Introduction to Protein Design and Protein Design Algorithms

Transcript of Part 2: Design of New Protein Functions

00:00:06.28	Hi.
00:00:08.04	I'm David Baker.
00:00:09.15	I'm a professor at the University of Washington,
00:00:11.12	and this is part 2 of my iBio seminar,
00:00:13.26	and today I'm going to be talking about
00:00:16.18	the design of new protein functions.
00:00:18.16	In the first part,
00:00:20.14	I spoke about designing brand new protein structures,
00:00:23.06	and now I'm going to show you, today,
00:00:25.12	how we can go beyond designing structure,
00:00:29.11	to designing new protein functions.
00:00:33.16	The motivation for this is really presented by nature.
00:00:38.24	The exquisite functions
00:00:40.04	of naturally occurring proteins
00:00:41.28	really solved the challenges
00:00:43.22	that were faced during biological evolution remarkably well.
00:00:46.20	So, if you think what living things are able to do,
00:00:50.11	they're able to capture energy from the sun,
00:00:53.01	they're able to use that energy to build up molecules,
00:00:55.25	build up complex organisms,
00:00:58.07	and eventually to think,
00:01:00.04	and for me to talk and listen to you...
00:01:02.04	and for you to listen.
00:01:04.03	So, in all those processes...
00:01:07.21	they are largely mediated by proteins.
00:01:12.04	In our genomes,
00:01:14.02	of course, are genes,
00:01:16.09	and those genes give the blueprint for life,
00:01:18.29	but they do so by encoding proteins.
00:01:21.12	Proteins are what actually do the work.
00:01:23.07	And again,
00:01:25.11	the protein complement we have in our bodies,
00:01:27.27	and the other living things currently existing on earth,
00:01:30.23	are really exquisitely tuned by natural selection
00:01:34.02	to solve the problems
00:01:35.21	that were relevant during evolution.
00:01:37.21	However, in today's world
00:01:39.14	we face challenges that were not faced
00:01:42.00	during natural evolution.
00:01:43.14	There are diseases like cancer and Alzheimer's
00:01:45.25	that were not really issues during evolution
00:01:47.25	because we didn't live long enough.
00:01:49.29	We're heating up the planet,
00:01:51.23	we're running out of fuel,
00:01:55.13	and there are new types of viral epidemics
00:01:57.18	that are coming around,
00:02:00.25	and one can have reasonable confidence that
00:02:03.02	if we had another billion years to wait,
00:02:04.24	and there was adequate selection pressure,
00:02:06.15	that all of these problems would be solved
00:02:08.06	beautifully by natural selection.
00:02:10.14	But most of don't have a billion years to wait,
00:02:12.12	and so what if we could design
00:02:14.24	a whole new world of synthetic proteins
00:02:16.26	that solved today's problems
00:02:19.01	as well as naturally occurring proteins
00:02:22.02	solved the problems that arose during evolution.
00:02:26.19	And that's really the grand challenge of protein design.
00:02:31.13	The methods
00:02:34.11	that are used in the calculations
00:02:36.16	I'm going to tell you about today
00:02:38.11	I reviewed in part 1 of my iBio seminar,
00:02:40.19	but I'll go over the basic ideas quickly again now.
00:02:44.17	The basic principle is that
00:02:46.16	proteins fold to their lowest-energy states,
00:02:48.24	and so if we want to design new proteins
00:02:50.10	that fold up into new structures
00:02:52.05	that carry out new functions,
00:02:54.01	we have to be able to calculate energies
00:02:55.26	reasonably accurately
00:02:57.18	and we have to be able to sample through
00:02:59.21	the different possible protein conformations
00:03:01.15	to find the lowest-energy state.
00:03:03.14	And, over the years, my group,
00:03:04.29	in collaboration with many groups around the world,
00:03:06.25	has developed the Rosetta protein design software...
00:03:09.21	protein structure modeling software
00:03:11.18	to carry out these calculations.
