DNA-based Molecular Engineering: Complex 3D Structures and Applications
1 Video View·Sep 5, 2023
It is my great pleasure to introduce today's speaker, Dr. Yonggong Ki.
Dr. Ki is currently a post-doctoral researcher at the Dana-Farber-Kemp Institute and the
Wyss Institute for Biology, Inspiration, and Engineering at Harvard.
And he's currently working with Professor William Chi and Professor Sun Ying as well.
And so, he received his Bachelor of Science in Chemistry from Peking University in 1999
and a Ph.D. in Chemistry in Atlanta State.
So, his research lies in the area of DNA nanotechnology with special interest in rational design and
assembly of nanostructure as well as for exploring the application for this kind of
structure for medicine and material research.
And his research kind of work has led to a set of high-profile papers.
Today, Dr. Ki is going to give us a talk entitled DNA-Based Modern Engineering.
So, please join me in welcoming him.
Thanks for a really kind introduction.
And it's my great pleasure to have this opportunity to come here.
It's the first time I've come to your beautiful campus and nice weather today
and to talk about our research at Harvard.
I changed the title a little bit, but anyway, it's basically the same thing.
So, today I'm going to present to you about DNA-Based Molecular Engineering,
especially how to make complex and large three-dimensional DNA structures
and how we are using them and how you might be using them for different applications.
So, first, a little bit of extra information about myself.
I joined Harvard about four years ago and initially I started work with Professor William Shee.
And about two years ago, I also started working with Professor Peng Yin's group.
And both groups actually are doing quite similar work.
We're trying to program molecules to self-assemble into structures.
We especially focus on biomolecular self-assembly.
And we do a lot of interesting research.
You can go to our website, check out our fantastic research there.
But at the heart of our research work is trying to manipulate a DNA molecule
to make them to form artificial structures and use them for applications.
So, for today's talk, I think I'm trying to first cover a little bit of background information
about DNA self-assembly and then I'm going to move on to the second part,
which is going to be the major focus of today's talk is how to design DNA structure
and the state-of-art DNA self-assembly structure.
And last, I'm going to use my past research as an example to discuss with you
about a few applications using DNA structures.
So, the field of DNA nanotechnology, or DNA self-assembly,
is essentially inspired by a biologic system.
And if we look at biology, of course, biology has a super diverse and sophisticated structure and functions.
Just take a tiny cell as an example.
It's essentially a complex biologic machine.
It has thousands of small parts.
Each part has a specific structure and performs really delicate functions.
And what's also amazing to us is such complex machines are self-assembled from small molecules,
for example nucleotides or amino acids,
they assemble to DNA large molecules like DNA or protein
and eventually hierarchically assemble into the final cell structure.
So, based on this observation, one question, an important question to ask ourselves is,
can we design artificial DNA or protein or other biomolecules self-assembly
to try to achieve these two goals?
First of all, can we rationally design a complex structure?
That's the foundation to achieve functions.
Secondly, on top of that, can we add functionalities to those structures?
Of course, you can choose different molecules to do that, like protein or other biomolecules.
But we chose DNA, so why did we choose DNA?
Why are we particularly interested in the DNA molecule?
The basic reason is, first of all, if you forget about anything else,
in order to do a complex self-assembly, the first requirement, of course, is the molecule itself.
It has to be able to store a large amount of information in its sequence.
So DNA can do that.
We know DNA can store a large amount of genetic information
and eventually that information translates to complex all kinds of structures.
So DNA can store a large amount of information in its sequence.
The second important reason is DNA is relatively easy to program
and the structure is highly predictable.
This is especially true compared to protein.
Because amino acids, protein, can also store a large amount of information in its sequence.
So you can also use DNA to engineer a sophisticated structure.
But unfortunately, protein structures are much more complex than DNA structures.
For DNA, it's really simple.
There are only four DNA bases, as we all know.
There's only A, T, and G, C bases.
And according to Watson-Crick base pairing, A pairs with T, G pairs with C.
So we always follow that simple rule.
And secondly, DNA simply is a duplex structure in nature.
We know all the parameters of such structures.
So it's really simple.
It makes it easy to program, easy to construct.
So that's two most important reasons why we're interested in DNA.
But in addition to that, of course, DNA has its biologic relevance.
And also DNA is used to modify the many enzymes available.
But those are two most important reasons.
So we try to treat DNA simply as building material here.
Just forget about its biologic function for a moment.
So if we treat DNA as material, how do we design structure?
The whole field is founded by this guy from NYU, New York University, about 30 years ago.
He proposed this idea of using DNA as a smart molecule to engineer 2D or 3D structures.
So the basic idea is really simple.
So the natural DNA is simply a duplex.
So in order to build 2D or 3D structure, of course, you need to include more than two strands in your system.
And in addition to that, you want to generate branches of motifs.
In this case, this tiny structure consists of only four single-strand DNA.
But instead of fully complementary to each other, each single-strand DNA here has two binding domains.
For example, the green color strand has binding domain 1 and domain 2 star.
And domain 1 is complementary to the domain 1 star on the orange strand.
And domain 2 star is complementary to the domain 2 on the pink color strand.
And if you design a system this way, you can form a small two-dimensional object.
