Jerry H Brown. American Scientist. Volume 97, Issue 4. Jul/Aug 2009.

Nature has often inspired human creativity since the first humanrendered artifacts appeared some 70,000 years ago. In particular, the study of nature, at various scales of size, has led to numerous innovations in engineering. For most of our species’ existence, the structures available to person-as-engineer were those that could be seen with the naked eye. Then, in the 17th century, the invention of a practical light microscope opened up a new world of natural objects for examination and inspiration.

Studying burrs under a microscope in 1948, for example, inspired George de Mestral to appreciate the binding properties of arrays of hooks and loops, the underlying design that led to the invention of Velcro. More recently, D. W Bechert with the German Aerospace Center and colleagues and Konrad Koeltzch of Ohio State University and colleagues have studied sharkskin to reveal how a surface consisting simply of appropriately oriented riblets diminishes turbulence in fluids; this design is finding various applications including the surfaces of fuel-efficient airplanes. These discoveries of designs that are in retrospect quite simple in concept—and made more than three centuries after the invention of the light microscope—suggest that additional important basic discoveries of form and function are yet to be made. Biomimicry has much more to teach us.

Just as the light microscope redefined the natural world in the 1700s, access to nature recently expanded again: Biophysical techniques of the 20th century, especially x-ray crystallography but also nuclear magnetic resonance and to some extent electron microscopy, are exposing a diverse molecular world for the engineer to study. These and other techniques have shown that molecules come in a near-infinite variety of shapes, compositions, geometric arrangements and overall sizes, and knowledge of their structure and dynamics is crucial for understanding their chemical and functional properties. But of all the classes of molecules, proteins should prove most revealing for the engineer, and their study could be one of the most important subspecialties of biomimicry—enough so to earn it the name proteomimicry.

What’s So Special About Proteins?

As is well known to the protein specialist, but not necessarily to the wider scientific and engineering authence, the structure of proteins may be understood at different levels or hierarchies. Briefly, sequencing yields the primary structure of a protein chain, a one-dimensional view of the covalently linked amino acids. Although all protein (polypeptide) chains basically share the same chemical composition of the main-chain backbone, they differ both in the number of amino acid residues that are linked together, and in the composition and arrangement of the protruding side chains. At each amino acid position along the sequence, there are (at least) 20 types of side chains from which to choose (varying in size, shape, polarity, charge and so on). Segments of the protein chain nevertheless only form a few types of secondary structures, with compact alpha helices and elongated beta strands being most common. It is how (and whether) these segments are folded into a defined three-dimensional tertiary structure, and even how different folded chains combine to form a quaternary structure, that provides a particular protein with its form, and hence its function.

The world of proteins is vast. The human genome alone consists of about 20,000 genes that can, through alternative splicing, make perhaps a hundred thousand or more different human proteins. The genomes of other organisms encode for a nearly countless number of non-human proteins as well. Sir John Kendrew and his coworkers first determined the three-dimensional structure of a globular protein in 1958, when they showed how the chain of the oxygen-storage protein myoglobin is folded. Over the next 32 years, 450 protein structures (with 80 unique folds) were deposited in a database called the Protein Data Bank (PDB), and recent technical advances have swelled the number of deposits to more than 53,000 structures (and 1,100 unique folds).

This quickly growing collection of structures reveals a few properties of proteins that provide a unique source of ideas for the engineer. The first property was immediately recognized in the fold of myoglobin: the unexpectedly irregular manner in which its secondary structural elements are positioned relative to one another. Compared for example to the reassuringly symmetrical structure of the DNA double helix, the relative arrangement of the alpha helices in myoglobin seemed disappointing to some at first glance. Max Perutz, who shared the 1962 Nobel prize in chemistry with Kendrew, even called it “hideous.” Perhaps to some eyes, but it is this property—irregularity in globular proteins, as well as deviations from regularity found in fibrous proteins—that yields the vast repertoire of different structures for study.

