>> welcome to a very special lecture. the marshal w nirenberg lecture in the auditorium on the campus of the national institutes of health. and to the people who are watching on the video, wonderful to see the auditorium absolutely packed this afternoon, which i
think reflects the importance of the topic and the speaker that we are fortunate to have with us namely george church. before i introduce george, let me just also you say a word about marshal nirenberg. i am one of those who had the good fortune to know marshal,
and of course, he is the individual who figured out the genetic code back in the early 60's, beginning with that realization that that poly u codes for fen alnine and going on from that to fill out the table, something that we have had many occasions to reflect on
since marshal is one of our own, having spent almost his entire scientific career here. and if you have not seen the display about him and his work that's located over near the lipsett auditorium, in case you have not wandered over there, go have a look.
it's pretty interesting to see some of the original materials and equipment that he used back then more than 50 years ago to figure out this fundamental information, about how information gets translated from nucleic acids to protein. and marshal, as you know, did a
lot of that work in the area of molecular biology and in the second half of his career, became fascinateed with neurobiology and was one of those individuals with just boundless curiosity and i certainly remember a number of conversations with him as he was
not only chasing after the secrets of neurobiology but bringing into that a whole new set of ideas about chemistry and small molecule approaches, which were a great deal of fun to talk about. so we've had, since the last five years, this lecture to
honor him in particular, we have tried to make this an opportunity to bring in some of the most highly respected figures in the fields of molecular genetics and today we have actually done extremely well in that regard, because we have a speaker who is going to
speak about genetic codes. but also about brain codes. so capturing neatly both parts of marshal's career in the work of dr. church. he got his undergraduate degree at duke. he went on to get a ph.d at harvard in biochemistry and
molecular biology. after a brief stint at biogen then at ucsf, he landed back at harvard, where he has been since 1986 and is now a professor of genetics, health sciences and technology. he's director of the harvard nhgri center of excellence in
genomic science, having had that role now for all of 13 years. this is a major investment the genome institute has made into development of technologies and george has had a wonderful team that has been working on this and no doubt we'll hear something about that.
he's director of the personal genome project, the first effort to derive complete sequences on individuals and put all that information truly in the public domain, including his /oeown. he's a senior associate of the brode institute and he is also founding cofaculty and platform
lead in synthetic biology for biolodgically inspired engineer engineering. i am happy to say i think i've known george for 30 years and the way in which we in those days were trying to days figure out how /to sequence dna at the proper speed that it might
ultimately lead to the whole geno human genome and that ultimately succeeded and in no small part because of the creativity of george and others who continual continually invented new ways to speed up the possibility of getting rapid and accurate dna
sequence. in fact, he developed one of the first direct genomic sequencing methods. he's elected member of the national academy of sciences and also the /tpharnational academyof and he was cited as one of the major contributors, which i
suspect he may say something about, in sciences breakthrough of the year, namely crisper, where george has, played a very significant role in the resolution that probably almost everybody in /this room is now using in their own research. he has no less than 234070 pa
pageants. so we are fortunate indeed to have dr. church with us. please join me in giving him a warm nih welcome. /phra [applause] -- patents. >> thank you very much, francis
francis, and everybody. it's nice to see old friends here. you've already seen on the pre preslides my conflict of interests. that's what this is as well. laught[laughter] there isactually a full web page dedicated to it if
you want /to look at /ththat. and be thethere will be somedrugs mentioned, which is by no means the point of the talk. it turns out that marshal newer neweremberg and i have had some pretty close related passions. i've been working on genetic code since i was a teenager,
essentially my career career. initially on thecriystalography, which is transrna, which is the thing that breaks this code. we'll start on genetic code and how far it's come and where it's going. and then he had many papers on newuroblast /oepomas and inparticular
on transcription factors in newurobiology and our approachis transcription approaches in newurobiology. and he had an interest in science and society and i'll mention that briefly in wrapping up. so this is the genetic code four
years after his first codon was dissected which was u /srafu for fen algae. there is a lot /of ambiguity, especially the loeucine, tripto fan and argnine codons. but most of it was done by sin they sizing these nukcleotides. which was a big deal in 196
/1k35* they made especially -- essentially all of them. and polymers of these trinuke trinucleotides. this was -- and by the way, phil litter was one of the cospirit cospiritors in /this effort and he was my first department chair.
so we're constantly remind of the genetic code. this is where we are today. in fact, we have the numbers here are the number of instances of these -- each of these codon s in a complete genome and we have these from many different genomes.
