|Home | About | Journals | Submit | Contact Us | Français|
Simple surfactants achieve remarkable long-range order in aqueous environments. This organizing potential is seen most dramatically in biological membranes where phospholipid assemblies both define cell boundaries and provide a ubiquitous structural scaffold for controlling cellular chemistry. Here we consider simple peptides that also spontaneously assemble into exceptionally ordered scaffolds, and review early data suggesting that these structures maintain the functional diversity of proteins. We argue that such scaffolds can achieve the required molecular order and catalytic agility for the emergence of chemical evolution.
Cell membranes organize life's chemistry through an intricate mosaic of lipids and proteins [1-3]. The order achieved by a phospholipid membrane may be important for all cellular functions, but nowhere is this need more apparent than in the multi-enzyme reaction complexes that achieve photosynthesis . Oxidative phosphorylation similarly depends on the membrane to position appropriately the numerous electron transport chain components for burning food . While the phospholipid membrane is quite naturally viewed as the barrier that buffers the cell against environmental fluctuations, this confined fluid, most importantly, provides the principle organizational scaffold in which functional proteins “float” and self-assemble into large functional multi-component catalysts. Lipid rafts have now extended the role of the phospholipid membrane from a simple passive scaffold to a higher level of organization .
Nature's phospholipids then are known to self-assemble spontaneously into supramolecular structures in water, adopt diverse morphologies in response to their environment, survive harsh physical and chemical challenges, and maintain strong electrical and chemical potential gradients. Despite these essential physical capabilities of current cells, the structurally complex phospholipids are unlikely to have been present in the terrestrial prebiotic chemical inventory [7,8]. Other less complex amphiphiles more likely provided the thermodynamically accessible and functional superstructures necessary for the origins of chemical evolution. For example, simple fatty acid mixtures have been found in carbonaceous chondrites  and these structures can form stable [10-12] and functional protocell membranes [13-16].
Amino acids represent another likely component of prebiotic inventories. The Miller-Urey reactions [17,18] and meteorite deposition  both suggest potential sources for amino acids on a prebiotic earth. Simple polypeptides form via dehydration-rehydration cycles , at hydrothermal vents , in salts  on mineral surfaces [23,24], and under simple thermal heating cycles . Peptides composed of alternating hydrophobic/hydrophilic residues can assemble into “membrane-like” β-sheet bilayers by dimerizing to sequester the hydrophobic faces and expose the hydrophilic residues to solution [26,27]. Four to 20 residue peptides can also assemble into a wide-range of morphologies, including fibers [28,29], hollow tubes [28,30-33], ribbons , and vesicles [30,33,35]; all structures reminiscent of the diversity seen in lipid surfactants. However, the peptide backbone and side chains carry much greater functionality than amphiphilic alkanes, and likely have very different structures from lipid membranes.
In aqueous solutions, simple amphiphile assembly is driven by dehydration of hydrophobic domains to create tail-to-tail bilayers stabilized by an exposed hydrophilic surface . Zhang and co-workers have shown that amphiphilic peptide sequences with up to two charged residues at the C-terminus  and a stretch of 6-8 hydrophobic residues, including alanine, valine, leucine [30,37], and glycine , can assemble into worm-like micelles and nanovesicles with diameters ranging from 30 to 50nm . These assemblies display a thickness similar to biological phospholipids, and structural models propose an extended peptide backbone conformations with the peptide axis oriented perpendicular to the tube wall [38,39] forming a bilayer  similar to lipid tubules .
The initial fluid mosaic model  of biological membranes viewed the lipid interior as a confined fluid that orients and stabilizes associated proteins by burial of the hydrophobic residues within the bilayer and directing the polar groups into the aqueous medium. The inherent fluidity of the structure  allows the dissolved components to diffuse laterally [2,41], interact within or outside the membrane, and associate into the assemblies required for such complex functions as photosynthesis and oxidative phosphorylation. Simple amphiphilic peptide surfactants (e.g. AAAAAAK and VVVVVVD) have now been shown to stabilize a variety of membrane bound proteins including glycerol-3-phosphate dehydrogenase , bovine rhodopsin , and the 36 protein complex photosystem I (PS-I) as effectively as commercially available detergents. In fact, peptide stabilization of the PS-I complex showed a nearly 8-fold increase in O2 consumption and remained active for more than 2 months . Interestingly, the degree of PS-I stabilization depends upon the amino acid sequence, specifically the position, number, and type of charged residues. Therefore, these early experiments suggest that simple peptide surfactants can be as effective as lipid surfactants in stabilizing typical membrane protein complexes, and the peptide sequence dependence may indicate that higher-order specific molecular interactions, beyond the simple fluid mosaic, are playing a role in these assemblies.