00:03:15.19	If we want to design proteins
00:03:17.06	with new functions,
00:03:18.28	we need hypotheses about
00:03:21.15	the shape of the protein, the configuration of atoms,
00:03:23.15	that would best carry out that function.
00:03:25.25	And the final point is the most important one:
00:03:27.24	we can design new models of new molecules
00:03:30.13	as much as want on the computer,
00:03:32.12	but if we don't go to the lab and test them,
00:03:34.09	they remain purely science fiction,
00:03:36.27	so the final step in everything which I tell you about
00:03:40.18	is to... after doing the protein design calculation,
00:03:43.16	coming up with a new amino acid sequence
00:03:46.26	that encodes a protein
00:03:49.05	that's predicted to have the desired function,
00:03:51.02	the final step is to
00:03:53.08	manufacture a synthetic gene
00:03:55.19	encoding that new protein,
00:03:57.16	a brand new protein that never existed before,
00:04:01.07	and then take that synthetic gene,
00:04:03.00	put it into bacteria, make the protein,
00:04:05.02	and then see whether the protein does
00:04:06.28	what it was designed to do.
00:04:10.04	The way the protein design calculations work
00:04:13.22	is shown very schematically here
00:04:15.26	for the simplest possible case.
00:04:17.24	This is the problem
00:04:19.25	where we have a protein backbone we want to make,
00:04:23.25	and we want to find a sequence
00:04:26.03	which is very low energy in this backbone.
00:04:29.00	So, we keep the backbone fixed
00:04:31.05	and we search through the different combinations of amino acids
00:04:33.08	for an amino acid sequence
00:04:34.28	which is very low in energy in this structure.
00:04:37.25	Then, as I said, once we have that sequence,
00:04:40.03	we can go to the lab and make it
00:04:41.22	and experimentally test it.
00:04:44.15	So, the first example
00:04:46.03	I'm going to give you
00:04:48.07	concerns the influenza virus.
00:04:49.22	A schematic of the influenza virus
00:04:51.05	is shown on the upper left,
00:04:52.25	and then in the middle two panels
00:04:55.19	is a blow-up of a surface protein on the influenza virus
00:05:00.13	called the hemagglutinin,
00:05:02.17	and in yellow in the middle panel
00:05:05.10	are two parts of that viral surface protein,
00:05:09.00	this hemagglutinin,
00:05:10.23	which are very highly conserved during evolution.
00:05:12.28	The virus is constantly mutating
00:05:14.20	to evade our immune systems,
00:05:16.09	that's why we need new vaccines every year,
00:05:18.13	but there are certain regions
00:05:20.03	which absolutely don't change
00:05:21.29	because they're critical to the function of the virus.
00:05:24.16	There's a region I'll refer to as the stem region,
00:05:27.14	in the middle of the structure,
00:05:29.07	and then on the top,
00:05:31.10	where the protein is actually attaching
00:05:33.08	to cells in our bodies,
00:05:35.07	this is how the virus gets into our cells,
00:05:36.26	there's a second site called the receptor-binding site.
00:05:39.13	What I'm going to tell you about today
00:05:42.07	is the design of proteins
00:05:44.04	which bind to these sites shown in yellow
00:05:46.07	and block the virus function;
00:05:47.26	they prevent the virus from getting into our cells.
00:05:50.20	So, using the methods that I briefly outlined,
00:05:53.16	we've designed proteins which block the virus
00:05:56.20	that bind at both the site in the stem region on the side
00:05:59.10	and then on the surface,
00:06:01.22	but I'm going to tell you in detail
00:06:03.15	about the ones that bind at the stem site today.
00:06:07.29	So, the design process has two steps,
00:06:10.16	and I'm going to illustrate them for you here.
00:06:13.15	On the left
00:06:15.17	you see a blow-up of that stem region
00:06:18.28	of the influenza virus hemagglutinin,
00:06:22.03	that was the region that was in yellow
00:06:24.12	on the previous slide in the middle of the slide...
00:06:28.08	in the middle of the protein...