Of course, this is really simple, but it delivers an important idea of how you design 2D or 3D structure.
Imagine you have a system.
Without knowing how to do that, imagine you have a system that contains hundreds of strands.
You can make larger 2D or 3D structures.
So the field sometimes is called structured DNA nanotechnology.
At the heart, the center of the field, it's trying to design artificial structures.
And it has evolved into four different research branches.
The first research branch, of course, is to develop design strategy,
allow us to control the shape, the size, and the sequence of our DNA structure.
And in addition to that, we're also trying to study the assembly,
trying to improve the folding yield and folding structure stability,
and try to understand the kinetics behind DNA self-assembly.
And in addition to those two research branches,
we're also trying to use DNA structure to template other molecules.
The idea is to add additional functions to our DNA structure.
We can template biomolecules, such as proteins,
and small molecules, organic molecules, inorganic molecules, and nanoparticles.
Now you can have the functionalized DNA structure.
Once you have those functionalized DNA structures,
you can then address many applications,
ranging from biophysics study to making biomedical devices,
even making nanoplasmonic devices, and many other applications.
By looking at this picture, you can roughly divide this four research branches
into two categories, as I mentioned moments ago.
The first challenge we face is trying to achieve rational design complexity.
So these two research branches are trying to address that challenge,
trying to design structure, trying to understand self-assembly.
And the templating application part, of course,
is trying to achieve functionality and trying to address applications.
Of course, after 30 years, the field looks quite different
to what it looked like when Nassim first founded the field.
Now we have much better control in terms of designing DNA structures.
Without discussing more details here,
I just want to emphasize today the structure we can build
can be arbitrarily shaped 2D or 3D structure up to 8K to 12,000 base pairs.
That's roughly $5 to $8 in terms of molecular weights.
And in addition to that, those structures are fully addressable.
So it contains hundreds of unique DNA strands.
It means we can address other materials at nanometer precision.
That's really important for applications.
With those sophisticated structures, now we can use them for many applications.
It's really hard to cover all these applications.
Instead, here I highlight a few papers,
high-level, high-profile papers from last year.
The idea is just to give you a brief taste about applications using DNA structures.
And no matter what we do, you can focus on these two things.
All the application relies on these two things.
One is how do we control the molecular shape?
How do we design our structure?
And in addition to that, how do we site-specifically functionalize our structure?
How do we attach other molecules to that structure at specific locations?
So the first paper I want to show to you today
is an artificial membrane channel, or sometimes called nanopole.
This is without discussing more details.
Again, this whole structure is built using DNA.
Each cylinder here is a DNA helix.
It's a three-dimensional structure with a channel in the middle of the structure.
And the structure contains a group of cholesterol molecules on one end of the structure.
And those cholesterol molecules can insert themselves into a liposome particle.
This is a liposome, a lipid particle here.
They can insert themselves into a lipid bilayer and form this artificial membrane channel.
So this is one example using DNA structure to make biomimetic structures,
also for making sensors.
And another example is a multi-dynein kinase protein molecule device.
So the idea here is it also shows it can immobilize a group of dynein or kinase molecules
on a cylinder-shaped, a tube-shaped DNA structure.
And they can control the position of those molecules and the ratio of those molecules.
By doing this, they can study how this device moves on a microtubule.
This is one example using DNA structure to build an artificial device to study biophysics in vitro.
And the third example is a delivery vehicle, sometimes called a nanorobot.
In this case, it's essentially a cage structure.
Inside the cage, it also immobilizes a group of antibodies.
And the cage has two triggers here.
When a certain cell line expresses a certain antigen, it actually triggers the opening of this cage
to expose antibodies and kill the cell.
That's an example potentially using DNA for biomedical research or applications.
The last example is a device based on plasmonic effect.
So in this case, again, a tube structure was used.
And it also shows it can immobilize a group of particles to form this chiral shape.
And by doing this, because we can control the position of those core particles,
this shows it can tune the optical property of such a device.
This is one example of using DNA structure for immobilizing or manipulating inorganic material
to achieve different functions for making different devices.
So for my own research, of course, it also can divide into two categories.
The most important part of my research is actually just making DNA structures to study DNA self-assembly.
I have done a lot of research work, and you can see a lot of publications here.
But today we're going to just focus on this one.
This is a newly published paper from last year.
It allows us to design really sophisticated three-dimensional discrete objects
or large micro-scale periodic arrays.
And the second part of my research is using them for applications.
It can, again, roughly be divided into two sub-categories.
One is programming biomolecules.
Secondly is manipulating inorganic materials.
Today we're going to focus on this part.
We're also going to discuss a few applications manipulating biomolecules really quickly.
So now we move on to the second part, which is DNA self-assembly.
So I don't want to spend too much time to discuss all the publications in the field.
One thing I want to mention to you is even though the field has developed many sophisticated structures,
as shown here, one thing missing is a modular approach for designing large and complex three-dimensional structures.
A large approach has many advantages.
We will see it in the following slides.
But in order to do this, what essentially we want to achieve are these three things here.
In order to do so, we want a method to use more standard components, or you can call it bricks.
And in addition to that, the assembly needs to follow standard local interaction rules.