this happens to have /* -- to be eco. coli. and now they are the ones that we're targeting for removal. coinciincidental. so they are loeucine, c /r-fplrargue arguenine. we have synthesized the 4 million base pairs, for which
this represents for each type of amino ajihad. and in so doing, we've he will eliminated either one or 7 cod codons genome-hon wide. and i'll give reasons in a moment why we did this. but the thing that's significant here is we normally reading
essentially the codons, but also now writing them. and we can do this radical re recoding for reasons that we'll get to in a moment. and the reason that we can do this, the reason that we can read whole genomes and in fact /tpwaebgt write them assistant
at the billion base pair scale is that it's ex/ponential, andit was breath taking, even back in the 1980s. we didn't fully recognize it was going at /tthat rate and then around 2003 and 2004, it changed slope, which tells you to be suspicious of these straight
lines. so this is factors of 10 on the y /ab-axis. and now it's been about a 3 million-fold improvement in sequencing and about a billion- billion-fold improvement in sin they sis of nucleotides. so i'll speculate on what
happened in 2003 to 2004 in an extremely immodest way by focus focusing on papers that came from my lab in 2003, 2004. but it also -- and the key was going to chips, essentially perfectly flat surfaces, way loud us to minuiaturize, multi multiplex and start a new code
that was much faster. and this was -- we were using chips to sin /tynthesize dna tonuse them the way they were conventionally use but to strip off /the dna to lost its structure so that we could a assemble large genomes. and there is an interplay
between sequencing and/stheupsynthesis synthesis. it's called sequencing by sin now it if we go from sin/tynthesize synthesizing short pieces to put pee together large genomes like the one that i briefly mentioned mentioned, this is often called he had -- editing.
i think there is a version of this is you which is called writing that you edit so much that you have to sin /tynthesizehuge blocks of 50, 100 kilobases, a megobase. but the he heditors, there is alot of obsession recently and francis alluded to this with
crisper, which is in the upper right here. but i want to remind everybody there are at least nine and there probably will be more and each one /-of them has been dis displaced or found its niche. and they have three different ways of scanning the genome,
this 6 billion base pairs in the case of human genome, to find a unique place to make an /aedit o oair cut, and that scanning head essentially, the information interface, is either dna, rna, o oor protein, examples of protein scanners. nucleases and rna as the crisper
cast 9 and dna scanning are things like the beta recombinase i'll show you in a moment. now, there are problems with crisper and i will, even though my lab was among the labs that invented it, i'll be the first to point out the problems of all of our inventions.
and this one has off-target issues and on target issues. the off-target issues have been many solutions we won't go through them. but all of them claim to be -- or many of them claim to be so low that you can can't detect the off-target but that's
typically because they've used the trick that was used in the very beginning here 2013, which is computational, avoiding being off by more than 3. so you have no instances where, where you have one or two dis dismatches. there is now one /waway, if youare
interested of general way of getting single nucleotide /kpwroeul morphism changes specificity and that's published in /this top-rateed journal called bioarchive, which i en encourage you all to publish in. from my group. so that's single nucleotide-
nucleotide-specific crisper. i can't -- and then there is the on target problems and this comes from all the methods that use double strand break. talons and crisper. when you make a double strand break, it's a race between the cell fixing it and getting your
donor dna to fix it how you want want. and there are two alternatives we've been working on longer than crisp er aer and these areas many things. these are harvested from the mi mi/krocrobial world inparticular. ecolea lamd recombinase and a
avoid the double strand break dilemma. so the various int grasses require paring of donor, editing molecule and the genome that you are trying to engineer, and the beta recombinase works at the lagging strand to pretend to be a fragment.
and so you are not making a break here. you are just in/krorcorporateing ing a mismatch or it could be a very large mismatch. so we've actually used those two non-crystal methods. so i call this beyond crisper, to do the largest and most
radical genome engineer to date, which is changing the genetic code, back to marshal newerem neweremberg, of eco. coli andhope hopefully many other organisms soon. and reasons are given here. and we've changed it first to get rid of one codon genome wide
wide. so there is 64 triplet codons and all organisms share one thing. they don't all share exactly the same genetic code, although it's similar but they share that they use all 64 codons somehow. and in nature and synthetically
part of this work there was none that used 63 or less. so now there is one that uses 63 63, and it shows all the things, all the goals that we have. it can use non-standard amino acids, not original to the set that all organisms have, 20, 21. it is genetically isolate sod
this could be used for bio biocontainment. and finally has multivirus re resistance. i think this is very interesting interesting, both practically and philosophyically that you can make an organism resistanant to all viruses in the world that
you have never studied before, because they all expect a gent genetic code to be provided by the host. and we can change that in a more radically the more viruses. but we were surprised that even changing one codon was enough to make it resist /tao*pt most
classes of viruses. now, you might say well, why do we need new amino acids /stpheurgs mean, isn't 209 add adequate? quickly justified -- justify that, one of the four goals. these are some of the ones that we and others have used to
advantage. you can put in fluorescent amino acids and you can put in multi multiple fluorescent amino acids acids, which is more faster. redox. there is a 10-fold /proeufplt in redox in some cases. rick place disulfides with disul
disulin ides. you can get optical wave length ishe usomerization. you can get peptic nucleotide acids but inexpensive root. and finally you get conjugates where you can make quick /kphaeupblgstry coupleing and this is already in use in the
pharmaceutical industry to make pegal aated human proteins like this in case human growth horm one. now, in order to do a wholeiem, the /stheusynthesis is not sobad and it's getting better exponential exponentially. the /rareal trick is /-the /*debug
debugging the biology. so we wanted to derisk this. rather than build a whole genome that was broken in 100 different ways and then trying to figure out what was wrong with it. we would break it up into pieces and debug all the pieces. and interestingly, when we did
the one codon, it was everything essentially all 320 changes worked. so that was a case where we were pleasantly surprised when we did the -- what we thought to be the two hardest codons, they were in indeed hard. we changed -- these are the uag
is /the meris codon and ag ganda are the they couldnsecond andthird rarest ones. and there were 13 examples where those were problematic and we had to debug it and eventually came up with work around fors all of them. then we did 13 codons for /aahuge
fraction of the ribosomal genes and here we are the only one with ungot you in changing those 30 and we engineered around that that. so all of these things we engineered around and that a allowed to us move arouon to the whole genome, which is now 100%
sin /tynthesizeed and aserved in yeast. it's 86 percent -- these numbers are a little out of date but close enough. 86% in edolei and valid /a*ateand in/krorcorporated them into the genome. and this was a team effort, led
mostly by nelly, who is a post doctoral fellow in the lab, as well as michael landen. now our goal is not just to have some genome that barely saisalive alive. we actually want it to be faster than the laboratory strains. and here you can see is our in
inspiration for this. on the left ha-hand side youwill find an organism that is disturb disturbingly faster than eco. coli, which is faster than other organism on the planet. this is called dab rio netrogens and it's woking -- becoming the work horse for molecular biology
that may re/phraeplace ecol.coli and we will be sin -- recoding its genome or making owe coli have the properties that we want. so to the second topic, having to do with marshal nuremberg, was brain codes. and i capital /aoeize the wordbrain for some of you that may know.
