of the Gly-Pro-Pro sequence 1 (Gly: glycine, Pro: l-proline) and Ag⁺ ions, a crystalline coordination network was obtained (Figure 2c).¹³ As the coordinative part on both peptide termini, 3-pyridyl (py) groups were respectively introduced through amide bond linkages. We consider the reversibility of the Ag–py bond is suitable for cooperation with transient short peptide folding. We chose 3-py instead of 4-py groups because metal coordination through 3-py groups could allow various angles. We expected such structural flexibility expanded possible conformations of metal–peptide strands, which might increase the successful rate of nanostructure formation in the F&A system. Indeed, successful examples of 3-py-based ligands are summarized in this account review.

Figure 2.

(a) Design of a metal–peptide strand, (b) complexation scheme of Ag⁺ ions and tripeptide ligand 1, (c) the entangled network structure, and (d) illustration of entanglements.

To obtain a collagen-mimic coordination network, we carried out the complexation as a crystallization process. Single crystals were formed by the layer diffusion of AgBF₄ in water and tripeptide ligand 1 in ethanol. Crystallographic study confirmed a F&A nanostructure, in which the short peptide region of 1 was folded into the polyproline II helix (P_II-helix) conformation¹⁴ and sequentially connected by Ag⁺ ions. Ligand 1 has a directionality of the N→C termini, yet all of 1 in the crystal arranged in the head-to-tail manner via Ag⁺ crosslinking. Each polymeric metal–peptide strand entangled with each other, which resulted in a honeycomb network (Figure 2d). The huge 2.2 nm-sized pore was derived from a hexagonal entanglement of six metal–peptide strands, which exhibited chiral recognition properties for inclusion of BINOL and oligosaccharides. Such a hexagonal entwining behavior was also observed in a 2D material made of oligoproline building blocks.¹⁵

When silver bis(trifluoromethanesulfonyl)imide (AgNTf₂) was used for the complexation, an analogous tetragonal network was obtained.¹⁶ The P_II-helical scaffold was the same, but in this case, packing of NTf₂⁻ ions induced not hexagonal but tetragonal entanglement (Figure 3a). The 1.5 nm-sized pore could be modified with functional groups or reduced to a lower symmetry, which revealed both fidelity and flexibility of the P_II-helix conformation. Deformation of the peptide conformation by dimethyl substitution of the Gly Hα (Aib-Pro-Pro sequenced ligand 2 in Figure 3b; Aib: 2-aminoisobutyric acid) afforded a narrower pore, which was more suitable for trapping a guest and inducing a reaction.¹⁷ For example, methyl pyruvate was converted to the unstable hemiacetal within the pore as a chiral form, which was clearly observed by X-ray diffraction study (Figure 3c). Thus, functional porous crystals could be constructed by three-dimensional entanglements here. In the field of porous crystals, attention has been increasing for using metal ions and short peptides.¹⁸

Figure 3.

(a) A tetragonal network structure constructed by AgNTf₂ and tripeptide ligand 1. (b) Structure of ligand 2 and (c) chiral hemiacetal formation observed within a pore of Ag•2 strands.

3. Discrete Topological Molecules

3.1 [4]₁₂-Catenane

We next attempted construction of a discrete molecular structure. For this purpose, we believed utilization of a loop or turn conformation would be important as seen in natural globular protein structures.¹⁹ Contrary to our initial expectation that finding a suitable sequence would take considerable efforts, we would soon discover one. When tripeptide ligand 3 of the Pro-Gly-Pro sequence, the sequence isomer of 1, was mixed with AgBF₄ in aqueous methanol, the Ag₁₂(3)₁₂ spherical complex (molecular weight: 7742) was easily crystallized²⁰ (Figure 4ab). The molecular structure had a 12-crossing [4]catenane topology (hereafter, called as [4]₁₂-catenane 4) although its topological analysis by using knot theory²¹ was very difficult (Figure 4c). To consider a topology here, connections of metal–peptide arrays are replaced with flexible strings, which are allowed to deform freely except for detachment. The topological complexity is described by using a crossing number, which means the minimum number of crossing points in the two-dimensional plane where the topology or any deformed topologies is projected (Figure 4d).²¹ The ring threading manner in 4 was analyzed many times using flexible wires, and we confirmed it as a 12-crossing number topology at last. In the field of chemical topology, unique topologies as represented by a trefoil knot²² (a single ring with 3-crossings) and Borromean rings²³ (3 rings with 6-crossings) have been synthesized, yet a complex topology with more than 10-crossings has not been synthesized. No link table, a topological classification list by a mathematician, was available for such high crossing number topologies because the number of topologies significantly increases as the crossing number increases.²⁴ The topological notation for 3 we applied is described in the next section 3.2. The F&A strategy achieved such a very complex entanglement as a discrete molecule.

