 So, thank you Reshma and also I'd like to thank the organizers for inviting me to speak here today. It's a pleasure to be here and at the IHES. So just in brief, my lab works at the interface of biology, engineering and technology development where we are particularly interested in developing molecular technologies and GM technologies to reprogram genomes as a platform to probe biological function as well as expand biological function. So on the topic of genomics, one of the first sort of set the stage by first acknowledging the quite the revolution that we've been, we've all been experiencing over the past decade or so in the advent of next gen sequencing and impact it's had on biological research. In particular, with the ability now to generate scores of data sets, we can really gain a much better understanding in genetic variation, in particular starting to get a better sense of the genetic variants that are associated with specific phenotypes or disease. Clearly this is fueling a tremendous amount of effort and insights stemmed from genome wide association studies as well as impacting many other areas of biological research. But what's interesting is that the data that's generated from these types of next gen sequencing efforts really establish long lists of sequencing or genetic mutations that are implicated or associated with certain phenotypes. And so there are opportunities to really drill down and understand and a much better understanding the impact or the causal relationship between genotype and phenotype. And this is where we believe that the prospect of writing genomes would immediately be able to allow us to test all of the hypotheses that are being generated from these next gen sequencing data by being able to functionally create these genetic modifications and their combinatorial variants to really gain a much deeper connection between genotype and phenotype ultimately leading to a causal understanding. Clearly there's many other implications and applications that you can harness from the ability to write genomes, those being possibly the treatment of human disease and we're seeing lots of that clearly being discussed over the past few years. And the context of all the work that we're doing using this really is a platform to engineer synthetic biological systems. And in many respects I'd like to sort of invoke this nice quote by Saul Bellow which nicely encapsulates the spirit of research in the lab where we sort of think ourselves as writers who are readers, they're moved to emulation. So as you've heard throughout the course of the week, many of us also think about biology as a technology, really engineering and harnessing the diversity and the power of biology as a platform to change really the basic fabric of society, the way we produce medicines, food, energy, chemicals and materials. So really I won't harp on this but really acknowledge that a lot of this really stems from recombinant DNA technologies that were primarily developed in the 1970s and until recently really have not seen much improvement. However, they have led to a lot of production of high value compounds. Clearly one of the hallmark examples being the first introduction of recombinant insulin in the late 70s which obviously has had a big impact on treatment of diabetes. A number of instances throughout the past 40 years or so to really harness some of the power of recombinant DNA technologies. Over the past 10, 15 years, people have used these types of approaches to co-opt or re-engineer native metabolism to convert cells into little factories for producing high value compounds. Here's an example that stemmed from a collaboration between Dupont and Gencor in the late 1990s to convert E. coli into a factory for producing a new type of a material, in this case 1.3 propane diol and we heard a little bit about this from Sophan yesterday but in brief what they basically did is they ultimately isolated an engineered strain of E. coli containing about 27 changes that was quite successful in converting 90% of theoretical yield of glucose into 1.3 propane diol. This is great however along the way they went through literally thousands and thousands of strain variants to isolate this top strain and this really took over a decade and tremendous resources to achieve these goals. So in brief one thing I want to just impress upon you is that engineering cells can make stuff whether it's 1.3 propane diol or scores of other things that we've heard about throughout the course of this week or in the news really takes a tremendous amount of time and effort and resources to get there. So this kind of sets the stage for how we think about some of the key challenges in engineering biological systems. The first clearly being that biology is complex and I think it's important to appreciate the complexity and how much we've yet to understand about how these systems function collectively to confer the phenotypes that we're observing. This is confounded in the context of engineers who are trying to introduce new biological function really co-opting these natural systems and really fighting that upstream process of evolution. Those few examples from the disorbite technology really highlight our inability and the costs associated with really converting cells into factories for producing high value products and in particular I'd like to emphasize that I think a lot of these efforts are really limited by technologies that we have to probe and engineer organisms and by this I mean being able to really develop and deploy large scale technologies to create scores of target modifications across genomes, being able to at will modulate the expression of any gene or alter metabolism, complementary to those with being able to detect and