a few of my colleagues and i had an /obsession with brainactivity mapping around 2011, which managed to make it to the attention of the os tp and nih and other variati/skpaogsz itbecame a real thing, which is now called brain research through advancing oninnovativeneweurotechnologies.
so that's the akrcronym and i really loved it when it was announced, because it included the words oninnovative and technologies, which is not something you often get from a top-down project. and it shaped up have /* to be a very remarkable project and i'll
tell you a little bit about our little piece of it. so marshal nuremburg wormked on krebs and so forth and i'll tell you what we're doing now and in the near future that specifically involves building brains, not just analyzing them but both.
so it's the whole reading and writing mantra that we have for the genetic code and the genome. we can also do for these ep ep/skwrepigenetic code. and as you will see in a moment, we tried various ways of doing the ep/skwrepigeneticreprogramming, including using crisper, but we
found that in direct side-by- side-by-side comparisons, we could get a lot more mileage right now using transcription factors. and we were shocked to find out that despite the fact that we've got c /tk-fpdna /stkwaoefpbltssfor many human cnas and there was no
collection of human genes or human transcription factors that was adequate, even close to /ad adequate. so alex, as /aa graduate student in the lab, put together 1576 human transcription factors. we think this is at least one per gene that is classified even
in /aa broad classification of what a transcription factor is. 88 of them have alternative splice forms and those are quite useful and we'll probably keep expanding that. but there is at least one transcription factor per transcription factor gene. and
and we've had many apdataba-- ap applications. even though it hasn't been published, we have eight collaborators that found out about it through talks like this this. and here is an example of where we screened it for loss of pluri
pluripotency in /aea human stem cell line. more about that in a moment. and what you can find is that one end /-of the spectrum are things that stabilize pluripoet pluripotency and at the other end /-of the spectrum arefactors that take it to differentiate
into something. and it is an interesting open question, but we've used this to find a whole variety of new path pathways, some of which are extraordinarily efficient and rapid. so for example, when some of the protocols we were looking at for
some of the tissues required -- the protocols required 120 days and maybe have single digit efficiencies. while here we have now numerous transcription factors that will give us gust 98 -- give us 98% yield of newurons andessentially we don't have to /remove thestem
cells that are completely gone and the 2% that aren't newurons are not stem cells either. but the point is it's very efficient and short period of time. and here is the two methods that i've alluded to /tthat we'vetried tried.
we developed one of the most efficient crisper base. so this is dead cast 9 with three different activation do domains attached to a c terminis terminis. and this is about 20,000 times more active -- can increase gene expression by 20,000 times, and
it's 100 times more active than the first crisper activators that we and others developed. anyway, when we hooked this up to newurojegenic, we get about7% field /* yield, but we /- when we go to newuroojegenuinedirectly to transcription factor, we get 98% yield.
and i want day we're already seeing this huge transformation and sort of this gregarious pile of stem cells, which -- to these elong aated bipolar neweurons,where you have exactly two processes. now, one of the challenges, when we developed these protocols, where we get a new developmental
fate is /tkw-- figuring out whatit corners to but despite all the transcript /o*eomics that'soccurred in the world, we still don't have a list of all the transcript ohms for all the cell types in all the stages of development so we don't know what exactly this cell
corresponds to yet. but we and others are working on a cell atlas for human and mouse that will cover hopefully every cell state and every stage of development. it /tkwrufpblts -- most of the transcript ohms have been done on blocks of tissues that have
specific cell types. that's the conversion. it's very fast and efficient and we've done it for /aa largenumber of cell types. newurons, myeiocytes,, and here are some of the cases where on the top are the genes, the transcription factors that we've
used and down below are some of the antibodies that we used to show what cell type it is or to do fact sorting. and then we follow up with transcript /o*eome on these cell types. and /thaos work from alex, again again, and /aa post doc in thelab
lab. now the next step /up from cell types, tissue types, is organis or organioids, as they aremodest modestly called. we haven't gotten perfect organiorgans yet. and there is pioneering work
from lan /kcaster in 2013, where they actually used these two /* to study mi/kcroselcephaly genegent genetic influences that show cause and effect in variation in a way that's very hard to do in mice. it's hard to get because of the an /toatomical structures arenot
present. but they actually got good an tomical structures. you can see this beautiful layered effect. but if they get more than half a millimeter, you start getting ne necrosis because there is no vas clatour.