00:06:30.12	and you can see that there's kind of
00:06:33.17	a deep groove that  we decided we would try
00:06:36.09	and design proteins to bind into.
00:06:39.17	The design calculation has two parts.
00:06:41.28	The first part
00:06:43.29	consists of placing amino acid sidechains
00:06:48.01	into the groove
00:06:49.25	in ways that they make very good interactions.
00:06:52.06	An analogy for our approach
00:06:54.24	is to think of this like a climber
00:06:58.11	would think about a climbing wall,
00:07:00.12	where there's some region
00:07:02.09	that you want to hold onto, like this groove,
00:07:04.18	and the first problem is to find handholds and footholds
00:07:06.21	that allow you to really get a grip on this,
00:07:08.17	and then you have to figure out
00:07:10.10	how you're going to place your body
00:07:12.16	so that you can have your hands and feet
00:07:14.06	in all the good places for them
00:07:16.02	at the same time.
00:07:17.17	So, we start by figuring out where the handholds and footholds are,
00:07:19.21	that is, where we can place disembodied amino acids
00:07:22.17	into this cavity
00:07:24.25	to make really good interactions,
00:07:27.19	and the second part is to place the body,
00:07:30.15	and this can either be a protein
00:07:32.01	that we designed from scratch
00:07:33.27	or one that we design de novo.
00:07:36.17	And, so what you see here again
00:07:39.04	in sort of the solid surface representation
00:07:41.15	is the flu virus protein,
00:07:44.04	and you see the sidechains
00:07:46.09	that we placed in the preceding slide
00:07:48.28	docked up against the surface,
00:07:51.00	and now the ribbon-y thing
00:07:52.21	is a brand new designed protein that we've made
00:07:56.20	that holds these critical side chains
00:07:59.02	up against the virus in exactly the right orientations.
00:08:03.21	There are...
00:08:05.06	one of the components of the calculations of the design
00:08:08.07	are electrostatic interactions,
00:08:10.11	favorable interactions between positive atoms and negative atoms,
00:08:13.28	so on the right
00:08:16.12	you see a very red region on the virus,
00:08:19.05	that's negatively charged,
00:08:20.28	and we're putting a blue side chain, which is positively charged,
00:08:22.29	right into that to get more binding energy.
00:08:27.28	The two designs that I'm going to tell you about
00:08:31.02	are shown here, again,
00:08:32.27	with the influenza virus in yellow
00:08:34.10	and the design in magenta.
00:08:36.16	You see the sidechains
00:08:38.23	fitting into that pocket on the virus,
00:08:42.01	and you see the backbone of the designed protein
00:08:44.09	in the ribbon diagram.
00:08:46.17	Something that's important for me to emphasize
00:08:48.15	is that when we do these calculations,
00:08:51.05	only a fraction of the computed designs
00:08:55.01	that are predicted to bind the virus
00:08:56.26	actually fold up to fold up to structures
00:08:59.16	that, when we test them,
00:09:01.17	bind the virus experimentally.
00:09:03.18	These two proteins
00:09:05.14	bind the virus and they bind quite tightly,
00:09:07.21	but most of the designs in fact don't,
00:09:10.11	and it turns out the reason that they don't
00:09:12.13	is probably because these sequences don't fold up,
00:09:15.08	don't really fold up to these structures.
00:09:17.20	Our calculations
00:09:19.12	are not quite good enough,
00:09:21.08	so that we get some designs
00:09:23.03	which simply don't fold properly,
00:09:24.26	but the thing that's very powerful now
00:09:28.27	is it's very easy to synthesize synthetic genes,
00:09:32.27	so we can make many, many, many different designs
00:09:36.02	that have been found in these computer calculations
00:09:40.13	and test them all,
00:09:41.28	and identify those which actually function.
00:09:45.02	Now, I told you that those two proteins
00:09:47.25	in fact do bind the virus,
00:09:49.19	but it's important to know how they bind the virus
00:09:51.18	and how similar it is to the way that we designed them to bind the virus.