respond to changes in important expressions of genes or changes in metabolites for example and then being able to engineer and regulate in response to that. And the last thing I'd point out is with the fast pace that the feel of synthetic biology is moving it's important to also consider safety and security aspects of GMOs. And here I've sort of raised two examples of what we're thinking about with regard to safety and security. First being the stability of bioaffirmation processes. So a few years ago in the summer of 2009, Genzyme, one of the world's leading orphan drug manufacturers in Cambridge, Massachusetts faced a pretty daunting problem where their biomanufacturing pipelines for two orphan drugs was compromised by viral infection wiping out the world supply of these orphan drugs for the better part of six to nine months. This is all due to basically virus contamination. So clearly there's a need to create bioaffirmation processes that are more resistant to contamination and more stable. Similarly, as efforts in synthetic biology moved and transitioned from more physically contained systems to applications in open systems such as the clinic or the environment, there's a growing need to consider addressing challenges of intrinsic biocotainment or basically limiting the growth of engineer organisms to synthetic defined environments. So what I'll now do is sort of transition to focusing on some of the technologies that we've been developing over the past few years with an eye towards addressing some of these challenges that I just outlined. So the first being is how do we really improve our ability to modify genetic material? And clearly I mentioned briefly most work until the past five or six years has really stemmed from nuanced improvements of conventional or common DNA technologies that date back to the 1970s where you effectively rely on really serial and inefficient modification of DNA allowing single or few genetic modifications. To a certain extent you can paralyze your efforts but the amount of sequence space that you could sample is very limiting. And so what we've really been thinking about and driving is the idea of genome engineering with the goal of introducing in parallel site-specific efficient changes across whole genomes. And in the process exploring combinatorial genomic sequence space. So both as a way to re-engineer and reprogram cells to explore a new function but also as a technology to really gain a much deeper understanding and establish causal relationships between genotype and phenotype. So just to sort of think about this a little bit more we think about being able to introduce large scale gene modifications as a way to lead to disruptive and novel fitness landscape. So if you consider sort of your gene type space over here and you can think about what evolution is trying to do which is maximize fitness or what the engineer is trying to do is to derive towards a very well-defined prescribed phenotype. You can think about sort of the scale of what you're able to do in the case of a single gene type of engineering problem. You can really sort of sample a very small piece of that genetic or phenotypic landscape. As you move towards a network you can start to explore larger landscapes but it's really the prospect of being able to modify the whole genome that really allows you to really sample entire genotype phenotype landscapes. And so this is where we've witnessed a lot of really powerful technologies being presented over the past few years. I like to sort of bucket them into two key categories. One is synthesis and one is editing. And so you get obviously tremendous amount of effort has been presented over the past few years out of both the venture institute as well as we hear later this afternoon in the context of some of the yeast 2.0 efforts to effectively re-synthesize entire genomes. Others including our lab have worked to develop technologies such as multiplexed automated genomes or mage as well as obviously the advent of CRISPR-Cas9 technologies really allow you to make target modifications across genomes of living cells. You really think about using the chromosome of living cells as the template for modification. So in thinking about this as well as well as also being inspired by some of the discussion from yesterday afternoon, the way we like to think about engineering systems is really rooted in one where it's a biologically inspired approach. So you can imagine sort of context of natural selection evolution. You've got your populations that could be subject to genetic variation, can create some altered genotype that could lead to some sort of ultrafenotype that's clearly subject to some sort of a violent with the selection that will select for cells that are most fit and ultimately those will survive. You can think about creating, if you will, artificial Darwinian systems where you could now develop technologies that are inspired by this process of natural selection that allow you to drive rapid and continuous modifications of whole genomes really of living cells and then deploy high throughput analytics. Clearly you could use more instrumentation based approaches such as sequencing, mass spec facts, et cetera, but really thinking about embedding in cells systems of biomolecular sensors and detectors that can respond to these changes and then you could think about taking that a step further by endowing them with the ability to self select for desired phenotypes. So this is sort of like a more of a biologically inspired approach and really the way we like to think about it as well is really engineering in the context of evolution. So I think you can have it both ways. You can think about developing models to really inform rational design but also