>> Amy Draves: Thank you for coming. My name is Amy Draves and I'm here to introduce
George Church who is joining us as part of the Microsoft Research Visiting Speaker Series.
George is here today to discuss his book Regenesis, How Synthetic Biology Will Reinvent Nature
and Ourselves. We are at a threshold of a new era of bioengineering. Imagine resurrecting
extinct species or making humans immune to all viruses. A few years ago this would have been
science fiction, but today these scenarios are quickly becoming reality. It's a new world filled
with possibilities and perhaps perils. George Church is a professor of genetics at the Harvard
Medical School and the visionary behind the Personal Genome Project. His many honors
include the Franklin Laureate for achievements in science. Please join me in giving him a very
warm welcome.
[applause]
>> George Church: Thank you everyone. It's nice to be here. This is my conflict of interest slide
so these are some of the companies, some of the organizations that have helped take some of
our primitive ideas and develop them and get them so that they are out in the real world. The
big question that you should be asking of anybody like me and of this book is why, why are we
doing some of the things that we are doing? Some of the things that we are doing are quite
exotic, and I won't be able to talk about all these things but you are certainly welcome to ask
me during the questioning or afterwards why should each of you do your personal genome.
That is a big question. Why should we do genome engineering as opposed to genetic
engineering? Why mirror life? Why should we reverse extinction for some key species? Why
should we store ten to the 24th bytes or more? And why should we study every neuron and
talk to it? First a little history, Bill and Paul and I did almost the same thing in 1968 which was
we worked on a teletype that was hooked up to a GE 635 mainframe. Since then we've kind of
converged a little bit on our interests in terms of public health and biology and neuroscience
respectively. Both Paul and Bill have funded me and I've gotten to know them, but the key
point I'm making here is about back then we borrowed mainframe computers. We time shared.
Eventually we made a decision to get our own and my, you know, I had borrowed many
computers by the time this TRS model 100 had come out in 1983, and it of course it had
Microsoft inside and it was listed at 1100 and when it came down to 380 my student wallet
opened up and I got it and it is still running programs in my office and whenever journalists
come to my office and they see these computers are running programs, they get all misty eyed
about how they used to phone in at 300 baud their news stories on this thing, so it's actually
had quite a impact. It was 20 hours supposedly on AA batteries and it had an awesome 8
kilobytes of RAM. It's amazing what you can do with that much RAM [laughter] and no disk
drive. So I'm going into the technicality of this because it represents how do we decide when
we get our own genome and how do we decide when we make a lot of technological leaps? So
one of the deciding factors is cost. As cost comes down enough even if there isn't a lot of
application software you will do it, and I should mention almost everything that I'm talking
about here is software one way or the other. We might write it in the hardware. We might
write it in DNA. Anyway if you extrapolate, and extrapolation is always a dangerous business
and it's especially dangerous if you extrapolate exponentially. Optimistic exponential
extrapolation was to say that DNA since the ‘60s was on the same curve as Moore's law since
the ‘60s. A very nice fit to both of them at a 1.5 X exponential per year, so you can say well, we
did the first genome pretty badly for about $3 billion. When would it be accurate and available
for all of us, affordable sort of in the low thousands? And the estimate was 50+ years you could
see this extrapolate somewhere out here past 2040. Fortunately, we had an earlier arrival
where the exponential changed. It's still an exponential, but it's a much deeper one. I'm happy
to say I was involved in that. I don't know whether we should over interpret the fact that it's
flat now, but you could argue that if it's not coming down, it might be a good time to buy your
genome now. It's in the low thousands. It's always hard to buy something when the price is
exponentially decreasing. I won't go through all of the details as to why this arrived in six years