we set out to solve the problem using the same kind of pluripoet pluripotent stem cells, human ip ips, that we were using before, but now using the transcription factors in multiple combinations to get at vas /khculature. now the challenge was here if you just take newurons or nurown
ownal pre/kucursors inendo/thaoethelial cells and mix them together and hope for the best, they will separate like oil and water. that has been our experience. but if you put in induced pluri pluripotent stem cells that are un/tkedifferentiated, they loveto aggregate and if you have a mix
mixture where they are programmed in advance to do different things when you add the deoxy cyclin induceer, you will get the aggregation and get inter/spespersed self-assemblingnow vascularized organioid s. the paper had organioided butnow you have this organize --
organization with vas/khculature mixd in. and here the capillaries have been here labeled with anti- antibodies. /srefrplt cad heroin and are nicely inter/spespersed with the various newuronal types. we've alreaso, in addition to/-the
cap /hrillaries and thenueurons, we have other non-nur /o*euronal components. the next tep is hooking this up either to mi/kcropumps such asthe ones that are already common in our lab and in jennifer liewis's or getting it to pump itself with a cardiac muscle that you
will see in /aa subsequentslide. now, just /aa slight digression all of the humanep/skwrepigenetics that we've shown so far is due to the personal genome project cohort that francis mentioned in his introduction and all the slides i'll show for the rest of the talk are also from the same
cohort. a good number of them are from begguinea pig number one, whichis me. and it's had advantages. take two examples here. the nih has an en/kocodeproject, which is an encyclopedia dna element, which is aimed at
observing theep/skwrepbigenetics in every way you can, looking at factors, proteins that bind to different base pairs throughout the genome, looking at rna levels and so forth. but the project was mostly pred predicated on whatever cell lines were convenient at the
time and sort of represented various tissue types. but this is the only one that i know of where we had isojegenic sets such as fibroblasts where these all had the same genome but different epigenomes. so i think that made a particularly nice set and that's
publicly available, the data from all the labs that were involved in the nih code project project. another project was a collaboration of this and the fda. this may be the first time and they established genome in /aa
bottle to establish standards for the world for providing genomes and cells and so forth. and they looked around /theworld for /aa cohort that had been properly consented for this sort of thing and ours was the only one that had been consented for possible disclosure of their
identity and also for commercial use, because these standards have to be used commercially. so anyway, this project is now international. boston, toronto, london, vienna and also in new york and in case of off taustria, it issynonomous with the entire international
genome project. it's called genome off taustriaand in collaboration with many other omics. and any of you are invited to join it. this is an -- another example of use of these p /tkpw-fpg /p-fplpcell lines. this is for getting at variance
of unknown significant. it i willustrates run it one/waeway and you have gene therapy and you otrun another another wayand you get a cause and effect of individual point mutations dell did the clinical genetics on this card /kwro /phaoiomyopathy,which he thought might be a mutation in
the tass gene on the x chrome chromosome of this little boy. it might be this quguanine. now the hypotheses are not this narrow. there is hundreds of potential candidate alleles. but this is a nice case. so the x chromosome.
there is only one copy of this gene. and only one base pair. so this is one base pair in the whole 6 billion we thought might be causative. but up to /tthat point we didn't even have correlation data because this was the end of one
study. well, what can you do? the answer is quite a bit on this particular case. we turned you it into cause and effect by taking these begguinea pig number one cell line, making it -- putting in the crisper, the cat humanized cast /50eu69d
along with our pair. if you don't include, you get this mess of that some people call editing but i call genome valndalism. and but if you do include the re repair, then you get with fairly high yield sort of in the 30 to 80% range, depending on the
tissue type. but you'll get a deletion of the an example of this genome valen vandalism is an /euinsertion ofa rand /o*om base pairs that haveno match, obvious. anyway, so you can make this one one-gene deletion. and then you can makeorganioids.
so you've done genetic change and then you doep/skwrepigenetic re reprogramming and that's what we have here is /the one with parker. we've made cardiac muscle here and you can see this beautiful repeating motif of the organization of this cardiac
muscle, which /will contract normally with sort of a 6/0-beat per minute rate. and then one base pair in the whole genome, just changing that one base pair causes abnormal biokchemistry, morphology and fiphysio/hropbllogical beating. as does inserting eight base
pairs. they are both out /of frame for the tass protein. now you might say well, you've talked about off-targets and on target problems with crisper. how do we know that this is really what you say it is? and answer is that my graduate
students think nothing of when they change one base pair of sequencing all 6 billion base pairs to make /shaosure they didthe right thing because it is in inexpensive now. and that might be a lesson that someday we'll think nothing of sin /tynthesizing all 6 millionbase
pairs just to change it if it's cheap enough. stranger things have happened. in any case, this is kind of proof in /aa ro1 study that you have cause and effect of that single quguanine for that cardio mi/yopathy. so back to brains.