So every brick only interacts with its neighbors.
And at last, of course, the method has to be able to produce large and complex structures.
Okay, I'll quickly go over this slide.
So even though we're going to discuss this newly developed modular design self-assembly strategy,
I want to mention another technology available out there, which is really popular.
We use it a lot for many applications.
It's called scaffold DNA self-assembly, or sometimes referred to as DNA origami.
Basically, you take a really long single-strand DNA, which we call scaffold,
mix it together with hundreds of short synthetic oligos, which we call staple strand.
Each staple here has a specifically designed sequence.
Allow it to hybridize to multiple locations onto the scaffold.
And they work collectively in solution and fold this scaffold into a pre-scrapped two-dimensional shape.
So Paul Rossmann invented this method in 2006.
He demonstrated many sophisticated two-dimensional shapes using M13 genomic DNA,
which contains 7,249 nucleotides.
He made many sophisticated 2D shapes.
And he liked to refer to this method as DNA origami, which is a really catchy name.
I really like it because I really like to compare the capability of making arbitrary shapes to the art of paper origami.
And soon after the 2D structure, people extended this approach in making three-dimensional structures,
including cage structure like this cube, tetrahedral, and later more sophisticated container shapes.
Another type of three-dimensional structure I developed at Harvard from our group,
what we call the closely packed three-dimensional structure.
In this case, you pack the helix into this parallel lattice, in this case a honeycomb lattice.
It's a square lattice.
Each cylinder here represents a DNA helix.
So you can build any arbitrary three-dimensional shape using this method.
You can even generate really sophisticated curvature in your structure.
We use it for many applications, as I mentioned.
But as great as the method is, it's not a modular approach,
which makes it complicated to design some sophisticated large structure.
The reason for that is really simple.
I like to compare that method to two simple toys.
One is called Magic Ruler.
The other one is called Cody.
Essentially, it's a really long plastic rod, and you can fold it into different shapes.
These are great and really cute shapes you can make from here.
But if you want to make large structures, people often use another toy, Lego bricks.
With Lego bricks, you can make really sophisticated structures like these buildings.
The reason for that is really simple, because Lego is a modular toy.
So in order to assemble this, you don't have to follow any particular order.
Also, if you make any mistake, you can really easily change it.
Mathematically, it's possible to build this with a really long rod.
But you can imagine it's going to be extremely hard.
And in the real world of self-assembly, a really complicated routing
means a lot of mis-assembly during the self-assembly process.
So we really want to develop this modular design self-assembly strategy.
It has many advantages.
First of all, it's modular design. It's much easier.
And in addition to that, I want to mention,
because we don't have to rely on that M13 scaffold I mentioned in the last slides,
now we have much better control in terms of size and sequence,
because it's really hard to get a really long genomic DNA.
So now with just the short DNA molecules,
we can make much better control in terms of size and sequences.
And in addition to that, we can even choose unnatural nucleosides in addition to DNA.
So we can make other structures also.
So I'm going to skip all the technical details here.
If you have questions later, you can ask me later after the talk.
But the basic thing here for this design is eventually we test many designs,
and eventually we figure out one particular design is each DNA break
equal to just a 32 nucleotide single strand DNA.
And another important design feature is we engineer each break
to have a particular sequence here.
And between each two neighboring break, as shown here in different colors here,
they form and only form an 8-base pair duplex.
That 8-base pair duplex, if you calculate it,
corresponds to a 90-degree dihedral angle between the two DNA breaks.
You're perhaps wondering what happens to this other part,
because you also have other breaks here.
So each break here interacts with four other DNA breaks.
So every part is forming this helical structure.
And this structure is great, but it's really hard to understand
if you have hundreds of them.
So fortunately for us, we can use a much simpler LIGO model
to represent our DNA breaks and also to indicate their interaction.
You can imagine we have two breaks coming together
to simply form this kind of 8-base pair connection.
So by using this design, we already fulfilled two requirements we mentioned.
One is we have the standard small component, 32 nucleotide DNA breaks.
We also have the standard local interaction rule.
Every two neighboring breaks form an 8-base pair
and form a standard degree dihedral angle.
So what about the third point?
Can we make large and complex structure?
At the end of the day, that's the most important thing.
If you can have a really sophisticated design,
but you cannot make a large structure, what's the point?
So we test many structures with different sizes.
One particular structure we test is this DNA cube structure
that contains 8,000 base pairs.
It's a 10-helix by 10-helix by 80-base pair cube structure.
It is assembled from more than 500 unique breaks.
In order to form this structure, we simply perform a linear process.
You put everything together and raise the temperature to about 80 degrees
and gradually cool it down to room temperature,
and you get your final structure.
We're also testing different linear methods right now,
but we can form structure using this method.
And after this, we use a transmission electron microscope
to image our structure here.
You can see multiple particles here.
If you zoom in, you can see two types of particles,
and each one here corresponds to one projection view for this structure.
For example, this one actually corresponds to looking down to that direction,
so you can see each individual helix.
Each single dot here corresponds to a single DNA helix.
So we can make large structures.
So how about complex shapes?
That's another important thing.
We decided to do something more interesting here.
We decided to fully demonstrate the power of modular design.
