Self-Assembly of Cyclic Peptides in Hydrogen-Bonded Nanotubes Part 1(Nanotechnology)

INTRODUCTION

The numerous potential applications of tubular molecules or supramolecules in such fields as biosensing, pharmacology, or catalysis continue to drive the search for new strategies for their construction. This article reviews the design, preparation, characterization, and application status of nanotubular structures obtained from cyclic peptides by self-assembly.

OVERVIEW

Nanotubes and Supramolecular Chemistry

The novel properties of nanostructures have given rise to enormous current interest in the development of nano-structured devices and materials for equally novel applications.[1-3] Nanotubes[4,5] are of particular interest because of their potential applications—sometimes inspired by the biological roles of natural tubules—in fields including ion sensing, molecular inclusion and separation, catalysis, nanocomposites, optics, electronics, chemotherapy, transmembrane transport, and drug delivery. In these contexts, nanotubes have been constructed that consist of, or are based on, zeolites, graphite, inorganic, lipids, and cyclodextrins.[6-11] In most cases, however, it is not yet possible to guarantee uniform prespecified internal tube diameter, a property that, together with its length, will be essential for success in many potential applications.

Nanotubes composed of the materials mentioned above are mostly covalently bonded. Although great advances have been made in this area, noncovalently bonded nanotubes offer significant advantages, including high synthetic convergence, built-in error correction, control through unit design, and, most importantly, self-organization. In noncovalently bonded nanotubes the component molecules are held together by mechanisms such as hydrogen bonding, p-p stacking, electrostatic or van der Waals forces, and hydrophobic interactions.1-12,13-1 These interactions are much weaker than covalent bonds but determine many of the physical properties of molecules, such as solubility and organization in aggregates, and their exploitation for the formation of structures with well-defined composition, shape, and chemical, physical, and biological properties can constitute a viable alternative to covalent synthesis. Lehn coined the term ”supramolecular chemistry” for this field,[14] which is beginning to have an enormous impact on materials science and nanotechnol-ogy. In supramolecular chemistry, molecules are designed and synthesized for their ability to interact specifically with more or less well-specified target molecules (”molecular recognition”) or spontaneously to form larger aggregates with well-defined pattern or structure (”self-assembly” or ”self-organization”).[15-17]

In this article we review the history and current situation of one of the most successful approaches to the preparation of noncovalently bonded self-assembling nanotubes, to wit, the aggregation of cyclic polypeptides in stacks stabilized by hydrogen bonds.

Nanotubes Designs and Peptide Nanotubes

Programs for the preparation of noncovalently bonded self-organizing nanotubes have aimed at a variety of different structures (many inspired by natural structures), including hollow bundles of rod-like units, helically folded linear species, rolled-up sheets, helically juxtapo-sitioned truncated wedges, and stacked rings (Fig. 1A-E, respectively).1-11-1 The last of these schemes (Fig. 1E) is particularly attractive in that the internal diameter of the nanotube is determined solely by that of the unit rings, and can therefore be controlled more easily than in the other cases.

Self-assembling peptide nanotubes (SPN)[18] are formed by stacking cyclic peptides, the stacking interactions being backbone-backbone hydrogen bonds (Fig. 2). Suitable peptides are those in which the ring can adopt a flat conformation in which all the amino acid side chains have a pseudo-equatorial outward-pointing orientation and the carbonyl and amino groups of the peptide bonds are oriented perpendicular to the ring, the carbonyl of each peptide bond on one side of the

Fig. 1 Various strategies for the self-organization of simple components as tubular structures.

Fig. 2 Schematic representations of the general strategy for nanotube formation stacking cyclic peptides.