appreciating and leveraging the power of biology's ability to adapt and evolve and select for phenotypes. So in many respects the mage technology encapsulates this spirit where the idea is to basically have design synthetic pools of single-stranded pieces of DNA where you can effectively design these such that they can introduce target modifications across whole genomes. This was first developed in E. coli with actually the goal of whole genome recoding which I'll talk about in a minute and what this allows us to do is site specifically introduce target mutations such that we can accumulate many mutations in a single genome as well as sample combinatorial arrangements across the whole population. And this really allows us to target modification efforts at the gene level to thinking about doing for example molecular evolution being able to target things on the network or pathway level so allowing you to basically make to multiplex a number of modifications across entire pathways as well as having the whole genome as a template for modification. How does this technology work? Just want to take a minute to explain these to you. This really was inspired by paper that I read out of Don Quartz group where he basically described this process where they introduced single-stranded DNA into E. coli cells that were modified with the ability to express a beta recombinase that would bind these single-stranded pieces of DNA protected from a nucleus degradation as well as positioned at the replication fork such that it could basically slide in, anneal right there, introduce its target mutation and effectively prime like an Okizaki fragment and basically incorporate its modification at a very high efficiency. With this I asked the question, can we basically scale this, improve this and really make this a powerful technology to write genomes? And so that really led basically to the development of the mage technology which allows us to basically take large pools of DNA and then effectively introduce them in a cyclical manner to populations that allow us to drive large-scale genome modifications. What's powerful about this technology is this is done in a cyclical manner so it can be done in a very iterative quick way where basically each cycle is about two to two and a half hours so you could basically work if you go 24-7 to do this 10 cycles per day or as many of you know we've also have developed and are continuing to work on automation solutions so that this can be done in an autonomous way. Importantly you can introduce diverse sets of mutations at high efficiencies and depending on the way you design the experiment and the size of your oligopopulation as well as your cell population we've shown you could basically generate billions of genomic variants per day and really using the chromosome as that edible template for engineering and evolution. So now armed with the ability to write genomes what would or could we sort of compose if you sort of will be inspired by let's say Beethoven's Symphony No. 9? So clearly we could think about as you've heard throughout the week using these technologies to optimize and enhance the discovery of natural products use this for biobased production, chemicals, materials, fuels, drugs, etc. and actually we showed the power of Mage to re-engineer metabolic pathways to enhance production of lycopene similar in spirit to the 1-3-pipin dial case and obviously there's lots of efforts now thinking about engineering organisms towards use for clinical applications as well as environmental. So what I'd introduce you is the notion of using these technologies to create new genetic codes and before I go into what we've done just want to remind everyone about the genetic code and some of the key properties. So clearly this is the canonical genetic code you've got your 64 codons the 20 amino acids that they encode and the stop functionality. So it's important and as you can see by this codon table is there's redundancy or degeneracy in the code and that you've got more than one codon for example these four plus these additional two codons will code for serine and so basically you've got this redundancy in the code. Importantly the code is largely conserved across all domains of life with a few exceptions but basically that really allows you to exchange genes between species and really drives the ability for horizontal gene transfer events and from a more biotechnology perspective it allows us to basically expand and leverage the power of recombinant DNA technology. Now as we've heard a little bit in the previous talk and by many others there's been a lot of efforts to expand the genetic code with really a focus on engineering the components of translation and it's really to encode the non-standomino acids. I like to think about these non-standomino acids also known as non-canonical that fall really into two main buckets. Those are purely synthetic or unnatural that don't exist in nature and those that can exist but only from enzymatic post-transitinal modifications of a natural amino acid. And people have basically, for example, all the labs of David Terrell have basically replaced a residue from a native amino acid across the entire proteome with a close analog to encode these new amino acids in that manner. Others such as Pete Schultz, Lee Wang, Zesron Shin, my colleague Dieter Sol have also used site-specific incorporation of non-standomino acids to introduce new chemistry into proteins all really with the goal of expanding the chemical repertoire of proteins changing structures, stability, enzymatic activity, etc. I'm not going to go into the details because we just heard a good description of protein translation but here's just a nice little graphic that shows natural protein