One fantastic thing about LEGO toy is
if you have a cube shape made of LEGO like this,
you can easily change it to any other shapes
if you just remove some LEGO bricks.
So we decided to do something similar here.
Before the assembly, you can imagine you can remove some brick from this cube.
Now that would allow you to assemble any structure smaller than this cube.
So that's why we call this cube a 3D canvas.
So we essentially can design any shape based on this cube.
And to facilitate it, to make the design process easier,
we actually wrote a computer program.
I spent a lot of time on that, but I'm not going to discuss it.
Eventually, we were able to convert this actually physical cube
to a virtual 3D canvas in the computer program.
So this virtual 3D canvas contains 1,000 voxels.
A voxel is a 3D pixel.
In this case, a voxel is equal to an 8-bit pair.
So using this canvas and with the computer program,
you can easily design any arbitrary shapes
by selecting and removing the voxels from the canvas.
This is one of my favorite slides.
It shows the whole design process of using this method.
And what's very interesting in this method
reminds me of a popular technology called 3D printing.
So in order to make any arbitrary shape,
we start with our building material.
In our case, it's our DNA strand of the cube.
You don't need to order any new materials.
So we have ordered the DNA strand
and sitting there is our building material.
Then we use our design software,
which starts with a pre-made 1,000-voxel matrix.
And using the program, you can quickly select
and remove the voxels from the canvas.
With this program, you can easily, quickly design
any arbitrary shapes in a few minutes' time,
comparing to, like, if you use origami,
typically it takes, like, hours, if not days,
to design one particular shape.
And another nice thing about this method is
you don't need to understand how the DNA...
how to design DNA sequence,
how to design the interaction between DNA strand,
because you're simply working on a simple 3D model.
And after that, the program actually
transfer, based on the shape,
decide what breaks are necessary for this particular shape
and send that information to a pipetting robot.
So in this case, the pipetting robot is our 3D printer.
What it does is select certain breaks from the plate here
and mix them together in a tiny tube.
Just one thing I want to remind you,
even though I call it 3D printing
because it's similar, the way it works,
similar to 3D printing, but it's actually quite different.
3D printing is an additive method.
You are adding material layer by layer on a surface.
But in our case, it's a self-assembled structure,
so you do have to perform one extra step,
which is a cleaning process.
A couple days later, you get billions of copies
of this nanoscale object,
which looks exactly like what you design in the software.
With the program, we designed 100 sophisticated structures.
Most of them are much more complex
than a structure available before.
And the nice thing here is
essentially a really rapid prototyping 3D nanoscale method
allows you to design objects
with 2 to 3 nanometer resolution.
We can design a structure quickly and easily,
but however, one thing we do not understand
is what's happening in the solution.
The assembly process is really hard to understand.
We don't have any nice method to understand that.
But Gail from Harvard,
he made this really nice animation for us
to demonstrate what could be happening,
what we think is going on in the solution,
but it's not exactly accurate.
So initially, each DNA break, of course,
is a single-strand floppy, single-strand DNA,
and a few of them, several of them,
coming together form a small structure
and gradually grow larger and larger
by recruiting more single-strand DNA from a solution.
And just to remind you again,
each single-strand DNA here in the structure
has a unique sequence.
Each single-strand DNA here are different.
That's how they know where to go in the final structure.
So we can make a nanoscale object,
which typically is smaller than 100 nanometer.
That's great for many applications.
However, for many other applications,
you do need larger structure.
So one way to get larger structure
is assemble them to crystals
forming periodic lattice or crystal structures.
So in the following work we're trying to do here,
we're trying to use this modular self-assembly
to make micro-scale crystal structure.
Without discussing more technical details,
the idea is simple.
You take a small...
take a modular object,
like in this case a tiny cube structure,
and you modify this sequence on the surface.
Now you can connect the individual structure together
to form periodic crystals.
But it depends how you grow it.
Of course, you can grow it to one-dimension,
two-dimension, or even three-dimensional crystals.
And in this particular work,
we focus on making two-dimensional crystals.
And even for that,
there's two different ways to grow your crystal.
One way is grow to this two-direction.
And in this two-direction,
then you get this multi-layer crystal.
So depth of the crystal is defined
by how many layers you have in the initial building object.
And you can also make this two-dimensional,
what I call two-dimensional helix forest,
if you grow a crystal along this direction and this direction.
And you grow a crystal like this,
and each cylinder here is a DNA helix.
The depth of such crystal is defined
by the length of this initial structure,
the helix in the initial structure.
And the reason why we think...
By doing so, remember,
we can change the initial building structure.
By changing that, we can program our crystal
to have different depths,
from 2 nanometer all the way to 80 nanometer.
And in addition to that,
we can also program nanoscale patterns.
If you use different shapes instead of a cube,
you can generate sophisticated patterns.
So why that is interesting,
just to remind you again,
actually before this, there are many DNA crystals,
but most of them are a simple...
a simple single-layer crystal
with 2 nanometer depth.
And in addition to that,
another type of crystal is made from DNA origami.
It's also a single-layer crystal,
so there's no control of crystal depth.
This is the first time we have a method
that allows us to define the depth of crystal.
In addition to that,
it allows us to design nanoscale features.