fire moment here. we are not just interested -- for /aa moment here. we are not just interested in making brains. we also want to be able to make sure that when we make these organioids, they are fizphysio physio/hropblogically reasonableand
if they are not, we want /to de bug them with all the tools we can and we call this a through e activity, which is typically done with castlcium imaging. but we can -- we are developing methods where we can do it in c situ and time does not permit. but questions can ask.
then we monitor behavior you in the case often video in/pput for visual tass and video of what the response, motor responses are. then the connect i'll have and how the synapses are connected. the developmental lineage. how each cell in the body got it
where it was from the fertilized egg and finally expression analysis, so rna protein. this is funded by irpa as part of the brain initiative and these are some of my colleagues at harvard and columbia. and mitt. and the main thing is getting
this connect oome andcorrelating it with the activity map with castlcium. this is -- this and the essentially all of the things on our a through e list are helped by the fact that next genome sequencing is actually an imagining method.
when we first developed it in 19 1999, we literally used a micro microscope. but conventional sequence ugrand ugrandomize the cell contents and display them on a flat surface. the original concept, even before next gen sequencing, now
called conventional sequencing and what we have now implemented implemented, called insitu sequence, is to leave the cells alone on a surface. they can be fairly thick. tens of microns thick, and sequence them where the rnas and dnas and proteins lie in the
first place. but in both cases, it's just a series of cycles where you use four colons -- codons per cycle and you build up and get an information content of 4 to the n, where n is /the number of cycles and it's easy to do, hundreds of cycles with next gen
sequencing, whether it's in situ or conventional. we tend to aim for 30 cycles or less, because you can get all the information you need from a short tag. now i am going it show you multi multicell /phraurlt sequencing and later subcellular, because r
rnas and proteins are not uniformly distributed and you can't take them out and analyze single cells. you have to see them in context. and where there is four colors, each neuronur -- nur /o*on has aspecific color because we've bar coded the entire nur /o*ons.
there will be thousand /s ofdots but for this purpose, we are bar coding these nurons so we know how /to connect over sometimes centimeter distances and this is work from tony zaider's group. watch on the side we are monitor monitoring one and two and building ape bar code.
and the cells, all the processes will stay in place but the colors will change. and you can see all the ax/o*ons and dendrites and you can see the bar codes building up on the side. now this is subcellular in situ sequencing.
this is each dot there is a sing single rna molecule -- molecule and the dots stayed in place but the colors kept changing. and then we can record all those colors and play them back out of the plane as /aa bar code. so every position in there had a bar code, which was built up the
same way next gen sequencing, except we didn't bother to throw away the morphology and the 3- 3-dimensional coordinates of each of the rnas. but we can also use dna and protein, as you cwill see in the next couple of slides. and we can do souper resolution
as well. one is called storm, which is fiphysics and optics strategy where you query over and over the same failuluoroflora and youget a better and better centroid for it. so here is conventional micropsy and the same magification and
getting the 5 to 20 nan /ometer resolution. and so it's on the order of nuke nucleosomes and this is sweeping through the field. it's work from professor at harvard medical school. so that's rna and dna. now proteins seems a little hard
harder. and it is. because we're limited by anti- antibody resources. but if you have these purple proteins and gray antibodies, these y-shaped molecules, tagged with a nucleotide, coupled to a poll mer.
so still straits two things. one is how /to tag proteins and turn it into a nucleic acid problem. and the other is how /to get ami microp /sp*eu where you literal literally take the cell and ex expand it by -- this is polymer acryl aate, the activeingredient
of disposable diapers, which /hr will absorb a lot of water. if you do one round of expansion and 100 squared. and so for example, here is a pre/s- andpost-significanynaptic proteins that are normally not resolved that are beautifully resolved and help us to determine where
the synapses are and which direction they go. and this is work that we've done for /aa few years now with edboyd boyden's group. now, integrated with expansion that's called expansion sequence sequencing and expansion phish, for ways integrated with other
nucleic acids. this is a busy slide, but the take-home here is we have nine different ways, which is like nine he heditors, different waysof detecting metabolites where we can essentiallytarian still aheady metabolite level into a protein level by stabilizing
that protein. and you have already seen that we can detect proteins in situ. so we are pursuing this as /aeaway of doing met blomic s s in situ. in summary with that part we have omniomics, which is pretty cool by itself. but you add the word in situ to
it and then you can do proteome s in situ, where you have all the advantages of my crosspi and morphology. so the last thing -- and i think we all should do and marshal did is to engage with society as a whole. think about where this is all
going and certainly many of the technologies i have brought up raise issues. and since the technologists see them first, we have a special responsibility of pointing out the down /skpaoeutdz solutions to the down sides and then the problems that the solutions
create and then the further tweaks, which makes people nervous, but we need to do it. so first thing i'll just jump right to where is everybody as genetically modified humans. there are already tens of thousand /s of genetically modified humans because 2300
clinical /traoeutrials over/-the years for gene therapy. you may not say that's not a genetically modified human but is /* it is. one /of them is approved and the delivery vehicle for that is associated virus, which explains why so many people are excited
about a ad because it's approved and it's fairly benign in every way. but it's really terribly small. it's terrible in that regard. it is hopefully the price will drop. as with orphan drugs, you have almost the same cost of
development as /aa regular drug but a smaller market. and for. these genetic counseling is an alternative. okay so you are saying that's not really genetically modified humans. it's about therapy.
