Nanomedicine, Volume I: Basic Capabilities

© 1999 Robert A. Freitas Jr. All Rights Reserved.

Robert A. Freitas Jr., Nanomedicine, Volume I: Basic Capabilities, Landes Bioscience, Georgetown, TX, 1999


2.3.1 Biotechnology

The first broad category of enabling technology for molecular manufacturing in the mechanical tradition is biotechnology.10,322 Molecular biologists study and modify systems of molecular machines, and genetic engineers reprogram these systems, sometimes to build novel molecular objects having complex functions. By 1998, biotechnologists could make almost any DNA sequence or poly-peptide chain desired (though not extremely long aperiodic ones), assemble "devices" such as artificial chromosomes and viruses, and were actively pursuing gene therapy, all of which are examples of molecular engineering. While earlier writers had suggested that biomolecules could be used as mechanical components,2414 Drexler182 evidently was the first to point out, in 1981, that complex devices resembling biomolecular motors, actuators, bearings, and structural components could be combined to build versatile molecular machine systems analogous to machine systems in the macroscopic world (Table 1.3). "Development of the ability to design protein molecules," Drexler wrote, "will open a path to the fabrication of devices to complex atomic specifications, thus sidestepping obstacles facing conventional microtechnology. This path will involve construction of molecular machinery able to position reactive groups to atomic precision."

Perhaps the best-known biological example of such molecular machinery is the ribosome (Fig. 2.2; see also Section, the only freely programmable nanoscale assembler already in existence.* In nature, ~8000 nm3 ribosomes act as general-purpose factories building diverse varieties of proteins by bonding amino acids together in precise sequences under instructions provided by a strand of messenger RNA (mRNA) copied from the host DNA, powered by the decomposition of adenosine triphosphate (ATP) to adenosine diphosphate (ADP). Each ribosome is a compact ribonucleoprotein particle consisting of two subunits, with each subunit consisting of several proteins associated with a long RNA molecule (rRNA). Biotechnology can employ the ribosomal machinery of bacteria to produce novel proteins, which proteins then might serve as components of larger molecular structures.182

* D. Heidel notes that there are three distinct types of ribosome -- eubacterial, archaebacterial and eukaryotic -- and multiple subvarieties among the ~108 different species on Earth; additionally, other bioassembler units such as polyketide synthases3224 also exist but these enzymes generally only make one type of product per assembly.

Of course, such biological mechanisms have been disparaged as "slipshod devices working in spite of haphazard design." Dan Heidel, a University of Washington biomimetics engineer, agrees that biological systems are sloppy but observes that they are designed to tolerate it, and offers high praise for the lowly ribosome:

"The average ribosome sits in an aqueous environment surrounded by thousands of cosolutes at concentrations just a hair's breadth from precipitating out of solution. The ribosome itself consists of three RNA molecules that spontaneously fold into floppy spaghetti noodle piles of chemically active groups surrounded by >50 proteins of similarly dubious structure. The other solutes are bombarding this motley assemblage at tens to hundreds of miles an hour and the solvating water molecules are a constantly shifting and unpredictable network of hydrogen bonds and polar ionic charges constantly impinging upon the ribosome. The ribosome must recognize one particular cognate transfer RNA (tRNA) for the codon in its A-site out of ~60 other nearly identical tRNAs. This particular tRNA must be discriminated from all its tRNA brethren along with whatever other solutes can fit into the A-site pocket using only three base pairs for discrimination. In fact, most codons use only two of the three base pairs for recognition so some tRNAs are recognized by as few as 4 hydrogen bonds. All of this requires sub-Angstrom precision in the active sites but since the ribosome is a massive complex of subunits held together by only ionic, hydrogen, hydrophobic and Van der Waals bonds, it's like getting submillimeter accuracy from points on a bunch of taped-together water balloons being rolled down a staircase."

"This ridiculous machine can assemble proteins at a 20 Hz subunit-incorporation frequency with a 103 error rate. Ribosomes are even more impressive when one considers that the number of amino acids in each ribosome's own protein subunits combined with the 103 protein assembly error rate ensures that each ribosome has several errors in it and that each cell almost certainly has no two ribosomes exactly alike. I dare any engineer on this planet to rationally design an assembler at any scale that is able to operate at a comparable level of performance under similarly adverse operating conditions."