A second important property of proteins, confirmed as the first few dozen structures were determined, is the hierarchal arrangement of primary through quaternary structures described above; this allows the protein engineer to conceptually manage the very large number of structures and to develop an understanding of common features and relationships between their folds. The third significant property, appreciated as a result of comparing the structures of the same protein under different conditions, is that many proteins consist of relatively “rigid” domains (often tertiary folds), connected to each other by flexible joints. This feature led Jonathon Howard, now with the Max Planck Institute, to refer to the “mechanics” of these proteins. The chemical forces and engineering principles responsible for modulating the motions of the domains are beginning to be deciphered. A fourth property is that, despite the intrinsic weakness of the noncovalent forces that hold a protein together, some proteins use clever designs to enhance rigidity and resistance to melting or mechanical stress. These designs may provide especially useful lessons for engineers who need to deal with intrinsically weak or soft components.

Some of the designs controlling the mechanical properties of proteins appear to have well-recognized analogues in the engineering world. For example, Daniel FIimmel, now at Rutgers University, and Carolyn Cohen, of Brandeis University, and their colleagues have likened the motor domain of myosin to an automobile engine, with elements that resemble a piston and a clutch; David DeRosier of Brandeis University has observed that the bacterial flagellum uses a corkscrew-shaped filament to convert this rotary motor’s torque to thrust; Zhisong Wang and his colleagues at Fudan University have described kinesin as a “ratchet-and-pawl” device that provides directionality to its movement on microtubules; and Gwangrog Lee, while a graduate student at Duke University, and colleagues noted that ankyrin repeats appear to have the mechanical properties and shape of a (Hookean) spring.

In this article I show, using specific examples, that the world of proteins may provide novel designs, or useful variations of well-known designs, for the mechanical engineer to apply in the macro—as well as microscopic worlds. Some of the specific designs described herein function to strengthen joints; others modify the dynamic properties of structures.

Joints of the Alpha-helical Coiled Coil

One of the simplest structural designs in the macroscopic world is the tenonmortise joint. The molecular analogue of this joint, the so-called knob-into-hole interaction, provides the major physical basis for the alpha-helical coiled coil. (All uses of “coiled coil” in this article refer to the alpha-helical coiled coil.) This type of structure is one of the most frequently encountered motifs that joins (or dimerizes) two polypeptide chains. First modeled by Francis Crick in 1953, the alpha-helical coiled coil also has the most well-understood relationship between sequence and three-dimensional structure. Because of its relative simplicity, the alpha-helical coiled coil, perhaps more than any other type of protein structure, is ideally suited for deriving simple engineering principles from the shapes and patterns of its amino acid components.

The sequences of (most) coiled coils are simply characterized by a sevenresidue (heptad) repeat of the form ah-c-d-e-f-g, that spans nearly two turns of an alpha helix, where the side chain at every third, then fourth, position (a and d) is generally apolar. The stability of the alpha-helical coiled coil depends on the types of amino acid residues that occupy these heptad positions. Two types of interactions appear most critical in stabilizing the association of the two helices: salt links and apolar interactions. Interhelical salt links may be made from a number of heptad positions, and those from the g and e’ positions (which flank the core) appear most stabilizing. Apolar interactions between the alpha helices usually involve the core a and d positions. As proposed by Crick, and later verified by Erin O’Shea at the Howard Hughes Medical Institute based on crystal structures, a nearly parallel, gently left- handed winding arrangement of the alpha helices relative to each other permits each of the side chains from the a and d positions to form a knob that directly faces into a cavity or “hole” between four side chains in the partner helix. In most two-stranded coiled coils, leucines are common in the d positions and isoleucines or valines in the a positions, as the branched ends of these side chains snugly fit into the diamond-shaped holes. The existence of repeating knob-hole interactions in extended alpha-helical coiled coils helps stabilize the structure. By analogy, in the world of woodworking for example, the protruding a and d side chains are “tenons,” the holes are “mortises,” and the tenon-mortise joint is the fundamental unit that stabilizes the connection between the two pieces. Reinforcements, such as salt links between the alpha helices, and glue or nails between the pieces of wood, in addition to multiple tenons and mortises, can produce a relatively stable joint.

