Foresight Update 28
Page 3
A publication of the Foresight Institute
 Recent
Progress: Steps Toward Nanotechnology
by Jeffrey Soreff
 Advances in Parallel Techniques
Advances in nanotechnology require techniques for creating
complex, 3D, atomically precise structures. One general strategy
towards forming these structures is to construct many molecules
in parallel, synthesizing macroscopic amounts of materials which
are designed to assemble themselves into the desired structures.
The following four papers describe advances in these techniques.
Designing high-affinity ligands
In the first paper on techniques for synthesizing macroscopic
quantities of materials, S. B. Shuker et. al., writing in [Science
274:1531-1534 29Nov96] describe a systematic method for
designing high-affinity ligands using information from NMR. Their
technique essentially builds up a composite ligand piece by
piece, with excellent control of the detailed geometry of the
protein/ligand interface. In this paper they built composite
ligands out of two pieces, one of which bound to the protein with
Kd = 2.0 mM and the other of which bound with Kd
= 100 mM, to yield 5 composite ligands with affinities in the
nanomolar range. Their procedure ensures that they can identify
initial ligands that bind in two different places, then
synthesize a link that will not interfere with the binding in
either area. Their method has five steps:
- Screen compounds for a ligand that binds to the proteins.
In all cases, binding is evaluated by using 15N
labeled protein together with an NMR spectroscopy
technique called 15N-heteronuclear single
quantum correlation (HSQR). "These spectra can be
rapidly obtained, making it possible to screen a large
number of compounds." Because this technique only
"sees" atoms near the 15N atoms,
unbound ligand is invisible to it and even weak binding
can be detected without prohibitive background signal
from unbound ligand.
- Generate and evaluate derivative compounds to optimize
the binding of the first ligand.
- In the presence of enough of the first ligand to saturate
its binding site on the protein, screen for ligands that
bind to another site on the protein. The saturation by
the first ligand ensures that the second ligand will bind
to a different place on the protein than the first
one.
- Generate and evaluate derivative compounds to optimize
the binding of the second ligand in the presence of
saturating quantities of the first one.
- Use structural information from the NMR spectra of the
doubly bound protein to find the relative positions of
the ligands, allowing design of a link that spans nearby
positions on them and that does not collide with the
protein.
From a molecular engineering perspective, this technique
allows expanding an existing stable structure step by step. At
each step the atoms of the starting structure set up a coordinate
system which the NMR spectra extend to the new ligands. This
might allow an incremental solution to the fold design problem,
by designing layer after layer of ligands to bind to an existing
stable structure, introducing covalent links after the NMR
spectra had determined the geometry of the noncovalent binding.
This approach would not be sensitive to errors in fold prediction
techniques, since it would rely on experimental data at each
step.
Self-assembling cage accelerates
Diels-Alder reaction
An article in [C&EN p50 20Jan97] describes the
presentation of the James Flack Norris Award to Julius Rebek Jr.
Amongst other notable work, the article describes Rebek's work on
hydrogen bonded organic structures with cavities. C&EN says
that "Such self-assembling superstructures are of tremendous
interest for nanotechnology."
An example of the properties of one of these structures is
described in a recent article by Rebek and Kang [Nature 385:50-52
2Jan97], the second article described in this section. They
describe the acceleration of a Diels-Alder reaction by
encapsulation of the reactants in a dimeric capsule. The capsule
is assembled from a rather complex compound, containing a
primarily linear array of 14 fused rings with quite a few
functional groups, primarily containing amides but also
containing two hydroquinones. In 3D space, the molecule curves
into a "C" shape. Two of these molecules form the
capsule, where "Intermolecular hydrogen bonds hold the two
subunits together in much the same manner that the stitches along
the seam hold a baseball together." Previous work by the
authors had shown "that two molecules of solvent benzene are
accommodated inside [the capsule], [which] raised the possibility
of the use of these capsules as chambers for bimolecular
reactions." The particular reaction discussed is the
Diels-Alder addition of p-quinone to cyclohexadiene. When both
reactants are present at 4mM concentrations and the capsules are
present at 1mM concentration the p-quinone is rapidly detectable
as an encapsulated species via NMR spectroscopy. Under these
conditions the p-quinone reacts to form the Diels-Alder adduct
with a time constant of about one day. By contrast, the same
concentrations of the two reactants, in the absence of the
capsules, have a reaction time constant of roughly a year. The
reaction is accelerated roughly 200-fold.