rather than 60 years, but I will give you one totally from the cutting edge of where it's going
next, where it might start this exponential again and I don't know at what slope and that's
called nanopores. Nanopores are basically a direct interface with single molecules where each
single protein pore through its ions are going, sodiums or potassiums in a very high resistance
sort of giga to tera ohm resistance bi-layer and as a single polymer chain of DNA, a single strand
goes through that it will selectively block the ion conductance. What you see here is a pico
ampere on the Y axis and time in milliseconds along here, and here are little blocks of 3 G's in a
row and 3 T’s in a row and you can see that there is very nice discrimination between the four
different kinds of basis, but there's not such great discrimination between a run of bases in a
row. It's hard to tell if that is three or more. You can get on the order of 4% error rate with
these and the good news is that instruments have come down in size from kind of, you know,
mainframe size sequencing devices to handheld ones. This is not yet commercially available
but it is from a company, Oxford Nanopore. So you can either measure the polymers, meaning
those long chains, or you can break it into monomers into A’s, C’s, G’s and T’s and throw one at
a time and that might solve this problem. Or you can tag them; you can actually put tags on
each of the bases and monitor them and that's probably the winning combination. This is the
slowest of the new technologies to arrive. I've been working on this since the early 1990s, but
here you can see how cleanly separated -- this is a histogram actually. It looks, you typically
don't get A tags that are anywhere close to the C’s and G’s and T’s. These are very well
separated baseline separated and the error rate for detection of these things is on the order of
-- these are single molecule events with error detection around 5x10 to the eighth. That's a
really, that's not a sequencing error rate; that's an analytic chemistry error rate, but it can turn
into a raw single molecule sequencing error rate which would be unprecedented and
wonderful. Now treating you all as both geeky consumers you want to know what kind of
genome should you get. So you can get sort of, you know, in the sub hundred dollars you can
get sort of the SNP chip, single nucleotide polymorphism, but this is only 1/3000th of your
genome and it's not actually the best one. It's not even that informative of a fraction. The
exome is specifically the parts of the genome that are most informative from a medical genetic
standpoint, but it's only 1% of the genome and there are all kinds of things it misses. Whole
genome sounds pretty good. It's only 95% but for some reason we call it whole or complete
genome. No one has ever finished actually a human genome or any major animal genome, but
it this is something that a phased genome will give you. So if you're going to go out and buy it
which I think you should, you should get a phased genome. And this means two hits in one
gene can be distinguished from a hit in one gene and a hit in another, so each of you has two
copies of every gene and if you observe with these short reads that you have two locations in
that gene. You don't know unless you phase it whether it's just two hits in one gene copy and
the other gene is fine and doing its job and handling, making you fine, or if you have a
deleterious hit in both copies then you have nothing from that gene and that gene’s important
then you potentially have got a time bomb going off. So get a phased genome is the take home
there. Second take home is that it's not that easy to go from your personal genome straight to
traits, to determine the exact hour of your death and what you're going to die of, but there are
all kinds of cool things that you can do in between if you have a good cohort to help you
interpret them and we have this awesome cohort@personalgenomes.org and we collect all of
these data, all these intermediate and environmental data because in part the cost of your
inherited genome has come down a million fold. So has the cost of collecting big data from
other kinds of omics, they are called, genomics, immunomics and so forth, and these are
environmental, so it's how your immune system responds to microorganisms and how you
develop your RNA and protein is all big data that we can collect now a million times cheaper.
This is now an international project. There are over 2000 people in the Boston branch alone.