Others have argued that all proteins assemble into β-sheet rich amyloid under appropriate conditions [44,45]. This prediction is particularly relevant for short peptides, <15 residues, which can easily sample the extended β-sheet secondary structure in water . Amyloidogenic peptides, such as the amphiphilic nucleating core of the Alzheimer's Disease Aβ peptide, Aβ(16-22) (KLVFFAE) , has also been shown to assemble into hollow nanotubes at pH 2 (Fig. 1A, B) with homogenous diameters of 52nm and wall thicknesses of ~4nm . Since the extended peptide length of Aβ(16-22) is ~2 nm, accounting for roughly half the tube wall thickness, initial models predicted a membranous bilayer arrangement with the lysine ammonium ion side chain localized to the surfaces of the nanotube [31,47]. However, isotope-edited IR and NMR experiments on the nanotubes indicated that the peptides were precisely arranged as anti-parallel β-sheets with the registry of the strands within the sheets shifted by a single amino acid residue as illustrated in Fig. 1C .
These β-structural elements then appear to exist in the peptide membranes, suggesting that peptides utilize the forces of protein folding as opposed to the non-specific hydrophobic interactions of alkane phospholipid membrane bilayers. For example, while the Aβ(16-22) membranes discussed above expose half their lysine residues along the solvent exposed surface, the structural model predicts the other half of these positively charged side chains must be packed within the peptide bilayer interface. Non-covalent cross-strand pairing [28,48,49] interactions along the β-sheet surface, including phenylalanine π-π interactions [50-52], K-E salt bridges , and β-branched residue (Val, Ile) packing , must contribute not only to strand arrangements, but also to stabilize these leaflet packing morphologies.
Side-chain functionality also regulates assembly in the presence of metal ions in a manner very similar to proteins. Aβ(10-21), 10YEVHHQKLVFFA, which includes the nucleating core residues discussed above, assemblies very rapidly in the presence of Zn2+ . In the absence of metals, Aβ(13-21)K16A 13HHQALVFFA, forms 5 nm fibers, but in the presence of Zn2+, the peptide assembles into ribbons and nanotubes with diameters two orders of magnitude larger than the metal-free fibers. His-ZnHis ligation between adjacent β-sheets is responsible for the observed increase in sheet stacking (Fig. 3B) [34,47]. Co-assembly of HAQKLVFFA with Cu2+ (Fig 3D) or Zn2+ accelerates assembly and each metal is specifically coordinated to side chain functionality [34,54]. Therefore, many of the architectures that are so prevalent in present day proteins, specific secondary structure elements, transition metal binding sites, and environmental switches, can be incorporated into these peptide bilayer membranes.
With the HHQALVFFA peptide nanotube as a model (Fig. 3B), molecular dynamics simulations predicted cytosine i-motif base pairs as stabilizing elements between adjacent β-sheets. Indeed, when the base was incorporated as a β-(cytosine-1-yl)-alanine (c) into ccQALVFFA and allowed to assemble under conditions favorable for cytosine i-motif formation (Fig. 3C), fixed diameter hollow nanotubes formed rapidly . Characterization by solid-state NMR positioned the peptides as parallel β-sheets, consistent with i-motif base pairing both stabilizing parallel strand arrangements in each sheet and parallel β-sheet lamination . The 3.3 nm wall thickness, determined by small angle x-ray scattering , and the length of the ccQALVFFA peptide (3.1 nm), indicates a monolayer nanotube with distinct inner and outer surfaces; one surface lined with residues at the N-terminus and the other lined with residues at the C-terminus as shown in Figure 3C . Such structures are not accessible in present day biological membranes, and could provide a simple mechanism of segregating functional entities on the inside or outside surface of the tube.