translation. What I want to do is contrast that to what others have been doing in the context of basically developing orthogal translation systems that will allow one to site-specific encode these non-standomino acids. It really comes down to having orthogonal amino acid or synthetases that will uniquely bind to these non-standomino acids and charge in orthogal tRNA that will then interact with EF2U and allow it to be encoded at the rev zone during the process of translation. Importantly, these are designed to be orthogal, meaning that they won't cross-react with native transitional components or bind to the natural amino acids. Now, some of the challenges in this space is that, one, there's been no codon dedicated for 21st immunoacid and so what's commonly used is using the UIG stop codonine as we just heard in the last talk, but one of the challenges at this space is it competes with the release factor one at the ribosome which will basically drive the cleavage of the polypeptide and the release of the protein. A lot of the efforts really have been compromised by truncated proteins. On the flip side, what people have been basically doing is introducing high copy numbers of these systems, dumping a lot of the amino acids to really drive the system, but then what that can do is that can basically cause the misincorporation of these non-standomino acids at the greater than 300 native UIG sites across the genome which causes a lot of toxic effects which I'll touch upon later. Moreover, a lot of the orthogonal transition systems suffer from poor performance or really low enzymatic activity of these synthetases and I'll touch more upon that later, but just to give you a sense, the typical catalytic activity of these synthetases is in the order of two to three logs below the natural amino acid, amino acid tannery synthetases. So partly inspired by this, set out about 10 years ago with the goal of recoding the genetic code of E. coli. So here's the coding usage table for E. coli, the function that those codons encode as well as the frequency of those codons throughout the genome. And the idea here is to really leverage some of the key principles of the code, conservation and degeneracy, to see if it could basically perform a genome-wide codon reassignment. And the specific experiment that we first did was to basically try to eliminate TAG by reassigning it to the synonymous TA across the whole genome, creating an organism that's got 63 codons and has eliminated TAG from its genetic code. So you might be thinking about, why would we want to be doing this? So there's a number of key reasons why. One, from a more of a biological perspective, is to really use things as a way to test the availability of the genetic code. Could we actually, based on our understanding of the code, perform a whole genome-wide codon replacement and, moreover, be able to then knock out the function of that codon in that cell, which I'll talk about in just a minute. And then, as I mentioned just a few minutes ago, inspired by some of that work on engineering orthogonal translation systems, could we now then reintroduce that codon as a dedicated open coding channel and specifically incorporate non-stratomial acids at that position, removed from some of the challenges that I described just a minute ago. And then we had a few other goals and hypotheses. By virtue of having an alternate genetic code, would that allow us to create basically these genetic firewalls such that we could genetically isolate these organisms from other organisms or viruses in the environment? So, for example, what would happen if we basically took viruses and tried to infect a recoded genome? Could that viral genome be properly expressed in the context of a recoded genome? And then lastly, could we then take this a step further by then creating dependencies on synthetic amino acids for these organisms to grow as a way to address some of the challenges of intrinsic biocontainment? Okay, so how do we go about doing this? Just to give you a sense of what this project entailed looking at recoding and coli by the numbers and biological constraints. First thing I want to point out is that really to think about this, we're considering both genome considerations as well as some of the translational considerations. So, on the genome side, there are about 321 TAG codons dispersed across the genome, about one TAG codon every 14 kilobases. Seven of those genes are essential, and 39 of these TAGs reside in overlapping genes. So, that means that these could basically change the open reading frame of 39 other genes. So, these are some of the challenges on the genome side. At the translational side, this is very important, is that when you think about recoding, you have to basically identify a mechanism and a path such that when you basically reassign your UAG to a UAA codon, you can basically eliminate the translational machinery that decodes that codon at the ribosome. And this is where we hypothesize that by recoding UAG to UAA it would render RF1 functional, allowing us to delete it, while still retaining two-stop codons that could be faithfully decoded by RF2. So, that leads to the underlying key hypothesis behind this whole experiment, is that re-signing all UAG codons to UAA will render RF1 non-essential, eliminating natural UAG function, creating an alternating code with 63 codons and a free codon that can be dedicated to a 21st amino acid. So, with that, I'll just summarize what the key experiment involved. So, it's really three steps. So, you start with your wild type E. coli with the full 64 codons, focusing on the stop codons that are decoded by RF1 and RF2. The first question is, could we basically perform a whole GMY codon replacement re-signing all UAGs to