The Canadian version is, can handle the rest of the world and there are many other countries
queuing up as well. We have an entrance exam. You have to get 100% on it. Now this is, as it
turns out one of the best projects for establishing standards. And so the National Institutes of
Standards and Technology and the FDA have teamed up and this is a great teaming up to make
genome in a bottle, which means they are going to take a few genomes and make thousands of
copies of them and send them out to every company that wants to develop new diagnostic
tools, new instruments, anything that might involve genetics, and so when they look through all
of the different kinds of samples that are present in the world, they have quickly settled on ours
as the main one, the only one that really properly has the consent of the people involved for reidentification and commercial use. Either one of those could be a showstopper. How is it that
we have this unique cohort? Again, it's because everyone has to get a hundred percent on an
entrance exam. We educate first. This is the world's only open access data set meaning we do
not promise the identification because we know that everybody can be re-identified and there
are probably people in this room that understand better than anyone how in this world, this
highly connected world how easy it is from either your traits or your DNA or both in this case
you can identify people. We make stem cells available and many other cells available through
bio banks so people can task the hypotheses coming out of these projects, and we are
collecting even bigger data sets of all kinds of traits, not just highly medical traits, but almost
anything that could impact our way of living, anything that can be measured, we are trying to
measure it, and this includes skin resistivity, all sorts of cardiovascular components. This is a
portable ultrasound. EEGs, not necessarily the best example, and this is the one that isn't
portable which is FMRI but we think it's a terrific stopgap for measuring the brain. This is
actually slices through my brain; fortunately they were virtual slices rather than real ones
[laughter]. One of the reasons that people have hesitated in addition to price is that they think
I am not going to learn anything about my genome. I'm just looking at this generic health
advice and it's not going to tell me anything and if it did there's nothing I can do about it, so I'm
just going to take this advice, exercise, drink milk, greens, grains and so forth. Well, it depends
on your genetics. These aren't necessarily, if you exercise and you have hypertrophic
cardiomyopathy you can be dead. There are things you can do to handle it in advance,
preventively, if you know that you have it but it's not part of the standard medical diagnosis. If
you have lactose intolerance, that's not good for milk. Green beans if you have favism which is
a G6PD mutation, that can be a killer, and many drugs as well. Certain grains will set you off if
you have celiac disease. Hemochromatosis is really bad if you have iron in your diet. Some
people do take iron supplements not knowing they have hemochromatosis, and so forth. I'm a
narcoleptic, so rest is not a problem. And that's just a short list. There are 2700 of these that
are highly predictive and medically actionable. These are not things that there are weak
effects. There are plenty of things that are weak effects. There are plenty of things we don't
know anything about, but that's not what medicine is about. Medicine is about what works
really well and there are 2700 that worked really well. So here is audience participation. How
many people here do not have any insurance whatsoever? Okay. It can cost. This is kind of
group insurance rates. I don't know how much, I won't ask how much you're spending on
insurance, but the fact is if you get to the end of your life and you've been very, very healthy
and you suddenly die and your house, nothing ever happened to it. You never had an auto
accident. You didn't need long-term care, didn't have any fires or anything, you could say you
wasted $9000 a year. But you didn't know that in advance. So the question is why don't you
have your genome sequenced? Because it could be that it doesn't tell you anything, but you
don't know that until you look at it, right? That's the end of my argument. And here's some
examples of actionability. Not just actionability kind of in conventional medical genetics, but
this is the new way to do way to do medical genetics where people are getting the whole
genome sequenced. They're not looking for a particular gene which is kind of the way that
we've been doing it since the ‘80s. In this case little Nick Volker, three years old, his guts were
literally, fecal matter was leaking into his abdomen. He got dozens of intestinal surgeries.
Clearly this was not going well. In desperation they decided to sequence him. They found this
gene here XIAP and they completely switched the therapy to a little cord blood transplant. He
was fixed within weeks and he is around this time celebrating his eighth birthday. The Beery
twins misdiagnosed with cerebral palsy. They eventually sequenced their whole genome and
found this gene and then they changed the diet to include serotonin and dopamine precursors,
just a dietary change. Doctor Wartman actually studied leukemia and ironically got leukemia,
no connection, just coincidence and then his genetics told him what drug would be effective.
Gilbert was an example. Many people who get either mastectomies or ovarectomies if they
have BRCA one or two alleles it's a preventive thing, but it's increasingly applied. And here's a
case of a whole family had a -- not breast cancer predisposition but stomach cancer. This was
killing the family off it early age and 10 out of 18 people that were tested in this family had it
and had their stomach surgery to come preventively. So there are things that you can do, some
of them relatively simple from a whole genome sequence. You don't necessarily need to know
exactly which genes you are looking at. And this raises an interesting new way of doing science
as well as doing medicine, which is we've been building up these gigantic cohorts where to even
draw a single weak correlation, we have to have 10,000 people, but what if we had one person?