Thus ribosomes can manufacture, with >99.9% fidelity to arbitrary specifications, linear strings of amino acids of great length which may then fold up to produce three-dimensional protein structures performing a broad range of highly differentiated functions, providing good analogs for working nanomachines capable of forming self-assembling systems. Much of the great diversity of protein activity flows from the selectivity with which they bind to specific receptor sites on other molecules. Some of these binding mechanisms might be copied to construct molecular robotic arms, while other devices could be made by adapting naturally-occurring proteins, or by synthesizing new, custom-built ones.2393 Protein engineering might be used to create proteins with desired qualities as building blocks for constructing the first primitive assemblers.

The field of protein engineering has its own journal of the same name, and has already achieved remarkable results in the synthesis of novel structures. Drexler10 cites the creation of the first de novo structure by DeGrado,2394 the development of synthetic branched protein-like structures that depart substantially from protein models,2395 and the engineering of a branched, nonbiological protein with enzymatic activity.2396 Since then, numerous alpha-helical peptides* have been designed de novo,2410-2413 a 20-residue artificial peptide exhibiting designed beta-sheet secondary structure has been created,2398 and various peptide analogs have also been synthesized.2399 By 1998, advances in computational techniques allowed the design of precise sidechain packing in proteins with naturally occurring backbone structures, and the study of de novo backbone structures had begun.2400 The catalytic task space766 was being laboriously investigated at the molecular level -- for example, one study of comparative enzyme structures revealed that changing as few as four amino acids converted an oleate 12-desaturase to an oleate hydroxylase, and as few as six substitutions could convert an oleate hydroxylase into an oleate desaturase.2401 To further expand the toolkit, Mills2431 and Fahy322 proposed exploiting the degeneracy of the genetic code (wherein amino acids are specified by up to 6 different codons; Chapter 20) by assigning some codons to unnatural amino acids, allowing the creation of artificial proteins with unprecedented structural or catalytic properties. With 61 usable codons available, there is room for up to 41 novel amino acids in addition to the natural 20 amino acids; unnatural amino acids have been introduced into beta-lactamase and T4 lysozyme enzymes in a site-directed fashion, using artificial tRNAs and mRNAs in an in vitro translation system,2402,2403 and into an engineered peptide as a reporter group.2412 Additionally, artificial novel base pairs can be incorporated enzymatically into DNA2404,2607,2932 and other schemes are possible.2676 By incorporating one additional base pair, the number of possible codons is expanded from 43 = 64 codons to 63 = 216 codons, potentially allowing up to 193 novel amino acids to be incorporated into artificial protein structures -- or at least 48 new amino acids assuming that present levels of redundancy are retained and the existing 4-letter code is kept as a subset for "upward compatibility".2431 If the genetic alphabet is increased to eight bases (83 = 512 codons), the number of available codons for novel amino acid-like molecules increases to as many as 489.322

* Chains of amino acids shorter than ~100 residues are customarily called peptides; longer chains are proteins.

One of the greatest challenges in protein engineering (where one objective is to create functional, atomically-precise 3-D aperiodic structures) has been the difficulty of protein fold design, because individual amino acids have no strong, natural complementarity.2392 In 1998, Duan and Kollman2324 successfully folded a solvated 36-residue (~12,000-atom) protein fragment (i.e., a peptide) by molecular dynamics simulation into a structure that resembles an intermediate native state. During a 1-microsec simulation, the chain folded during 150 nanosec into a compact structure resembling the intermediate native state, as known by NMR, then unfolded and refolded again for a shorter period. In 1998, whole-protein folding to the final stable native state had not yet been computed deterministically using molecular dynamics -- the real protein will fold and refold hundreds of times before it stumbles into the stable conformation with the lowest free energy2405 -- but research pursuing this objective was active and ongoing.2406,2407 Structural changes that occur during protein function (e.g., enzymatic action) were also being avidly studied.2408

A different protein-based approach to near-atomically-precise 3-D structures avoids the folding problem by making use of semi-rigid protein nanoshells of deterministic size and shape to guide the ordered assembly of inorganic particles. For example, the capsid of the cowpea chlorotic mottle virus, a protein container with a 28-nm external diameter and an 18-nm internal cavity diameter, is composed of 180 identical coat protein subunits arranged on an icosahedral lattice. Each subunit presents at least nine basic residues (arginine and lysine) to the interior of the cavity, creating a positively charged interior surface which provides an interface for inorganic crystal nucleation and growth, while the outer capsid surface is not highly charged. In one experiment,2391 the empty capsid shell was found to act as a spatially selective nucleation catalyst in paratungstate mineralization, in addition to its role as a size- and shape-constrained reaction vessel. Noted the authors: "The range of viral morphology and size allow great flexibility for adapting this methodology to control the size and shape of the entrapped material, which is limited only by access to the protein interior and could include inorganic and organic species. The electrostatic environment within the cavity could be altered by site-directed mutagenesis to induce additional specific interactions."