Further inspection of various alphahelical coiled coils reveals particular arrangements or relative sizes of the knobs and holes that appear to impart special properties to this motif.

Economizing Production. One pattern observed in the alpha-helical coiled coil increases the efficiency of joint production: the azimuthally (laterally) reciprocal nature of the knob-hole interactions. Had an interaction between two hypothetical chains involved one chain that only has knobs and the other chain that only has holes, then two different polypeptide chains would be required, encoded by two different genes; by analogy, such a tenon-mortise joint would require the production of two types of components. Each alpha-helical chain, however, offers both knobs and holes, with diamondshaped holes positioned azimuthally adjacent to the knobs. The hermaphroditic nature of this alpha helix permits two identical polypeptides to interact reciprocally with each other, using pairs of knob-hole interactions at each layer, and thus homodimeric coiled coils requiring only one gene are possible. A lateral reciprocal pairing of tenon-mortise joints can similarly simplify production of components for woodworking and other projects. In fact, such a concept has already been applied in the design of a seismic sensor shroud package (U.S. patent 6,385,132); each half of its outer shell is identical, containing both tenons and mortises that are used to reciprocally join the two halves together, a feature which is noted for reducing inventory issues.

Strengthening the Joint: Face-centered Pattern. The relationship between consecutive pairs of knob-hole interactions in the alpha-helical coiled coil shows another specific pattern (one not presented in the above-mentioned patent) that may be especially suited for stabilizing a joint. As noted above, the strength of the interaction between two components is commonly increased simply by including multiple (three shown in the patent) pairs of tenon-mortise joints rather than just one pair. Often (as in the patent), all of the tenons are lined up on one side of each component, and the mortises on the other side. (This arrangement is somewhat analogous to the uncommon 3.10 helix in proteins, which because of the near-integral number of residues per turn has its side chains axially aligned.) The stability of a dimeric structure with this arrangement may, however, be affected by a strong axial force; the relatively short distances between consecutive mortises (holes), as well as the alignment of the mortises, could create a weak link, in particular if the material used is not especially sturdy.

Along a face of an alpha helix, however, the positions of the knob and hole are laterally switched in any one layer (for example, the d- layer) relative to their positions in the next axial layer (the a- layer). (This is because there are [nearly] a half-integral number of residues [3.6] per turn in an alpha helix.) The alpha-helical coiled coil thus displays a sort of face-centered pattern of knob-hole interactions. Douglas Root, of the University of North Texas, and his colleagues recently performed a molecular mechanical simulation of an alphahelical coiled coil subjected to axially directed forces. Their results suggest that the hydrophobic residues that interact between the chains are relatively resistant to separation, which may be due in part to the pattern of knob-hole interactions. A face-centered pattern applied to tenon-mortise joints would yield relatively large distances and non-identical vectors between consecutive mortises, leading one to expect that they would be less likely to be dislodged from each other by an axially directed force than with the pattern wherein the mortises (and tenons) were aligned.

Also note that, as mentioned above, the two chains in the alpha-helical coiled coil are not precisely parallel, but gently coil around each other. (Again, the structure coils because the number of residues per turn is not exactly half-integral). Such coiling can provide stability regardless of the direction of an external force.

Semi-flexible Bends of Tropomyosin. Whereas alpha-helical coiled coils in general reveal a design for stability, my colleagues and I have found that the crystal structures of certain members of this class of proteins, in particular tropomyosin, reveal features for semiflexibility.