|
"Self-assembling
superstructures...of
tremendous interest for nanotechnology." |
|
The authors did a number of control experiments to exclude
several possible effects of the capsules other than geometrical
confinement. They excluded the possibility of direct hydrogen
bonding of one of the capsule molecule's hydroxyl groups to the
reactants by synthesizing a molecule similar to the capsule
molecule (with essentially the same functional groups) but with
different global stereochemistry (an "S" shape instead
of a "C" shape) that precluded formation of a capsule.
This variant did not accelerate the reaction. The authors also
methylated the hydroxyls of the capsule molecule, preventing the
formation of the hydrogen bonds that held the capsule together
and again suppressed the acceleration. A variant of the reactants
was also examined. Replacing the p-quinone with napthoquinone (a
larger molecule that does not fit in the capsule) eliminated the
acceleration.
This work is applicable to nanotechnology because it illustrates
the control of a reaction forming covalent bonds via nonbonded
geometrical constraints on the reactants without the use of
enzymes or other biopolymers. The authors write that the
"interior of cage-like molecules can be considered to
provide a new phase of matter, in which it become possible to
stabilize reactive intermediates and to observe new forms of
stereoisomerism." This "new phase" has an absence
of solvent molecules or other non-reactant species and the
presence of a constraining network around the reaction site. Both
of these features are analogous to conditions during machine
phase chemistry in molecular manufacturing.
Stable helical structures with
beta peptides
The third article on parallel techniques discusses beta
peptides. Natural proteins are composed of alpha amino acids,
compounds where the amino nitrogen and carboxylic acid group are
bound to the same carbon atom. Writing in [Nature 385:113-114
9Jan97], B. L. Iverson describes work by the research groups of
S. Gellman and D. Seebach on oligomers of beta amino acids,
compounds where the amino and acid groups are attached to
adjacent carbons. Gellman's group found that a hexamer forms
"well-defined helices in methanol solution and in the solid
state." Seebach's group also found evidence of a helix in
solution. The helices formed are analogous to alpha helices in
natural proteins. In these oligomers they are "stabilized by
hydrogen bonds between every third unit. This folding pattern
results in 14-atom 'rings' being formed by the hydrogen bonds, so
Gellman has called the structure the 14-helix." From an
engineering viewpoint, the attractive feature of these compounds
is the stability of their secondary structure even in very short
polymers. By comparison, the shortest alpha peptide that I am
aware of that forms a well-defined helix is 23 residues long [Science
271: 342-345 19Jan96]. Perhaps the fold prediction and
fold engineering problems will prove to be intrinsically easier
with beta peptides than with alpha peptides. These experimental
results are "extraordinary, especially when one considers
that the extra -CH2- group of the beta-amino acids
might be expected to make the resulting resulting beta-peptides
more flexible than alpha-amino-acid peptides, not more
structured." Each of the two non-carboxylic carbons in a
beta peptide residue can carry a side chain, so the potential
design freedom for these peptides is higher than for alpha
peptides. Further work will tell how much of this design freedom
can be exploited while retaining the stable 3D structures
demonstrated by Gellman's and Seebach's groups.
Trends in combinational chemistry
In the fourth article on parallel methods, J. C. Hogan,
writing in [Nature (supp) 384:17-19 7Nov96]
describes trends in combinational chemistry for drug development.