The cyclic peptide approach to nanotubes has two crucial advantages over all others that have so far been tried: firstly, the diameter of the polypeptide rings, and hence the internal diameter of the nanotube, is easily controlled by varying the number of amino acid residues in each ring; and secondly, the properties of the outer surface of the nanotube can easily be modified by varying the amino acid side chains. For example, self-assembly in lipidic or aqueous media can be promoted by endowing the cyclic units with hydrophobic or hydrophilic groups, respectively, so as to stabilize the flat conformation required for ring stacking. Appropriate unit design and optimization of conditions for self-assembly allows the properties of the resulting nanotubes to be tailored for specific applications. Thanks to this, SPN can be used as or in porous solid materials, soluble cylindrical super-molecules, ion channels and other transmembrane pores, solid-supported ion sensors, antimicrobial and cytotoxic agents, and nanocluster composites.[11'20]

Fig. 3 Schematic representations of nanotubes (bottom) formed by self-assembly from cyclic d,l-a- or P-peptides (top left and top right, respectively) (for clarity most side chains are omitted).

TYPES OF PEPTIDE NANOTUBE

Introduction

The assembly of nanotubes by stacking cyclic peptides was first suggested in 1972 by Hassall, who envisaged that cyclic tetrapeptides composed of alternating a- and p-amino acid residues would assemble through backbone-backbone hydrogen bonding.[21] This hypothesis was verified in 1975 by the analysis of the crystal structure of the tetrapeptide cyc1o[l-Ser(O-t-Bu)-p-Ala-Gly-l-p-Asp(OMe)-].[22] However, although the peptide rings were found to adopt the required conformation and stack above one another in the crystal lattice, neighboring rings were linked by only two hydrogen bonds instead of four. The use of hybrid rings composed of mixed a- and p-amino acid residues is currently in abeyance, and in this review we concentrate on cyclic d,l-a-peptides, cyclic p-peptides, and other peptide rings.

Tubular Ensembles of Cyclic D,L-a-Peptides

Preliminary studies

In 1974, on the basis of theoretical analysis, De Santis et al. concluded that peptides composed of an even number of alternating d- and l-amino acid residues would form closed rings capable of stacking through backbone-backbone hydrogen bonding.[23] The resulting supramolecular tubes would be closely related to the p-helical conformations adopted by linear d,l-peptides. Initial attempts to verify these predictions experimentally were inconclusive because of the poor solubility of the peptides employed.[24] For example, in 1989 an X-ray crystallo-graphic study of diverse cyclic peptides by Lorenzi and coworkers failed to find the expected associations between the basic units, each peptide instead being hydrogen-bonded to several cocrystallized solvent molecules.[25] However, in 1993 Ghadiri and coworkers took advantage of the pH-dependent ionization of the glutamic acid side chain to control nanotube formation, as described in the next subsection.[18]

Solid-state ensembles: microcrystalline peptide nanotubes

The first well-characterized peptide nanotube was prepared using the octapeptide cyc1o-[(l-Gln-d-Ala-l-Glu-d-Ala-)2], which was chosen because it would favor solubility in basic aqueous solution, where coulombic repulsion among its negatively charged carboxylate side chains would prevent premature ring stacking.[18] Controlled acidification of the basic solution of this peptide led to microcrystalline aggregates that were characterized by transmission electron microscopy (TEM), electron diffraction, Fourier transform infrared spectroscopy (FT-IR), and molecular modeling. These analyses showed, as expected, ordered hollow tubes formed by cyclic units stacked through antiparallel p-sheet-type hydrogen bonding, the between-ring distance being 4.73 A and the internal tube diameter 7.5 A (Fig. 4).

Proton-controlled self-assembly also allowed the preparation of microcrystalline aggregates of nanotubes with an internal diameter of 13 A composed of cyc1o-[(l-Gln-d-Ala-l-Glu-d-Ala-)3] units,[26] which confirmed that the internal diameter of the nanotube could be determined just by varying the number of amino acid residues in the cyclic unit. Ghadiri’s group has also prepared solid nanotubular assemblies using various uncharged cyclic octapeptides to explore the effects of intertubular hydrophobic packing interactions on crystal formation; FT-IR, cryoelectron microscopy, and electron diffraction analyses show the expected cylindrical structures, with between-ring distances of about 4.8 A and all the characteristic features of an antiparallel p-sheet-like structure.[27] More recently, Lambert and coworkers have employed pH-controlled self-assembly to synthesize nanotubular microcrystals of the cyclic d,l-a-octapeptide cyc1o-[(l-Asn-d-Phe-l-Asp-