I'm going to quickly show you a few examples,
and then we can move on to the next section,
which is applications using DNA structure.
I'm going to show you three crystal examples
in the following slides.
The first one is a DNA crystal with parallel channels.
So on the left side,
we have the 3-dimensional model,
and on the right side,
we have the actual TEM image of the crystal.
A crystal can grow to about 10 micrometer size
or slightly larger than that.
And if you take a look at the zoom-in picture here,
you can clearly see the well-folded pattern
of this parallel channel.
Just to remind you, the channel here is really small.
It's only 5 nanometer diameter,
and each neighboring channel
is separated only by 10 nanometer distance.
The second example is a membrane-like DNA crystal.
It's made from this DNA forest method.
It looks quite like a membrane with all these channels,
all the holes on the structure.
And if you look at the real TEM image,
you can clearly see the long-range order
of such crystal is really great.
Again, if you look at the zoom-in picture here,
you can see in the video,
because you are seeing the projection view of such crystal,
so each single dot here corresponds
to a single DNA helix.
The last one is one of my favorites.
It's like a virus-like crystal.
It's not a two-dimensional crystal.
It's a one-dimensional tube structure.
Each tiny cylinder here
is a 32 base pair DNA helix.
And if you just simply look at the zoom-in picture here,
you can clearly see the well-folded tube
with a cavity and a short helix here.
The tube can actually grow above 10 micrometer length.
I call it virus-like because it reminds me
of this TMV virus particle, tobacco mosaic virus.
Of course, in the case of virus,
the tube actually is made from a protein shell,
and inside the cavity, it carries the RNA sequence.
So just to summarize the second part of today's talk,
what I present, what I bring to you today
is truly the state-of-the-art
three-dimensional DNA self-assembly method
that allows you to make both discrete 3D objects
and large micro-scale crystals.
And with those new structures,
we think it opens up a lot of possibilities
for future applications.
So for the next section,
I want to use some projects I have done in the past.
Some of the work I have done during my PhD work.
I'm trying to use those as examples
to show you how we use DNA structures
to address applications
and also hopefully give you an idea
to think about how can you use DNA structure
for your own research interest.
So each project here are quite different.
However, they all rely on these two things I mentioned
at the beginning of this talk,
is how we design the micromolecular shape
and how we site-specifically functionalize
our DNA structure at nanoscale precision.
So the first project I want to quickly discuss with you
is high-sensitivity, high-specificity
divalent protein binding.
So the binding sensor.
So in this case, the multi-valence actually
is a really common phenomenon in biology.
As we know, the interaction between two biologic entities
often is through multiple interactions
between receptors and ligands.
So it's a really important and interesting phenomenon.
So the motivation here is we're trying to use DNA as a template
to screen strong divalent binding
and try to construct a biocompatible divalent sensor.
Of course, we're trying to start with something
as simple as possible.
So in this case, we choose this human thrombin protein.
It has two binding ligands, aptamer A and aptamer B.
Aptamer is a DNA sequence that has binding affinity
to a certain protein.
In this case, it has two aptamers.
So in order to build a divalent sensor,
ideally, it's really easy.
You just want to construct some structure,
and you want to immobilize these two aptamers
on this structure
and make them to interact with protein at the same time.
But it sounds easy, but it's not that easy to do
because in order to achieve this,
you have to be able to tune the distance
between these two aptamers with nanometer precision.
And with DNA structure, we can easily do that.
So here I'm showing you a small structure.
The blue structure here contains four DNA helix.
We designed a whole group of DNA structure
that contains either just one aptamer,
this red color single-stranded DNA here,
or this green color single-stranded DNA here.
All the structure contains two aptamers,
but with different distance to separate these two aptamers
from 2 nanometer to 4 nanometer to 6 nanometer.
And then we incubate those DNA structure
with the thrombin protein,
and we run a native gel to see the result.
At this particular condition, you can see
with just one aptamer,
there's no change with or without the protein.
But when you have two aptamers
and increase the distance to 4 nanometer or 6 nanometer
between the two aptamers,
you start to see a new band
which can correspond to this protein and DNA complex.
And this band is stronger than this band,
meaning the 6 nanometer band
is even stronger than this structure
that contains the aptamer separated by 4 nanometer.
If you further move this two aptamers
beyond 6 nanometer distance,
this band intensity actually decreases.
The result is not shown here.
But by doing so, essentially we already achieve
a really strong divalent binding sensor
using DNA structure.
But the next thing we're trying to do
is trying to visualize this at single-molecule level.
Just to remind you,
that's another advantage of using DNA structure.
It often allows you to visualize,
to say something at a single-molecule level.
So in this case, we want to say
that we have a divalent binding at a single-molecule level.
So what we do here is we mix our thrombin protein
together with a two-dimensional
rectangular-shaped DNA structure
with a dimension about 90 nanometer by 60 nanometer.
And on the surface of this two-dimensional DNA structure,
we design multiple features.
On the left corner here,
we have a group of DNA loops
which we call the index
that tell us orientation of such structure.
And in the middle part of the structure,
we have two columns,
aptamer A and aptamer B,
that are separated by 20 nanometers
to avoid any crosstalking between the aptamers.