Can we do anything that's maybe even more powerful than correlation which has to do with
causality? Furthermore, I'm going to take this other strange direction which is are we just
talking about rare deleterious alleles; are there where protective alleles that give us
superpowers or interesting properties or maybe mixture of good and bad? And here's an
example where a double null meaning knocking out both copies of the myostatin gene results
in high muscle growth and decreased body fat. Flex Wheeler does work out, but he obviously
has an advantage and the causality here can be proven in animal models. Here are three
different animal species, mouse dog and cow that all have the same effect when they have both
copies of their gene knocked out, so that's myostatin. But there are many other ones and this is
just a partial list. LRP5 produces bones that are so strong it will break surgical drill bits. PCSK9
reduces your coronary artery disease by about 90% and is an inspiration for new antibodybased drugs. CCR5 is an inspiration for drug that we will see in the next slide. FUT2 and CCR5
both result in virus resistance. FUT2 double null results in Nora virus resistance and then here's
my favorite. APP gives you decades of extra life without Alzheimer's. In addition, APP4 is an
Alzheimer's risk that's about 20 times higher risk than if you have two copies of APP3. I
fortunately have very low Alzheimer's risk and I know that. That was of significance to my
family because my father had had problems. So more than this equals one. This n equals 1
patient happened to have, he happened to be one of the most unlucky people on the planet.
He had both leukemia and AIDS. They treated his leukemia in a fairly conventional way which
was to find a stem cell donor who was matched for his HLA so wouldn't reject the graft, but
they happened to also to look out and they could find a donor who not only matched for not
rejecting, but it had a double null, both copies knocked out of the CCR5 gene, which is how the
AIDS virus gets into the cell. It binds to the CCR5, so if you knock out both copies there is no
CCR5 on the T cells and this, he got cured of both diseases very quickly and five years later he's
still disease-free as far as I know from them both diseases. And that inspired a clinical trial
which went through phase 1 and now phase 2 which means it's low toxicity and efficacious
where you take the cells out of a person, treat their own cells with a molecular machine of zinc
finger nucleus it's called which will cleave specifically just in the CCR5 gene, so it's not putting
some random package anywhere around the genome where it can cause trouble. It's
specifically making a double null in CCR5. You put it back in the patient and now they have a
bunch of T cells that are resistant to the AIDS virus. This is just what those zinc fingers look like.
They're called fingers because they have these alpha helices. This is sort of nanometer scale.
Here's a little zinc atom that holds it in place next to this DNA sequence so you have a very
specific code where you can tell how this protein will recognize the DNA, and then you have a
similar code now for newer class of proteins called tal or talins. As I mentioned, we are doing
big data on the various omes including microbiome and immunomes. We can not only detect
antibiotic resistance, but there are consequences in antibiotic treatment where you can, if you
knock out enough of your microorganisms one will stay behind which is clostridium difficile and
this will cause all kinds of gastrointestinal problems. One of the solutions which is actually
pretty low-tech is called fecal transplants, so you can actually change your microbiome and it's
probably affecting obesity and diabetes as well as well as this. You can change your immune
system as well. We go around the world and we can find people who are resistant to HIV. They
are very rare. I mentioned one already, one way that people can rarely be resistant is by
removing the receptor, but the other way is they can actually make an antibody that is
neutralizing. It's not easy to do this with a vaccine, but you can do it with, you can find a rare
person that happens to have hit the jackpot, not just because they had vaccine or the right
virus, but because they had luck with their immune system. And you can take that and put it
into other people. So far this is at the rodent preclinical testing stage. Another way we can
deliver antibodies aside from having your muscle cells make antibodies which is what's
happening with those HIV neutralizing antibodies is we can make these nano robots which are a
little bit smarter than an antibody, yet they contain antibodies. These little gray things are
antibodies. They are color-coded purple here, and all of the rod shaped things are DNA double
helixes. You can design this basket basically which is held together when these two come
together, these long rods; there are just two points that are holding this basket together.