00:09:55.11	So, on this slide
00:09:57.02	I show crystal structures,
00:09:58.26	determined in the laboratory of Ian Wilson at Scripps,
00:10:01.04	where the influenza virus protein
00:10:03.02	is shown on the left in magenta and cyan
00:10:09.17	and the design model is in purple,
00:10:12.12	and it's binding, again, in the middle of the influenza virus protein
00:10:15.05	in that stem region,
00:10:17.19	and in red is the crystal structure.
00:10:20.29	What you can see is that the crystal structure...
00:10:24.17	in the crystal structure,
00:10:26.22	this protein we've designed,
00:10:28.11	this one is called HB36 on the left,
00:10:30.13	is binding to the virus
00:10:34.23	exactly like we designed it to bind,
00:10:37.20	and in that inset there in the middle
00:10:40.20	you can see that even the designed side chains
00:10:42.20	in the crystal structure are exactly
00:10:46.01	where they were supposed to be.
00:10:47.27	And the same thing is true for the other designed protein
00:10:49.24	that I described, called HB80.
00:10:52.19	The crystal structure is, again,
00:10:54.24	nearly identical to the design model.
00:10:56.29	So, while I told you that
00:10:59.02	a large fraction of our designs simply don't bind at all,
00:11:01.04	the ones that do bind
00:11:03.13	bind to the virus in essentially exactly the same way
00:11:07.23	that they were supposed to bind the virus.
00:11:10.07	The proteins,
00:11:11.10	after some experimental optimization of the sequence,
00:11:14.11	bind with picomolar affinity to the virus,
00:11:17.18	they're very tight binding proteins,
00:11:22.25	and our collaborators Merika Treats,
00:11:25.13	a graduate student in the laboratory of Deb Fuller,
00:11:27.17	has some very exciting results now
00:11:29.12	showing that mice
00:11:32.12	who would die from a lethal infection from the flu virus
00:11:36.17	are completely protected
00:11:39.00	when these designed proteins,
00:11:40.15	actually the one that was on the left,
00:11:42.14	and given to them,
00:11:44.15	and the protein can be given to them
00:11:46.23	up to 24 hours before or 24 hours after
00:11:49.29	they are infected with the virus.
00:11:51.28	So, we're very excited now about the possibility
00:11:54.05	that this could become a new type of flu therapeutic
00:11:56.25	where either you're going into an area that's infected
00:11:58.28	or you've just been infected.
00:12:01.01	Such designed proteins
00:12:02.29	might be a future treatment for the flu.
00:12:08.09	We're designing proteins now,
00:12:11.10	using the techniques that I've described,
00:12:13.10	to bind to
00:12:15.13	not only other pathogens
00:12:17.12	but to proteins on the surfaces of cancer cells
00:12:21.12	and normal cells
00:12:23.20	to modulate biological function.
00:12:27.05	I don't have time today to tell you about that,
00:12:29.23	but we're able to make proteins
00:12:33.04	that are also useful for figuring
00:12:35.17	some fundamental biological questions,
00:12:37.14	because we can design proteins
00:12:39.18	that knock out specific interactions,
00:12:41.17	and so that allows biologists, then,
00:12:43.03	to probe what the function of that interaction is.
00:12:45.12	But now I'm going to switch gears
00:12:47.07	and talk about the design of proteins
00:12:49.20	to bind small molecules,
00:12:51.21	and we use a very similar approach.
00:12:53.18	On the left is the structure of
00:12:57.04	a small molecule called digoxigenin,
00:12:59.06	which is used as a therapeutic
00:13:02.15	to treat heart patients, some heart conditions,
00:13:05.03	but if you get too much of it
00:13:07.14	it's very, very dangerous and patients can die.
00:13:09.24	So, we were interested in trying to design a protein
00:13:11.22	that could essentially be a therapeutic sponge
00:13:13.16	and soak it up.
00:13:15.07	The designed protein is shown on the bottom right.