And on the right side of the structure,
we have pairs of aptamer A and B.
Each pair of aptamer A and B
are separated by a 6-nanometer distance.
That's the optimal binding distance
we figured out from a previous experiment.
So we expect the human thrombin protein
going to prefer binding
to the two aptamer positions.
So after incubation,
we use an atomic force microscope
to image our structure.
And this two-example image here,
we do see the protein actually
sit on those two aptamer positions.
You can see individual proteins,
in this case,
corresponding to those white dots here.
And we count the binding events.
Eventually, we figure out,
based on our counting method,
we figure out we roughly see
a 100-fold increase of binding affinity
if you compare this two aptamer position
to single aptamer binding affinities.
The next paper, next example,
application I want to show you
is a DNA nanochip
for multiplex single molecule
DNA and RNA sensor.
The idea here
actually is really simple.
We are familiar with microarray,
which uses a complementary DNA probe
to detect certain gene sequences.
And they use a fluorescent signal.
But one thing here is
you don't have really
particular control
in terms of the probe density
and the locations
on such a microarray.
So here, we try to use our DNA structure
to demonstrate two things.
One is we demonstrate
we can build this biocompatible
single molecular sensor
using DNA structure.
It's a small DNA structure.
It's made of biocompatible material
of DNA, and also it's small.
That's a great potential to be used
to demonstrate with this DNA structure
because it's totally addressable.
We're trying to demonstrate we can precisely
control the position and density
of probes on such a
small hybridization array.
That allows us to
achieve optimal hybridization
efficiency.
So here again, we choose the same
rectangular-shaped DNA structure
we used for the
divalent binding study.
It's again a simple rectangular-shaped structure.
The initial design is like this.
On the left corner here,
again, we have this index to tell us
the orientation of the structure.
On the other part of the structure, we have
different design features.
We have six DNA probes
containing random sequences
used as control probes.
At different locations,
we have columns
of DNA probes for detecting
three mammalian genes,
Reg1, Semic, and Beta-actin.
Each color here represents one type
of DNA sequence.
Each probe here is simply a single-strand DNA.
When we add
our target DNA
into the solution, we expect
the target DNA,
shown as the pink strand here,
is going to bind to the single-strand probe
and form a duplex.
That duplex should be able
to be sensed by the
AFM cantilever. Here, we're using AFM
to detect the
hybridization signal.
We test the initial design,
and when we add all three targets
into the solution, we do see the hybridization signal
on our DNA structure.
One really annoying thing is,
no matter how many times we try
to change our design, how many times we try
to repeat our experiment,
we always see this annoying positional effect,
which is we always get the best binding
efficiency, hybridization efficiency,
for this probe, the Beta-actin probe,
and lowest binding efficiency
We did a lot of studies.
Eventually, we figured out it's because
this is such a small scale,
the position actually is really important
because when you have a probe in the middle
part of the structure, it's harder for those
targets to get access to the middle part
of the structure. That's why you get much lower
binding efficiency.
This is something you can only notice
using a single-module kind of study.
This is interesting, but it's
annoying, too. How do you do
multiplex detection if
you have the positional effect?
We come up with a really simple idea.
Instead of using just one structure,
we come up with this
barcode idea
in which we use four
rectangle-shaped structures.
They have the same geometry, but on the left
side of the structure, we put different combinations
of DNA
duplexes to form
what we call barcodes.
They allow us to identify, distinguish
each structure. Now we can mix
four structures together. Each one only contains
one set of probes
for detecting just one gene.
We can mix them all together
and we can detect all three different
sequences simultaneously.
I showed you two projects
so far.
It's really interesting, but
you probably ask, maybe I don't care about
making nanoarray, nanochip,
or don't care about making
surveillance sensors, or whatever.
One thing I want to remind you,
in this project, what I'm trying to
convince you is, I think DNA
what I want to convince you is
it's a really great, very effective
molecular pegboard.
You can easily tune
different biomolecules
on the structure, put them at different positions.
That's really important for
studies, for example, for biomolecular signaling
and those kinds of studies.
Just to remind you, another thing
I use in this project,
an ongoing project, I'm writing a paper
on this, is not only
we can make a static DNA structure,
we can only make dynamic structures.
We can change our DNA structure,
make a nanorobot, a nanomachine,
make them perform certain functions.
This is one simple example.
The idea is really simple. A card is
GFP protein, green fluorescent protein,
into two halves, the C-terminal
and N-terminal part.
When I put them close
to each other, they are going to show
this strong fluorescent signal.
I put these two separate parts onto
our DNA structure.
It's a square-shaped DNA structure.
We can change the conformation
without discussing details here.
We can make the structure open
and make it close whenever we want.
When it's close,
we can observe a strong
fluorescent signal. This is an example
to show we can use a dynamic DNA device
to regulate biomolecular
behavior of biomolecules.
I'm not going to show you
fluorescent data here.
I'm just going to show you one gel here
so that we can control the
structure to have different
conformations from fully open
to slightly closed
to all the way closed structure.
You can also observe such kind of
conformational change on a native gel here.
You can see when it's closed, it's actually
moving much faster on the gel.
The last
slide here
is not really a single
paper or single application.