Solution-phase studies of dimerization

The basic step in all the above self-assembly processes, the association of cyclic peptides by antiparallel-p-sheet-type hydrogen bonding, has been investigated in water by fluorescence quenching methods, which confirm that ring-ring association with the previously reported stacking parameters occurs in solution and is not just a consequence of crystallization.[29] To obtain a better understanding of this stacking interaction, and to carry out preliminary exploration of various ways of stabilizing the resulting nanotube structures, toy systems were designed and studied in which complications associated with unlimited stacking (such as poor solubility) are avoided by allowing only the formation of two-ring structures (hereinafter, ”tubelets”). This restriction is achieved by selective alkylation of the amino groups on one face of the peptide ring (those of amino acid residues with the same chirality).[30] For example, in deuterochloroform cyclo-[(l-Phe-d-MeN-Ala-)4] associates as dimeric tubelets with an association constant of about 2540 M~1, but these structures grow no further because their N-methyl groups prevent hydrogen bonding to a third peptide ring (Fig. 5).[31] X -ray analysis shows that crystals obtained in methylenechloride/hexanes are composed of similar dimers, the hydrophilic nature of the cavity of which is confirmed by their containing water molecules. Solution-phase nuclear magnetic resonance (NMR) and FT-IR studies of this and other N-methylated peptides support the proposed structures and have begun to elucidate the thermodynamics of nanotube formation. These studies suggest that each hydrogen bond contributes 0.5-0.8 kcal mol~1 to the enthalpy of formation. N-Methylated cyclic hexapeptides such as cyclo-[(l-MeN-Leu -d-Leu-)3] dimer-ize with smaller association constants[32], but deca- and dodecapeptides fail to do so because of the difficulty to adopt the required flat conformation.[29-31] N-Propyl and allyl groups can also be used to prevent ring stacking from proceeding beyond the dimer stage.

Fig. 4 Dimensions of cyclo-[(l-Gln-d-Ala-l-Glu-d-Ala-)2]-based nanotubes in crystal arrays, as deduced by TEM, electron diffraction, FT-IR, and molecular modeling.

Fig. 5 Parallel (p-q) and antiparallel (p-p and q-q) P-sheet structures of dimeric tubelets composed of cyclo-[(l-Phe-d-MeN-Ala-)4] (p) and/or cyclo-[(d-Phe-l-MeN-Ala-)4] (q), with the corresponding dimerization constants.

Studies of sidechain-sidechain interactions have shown that g-branched side chains are more favorable for dimerization than unbranched chains. Recently, aromatic side chain-side chain interactions in cyclic peptides containing homophenylalanine were used to induce crystal growth orthogonal to the tubulet axis through the prevalent effect of dimer formation in the crystal nucleation.[33]

Dimeric tubelets have also allowed easy comparison of the stabilities of parallel and antiparallel P-sheet structures in solution.[34] Using the enantiomeric cyclic peptides cyclo-[(l-Phe-d-MeN-Ala-)4] and cyclo-[(d-Phe-l-MeN-Ala-)4], it was found that antiparallel P-sheet structures (p-p and q-q in Fig. 5) are more stable than parallel structures (p-q) by 0.8 kcal mol~1.