00:13:18.01	In magenta is this dig molecule,
00:13:22.17	I'll call it for short,
00:13:25.24	and in green is a protein we've designed
00:13:28.13	which makes very complementary interactions,
00:13:32.23	those are hydrogen bonding interactions
00:13:34.27	shown in the dashed lines,
00:13:37.17	and it surrounds the dig.
00:13:41.10	Another view of it is shown in the upper panel,
00:13:43.24	where you can see a space-filling view of the designed protein,
00:13:46.06	and you can see it really snugs the surface
00:13:48.04	of the small molecule.
00:13:50.08	So again, this is purely a computer calculation,
00:13:53.21	but we then go to the lab and make the protein...
00:13:57.01	and we make the protein...
00:13:59.07	and when we made the protein
00:14:01.06	we found it bound the small molecule,
00:14:03.17	and Barry Stoddard's group
00:14:05.23	was then able to solve the crystal structure,
00:14:07.23	and that's shown here.
00:14:09.29	In cyan is the...
00:14:11.12	sorry, in magenta is the designed model,
00:14:13.07	that's what I already showed you,
00:14:15.10	it's the designed model of the designed protein
00:14:17.04	bound to this small molecule,
00:14:19.16	and in cyan is the crystal structure,
00:14:21.23	and you can see that the small molecule...
00:14:24.04	first of all, you can see that
00:14:25.28	this designed protein has the correct structure,
00:14:28.04	and second, you can see that the designed molecule
00:14:30.09	binds to that structure
00:14:32.07	in almost exactly the way that was designed,
00:14:34.04	making those same hydrogen bonding interactions.
00:14:36.07	And the left panel shows you the
00:14:39.04	shape complementarity in the crystal structure
00:14:41.09	of this small molecule with the protein.
00:14:44.18	This design was very exciting
00:14:47.02	because it, again, binds the small molecule
00:14:52.03	with picomolar affinity,
00:14:54.10	and we are now using this method
00:14:56.00	to design proteins which bind
00:14:58.04	a number of different types of molecules,
00:15:00.03	both toxins and other types of drugs,
00:15:05.05	and these types of designed proteins
00:15:06.26	could be useful
00:15:08.17	not only for soaking up dangerous molecules
00:15:10.19	in the body,
00:15:12.06	but also for detection of molecules and other purposes.
00:15:16.04	And, I'm going to conclude
00:15:18.08	by telling you about
00:15:20.27	our work on designing new materials.
00:15:23.24	So, many of the materials that you're familiar with,
00:15:26.05	like silk and wool,
00:15:28.01	are made out of proteins,
00:15:29.23	and biology has lots of examples
00:15:32.07	of more specialized sort of nanomaterials,
00:15:35.02	like viruses
00:15:37.25	have these very elaborate and beautiful coat structures
00:15:40.25	with which they use to protect their DNA,
00:15:48.05	and the principle of all these materials in biology
00:15:51.23	is self-assembly,
00:15:53.06	where there's a subunit that's made,
00:15:55.13	that's encoded in a gene,
00:15:57.06	and then that subunit interacts
00:15:59.21	with other copies of itself
00:16:02.02	to make a larger structure.
00:16:03.21	And, I'm going to show you now
00:16:05.20	how we can design brand new proteins
00:16:07.11	which self-assemble with other copies of themselves
00:16:10.24	to make larger structure.
00:16:13.18	So, in this first example,
00:16:17.09	what we've done is to take a protein that's shown on the left,
00:16:21.06	and place it on the corners of a cube.
00:16:24.25	And so, there are eight corners on a cube,
00:16:26.22	so we've taken eight copies of this protein
00:16:28.27	and arranged them on the corners of the cube
00:16:30.27	in such a way that
00:16:34.25	the surfaces of these different copies
00:16:37.03	on the different corners
00:16:39.06	touch each other.
00:16:42.11	And we then designed the sequences
00:16:44.19	of these interfaces where they touch
00:16:47.10	so that the proteins...