Figure 4.

(a) Complexation of Ag⁺ ions and tripeptide ligand 3 and (b) structure of [4]₁₂-catenane 4. (c) Topology and (d) its diagram of 4. In (d), arrows on each ring indicate the N→C direction within each peptide ligand. (e) A doubly twist structure observed in 4.

The structure and topology of 4 are explained as follows. In 4, the peptide Pro-Gly-Pro sequence of ligand 3 adopted a specific loop conformation, which we called an Ω-shaped loop (Figure 4a). This was connected in a head-to-tail manner by Ag⁺ linkages, which afforded macrocyclic Ag₃(3)₃ rings. Four sets of the equivalent Ag₃(3)₃ ring threaded once with each other, which resulted in a T-symmetric assembly. The nanostructure of 4 was stabilized through multiple non-covalent interactions. At the interfaces between any two rings, both a quadruple π-stacking structure of py groups (3.6–3.9 Å) and Ag⋯O=C interactions at the apical site of each Ag ion (2.6–2.7 Å) were observed (Figure 4e). In addition, at the voids among rings, a total of nine BF₄⁻ were packed through multiple H-bonds with amides of the peptide backbone.

The nanostructure of 4, revealed by crystallographic study, was also confirmed in solution state. Although 4 precipitated as single crystals in aqueous alcohol, the solution structure of 4 was successfully studied in nitromethane. This is because this non-coordinating solvent²⁵ did not interrupt the formation of the reversible Ag–py bond. NMR and MS analyses in solution state enabled detailed studies on the F&A behaviors: for example, a small structural change to a four-membered ring analogue of the N-terminal Pro residue on 3 led to the quantitative formation of the [4]₁₂-catenane structure. This useful solvent was thereafter used for construction of topological molecules described in subsequent sections.

It is also noteworthy that Au⁺ could be used instead of Ag⁺ here. In this case, quantitative formation of 4 was confirmed by complexation of Au⁺ and unmodified ligand 3 in coordinative solvents such as acetonitrile, which was owing to the strong Au–py coordination bonds.

3.2 Another [4]₁₂-Catenane

Another [4]₁₂-catenane topology appeared after exploration of tripeptide sequences. Thr-Pro-Pro sequenced ligand 5 (Thr: l-threonine) was examined in view of introducing a H-bonding side chain on the initial Gly-Pro-Pro sequence. Complexation of ligand 5 with AgOTf in nitromethane afforded [4]₁₂-catenane 6 quantitatively, which was also successfully crystallized²⁶ (Figure 5a). Again, we encountered a 12-crossing topological framework of the Ag₁₂(5)₁₂ composition composed of four sets of the Ag₃(5)₃ ring (Figure 5b). Although both ring- and crossing-numbers were the same as 4, the topology of 6 was different (Figure 5cd). The difference from 4 was the tripeptide conformation, not an Ω-shaped loop but a P_II-helix, which led to the different topology of ring threading. The entangled structure of 6 was derived from a three-crossing substructure (5)₃ through folding into the P_II-helical conformation (Figure 5e). On the other hand, in 4, the interlocking substructure (3)₂ formed through interdigitation of the Ω-shaped loop conformation of 3 (Figure 4e). These observations were also useful for their topological descriptions. Starting from a tetrahedron, the topology of 6 was constructed by replacements, in which every vertex and edge were converted to three-crossed lines and two separated lines, respectively (Figure 6, right). In the case of 4, different replacements were used, in which every vertex and edge on a tetrahedron were converted to three separated lines and doubly twisted (T₂) double lines, respectively (Figure 6, left). The former is called the three-crossed tetrahedral link and the latter is the T₂-tetrahedral link.²⁷ The existence of such entangled topologies tracing the shapes of polyhedra, called regular polylinks or polyhedral links, was discovered in the 1970s–1980s²⁸ and has also been utilized for topological descriptions of DNA polyhedra.²⁹ After exploration of tripeptide sequences in this F&A system, we revealed that the Ala-Pro-Pro (Ala: l-alanine) sequence also gave the three-crossed type of [4]₁₂-catenane framework, on the other hand, both the Ile-Pro-Pro (Ile: l-isoleucine) and the Val-Pro-Pro (Val: l-valine) sequences gave the T₂-type of [4]₁₂-catenane frameworks. Thus, the two types of topologies with the same ring- and crossing-numbers were selectively constructed depending on the difference in only one residue. As a supplement, there are in total 100 topologies for interlocking 4 rings with 12-crossings.²⁴