Finally, further confirmation of P-sheet-type hydrogen bonding has been obtained by covalent consolidation of noncovalently constituted cyclic peptide dimers. In nonpolar solvents the olefin-bearing peptide cyclo-[(l-Phe-d-MeN-Ala-l-Hag-d-MeN-Ala-)2] (X = CH2-CH=CH2 in Fig. 6; Hag=homoallylglycine) forms equal proportions of two kinds of dimer, one of which undergoes selective hydrogen-bond-mediated olefin metathesis to afford covalently stabilized P-sheet peptide tubelets (see legend in Fig. 6).[35] Subsequent work has extended this general strategy to other covalent links, such as disulfide bonds:[36] for example, with cyclo-[(l-Phe-d-MeN-Ala-l-Cys-d-MeN-Ala-)2] (X=SH in Fig. 6) S-H – S hydrogen bonds appear to stabilize the reactive dimer prior to formation of the disulfide bond. Studies of this kind, using dimers as toy systems, are paving the way for the preparation of stable polymeric nanotubes composed of cyclic peptides with backbone-unalkylated amino groups.

Artificial transmembrane ion channels

As was emphasized above, the major advantage of peptide nanotubes over others is the ease with which their external surface properties and internal diameters can be controlled by the use of cyclic peptides with appropriate side chains and numbers of amino acid residues. In 1994, Ghadiri’s group created the first transmembrane nanotube to self-assemble in a lipid bilayer, a feat that depended on the hydrophobic side chains of the cyclic d,l-a-peptide cyclo-[l-Gln-(d-Leu-l-Trp)3-d-Leu-] (Fig. 7, x = H).[37] Since then, several similar structures have been studied by liposome-based proton transport assays, computational methods, FT-IR spectroscopy, grazing-angle reflection/ absorption, polarized attenuated total reflectance (ATR), and single-channel conductance measurements.1-38’39-1 In all cases, when the amino acid sequence is dominated by hydrophobic amino acids, the resulting nanotube is oriented nearly parallel to the lipid alkyl chains, as required for a transmembrane channel.1-39-1 While peptide nanotubes formed from amphipathic cyclic peptides lay parallel to the membrane plane. Increasing the size of the cyclic peptide, for example by using cyclo-[l-Gln-(d-Leu-l-Trp)4-d-Leu-], provides channels with larger pores. Both cyclo-[l-Gln-(d-Leu-l-Trp)3-d-Leu-]- and cyclo-[l-Gln-(d-Leu-l-Trp)4-d-Leu---based channels have prodigious alkaline ion transport activities. Molecular dynamics simulations suggest that these high rates of transport may be due largely to the way in which water molecules are ordered in the lumen.[38- However, even larger cationconductances (10-15% larger for 20 mM K+) can be achieved by oriented heterodimer-based nanotubes in which the ring at the end facing the K+ source, cyclo-[(l-Glu-d-Leu-)4-, is negatively charged at the working pH.[40-Ab initio calculations have suggested that band conduction may occur through the between-ring hydrogen bonds of peptide nanotubes as a result of delocalization of electrons and holes toward the tube axis.[41]

Fig. 6 Consolidation of cyclic peptide dimers by covalent bonding between their constituent rings using olefinic or thiolic side chains. Of the two dimers formed by cyc/o-[(l-Phe-d-MeN-Ala-l-Hag-d-MeN-Ala-)2] (Hag=homoallylglycine; X = CH2-CH=CH2 in the figure), only the isomer in which CH2-CH=CH2 groups of different peptide rings face each other can undergo covalent stabilization.

Tubular Ensembles of Cyclic p-Peptides

p-Peptides are an emerging class of nonnatural biopoly-mers that in solution adopt a variety of secondary structures analogous to those of a-peptides, including pleated sheets, turns, hairpins, and a variety of helical conformations of differing chirality and radius.[42] Their ability to resist enzymatic degradation and their favorable conformational properties suggest that b-peptides may have interesting applications. Molecular modeling and X-ray crystallography have shown that in the solid state, cyclic tetrapeptides composed of chiral b3-amino acid residues can adopt flat-ring conformations and stack to form nanotubes in the same way as cyclic d,l-a-peptides, and that in the case of P-peptides this is possible with peptide rings composed of homochiral P-amino acid residues as well as with rings of residues of alternating chirality (Fig. 3).[43]

Fig. 7 Schematic representation of a peptide nanotube self-assembled in a lipid bilayer by association of cyclic peptide units. Peptides used by Ghadiri and coworkers in transport experiments are shown at the left (x = H)[38'39].