When it opens up the delivers this payload which can either kill specific cancers without killing
any other cells or it can do immune modulation. It only opens up when both of these double
helixes breathe enough that one half of them will bind to some protein on the surface of a
cancer cell and recognize it and when both A and B do that, they no longer can bid to their
partners and so there's nothing holding this hinge thing closed anymore and it opens up dumps
its payload. So this satisfies some of the main criteria as for what we would call robot in that it
has sensors which are these actimers [phonetic] that make a transition. It has logic, so it's A
and B and you can have more complicated Boolean logic and it has actuators which open up
and deliver payload which can be toxic or modulating. We're going to come back to nano
robots and DNA encoding in just a moment. The bio bank that I mentioned that we have for
stem cells and the personal genome project can also be used on organs on chips, so this is
getting pretty sophisticated where you can test drugs for their toxicity on human cells rather
than animal cells but in a much more realistic way. So for example, here is a heart that has this
contractility and a lung that has air epithelium endothelium and blood and it flexes with the air
going in and out, so these always mechanical forces are much more realistic. And there are 10
of these including a neural and blood brain barrier which brings us to our last topic which is
brain activity map. We want people to be -- there is a whole variety of neurative diseases,
traumatic injuries. This particular woman is tetraplegic, can't control arms or leg, Parkinson's,
Alzheimer's, many diseases that have a major neural component and we'd like to be able to
input and output to computers. In this case she has a small number of electrodes going into
her brain at this point. For the first time in her life she's been able to manipulate, give herself a
drink by manipulating through her brain waves with this robotic arm. This is some of the work
for my close colleague John Donohue, but the point is that this is still very crude and slow and
by looking at only a few neurons it's amazing that we can do anything. What we could do with
a larger number of neurons. One of the problems with cramming more electronics into a brain,
even if they are just wires, is that the efficiency of integrated circuits in terms of the amount of
energy that they need to operate, if you wanted to make this say wireless or in a variety of
senses either record or to input and output, there's been this Moore’s law not just for density
of circuitry but also for energy efficiency and you can see that we are off by about 10 hours of
magnitude from where biological systems can do operations at about almost 10 to the 26
operations per kilowatt hour. An example of one of these that we are harnessing is DNA
polymerase. This can go up to about 1 kHz depending on the polymerase and these
polymerases we've shown and others have shown that they are responsive to ions. Remember
I showed you earlier a nanopore where you can convert DNA into modulation of ion channels.
Here you can take ion channels and use it to modulate DNA synthesis so we can actually get
missing corporation of bases making a tickertape essentially or a long digital recording where
DNA is the digital media. So this is very compact. It's about a billion times more compact than
any other digital media and it has very low energy components as I mentioned just a moment
ago. So here's a paper we just published in Science where we did another kind of digital
encoding. That was one intended for monitoring neurons. This one was purely digital where
we did an inkjet using organic chemistry to encode a book, 5.2 megabits which were 5.2
megabase pairs of DNA and this kind of exceeded the density in terms of its per cubic
millimeter of most other technologies. And the next that we would like to do is we are
developing, this was digital to digital, essentially digital, zeros and ones to As, Cs, Gs and Ts and
next we are going analog to digital so taking photos and audiovisual data and converting it
directly to DNA with no mechanical or electronic components, purely photons to DNA
sequence. Here is Stephen holding a book that you may recognize and holding 20 million
copies of that book in DNA form which is actually a tiny dot in the middle of that red ring here.
If you missed this episode he immediately wanted to eat it [laughter]. I stopped him. Not that
it's that particularly valuable, 20 million copies here and there doesn't really matter. We made
70 billion of them. I covered some of these topics here ending a tiny bit early here and
hopefully I stimulated some questions. I think I gave you a few reasons why I think personal
genomes is now a very low cost. It's like insurance and you should get it whether you are going
to need it or not. Genome engineering, I barely got to touch on, but we use this for engineering
chemicals and fuel now being produced at about $1.28 a gallon. Mirror life that would be very
hard to explain as well as the de-extinction. Yottabytes, I think the audiovisual analysis, goal
that we have you can imagine having ubiquitous audiovisual recording where you would be
basically archival storage where you wouldn't read it unless something interesting, exotic or
really good or really bad happened and then you would go and you would selectively read just a
few giga pixels from what would be relevant. Period for stop [laughter] thank you. [applause]
>>: I'm actually reading your book. It's a great book, however, I was a little discouraged by
your last comment. I was a little bit lost about mirror life, so is there any way you can give just
a 30-second layman terms as to what that is. I was lost on that.
>> George Church: Okay. It's really counterintuitive. So let's see, molecules in the world can,
have handedness. Not all of them do, but some of them do, so like your left and your right
hand when you hold it up to a mirror you get a mirror image of it, but you can't rotate it, so I
can't rotate my right hand and make it fit on top of my left. The only way I can make it look like
my left is through a mirror, so that's true with molecules too. Almost every molecule in your
body, not all, but almost all is chiral; it has a handedness to it.