Figure 5.

(a) Complexation of Ag⁺ ions and tripeptide ligand 5 and (b) structure of [4]₁₂-catenane 6. (c) Topology and (d) its diagram of 6. In (d), arrows indicate the directionality of each ring, whose definition is same as Figure 4d. (e) A three-crossing structure observed in 6.

Figure 6.

Topological description of two types of [4]₁₂-catenanes 4 and 6.

3.3 A Capsid-Like [6]₂₄-Catenane

Incorporation of an unnatural amino acid residue was also examined. We considered the π-stacking of py groups on the peptide of both termini often worked for stabilization of the entangled frameworks. Accordingly, we conceived an idea of introducing an aromatic γ-amino acid, 3-aminobenzoate (hereafter, x), at the middle of a peptide sequence. Complexation of ditopic ligand 7 of the Pro-Pro-x-Ala-Pro sequence and Ag⁺ ions afforded 3.7 nm-sized huge capsule 8 of the Ag₂₄(7)₂₄ composition (molecular weight: 21011)³⁰ (Figure 7ab). The nanostructure of 8 was a [6]₂₄-catenane, in which six sets of the Ag₄(7)₄ ring were interlocked to form a cubic shape with 24-crossings in total (Figure 7cd). The topology of this nanostructure can be called a T₂-hexahedral link according to the polyhedral link notation. The crossing number 24 is very large for not only chemists but also mathematicians: no one can exactly predict the number of existing topologies to our knowledge.²⁴

The x residue at the middle of ligand 7 connected both the turned Pro-Pro sequence and the gently curved Ala-Pro sequence, which led 7 to a S-shaped conformation in the framework of 8. This unique conformation was suitable for creating both three-crossing and doubly twist entangling elements through π-stackings, H-bonds, Ag–π interactions, and Ag⋯O=C interactions (Figure 7e). Owing to these non-covalent interactions, the framework of 8 was retained even at 50 °C. One Ala residue in the sequence of 7 was necessary: neither rigid Pro nor flexible Gly could be used instead of the Ala residue. Pentapeptide was the appropriate peptide length because we have observed longer heptapeptide ligand resulting in a simple [2]catenane through Ag-linked intramolecular cyclization.³¹

Figure 7.

(a) Complexation of Ag⁺ ions and pentapeptide ligand 7 and (b) the structure of [6]₂₄-catenane 8. (c) Topology and (d) diagram of 8. (e) Two types of entangled elements observed in 8. (f) Surface representation of the inner cavity of 8 and (g) the geometry-optimized structure of guest encapsulation structure of 8′.

Fascinating entanglements of as many as 24-crossings sustained the huge hollow structure of ca 3200 ų inner volume as a function (Figure 7f). Although there is a difference in scale, construction of an isolated cavity by peptide entanglements is a common feature with the natural virus capsid described in the introduction of this account review. The methyl side chain of each Ala residue in 8 exquisitely protruded from the inner cavity wall. By changing this side chain, modification of the inner surface could also be achieved. For example, when the Pro-Pro-x-Cit-Pro sequence (Cit: l-citrulline) was used for the construction ([6]₂₄-catenane 8′), 1,3,5-benzenetriacetic acid was captured within the inner cavity through multiple H-bonds between carboxylic groups on the guest and urea groups densely integrated on the inner surface of 8′ (Figure 7g). Such a guest encapsulation within the multiply entangled capsule was possible because the capsular shell was based on the reversibility of Ag–py coordination bonds as well as the other non-covalent interactions.