UAAs? Followed by addressing the question, could we eliminate UAG termination or its natural function by deleting release factor one? And then lastly, could we reassign UAG from a stop codon to a sense codon by basically introducing these orthogonal translation systems? So, this first challenge really inspired and led to the development of these technologies, a mage of which I described earlier, a cage which I won't describe, but in the interest of time, I'll just use this picture to illustrate what this experiment looked like. Effectively, what we did is we started from a wild type strain divided up into 32 pools where we basically used mage to make 10 codon changes across each of those 32 strains simultaneously to ask the question, are all of these mutations permitted? And are any of them deleterious or synthetic? Once we confirmed that all those changes were achievable, we then basically developed a conjugative assembly genome engineering technology that allowed us to hierarchically assemble the recoded regions across all 32 of these strains through five steps into the chimeric genome that contained the full complement of the recoded genome. And ultimately, we were able to show that we were able to recode all 321 codons from UAG to UAA. That led to the really important question is could we eliminate wild type UAG function? That is where we basically did the experiment to try to delete release factor one from a genome where all 321 codons were recoded. And we basically benchmarked this against some other strains that we and others had developed in the literature. One were basically Lee Wang's group mutated RF2, so it had gain of function so that could basically decode UAG. And what you can basically see here, and also there's a minimal set of genes that have been recoded, is seven essential as well as six others that were shown to improve fitness. But what's important here is if you look at a ratio of cells that contain release factor one versus those that don't, and look at their doubling times, only the strains that have been fully recoded really should know impact on growth. So this really allowed us to conclude that the only essential function of release factor one was the peptidol cleavage at UAG. And it confirmed our hypothesis that recoding all UAGs to UAA then allowed us to eliminate wild depth UAG function. That led to the next experiment. Could we now build off of the work by Pete Schultz, David Terrell and colleagues and introduce orthogonal transition systems and reintroduce that UAG codon. And what we found here, similar to what I just showed you a minute ago, is that only in the context of a recoded genome and then here we're looking at doubling time as well as max OD, again ratios of release factor one to release factor minus cells, do we basically see that the fully recoded genomes are able to really faithfully encode these amino acids without impairing growth. Or you could see these other solutions so significantly reduced fitness. And what we've later showed in the paper that described this is that by and large what we actually have observed through mass spec is a large number of suppression events by encoding these non-strand amino acids at the many scores of UAG codons that reside in wild type backgrounds. Now if you look at our ability to use us to set specifically encode these amino acids to proteins, what you could basically see is the recoded organism has the ability to encode really scores of many new amino acids. So you're looking at 10, 20 or 30 site-specific incorporations into a protein or a polymer, you can benchmark that against your wild type tyrosine control, you can basically see we can produce these levels that are on par with the wild type controls versus your wild type E. coli shows significantly attenuated ability to do so by and large due to the presence and competition from release factor one. So now what we've been doing is asking the question what kinds of new biological function can we extract from organisms that have been recoded. And this is the largest opportunity to sort of think about some of the goals of recoding. It really inspired and led to the development of some very powerful genome engineering technologies and now we're able to start addressing problems in genetic isolation, biocontainment, as well as the production of synthetic proteins and polymers. So first I'll touch upon genetic isolation. So as I mentioned one of the goals here is to see if we can engineer organisms that are isolated. One of one key experiment in doing so is to test the ability of phage to infect these recoded organisms. Here what I'm showing you is data from experiment looking at plaque area. We're looking at the T7 infection using your wild type E. coli as a control. This is a metric of plaque area. We looked at a number of other metrics as well. This gives us a baseline. When we look at the recoder organism that contains the release factor one, it shows plaque areas that are basically equivalent to the wild type. However, when you knock out release factor one you see a striking attenuation showing basically that when you eliminate UAG function, it basically isolates, creates this genetic isolation and leads to basically virus resistance. And in other experiments that we've done of the past year or so, we've actually shown this effect scales to multiple viruses and hopefully that work will be coming out over the next few months. From an applied perspective, what we're thinking about now is really using this as a way to apply this to address some of the challenges of contamination and bioaffirmation that I mentioned earlier in my talk. Now, thinking about this in the context of producing synthetic proteins and polymers, I showed you this before. We could basically