>>: So does handedness literally mean the physical layout of the molecule, like [inaudible]?
>> George Church: So think of it like tetrahedron, right, where each of the balls on the vertices
of the tetrahedron are labeled it in different. That's what the molecules are like and it's really
because carbon has tetrahedral symmetry and has this, you know, if everything were made up
of simpler bonds then we wouldn't have this problem.
>>: So the idea is to make a form of life that which you could keep under control because it
would be completely separate from our ecosystem?
>> George Church: The separate from our ecosystem is partially true and the keeping under
control is only if we intentionally do that because here is a scenario which is if you create, if we
were to create a mirror life, it wouldn't be completely separate from the ecosystem because
there are a few foods that are common. So acetate, glycerol and glycine are three top ones,
and so they could steal all of that and incorporate it into their body and they could not be eaten
by anything else alive today because all of the enzymes, all of the viruses and everything have a
certain handedness they expect, so you'd have to be very intentional about this. And in fact,
many of the things that I raised in the book are not because I'm into science fiction or anything,
they are actually approaching much more quickly than we think and we need to talk about
them a little bit in advance so we don't just accidentally create something that will suck down
all of the carbon dioxide.
>>: [inaudible] for creating biofuel. One of the problems is they get sick. Presumably this
would create a form, you could do that for all algae and it wouldn't be infected by anything
living.
>> George Church: That's exactly right, but you need to be very careful. You need to have it on
a leash, so algae are particularly hazardous because unlike say E. coli which is another industrial
micro organism that we work with a lot. That one actually has to eat food, but algae is just
using minerals and carbon dioxide, all inorganic stuff and so literally it could suck down all of
the carbon dioxide in the atmosphere, so we may think we have too much carbon dioxide right
now, but if we lost it all, we’d have an opposite side to global warming. I'm not sure I answered
your question completely.
>>: So the significance, like if you had a molecule and you create a mirror of it what is the…
>> George Church: You do that chemically.
>>: For what purpose? Why…
>> George Church: So why? Here is one description, if we can keep them under control we can
increase productivity and done right we can increase safety. Imagine that everything in the
world that rots, rope, clothing, wood and so forth. These are actually good materials, but they
are displaced by materials, to the extent that they can be displaced that don't rot, but you
could make biological materials that would persist in the human body without causing immune
response or a much lower immune response and they won't degrade, so artificial molecular
devices I described, or even physically large devices, some implants and so forth. And it also
brings a whole new set of chemistry. It's not just twice as much chemistry, right and left
handed; it's all, at every position in the polymer, you can think of it as a long tape it can be
either right or left handed, so it's not just two; it’s two to the n where n is the length of your
polymer.
>>: So in the science fiction [inaudible] starfish, it's about mirror life organisms that just
happen to have a faster metabolism so they eat us out of house and home [laughter].
>> George Church: They don't even have to have faster metabolisms. All they have to do is not
die.
>>: I'm going to ask what the benefits other than the doomsday scenario would be [inaudible]
and I think you answered in terms of things I never thought about. Would that then imply that
you would create a mirror life ability to break that down, so that otherwise you just have super
plastics that don't ever break down?
>> George Church: I think we have to be careful about that. We have to be careful not only
about the things we create but the things we use to control them because let's take antibiotics.
There are quite a large number of them and one by one we are running out of them, right? I
mean maybe we had 12 major classes, 100 minor classes and as they develop drug resistance,
they are no longer useful. They just get crossed off the list. We only have two kinds of chirality,
right and left hand. If we use up the second one, that's it. So that's one way to do it. The other
way to do it is put them on a metabolic leash. That is to say they are dependent upon
something that only we have in the laboratory, right? And that's even easier to do for mirror
life, but it's something, there are ways we can make virus resistance without mirror, so we
change the genetic code. That's another way to get virus resistance. But all of these things we
developed we are practicing under controlled conditions how we can keep them on a leash and
really testing that, not just assuming because it looks good on paper that it's going to work. So
we have physical isolation, genetic isolation, every kind of isolation we can so we can do a test
on a small scale first.