Instead, I put
more like a summary slide
here just to remind you another
important direction using DNA structure.
Using DNA structure, fabricate
inorganic materials. The idea is
to combine the functionality
of inorganic materials
together with the power of DNA
self-assembly.
Here, I just want to remind you
four available methods to do
that kind of fabrication.
The most popular way
of manipulating
inorganic material is called
templating method. The idea is you
assemble your DNA structure
and you use it to immobilize
materials, especially gold particles.
Here,
most publications in the field are using
this method. We can
arrange, in this case, gold particles
into a periodic 2D array
or this kind of parallel pattern
of gold particles. Another method
is just invented
this year is called etching method.
The idea here is we
put a layer of graphene
on a solid support, a silicon
wafer, and we drop
our DNA structure, in this case
a circle shape or other different shape
DNA structure onto the
graphene layer and we
metallize
our DNA structure and then use
ion etching method to blast off all
this uncovered graphene on
the silicon wafer.
So what's left there is a shape
which looks identical to a DNA structure
but is made of graphene.
So that's how you make different graphene shapes.
And the third method, again, is
only published in
this year or maybe from last year. There's another
publication before this. It's called
coating method. In this case, you again use
DNA structure. Then you can grow a layer
of different dioxide,
a different oxide on the surface of a structure.
In this case, it's actually
silicon dioxide.
And the last method is unpublished work
from my colleague Wei.
Wei, it's called
casting. So in this case, you build a
DNA cage structure and you
immobilize a tiny
gold particle seed inside of the
cavity here and you can grow
that seed into different shapes.
Because the growth
is confined by the DNA
structure, so now you can grow the structure
to different irregular
shapes which you typically cannot achieve
with other chemical synthesized
methods. And with those
inorganic materials, of course, you can
use them for many applications like
making biosensors, making different
devices.
So to sum up the first part
and second part, I think
in my mind, DNA nanotechnology offers
one of the few most
effective ways to organize and
manipulate matters at this
single molecule level with just
nanometer precision. I think that opens
up a lot of research possibilities
for myself and for a lot of people
in the field.
I put one more
slide here to sum up
today's talk. I want to
talk about my future
research interest.
In general, there's two
directions in our field.
One is developing DNA structure,
DNA self-assembly, trying to understand DNA
assembly, trying to predict
how DNA behaves in solution.
And I like to call those
enabling technologies. So you develop
your structure and it enables
you to address applications.
And to that direction
there's three big challenges in my mind.
One is sequence design.
And the second thing is
how to stabilize your structure
at particular conditions.
Especially, for example,
biologic conditions or other extreme
conditions if you want to use DNA to build
inorganic materials.
And the last challenge
is here, of course, design
assembly strategy.
We're not going to discuss details here.
In terms of applications,
I think in general there's two different
directions. One is using DNA
as tools to study
biology or to make
biomedical devices
for biomedical engineering.
That's more like a biology-inspired
direction. The second direction
is using DNA to manipulate
inorganic material. That's more towards
material science and
inorganic materials devices.
Those directions. Personally,
I'm interested in the first direction
which is using DNA to develop
tools for biologic and biomedical
engineering. I think there's three
particular directions I'm most interested
in. One is trying to
use DNA as an artificial
system, trying to mimic and
study biologic systems in vitro.
One example is that
a motor
protein project
I showed you at the beginning of this
talk is one example of that
direction. A second example is using
DNA to build devices for
studying biophysics
or to control
how biomolecules behave.
The third direction,
which is really a big direction
and a lot of money
going to that direction, is
trying to develop biomedical
devices like sensors, imaging tools
and using it to build
delivery vehicles for drug delivery
cancer treatment
and those kinds of research.
I'd just like to thank
my advisor, William
Shi and Peng Ying and all the group members
who gave me a lot of help.
Also, I'd like to thank all my
collaborators, especially
my PhD advisor,
Professor Hao Yan from Arizona
State University. Also, I'd like to thank you all
for attending today's talk.
I'm happy to take any questions you may have.
Thanks.
How can you tell
and can you build an algorithm?
There are different ways.
When you talk about quality, actually, it's a really
complicated question.
Let's start with something simple.
I want to make this shape. For example,
a letter A.
I want to make that letter A.
I just want to see if I can make that shape
roughly so I can use imaging.
You can use TME images so I can count
thousands of particles
and count how many percentage
particles actually show me this letter
A. Of course, you
can argue that's really subjective.
It's not a perfect method.
A lot of people
really care about
molecular level, what kind of quality you can
get for the structure. Instead of getting
the rough shape, for example,
I have hundreds of strands.
For argument's sake, let's say I have 300 strands.
Do I have all the 300
strands in there?
For many applications, that's not
necessary. If you have 280
strands there, maybe it's enough for many
applications. But it's
important to be able to answer
this question. If I have a structure
containing 300 strands, do I have all 300
strands there? Unfortunately, there's
no available method to do
that yet, to study if
I have all the strands there at the right position.
Some possible
way to do that is, for example, you can
use a flat experiment.
You can label some DNA strand with
flat pairs
and you can study it.
But that typically allows you to
only understand if particular
strands are there or not. So you can basically
do a few strands each time.