there are already people who have received mitochondrial dna therapy. somehow that isn't always classified as therapy. i think it should be. there is a possibility of using experimentic stem cells, which is not always discussed.
typically this is about embryos, which is problematic because of month sakeism, where different cells have different types and so you can't really determine whether you are crisper or other gene therapy worked. but with stem cells you can get close to 100% correct editing by
the method i described earlier for the cardiac study by analyze analyzing stem cell clones. and i think if -- this will deal with /euinfertility but also if it's applied to /recessivegenes, such as t sacks, which already are being treated by either abortion or ibfpgd, which some
would clarissify with as evenmore embryo at risk, you might have that idea of p /tkpw-fpgd and80% at risk with it. so the irony is that germ-line therapy might reduce embryo harm rather than increase it. so i think we need to think, be not too difference? in when we
have to think carefully about -- dismissive. amp amping it up further in terms of alarm level. there is this tississue of canwe alter cog /tphenitive rates? dismissing is not a healthy thing for either side.
because it's hard to see it in -- varying with alleles in natural populations. but in synthetic situations, just like you will see a /* big impacts -- sorry, in rare cases you will find big impacts and also in synthetic cases in a moment.
for example, effects on al alzheimer's and some that affect disabilities. but then there is gene therapy trials that are based on some somewhat abnormal, not found in human populations, but based on genes that are in the genome. and so for example, there are
some gene therap trials and either preclinical animal or human trials. and there are even some that significantly enhance performance and cog /tphenit /stkpwhraoeufrpblgsz showing the black ones, where you have a significant increase.
very specific cog /tphenitivetasks. maybe sometimes multiple ones. and the reason that we might not be able to dismiss it as multi multigemult multijgenics and complex,meaning so many genes in environmental components is that even though there are maybe 10,000 single
nucleotide morphism that's affect things like height, the example of /aa complex straight with genetic environmental it's not all about the natural variation in the bell curve. it's about the ends of the bell curves and at the end are things that affect growth haorm ononeand
receptors. and when you find those, they're very often not onlyin/tporpformative about the pathways, but also they provide therapies. and here is an /kpw-rl growth ha horm oone therapys work onalleles and here is an /kaeuindicationwhere
growth hormone is used as /aadrug drug. even more alarming is people -- so we've gone from genetically modified humans to germ-line to complex traits. to now do it yourself gene therapy and this is a real thing it's been /roreported in two
articles in "technology review," among others. i can give their names because they are public. and they are do it yourself crisper kits for $150, assuming no prior equipment whatsoever. and these two did gene therapy by outsourceing it.
one /of them doesn't have any biological training whatsoever. these were tested in animals but not exactly the same gene therapies. so and in /this case i don't think is a good thing. the other one that i mentioned earlier is fine.
but these were tested in all these different animals. and they are trying to reverse aging and it's not completely out to lunch to do that. there is a great deal of information known about aging and model organisms. and in fact, our group has taken
45 of these into gene therapies for vet near preclinical /stkwraoeurblgsz could then lead to human. and in fact the fda process for vet near is much faster. and i think this is a more cautious approach. but we'll see.
another application, which is /* -- which has caused some concern 15 years ago there was concern that transplanting pig organis into humans would in addition to immune problems, would have in in/tkopbendogenousre/trtroviruses that will infect human cells and it is known that they do infect
human cells. and so especially immune com compromised host would not be great. so when we brought crisp tore bear on this problem 15 years later and it was about a $15 billion investment and it dropped because of per vs dish
didn't come up with theakrcronym -- we knocked out all 62 at once and it was not heroic. 14 days in an ink /skpwaeurt little bit of pcr is all it was. and we knocked out most of these 45 genes that were involved in the clotting in/kpabts with so now these were fetuses.
this is not published. we published the insidedodgenous re/ttrovirus, which as a as a record for crisper 62 at a time where two or three genes at once was challenging. we have positive ultrasounds, but i won't believe it until the piglet are healthy and are used
to transplant to -- into non- non-human primates. hopefully later this /kwrayear. and a /hrwill the of this wasled by yang, who was a post doc and is now leading this company called egenesis. another troubling -- positively and negatively -- each of these
has very positive components as well -- is this idea of gene drives. this could be a real boon to public health, especially in countries that cannot afford other medicines and even vax vaccines are hard to distribute into remote areas or war zones.