It may also be possible to borrow various existing protein devices (Table 1.3) and apply them to new uses. The bacteriophage DNA injection system (Section 9.2.4) is self-assembling and could be engineered for other applications. Vogel2425,2426 has attached kinesin motors to flat surfaces in straight grooves; when ATP was added, the motors were activated, passing 25-nm wide microtubules hand over hand down the line in the manner of a ciliary array (Section 9.3.4). Montemagno2278,2426 genetically engineered the 12-nm wide ATPase molecular rotary motor so that one end would adhere to a metal surface and the other end would provide an attachment site for 1-micron fluorescent streptavidin-coated bead payloads. When ATP was added, each bead individually attached to a biomotor in the motor array began twirling at ~10 Hz, generating a >100 pN force. Bead rotation was maintained continuously for more than 2 hours, indicating the high reliability of these ~100% efficient self-assembling biomotors. Montemagno explains that his laboratory's "long-term goal is the integration of the F1ATPase biological motor with nanoelectromechanical systems (NEMS) to create a new class of hybrid nanomechanical devices." Other natural nanomotors such as the self-assembling bacterial flagellar motor578-581 and the hexagonal "packaging RNA" or pRNA pumping mechanism that packs DNA into the capsid shell of the bacteriophage Phi 291723 are also being intensively studied.

Yet another material which allows moderately well-defined three-dimensional nanoassemblies is DNA. The ideas behind DNA nanotechnology have been around since 1980,2414 but activity in this field accelerated in the 1990s after numerous experimental difficulties were surmounted. Horn and Urdea2606 reported branched and forked DNA polymers. Niemeyer1905 and Smith2430 exploited DNA specificity to generate regular protein arrays, and both2430,2444,2544 suggested using self-assembled DNA as an early material for molecular nanotechnology. Shi and Bergstrom2416 attached DNA single strands to rigid organic linkers, showing that cyclical forms of various sizes could be formed with these molecules. Mirkin's group1904 attached DNA molecules to 13-nm colloidal gold particles with ~6 nm particle spacings, with the goal of assembling nanoparticles into macroscopic materials; Alivisatos, Schultz and colleagues2415 used DNA to organize 1.4-nm gold nanocrystals in arrays with 2-10 nm spacings. In either case, the specificity of DNA pairing should allow the construction of complex geometries since the use of colloidal particles can potentially add structural elements (which are more rigid than any polymer strand) to the set of building blocks for nanometer-scale structures -- although to take full advantage of this capability would require particles which are atomically precise and which possess several chemically distinct anchoring points on each particle. Damha has synthesized V-shaped and Y-shaped branched RNA molecules,2566,2567 branching RNA dendrimers with "forked" and "lariat" shaped RNA intermediates,2568 and even trihelical DNA.2569 Henderson and coworkers2706,2707 have designed a simple DNA decamer that can form an extended linear staggered quadruplex array reaching lengths of >1000 nm, and have made branched oligonucleotides that template the synthesis of their branched "G-wires"; depending on the ratio of linear to branched building blocks, extensive DNA arrays with differing connectivities but irregular interstices can be created.2704

In 1998 the most intensive work on three-dimensional engineered DNA structures was taking place in Nadrian Seeman's laboratory in the New York University Department of Chemistry. Seeman originally conceived the idea of rigid 3-D DNA structures in the early 1980s,2414,2419 while examining DNA strands that had arranged themselves into unusual four-armed Holliday junctions.3154 Seeman recognized that DNA had many advantages as a construction material for nanomechanical structures.1916 First, each double-strand DNA with a single-strand overhang has a "sticky end," so the intermolecular interaction between two strands with sticky ends is readily programmed (due to base-pair specificity) and reliably predicted, and the local structure at the interface is known (sticky ends associate to form B-DNA). Second, arbitrary sequences are readily manufactured using conventional biotechnological techniques. Third, DNA can be manipulated and modified by a large variety of enzymes, including DNA ligase, restriction endonucleases, kinases and exonucleases. Fourth, DNA is a stiff polymer in 1-3 turn lengths2421 and has an external code that can be read by proteins and nucleic acids.2422

During the 1980s, Seeman worked to develop strands of DNA that would zip themselves up into more and more complex shapes. Seeman made junctions with five and six arms, then squares,2417 stick-figure cubes comprised of 480 nucleotides,1914 and a truncated octahedron containing 2550 nucleotides and a molecular weight of ~790,000 daltons.1915 The cubes (Fig. 2.3) were synthesized in solution, but Seeman switched to a solid-support-based methodology2418 in 1992, greatly improving control by allowing construction of one edge at a time and isolating the growing objects from one another, allowing massively parallel construction of objects with far greater control of the synthesis sequence. By the mid-1990s, most Platonic (tetrahedron, cube, octahedron, dodecahedron, and icosahedron), Archimedean (e.g., truncated Platonics, semiregular prisms and prismoids, cuboctahedron, etc.), Catalan (linked rings and complex knots), and irregular polyhedra could be constructed as nanoscale DNA stick figures.