drive many corporations of amino acids. But when you take these orthoglutrientization systems and you introduce it into the chromosome what you see is a striking decrease in the ability to encode and produce those GFP proteins. And really what this experiment does, it highlights the poor catalytic activity of these amino acid synthesis, highlighting another major challenge in the field. And so this is where we turned to the development basically of a platform that starts with a crystal structure that can inform targets from eugenesis. Then we use MAGE, now here as a molecular evolution technology tool to evolve the synthetase that's in the genome. Specifically targeting the amino acid binding pocket as well as the domains that interact with the tRNA. And then basically embed in this recorded organism selections that will select against any cross reactions with the natural amino acids. And then we feed that basically into a high throughput fact sorting experiment to enrich for those variants that have increased GFP suppression. And that's followed up by detailed biochemistry mass spec validation. To make a long story short, basically what we basically did is we targeted the molecular evolution of those two domains and we find, for example, eight fold improvement in the amino acid binding pocket, more of a moderate one and a half fold improvement in the domain that interacts with the tRNA in particular the anti codon loop. You put them together, you get a synergistic improvement of 17 fold and we show in a separate for a separate mass of a 25 fold improvement. And we then further validate this through biochemical characterization. What this importantly allowed us to do is to drive site specific incorporation of many of these amino acids into proteins at high purity and yield. So now we're able to, for the first time, produce new types of synthetic polymers. So what we basically did have a postdoc who was trained previously in studying elastolite polypeptides which have been basically used for applications in drug delivery and basically these are basically these repeat pentapeptides where you've got these guest residue positions. And what we basically wanted to do is encode many instances of this dopa amino acid which strongly binds metals and organic materials. And effectively what we're showing you here is a new type of a synthetic polymer. And what this video is going to show you is a purified dopa elp solution that has many instances of this dopa. You can see it can be easily pipetted up here. But once you expose it to this iron nitrate solution it undergoes this immediate this immediate aggregation and effectively forms this strong adhesive that's shown here. We're currently doing a lot of the detailed biophysical characterization but we think this has really powerful applications as a new type of biocompatible adhesive. And we're actually producing a whole class of polymers that are endowed with entirely new properties. What we're excited about is using recoded organisms really as living foundries for producing sequence-defined polymers that can't be made in natural biological systems which are largely limited by the 20 canonical amino acids but are powerful in that they're very precise and can basically produce polymers that are template-directed, especially at the ribosome. On the contrast, chemistry can produce lots of polymers with unlimited diversity but they can't be done in a template-directed manner. What we want to basically do is in the context of recoded organisms is really leverage the powers of both of these fields and start to produce entirely new classes of polymers in a template-directed manner for various applications. And I'll just close for the next few minutes in explaining how we've also leveraged recoded organisms to address problems in biocontainment. And basically here it's motivated by the prospect of the field of synthetic biology, more broadly industrial biotechnology, to transition from more closed system use to open systems where you can imagine engineering new types of probiotics or engineering organisms to address problems in environmental applications such as bioremediation, which require biocontainment measures such that you can restrict their growth to define synthetic environments. As we're starting with this recoded organisms that have these orthological transition systems, what we ask ourselves is could we basically reintroduce these UEG codons into multiple essential genes and then link their viability to an exogenous supply of synthetic amino acids such that the production of these essential proteins requires the incorporation of these synthetic amino acids. And so we first employed approaches where we basically just try to introduce these synthetic amino acids right at the amino terminal of proteins as well as using some computational design to identify tolerant sites that allowed us to basically achieve escape frequencies in order of 10 to the minus 3, 10 to the minus 7, which are good but really don't show significant improvement over current state of the art or the kind of containment that we really need to address those other problems. Clearly creating higher combinations of these TG codons drive those escape frequencies down to about 10 to the minus 9, which is going in the right direction but still not where we want to be. Ultimately we converge in a solution where we basically targeted conserved functional residues of multiple essential proteins where for example if you focus on DNA which is an essential protein that initiates the process of DNA replication it forms dimers in order to towards their functional state if we basically replaced a conserved tryptophan residue in the dimerization site with a paraxylphenylalanine it allowed us