>>: So the last decade was gray goo and this decade is like ambidextrous pink goo is about it.
[laughter]
>> George Church: Yeah, and gray goo too.
>>: One other question. I read recently, a subject in one of the [inaudible] were aware of it
that there is some discovery that there was an RNA marker that any cell that was infected with
a virus they could use to detect that cell has a virus and then cause it to self-destruct. Have you
heard of this? It's pretty remarkable.
>> George Church: Yes. We have heard of it and we use it as a tool. Almost every, so a lot of
biologists whenever they see a new tool they apply it to their biology. Whenever we see new
biology, we apply it as a tool, okay? And I think what you are referring to is the crisper system
but there are a few like that that there are basically two kinds, two major kinds of [inaudible]
response. One is the one I described which is protein antibodies that float throughout your
blood system and the other ones are kind of intracellular which are RNA that you get interferon
response, so they have all kinds of ways of recognizing that something is wrong with the nucleic
acid inside the cell. And these are great tools. You can use them not only for diagnosis and
fighting viruses as you mentioned, but you can also use them for engineering genomes which is
one of my favorite things.
>>: Is there, are they good viruses or is this like -- it seemed like it is pretty remarkable that any
virus that you could get rid of out of your system like is there a downside to that?
>> George Church: There are definitely good bacteria, right? So I mentioned a case where if
you don't have a good set of bacteria, clostridium difficile will take over your gastrointestinal
and that you do not want. There are very few examples of good viruses. Probably we are
ignorant, but not that ignorant. There are satellite viruses that kind of keep the other viruses in
check. There are not as many good examples of good viruses as there are bacteria.
>>: Is this is bigger panacea as it sounds? I mean, I read that and it's like holy cow that's…
>> George Church: So the tests that we have done, we meaning the field, is that you can grow
mice in completely germ-free conditions, multiple generations and they are okay. If anything
they are missing a few bacteria that helps the vascularization of their gastrointestinal system,
so there is some developmental feedback there, but there is no known consequence of having
no viruses. Now we have viruses in our own genomes, so these are endogenous retroviruses
and other things like that and they go wacky in your brain, so during brain development you get
a lot of these things jumping around and it's still an open question whether these are good for
your brain, bad for your brain or neutral. And so the only way to test them is using genome
engineering to get rid of them from say the mouse genome. And this is very significant in the
sense that when these things, if we are re-creating extinct species or doing transplants between
pigs and humans, one of the concerns is that one of these retroviruses will jump species
because of this intimate, this new intimate relationship.
>> Amy Draves: We have one of the online questions.
>> George Church: Sure.
>> Amy Draves: There is a lot of interest in the personal genome, so a question about how do
you get to the phased genome?
>> George Church: Yes, right, so we published this in Nature just this July. I may have the
reference here at the bottom. Yeah, I don't, anyway it was in Nature in July in collaboration
with Complete Genomics. The way it's done is really cute. You take as few as 10 cells, 5 to 10
cells, you break them up so the DNA gets a little bit broken and then you spread it out into a
large number of small wells, so 384, but it could be anything around that. In each well is about
1/10 of a genome equivalent and then you amplify that up with an enzyme and now in that well
is that fraction of the genome and now the chances are that you didn't get two copies of each
piece of DNA, the one from your mother and one from your father, so you actually get these
long things called haplotypes where you get basically you know that all of the DNA came from a
piece of one of the two chromosomes. And so you don't get confused between the two, that
eliminates all kinds of errors and it establishes phase that you know that all of these came from
your father’s chromosome or from your mother's chromosome. And so you get two for one.
It's about one error in 10 million. I forgot to mention this, and so you get both, about 100 times
better accuracy and you get phasing, so look in July Nature.
>> Amy Draves: There's another question about what you think about the privacy aspect. It's
one thing to do it for yourself or what when the insurance companies start to get a hold of it…
>> George Church: Right. So privacy has changed little bit since we started this project in 2005.