The analysis of oriented KLVFFAE peptide nanotubes and fibers by diffraction reveal their remarkably well-defined paracrystalline order. Reflections from the hollow Aβ(16-22) nanotube (Fig. 2B) showed a doubling of the traditional cross-β 4.7 Å and 9.9 Å d-spacings, arising from orthogonal H-bonded peptides and stacked β-sheets arrays. The doubling originates from distinct cross-β patterns from both top and bottom surfaces of flattened tubes, offset by 23±3° from the tube axis . These repeating cross-β arrays of peptide membranes provide many closely packed binding sites for small molecules. Such arrays have been used to explain the unusual optical properties of the dyes Congo red (CR) [56,57] and Thioflavin T [58,59] bound to amyloid. The precisely patterned surfaces of the Aβ(16-22) nanotubes are composed of positively charged lysine (blue) knobs and leucine-rich hydrophobic groves (gray) as shown in Fig 4B . An interesting outcome of this patterning is seen at CR saturation, which sufficiently passivates the surface and bundles the hollow nanotubes in a manner similar to sulfate  into higher order millimeter long supramacromolecular fibers similar to those shown in Fig. 1B.
Combining the insights gained from CR binding, the fluorescent dye Rhodamine 110, which does not bind to the cross-β template of Aβ(16-22), was covalently substituted for the N-terminal lysine of this peptide  and co-assembled with the unmodified KLVFFAE. The resulting fluorescent nanotubes (Fig 4A) readily bound the sulfonated Alexa 555 dye, a structure similar to CR. Co-localization was evaluated via fluorescence resonance energy transfer (FRET) across the well-defined grid of the nanotube surfaces (Fig. 4C) . This experiment represents an early step in functional light energy capture and transfer along the peptide bilayer surface, in a manner similar to that seen in other amyloids [62,63].
The yeast prion Sup35, which serves naturally as a translational terminator, carries a conserved N-terminal amyloid domain that allows for epigenetic control of its function through fiber assembly . This general concept of organizing active catalysts with simple amyloid domains has been synthetically tested with a variety of other proteins , including green fluorescent protein [65,66], barnase , carbonic anhydrase , glutathione S-transferase [66,67], and cytochrome C , confirming that amyloid can provide a robust scaffold for reversibly organizing large functional proteins. Is it possible then that this conserved Sup35 prion domain is an ancestral amyloid co-opted as a present day regulatory element? Might such assemblies be remnants of an earlier age rather than simply disease-causing protein misfolding errors?
We argued above that many peptides have the generic propensity to self-assemble and to be selected by an ability to form ordered peptide membranes. Building on insight provided by the CR binding site , such membranes could be selected for binding to and production of other membrane components, allowing the membranes to grow and “reproduce”. With such functional selection, autocatalytic cycles , like those seen for helical peptides [70-73] and coiled-coil fibers , could then efficiently produce these peptide membranes. Indeed, preliminary experiments have shown amyloid fibers can self-template the ligation of fragment in a similar manner [75,76]. Such a self-sustaining chemical system would benefit from very short peptides achieving micron scale organization of membranes containing typical protein catalytic features that include metal binding arrays [34,53-54], substrate binding sites [57,59], and even large catalytic domains [64-68]. Membrane and/or fibril localization of multiple copies of catalysts arrayed in close proximity could even grow to a “metabolism” of expanding structural complexity.
To the extent that chemical evolution is simply the nonrandom survival of randomly encoded information , these peptide membranes may represent a low fidelity strategy capable of emerging complexity from a well-recognized prebiotic inventory. The extent to which these molecular surface grids, arrays of catalytic protein domains, and functioning nanoscale molecular barriers can be selected is just now being explored. But just as RNA's catalytic function [78,79] suggested a common molecular ancestor that unified information storage and chemical catalysis for an origins of life hypothesis [78,80-82] the peptide membranes may reveal the earliest roots of a prebiotic chemical evolution.
We are indebted to Jeannette Taylor, Hong Yi, and the Robert P. Apkarian Microscopy Core, Emory University for training in TEM, electron diffraction, and Cryo-etch HR-SEM, and gratefully acknowledge funding from the U.S. DOE (ER15377), NSF CHE-0404677 and NSF-CBC-0739189.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
*These ideas were first discussed at the 3rd Annual Advances in Biomolecular Engineering: Protein Design Symposium at the New York Academy of Sciences on June 12, 2009