to retain the function of that protein which now depends on a synthetic amino acid. Upon doing that against a few other proteins we basically showed that we could push the limit of escape frequency detection below actually our limit of detection which is about a trillion cells. And this were planning lots and lots of cells and a lot of hard work by the people who were doing that. But what we found is that after about two days these really the small colonies with poor cells started to form. And when we see those genomes we saw the formation of tyrosine abrasopressors. So then that led to the hypothesis that if we basically knocked out a tandem set of tyrosine abrasopressors it would basically leave the cell with a single tyrosine abrasopressor that would be essential to retain tyrosine incorporation to your protein synthesis. And upon doing that we basically observed no escape mutants in the context of about trillion cells over basically a seven day experimental plates as well as about three weeks in liquid culture. So looking ahead we really think about now these recorded organisms as a new type of GMO that really addresses safety concerns. They're stable and most importantly they're immune to rescue by metabolic crossfeeding which has compromised many types of biocontainment strategies prior to this. What's interesting about this particular work is that it's now by creating these dependencies on these synthetic amino acids it really requires these new types of synthetic biochemical building blocks for viability. And what's really exciting for us is could this really establish a new basis for synthetic molecular language capable of sampling new types of evolutionary landscapes. So with that I'll just close by summarizing some of the take on messages from our recoding efforts. One is that recoding is achievable. We basically were able to show that we could demonstrate the elimination of a wild type codon function and reassignment to new function. And moreover we're now using these recorded organisms as little factories or living foundries for producing natural life sequence find polymers for an array of new types of applications that we're really excited about and also show it how we could use these recorded organisms to address problems and biocontainment, genetic isolation and virus resistance. More broadly what you can see from the work that we're doing is we're developing a whole host of new enabling technologies. I focus a lot of the genomic ones but other types of biomolecular technologies with applications for engineering pathways and genomes as well as continually doing more recoding towards producing new and orthogonal biological systems that now are going beyond just E. coli into other microbes communities as well as eukaryotic systems. With that, thank some of the key people who were involved in this work. I first started the recoding effort while I was a postdoc with George Church with some collaborators down here and since that time some students and postdocs of mine have really expanded those new directions I just recently described. So Alexis Rovner was important in the finishing of the recoding work as well as the biocontainment efforts in collaboration with Adrian Heimovich who worked with Mira Amiram in some of the efforts on being able to evolve the orthogonal transition systems towards the encoding of many amino acids and Natalie has been working on the genetic isolation and virus resistance work. And with that I would like to ask all of you for your time and attention. I think we're running a couple minutes ahead of schedule so why don't we take a few questions for Farron on this topic. It's a great talk and you introduced some very powerful tools for coaching and editing and as you said that allows you to access a much wider or a much larger fraction of the sequence space. I was wondering if you can comment on the general strategy on how you're actually going to screen through that sequence space if the thing that you're looking for is not tied to survival. Yeah that's a good question and so that sort of speaks to that one slide I spoke about earlier where really think about biologic inspired solutions to engineering and so if you think about that with some of these genome and genetic technologies you're right we're really able now to drive large scale modifications across genomes creating lots of genetic diversity and that by extension will create a lot of different phenotypic outcomes and then the next challenges really are being able then to embed selection systems inside those cells where you could basically link those to desired phenotypes and then self select. So we are working on strategies based on protein evolution and engineered RNA molecules and we have some promising results but nothing that really is earth-shattering just yet but that's how we're thinking about going after the problem that you raised which I think is sort of like the next technological challenge that we're facing right now but there's still lots of things that we can still achieve in the absence of having those selection strategies in hand. So the mage technology allows you to show this about 30% or more than 30% efficiency. So I was wondering how finally you could replace all these Ampere Stock Covalent and solve this discrepancy to then allow 100% efficiency. So what was the trick? Well when we report the efficiencies of the mage technology it's think about it about in the context of let's say you have one all ago and you introduce that in the population cells we have 30% of the population will undergo that genetic modification. When you multiplex you can increase that overall efficiency but it will decrease a little bit on a per all ago basis. So what we basically do is we just basically