He describes past attempts to generate drug leads from peptide
and oligonucleotide libraries, but describes them as having been
unproductive pharmocologically. He also mentions a wide variety
of other oligomer libraries which have been produced by solid
phase synthesis, writing: "the synthesis of oligomeric
N-alkyl glycines shown in Fig. 1 is an excellent non-peptidic
example. Using this [solid phase, with a single coupling
chemistry] approach, libraries of oligocarbamates, peptide
phosphonates, vinylogous polypeptides and other oligomeric
scaffolds have been produced." This spectrum of oligomers
would seem to provide a good selection of candidates for stably
folded strands with well defined 3D structures for both drug and
machine part applications, however Hogan goes on to write:
"These molecules generally possess flexible backbones, which
can weaken target binding and can also hinder the application of
structure-guided techniques."
Hogan goes on to describe new techniques where "multiple
variable groups are arranged about a central scaffold or
core." In the examples that he shows, the cores appear to be
fairly rigid in two of the three cases, with aromatic rings
present and only two or three torsional degrees of freedom in the
core. Unfortunately, the variable substituents are joined to the
core with single bonds in all of the cases that he cites, so the
molecules as a whole have quite a bit of flexibility. From a
machine perspective, the current trend in synthetic libraries
would become much more helpful if cyclization reactions between
the substituents were incorporated.
Advances in Sequential Techniques
The other major strategy for building precise structures relies
on sequential operations at some point in the techniques. This
may include sequential scanning probe tip placements in a
tunneling or force microscope or it may involve electron beam
flashes in e-beam lithography. In general, these techniques allow
more predictable geometrical control at the cost of slower
fabrication than in the parallel methods. The three papers in
this section describe advances in these sequential techniques.
STM positioning of buckyballs on a
surface
In the first paper M. T. Cuberes, R. Schlittler, and J. K.
Gimzewski, writing in [Appl. Phys. Lett. 69:3016-3018
11Nov96], describe the reversible
positioning of individual C60 molecules adsorbed on to
a step on a Cu(111) surface at room temperature. In a sense,
this work is complementary to this group's earlier work on
porphyrin manipulation. In that work the structure and
flexibility of the mobile molecule, the porphyrin, helped control
the binding to the surface, permitting controlled motion while
avoiding thermal diffusion. In this case, the C60
molecules are rather rigid, but the Cu(111) surface contains a
step edge that confines them to motion in one dimension. In
addition "small kinks [in the step edge] are noticeable from
the misalignment of the C60 molecules with respect to
their neighbors ... The formation of kinks at the Cu steps around
the C60 molecules increases the coordination of the
molecule with Cu and hinders its diffusion along [emphasis
added] the step edge." The authors demonstrated that an
"STM tip can separate a C60 molecule from a
molecular chain adsorbed at a monatomic Cu(111) [step] and shift
it controllably and reversibly back and forth without
significantly altering the position of the other atoms in the
chain." This implies that STM tips can be remarkably clean,
with sufficiently sharp surfaces on both sides that
"pushing" molecules from either side can be done with
little disturbance to adjacent molecules. This is not what one
would expect, for instance, if STM tips always depended on a
single critical atom on what otherwise was approximately a 100 nm
sphere. This is a hopeful sign for many types of attempts to use
scanning tips to build precise structures.
|
"...our results show
promise for further advances in
bottom-up fabrication and operation of devices..."
|
|
There are limitations to the positioning in this system. The
authors found that "only single molecules can be
controllably repositioned in the current system. Attempts to move
more than one molecule at a time distort the molecular
rows." This, however, sounds like it is only a mechanical
instability in a compressed row of spheres rather than a
limitation on the precision with which the tip can apply forces.
The description of the applicability to nanotechnology is best
left in the original authors' words. They conclude:
"Therefore, STM-aided manipulation can be used to fabricate
a functional counting device based on the abacus mechanism at the
molecular scale. Although the use of nanomechanics at the
molecular level is still at an early stage of development, our
results show promise for further advances in bottom-up
fabrication and operation of devices with dimensions on the level
of several nanometers."
Lithography-patterned template
controls 3D assembly of particles
In the second paper, A. van Blaaderen, R. Ruel, and P.