Like their d,l-a counterparts, cyclic p3-peptides with appropriate hydrophobic side chains can associate in lipid bilayers to form efficient self-assembled artificial trans-membrane ion channels.[44] In the cases of cyclo-[(p3-HTrp-)4]- and cyclo-[(P3-HTrp-P-Hleu-)2]-based channels assembled in planar lipid bilayers, single-channel conductance measurements have shown activities similar to those of d,l-a-peptide-based channels, with K+ transport rates of 1.9 x 107 ions sec~1, greater than that of gramicidin A under similar conditions. It is thought that the unnatural backbones of channel-forming cyclic P3-peptides will have interesting consequences for the corresponding channels and channel formation processes. In particular, as linear P-peptides resist digestion by proteases because of their unnatural backbone, the same is expected to be true of cyclic p3-peptides, which should therefore be able to form ion channels with potential antimicrobial effects in the presence of these enzymes; a mechanism of this kind is thought to be responsible for the known antimicrobial activity of helical linear p-peptides.[45]

Tubular Ensembles of Cyclic Peptides Containing g-Amino Acid Residues

The inner faces of d,l-a- and p-peptide-based nanotubes are hydrophilic (and some of their potential applications depend on this). It is not possible to modify their pore properties by introducing functional groups on their inner faces, because all their amino acid side chains point outward and substitution at Ca or Cp would interfere with nanotube assembly. However, this shortcoming disappears if cyclic hybrid a,g-peptides are used as the basic units for nanotube construction. Our group has recently designed, synthesized, and characterized a new cyclic peptide, composed of alternating a-amino acids and cis-3-aminocyclohexanecarboxylic acid (g-Ach), that can adopt the flat conformation required for the formation of nanotubes consisting of hydrogen-bonded stacks of alternately oriented rings (Fig. 8).[46] In these rings the p-methylene of each cyclohexane projects into the lumen, creating a hydrophobic region. That these rings can indeed stack as desired was verified by using the backbone methylation strategy (see ”Solution-Phase Studies of Dimerization”) to prepare hydrogen-bonded dimers of each of the two types required for nanotube formation (one bonded via the a-amino acid residues of each component cyclic peptide, the other via their g-Ach moieties; Fig. 8a, frames A and B). Nuclear magnetic resonance, Fourier transform infrared spectroscopy, and X-ray diffraction studies conclusively confirmed the formation of these dimers and showed that the dimer bonded via a-amino acid residues are extremely stable in nonpolar solvents, with Ka’s> 105. It should be possible to functionalize the inner surface of nanotubes composed of these hybrid a,g-peptides by introducing substituents on the p-methylene, something that, as noted above, is impossible for nanotubes based on cyclic a- or p-peptides. The ability to functionalize nanotube cavities would give the chemist greater capacity to design selective ion channels, selective molecular inclusion devices, or molecular and catalytic containers.

Fig. 8 A cyclic peptide composed of alternating a-amino acids and cis-3-aminocyclohexanecarboxylic acid (g-Ach) that can stack to form nanotubes with hydrophobic regions in the lumen around the cyclohexane p-methylenes. a) The two types of hydrogen bond involved in nanotube formation (left, A and B) and their separate realizations in N-methylated dimeric tubelets (right). b) Silicon representations (1, top view; 2, side view) of the crystal structure of dimeric cyc/o-[(d-Phe-(1R,3S)-MeN-g-Ach-)3], which crystallized in association with five molecules of chloroform, one of them occupying the central cavity of the dimer.

Fig. 9 Self-association of a rigid lactamic g-tripeptide as a nanotube.

Fig. 10 Self-assembled nanotubes investigated by Ranganathan and coworkers. a) Self-assembly by association of cystine-based macrocyclic bisureas. b) Self-assembly through p-p interactions between interdigitating aromatic rings included in the backbones of the cyclodepsipeptides constituting neighboring nanotubes.