The sequence of tropomyosin shows that in the core positions, regions rich in leucine-sized residues alternate with clusters of small alanine residues. In the crystal structures, the leucine-sized residues fit snugly into and span a diamond-shaped hole formed by four side chains of the neighboring helix. As a result, a relatively tight fixed joint between two nearly in-register helices is produced. In contrast, alanine is too small to make contact with all four side chains of the neighboring helix. The main chains at the alanines are thus axially staggered relative to one another by about 1.2 Angstroms, so that the alanine side chain is positioned into one end of the diamond shaped hole, making firm contact with three of the side chains. By analogy, the alanine side chain is a “tenon-wannabe” that is smaller than its mortise. The alanine interaction may thus be better described as a “pin-in-slot,” where the pin is stabilized at one end or the other of the slot.

The joining of the in-register and axially staggered regions results in locally more helix on one chain than on the other at the junction; that is, there is a wedge, and correspondingly there is a relatively sharp (around 6-degree) “inplane” bend in the axis of the tropomyosin coiled coil away from the wedge. The ability of tropomyosin to bend is crucial to its biological function: It must wind around actin to regulate the interaction between myosin and actin. But tropomyosin also demonstrates a general design principle: how a pinin-slot design can be used to create a wedge bend.

The generally homodimeric nature of tropomyosin, moreover, imparts dynamic characteristics to these mechanical designs. As described above, each alpha helical chain of tropomyosin has both knobs and holes, and thus at any one layer there is a reciprocal pair of knobs-into-holes interactions. At the leucine-rich symmetric segment, the two interactions are identical. At the alaninerich segment, however, the interactions are not the same: If the alanine knob from the “right” helix interacts with the top of the diamond-shaped hole in the “left” helix, then the corresponding alanine knob from the left helix interacts with the bottom of the hole of the right helix. But since the sequence of the two chains of (“alpha/alpha”) tropomyosin are identical, the alternative stagger, left chain up and right chain down, is intrinsically equally likely.

David Goodsell and Arthur Olson of the Scripps Research Institute have made the general point that the ability to adopt multiple conformations is a necessary consequence of designing asymmetric interactions between identical objects. Tropomyosin thus suggests a reciprocal pin-in-slot “switch” between two axial positions. In addition, as a result, at the junction of the alanine-rich and leucine-rich segments, tropomyosin can bend with equal likelihood in two opposite directions. Tropomyosin thus also shows how a bistable pin-inslot switch can be used to produce an object with semi-flexible bends.

The Lever and Myosin

Alpha-helical coiled coils have taught us about the way tenons and mortises (or pins in slots) may be advantageously arranged; now we turn to the head of myosin for possible uses of the lever. The heads of most muscle, as well as non-muscle, myosins have similar architecture employed to convert chemical energy to mechanical work. The motor domain portion of the head consists of four relatively rigid globular subdomains, linked to each other by flexible single-stranded joints. The consumption of adenosine triphosphate (ATP) in, or the binding of actin to, the motor domain results in comparatively small changes in the relative orientations and positions of its upper 50 kilodalton (kDa), lower 50 kDa and N-terminal subdomains. Taro Uyeda’s group at the Japanese National Institute for Interdisciplinary Research, and more recently Stefan Fischer, at Universität Heidelberg, and his colleagues, have shown that these changes are transmitted through a nearby fulcrum located at the boundaries of the converter, N-terminal and lower 50 kDa subdomains to create a relatively large swing of the light-chainbearing domain, which muscle scientists also call the “lever arm.”

The Conventional View of a Lever. The lever is probably most commonly known for its ability to amplify effort. This “mechanical advantage” arises when the length of the effort arm—that is, the distance between the location of applied effort and the fulcrum—is greater than the length of the load /resistance arm, which is the distance from the fulcrum to the point of resistance. In this way Archimedes can move the world or, more humbly, a claw hammer can be used to extract a stubborn nail. It follows that levers can also be used to amplify distance (and speed), when the load arm is made to be longer than the effort arm. In this way, contraction of one’s biceps over a relatively short distance can be made to move the forearm and the hand (and anything it might be holding) a relatively long distance, albeit with diminished output force; a similar principle applies to the catapult. The head of myosin also uses the “distance amplification” option of the lever on the molecular level to promote a relatively speedy contraction of the sarcomere, the chassis to which myosin is attached. This increased speed results in part because myosin’s lever arm domain, which in engineering parlance is a “load arm,” is long relative to myosin’s effort arm, thus allowing myosin to take relatively long steps along actin.