The success in constructing T₂-tetrahedral and T₂-hexahedral links inspired us to the next target of a [12]₆₀-catenane, namely T₂-dodecahedral link (Figure 8). Furthermore, considering the relationship between 4 and 6 described in section 3.2, another series, [6]₂₄- and [12]₆₀-catenanes of the vertex-crossed type, can also be the next target. In addition, the F&A strategy is not necessarily unique for such high symmetric polyhedra. For example, we have recently obtained a [5]₁₈-catenane of the T₂-triangular prism link.³²

Figure 8.

(a) Examples of T₂-series polyhedral links in a C₃ axis view.

3.4 Torus Knots

Pro-rich sequences have successfully given a variety of highly entangled nanostructures. A simple question came up: what happens if a simple oligomer of glycine, the simplest amino acid, is used? Such an idea prompted us to examine ligand 9 of the Gly-Gly-Gly sequence (Figure 9a). As a result, highly entangled nanostructures appeared again. In other words, we revealed peptide strands have a latent entangled nature. The F&A of ligand 9 and Ag⁺ ions afforded an equilibrium mixture of the Ag₇(9)₇ and Ag₈(9)₈ assemblies (hereafter 10 and 11, respectively), that were new chemical topologies of a 7-crossing torus knot and an 8-crossing torus link, respectively³³ (Figure 9b). In torus topologies, crossing in an odd number ties a single macrocycle (i.e., 10 is a [1]₇-knot) and that in an even number entwines two macrocycles (i.e., 11 is a [2]₈-catenane). These molecules were selectively crystallized by changing the counter anion of Ag⁺ and thus successfully characterized by crystallographic study. Note that counter anions were not templates for each torus nanostructure as confirmed by the ¹H NMR signal ratio. These entangled nanostructures were derived from a specific loop conformation of the triglycine sequence and cooperation of non-covalent interactions such as Ag–py coordination bonds, π-stacking of py groups, Ag⋯O=C interactions, and amide H-bonds (Figure 9c). The peptide conformation of 10 and 11 could be well overlayed and thus almost no energy difference was found for each Ag₁(9)₁ unit in molecular mechanics calculation study. Although torus knot/link topologies have been receiving increased attention,^9,34 both topologies of 10 and 11 are the first reports to our best knowledge.

Figure 9.

(a) Complexation scheme of Ag⁺ ions and tripeptide ligand 9, (b) structures of torus knot 10 and torus link 11, and (c) a doubly twisted structure observed in 11. In (b), each one enantiomer is shown. (d) Possible isomers for 11, considering topological chirality and strands’ directions.

The F&A of Ag⁺ ions and ligand 9 occurred in a highly stereoselective manner. Not only all the head-to-tail linkages in the metal–peptide strand(s), but further structural selectivity can also be discussed. For example, even assuming that all the metal ions and peptide ligands bind perfectly in a head-to-tail manner, there are six structural isomers for 11 (Figure 9d). They are derived from the combination of the topological chirality (M or P, for torus frameworks) and the directionality of each metal–peptide strand. The F&A structures converged to a single enantiomeric pair in the case of ligand 9 and the single P-form in the case of analogous ligand of the Ala-Gly-Gly sequence. Molecular mechanics calculation study revealed that the frameworks of each structural isomer in Figure 9d could be modeled with almost no energy strain compared with 11. However, no amide H-bond formation was found in them. Hence, presence of amide H-bonds between peptide strands in 11 would contribute the structural selectivity. These observations support our description in the introduction paragraph, in which entangling behaviors are likened to fitting jigsaw puzzle pieces. Synthesizing interlocked molecules with high stereoselectivity is an important issue in the field.³⁵ Also, in terms of coordination chemistry, construction of asymmetric nanostructures is a hot topic.³⁶