here you can aim it to /skprerbgts the vectors spread -- spread the good package themselves and the package in this case might be anti-ma anti-ma/hraeanti-malarial multiple anti-ma/hraelarialpeptides or antibodies that spread through the marriage certificate
population -- mosquito population, not at the normal rate of 50%, which doesn't really change. but at 100%. and here is an example of /aa standard drive. and one of the things we did in this case is we got ahead of the
game and talked about the safety issues and the reverseibility and keeping things local before we started the /speurplts rather than announceing it like gain of function flu after we are all done. and then we did the experiments in /aa variety of organizsmsrange
ranging from yeast to /thel/-the ma ma/hraelaria in mosquitoes. and idea is that crisper cuts, you have -- in an essential and green transfers over to a copy of the gene drive, which includes the whole crisper machinery, as well as your pay load, which /will fight the ma
ma/hraelaria and make /-hemosquito re resist aant. and it in /aa standard drive it gets cut and repaired and now you have two copies so all the offspring, 100%. and this is both the calculated and observed rate of spread, as it goes to fixation.
and we've come up with a disaisy drive. i'm not going to go through the details, unless you want to in the questions, where crisper one cuts crisper two. and crisper two cuts a and then all decay because there is a slight disadvantage in /this
model. but a goes close to fixation and could even go to fixation, depending on how many stages of crisper and how you rig this. but it can control things in time and in space. this is again in /this lovely journal called "bioarchive."
so in wrapping up, i want to say three thank yous. one is to p /tkpw-fpg ed.org forthe /skphoepbt education of this. they do congressional briefings without lobbying quarterly. they reach out to minority high schools and work with screen writers and then they have these
online game tools like map ed. and this is initiated by team woo in 2006. and janetin is an ethexist has been working with us ever since our firster grant and had an lc component in it and we've published many papers together, about 20 or so.
and finally i want to thank the nih for funding our centers of excellence for genomic science. in this case, our third on three different topics. first one was next gen sequence sequencing in 2004. then one on zinc fingers, which we didn't deliver but we did
deliver crisper. so i figured they were okay with and also on the first one we didn't deliver the 1.7 base pair bacterial genome that we promise promised but delivered 5 million genomes in next gen sequence. so we're getting away with murder on these grants.
but the next one is on organis. and i mentioned the i orpa project and didn't mention two transformative awards for molecular re/* recording and using this in situ method on human samples. so questions? thank you.
>> so please go to the microphones to pose questions so that everybody watching on video can hear. while they are thinking about it it, george, let me ask. i was very struck with this effort to find expressable versions of all the
transcription factors and that you are seeing some pretty surprising things. and obviously one wants to go from single factors to mixing them up /sand if you were goingto do two at a time, that's 2 million /123er789s. is that feasible?
>> right. so what we're doing with a sing single factors through the library is we will take a target target, let's say, and do rna sequencig on single cells so we know what the levels of the transcription factors are. we take the top ones, and we put
those in /as /aea mixture sowe're not doing full. and then we'll put in the whole library as the unprogrammed part in case we missed something or there is some thing synthetic that can help us out. that's our current plays. but we wanted to know what they
all did individually so we can put that into the model. pairs, we can probably do the whole pair-wise. triples i think is getting out of range and that's where you use this other method where you put in as many as you can guess at and then put in the whole
library. >> makes sense. over here. >> from past experience in ped pedia trick pathology i can tell you that the soxix can re regenerate all of the organisand essentially all of the tissues because i've seen a few of them.
i can tell you that it's always a question why you wouldn't just want /to inject some stem cells into people. generate it. and my understanding is that you put stem cells to a critical mass and they will automatically convert to a toma and you can
pick the cells out there have to do your engineering rather than to go through all of this pluri pluripotent stem cell et cetera. it seemed that you have perfectly viable neurotissue. you have all the tissue -- skin tissue. so there is there any answer to
that or is that just /aeunot a viable idea? >> yes. that's a great question. in fact, i started my work with gail martin, who worked in tera teratomas in the 80s. i do /* did my post doc in her lab in 198 /1k35* she was one of
the first practitioners of em embryonic stem cell manipulation manipulations. back before it was popular. the -- what we're trying to do is we're trying to make organized organis, not just tissues. we want /to know the rules andwe
want /to know what we can do and we want -- the closer the arts of fiphysio/hropbllogicalorganizs. so the teratome /kwraas far frommorph they will have local pieces that are quite interesting but they don't have the right -- so that's our goal. and i showed you an example
where we're vas /khclaizing and putting in glial cells into the previously fairly successful cerebral organioids, which were instrumental in figuring out mi mi/kcroselcephaly. but a /hrwill the of this has todo with the basic research of figuring out how organizsassemble
assemble. >> okay. >> it's an unusually shy audience. don't be shy, folks. there are a lot /of issues wecan talk about while they're thinking. you showed the data on what/uyou
are doing in terms of preparing pigs to be able to produce organs that would be acceptable to humans both because you have knocked out thein/tkopbendogenous virus viruses and tried /to make them less endojgenic. obviously, the alternative, which other people are pursuing,
is /the idea of making chimee erick embryos, where you end up with a purely human organi in a pig host because the pig is engineered so that it cannot make its own heart because of the paper describing this. so what is your view here of the pros and cons of othose a
approaches, both of which have ethical issues that are quite interesting and vexing? >> so i would add a third a approach, which is making human organis entirely outside of anan mal. but let's talk about those two. the problem with the chym /eras
-- first of all /skwr-rb, it'sethically more problematic. but in addition to /tthat, when you take out one organi, you re remove it from the developmental potential, and there is essentially a gap filled by the human stem cells. you really typically have gotten
rid of a small number of cell types and there are many others that are still in there -- nerves, cap /hrillaries, various muscle types, smooth muscle, et cetera -- and if you don't re re/phraeplace all of them, thenyou are going to have a host re rejection of that part.