Amplifying Distance Amplification. Conventional distance amplification with a lever is only part of the story of how myosins take long steps along actin. Fischer’s group has modeled the transition between the beginning and end points of the recovery stroke of myosin (and by implication of the power stroke) as consisting of two “phases” or types of motion. The first, so-called “see-saw,” phase uses the conventional distance amplification feature of the lever described above: As one side of the see-saw (the N-terminal segment of the relay helix and switch ?) is being pulled “down” toward the nucleotide in the active site, the other side of the see-saw (the C-terminal segment of the relay helix, followed by the converter and the light chain “lever arm” domain) swings “up” by about 25 degrees. Here, the relay helix pivots about a fulcrum that is made up of interlocking bulky residues. But as the nucleotide in the active site is hydrolyzed, and the N-terminal segment of the relay and switch ? are pulled even closer to the nucleotide, strain seems to build up near the fulcrum. In response, one turn of the relay helix near the fulcrum unwinds by one H-bond along the helix. In this “unwinding phase” there is an additional 40-to-50 degree rotation of the C-terminal segment of the relay and the attached converter and lever arm. Myosin thus shows that distance amplification can result not only from the appropriate positioning of the fulcrum on the lever, but also by the introduction of a flexible joint within the lever itself.

Although this combination of lever swing about the fulcrum and bending of the lever about a joint near the fulcrum appears to be sufficient to account for the step size of most myosin isoforms, bending about an additional joint farther up the load arm, in the converter, appears necessary for the myosin VI isoform to take long steps on actin. This bending at the end of its power stroke helps avoid a steric clash during the power stroke that would otherwise occur between the converter and a region of the N-terminal subdomain.

Myosins have thus found a way to increase the distance traversed by a lever without merely increasing its length. Engineers have also sought solutions to this problem, and a card ejecting mechanism in an IC memory card connector in U.S. patent number 5,383,789 is one clever example. A constraint here is the desire to amplify force (to extract the card from a sticky connection) as well as distance (to ease grasping of the ejected card), and the solution offered is to use two simple (rigid) appropriately placed fulcrums to allow both types of amplification to occur in a sequential manner. Our analyses of myosins reveal an alternative solution: introduction of strain into the interaction between a fulcrum and the lever arm, the relief of which triggers formation of and bending about a flexible joint or joints within the lever itself.

Resistance Amplification. Apart from distance amplification, studies of myosin can extend our understanding of the force amplification function of the lever as well, although amplification of effort is a well-recognized function of levers. My colleagues and I recently compared myosin from muscle with myosin V, which is involved in transporting cargo, showing resistance amplification as another function for the lever.

All myosins share a general crossbridge cycle: a power stroke promoted by the strong binding to actin and a recovery stroke fueled by the hydrolysis of bound ATP while the head is detached from actin. Myosins differ, however, in the relative amount of time they reside in actin-attached and -detached states. In muscle, the tails of hundreds of molecules are tethered to each other in the thick filament, and at any one time most of the myosin heads must be detached from the actin thin filament to permit efficient contraction. By contrast, cargo-transporting myosins such as myosin V are dimeric, and it appears that a relatively high actin residence time is required to prevent the molecule from irreversibly dissociating from its actin track.

By comparing crystal structures of the head of muscle myosin, as determined by Carolyn Cohen and colleagues at Brandeis University, to those of Myosin V, as determined by Anne Houdusse and colleagues at Institut Curie, we have shown that in muscle myosin, but not in myosin V, there are interactions between the lever arm and the N-terminal subdomain of the motor domain in certain actin-detached-specific orientations of the lever arm. These interactions are not intrinsically strong but instead use just a few relatively weak noncovalent forces. Nevertheless, the ability of these interactions to resist the effort being applied at the active site appears to be amplified by their location on the lever (load) arm relatively far from the fulcrum (albeit not as far as the tip of the lever arm domain). These interactions may therefore contribute to slowing down the rate of nucleotide product release and thus delay progression of the cross-bridge cycle from actin-detached to actin-bound states. Such a design may suggest a method for stabilizing a composite structure made out of weakly attracting materials.