Seeman's DNA strands that formed the frame figures were strong enough to serve as girders in a molecular framework, but the junctions were too floppy. In 1993 Seeman discovered the more rigid anti-parallel DNA "double crossover" motif,1920 which in 1996 he used to design and build a stiff double junction to keep his structures from sagging.2420 The next goal was to bring together a large number of stick figures to form large arrays or cage-shaped DNA crystals that could then be used as frameworks for the assembly of other molecules into pre-established patterns. These DNA molecules would serve as the scaffolding upon which new materials having precise molecular structure could be assembled.

In 1998, Erik Winfree and colleagues in Seeman's laboratory reported the design and construction of two-dimensional DNA crystals using self-assembling double crossover molecules.1970 Repeating array units were approximately 2 nm x 4 nm x 16 nm in size, and examination of the array with an Atomic Force Microscope (AFM) revealed domains of up to 500,000 interconnected units, showing that the self-assembly process (Section 2.3.2) could be very reliable under ideal conditions. While the work described the construction of two- and four-unit lattices, the number of component tiles that could be used in the repeat unit did not appear to be limited to such small numbers, suggesting that complex patterns could be assembled into periodic arrays, which might also serve as templates for nanomechanical assembly. Noted the authors: "Because oligonucleotide synthesis can readily incorporate modified bases at arbitrary positions, it should be possible to control the structure within the periodic group by decoration with chemical groups, catalysts, enzymes and other proteins, metallic nanoclusters, conducting silver clusters, DNA enzymes, or other DNA nanostructures such as polyhedra." In these experiments, DNA hairpin turns incorporated within selected units inside their structures were visible as topographic features under AFM imaging, proving the ability to build predictable, atomically precise 2-D crystals with design control over every lattice point. Since the density of lattice points was much sparser than atomic spacings, this DNA self-assembly technology could be paired with atomically precise synthesis of diverse nanoscale subassemblies so that each unique double crossover unit in a pattern could be decorated with a unique subassembly of comparable size, possibly allowing the construction of mechanical nanocomputers1970 and other nanomachinery.

In early 1999, Seeman reported yet another breakthrough -- the construction of a mechanical DNA-based device that is a possible prototype for a nanoscale robotic arm.2409 The mechanism has two rigid arms a few nanometers long that can be rotated between fixed positions by introducing a positively charged cobalt compound into the solution surrounding the molecules, causing the bridge region to be converted from the normal B-DNA structure to the unusual Z-DNA structure. The free ends of the arms shift position by ~2.0 nm during the structural conversion. Explained Seeman: "Using synthetic DNA as a building material, we have constructed a controllable molecular mechanical system. In the long-term, the work will have implications for the development of nanoscale robots and for molecular manufacturing."

DNA can also serve as an assembly jig in solution phase. Bruce Smith and colleagues2423 are devising a method for the assembly and covalent linkage of proteins into specific orientations and arrangements as determined by the hybridization of DNA attached to the proteins, called DNA-Guided Assembly of Proteins (DGAP). In this method, multiple DNA sequences would be attached to specific positions on the surface of each protein, and complementary sequences would bind, forcing protein building blocks (possibly including biomolecular motors, structural protein fibers, antibodies, enzymes, or other existing functional proteins) together in specific desired combinations and configurations, which would then be stabilized by covalent interprotein linkages. This technique could also be applied to nonprotein components that can be functionalized at multiple sites with site-specific DNA sequences, although proteins, at least initially, may be more convenient building blocks due to their size, their surface chemistry, the wide variety of functions and mechanical properties they can confer on the resulting assemblies, and the many existing techniques for introducing designed or artificially evolved modifications into natural proteins of known structure. (In 1998, custom DNA and peptide sequences could be ordered online.)2424 Methods for covalently attaching functional proteins to a DNA backbone in a specified manner at ~8.5 nm (25 base-pair) intervals,2430 addressable protein targeting in macro-molecular assembly,2544 and "protein stitchery"2848 were being explored by others. Drexler3208 notes that evolution has not maximized the stability of natural proteins, and that substantially greater stability may be engineered by various means (e.g., increasing folding stability by >100 Kcal/mole.


Last updated on 6 April 2003