So if you want to do
hundreds of strands, it's really hard
to do that.
Another way people are using
cryo-EM
to study a structure.
But again, in order to
reconstruct a 3D
image of a
particular design, you have to
collect images of multiple
particles and average all the pictures
together. So what you end up with is still
average
quality, average structure.
So probably, if your
structure is always missing one strand,
you will see that using this method.
But if your structure is missing a strand,
you will not see that.
I hope that answers your question.
But if you have more questions, we can talk more later.
So for the structures, they are
all three-dimensional, but the TM
image can only show us 2D projections.
Yes.
Cryo-EM certainly
is a way to go.
We do cryo-EM.
I don't know if I have
pictures here.
Maybe I have one picture.
Just one moment.
This is cryo-EM.
This one
is a picture
from cryo-EM. Of course, this is still
showing just a projection view, but
we did the 3D reconstruction
so you can see the 3D structure.
So cryo-EM, if you really
want to see the 3D
confirmation of the structure,
that's the way to go.
But even with just the TM,
you still can get a rough idea
because you can collect
an image from multiple angles.
That doesn't give you the whole picture,
but you have really good confidence
if you see it's really
forming the right
confirmation.
So those
two are cryo-EM images.
Thank you for your nice work.
Thanks.
So
that's a good question.
The question I guess is
compared to a top-down approach,
traditional top-down approach, if you want to
manufacture new organic material,
what's the advantage here?
At least what I show here is only micrometer
size compared to
maybe millimeter size device
that you can make with top-down approach.
Also the cost issue. Maybe this
looks like it's going to be more expensive
than the top-down approach.
So that's a really good question.
I think right now
the main focus
that direction, the new organic material
fabrication, there's two different
directions. One is trying to make some structure
that cannot be done yet with top-down approach.
So we have a few
examples. At least
it cannot be easily done using top-down approach.
Of course we cannot do that either,
but we're moving toward that direction.
That's just for
fundamental research purpose.
Then you don't care about the cost
or those issues. If you can make something
nobody can do without the method, that's okay.
And secondly,
we're trying to
Okay.
Let me give you one
example.
I probably don't have
pictures here.
But again, it's not like
we can make it already.
It really depends on
do you really believe it or not.
So let's just say
Sorry about this.
First of all,
let's take a look at this one.
Not this one.
Okay, this one. So I talked to people
from physics
department. So they are interested to make this
small cavity
structure.
To study how the light
movement is confined in a small space.
So in this case, we're trying to
cover this structure outside
of this tiny tube structure
with gold particles.
And then we're going to
immobilize quantum dot
or fluorophore inside of the structure.
So you basically
get a tiny
tube covered by gold particles.
I think that's already something
not easy to do with
conventional method.
Remember the dimension here is only
about 30 nanometer.
So you're going to have a 30 nanometer
diameter tube
covered with gold particles.
As far as I know, I don't think there's any
easy way to do that yet.
Of course, this is just one simple example.
If you know someone can do that, let me
know. I'll be happy to know that.
In addition to that, again, this is just a rough idea.
Since we can make it really thick
crystal, again, I don't have the real
structure here, but imagine you can make
a crystal. Now you can design 3D
features. So you can assemble
in theory, you can assemble
3D arrangement of
inorganic materials into three-dimensional
patterns. That probably
can be done with top-down
approach, but that's not easy. And also
remember we have, at least
in theory, the position and control
we're going to have is, we're going to have
nanometer precision control in
all three dimensions, especially
the Z dimension. I think
with top-down approach, at least at this
point, it's hard to do that, to
generate some kind of complex
3D features with
nanometer precision control,
especially at the Z direction.
That's something I think probably
not easy to do with top-down approach.
Then move on to the second issue, the cost.
If I want to, let's just say
I really want just mass production
a lot of materials. Can we use
DNA? The cost
issue, so that
major, that boils down to the cost
and the size, right? So right now we can
only make maybe micrometer size.
But you can also argue maybe with
some development, we can
make it to larger size. That's
totally possible. We have
one really simple example. They made
a millimeter-sized
crystal. It's not really
complex, but they showed you it's
possible to grow a crystal larger.
So that's the size. And the second
question is the cost.
I think there's, first of all
it depends
what structure you're trying to build.
The DNA structure is not always expensive.
If you want to build just a single
layer structure, it's not going to cost
that much. It also
depends how many DNA strands you
how sophisticated, how
complex your structure is. Sometimes
you just want to generate really simple patterns
like, for example, some
of those patterns here. You only
need a few strands.
So it's not that expensive in that
case. But you can also
argue it's still more expensive than
top-down approach. But you can also argue
with top-down approach, you have to have all
these big machines. Let's forget
about that.
Bottom line is the cost
of DNA is gradually
lower and lower each year.
So we are working
on trying to
mass produce DNA using
enzymatic method. Or
eventually, we would love to
that's a big project, a big
direction to go, is trying to use
bacteria to grow
DNA for you instead of
synthesize it. In that case
it's going to be really cheap if we can do that.
So there's a lot of room to improve
in terms of cost
and size of structure.
Alright, so maybe
last job. You guys, if you have
questions, you can ask in person. So let's
thank Don Keith for the panel.