00:16:49.20	to make very low energy interactions,
00:16:51.20	so that when this protein is made in cells,
00:16:53.22	what we hope is that
00:16:55.22	it will self-assemble into the cubic structure,
00:16:57.19	stabilized by these designed interactions
00:16:59.21	that we've made.
00:17:01.13	And, in the lower panel here,
00:17:03.09	you can see an electron micrograph of cells
00:17:05.16	that are making this designed protein,
00:17:07.24	and you can see that these cells
00:17:09.20	are filled with these cubic structures,
00:17:12.29	and the averages of these images
00:17:14.25	are shown on sort of the right column of this panel,
00:17:17.10	and you can see they look quite a bit like the designed model.
00:17:20.22	They look like little dice.
00:17:22.20	In fact, what we'd like to be good enough to do
00:17:24.07	is be able to put different numbers on different sides.
00:17:26.14	We're not quite there yet.
00:17:28.19	When the crystal structure was solved
00:17:31.10	in Todd Yates' lab,
00:17:33.10	it was found to be nearly identical to the designed model,
00:17:37.01	which we were very excited about.
00:17:38.26	So, we can make these types of nanomaterials
00:17:40.19	and enclosed structures
00:17:42.14	with very high accuracy.
00:17:44.27	This shows another view.
00:17:47.28	The left three columns
00:17:50.01	show the same design I just described,
00:17:52.18	but now viewed down the different symmetry axes of the cube.
00:17:56.21	So for example, the third column
00:17:58.16	is the four-fold axis of a cube,
00:18:00.22	and in the upper row
00:18:02.21	is the designed model,
00:18:04.08	what we were trying to make,
00:18:05.29	and in the lower row is the crystal structure,
00:18:08.08	those are the structures that we actually found experimentally,
00:18:10.25	and you can see they're essentially identical.
00:18:13.25	On the right is a second example
00:18:15.12	where we were trying to design proteins
00:18:17.13	to come together to form a tetrahedron,
00:18:19.10	and again you can see that
00:18:22.03	the designed models in the top row
00:18:23.26	are very similar to the actual crystal structures
00:18:26.17	that were solved experimentally in the bottom row.
00:18:32.29	And Yang Hsia, a graduate student in the lab,
00:18:35.03	has more recently used this approach
00:18:36.26	to try and make even bigger structures
00:18:39.14	like the icosahedron shown on the top left.
00:18:44.21	This is more or less
00:18:46.14	like the play structures that they have in some playgrounds,
00:18:49.28	except this is a complete icosahedron.
00:18:53.17	And, when Yang made this protein in the lab,
00:18:55.24	very recently,
00:18:57.16	he was excited when Shane Gonen,
00:18:59.17	who he sent the protein to to do electron microscopy,
00:19:02.27	sent back the pictures that I'm showing you here.
00:19:04.28	You can't quite see the whole icosahedron
00:19:07.00	but, for example
00:19:08.27	in the lower row on the middle panel,
00:19:11.05	you see something that looks very much like it.
00:19:13.09	So, we're currently trying to solve the high-resolution structure.
00:19:17.08	So, these were materials
00:19:19.02	that were made out of just one component
00:19:20.20	that was identical
00:19:22.08	that was then interacting with other copies of itself.
00:19:24.16	We can make this more sophisticated
00:19:26.14	by, instead of having one component,
00:19:28.17	we can have two components.
00:19:31.01	So, in panel A here, I'm showing two tetrahedra
00:19:34.02	that are inverted relative to each other,
00:19:36.08	one green and one blue.
00:19:40.19	And so, what we're doing here is
00:19:43.06	we're taking one building block, the green one,
00:19:45.13	and putting it at the corners of the green tetrahedron,
00:19:48.02	and another building block, the blue one,
00:19:50.28	and putting it at the corners of the blue tetrahedron,
00:19:53.14	and then as shown in the middle panel here,
00:19:55.18	we can move them...