A Myosin-inspired Expanded View of the Lever. A fulcrum need not be merely a passive place on which to rest and pivot a lever. Myosins appear to enrich both the distance and force amplification functions of a lever by expanding the role of the simple fulcrum to that of a “control panel” or fulcrum board. In this view, the lever system consists of a fulcrum board that, depending on the types of features it contains, can interact with the lever arms in a variety of ways. The arm-interacting features that have so far been identified on the fulcrum board include simple fulcrums, flexible joint producers, effort arm attractors and resistance arm attractors. The fulcrum board rigidly connects these features, which is crucial for the functions of these levers. In myosins, the N-terminal subdomain serves as the fulcrum board for this molecular lever. This board-based picture of the lever provides an easy-to-modify template: One can simply add on to the fulcrum board various combinations of the features in particular geometric arrangements to accomplish a desired and perhaps novel mechanical effect.

Mechanically Robust Structures

Over the past few years, techniques such as atomic force microscopy and molecular dynamics have begun to reveal the sequence of events that occur when proteins are put under mechanical stress. These methods are, of course, incompletely developed; they are limited to experiments that are currently feasible (such as pulling from the N and C-termini) and do not catch the full range of mechanical stresses a protein would encounter naturally. Nevertheless, these experiments show that proteins undergo a variety of responses to mechanical stretching, with one of the found observation being a sequential breaking of certain types of interactions.

The resistance within a protein to mechanical unfolding varies between different types of domains, secondary structural elements and bond types. For example, titin, the elongated multidomain constituent of the vertebrate striated-muscle sarcomere, consists of domains with varying unfolding rates; Kathryn Scott, of the University of Oxford, and her colleagues have suggested that the location of strong domains adjacent to weak ones reduces the chances of misfolding upon release of the external force. Beta-sheet domains are mechanically stronger than other types, and within a beta sheet the terminal parallel beta strands are particularly resistant to unfolding. Root’s simulations of mechanical stress on an alpha-helical coiled coil from myosin suggest that main-chain ?-bonds break before hydrophobic bonds. Certain salt links are the last to succumb to mechanical disruption within a chain of this myosin alpha-helical peptide, and Nathan Duff, N.-H. Duong and Daniel Lacks of Case Western Reserve have shown this connection between beta strands of the Ig27 domain of titin. Of course, these computer simulations need experimental verification; nevertheless, the attraction of oppositely charged long side chains such as between glutamate and arginine or lysine may flexibly but strongly tether two main-chains to each other. As Theodor Ackborow and colleagues at MIT have noted, proteins can thus display a hierarchal system of interactions to accommodate different levels of mechanical stress.

Taken together these observations begin to suggest that “robust” proteins are often designed to place elements with varying types of responses to mechanical stress vrithin close proximity of one another. This type of design permits proteins to access both rigid conformations, necessary, say, for enzymatic functions, and to reversibly access flexible or partially unfolded conformations necessary for allosteric effects, motile activities and elastic integrity. A goal of constructing structures that form a well-defined shape under normal operating conditions and that bend but do not break under conditions of high stress is one that can be of use at various scales of size.

Into the Mainstream

Biomimicry can now be extended to include proteomimicry as a subspecialty. Just as examination of sharkskin suggested how simple changes to the arrangement and orientations of riblets can affect fluid dynamics, examination of protein structures described here suggests how changes in the pattern and sizes of tenons and mortises or changes in the positions of elements of a lever can alter the strength and flexibility of joints. These examples chronicle the types of investigative pathways that can be used to extract novel engineering principles from the structures of proteins. Verifying these mechanical designs, investigating their applications and searching for new designs in other proteins are subjects of future endeavors.