In 2008 Congress passed a law called the Genetic Information Nondiscrimination Act or GINA
and that prevents employers and health insurance from discriminating based on genetics to
matter how they get it, whether they get from your sister or from you or whatever, they can't
use it. And I think that that's got teeth in it and I can't imagine why they would mess around
with it. It doesn't affect long-term care and its life insurance, but I think right now even though
there's no law, no insurance company that I know of is doing anything to discriminate based on
that, so probably what's happening is the opposite, which is people like me that know that they
are APP3 homozygous and are probably not going to get Alzheimer's, why do I have any longterm care? Right, you know? So what will happen is most of the people that will stay in that
game will be people that are at very high risk of getting Alzheimer's and other things that
require early long-term care, so I think it's going to work the other way around. Privacy, you
know, I think it's one of these questions, kind of an open question, is is our genome going to be
like our face? Because our face right now is we have kind of contract with society that we are
going to share our face and our face tells you all kinds of things about us that you might not
want to share. It tells us pretty accurately your age, how healthy you are, what emotional state
you are in, your ancestry, just all kinds of stuff, right, and if you covered your face you could
keep that stuff private. But that's not really acceptable social behavior for most of us. And it
may or may not be the case for genomes. It's like we are in an historical moment where we are
deciding whether it's going to be like faces or whether it's going to be like something
supersecret.
>>: Another question. Could you describe the process of actually, you know, I'm a 40-year-old
male, and like if I want to change a gene, what is the process today of changing that and is it
just you change a little bit of the gene, like a segment of the gene in like a few cells, or is it your
entire body and does it stay locked in?
>> George Church: Yes. So gene therapy was very bad 10 years ago. You would take a virus
that would deliver a new gene you are missing and it would deliver it all over your body or, and
it would misdeliver it quite frequently and would cause cancer. Instead of curing a disease, it
causes a disease. I wasn't good and so that set the field back about 10 years. Now we've got
this new generation of technologies where we have these awesome zinc finger nucleuses.
These are the talons that would go to a specific place and they go -- you can think of the kind of
like a little nano robot. It goes to the right place and makes the fix. It can make it to one gene
or both copies of that gene if you have to knock out both copies. Efficiency needs to be
improved.
>>: Another question. So does it take along with it that section of the base pairs and cut and
insert or what?
>> George Church: It can. That particular conical study that I described from Singamo
[phonetic] which is a little company that I had a small role of starting in the ‘90s, in that case
they are just trying to screw up the gene so it makes a double strand break and then it lets the
body kind of, or the cell fix it with an inter-prone way.
>>: So essentially you're just cutting it out.
>> George Church: But there are plenty -- the technology that we use all of the time in the lab
for bringing in a piece of DNA and patching it in perfectly, so you can actually replace a piece of
DNA with another piece of DNA. It just hasn't made it to clinical trials yet but it's certainly, we
do it on human cells and, in fact, I have a center that I run that's entirely dedicated to human
genome engineering of that sort.
>>: And does it, sorry, does it lock in? I mean once you change it are all of your cells…
>> George Church: It's locked. It's not temporary. Well, it's not all of your cells, so getting all of
your cells requires efficiency and delivery that we don't currently have, but we have pretty
good efficiency and we are getting better at delivery. Right now if you want to make sure that
you've got it, you take the cells out of your body, you change maybe half of them and you get
rid of that half that weren't changed and you put those half back where they belong and, you
know, you have to help it out a little bit, but I fully expect as fast as this technology is going we
will be able to get any tissue to any degree of replacement.
>>: Is GINA worded in such a way that genetic information [inaudible] is it worded in a way that
is not discriminatory against the genes that you get naturally versus inverse discrimination? If
you [inaudible] and you start artificially creating genes for competitive advantage, does the law
discriminate against [inaudible]?
>> George Church: I think that the law does not address that question at all.
>>: So basically the genes they mean are the ones you're born with as opposed to 10 years
from now.
>> George Church: Yes. That is what they mean. But this was just like a first step. I think
there's going to be a flurry of legal activity in the future on this both positive and negative, but
so far it was almost addressing a nonissue. There were really no examples of genetic
discrimination either in health insurance or employment. The closest you came was
occasionally a basketball team would require somebody to get a test for hypertrophic
cardiomyopathy which is actually for their own good, but that was considered discrimination.
>> Amy Draves: I think we will leave it with that. Thank you so much for coming. [applause]