Wiltzius, writing in [Nature 385:321-323 23Jan97],
describe a technique for controlling the 3D assembly of colloidal
particles by initiating the assembly on a 2D patterned template.
Their template was "a 500-nm-thick fluorescent polymer, with
holes made with electron beam lithography." They deposited
fluorescent silica spheres with a radius of 525 nm by gravity on
their polymer template. In the absence of a template,
"hard-sphere-like colloidal dispersions are known to
crystallize with a random stacking of close packed planes."
In this experiment, a (100) pattern of holes in the polymer
caused the formation of a pure face centered cubic (FCC) crystal.
The 2D template controlled the long range order of the 3D
crystal. The (100) slice through the crystal was chosen rather
than a denser (111) slice because close packed (111) layers can
stack on top of each other in either of two possible positions,
creating the possibility of twinning at each layer. In contrast,
a (100) layer of hard spheres only permits one possible position
for the next (100) layer. "In other words, there is no
twinning possibility along this growth direction."
The authors also experimented with mismatched lattices, showing
that templates with mismatched spacings generated defects which
gradually converted the lattice to a random close packed one.
Another experimental variation was to leave a gap between two
patterned regions. When the gap was 11 diameters wide, a
hexagonal region appeared between the two FCC regions.
Manipulations like this might inject a considerable amount of
information into the crystal. For example they might introduce
twinning boundaries (at an angle to the template) at selected
locations in the crystal. The authors suggest that their
technique could be extended to charged spheres as well, using a
charged template rather than purely hard-sphere-like repulsion.
While the present work does not create an atomically perfect
structure, the substitution of virus particles, with well defined
structures (and presumably with well defined interparticle
contacts in a crystal) would permit lithographic control of a
long range atomically perfect structure. This might permit
molecular manufacturing to take advantage of the large investment
that has been made in fine-line lithography.
Exponential replication of molecular bilayers
The third paper is not directly about an advance in a
sequential technique, but rather describes an amplification
technique which might improve our ability to exploit sequential
techniques. Writing in [Nature 384:150-153
14Nov96], R. Maoz et. al. describe exponential replication of a
stack of partially condensed alkyl siloxanol layers. At each step
they have bilayers of n-octadecylsiloxanol (formally CH3-(CH2)17-Si(OH)3,
but with partial dehydration of the -Si(OH)3 groups, forming
lateral covalent Si-O-Si bridges within the layers). Alternate
layers have their hydrophilic -Si(OH)3 groups pointing
up and pointing down. Each replication is done in two steps.
First the stack is treated with wet acetone. This places water
molecules between the hydrophilic siloxanol groups. Next, the
stack is treated with n-octadecyltrichlorosilane, CH3-(CH2)17-SiCl3.
This enters the stack and reacts with the water to form
additional n-octadecylsiloxanol, which inserts a new bilayer in
between each existing pair of bilayers. The authors write that
"A stack of preformed bilayers thus functions as a set of
independent template units which define the discrete spatial
distribution of the water incorporated into the film, while
providing a succession of distinct polar interfaces, each of
which is capable of sustaining the spontaneous self-assembly of a
similarly structured bilayer." The authors also found that
these layers were mechanically robust, with AFM examination
showing "no defects, such as holes or steps", and
finding that "no surface defects could be induced by the
tip."
Ideally, it would be helpful if this work could be extended to
allow the accurate complementary replication of layers containing
mixed siloxanols. This would essentially yield a 2D analog to
PCR. For example, if several different kinds of hydrophilic
groups could bind selectively at the polar/polar interface of a
bilayer, then an initial lateral pattern could be exponentially
amplified to macroscopic quantities. This would greatly enhance
the usefulness of the atomically precise patterns that can be
produced sequentially with STMs today. One could program them and
amplify them much as nucleic acids are handled today, but one
would not be faced with predicting the folding of a 1D amino acid
sequence into a 3D protein in order to produce a useful
structure.
Foresight Update 28 was originally
published 30 March 1997.
|