do continuous cycling such that we can accumulate all the desired changes in the single genome and that's how we get to all codons. We were working on strategies of getting those efficiencies higher and we have we first reported that but it's it would be very difficult to get to 100% although we're getting closer. Maybe related to that question there's this fast development of techniques in genome engineering or genome editing so would you do everything as you did in the first place if you would start today to change your 321 TAGs or would you go for Cas9 based systems? So it's a good question. So first of all in terms of the ability to drive massively parallel modification mage has a much higher ability to multiplex than Cas9 particularly in bacteria. If you go to maybe eukaryotes maybe you might opt for Cas9 or you'll see coming of our lab over the next year or so a mage type of technology in eukaryotes. We would however change the way we designed the experiment so you could see I described it as we basically divide the genome into 32 segments and recoded 10 at a time. We would change the way we do that we probably do that approach across probably four strains and we could probably achieve that same recoding effort in under 6 months just that the when we start this particular effort these technologies didn't exist we developed them along the way but we would still use the same mage technology which allows us to drive massively parallel modifications and really explore combinatorics at a scale that goes beyond what CRISPR can do today. Maybe CRISPR can get there ultimately but we would still opt for this in the context of E. coli. That was a great talk kind of like Kirsten's talk earlier I was curious for T7 violence then how many TAGs is it in a genome where a few irradiated T7 have hydrogenized it how many escapements were you able to get? It's a good question. So in T7 there are four genes that contain TAG and we think it really comes down to one of them which is really essential in the vial formation of the envelope but we partly answer the second part of your question and a few other viruses so for example we did similar experiments with Lambda M13, MS2, P1 and showed even greater degree of viral immunity and what we did then is two important experiments. One is we created lysogens, recoded those viral antibodies to TAA and showed that that then allowed them to easily infect the recoded background and the other experiment we did is we just then we subjected the created this adaptive evolution experiment where we continuously exposed the cell to the virus and ultimately because of high mutation rates in viruses the virus basically mutated different escape routes. So we observed one some TAG codons in the virus convert to TAA and in other instances the ultimate codon in the virus was converted to a TGA stop codon so that experiment is important for a few reasons one is that it's showing that the recoding does create these barriers and that currently viruses because we're only recoding a single stop codon can overcome them but it does establish the proof of concept that recoding if you do additional forms of recoding will create much stronger barriers to viral infection so I would say stay tuned for those experiments yeah my question is related to that part of your work which is related to artificial extended genetic code so I mentioned the genetic code is a dander probably it's not clear, at least a priori it's not clear at all why it is redundant in that way and in some other way and there is at least it's where what works trying to somehow to explain and understand this redundancy and some of this were based on numerology I would say but some of them are quite pretty much interested from mathematical viewport so they try to model or explain it using some pretty much abstract mathematical machinery like representation of groups or things like that so I'm wondering how this extended code fits this picture so maybe we try to ask more concrete question so when you modified genetic code you were driven by purely biological considerations or maybe more by some more abstract considerations in the calcimetry of some sort or something like that we're sort of proposing some interesting philosophical questions in some ways in terms of our motivation we're motivated from a biological perspective based on our understanding of the code and the redundancy and the degeneracy that you described is recoding possible and clearly we show that we were also really motivated by some of the new kinds of biology that we're uncovering with organisms that have been recoded so in many instances it's a really interesting platform to maybe start asking or start really answering some of the questions that you're posing on the code we haven't really done that yet and would be happy to talk to you about ideas that you have right now what we've been really focusing on is really leveraging the new types of properties that these organisms have clearly with the ability now to make large-scale genome modifications and showing that recoding is possible you could start to design experiments to go after some interesting questions about conservation of the code codon usage to what extent can you really drive large-scale recoding across not just one but let's say a few to maybe let's say a dozen codons and how does that impact the code, the fitness of the organism expression of genes RNA secondary structures, all sorts of questions that you can structure address now in the context of understanding what led to the canonical code, what are the constraints and how far can we push the system and really more towards new types of functions so I think there's all sorts of questions that you can address we're just to the beginning