3.5 Poly[n]catenane

Ligand 9 has another F&A pathway to give poly[n]catenane 12, which was confirmed by crystallization in water/2,2,2-trifluoroethanol solvents in high concentration³³ (Figure 10ab). In contrast to previous examples,^9b,37 we could construct a well-defined poly[n]catenane structure from directional rings here. In 12, structural selectivity was again observed: Ag⁺ connected 9 in the head-to-tail manner, and furthermore, all the metal–peptide rings orientated the same direction (Figure 10c). The Gly-Gly-Gly sequence in the structure of 12 adopted enantiomeric forms of a specific curved conformation: one is called conformation A and the other conformation A* here. Each Ag₂(9)₂ ring was composed of either A–A* or A*–A pair (Figure 10d). In other words, inter-ring interactions such as π-stackings, Ag⋯O=C interactions, and amide H-bonds worked selectively for the homochiral association of the A–A or A*–A* pair (Figure 10e). In general, diastereomerism exists for a poly[n]catenane formation made of directional rings, yet only one meso-type stereostructure appeared here owing to the F&A mechanism. As for the topology, torus knot 10 and link 11 described in section 3.4 were arrays of a 1-crossing unit, on the other hand, poly[n]catenane 12 was an array of a 2-crossing unit. Both entangled elements were in an equilibrium state. In solution state, formation of 12 was suggested by ¹H NMR studies, in which new peaks appeared besides those of 10 and 11 at higher concentration conditions. It was not easy to estimate the degree of polymerization for 12 in solution state, but around n = 10 was suggested according to the diffusion-ordered NMR study.

Figure 10.

(a) Complexation scheme, (b) crystal structure, and (c) topological diagram of poly[n]catenane 12. (d) Enantiomeric pair of rings. (e) Inter-ring interactions observed in 12.

4. Summary and Perspectives

As described in this account review, highly entangled nanostructures have emerged one after another from the simple idea of a metal–peptide strand. Since the number of combinations of peptide sequences and metal salts is enormous, we believe the results shown here would just be the tip of the iceberg. As described in the introduction paragraph, concomitant processes of self-folding and self-assembly are important for constructing highly entangled nanostructures and, at the same time, controlling the stereoselectivity on multiple ring threading. Regarding the peptide sequences, the sequence for entangled nanostructures is not necessarily unique for the proline-rich sequences mainly described here. For example, a β-barrel³⁸ and double β-helices³⁹ formed by F&A of β-strand sequences have also entangling elements. Rational synthesis of entangled nanostructures will be possible by crosslinking of both termini on such peptidic nanostructures. Conformational preferences of the sequence will give a hint for designing entangled nanostructures.

With progresses in constructing topologies with high crossing-number, the classification of them will become important. Such an attempt has already begun for those with high symmetries.⁴⁰ Both approaches of prediction/design and screening/discovery will expand the scope of entangled nanostructures. In addition, attention to entangled nanostructures will also be spread to cages and polyhedral frameworks.⁴¹ Finally, considering that simple topological elements have been utilized in material sciences,⁴² the pursuit of functions of highly entangled topologies will also become a significant topic in the future. For example, it is exciting to find differences in a physical property among topological isomers that have the same ring- and crossing-numbers. Thus, we expect exploration for highly entangled molecular structures will be more and more active in the near future. We hope the work described in this account review will contribute to such a challenge.

Acknowledgment

All studies described in this account review could not have been done without support of collaborators. The authors sincerely thank all of collaborators for their contributions and colleagues for valuable discussions and generous support. TS acknowledges the financial supports by the Japan Society for the Promotion of Science (JP19H02697) and Japan Science and Technology Agency (JPMJPR20A7).

Tomohisa Sawada 【Award recipient】

Tomohisa Sawada received his Ph. D. from the University of Tokyo in 2010 under the guidance of Prof. Makoto Fujita. He then worked as a JSPS postdoctoral fellow at University of Wisconsin, Madison (Prof. Samuel H. Gellman’s group). In 2011 fall, he became an assistant professor at the University of Tokyo (Prof. Makoto Fujita’s group). He was promoted to a lecturer in 2016 and an associate professor in 2018. His current research interests include metal-induced folding and assembly, and molecular recognition.

Makoto Fujita

Makoto Fujita graduated from Chiba University (M. Eng., 1982) and received his Ph. D. degree from Tokyo Institute of Technology in 1987. After beginning his academic career at Sagami Chemical Research Center, Chiba University, Institute for Molecular Science (IMS), and Nagoya University, he moved to the University of Tokyo as a full professor in 2002. In 2018, he was concurrently appointed to IMS as Distinguished Professor. He was awarded the titles of Distinguished Honorary Professor from Chiba University (2018) and University Distinguished Professor from the University of Tokyo (2019). He is a laureate of the Wolf Prize in Chemistry in 2018.