00:19:57.25	we can slide them closer and further away
00:20:00.04	from the center of these tetrahedra,
00:20:02.24	and we can also rotate each one,
00:20:05.16	and we do this
00:20:07.17	until we find a way in which these fit together
00:20:10.13	in a very shape-complementary way,
00:20:12.13	and that's shown in panel C.
00:20:14.29	At this point it becomes a calculation
00:20:17.09	very similar to what I showed in that movie
00:20:19.11	that I showed at the beginning of my talk,
00:20:21.17	where we now have to design...
00:20:23.22	find an amino sequence...
00:20:25.16	amino acid sequences on both sides,
00:20:27.00	on both the green side and the blue side,
00:20:28.22	which fit together very well
00:20:30.17	and make very strong interactions.
00:20:33.07	And, when we've done that,
00:20:35.15	we again order synthetic genes,
00:20:37.18	or make synthetic genes,
00:20:39.10	that encode both proteins.
00:20:41.03	We make them in bacteria
00:20:42.24	and then we look to see
00:20:44.27	whether there's anything that's assembled,
00:20:47.00	and I'm going to show you the results on the next slide.
00:20:50.01	These are electron micrographs
00:20:51.27	of two of these materials.
00:20:53.29	These are, again, two components,
00:20:55.21	with a green component and a blue component,
00:20:58.13	and the designed models are shown
00:21:01.25	on the lower part of the slide,
00:21:05.24	with one component in green
00:21:07.21	and one component in blue.
00:21:09.17	In the upper panels
00:21:11.02	are electron micrographs of what we get out of E. coli cells,
00:21:13.22	bacterial cells
00:21:15.15	that are expressing these two proteins,
00:21:17.13	and you can see that...
00:21:18.25	first of all what you can see is that, for each design,
00:21:20.19	we get remarkably homogeneous particles,
00:21:22.25	so all the particles in these images
00:21:24.17	look essentially identical,
00:21:26.11	and if you look closely you can see that,
00:21:28.02	for the different shaped designs,
00:21:30.12	we get different shaped structures
00:21:31.29	and they correspond
00:21:34.05	to the shapes that we're trying to design.
00:21:36.06	So, I think in the middle panel,
00:21:38.04	you can the that the holes are a little bit bigger
00:21:40.13	than in the particles on the left panels.
00:21:44.14	And, what's exciting about this
00:21:47.24	for the applications I'll describe
00:21:49.24	is not only that the shapes are coming out right,
00:21:52.00	as we designed,
00:21:53.18	but that every particle is the same.
00:21:55.03	So, for example,
00:21:56.18	if you wanted to make a new type of drug delivery vehicle,
00:21:59.14	there are various ways of making particles
00:22:01.16	for drug delivery now,
00:22:03.18	so say you want to target a toxic compound
00:22:07.09	specifically to the tumor you want to kill,
00:22:10.06	but those methods always...
00:22:12.25	when you look at the particles
00:22:14.10	they're always very heterogeneous,
00:22:15.21	so it's hard to predict what they'll do inside the body.
00:22:17.20	With this technique,
00:22:19.14	we can make particles that are very precise
00:22:21.15	and each one is identical to each other one.
00:22:25.16	So, Todd Yeates' group
00:22:27.10	was again able to solve crystal structures
00:22:29.16	of these two-component materials.
00:22:31.14	So, in the upper rows are the designed models,
00:22:34.09	shown down the different symmetry axes...
00:22:36.17	two of the symmetry axes of these particles,
00:22:39.04	and in the lower rows
00:22:40.27	are the crystal structures of these designs.
00:22:42.26	So again, the process is,
00:22:44.21	you have the computer model, which is what's on the top row,
00:22:48.28	then you order a synthetic gene
00:22:50.26	which encodes both of the designed proteins,
00:22:53.27	you put these synthetic genes into bacteria,
00:22:56.06	you make the proteins,
00:22:58.02	and then you purify them out of E. coli
00:22:59.24	and you look to see what you've got.