and that's going to result in ne necrosis probably. so and most of these have been over a /* very short distances like rat-to-mouse kind of distances. and i am not saying it's /eupl impossible. these are some of the limitation
limitations. so if you are going to humanize them so that they can take on -- engineer them to take on human cells, you might as well human humanize the whole thing because the third problem is that if you have -- if you have engineered one organi, say the heart, the
rest of the pig is useless. but if you humanize the entire pig, then you can get ten different, sometimes more trans transplants. all the same tissues you will get from a human, you can get from a pig. >> over here.
>> hello. the first is a very small question. you showed he had -- you show he heed at the beginning the abundance of the codons in the could you talk about how that relate to the abundance of the different t /r-fpls /tphrnas?
and sorry and that relative abundance in different cells or tissues? so this codon table is very cool cool. i know there are cell types and it does var from -- vary from station aary phase to rapidly represelicating.
for the most part the number of trnas corresponds to the abundance in /this table. the exception is methiannine, which is especialspecial. but there is a correlation. and so when we recode, if we want /to change it from one sin synonym to another, then we need
to increase the number of trnas or release factors to compensate for that probably. but they are definitely used for regulation. but it's kind of -- it's a nuance so we can layer on top of >> and in different -- i also have a second question.
in different cells you find the same expression of the different t /r-fprnas? >> no. t /r-fprnas are levels dependingon cell type. the second question is broader broader. how do you see all these genomic
things converting into personal medicine? so in terms of public versus private research and how much these things cost and how it's going to develop? >> yeah, that's a pretty big question but i'd love to talk to you or a bunch of you afterwards
afterwards. but i think that we have a huge opportunity here with the technologies still going ex exknownst, where i think we're going to very soon have? addition to cameras on your phone, a sequencer on your phone phone, too, that the nan /oport
technology is really going into that direction of portability and low cost so you can analyze your environment, which is something hugely missing from the studies so far. we'll take two more questions. >> hi, thank you for the talk. i have a question regarding
building the organioids. one of the projects is building vas /khculature into theorganizs and how that was a challenge. my question is do you think building the entire, say, fulfill function /aal brain,each of the sort of units have a re re/kucursive programming,meaning
that they self-grow and they sort of in organizing to /-the beautiful structure of the brain brain? or do you think that at some point there are directive or plan that sort of shapes the final product? so whether this is more
individual blocks formed together into complexity or more a massive plan directing? >> right the court:^ so another big question, but it's an open one. and it's even more open than it might seem because we have the way it normally happens and then
what synthetic biology can do we've been pleasantly surprised now with our synthetic developmental biology that we can take things that normally take hundreds of days in vivo and hundreds of days in most protocols and shorten them to four days.
in fact, four days is kind of our standard for many of these protocols. we need to be individuvigilantthat what we're getting is either natural or acceptable. and in particular, making a human organi takes, anadult-size organi takes decades.
are there short cuts, or are we going to have to go -- clearly the pig does it in less time than we do. and that's something that's quite large. so it's a really interesting can we use synthetic approaches to making organis faster and
possibly better? by better, i mean reduced cancer cancer, he deuceed viral and re reduced senescence. that's something that will be hard to do in humans but -- so essentially the approval pressures are against it in humans, but they will be in
favor of it when you are dealing with organis. so i think there are some really interesting opportunities. >> last question. >> you covered the world of genomics and also ultimately the human fronteers of brain. so you think someday we should
be able to identify clusters of genome for higher intellectual ability of cog /tphenitiveprocessing and all the other nur /o*euronal functions? >> well, there is two shoulds one is technically and another is societally. my guess is that both will
happen. and part of it is there is going to be high motivation to deal with alzheimer's and othernewueuro neurode/skwrerpgeneric diseasesand some of those solutions will be effect /kwreuive in people whoare not already de/skwrerpgenerating. they will be enhancing.
in the same sense the vaccines are enhancing. they make us souper immune to a variety of path gens. pathogens. and we've already seen in mice that if you have a specific set of cog /tphe!nitive tasks inmind, you can, with one or two genes, make
a huge change. so i don't think it's going to be fundamentally different from other things. and we shouldn't use natural variation of population as our ouronour only guide post because there are synthetic things you can do.
most of fapharmacology is not natural and you end up with things that are in some cases far beyond the native state/-ofib human beings. but we need to think about this very carefully and talk about it in advance, as we are doing today.
>> congratulations, thank you. >> so what a wonderful romp we've had through technology, /sreugsz of the future, ethics that have to be very much in in/krorcorporated into all ofthis, which george has emphasized throughout all of these adventures.
i think if marshal nuremberg had been here, he would have loved we can join now in the library for some /skpoeuf cookies, but -- coffee and cookies, but please, let's thank george again again.
0 comments:
Post a Comment