00:23:01.29	And then, in this case, go one step further
00:23:04.27	to determine the X-ray crystal structures,
00:23:06.26	and what you can see here
00:23:08.28	is that these designed proteins are again...
00:23:11.09	the crystals structures are essentially identical
00:23:13.03	to the designed models.
00:23:14.18	So, we can make these designed nanomaterials
00:23:16.23	very, very precisely.
00:23:20.12	So, the different types of nanostructures
00:23:23.27	that I've described so far
00:23:26.15	are the ones on the left, and I already mentioned...
00:23:28.29	so, the question is, what good could they be for?
00:23:30.28	One very exciting possibility
00:23:32.20	is targeted drug delivery,
00:23:34.14	where, as I mentioned, you could put a chemotherapy agent
00:23:36.24	inside the cage
00:23:38.16	and then target it to the tumor,
00:23:40.01	so you don't have to take it systemically.
00:23:41.29	You can also put targeting domains on the outside
00:23:44.12	so that it goes exactly where you want it to go,
00:23:47.08	and we're now...
00:23:49.11	a first-year student in the lab
00:23:50.24	is now exploring different ways of putting nucleic acid
00:23:52.29	inside these
00:23:54.25	to make synthetic viruses,
00:23:56.09	not for bad purposes but for good purposes,
00:23:58.14	so we can deliver, say,
00:24:01.00	for gene therapy or for other types of therapy,
00:24:04.18	deliver RNA or DNA molecules
00:24:06.29	exactly in the body
00:24:08.21	where they would be good to go.
00:24:11.00	Another application is to vaccines.
00:24:13.16	We can display...
00:24:15.18	one of the things we're trying to display now
00:24:17.16	is the HIV coat protein...
00:24:19.23	we can display it on the outside of these cages,
00:24:21.28	it will be there in many copies,
00:24:24.02	and hopefully trigger a strong immune response.
00:24:27.21	We can also put molecules called adjuvants
00:24:30.04	inside these cages
00:24:32.01	to stimulate a stronger response.
00:24:33.17	Now, there are other types of particles,
00:24:35.09	other types of nanomaterials that we can design.
00:24:37.28	For example, the wire on the right side.
00:24:41.15	You could imagine things like
00:24:45.01	being useful for transporting ions
00:24:47.07	or maybe even electrons
00:24:49.01	in some sort of nanoelectronic device.
00:24:50.28	And, my last example today
00:24:52.23	is going to be for what you see in the middle
00:24:55.06	- a designed, repeating, 2-dimensional layer,
00:24:59.17	and this is the work of graduate student Shane Gonen.
00:25:03.14	Here is his design.
00:25:06.16	It's a hexagonal lattice
00:25:09.10	where these proteins are designed to
00:25:11.29	assemble first into hexagons,
00:25:13.22	which then interact with other copies of themselves
00:25:15.14	to tile the plane,
00:25:17.13	and when he makes this protein in E. coli
00:25:19.15	he gets this... this is straight out of a broken E. coli cell.
00:25:23.13	He sees these large arrays that correspond...
00:25:28.21	that have the geometry one would expect for his design,
00:25:32.25	and if he averages his data, the...
00:25:39.05	and then, sort of a representation of a map,
00:25:43.02	a density map that comes from this data
00:25:45.07	is shown in the lower-left panel,
00:25:47.10	and you can see that his model
00:25:49.09	fits into that quite well.
00:25:51.06	But, as you can imagine,
00:25:52.20	we really aren't satisfied until we've determined
00:25:54.13	the high-resolution structure,
00:25:56.10	which Shane is currently working on.
00:25:58.14	I've been very fortunate to have absolutely outstanding colleagues
00:26:01.07	that actually did all the work that I described.
00:26:03.13	Their names are listed on this slide
00:26:05.05	and, more generally,
00:26:08.08	I hope I've given you a sense, today,
00:26:10.04	for the potential of protein design
00:26:12.23	to create a whole new world of designed proteins
00:26:15.27	to solve challenges
00:26:20.02	that we collectively face today.